linux/arch/powerpc/kernel/cacheinfo.c

935 lines
23 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
/*
* Processor cache information made available to userspace via sysfs;
* intended to be compatible with x86 intel_cacheinfo implementation.
*
* Copyright 2008 IBM Corporation
* Author: Nathan Lynch
*/
#define pr_fmt(fmt) "cacheinfo: " fmt
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
#include <linux/cpu.h>
#include <linux/cpumask.h>
#include <linux/kernel.h>
#include <linux/kobject.h>
#include <linux/list.h>
#include <linux/notifier.h>
#include <linux/of.h>
#include <linux/percpu.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
#include <asm/prom.h>
#include <asm/cputhreads.h>
#include <asm/smp.h>
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
#include "cacheinfo.h"
/* per-cpu object for tracking:
* - a "cache" kobject for the top-level directory
* - a list of "index" objects representing the cpu's local cache hierarchy
*/
struct cache_dir {
struct kobject *kobj; /* bare (not embedded) kobject for cache
* directory */
struct cache_index_dir *index; /* list of index objects */
};
/* "index" object: each cpu's cache directory has an index
* subdirectory corresponding to a cache object associated with the
* cpu. This object's lifetime is managed via the embedded kobject.
*/
struct cache_index_dir {
struct kobject kobj;
struct cache_index_dir *next; /* next index in parent directory */
struct cache *cache;
};
/* Template for determining which OF properties to query for a given
* cache type */
struct cache_type_info {
const char *name;
const char *size_prop;
/* Allow for both [di]-cache-line-size and
* [di]-cache-block-size properties. According to the PowerPC
* Processor binding, -line-size should be provided if it
* differs from the cache block size (that which is operated
* on by cache instructions), so we look for -line-size first.
* See cache_get_line_size(). */
const char *line_size_props[2];
const char *nr_sets_prop;
};
/* These are used to index the cache_type_info array. */
#define CACHE_TYPE_UNIFIED 0 /* cache-size, cache-block-size, etc. */
#define CACHE_TYPE_UNIFIED_D 1 /* d-cache-size, d-cache-block-size, etc */
#define CACHE_TYPE_INSTRUCTION 2
#define CACHE_TYPE_DATA 3
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
static const struct cache_type_info cache_type_info[] = {
{
/* Embedded systems that use cache-size, cache-block-size,
* etc. for the Unified (typically L2) cache. */
.name = "Unified",
.size_prop = "cache-size",
.line_size_props = { "cache-line-size",
"cache-block-size", },
.nr_sets_prop = "cache-sets",
},
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
/* PowerPC Processor binding says the [di]-cache-*
* must be equal on unified caches, so just use
* d-cache properties. */
.name = "Unified",
.size_prop = "d-cache-size",
.line_size_props = { "d-cache-line-size",
"d-cache-block-size", },
.nr_sets_prop = "d-cache-sets",
},
{
.name = "Instruction",
.size_prop = "i-cache-size",
.line_size_props = { "i-cache-line-size",
"i-cache-block-size", },
.nr_sets_prop = "i-cache-sets",
},
{
.name = "Data",
.size_prop = "d-cache-size",
.line_size_props = { "d-cache-line-size",
"d-cache-block-size", },
.nr_sets_prop = "d-cache-sets",
},
};
/* Cache object: each instance of this corresponds to a distinct cache
* in the system. There are separate objects for Harvard caches: one
* each for instruction and data, and each refers to the same OF node.
* The refcount of the OF node is elevated for the lifetime of the
* cache object. A cache object is released when its shared_cpu_map
* is cleared (see cache_cpu_clear).
*
* A cache object is on two lists: an unsorted global list
* (cache_list) of cache objects; and a singly-linked list
* representing the local cache hierarchy, which is ordered by level
* (e.g. L1d -> L1i -> L2 -> L3).
