linux_old1/tools/perf/util/symbol.c

1918 lines
43 KiB
C
Raw Normal View History

#include "util.h"
#include "../perf.h"
#include "session.h"
#include "sort.h"
#include "string.h"
#include "symbol.h"
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
#include "thread.h"
#include "debug.h"
#include <asm/bug.h>
#include <libelf.h>
#include <gelf.h>
#include <elf.h>
#include <limits.h>
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
#include <sys/utsname.h>
#ifndef NT_GNU_BUILD_ID
#define NT_GNU_BUILD_ID 3
#endif
enum dso_origin {
DSO__ORIG_KERNEL = 0,
DSO__ORIG_JAVA_JIT,
perf record: Introduce a symtab cache Now a cache will be created in a ~/.debug debuginfo like hierarchy, so that at the end of a 'perf record' session all the binaries (with build-ids) involved get collected and indexed by their build-ids, so that perf report can find them. This is interesting when developing software where you want to do a 'perf diff' with the previous build and opens avenues for lots more interesting tools, like a 'perf diff --graph' that takes more than two binaries into account. Tunables for collecting just the symtabs can be added if one doesn't want to have the full binary, but having the full binary allows things like 'perf rerecord' or other tools that can re-run the tests by having access to the exact binary in some perf.data file, so it may well be interesting to keep the full binary there. Space consumption is minimised by trying to use hard links, a 'perf cache' tool to manage the space used, a la ccache is required to purge older entries. With this in place it will be possible also to introduce new commands, 'perf archive' and 'perf restore' (or some more suitable and future proof names) to create a cpio/tar file with the perf data and the files in the cache that _had_ perf hits of interest. There are more aspects to polish, like finding the right vmlinux file to cache, etc, but this is enough for a first step. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-28 07:37:06 +08:00
DSO__ORIG_BUILD_ID_CACHE,
DSO__ORIG_FEDORA,
DSO__ORIG_UBUNTU,
DSO__ORIG_BUILDID,
DSO__ORIG_DSO,
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
DSO__ORIG_KMODULE,
DSO__ORIG_NOT_FOUND,
};
static void dsos__add(struct list_head *head, struct dso *dso);
static struct map *map__new2(u64 start, struct dso *dso, enum map_type type);
static int dso__load_kernel_sym(struct dso *self, struct map *map,
struct perf_session *session, symbol_filter_t filter);
static int vmlinux_path__nr_entries;
static char **vmlinux_path;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
struct symbol_conf symbol_conf = {
.exclude_other = true,
.use_modules = true,
.try_vmlinux_path = true,
};
bool dso__loaded(const struct dso *self, enum map_type type)
{
return self->loaded & (1 << type);
}
bool dso__sorted_by_name(const struct dso *self, enum map_type type)
{
return self->sorted_by_name & (1 << type);
}
static void dso__set_loaded(struct dso *self, enum map_type type)
{
self->loaded |= (1 << type);
}
static void dso__set_sorted_by_name(struct dso *self, enum map_type type)
{
self->sorted_by_name |= (1 << type);
}
bool symbol_type__is_a(char symbol_type, enum map_type map_type)
{
switch (map_type) {
case MAP__FUNCTION:
return symbol_type == 'T' || symbol_type == 'W';
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
case MAP__VARIABLE:
return symbol_type == 'D' || symbol_type == 'd';
default:
return false;
}
}
static void symbols__fixup_end(struct rb_root *self)
{
struct rb_node *nd, *prevnd = rb_first(self);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
struct symbol *curr, *prev;
if (prevnd == NULL)
return;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
curr = rb_entry(prevnd, struct symbol, rb_node);
for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) {
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
prev = curr;
curr = rb_entry(nd, struct symbol, rb_node);
if (prev->end == prev->start)
prev->end = curr->start - 1;
}
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
/* Last entry */
if (curr->end == curr->start)
curr->end = roundup(curr->start, 4096);
}
static void __map_groups__fixup_end(struct map_groups *self, enum map_type type)
{
struct map *prev, *curr;
struct rb_node *nd, *prevnd = rb_first(&self->maps[type]);
if (prevnd == NULL)
return;
curr = rb_entry(prevnd, struct map, rb_node);
for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) {
prev = curr;
curr = rb_entry(nd, struct map, rb_node);
prev->end = curr->start - 1;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
}
/*
* We still haven't the actual symbols, so guess the
* last map final address.
*/
curr->end = ~0UL;
}
static void map_groups__fixup_end(struct map_groups *self)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i)
__map_groups__fixup_end(self, i);
}
static struct symbol *symbol__new(u64 start, u64 len, const char *name)
{
size_t namelen = strlen(name) + 1;
struct symbol *self = zalloc(symbol_conf.priv_size +
sizeof(*self) + namelen);
if (self == NULL)
perf_counter tools: Add 'perf annotate' feature Add new perf sub-command to display annotated source code: $ perf annotate decode_tree_entry ------------------------------------------------ Percent | Source code & Disassembly of /home/mingo/git/git ------------------------------------------------ : : /home/mingo/git/git: file format elf64-x86-64 : : : Disassembly of section .text: : : 00000000004a0da0 <decode_tree_entry>: : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 3.82 : 4a0da0: 41 54 push %r12 : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.17 : 4a0da2: 48 83 fa 17 cmp $0x17,%rdx : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 0.00 : 4a0da6: 49 89 fc mov %rdi,%r12 0.00 : 4a0da9: 55 push %rbp 3.37 : 4a0daa: 53 push %rbx : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.08 : 4a0dab: 76 73 jbe 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dad: 80 7c 16 eb 00 cmpb $0x0,-0x15(%rsi,%rdx,1) 3.48 : 4a0db2: 75 6c jne 4a0e20 <decode_tree_entry+0x80> : static const char *get_mode(const char *str, unsigned int *modep) : { : unsigned char c; : unsigned int mode = 0; : : if (*str == ' ') 1.94 : 4a0db4: 0f b6 06 movzbl (%rsi),%eax 0.39 : 4a0db7: 3c 20 cmp $0x20,%al 0.00 : 4a0db9: 74 65 je 4a0e20 <decode_tree_entry+0x80> : return NULL; : : while ((c = *str++) != ' ') { 0.06 : 4a0dbb: 89 c2 mov %eax,%edx : if (c < '0' || c > '7') 1.99 : 4a0dbd: 31 ed xor %ebp,%ebp : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 1.74 : 4a0dbf: 48 8d 5e 01 lea 0x1(%rsi),%rbx : if (c < '0' || c > '7') 0.00 : 4a0dc3: 8d 42 d0 lea -0x30(%rdx),%eax 0.17 : 4a0dc6: 3c 07 cmp $0x7,%al 0.00 : 4a0dc8: 76 0d jbe 4a0dd7 <decode_tree_entry+0x37> 0.00 : 4a0dca: eb 54 jmp 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dcc: 0f 1f 40 00 nopl 0x0(%rax) 16.57 : 4a0dd0: 8d 42 d0 lea -0x30(%rdx),%eax 0.14 : 4a0dd3: 3c 07 cmp $0x7,%al 0.00 : 4a0dd5: 77 49 ja 4a0e20 <decode_tree_entry+0x80> : return NULL; : mode = (mode << 3) + (c - '0'); 3.12 : 4a0dd7: 0f b6 c2 movzbl %dl,%eax : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 0.00 : 4a0dda: 0f b6 13 movzbl (%rbx),%edx 16.74 : 4a0ddd: 48 83 c3 01 add $0x1,%rbx : if (c < '0' || c > '7') : return NULL; : mode = (mode << 3) + (c - '0'); The first column is the percentage of samples that arrived on that particular line - relative to the total cost of the function. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-06-06 21:48:52 +08:00
return NULL;
if (symbol_conf.priv_size)
self = ((void *)self) + symbol_conf.priv_size;
perf_counter tools: Add 'perf annotate' feature Add new perf sub-command to display annotated source code: $ perf annotate decode_tree_entry ------------------------------------------------ Percent | Source code & Disassembly of /home/mingo/git/git ------------------------------------------------ : : /home/mingo/git/git: file format elf64-x86-64 : : : Disassembly of section .text: : : 00000000004a0da0 <decode_tree_entry>: : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 3.82 : 4a0da0: 41 54 push %r12 : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.17 : 4a0da2: 48 83 fa 17 cmp $0x17,%rdx : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 0.00 : 4a0da6: 49 89 fc mov %rdi,%r12 0.00 : 4a0da9: 55 push %rbp 3.37 : 4a0daa: 53 push %rbx : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.08 : 4a0dab: 76 73 jbe 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dad: 80 7c 16 eb 00 cmpb $0x0,-0x15(%rsi,%rdx,1) 3.48 : 4a0db2: 75 6c jne 4a0e20 <decode_tree_entry+0x80> : static const char *get_mode(const char *str, unsigned int *modep) : { : unsigned char c; : unsigned int mode = 0; : : if (*str == ' ') 1.94 : 4a0db4: 0f b6 06 movzbl (%rsi),%eax 0.39 : 4a0db7: 3c 20 cmp $0x20,%al 0.00 : 4a0db9: 74 65 je 4a0e20 <decode_tree_entry+0x80> : return NULL; : : while ((c = *str++) != ' ') { 0.06 : 4a0dbb: 89 c2 mov %eax,%edx : if (c < '0' || c > '7') 1.99 : 4a0dbd: 31 ed xor %ebp,%ebp : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 1.74 : 4a0dbf: 48 8d 5e 01 lea 0x1(%rsi),%rbx : if (c < '0' || c > '7') 0.00 : 4a0dc3: 8d 42 d0 lea -0x30(%rdx),%eax 0.17 : 4a0dc6: 3c 07 cmp $0x7,%al 0.00 : 4a0dc8: 76 0d jbe 4a0dd7 <decode_tree_entry+0x37> 0.00 : 4a0dca: eb 54 jmp 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dcc: 0f 1f 40 00 nopl 0x0(%rax) 16.57 : 4a0dd0: 8d 42 d0 lea -0x30(%rdx),%eax 0.14 : 4a0dd3: 3c 07 cmp $0x7,%al 0.00 : 4a0dd5: 77 49 ja 4a0e20 <decode_tree_entry+0x80> : return NULL; : mode = (mode << 3) + (c - '0'); 3.12 : 4a0dd7: 0f b6 c2 movzbl %dl,%eax : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 0.00 : 4a0dda: 0f b6 13 movzbl (%rbx),%edx 16.74 : 4a0ddd: 48 83 c3 01 add $0x1,%rbx : if (c < '0' || c > '7') : return NULL; : mode = (mode << 3) + (c - '0'); The first column is the percentage of samples that arrived on that particular line - relative to the total cost of the function. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-06-06 21:48:52 +08:00
self->start = start;
self->end = len ? start + len - 1 : start;
pr_debug3("%s: %s %#Lx-%#Lx\n", __func__, name, start, self->end);
