linux_old1/tools/perf/util/util.c

697 lines
12 KiB
C
Raw Normal View History

#include "../perf.h"
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
#include "util.h"
#include "debug.h"
#include <api/fs/fs.h>
#include <sys/mman.h>
#ifdef HAVE_BACKTRACE_SUPPORT
#include <execinfo.h>
#endif
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <limits.h>
#include <byteswap.h>
#include <linux/kernel.h>
#include <unistd.h>
#include "callchain.h"
struct callchain_param callchain_param = {
.mode = CHAIN_GRAPH_REL,
.min_percent = 0.5,
.order = ORDER_CALLEE,
.key = CCKEY_FUNCTION
};
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
/*
* XXX We need to find a better place for these things...
*/
unsigned int page_size;
int cacheline_size;
bool test_attr__enabled;
bool perf_host = true;
bool perf_guest = false;
char tracing_events_path[PATH_MAX + 1] = "/sys/kernel/debug/tracing/events";
void event_attr_init(struct perf_event_attr *attr)
{
if (!perf_host)
attr->exclude_host = 1;
if (!perf_guest)
attr->exclude_guest = 1;
/* to capture ABI version */
attr->size = sizeof(*attr);
}
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
int mkdir_p(char *path, mode_t mode)
{
struct stat st;
int err;
char *d = path;
if (*d != '/')
return -1;
if (stat(path, &st) == 0)
return 0;
while (*++d == '/');
while ((d = strchr(d, '/'))) {
*d = '\0';
err = stat(path, &st) && mkdir(path, mode);
*d++ = '/';
if (err)
return -1;
while (*d == '/')
++d;
}
return (stat(path, &st) && mkdir(path, mode)) ? -1 : 0;
}
int rm_rf(char *path)
{
DIR *dir;
int ret = 0;
struct dirent *d;
char namebuf[PATH_MAX];
dir = opendir(path);
if (dir == NULL)
return 0;
while ((d = readdir(dir)) != NULL && !ret) {
struct stat statbuf;
if (!strcmp(d->d_name, ".") || !strcmp(d->d_name, ".."))
continue;
scnprintf(namebuf, sizeof(namebuf), "%s/%s",
path, d->d_name);
ret = stat(namebuf, &statbuf);
if (ret < 0) {
pr_debug("stat failed: %s\n", namebuf);
break;
}
if (S_ISREG(statbuf.st_mode))
ret = unlink(namebuf);
else if (S_ISDIR(statbuf.st_mode))
ret = rm_rf(namebuf);
else {
pr_debug("unknown file: %s\n", namebuf);
ret = -1;
}
}
closedir(dir);
if (ret < 0)
return ret;
return rmdir(path);
}
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
static int slow_copyfile(const char *from, const char *to)
{
int err = -1;
char *line = NULL;
size_t n;
FILE *from_fp = fopen(from, "r"), *to_fp;
if (from_fp == NULL)
goto out;
to_fp = fopen(to, "w");
if (to_fp == NULL)
goto out_fclose_from;
while (getline(&line, &n, from_fp) > 0)
if (fputs(line, to_fp) == EOF)
goto out_fclose_to;
err = 0;
out_fclose_to:
fclose(to_fp);
free(line);
out_fclose_from:
fclose(from_fp);
out:
return err;
}
int copyfile_offset(int ifd, loff_t off_in, int ofd, loff_t off_out, u64 size)
{
void *ptr;
loff_t pgoff;
pgoff = off_in & ~(page_size - 1);
off_in -= pgoff;
ptr = mmap(NULL, off_in + size, PROT_READ, MAP_PRIVATE, ifd, pgoff);
if (ptr == MAP_FAILED)
return -1;
while (size) {
ssize_t ret = pwrite(ofd, ptr + off_in, size, off_out);
if (ret < 0 && errno == EINTR)
continue;
if (ret <= 0)
break;
size -= ret;
off_in += ret;
off_out -= ret;
}
munmap(ptr, off_in + size);
return size ? -1 : 0;
}
int copyfile_mode(const char *from, const char *to, mode_t mode)
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
{
int fromfd, tofd;
struct stat st;
int err = -1;
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
char *tmp = NULL, *ptr = NULL;
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 (stat(from, &st))
goto out;
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
/* extra 'x' at the end is to reserve space for '.' */
if (asprintf(&tmp, "%s.XXXXXXx", to) < 0) {
tmp = NULL;