*/
struct cache {
struct device_node *ofnode; /* OF node for this cache, may be cpu */
struct cpumask shared_cpu_map; /* online CPUs using this cache */
int type; /* split cache disambiguation */
int level; /* level not explicit in device tree */
struct list_head list; /* global list of cache objects */
struct cache *next_local; /* next cache of >= level */
};
static DEFINE_PER_CPU(struct cache_dir *, cache_dir_pcpu);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
/* traversal/modification of this list occurs only at cpu hotplug time;
* access is serialized by cpu hotplug locking
*/
static LIST_HEAD(cache_list);
static struct cache_index_dir *kobj_to_cache_index_dir(struct kobject *k)
{
return container_of(k, struct cache_index_dir, kobj);
}
static const char *cache_type_string(const struct cache *cache)
{
return cache_type_info[cache->type].name;
}
static void cache_init(struct cache *cache, int type, int level,
struct device_node *ofnode)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
cache->type = type;
cache->level = level;
cache->ofnode = of_node_get(ofnode);
INIT_LIST_HEAD(&cache->list);
list_add(&cache->list, &cache_list);
}
static struct cache *new_cache(int type, int level, struct device_node *ofnode)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache *cache;
cache = kzalloc(sizeof(*cache), GFP_KERNEL);
if (cache)
cache_init(cache, type, level, ofnode);
return cache;
}
static void release_cache_debugcheck(struct cache *cache)
{
struct cache *iter;
list_for_each_entry(iter, &cache_list, list)
WARN_ONCE(iter->next_local == cache,
"cache for %pOFP(%s) refers to cache for %pOFP(%s)\n",
iter->ofnode,
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
cache_type_string(iter),
cache->ofnode,
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
cache_type_string(cache));
}
static void release_cache(struct cache *cache)
{
if (!cache)
return;
pr_debug("freeing L%d %s cache for %pOFP\n", cache->level,
cache_type_string(cache), cache->ofnode);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
release_cache_debugcheck(cache);
list_del(&cache->list);
of_node_put(cache->ofnode);
kfree(cache);
}
static void cache_cpu_set(struct cache *cache, int cpu)
{
struct cache *next = cache;
while (next) {
WARN_ONCE(cpumask_test_cpu(cpu, &next->shared_cpu_map),
"CPU %i already accounted in %pOFP(%s)\n",
cpu, next->ofnode,
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
cache_type_string(next));
cpumask_set_cpu(cpu, &next->shared_cpu_map);
next = next->next_local;
}
}
static int cache_size(const struct cache *cache, unsigned int *ret)
{
const char *propname;
const __be32 *cache_size;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
propname = cache_type_info[cache->type].size_prop;
cache_size = of_get_property(cache->ofnode, propname, NULL);
if (!cache_size)
return -ENODEV;
*ret = of_read_number(cache_size, 1);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
return 0;
}
static int cache_size_kb(const struct cache *cache, unsigned int *ret)
{
unsigned int size;
if (cache_size(cache, &size))
return -ENODEV;
*ret = size / 1024;
return 0;
}
/* not cache_line_size() because that's a macro in include/linux/cache.h */
static int cache_get_line_size(const struct cache *cache, unsigned int *ret)
{
const __be32 *line_size;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
int i, lim;
lim = ARRAY_SIZE(cache_type_info[cache->type].line_size_props);
for (i = 0; i < lim; i++) {
const char *propname;
propname = cache_type_info[cache->type].line_size_props[i];
line_size = of_get_property(cache->ofnode, propname, NULL);
if (line_size)
break;
}
if (!line_size)
return -ENODEV;
*ret = of_read_number(line_size, 1);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
return 0;
}
static int cache_nr_sets(const struct cache *cache, unsigned int *ret)
{
const char *propname;
const __be32 *nr_sets;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
propname = cache_type_info[cache->type].nr_sets_prop;
nr_sets = of_get_property(cache->ofnode, propname, NULL);
if (!nr_sets)
return -ENODEV;
*ret = of_read_number(nr_sets, 1);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
return 0;
}
static int cache_associativity(const struct cache *cache, unsigned int *ret)
{
unsigned int line_size;
unsigned int nr_sets;
unsigned int size;
if (cache_nr_sets(cache, &nr_sets))
goto err;
/* If the cache is fully associative, there is no need to
* check the other properties.