perf_counter tools: Add 'perf annotate' feature Add new perf sub-command to display annotated source code: $ perf annotate decode_tree_entry ------------------------------------------------ Percent | Source code & Disassembly of /home/mingo/git/git ------------------------------------------------ : : /home/mingo/git/git: file format elf64-x86-64 : : : Disassembly of section .text: : : 00000000004a0da0 <decode_tree_entry>: : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 3.82 : 4a0da0: 41 54 push %r12 : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.17 : 4a0da2: 48 83 fa 17 cmp $0x17,%rdx : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 0.00 : 4a0da6: 49 89 fc mov %rdi,%r12 0.00 : 4a0da9: 55 push %rbp 3.37 : 4a0daa: 53 push %rbx : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.08 : 4a0dab: 76 73 jbe 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dad: 80 7c 16 eb 00 cmpb $0x0,-0x15(%rsi,%rdx,1) 3.48 : 4a0db2: 75 6c jne 4a0e20 <decode_tree_entry+0x80> : static const char *get_mode(const char *str, unsigned int *modep) : { : unsigned char c; : unsigned int mode = 0; : : if (*str == ' ') 1.94 : 4a0db4: 0f b6 06 movzbl (%rsi),%eax 0.39 : 4a0db7: 3c 20 cmp $0x20,%al 0.00 : 4a0db9: 74 65 je 4a0e20 <decode_tree_entry+0x80> : return NULL; : : while ((c = *str++) != ' ') { 0.06 : 4a0dbb: 89 c2 mov %eax,%edx : if (c < '0' || c > '7') 1.99 : 4a0dbd: 31 ed xor %ebp,%ebp : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 1.74 : 4a0dbf: 48 8d 5e 01 lea 0x1(%rsi),%rbx : if (c < '0' || c > '7') 0.00 : 4a0dc3: 8d 42 d0 lea -0x30(%rdx),%eax 0.17 : 4a0dc6: 3c 07 cmp $0x7,%al 0.00 : 4a0dc8: 76 0d jbe 4a0dd7 <decode_tree_entry+0x37> 0.00 : 4a0dca: eb 54 jmp 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dcc: 0f 1f 40 00 nopl 0x0(%rax) 16.57 : 4a0dd0: 8d 42 d0 lea -0x30(%rdx),%eax 0.14 : 4a0dd3: 3c 07 cmp $0x7,%al 0.00 : 4a0dd5: 77 49 ja 4a0e20 <decode_tree_entry+0x80> : return NULL; : mode = (mode << 3) + (c - '0'); 3.12 : 4a0dd7: 0f b6 c2 movzbl %dl,%eax : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 0.00 : 4a0dda: 0f b6 13 movzbl (%rbx),%edx 16.74 : 4a0ddd: 48 83 c3 01 add $0x1,%rbx : if (c < '0' || c > '7') : return NULL; : mode = (mode << 3) + (c - '0'); The first column is the percentage of samples that arrived on that particular line - relative to the total cost of the function. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-06-06 21:48:52 +08:00
memcpy(self->name, name, namelen);
return self;
}
static void symbol__delete(struct symbol *self)
{
free(((void *)self) - symbol_conf.priv_size);
}
static size_t symbol__fprintf(struct symbol *self, FILE *fp)
{
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
return fprintf(fp, " %llx-%llx %s\n",
self->start, self->end, self->name);
}
void dso__set_long_name(struct dso *self, char *name)
{
if (name == NULL)
return;
self->long_name = name;
self->long_name_len = strlen(name);
}
static void dso__set_basename(struct dso *self)
{
self->short_name = basename(self->long_name);
}
struct dso *dso__new(const char *name)
{
struct dso *self = zalloc(sizeof(*self) + strlen(name) + 1);
if (self != NULL) {
int i;
strcpy(self->name, name);
dso__set_long_name(self, self->name);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
self->short_name = self->name;
for (i = 0; i < MAP__NR_TYPES; ++i)
self->symbols[i] = self->symbol_names[i] = RB_ROOT;
perf report: Adjust column width to the values sampled Auto-adjust column width of perf report output to the longest occuring string length. Example: [acme@doppio pahole]$ perf report --sort comm,dso,symbol | head -13 12.79% pahole /usr/lib64/libdw-0.141.so [.] __libdw_find_attr 8.90% pahole /lib64/libc-2.10.1.so [.] _int_malloc 8.68% pahole /usr/lib64/libdw-0.141.so [.] __libdw_form_val_len 8.15% pahole /lib64/libc-2.10.1.so [.] __GI_strcmp 6.80% pahole /lib64/libc-2.10.1.so [.] __tsearch 5.54% pahole ./build/libdwarves.so.1.0.0 [.] tag__recode_dwarf_type [acme@doppio pahole]$ [acme@doppio pahole]$ perf report --sort comm,dso,symbol -d /lib64/libc-2.10.1.so | head -10 21.92% pahole /lib64/libc-2.10.1.so [.] _int_malloc 20.08% pahole /lib64/libc-2.10.1.so [.] __GI_strcmp 16.75% pahole /lib64/libc-2.10.1.so [.] __tsearch [acme@doppio pahole]$ Also add these extra options to control the new behaviour: -w, --field-width Force each column width to the provided list, for large terminal readability. -t, --field-separator: Use a special separator character and don't pad with spaces, replacing all occurances of this separator in symbol names (and other output) with a '.' character, that thus it's the only non valid separator. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <20090711014728.GH3452@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-11 09:47:28 +08:00
self->slen_calculated = 0;
self->origin = DSO__ORIG_NOT_FOUND;
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
self->loaded = 0;
self->sorted_by_name = 0;
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
self->has_build_id = 0;
}
return self;
}
static void symbols__delete(struct rb_root *self)
{
struct symbol *pos;
struct rb_node *next = rb_first(self);
while (next) {
pos = rb_entry(next, struct symbol, rb_node);
next = rb_next(&pos->rb_node);
rb_erase(&pos->rb_node, self);
symbol__delete(pos);
}
}
void dso__delete(struct dso *self)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i)
symbols__delete(&self->symbols[i]);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (self->long_name != self->name)
free(self->long_name);
free(self);
}
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
void dso__set_build_id(struct dso *self, void *build_id)
{
memcpy(self->build_id, build_id, sizeof(self->build_id));
self->has_build_id = 1;
}
static void symbols__insert(struct rb_root *self, struct symbol *sym)
{
struct rb_node **p = &self->rb_node;
struct rb_node *parent = NULL;
const u64 ip = sym->start;
struct symbol *s;
while (*p != NULL) {
parent = *p;
s = rb_entry(parent, struct symbol, rb_node);
if (ip < s->start)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&sym->rb_node, parent, p);
rb_insert_color(&sym->rb_node, self);
}
static struct symbol *symbols__find(struct rb_root *self, u64 ip)
{
struct rb_node *n;
if (self == NULL)
return NULL;
n = self->rb_node;
while (n) {
struct symbol *s = rb_entry(n, struct symbol, rb_node);
if (ip < s->start)
n = n->rb_left;
else if (ip > s->end)
n = n->rb_right;
else
return s;
}
return NULL;
}
struct symbol_name_rb_node {
struct rb_node rb_node;
struct symbol sym;
};
static void symbols__insert_by_name(struct rb_root *self, struct symbol *sym)
{
struct rb_node **p = &self->rb_node;
struct rb_node *parent = NULL;
struct symbol_name_rb_node *symn = ((void *)sym) - sizeof(*parent), *s;
while (*p != NULL) {
parent = *p;
s = rb_entry(parent, struct symbol_name_rb_node, rb_node);
if (strcmp(sym->name, s->sym.name) < 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&symn->rb_node, parent, p);
rb_insert_color(&symn->rb_node, self);
}
static void symbols__sort_by_name(struct rb_root *self, struct rb_root *source)
{
struct rb_node *nd;
for (nd = rb_first(source); nd; nd = rb_next(nd)) {
struct symbol *pos = rb_entry(nd, struct symbol, rb_node);
symbols__insert_by_name(self, pos);
}
}
static struct symbol *symbols__find_by_name(struct rb_root *self, const char *name)
{
struct rb_node *n;
if (self == NULL)
return NULL;
n = self->rb_node;
while (n) {
struct symbol_name_rb_node *s;
int cmp;
s = rb_entry(n, struct symbol_name_rb_node, rb_node);
cmp = strcmp(name, s->sym.name);
if (cmp < 0)
n = n->rb_left;
else if (cmp > 0)
n = n->rb_right;
else
return &s->sym;
}
return NULL;
}
struct symbol *dso__find_symbol(struct dso *self,
enum map_type type, u64 addr)
{
return symbols__find(&self->symbols[type], addr);
}
struct symbol *dso__find_symbol_by_name(struct dso *self, enum map_type type,
const char *name)
{
return symbols__find_by_name(&self->symbol_names[type], name);
}
void dso__sort_by_name(struct dso *self, enum map_type type)
{
dso__set_sorted_by_name(self, type);
return symbols__sort_by_name(&self->symbol_names[type],
&self->symbols[type]);
}
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
int build_id__sprintf(u8 *self, int len, char *bf)
{
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
char *bid = bf;
u8 *raw = self;
int i;
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
for (i = 0; i < len; ++i) {
sprintf(bid, "%02x", *raw);
++raw;
bid += 2;
}
return raw - self;
}
size_t dso__fprintf_buildid(struct dso *self, FILE *fp)
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
{
char sbuild_id[BUILD_ID_SIZE * 2 + 1];
build_id__sprintf(self->build_id, sizeof(self->build_id), sbuild_id);
return fprintf(fp, "%s", sbuild_id);
}
size_t dso__fprintf(struct dso *self, enum map_type type, FILE *fp)
{
struct rb_node *nd;
size_t ret = fprintf(fp, "dso: %s (", self->short_name);
ret += dso__fprintf_buildid(self, fp);
ret += fprintf(fp, ")\n");
for (nd = rb_first(&self->symbols[type]); nd; nd = rb_next(nd)) {
struct symbol *pos = rb_entry(nd, struct symbol, rb_node);
ret += symbol__fprintf(pos, fp);
}
return ret;
}
int kallsyms__parse(const char *filename, void *arg,
int (*process_symbol)(void *arg, const char *name,
char type, u64 start))
{
char *line = NULL;
size_t n;
int err = 0;
FILE *file = fopen(filename, "r");
if (file == NULL)
goto out_failure;
while (!feof(file)) {
u64 start;
int line_len, len;
char symbol_type;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
char *symbol_name;
line_len = getline(&line, &n, file);
if (line_len < 0)
break;
if (!line)
goto out_failure;
line[--line_len] = '\0'; /* \n */
len = hex2u64(line, &start);
len++;
if (len + 2 >= line_len)
continue;
symbol_type = toupper(line[len]);
symbol_name = line + len + 2;
err = process_symbol(arg, symbol_name, symbol_type, start);
if (err)
break;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
}
free(line);
fclose(file);
return err;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
out_failure:
return -1;
}
struct process_kallsyms_args {
struct map *map;
struct dso *dso;
};
static int map__process_kallsym_symbol(void *arg, const char *name,
char type, u64 start)
{
struct symbol *sym;
struct process_kallsyms_args *a = arg;
struct rb_root *root = &a->dso->symbols[a->map->type];
if (!symbol_type__is_a(type, a->map->type))
return 0;
/*
* Will fix up the end later, when we have all symbols sorted.