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
goto out;
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
}
ptr = strrchr(tmp, '/');
if (!ptr)
goto out;
ptr = memmove(ptr + 1, ptr, strlen(ptr) - 1);
*ptr = '.';
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
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
tofd = mkstemp(tmp);
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 (tofd < 0)
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
goto out;
if (fchmod(tofd, mode))
goto out_close_to;
if (st.st_size == 0) { /* /proc? do it slowly... */
err = slow_copyfile(from, tmp);
goto out_close_to;
}
fromfd = open(from, O_RDONLY);
if (fromfd < 0)
goto out_close_to;
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
err = copyfile_offset(fromfd, 0, tofd, 0, st.st_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
close(fromfd);
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
out_close_to:
close(tofd);
if (!err)
err = link(tmp, to);
unlink(tmp);
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
out:
perf tools: Avoid possible race condition in copyfile() Use unique temporary files when copying to buildid dir to prevent races in case multiple instances are trying to copy same file. This is done by - creating template in form <path>/.<filename>.XXXXXX where the suffix is used by mkstemp() to create unique file - change file mode - copy content - if successful link temp file to target file - unlink temp file At this point the only file left at target path should be the desired one either created by us or other instance if we raced. This should also prevent not yet fully copied files to be visible to to other perf instances that could try to parse them. On top of that slow_copyfile no longer needs to deal with file mode when creating file since temporary file is already created and mode is set. Succesfully tested by myself by running perf record, archive and reading the data on other system and by running perf buildid-cache on perf binary itself. I also did revert fix from 0635b0f that to exposes previously fixed race with EEXIST and recreator test passed sucessfully. Signed-off-by: Milos Vyletel <milos@redhat.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1433775018-19868-1-git-send-email-milos@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-06-08 22:50:16 +08:00
free(tmp);
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
return err;
}
int copyfile(const char *from, const char *to)
{
return copyfile_mode(from, to, 0755);
}
unsigned long convert_unit(unsigned long value, char *unit)
{
*unit = ' ';
if (value > 1000) {
value /= 1000;
*unit = 'K';
}
if (value > 1000) {
value /= 1000;
*unit = 'M';
}
if (value > 1000) {
value /= 1000;
*unit = 'G';
}
return value;
}
static ssize_t ion(bool is_read, int fd, void *buf, size_t n)
{
void *buf_start = buf;
size_t left = n;
while (left) {
ssize_t ret = is_read ? read(fd, buf, left) :
write(fd, buf, left);
if (ret < 0 && errno == EINTR)
continue;
if (ret <= 0)
return ret;
left -= ret;
buf += ret;
}
BUG_ON((size_t)(buf - buf_start) != n);
return n;
}
/*
* Read exactly 'n' bytes or return an error.
*/
ssize_t readn(int fd, void *buf, size_t n)
{
return ion(true, fd, buf, n);
}
/*
* Write exactly 'n' bytes or return an error.
*/
ssize_t writen(int fd, void *buf, size_t n)
{
return ion(false, fd, buf, n);
}
size_t hex_width(u64 v)
{
size_t n = 1;
while ((v >>= 4))
++n;
return n;
}
static int hex(char ch)
{
if ((ch >= '0') && (ch <= '9'))
return ch - '0';
if ((ch >= 'a') && (ch <= 'f'))
return ch - 'a' + 10;
if ((ch >= 'A') && (ch <= 'F'))
return ch - 'A' + 10;
return -1;
}
/*
* While we find nice hex chars, build a long_val.
* Return number of chars processed.