*/
if (nr_sets == 1) {
*ret = 0;
return 0;
}
if (cache_get_line_size(cache, &line_size))
goto err;
if (cache_size(cache, &size))
goto err;
if (!(nr_sets > 0 && size > 0 && line_size > 0))
goto err;
*ret = (size / nr_sets) / line_size;
return 0;
err:
return -ENODEV;
}
/* helper for dealing with split caches */
static struct cache *cache_find_first_sibling(struct cache *cache)
{
struct cache *iter;
if (cache->type == CACHE_TYPE_UNIFIED ||
cache->type == CACHE_TYPE_UNIFIED_D)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
return cache;
list_for_each_entry(iter, &cache_list, list)
if (iter->ofnode == cache->ofnode && iter->next_local == cache)
return iter;
return cache;
}
/* return the first cache on a local list matching node */
static struct cache *cache_lookup_by_node(const struct device_node *node)
{
struct cache *cache = NULL;
struct cache *iter;
list_for_each_entry(iter, &cache_list, list) {
if (iter->ofnode != node)
continue;
cache = cache_find_first_sibling(iter);
break;
}
return cache;
}
static bool cache_node_is_unified(const struct device_node *np)
{
return of_get_property(np, "cache-unified", NULL);
}
/*
* Unified caches can have two different sets of tags. Most embedded
* use cache-size, etc. for the unified cache size, but open firmware systems
* use d-cache-size, etc. Check on initialization for which type we have, and
* return the appropriate structure type. Assume it's embedded if it isn't
* open firmware. If it's yet a 3rd type, then there will be missing entries
* in /sys/devices/system/cpu/cpu0/cache/index2/, and this code will need
* to be extended further.
*/
static int cache_is_unified_d(const struct device_node *np)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
return of_get_property(np,
cache_type_info[CACHE_TYPE_UNIFIED_D].size_prop, NULL) ?
CACHE_TYPE_UNIFIED_D : CACHE_TYPE_UNIFIED;
}
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
static struct cache *cache_do_one_devnode_unified(struct device_node *node, int level)
{
pr_debug("creating L%d ucache for %pOFP\n", level, node);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
return new_cache(cache_is_unified_d(node), level, node);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
}
static struct cache *cache_do_one_devnode_split(struct device_node *node,
int level)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache *dcache, *icache;
pr_debug("creating L%d dcache and icache for %pOFP\n", level,
node);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
dcache = new_cache(CACHE_TYPE_DATA, level, node);
icache = new_cache(CACHE_TYPE_INSTRUCTION, level, node);
if (!dcache || !icache)
goto err;
dcache->next_local = icache;
return dcache;
err:
release_cache(dcache);
release_cache(icache);
return NULL;
}
static struct cache *cache_do_one_devnode(struct device_node *node, int level)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache *cache;
if (cache_node_is_unified(node))
cache = cache_do_one_devnode_unified(node, level);
else
cache = cache_do_one_devnode_split(node, level);
return cache;
}
static struct cache *cache_lookup_or_instantiate(struct device_node *node,
int level)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache *cache;
cache = cache_lookup_by_node(node);
WARN_ONCE(cache && cache->level != level,
"cache level mismatch on lookup (got %d, expected %d)\n",
cache->level, level);
if (!cache)
cache = cache_do_one_devnode(node, level);
return cache;
}
static void link_cache_lists(struct cache *smaller, struct cache *bigger)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
while (smaller->next_local) {
if (smaller->next_local == bigger)
return; /* already linked */
smaller = smaller->next_local;
}
smaller->next_local = bigger;
}
static void do_subsidiary_caches_debugcheck(struct cache *cache)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
WARN_ONCE(cache->level != 1,
"instantiating cache chain from L%d %s cache for "
"%pOFP instead of an L1\n", cache->level,
cache_type_string(cache), cache->ofnode);
WARN_ONCE(!of_node_is_type(cache->ofnode, "cpu"),
"instantiating cache chain from node %pOFP of type '%s' "
"instead of a cpu node\n", cache->ofnode,
of_node_get_device_type(cache->ofnode));
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
}
static void do_subsidiary_caches(struct cache *cache)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct device_node *subcache_node;
int level = cache->level;
do_subsidiary_caches_debugcheck(cache);
while ((subcache_node = of_find_next_cache_node(cache->ofnode))) {
struct cache *subcache;
level++;
subcache = cache_lookup_or_instantiate(subcache_node, level);
of_node_put(subcache_node);
if (!subcache)
break;
link_cache_lists(cache, subcache);
cache = subcache;
}
}