*/
sym = symbol__new(start, 0, name);
if (sym == NULL)
return -ENOMEM;
/*
* We will pass the symbols to the filter later, in
* map__split_kallsyms, when we have split the maps per module
*/
symbols__insert(root, sym);
return 0;
}
/*
* Loads the function entries in /proc/kallsyms into kernel_map->dso,
* so that we can in the next step set the symbol ->end address and then
* call kernel_maps__split_kallsyms.
*/
static int dso__load_all_kallsyms(struct dso *self, const char *filename,
struct map *map)
{
struct process_kallsyms_args args = { .map = map, .dso = self, };
return kallsyms__parse(filename, &args, map__process_kallsym_symbol);
}
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
/*
* Split the symbols into maps, making sure there are no overlaps, i.e. the
* kernel range is broken in several maps, named [kernel].N, as we don't have
* the original ELF section names vmlinux have.
*/
static int dso__split_kallsyms(struct dso *self, struct map *map,
struct perf_session *session, symbol_filter_t filter)
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
{
struct map *curr_map = map;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
struct symbol *pos;
int count = 0;
struct rb_root *root = &self->symbols[map->type];
struct rb_node *next = rb_first(root);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
int kernel_range = 0;
while (next) {
char *module;
pos = rb_entry(next, struct symbol, rb_node);
next = rb_next(&pos->rb_node);
module = strchr(pos->name, '\t');
if (module) {
if (!symbol_conf.use_modules)
goto discard_symbol;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
*module++ = '\0';
if (strcmp(curr_map->dso->short_name, module)) {
curr_map = map_groups__find_by_name(&session->kmaps, map->type, module);
if (curr_map == NULL) {
pr_debug("/proc/{kallsyms,modules} "
"inconsistency while looking "
"for \"%s\" module!\n", module);
return -1;
}
if (curr_map->dso->loaded)
goto discard_symbol;
}
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
/*
* So that we look just like we get from .ko files,
* i.e. not prelinked, relative to map->start.
*/
pos->start = curr_map->map_ip(curr_map, pos->start);
pos->end = curr_map->map_ip(curr_map, pos->end);
} else if (curr_map != map) {
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
char dso_name[PATH_MAX];
struct dso *dso;
snprintf(dso_name, sizeof(dso_name), "[kernel].%d",
kernel_range++);
dso = dso__new(dso_name);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (dso == NULL)
return -1;
curr_map = map__new2(pos->start, dso, map->type);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (map == NULL) {
dso__delete(dso);
return -1;
}
curr_map->map_ip = curr_map->unmap_ip = identity__map_ip;
map_groups__insert(&session->kmaps, curr_map);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
++kernel_range;
}
if (filter && filter(curr_map, pos)) {
discard_symbol: rb_erase(&pos->rb_node, root);
symbol__delete(pos);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
} else {
if (curr_map != map) {
rb_erase(&pos->rb_node, root);
symbols__insert(&curr_map->dso->symbols[curr_map->type], pos);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
}
count++;
}
}
return count;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
}
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
static int dso__load_kallsyms(struct dso *self, const char *filename, struct map *map,
struct perf_session *session, symbol_filter_t filter)
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
{
if (dso__load_all_kallsyms(self, filename, map) < 0)
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
return -1;
symbols__fixup_end(&self->symbols[map->type]);
self->origin = DSO__ORIG_KERNEL;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
return dso__split_kallsyms(self, map, session, filter);
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
static int dso__load_perf_map(struct dso *self, struct map *map,
symbol_filter_t filter)
{
char *line = NULL;
size_t n;
FILE *file;
int nr_syms = 0;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
file = fopen(self->long_name, "r");
if (file == NULL)
goto out_failure;
while (!feof(file)) {
u64 start, size;
struct symbol *sym;
int line_len, len;
line_len = getline(&line, &n, file);
if (line_len < 0)
break;
if (!line)
goto out_failure;
line[--line_len] = '\0'; /* \n */
len = hex2u64(line, &start);
len++;
if (len + 2 >= line_len)
continue;
len += hex2u64(line + len, &size);
len++;
if (len + 2 >= line_len)
continue;
sym = symbol__new(start, size, line + len);
if (sym == NULL)
goto out_delete_line;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (filter && filter(map, sym))
symbol__delete(sym);
else {
symbols__insert(&self->symbols[map->type], sym);
nr_syms++;
}
}
free(line);
fclose(file);
return nr_syms;
out_delete_line:
free(line);
out_failure:
return -1;
}
/**
* elf_symtab__for_each_symbol - iterate thru all the symbols
*
* @self: struct elf_symtab instance to iterate
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
* @idx: uint32_t idx
* @sym: GElf_Sym iterator
*/
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
#define elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) \
for (idx = 0, gelf_getsym(syms, idx, &sym);\
idx < nr_syms; \
idx++, gelf_getsym(syms, idx, &sym))
static inline uint8_t elf_sym__type(const GElf_Sym *sym)
{
return GELF_ST_TYPE(sym->st_info);
}
static inline int elf_sym__is_function(const GElf_Sym *sym)
{
return elf_sym__type(sym) == STT_FUNC &&
sym->st_name != 0 &&
sym->st_shndx != SHN_UNDEF;
}
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
static inline bool elf_sym__is_object(const GElf_Sym *sym)
{
return elf_sym__type(sym) == STT_OBJECT &&
sym->st_name != 0 &&
sym->st_shndx != SHN_UNDEF;
}
static inline int elf_sym__is_label(const GElf_Sym *sym)
{
return elf_sym__type(sym) == STT_NOTYPE &&
sym->st_name != 0 &&
sym->st_shndx != SHN_UNDEF &&
sym->st_shndx != SHN_ABS;
}
static inline const char *elf_sec__name(const GElf_Shdr *shdr,
const Elf_Data *secstrs)
{
return secstrs->d_buf + shdr->sh_name;
}
static inline int elf_sec__is_text(const GElf_Shdr *shdr,
const Elf_Data *secstrs)
{
return strstr(elf_sec__name(shdr, secstrs), "text") != NULL;
}
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
static inline bool elf_sec__is_data(const GElf_Shdr *shdr,
const Elf_Data *secstrs)
{
return strstr(elf_sec__name(shdr, secstrs), "data") != NULL;
}
static inline const char *elf_sym__name(const GElf_Sym *sym,
const Elf_Data *symstrs)
{
return symstrs->d_buf + sym->st_name;
}
static Elf_Scn *elf_section_by_name(Elf *elf, GElf_Ehdr *ep,
GElf_Shdr *shp, const char *name,
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
size_t *idx)
{
Elf_Scn *sec = NULL;
size_t cnt = 1;
while ((sec = elf_nextscn(elf, sec)) != NULL) {
char *str;
gelf_getshdr(sec, shp);
str = elf_strptr(elf, ep->e_shstrndx, shp->sh_name);
if (!strcmp(name, str)) {
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
if (idx)
*idx = cnt;
break;
}
++cnt;
}
return sec;
}
#define elf_section__for_each_rel(reldata, pos, pos_mem, idx, nr_entries) \
for (idx = 0, pos = gelf_getrel(reldata, 0, &pos_mem); \
idx < nr_entries; \
++idx, pos = gelf_getrel(reldata, idx, &pos_mem))
#define elf_section__for_each_rela(reldata, pos, pos_mem, idx, nr_entries) \
for (idx = 0, pos = gelf_getrela(reldata, 0, &pos_mem); \
idx < nr_entries; \
++idx, pos = gelf_getrela(reldata, idx, &pos_mem))
/*
* We need to check if we have a .dynsym, so that we can handle the
* .plt, synthesizing its symbols, that aren't on the symtabs (be it
* .dynsym or .symtab).
* And always look at the original dso, not at debuginfo packages, that
* have the PLT data stripped out (shdr_rel_plt.sh_type == SHT_NOBITS).