*/
int hex2u64(const char *ptr, u64 *long_val)
{
const char *p = ptr;
*long_val = 0;
while (*p) {
const int hex_val = hex(*p);
if (hex_val < 0)
break;
*long_val = (*long_val << 4) | hex_val;
p++;
}
return p - ptr;
}
/* Obtain a backtrace and print it to stdout. */
#ifdef HAVE_BACKTRACE_SUPPORT
void dump_stack(void)
{
void *array[16];
size_t size = backtrace(array, ARRAY_SIZE(array));
char **strings = backtrace_symbols(array, size);
size_t i;
printf("Obtained %zd stack frames.\n", size);
for (i = 0; i < size; i++)
printf("%s\n", strings[i]);
free(strings);
}
#else
void dump_stack(void) {}
#endif
void sighandler_dump_stack(int sig)
{
psignal(sig, "perf");
dump_stack();
exit(sig);
}
void get_term_dimensions(struct winsize *ws)
{
char *s = getenv("LINES");
if (s != NULL) {
ws->ws_row = atoi(s);
s = getenv("COLUMNS");
if (s != NULL) {
ws->ws_col = atoi(s);
if (ws->ws_row && ws->ws_col)
return;
}
}
#ifdef TIOCGWINSZ
if (ioctl(1, TIOCGWINSZ, ws) == 0 &&
ws->ws_row && ws->ws_col)
return;
#endif
ws->ws_row = 25;
ws->ws_col = 80;
}
void set_term_quiet_input(struct termios *old)
{
struct termios tc;
tcgetattr(0, old);
tc = *old;
tc.c_lflag &= ~(ICANON | ECHO);
tc.c_cc[VMIN] = 0;
tc.c_cc[VTIME] = 0;
tcsetattr(0, TCSANOW, &tc);
}
static void set_tracing_events_path(const char *tracing, const char *mountpoint)
{
snprintf(tracing_events_path, sizeof(tracing_events_path), "%s/%s%s",
mountpoint, tracing, "events");
}
static const char *__perf_tracefs_mount(const char *mountpoint)
{
const char *mnt;
mnt = tracefs_mount(mountpoint);
if (!mnt)
return NULL;
set_tracing_events_path("", mnt);
return mnt;
}
static const char *__perf_debugfs_mount(const char *mountpoint)
{
const char *mnt;
mnt = debugfs_mount(mountpoint);
if (!mnt)
return NULL;
set_tracing_events_path("tracing/", mnt);
return mnt;
}
const char *perf_debugfs_mount(const char *mountpoint)
{
const char *mnt;
mnt = __perf_tracefs_mount(mountpoint);
if (mnt)
return mnt;
mnt = __perf_debugfs_mount(mountpoint);
return mnt;
}
void perf_debugfs_set_path(const char *mntpt)
{
snprintf(debugfs_mountpoint, strlen(debugfs_mountpoint), "%s", mntpt);
set_tracing_events_path("tracing/", mntpt);
}
static const char *find_tracefs(void)
{
const char *path = __perf_tracefs_mount(NULL);
return path;
}
static const char *find_debugfs(void)
{
const char *path = __perf_debugfs_mount(NULL);
if (!path)
fprintf(stderr, "Your kernel does not support the debugfs filesystem");
return path;
}
/*
* Finds the path to the debugfs/tracing
* Allocates the string and stores it.
*/
const char *find_tracing_dir(void)
{
const char *tracing_dir = "";
static char *tracing;
static int tracing_found;
const char *debugfs;
if (tracing_found)
return tracing;
debugfs = find_tracefs();
if (!debugfs) {
tracing_dir = "/tracing";
debugfs = find_debugfs();
if (!debugfs)
return NULL;
}
if (asprintf(&tracing, "%s%s", debugfs, tracing_dir) < 0)
return NULL;
tracing_found = 1;
return tracing;
}
char *get_tracing_file(const char *name)
{
const char *tracing;
char *file;
tracing = find_tracing_dir();
if (!tracing)
return NULL;
if (asprintf(&file, "%s/%s", tracing, name) < 0)
return NULL;
return file;
}
void put_tracing_file(char *file)
{
free(file);
}
int parse_nsec_time(const char *str, u64 *ptime)
{
u64 time_sec, time_nsec;
char *end;
time_sec = strtoul(str, &end, 10);
if (*end != '.' && *end != '\0')
return -1;
if (*end == '.') {
int i;
char nsec_buf[10];
if (strlen(++end) > 9)
return -1;
strncpy(nsec_buf, end, 9);
nsec_buf[9] = '\0';
/* make it nsec precision */
for (i = strlen(nsec_buf); i < 9; i++)
nsec_buf[i] = '0';
time_nsec = strtoul(nsec_buf, &end, 10);
if (*end != '\0')
return -1;
} else
time_nsec = 0;
*ptime = time_sec * NSEC_PER_SEC + time_nsec;
return 0;