static struct cache *cache_chain_instantiate(unsigned int cpu_id)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct device_node *cpu_node;
struct cache *cpu_cache = NULL;
pr_debug("creating cache object(s) for CPU %i\n", cpu_id);
cpu_node = of_get_cpu_node(cpu_id, NULL);
WARN_ONCE(!cpu_node, "no OF node found for CPU %i\n", cpu_id);
if (!cpu_node)
goto out;
cpu_cache = cache_lookup_or_instantiate(cpu_node, 1);
if (!cpu_cache)
goto out;
do_subsidiary_caches(cpu_cache);
cache_cpu_set(cpu_cache, cpu_id);
out:
of_node_put(cpu_node);
return cpu_cache;
}
static struct cache_dir *cacheinfo_create_cache_dir(unsigned int cpu_id)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache_dir *cache_dir;
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
struct device *dev;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
struct kobject *kobj = NULL;
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
dev = get_cpu_device(cpu_id);
WARN_ONCE(!dev, "no dev for CPU %i\n", cpu_id);
if (!dev)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
goto err;
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
kobj = kobject_create_and_add("cache", &dev->kobj);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
if (!kobj)
goto err;
cache_dir = kzalloc(sizeof(*cache_dir), GFP_KERNEL);
if (!cache_dir)
goto err;
cache_dir->kobj = kobj;
WARN_ON_ONCE(per_cpu(cache_dir_pcpu, cpu_id) != NULL);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
per_cpu(cache_dir_pcpu, cpu_id) = cache_dir;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
return cache_dir;
err:
kobject_put(kobj);
return NULL;
}
static void cache_index_release(struct kobject *kobj)
{
struct cache_index_dir *index;
index = kobj_to_cache_index_dir(kobj);
pr_debug("freeing index directory for L%d %s cache\n",
index->cache->level, cache_type_string(index->cache));
kfree(index);
}
static ssize_t cache_index_show(struct kobject *k, struct attribute *attr, char *buf)
{
struct kobj_attribute *kobj_attr;
kobj_attr = container_of(attr, struct kobj_attribute, attr);
return kobj_attr->show(k, kobj_attr, buf);
}
static struct cache *index_kobj_to_cache(struct kobject *k)
{
struct cache_index_dir *index;
index = kobj_to_cache_index_dir(k);
return index->cache;
}
static ssize_t size_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
unsigned int size_kb;
struct cache *cache;
cache = index_kobj_to_cache(k);
if (cache_size_kb(cache, &size_kb))
return -ENODEV;
return sprintf(buf, "%uK\n", size_kb);
}
static struct kobj_attribute cache_size_attr =
__ATTR(size, 0444, size_show, NULL);
static ssize_t line_size_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
unsigned int line_size;
struct cache *cache;
cache = index_kobj_to_cache(k);
if (cache_get_line_size(cache, &line_size))
return -ENODEV;
return sprintf(buf, "%u\n", line_size);
}
static struct kobj_attribute cache_line_size_attr =
__ATTR(coherency_line_size, 0444, line_size_show, NULL);
static ssize_t nr_sets_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
unsigned int nr_sets;
struct cache *cache;
cache = index_kobj_to_cache(k);
if (cache_nr_sets(cache, &nr_sets))
return -ENODEV;
return sprintf(buf, "%u\n", nr_sets);
}
static struct kobj_attribute cache_nr_sets_attr =
__ATTR(number_of_sets, 0444, nr_sets_show, NULL);
static ssize_t associativity_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
unsigned int associativity;
struct cache *cache;
cache = index_kobj_to_cache(k);
if (cache_associativity(cache, &associativity))
return -ENODEV;
return sprintf(buf, "%u\n", associativity);
}
static struct kobj_attribute cache_assoc_attr =
__ATTR(ways_of_associativity, 0444, associativity_show, NULL);
static ssize_t type_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
struct cache *cache;
cache = index_kobj_to_cache(k);
return sprintf(buf, "%s\n", cache_type_string(cache));
}
static struct kobj_attribute cache_type_attr =
__ATTR(type, 0444, type_show, NULL);
static ssize_t level_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
struct cache_index_dir *index;
struct cache *cache;
index = kobj_to_cache_index_dir(k);
cache = index->cache;
return sprintf(buf, "%d\n", cache->level);
}
static struct kobj_attribute cache_level_attr =
__ATTR(level, 0444, level_show, NULL);
static unsigned int index_dir_to_cpu(struct cache_index_dir *index)
{
struct kobject *index_dir_kobj = &index->kobj;
struct kobject *cache_dir_kobj = index_dir_kobj->parent;
struct kobject *cpu_dev_kobj = cache_dir_kobj->parent;
struct device *dev = kobj_to_dev(cpu_dev_kobj);
return dev->id;
}
/*
* On big-core systems, each core has two groups of CPUs each of which
* has its own L1-cache. The thread-siblings which share l1-cache with
* @cpu can be obtained via cpu_smallcore_mask().