*/
static int dso__synthesize_plt_symbols(struct dso *self, struct map *map,
symbol_filter_t filter)
{
uint32_t nr_rel_entries, idx;
GElf_Sym sym;
u64 plt_offset;
GElf_Shdr shdr_plt;
struct symbol *f;
GElf_Shdr shdr_rel_plt, shdr_dynsym;
Elf_Data *reldata, *syms, *symstrs;
Elf_Scn *scn_plt_rel, *scn_symstrs, *scn_dynsym;
size_t dynsym_idx;
GElf_Ehdr ehdr;
char sympltname[1024];
Elf *elf;
int nr = 0, symidx, fd, err = 0;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
fd = open(self->long_name, O_RDONLY);
if (fd < 0)
goto out;
elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
if (elf == NULL)
goto out_close;
if (gelf_getehdr(elf, &ehdr) == NULL)
goto out_elf_end;
scn_dynsym = elf_section_by_name(elf, &ehdr, &shdr_dynsym,
".dynsym", &dynsym_idx);
if (scn_dynsym == NULL)
goto out_elf_end;
scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
".rela.plt", NULL);
if (scn_plt_rel == NULL) {
scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
".rel.plt", NULL);
if (scn_plt_rel == NULL)
goto out_elf_end;
}
err = -1;
if (shdr_rel_plt.sh_link != dynsym_idx)
goto out_elf_end;
if (elf_section_by_name(elf, &ehdr, &shdr_plt, ".plt", NULL) == NULL)
goto out_elf_end;
/*
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
* Fetch the relocation section to find the idxes to the GOT
* and the symbols in the .dynsym they refer to.
*/
reldata = elf_getdata(scn_plt_rel, NULL);
if (reldata == NULL)
goto out_elf_end;
syms = elf_getdata(scn_dynsym, NULL);
if (syms == NULL)
goto out_elf_end;
scn_symstrs = elf_getscn(elf, shdr_dynsym.sh_link);
if (scn_symstrs == NULL)
goto out_elf_end;
symstrs = elf_getdata(scn_symstrs, NULL);
if (symstrs == NULL)
goto out_elf_end;
nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
plt_offset = shdr_plt.sh_offset;
if (shdr_rel_plt.sh_type == SHT_RELA) {
GElf_Rela pos_mem, *pos;
elf_section__for_each_rela(reldata, pos, pos_mem, idx,
nr_rel_entries) {
symidx = GELF_R_SYM(pos->r_info);
plt_offset += shdr_plt.sh_entsize;
gelf_getsym(syms, symidx, &sym);
snprintf(sympltname, sizeof(sympltname),
"%s@plt", elf_sym__name(&sym, symstrs));
f = symbol__new(plt_offset, shdr_plt.sh_entsize,
sympltname);
if (!f)
goto out_elf_end;
if (filter && filter(map, f))
symbol__delete(f);
else {
symbols__insert(&self->symbols[map->type], f);
++nr;
}
}
} else if (shdr_rel_plt.sh_type == SHT_REL) {
GElf_Rel pos_mem, *pos;
elf_section__for_each_rel(reldata, pos, pos_mem, idx,
nr_rel_entries) {
symidx = GELF_R_SYM(pos->r_info);
plt_offset += shdr_plt.sh_entsize;
gelf_getsym(syms, symidx, &sym);
snprintf(sympltname, sizeof(sympltname),
"%s@plt", elf_sym__name(&sym, symstrs));
f = symbol__new(plt_offset, shdr_plt.sh_entsize,
sympltname);
if (!f)
goto out_elf_end;
if (filter && filter(map, f))
symbol__delete(f);
else {
symbols__insert(&self->symbols[map->type], f);
++nr;
}
}
}
err = 0;
out_elf_end:
elf_end(elf);
out_close:
close(fd);
if (err == 0)
return nr;
out:
pr_warning("%s: problems reading %s PLT info.\n",
__func__, self->long_name);
return 0;
}
static bool elf_sym__is_a(GElf_Sym *self, enum map_type type)
{
switch (type) {
case MAP__FUNCTION:
return elf_sym__is_function(self);
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
case MAP__VARIABLE:
return elf_sym__is_object(self);
default:
return false;
}
}
static bool elf_sec__is_a(GElf_Shdr *self, Elf_Data *secstrs, enum map_type type)
{
switch (type) {
case MAP__FUNCTION:
return elf_sec__is_text(self, secstrs);
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
case MAP__VARIABLE:
return elf_sec__is_data(self, secstrs);
default:
return false;
}
}
static int dso__load_sym(struct dso *self, struct map *map,
struct perf_session *session, const char *name, int fd,
symbol_filter_t filter, int kernel, int kmodule)
{
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
struct map *curr_map = map;
struct dso *curr_dso = self;
size_t dso_name_len = strlen(self->short_name);
Elf_Data *symstrs, *secstrs;
uint32_t nr_syms;
int err = -1;
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
uint32_t idx;
GElf_Ehdr ehdr;
GElf_Shdr shdr;
Elf_Data *syms;
GElf_Sym sym;
Elf_Scn *sec, *sec_strndx;
Elf *elf;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
int nr = 0;
elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
if (elf == NULL) {
pr_err("%s: cannot read %s ELF file.\n", __func__, name);
goto out_close;
}
if (gelf_getehdr(elf, &ehdr) == NULL) {
pr_err("%s: cannot get elf header.\n", __func__);
goto out_elf_end;
}
sec = elf_section_by_name(elf, &ehdr, &shdr, ".symtab", NULL);
if (sec == NULL) {
sec = elf_section_by_name(elf, &ehdr, &shdr, ".dynsym", NULL);
if (sec == NULL)
goto out_elf_end;
}
syms = elf_getdata(sec, NULL);
if (syms == NULL)
goto out_elf_end;
sec = elf_getscn(elf, shdr.sh_link);
if (sec == NULL)
goto out_elf_end;
symstrs = elf_getdata(sec, NULL);
if (symstrs == NULL)
goto out_elf_end;
sec_strndx = elf_getscn(elf, ehdr.e_shstrndx);
if (sec_strndx == NULL)
goto out_elf_end;
secstrs = elf_getdata(sec_strndx, NULL);
if (secstrs == NULL)
goto out_elf_end;
nr_syms = shdr.sh_size / shdr.sh_entsize;
memset(&sym, 0, sizeof(sym));
if (!kernel) {
self->adjust_symbols = (ehdr.e_type == ET_EXEC ||
perf_counter tools: Adjust symbols in ET_EXEC files too Ingo Molnar wrote: > i just bisected a 'perf report' bug that would cause us to not > resolve all user-space symbols in a 'git gc' run to: > > f5812a7a336fb952d819e4427b9a2dce02368e82 is first bad commit > commit f5812a7a336fb952d819e4427b9a2dce02368e82 > Author: Arnaldo Carvalho de Melo <acme@redhat.com> > Date: Tue Jun 30 11:43:17 2009 -0300 > > perf_counter tools: Adjust only prelinked symbol's addresses Rename ->prelinked to ->adjust_symbols and making what was done only for prelinked libraries also to ET_EXEC binaries, such as /usr/bin/git: [acme@doppio pahole]$ readelf -h /usr/bin/git | grep Type Type: EXEC (Executable file) [acme@doppio pahole]$ And after installing the 'git-debuginfo' package, I get correct results: [acme@doppio linux-2.6-tip]$ perf report --sort comm,dso,symbol -d /usr/bin/git | head -20 # # (1139614 samples) # # Overhead Command Shared Object Symbol # ........ ................ ......................... ...... # 34.98% git /usr/bin/git [.] send_sideband 33.39% git /usr/bin/git [.] enter_repo 6.81% git /usr/bin/git [.] diff_opt_parse 4.95% git /usr/bin/git [.] is_repository_shallow 3.24% git /usr/bin/git [.] odb_mkstemp 1.39% git /usr/bin/git [.] output 1.34% git /usr/bin/git [.] xmmap 1.25% git /usr/bin/git [.] receive_pack_config 1.16% git /usr/bin/git [.] git_pathdup 0.90% git /usr/bin/git [.] read_object_with_reference 0.86% git /usr/bin/git [.] show_patch_diff 0.85% git /usr/bin/git 0x00000000095e2e 0.69% git /usr/bin/git [.] display [acme@doppio linux-2.6-tip]$ I'll check what are the last cases where we can't resolve symbols, like this 0x00000000095e2e later. And I guess this will fix the problems Mike were seeing too: [acme@doppio linux-2.6-tip]$ readelf -h ../build/perf/vmlinux | grep Type Type: EXEC (Executable file) [acme@doppio linux-2.6-tip]$ Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-03 08:24:14 +08:00
elf_section_by_name(elf, &ehdr, &shdr,
".gnu.prelink_undo",
NULL) != NULL);
} else self->adjust_symbols = 0;
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) {
struct symbol *f;
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
const char *elf_name = elf_sym__name(&sym, symstrs);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
char *demangled = NULL;
int is_label = elf_sym__is_label(&sym);
const char *section_name;
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
if (kernel && session->ref_reloc_sym.name != NULL &&
strcmp(elf_name, session->ref_reloc_sym.name) == 0)
perf_session__reloc_vmlinux_maps(session, sym.st_value);
if (!is_label && !elf_sym__is_a(&sym, map->type))
continue;
sec = elf_getscn(elf, sym.st_shndx);
if (!sec)
goto out_elf_end;
gelf_getshdr(sec, &shdr);
if (is_label && !elf_sec__is_a(&shdr, secstrs, map->type))
continue;
section_name = elf_sec__name(&shdr, secstrs);
perf_counter tools: Add 'perf annotate' feature Add new perf sub-command to display annotated source code: $ perf annotate decode_tree_entry ------------------------------------------------ Percent | Source code & Disassembly of /home/mingo/git/git ------------------------------------------------ : : /home/mingo/git/git: file format elf64-x86-64 : : : Disassembly of section .text: : : 00000000004a0da0 <decode_tree_entry>: : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 3.82 : 4a0da0: 41 54 push %r12 : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.17 : 4a0da2: 48 83 fa 17 cmp $0x17,%rdx : *modep = mode; : return str; : } : : static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) : { 0.00 : 4a0da6: 49 89 fc mov %rdi,%r12 0.00 : 4a0da9: 55 push %rbp 3.37 : 4a0daa: 53 push %rbx : const char *path; : unsigned int mode, len; : : if (size < 24 || buf[size - 21]) 0.08 : 4a0dab: 76 73 jbe 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dad: 80 7c 16 eb 00 cmpb $0x0,-0x15(%rsi,%rdx,1) 3.48 : 4a0db2: 75 6c jne 4a0e20 <decode_tree_entry+0x80> : static const char *get_mode(const char *str, unsigned int *modep) : { : unsigned char c; : unsigned int mode = 0; : : if (*str == ' ') 1.94 : 4a0db4: 0f b6 06 movzbl (%rsi),%eax 0.39 : 4a0db7: 3c 20 cmp $0x20,%al 0.00 : 4a0db9: 74 65 je 4a0e20 <decode_tree_entry+0x80> : return NULL; : : while ((c = *str++) != ' ') { 0.06 : 4a0dbb: 89 c2 mov %eax,%edx : if (c < '0' || c > '7') 1.99 : 4a0dbd: 31 ed xor %ebp,%ebp : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 1.74 : 4a0dbf: 48 8d 5e 01 lea 0x1(%rsi),%rbx : if (c < '0' || c > '7') 0.00 : 4a0dc3: 8d 42 d0 lea -0x30(%rdx),%eax 0.17 : 4a0dc6: 3c 07 cmp $0x7,%al 0.00 : 4a0dc8: 76 0d jbe 4a0dd7 <decode_tree_entry+0x37> 0.00 : 4a0dca: eb 54 jmp 4a0e20 <decode_tree_entry+0x80> 0.00 : 4a0dcc: 0f 1f 40 00 nopl 0x0(%rax) 16.57 : 4a0dd0: 8d 42 d0 lea -0x30(%rdx),%eax 0.14 : 4a0dd3: 3c 07 cmp $0x7,%al 0.00 : 4a0dd5: 77 49 ja 4a0e20 <decode_tree_entry+0x80> : return NULL; : mode = (mode << 3) + (c - '0'); 3.12 : 4a0dd7: 0f b6 c2 movzbl %dl,%eax : unsigned int mode = 0; : : if (*str == ' ') : return NULL; : : while ((c = *str++) != ' ') { 0.00 : 4a0dda: 0f b6 13 movzbl (%rbx),%edx 16.74 : 4a0ddd: 48 83 c3 01 add $0x1,%rbx : if (c < '0' || c > '7') : return NULL; : mode = (mode << 3) + (c - '0'); The first column is the percentage of samples that arrived on that particular line - relative to the total cost of the function. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-06-06 21:48:52 +08:00