}
unsigned long parse_tag_value(const char *str, struct parse_tag *tags)
{
struct parse_tag *i = tags;
while (i->tag) {
char *s;
s = strchr(str, i->tag);
if (s) {
unsigned long int value;
char *endptr;
value = strtoul(str, &endptr, 10);
if (s != endptr)
break;
if (value > ULONG_MAX / i->mult)
break;
value *= i->mult;
return value;
}
i++;
}
return (unsigned long) -1;
}
int filename__read_str(const char *filename, char **buf, size_t *sizep)
{
size_t size = 0, alloc_size = 0;
void *bf = NULL, *nbf;
int fd, n, err = 0;
char sbuf[STRERR_BUFSIZE];
fd = open(filename, O_RDONLY);
if (fd < 0)
return -errno;
do {
if (size == alloc_size) {
alloc_size += BUFSIZ;
nbf = realloc(bf, alloc_size);
if (!nbf) {
err = -ENOMEM;
break;
}
bf = nbf;
}
n = read(fd, bf + size, alloc_size - size);
if (n < 0) {
if (size) {
pr_warning("read failed %d: %s\n", errno,
strerror_r(errno, sbuf, sizeof(sbuf)));
err = 0;
} else
err = -errno;
break;
}
size += n;
} while (n > 0);
if (!err) {
*sizep = size;
*buf = bf;
} else
free(bf);
close(fd);
return err;
}
const char *get_filename_for_perf_kvm(void)
{
const char *filename;
if (perf_host && !perf_guest)
filename = strdup("perf.data.host");
else if (!perf_host && perf_guest)
filename = strdup("perf.data.guest");
else
filename = strdup("perf.data.kvm");
return filename;
}
int perf_event_paranoid(void)
{
int value;
if (sysctl__read_int("kernel/perf_event_paranoid", &value))
return INT_MAX;
return value;
}
void mem_bswap_32(void *src, int byte_size)
{
u32 *m = src;
while (byte_size > 0) {
*m = bswap_32(*m);
byte_size -= sizeof(u32);
++m;
}
}
void mem_bswap_64(void *src, int byte_size)
{
u64 *m = src;
while (byte_size > 0) {
*m = bswap_64(*m);
byte_size -= sizeof(u64);
++m;
}
}
perf tools: Show better error message in case we fail to open counters due to EBUSY error Showing better error message in case we fail to open counters due to the EBUSY error. If we detect oprofile daemon process running, we now display following message for EBUSY error: $ perf record ls Error: The PMU counters are busy/taken by another profiler. We found oprofile daemon running, please stop it and try again. In case oprofiled was not detected the current error message stays: $ perf record ls Error: The sys_perf_event_open() syscall returned with 16 (Device or resource busy) for event (cycles). /bin/dmesg may provide additional information. No CONFIG_PERF_EVENTS=y kernel support configured? Also changing PERF_FLAG_FD_CLOEXEC detection code not to display error in case of EBUSY error, as it currently does: $ perf record ls Error: perf_event_open(..., PERF_FLAG_FD_CLOEXEC) failed with unexpected error 16 (Device or resource busy) perf_event_open(..., 0) failed unexpectedly with error 16 (Device or resource busy) The PMU counters are busy/taken by another profiler. We found oprofile daemon running, please stop it and try again. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: William Cohen <wcohen@redhat.com> Cc: Yann Droneaud <ydroneaud@opteya.com> Link: http://lkml.kernel.org/r/1406908014-8312-1-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-08-01 23:46:54 +08:00
bool find_process(const char *name)
{
size_t len = strlen(name);
DIR *dir;
struct dirent *d;
int ret = -1;
dir = opendir(procfs__mountpoint());
if (!dir)
return -1;
/* Walk through the directory. */
while (ret && (d = readdir(dir)) != NULL) {
char path[PATH_MAX];
char *data;
size_t size;
if ((d->d_type != DT_DIR) ||
!strcmp(".", d->d_name) ||
!strcmp("..", d->d_name))
continue;
scnprintf(path, sizeof(path), "%s/%s/comm",
procfs__mountpoint(), d->d_name);
if (filename__read_str(path, &data, &size))
continue;
ret = strncmp(name, data, len);
free(data);
}
closedir(dir);
return ret ? false : true;
}