*/
static const struct cpumask *get_big_core_shared_cpu_map(int cpu, struct cache *cache)
{
if (cache->level == 1)
return cpu_smallcore_mask(cpu);
return &cache->shared_cpu_map;
}
static ssize_t
show_shared_cpumap(struct kobject *k, struct kobj_attribute *attr, char *buf, bool list)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache_index_dir *index;
struct cache *cache;
const struct cpumask *mask;
int cpu;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
index = kobj_to_cache_index_dir(k);
cache = index->cache;
if (has_big_cores) {
cpu = index_dir_to_cpu(index);
mask = get_big_core_shared_cpu_map(cpu, cache);
} else {
mask = &cache->shared_cpu_map;
}
return cpumap_print_to_pagebuf(list, buf, mask);
}
static ssize_t shared_cpu_map_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
return show_shared_cpumap(k, attr, buf, false);
}
static ssize_t shared_cpu_list_show(struct kobject *k, struct kobj_attribute *attr, char *buf)
{
return show_shared_cpumap(k, attr, buf, true);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
}
static struct kobj_attribute cache_shared_cpu_map_attr =
__ATTR(shared_cpu_map, 0444, shared_cpu_map_show, NULL);
static struct kobj_attribute cache_shared_cpu_list_attr =
__ATTR(shared_cpu_list, 0444, shared_cpu_list_show, NULL);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
/* Attributes which should always be created -- the kobject/sysfs core
* does this automatically via kobj_type->default_attrs. This is the
* minimum data required to uniquely identify a cache.
*/
static struct attribute *cache_index_default_attrs[] = {
&cache_type_attr.attr,
&cache_level_attr.attr,
&cache_shared_cpu_map_attr.attr,
&cache_shared_cpu_list_attr.attr,
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
NULL,
};
/* Attributes which should be created if the cache device node has the
* right properties -- see cacheinfo_create_index_opt_attrs
*/
static struct kobj_attribute *cache_index_opt_attrs[] = {
&cache_size_attr,
&cache_line_size_attr,
&cache_nr_sets_attr,
&cache_assoc_attr,
};
static const struct sysfs_ops cache_index_ops = {
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
.show = cache_index_show,
};
static struct kobj_type cache_index_type = {
.release = cache_index_release,
.sysfs_ops = &cache_index_ops,
.default_attrs = cache_index_default_attrs,
};
static void cacheinfo_create_index_opt_attrs(struct cache_index_dir *dir)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
const char *cache_type;
struct cache *cache;
char *buf;
int i;
buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
if (!buf)
return;
cache = dir->cache;
cache_type = cache_type_string(cache);
/* We don't want to create an attribute that can't provide a
* meaningful value. Check the return value of each optional
* attribute's ->show method before registering the
* attribute.
*/
for (i = 0; i < ARRAY_SIZE(cache_index_opt_attrs); i++) {
struct kobj_attribute *attr;
ssize_t rc;
attr = cache_index_opt_attrs[i];
rc = attr->show(&dir->kobj, attr, buf);
if (rc <= 0) {
pr_debug("not creating %s attribute for "
"%pOFP(%s) (rc = %zd)\n",
attr->attr.name, cache->ofnode,
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
cache_type, rc);
continue;
}
if (sysfs_create_file(&dir->kobj, &attr->attr))
pr_debug("could not create %s attribute for %pOFP(%s)\n",
attr->attr.name, cache->ofnode, cache_type);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
}
kfree(buf);
}
static void cacheinfo_create_index_dir(struct cache *cache, int index,
struct cache_dir *cache_dir)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache_index_dir *index_dir;
int rc;
index_dir = kzalloc(sizeof(*index_dir), GFP_KERNEL);
if (!index_dir)
return;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
index_dir->cache = cache;
rc = kobject_init_and_add(&index_dir->kobj, &cache_index_type,
cache_dir->kobj, "index%d", index);
if (rc) {
kobject_put(&index_dir->kobj);
return;
}
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
index_dir->next = cache_dir->index;
cache_dir->index = index_dir;
cacheinfo_create_index_opt_attrs(index_dir);
}
static void cacheinfo_sysfs_populate(unsigned int cpu_id,
struct cache *cache_list)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache_dir *cache_dir;
struct cache *cache;
int index = 0;
cache_dir = cacheinfo_create_cache_dir(cpu_id);
if (!cache_dir)
return;
cache = cache_list;
while (cache) {
cacheinfo_create_index_dir(cache, index, cache_dir);
index++;