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (kernel || kmodule) {
char dso_name[PATH_MAX];
if (strcmp(section_name,
curr_dso->short_name + dso_name_len) == 0)
goto new_symbol;
if (strcmp(section_name, ".text") == 0) {
curr_map = map;
curr_dso = self;
goto new_symbol;
}
snprintf(dso_name, sizeof(dso_name),
"%s%s", self->short_name, section_name);
curr_map = map_groups__find_by_name(&session->kmaps, map->type, dso_name);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (curr_map == NULL) {
u64 start = sym.st_value;
if (kmodule)
start += map->start + shdr.sh_offset;
curr_dso = dso__new(dso_name);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (curr_dso == NULL)
goto out_elf_end;
curr_map = map__new2(start, curr_dso,
MAP__FUNCTION);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (curr_map == NULL) {
dso__delete(curr_dso);
goto out_elf_end;
}
curr_map->map_ip = identity__map_ip;
curr_map->unmap_ip = identity__map_ip;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
curr_dso->origin = DSO__ORIG_KERNEL;
map_groups__insert(&session->kmaps, curr_map);
dsos__add(&dsos__kernel, curr_dso);
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
} else
curr_dso = curr_map->dso;
goto new_symbol;
}
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (curr_dso->adjust_symbols) {
pr_debug2("adjusting symbol: st_value: %Lx sh_addr: "
"%Lx sh_offset: %Lx\n", (u64)sym.st_value,
(u64)shdr.sh_addr, (u64)shdr.sh_offset);
sym.st_value -= shdr.sh_addr - shdr.sh_offset;
}
/*
* We need to figure out if the object was created from C++ sources
* DWARF DW_compile_unit has this, but we don't always have access
* to it...
*/
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
demangled = bfd_demangle(NULL, elf_name, DMGL_PARAMS | DMGL_ANSI);
if (demangled != NULL)
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
elf_name = demangled;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
new_symbol:
f = symbol__new(sym.st_value, sym.st_size, elf_name);
free(demangled);
if (!f)
goto out_elf_end;
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
if (filter && filter(curr_map, f))
symbol__delete(f);
else {
symbols__insert(&curr_dso->symbols[curr_map->type], f);
nr++;
}
}
perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-08 00:48:56 +08:00
/*
* For misannotated, zeroed, ASM function sizes.
*/
if (nr > 0)
symbols__fixup_end(&self->symbols[map->type]);
err = nr;
out_elf_end:
elf_end(elf);
out_close:
return err;
}
static bool dso__build_id_equal(const struct dso *self, u8 *build_id)
{
return memcmp(self->build_id, build_id, sizeof(self->build_id)) == 0;
}
static bool __dsos__read_build_ids(struct list_head *head)
{
bool have_build_id = false;
struct dso *pos;
list_for_each_entry(pos, head, node)
if (filename__read_build_id(pos->long_name, pos->build_id,
sizeof(pos->build_id)) > 0) {
have_build_id = true;
pos->has_build_id = true;
}
return have_build_id;
}
bool dsos__read_build_ids(void)
{
bool kbuildids = __dsos__read_build_ids(&dsos__kernel),
ubuildids = __dsos__read_build_ids(&dsos__user);
return kbuildids || ubuildids;
}
/*
* Align offset to 4 bytes as needed for note name and descriptor data.
*/
#define NOTE_ALIGN(n) (((n) + 3) & -4U)
int filename__read_build_id(const char *filename, void *bf, size_t size)
2009-08-06 06:02:49 +08:00
{
int fd, err = -1;
2009-08-06 06:02:49 +08:00
GElf_Ehdr ehdr;
GElf_Shdr shdr;
Elf_Data *data;
2009-08-06 06:02:49 +08:00
Elf_Scn *sec;
Elf_Kind ek;
void *ptr;
2009-08-06 06:02:49 +08:00
Elf *elf;
if (size < BUILD_ID_SIZE)
goto out;
fd = open(filename, O_RDONLY);
2009-08-06 06:02:49 +08:00
if (fd < 0)
goto out;
elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
2009-08-06 06:02:49 +08:00
if (elf == NULL) {
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
pr_debug2("%s: cannot read %s ELF file.\n", __func__, filename);
2009-08-06 06:02:49 +08:00
goto out_close;
}
ek = elf_kind(elf);
if (ek != ELF_K_ELF)
goto out_elf_end;
2009-08-06 06:02:49 +08:00
if (gelf_getehdr(elf, &ehdr) == NULL) {
pr_err("%s: cannot get elf header.\n", __func__);
2009-08-06 06:02:49 +08:00
goto out_elf_end;
}
sec = elf_section_by_name(elf, &ehdr, &shdr,
".note.gnu.build-id", NULL);
if (sec == NULL) {
sec = elf_section_by_name(elf, &ehdr, &shdr,
".notes", NULL);
if (sec == NULL)
goto out_elf_end;
}
2009-08-06 06:02:49 +08:00
data = elf_getdata(sec, NULL);
if (data == NULL)
2009-08-06 06:02:49 +08:00
goto out_elf_end;
ptr = data->d_buf;
while (ptr < (data->d_buf + data->d_size)) {
GElf_Nhdr *nhdr = ptr;
int namesz = NOTE_ALIGN(nhdr->n_namesz),
descsz = NOTE_ALIGN(nhdr->n_descsz);
const char *name;
ptr += sizeof(*nhdr);
name = ptr;
ptr += namesz;
if (nhdr->n_type == NT_GNU_BUILD_ID &&
nhdr->n_namesz == sizeof("GNU")) {
if (memcmp(name, "GNU", sizeof("GNU")) == 0) {
memcpy(bf, ptr, BUILD_ID_SIZE);
err = BUILD_ID_SIZE;
break;
}
}
ptr += descsz;
}
out_elf_end:
elf_end(elf);
out_close:
close(fd);
out:
return err;
}
int sysfs__read_build_id(const char *filename, void *build_id, size_t size)
{
int fd, err = -1;
if (size < BUILD_ID_SIZE)
goto out;
fd = open(filename, O_RDONLY);
if (fd < 0)
goto out;
while (1) {
char bf[BUFSIZ];
GElf_Nhdr nhdr;
int namesz, descsz;
if (read(fd, &nhdr, sizeof(nhdr)) != sizeof(nhdr))
break;
namesz = NOTE_ALIGN(nhdr.n_namesz);
descsz = NOTE_ALIGN(nhdr.n_descsz);
if (nhdr.n_type == NT_GNU_BUILD_ID &&
nhdr.n_namesz == sizeof("GNU")) {
if (read(fd, bf, namesz) != namesz)
break;
if (memcmp(bf, "GNU", sizeof("GNU")) == 0) {
if (read(fd, build_id,
BUILD_ID_SIZE) == BUILD_ID_SIZE) {
err = 0;
break;
}
} else if (read(fd, bf, descsz) != descsz)
break;
} else {
int n = namesz + descsz;
if (read(fd, bf, n) != n)
break;
}
}
close(fd);
out:
return err;
}
char dso__symtab_origin(const struct dso *self)
{
static const char origin[] = {
[DSO__ORIG_KERNEL] = 'k',
[DSO__ORIG_JAVA_JIT] = 'j',
perf record: Introduce a symtab cache Now a cache will be created in a ~/.debug debuginfo like hierarchy, so that at the end of a 'perf record' session all the binaries (with build-ids) involved get collected and indexed by their build-ids, so that perf report can find them. This is interesting when developing software where you want to do a 'perf diff' with the previous build and opens avenues for lots more interesting tools, like a 'perf diff --graph' that takes more than two binaries into account. Tunables for collecting just the symtabs can be added if one doesn't want to have the full binary, but having the full binary allows things like 'perf rerecord' or other tools that can re-run the tests by having access to the exact binary in some perf.data file, so it may well be interesting to keep the full binary there. Space consumption is minimised by trying to use hard links, a 'perf cache' tool to manage the space used, a la ccache is required to purge older entries. With this in place it will be possible also to introduce new commands, 'perf archive' and 'perf restore' (or some more suitable and future proof names) to create a cpio/tar file with the perf data and the files in the cache that _had_ perf hits of interest. There are more aspects to polish, like finding the right vmlinux file to cache, etc, but this is enough for a first step. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-28 07:37:06 +08:00
[DSO__ORIG_BUILD_ID_CACHE] = 'B',
[DSO__ORIG_FEDORA] = 'f',
[DSO__ORIG_UBUNTU] = 'u',
[DSO__ORIG_BUILDID] = 'b',
[DSO__ORIG_DSO] = 'd',
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
[DSO__ORIG_KMODULE] = 'K',
};
if (self == NULL || self->origin == DSO__ORIG_NOT_FOUND)
return '!';
return origin[self->origin];
}
int dso__load(struct dso *self, struct map *map, struct perf_session *session,
symbol_filter_t filter)
{
2009-08-06 06:02:49 +08:00
int size = PATH_MAX;
char *name;
u8 build_id[BUILD_ID_SIZE];
perf record: Introduce a symtab cache Now a cache will be created in a ~/.debug debuginfo like hierarchy, so that at the end of a 'perf record' session all the binaries (with build-ids) involved get collected and indexed by their build-ids, so that perf report can find them. This is interesting when developing software where you want to do a 'perf diff' with the previous build and opens avenues for lots more interesting tools, like a 'perf diff --graph' that takes more than two binaries into account. Tunables for collecting just the symtabs can be added if one doesn't want to have the full binary, but having the full binary allows things like 'perf rerecord' or other tools that can re-run the tests by having access to the exact binary in some perf.data file, so it may well be interesting to keep the full binary there. Space consumption is minimised by trying to use hard links, a 'perf cache' tool to manage the space used, a la ccache is required to purge older entries. With this in place it will be possible also to introduce new commands, 'perf archive' and 'perf restore' (or some more suitable and future proof names) to create a cpio/tar file with the perf data and the files in the cache that _had_ perf hits of interest. There are more aspects to polish, like finding the right vmlinux file to cache, etc, but this is enough for a first step. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-28 07:37:06 +08:00
char build_id_hex[BUILD_ID_SIZE * 2 + 1];
int ret = -1;
int fd;
dso__set_loaded(self, map->type);
if (self->kernel)
return dso__load_kernel_sym(self, map, session, filter);
name = malloc(size);
if (!name)
return -1;
perf_counter tools: Adjust symbols in ET_EXEC files too Ingo Molnar wrote: > i just bisected a 'perf report' bug that would cause us to not > resolve all user-space symbols in a 'git gc' run to: > > f5812a7a336fb952d819e4427b9a2dce02368e82 is first bad commit > commit f5812a7a336fb952d819e4427b9a2dce02368e82 > Author: Arnaldo Carvalho de Melo <acme@redhat.com> > Date: Tue Jun 30 11:43:17 2009 -0300 > > perf_counter tools: Adjust only prelinked symbol's addresses Rename ->prelinked to ->adjust_symbols and making what was done only for prelinked libraries also to ET_EXEC binaries, such as /usr/bin/git: [acme@doppio pahole]$ readelf -h /usr/bin/git | grep Type Type: EXEC (Executable file) [acme@doppio pahole]$ And after installing the 'git-debuginfo' package, I get correct results: [acme@doppio linux-2.6-tip]$ perf report --sort comm,dso,symbol -d /usr/bin/git | head -20 # # (1139614 samples) # # Overhead Command Shared Object Symbol # ........ ................ ......................... ...... # 34.98% git /usr/bin/git [.] send_sideband 33.39% git /usr/bin/git [.] enter_repo 6.81% git /usr/bin/git [.] diff_opt_parse 4.95% git /usr/bin/git [.] is_repository_shallow 3.24% git /usr/bin/git [.] odb_mkstemp 1.39% git /usr/bin/git [.] output 1.34% git /usr/bin/git [.] xmmap 1.25% git /usr/bin/git [.] receive_pack_config 1.16% git /usr/bin/git [.] git_pathdup 0.90% git /usr/bin/git [.] read_object_with_reference 0.86% git /usr/bin/git [.] show_patch_diff 0.85% git /usr/bin/git 0x00000000095e2e 0.69% git /usr/bin/git [.] display [acme@doppio linux-2.6-tip]$ I'll check what are the last cases where we can't resolve symbols, like this 0x00000000095e2e later. And I guess this will fix the problems Mike were seeing too: [acme@doppio linux-2.6-tip]$ readelf -h ../build/perf/vmlinux | grep Type Type: EXEC (Executable file) [acme@doppio linux-2.6-tip]$ Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-03 08:24:14 +08:00
self->adjust_symbols = 0;
if (strncmp(self->name, "/tmp/perf-", 10) == 0) {
ret = dso__load_perf_map(self, map, filter);
self->origin = ret > 0 ? DSO__ORIG_JAVA_JIT :
DSO__ORIG_NOT_FOUND;
return ret;
}
perf record: Introduce a symtab cache Now a cache will be created in a ~/.debug debuginfo like hierarchy, so that at the end of a 'perf record' session all the binaries (with build-ids) involved get collected and indexed by their build-ids, so that perf report can find them. This is interesting when developing software where you want to do a 'perf diff' with the previous build and opens avenues for lots more interesting tools, like a 'perf diff --graph' that takes more than two binaries into account. Tunables for collecting just the symtabs can be added if one doesn't want to have the full binary, but having the full binary allows things like 'perf rerecord' or other tools that can re-run the tests by having access to the exact binary in some perf.data file, so it may well be interesting to keep the full binary there. Space consumption is minimised by trying to use hard links, a 'perf cache' tool to manage the space used, a la ccache is required to purge older entries. With this in place it will be possible also to introduce new commands, 'perf archive' and 'perf restore' (or some more suitable and future proof names) to create a cpio/tar file with the perf data and the files in the cache that _had_ perf hits of interest. There are more aspects to polish, like finding the right vmlinux file to cache, etc, but this is enough for a first step. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-28 07:37:06 +08:00
self->origin = DSO__ORIG_BUILD_ID_CACHE;
perf record: Introduce a symtab cache Now a cache will be created in a ~/.debug debuginfo like hierarchy, so that at the end of a 'perf record' session all the binaries (with build-ids) involved get collected and indexed by their build-ids, so that perf report can find them. This is interesting when developing software where you want to do a 'perf diff' with the previous build and opens avenues for lots more interesting tools, like a 'perf diff --graph' that takes more than two binaries into account. Tunables for collecting just the symtabs can be added if one doesn't want to have the full binary, but having the full binary allows things like 'perf rerecord' or other tools that can re-run the tests by having access to the exact binary in some perf.data file, so it may well be interesting to keep the full binary there. Space consumption is minimised by trying to use hard links, a 'perf cache' tool to manage the space used, a la ccache is required to purge older entries. With this in place it will be possible also to introduce new commands, 'perf archive' and 'perf restore' (or some more suitable and future proof names) to create a cpio/tar file with the perf data and the files in the cache that _had_ perf hits of interest. There are more aspects to polish, like finding the right vmlinux file to cache, etc, but this is enough for a first step. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-28 07:37:06 +08:00
if (self->has_build_id) {
build_id__sprintf(self->build_id, sizeof(self->build_id),
build_id_hex);
snprintf(name, size, "%s/%s/.build-id/%.2s/%s",
getenv("HOME"), DEBUG_CACHE_DIR,
build_id_hex, build_id_hex + 2);
goto open_file;
}
more:
do {
self->origin++;
switch (self->origin) {
case DSO__ORIG_FEDORA:
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
snprintf(name, size, "/usr/lib/debug%s.debug",
self->long_name);
break;
case DSO__ORIG_UBUNTU:
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
snprintf(name, size, "/usr/lib/debug%s",
self->long_name);
break;
case DSO__ORIG_BUILDID:
if (filename__read_build_id(self->long_name, build_id,
sizeof(build_id))) {
build_id__sprintf(build_id, sizeof(build_id),
build_id_hex);
2009-08-06 06:02:49 +08:00
snprintf(name, size,
"/usr/lib/debug/.build-id/%.2s/%s.debug",
build_id_hex, build_id_hex + 2);
if (self->has_build_id)
goto compare_build_id;
break;
2009-08-06 06:02:49 +08:00
}
self->origin++;
2009-08-06 06:02:49 +08:00
/* Fall thru */
case DSO__ORIG_DSO:
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
snprintf(name, size, "%s", self->long_name);
break;
default:
goto out;
}
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
if (self->has_build_id) {
if (filename__read_build_id(name, build_id,
sizeof(build_id)) < 0)
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
goto more;
compare_build_id:
if (!dso__build_id_equal(self, build_id))
perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K. Prasad <prasad@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Roland McGrath <roland@redhat.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
goto more;
}
perf record: Introduce a symtab cache Now a cache will be created in a ~/.debug debuginfo like hierarchy, so that at the end of a 'perf record' session all the binaries (with build-ids) involved get collected and indexed by their build-ids, so that perf report can find them. This is interesting when developing software where you want to do a 'perf diff' with the previous build and opens avenues for lots more interesting tools, like a 'perf diff --graph' that takes more than two binaries into account. Tunables for collecting just the symtabs can be added if one doesn't want to have the full binary, but having the full binary allows things like 'perf rerecord' or other tools that can re-run the tests by having access to the exact binary in some perf.data file, so it may well be interesting to keep the full binary there. Space consumption is minimised by trying to use hard links, a 'perf cache' tool to manage the space used, a la ccache is required to purge older entries. With this in place it will be possible also to introduce new commands, 'perf archive' and 'perf restore' (or some more suitable and future proof names) to create a cpio/tar file with the perf data and the files in the cache that _had_ perf hits of interest. There are more aspects to polish, like finding the right vmlinux file to cache, etc, but this is enough for a first step. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-28 07:37:06 +08:00
open_file:
fd = open(name, O_RDONLY);
} while (fd < 0);
ret = dso__load_sym(self, map, NULL, name, fd, filter, 0, 0);
close(fd);
/*
* Some people seem to have debuginfo files _WITHOUT_ debug info!?!?