cache = cache->next_local;
}
}
void cacheinfo_cpu_online(unsigned int cpu_id)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
{
struct cache *cache;
cache = cache_chain_instantiate(cpu_id);
if (!cache)
return;
cacheinfo_sysfs_populate(cpu_id, cache);
}
/* functions needed to remove cache entry for cpu offline or suspend/resume */
#if (defined(CONFIG_PPC_PSERIES) && defined(CONFIG_SUSPEND)) || \
defined(CONFIG_HOTPLUG_CPU)
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
static struct cache *cache_lookup_by_cpu(unsigned int cpu_id)
{
struct device_node *cpu_node;
struct cache *cache;
cpu_node = of_get_cpu_node(cpu_id, NULL);
WARN_ONCE(!cpu_node, "no OF node found for CPU %i\n", cpu_id);
if (!cpu_node)
return NULL;
cache = cache_lookup_by_node(cpu_node);
of_node_put(cpu_node);
return cache;
}
static void remove_index_dirs(struct cache_dir *cache_dir)
{
struct cache_index_dir *index;
index = cache_dir->index;
while (index) {
struct cache_index_dir *next;
next = index->next;
kobject_put(&index->kobj);
index = next;
}
}
static void remove_cache_dir(struct cache_dir *cache_dir)
{
remove_index_dirs(cache_dir);
/* Remove cache dir from sysfs */
kobject_del(cache_dir->kobj);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
kobject_put(cache_dir->kobj);
kfree(cache_dir);
}
static void cache_cpu_clear(struct cache *cache, int cpu)
{
while (cache) {
struct cache *next = cache->next_local;
WARN_ONCE(!cpumask_test_cpu(cpu, &cache->shared_cpu_map),
"CPU %i not accounted in %pOFP(%s)\n",
cpu, cache->ofnode,
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
cache_type_string(cache));
cpumask_clear_cpu(cpu, &cache->shared_cpu_map);
/* Release the cache object if all the cpus using it
* are offline */
if (cpumask_empty(&cache->shared_cpu_map))
release_cache(cache);
cache = next;
}
}
void cacheinfo_cpu_offline(unsigned int cpu_id)
{
struct cache_dir *cache_dir;
struct cache *cache;
/* Prevent userspace from seeing inconsistent state - remove
* the sysfs hierarchy first */
cache_dir = per_cpu(cache_dir_pcpu, cpu_id);
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
/* careful, sysfs population may have failed */
if (cache_dir)
remove_cache_dir(cache_dir);
per_cpu(cache_dir_pcpu, cpu_id) = NULL;
powerpc: Rewrite sysfs processor cache info code The current code for providing processor cache information in sysfs has the following deficiencies: - several complex functions that are hard to understand - implicit recursion (cache_desc_release -> kobject_put -> cache_desc_release) - explicit recursion (create_cache_index_info) - use of two per-cpu arrays when one would suffice - duplication of work on systems where CPUs share cache Also, when I looked at implementing support for a shared_cpu_map attribute, it was pretty much impossible to handle hotplug without checking every single online CPU's cache_desc list and fixing things up... not that this is a hot path, but it would have introduced O(n^2)-ish behavior during boot. Addressing this involved rethinking the core data structures used, which didn't lend itself to an incremental approach. This implementation maintains a "forest" (potentially more than one tree) of cache objects which reflects the system's cache topology. Cache objects are instantiated as needed as CPUs come online. A per-cpu array is used mainly for sysfs-related bookkeeping; the objects in the array just point to the appropriate points in the forest. This maintains compatibility with the existing code and includes some enhancements: - Implement the shared_cpu_map attribute, which is essential for enabling userspace to discover the system's overall cache topology. - Use cache-block-size properties if cache-line-size is not available. I chose to place this implementation in a new file since it would have roughly doubled the size of sysfs.c, which is already kind of messy. Signed-off-by: Nathan Lynch <ntl@pobox.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2008-12-24 02:55:54 +08:00
/* clear the CPU's bit in its cache chain, possibly freeing
* cache objects */
cache = cache_lookup_by_cpu(cpu_id);
if (cache)
cache_cpu_clear(cache, cpu_id);
}
void cacheinfo_teardown(void)
{
unsigned int cpu;
lockdep_assert_cpus_held();
for_each_online_cpu(cpu)
cacheinfo_cpu_offline(cpu);
}
void cacheinfo_rebuild(void)
{
unsigned int cpu;
lockdep_assert_cpus_held();
for_each_online_cpu(cpu)
cacheinfo_cpu_online(cpu);
}
#endif /* (CONFIG_PPC_PSERIES && CONFIG_SUSPEND) || CONFIG_HOTPLUG_CPU */