*/
if (!ret)
goto more;
if (ret > 0) {
int nr_plt = dso__synthesize_plt_symbols(self, map, filter);
if (nr_plt > 0)
ret += nr_plt;
}
out:
free(name);
perf tools: Fix dso__new handle() to handle deleted DSOs It is better than showing the map addr, this way at least we know that we can't get the symtabs because the DSO was deleted (system update) while an app still used such DSO. Yeah, don't do that, but if you do, you'll figure it out quicker this way. [acme@doppio linux-2.6-tip]$ perf report | head -15 # Samples: 3796 # # Overhead Command Shared Object Symbol # ........ ....... ................................................................... ...... # 23.55% pidgin /lib64/libglib-2.0.so.0.2000.4.#prelink#.Pd98lu (deleted) [.] 0x00000000038844 21.55% pidgin /lib64/libpthread-2.10.1.so.#prelink#.AFwK8Q (deleted) [.] 0x0000000000a42d 10.85% pidgin [kernel] [.] vread_hpet 7.85% pidgin /lib64/libgobject-2.0.so.0.2000.4.#prelink#.o1vpU7 (deleted) [.] 0x00000000014de8 3.35% pidgin /lib64/libc-2.10.1.so (deleted) [.] 0x0000000007a875 3.19% pidgin /lib64/libdbus-1.so.3.4.0.#prelink#.6mwgZP (deleted) [.] 0x0000000001d254 3.06% pidgin /usr/lib64/libgtk-x11-2.0.so.0.1600.5.#prelink#.511hAl (deleted) [.] 0x000000002334e7 2.90% pidgin /usr/lib64/libgdk-x11-2.0.so.0.1600.5.#prelink#.5qlMo1 (deleted) [.] 0x00000000037b2d 1.84% pidgin [kernel] [k] do_sys_poll 1.45% pidgin /usr/lib64/libX11.so.6.2.0.#prelink#.iR59Rx (deleted) [.] 0x0000000004c751 [acme@doppio linux-2.6-tip]$ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Luis Claudio R. Gonçalves <lclaudio@redhat.com> Cc: Clark Williams <williams@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Frédéric Weisbecker <fweisbec@gmail.com> LKML-Reference: <20090811200436.GA3478@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-12 04:04:36 +08:00
if (ret < 0 && strstr(self->name, " (deleted)") != NULL)
return 0;
return ret;
}
struct map *map_groups__find_by_name(struct map_groups *self,
enum map_type type, const char *name)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
{
struct rb_node *nd;
for (nd = rb_first(&self->maps[type]); nd; nd = rb_next(nd)) {
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
struct map *map = rb_entry(nd, struct map, rb_node);
if (map->dso && strcmp(map->dso->short_name, name) == 0)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
return map;
}
return NULL;
}
static int dso__kernel_module_get_build_id(struct dso *self)
{
char filename[PATH_MAX];
/*
* kernel module short names are of the form "[module]" and
* we need just "module" here.
*/
const char *name = self->short_name + 1;
snprintf(filename, sizeof(filename),
"/sys/module/%.*s/notes/.note.gnu.build-id",
(int)strlen(name - 1), name);
if (sysfs__read_build_id(filename, self->build_id,
sizeof(self->build_id)) == 0)
self->has_build_id = true;
return 0;
}
static int perf_session__set_modules_path_dir(struct perf_session *self, char *dirname)
{
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
struct dirent *dent;
DIR *dir = opendir(dirname);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (!dir) {
pr_debug("%s: cannot open %s dir\n", __func__, dirname);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
return -1;
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
while ((dent = readdir(dir)) != NULL) {
char path[PATH_MAX];
if (dent->d_type == DT_DIR) {
if (!strcmp(dent->d_name, ".") ||
!strcmp(dent->d_name, ".."))
continue;
snprintf(path, sizeof(path), "%s/%s",
dirname, dent->d_name);
if (perf_session__set_modules_path_dir(self, path) < 0)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
goto failure;
} else {
char *dot = strrchr(dent->d_name, '.'),
dso_name[PATH_MAX];
struct map *map;
char *long_name;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (dot == NULL || strcmp(dot, ".ko"))
continue;
snprintf(dso_name, sizeof(dso_name), "[%.*s]",
(int)(dot - dent->d_name), dent->d_name);
strxfrchar(dso_name, '-', '_');
map = map_groups__find_by_name(&self->kmaps, MAP__FUNCTION, dso_name);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (map == NULL)
continue;
snprintf(path, sizeof(path), "%s/%s",
dirname, dent->d_name);
long_name = strdup(path);
if (long_name == NULL)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
goto failure;
dso__set_long_name(map->dso, long_name);
dso__kernel_module_get_build_id(map->dso);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
}
}
return 0;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
failure:
closedir(dir);
return -1;
}
static int perf_session__set_modules_path(struct perf_session *self)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
{
struct utsname uts;
char modules_path[PATH_MAX];
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (uname(&uts) < 0)
return -1;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
snprintf(modules_path, sizeof(modules_path), "/lib/modules/%s/kernel",
uts.release);
return perf_session__set_modules_path_dir(self, modules_path);
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
/*
* Constructor variant for modules (where we know from /proc/modules where
* they are loaded) and for vmlinux, where only after we load all the
* symbols we'll know where it starts and ends.
*/
static struct map *map__new2(u64 start, struct dso *dso, enum map_type type)
{
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
struct map *self = malloc(sizeof(*self));
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (self != NULL) {
/*
* ->end will be filled after we load all the symbols
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
*/
map__init(self, type, start, 0, 0, dso);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
return self;
}
struct map *perf_session__new_module_map(struct perf_session *self, u64 start,
const char *filename)
{
struct map *map;
struct dso *dso = __dsos__findnew(&dsos__kernel, filename);
if (dso == NULL)
return NULL;
map = map__new2(start, dso, MAP__FUNCTION);
if (map == NULL)
return NULL;
dso->origin = DSO__ORIG_KMODULE;
map_groups__insert(&self->kmaps, map);
return map;
}
static int perf_session__create_module_maps(struct perf_session *self)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
{
char *line = NULL;
size_t n;
FILE *file = fopen("/proc/modules", "r");
struct map *map;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (file == NULL)
return -1;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
while (!feof(file)) {
char name[PATH_MAX];
u64 start;
char *sep;
int line_len;
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
line_len = getline(&line, &n, file);
if (line_len < 0)
break;
if (!line)
goto out_failure;
line[--line_len] = '\0'; /* \n */
sep = strrchr(line, 'x');
if (sep == NULL)
continue;
hex2u64(sep + 1, &start);
sep = strchr(line, ' ');
if (sep == NULL)
continue;
*sep = '\0';
snprintf(name, sizeof(name), "[%s]", line);
map = perf_session__new_module_map(self, start, name);
if (map == NULL)
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
goto out_delete_line;
dso__kernel_module_get_build_id(map->dso);
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
free(line);
fclose(file);
return perf_session__set_modules_path(self);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
out_delete_line:
free(line);
out_failure:
return -1;
}
static int dso__load_vmlinux(struct dso *self, struct map *map,
struct perf_session *session,
const char *vmlinux, symbol_filter_t filter)
{
int err = -1, fd;
if (self->has_build_id) {
u8 build_id[BUILD_ID_SIZE];
if (filename__read_build_id(vmlinux, build_id,
sizeof(build_id)) < 0) {
pr_debug("No build_id in %s, ignoring it\n", vmlinux);
return -1;
}
if (!dso__build_id_equal(self, build_id)) {
char expected_build_id[BUILD_ID_SIZE * 2 + 1],
vmlinux_build_id[BUILD_ID_SIZE * 2 + 1];
build_id__sprintf(self->build_id,
sizeof(self->build_id),
expected_build_id);
build_id__sprintf(build_id, sizeof(build_id),
vmlinux_build_id);
pr_debug("build_id in %s is %s while expected is %s, "
"ignoring it\n", vmlinux, vmlinux_build_id,
expected_build_id);
return -1;
}
}
fd = open(vmlinux, O_RDONLY);
if (fd < 0)
return -1;
dso__set_loaded(self, map->type);
err = dso__load_sym(self, map, session, self->long_name, fd, filter, 1, 0);
close(fd);
return err;
}
static int dso__load_kernel_sym(struct dso *self, struct map *map,
struct perf_session *session, symbol_filter_t filter)
{
int err;
const char *kallsyms_filename = NULL;
char *kallsyms_allocated_filename = NULL;
if (vmlinux_path != NULL) {
int i;
pr_debug("Looking at the vmlinux_path (%d entries long)\n",
vmlinux_path__nr_entries);
for (i = 0; i < vmlinux_path__nr_entries; ++i) {
err = dso__load_vmlinux(self, map, session,
vmlinux_path[i], filter);
if (err > 0) {
pr_debug("Using %s for symbols\n",
vmlinux_path[i]);
dso__set_long_name(self,
strdup(vmlinux_path[i]));
goto out_fixup;
}
}
}
/*
* Say the kernel DSO was created when processing the build-id header table,
* we have a build-id, so check if it is the same as the running kernel,
* using it if it is.
*/
if (self->has_build_id) {
u8 kallsyms_build_id[BUILD_ID_SIZE];
char sbuild_id[BUILD_ID_SIZE * 2 + 1];
if (sysfs__read_build_id("/sys/kernel/notes", kallsyms_build_id,
sizeof(kallsyms_build_id)) == 0) {
if (dso__build_id_equal(self, kallsyms_build_id)) {
kallsyms_filename = "/proc/kallsyms";
goto do_kallsyms;
}
}
build_id__sprintf(self->build_id, sizeof(self->build_id),
sbuild_id);
if (asprintf(&kallsyms_allocated_filename,
"%s/.debug/[kernel.kallsyms]/%s",
getenv("HOME"), sbuild_id) != -1) {
if (access(kallsyms_filename, F_OK)) {
kallsyms_filename = kallsyms_allocated_filename;
goto do_kallsyms;
}
free(kallsyms_allocated_filename);
kallsyms_allocated_filename = NULL;
}
goto do_vmlinux;
}
if (self->long_name[0] == '[') {
kallsyms_filename = "/proc/kallsyms";
goto do_kallsyms;
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
do_vmlinux:
err = dso__load_vmlinux(self, map, session, self->long_name, filter);
if (err <= 0) {
if (self->has_build_id)
return -1;
pr_info("The file %s cannot be used, "
"trying to use /proc/kallsyms...", self->long_name);
do_kallsyms:
err = dso__load_kallsyms(self, kallsyms_filename, map, session, filter);
if (err > 0 && kallsyms_filename == NULL)
dso__set_long_name(self, strdup("[kernel.kallsyms]"));
free(kallsyms_allocated_filename);
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
if (err > 0) {
out_fixup:
map__fixup_start(map);
map__fixup_end(map);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 14:29:58 +08:00
}
return err;
}
LIST_HEAD(dsos__user);
LIST_HEAD(dsos__kernel);
struct dso *vdso;
static void dsos__add(struct list_head *head, struct dso *dso)
{
list_add_tail(&dso->node, head);
}
static struct dso *dsos__find(struct list_head *head, const char *name)
{
struct dso *pos;
list_for_each_entry(pos, head, node)
if (strcmp(pos->name, name) == 0)
return pos;
return NULL;
}
struct dso *__dsos__findnew(struct list_head *head, const char *name)
{
struct dso *dso = dsos__find(head, name);
if (!dso) {
dso = dso__new(name);
if (dso != NULL) {
dsos__add(head, dso);
dso__set_basename(dso);
}
}
return dso;
}
static void __dsos__fprintf(struct list_head *head, FILE *fp)
{
struct dso *pos;
list_for_each_entry(pos, head, node) {
int i;
for (i = 0; i < MAP__NR_TYPES; ++i)
dso__fprintf(pos, i, fp);
}
}
void dsos__fprintf(FILE *fp)
{
__dsos__fprintf(&dsos__kernel, fp);
__dsos__fprintf(&dsos__user, fp);
}
static size_t __dsos__fprintf_buildid(struct list_head *head, FILE *fp)
{
struct dso *pos;
size_t ret = 0;
list_for_each_entry(pos, head, node) {
ret += dso__fprintf_buildid(pos, fp);
ret += fprintf(fp, " %s\n", pos->long_name);
}
return ret;
}
size_t dsos__fprintf_buildid(FILE *fp)
{
return (__dsos__fprintf_buildid(&dsos__kernel, fp) +
__dsos__fprintf_buildid(&dsos__user, fp));
}
static struct dso *dsos__create_kernel(const char *vmlinux)
{
struct dso *kernel = dso__new(vmlinux ?: "[kernel.kallsyms]");
if (kernel == NULL)
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
return NULL;
kernel->short_name = "[kernel]";
kernel->kernel = 1;
vdso = dso__new("[vdso]");
if (vdso == NULL)
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
goto out_delete_kernel_dso;
dso__set_loaded(vdso, MAP__FUNCTION);
if (sysfs__read_build_id("/sys/kernel/notes", kernel->build_id,
sizeof(kernel->build_id)) == 0)
kernel->has_build_id = true;
dsos__add(&dsos__kernel, kernel);
dsos__add(&dsos__user, vdso);
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
return kernel;
out_delete_kernel_dso:
dso__delete(kernel);
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
return NULL;
}
int __map_groups__create_kernel_maps(struct map_groups *self,
struct map *vmlinux_maps[MAP__NR_TYPES],
struct dso *kernel)
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
{
enum map_type type;
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
for (type = 0; type < MAP__NR_TYPES; ++type) {
vmlinux_maps[type] = map__new2(0, kernel, type);
if (vmlinux_maps[type] == NULL)
return -1;
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
vmlinux_maps[type]->map_ip =
vmlinux_maps[type]->unmap_ip = identity__map_ip;
map_groups__insert(self, vmlinux_maps[type]);
perf symbols: Add support for 'variable' symtabs Example: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } Added just after symbol__init() call in 'perf top', then: { u64 addr = strtoull(sym_filter, NULL, 16); struct map *map = map_groups__find(kmaps, MAP__VARIABLE, addr); if (map == NULL) pr_err("couldn't find map!\n"); else { struct symbol *sym = map__find_symbol(map, addr, NULL); if (sym == NULL) pr_err("couldn't find addr!\n"); else pr_info("addr %#Lx is in %s global var\n", addr, sym->name); } exit(0); } [root@doppio linux-2.6-tip]# grep ' [dD] ' /proc/kallsyms | grep ' sched' ffffffff817827d8 d sched_nr_latency ffffffff81782ce0 d sched_domains_mutex ffffffff8178c070 d schedstr.22423 ffffffff817909a0 d sched_register_mutex ffffffff81823490 d sched_feat_names ffffffff81823558 d scheduler_running ffffffff818235b8 d sched_clock_running ffffffff818235bc D sched_clock_stable ffffffff81824f00 d sched_switch_trace [root@doppio linux-2.6-tip]# perf top -s 0xffffffff817827d9 addr 0xffffffff817827d9 is in sched_nr_latency global var [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# [root@doppio linux-2.6-tip]# perf top -s ffffffff81782ce0 --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff81782ce0 is in sched_domains_mutex global var [root@doppio linux-2.6-tip]# perf top -s ffffffff818235bc --vmlinux OFF The file OFF cannot be used, trying to use /proc/kallsyms...addr 0xffffffff818235bc is in sched_clock_stable global var [root@doppio linux-2.6-tip]# So it works with both /proc/kallsyms and with ELF symtabs, either the one on the vmlinux explicitely passed via --vmlinux or in one in the vmlinux_path that matches the buildid for the running kernel or the one found in the buildid header section in a perf.data file. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260550239-5372-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-12 00:50:39 +08:00
}
return 0;
}
static int map_groups__create_kernel_maps(struct map_groups *self,
struct map *vmlinux_maps[MAP__NR_TYPES],
const char *vmlinux)
{
struct dso *kernel = dsos__create_kernel(vmlinux);
if (kernel == NULL)
return -1;
return __map_groups__create_kernel_maps(self, vmlinux_maps, kernel);
}
static void vmlinux_path__exit(void)
{
while (--vmlinux_path__nr_entries >= 0) {
free(vmlinux_path[vmlinux_path__nr_entries]);
vmlinux_path[vmlinux_path__nr_entries] = NULL;
}
free(vmlinux_path);
vmlinux_path = NULL;
}
static int vmlinux_path__init(void)
{
struct utsname uts;
char bf[PATH_MAX];
if (uname(&uts) < 0)
return -1;
vmlinux_path = malloc(sizeof(char *) * 5);
if (vmlinux_path == NULL)
return -1;
vmlinux_path[vmlinux_path__nr_entries] = strdup("vmlinux");
if (vmlinux_path[vmlinux_path__nr_entries] == NULL)
goto out_fail;
++vmlinux_path__nr_entries;
vmlinux_path[vmlinux_path__nr_entries] = strdup("/boot/vmlinux");
if (vmlinux_path[vmlinux_path__nr_entries] == NULL)
goto out_fail;
++vmlinux_path__nr_entries;
snprintf(bf, sizeof(bf), "/boot/vmlinux-%s", uts.release);
vmlinux_path[vmlinux_path__nr_entries] = strdup(bf);
if (vmlinux_path[vmlinux_path__nr_entries] == NULL)
goto out_fail;
++vmlinux_path__nr_entries;
snprintf(bf, sizeof(bf), "/lib/modules/%s/build/vmlinux", uts.release);
vmlinux_path[vmlinux_path__nr_entries] = strdup(bf);
if (vmlinux_path[vmlinux_path__nr_entries] == NULL)
goto out_fail;
++vmlinux_path__nr_entries;
snprintf(bf, sizeof(bf), "/usr/lib/debug/lib/modules/%s/vmlinux",
uts.release);
vmlinux_path[vmlinux_path__nr_entries] = strdup(bf);
if (vmlinux_path[vmlinux_path__nr_entries] == NULL)
goto out_fail;
++vmlinux_path__nr_entries;
return 0;
out_fail:
vmlinux_path__exit();
return -1;
}
static int setup_list(struct strlist **list, const char *list_str,
const char *list_name)
{
if (list_str == NULL)
return 0;
*list = strlist__new(true, list_str);
if (!*list) {
pr_err("problems parsing %s list\n", list_name);
return -1;
}
return 0;
}
int symbol__init(void)
{
elf_version(EV_CURRENT);
if (symbol_conf.sort_by_name)
symbol_conf.priv_size += (sizeof(struct symbol_name_rb_node) -
sizeof(struct symbol));
if (symbol_conf.try_vmlinux_path && vmlinux_path__init() < 0)
return -1;
if (symbol_conf.field_sep && *symbol_conf.field_sep == '.') {
pr_err("'.' is the only non valid --field-separator argument\n");
return -1;
}
if (setup_list(&symbol_conf.dso_list,
symbol_conf.dso_list_str, "dso") < 0)
return -1;
if (setup_list(&symbol_conf.comm_list,
symbol_conf.comm_list_str, "comm") < 0)
goto out_free_dso_list;
if (setup_list(&symbol_conf.sym_list,
symbol_conf.sym_list_str, "symbol") < 0)
goto out_free_comm_list;
return 0;
out_free_dso_list:
strlist__delete(symbol_conf.dso_list);
out_free_comm_list:
strlist__delete(symbol_conf.comm_list);
return -1;
}
int perf_session__create_kernel_maps(struct perf_session *self)
{
if (map_groups__create_kernel_maps(&self->kmaps, self->vmlinux_maps,
symbol_conf.vmlinux_name) < 0)
return -1;
if (symbol_conf.use_modules &&
perf_session__create_module_maps(self) < 0)
pr_debug("Failed to load list of modules for session %s, "
"continuing...\n", self->filename);
/*
* Now that we have all the maps created, just set the ->end of them:
*/
map_groups__fixup_end(&self->kmaps);
return 0;
}