asgardius/android_kernel_motorola_sm6225
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

[PATCH] kernel/cpuset.c, mutex conversion

Browse source 
convert cpuset.c's callback_sem and manage_sem to mutexes.
Build and boot tested by Ingo.
Build, boot, unit and stress tested by pj.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This commit is contained in:
Ingo Molnar 2006-03-23 03:00:18 -08:00 committed by Linus Torvalds
parent 6362e4d4ed
commit 3d3f26a7ba
1 changed files with 103 additions and 109 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

212
kernel/cpuset.c
View file

@ -53,7 +53,7 @@
#include <asm/uaccess.h> #include <asm/uaccess.h>
#include <asm/atomic.h> #include <asm/atomic.h>
#include <asm/semaphore.h> #include <linux/mutex.h>
#define CPUSET_SUPER_MAGIC 0x27e0eb #define CPUSET_SUPER_MAGIC 0x27e0eb
@ -168,63 +168,57 @@ static struct vfsmount *cpuset_mount;
static struct super_block *cpuset_sb; static struct super_block *cpuset_sb;
/* /*
* We have two global cpuset semaphores below. They can nest. * We have two global cpuset mutexes below. They can nest.
* It is ok to first take manage_sem, then nest callback_sem. We also * It is ok to first take manage_mutex, then nest callback_mutex. We also
* require taking task_lock() when dereferencing a tasks cpuset pointer. * require taking task_lock() when dereferencing a tasks cpuset pointer.
* See "The task_lock() exception", at the end of this comment. * See "The task_lock() exception", at the end of this comment.
* *
* A task must hold both semaphores to modify cpusets. If a task * A task must hold both mutexes to modify cpusets. If a task
* holds manage_sem, then it blocks others wanting that semaphore, * holds manage_mutex, then it blocks others wanting that mutex,
* ensuring that it is the only task able to also acquire callback_sem * ensuring that it is the only task able to also acquire callback_mutex
* and be able to modify cpusets. It can perform various checks on * and be able to modify cpusets. It can perform various checks on
* the cpuset structure first, knowing nothing will change. It can * the cpuset structure first, knowing nothing will change. It can
* also allocate memory while just holding manage_sem. While it is * also allocate memory while just holding manage_mutex. While it is
* performing these checks, various callback routines can briefly * performing these checks, various callback routines can briefly
* acquire callback_sem to query cpusets. Once it is ready to make * acquire callback_mutex to query cpusets. Once it is ready to make
* the changes, it takes callback_sem, blocking everyone else. * the changes, it takes callback_mutex, blocking everyone else.
* *
* Calls to the kernel memory allocator can not be made while holding * Calls to the kernel memory allocator can not be made while holding
* callback_sem, as that would risk double tripping on callback_sem * callback_mutex, as that would risk double tripping on callback_mutex
* from one of the callbacks into the cpuset code from within * from one of the callbacks into the cpuset code from within
* __alloc_pages(). * __alloc_pages().
* *
* If a task is only holding callback_sem, then it has read-only * If a task is only holding callback_mutex, then it has read-only
* access to cpusets. * access to cpusets.
* *
* The task_struct fields mems_allowed and mems_generation may only * The task_struct fields mems_allowed and mems_generation may only
* be accessed in the context of that task, so require no locks. * be accessed in the context of that task, so require no locks.
* *
* Any task can increment and decrement the count field without lock. * Any task can increment and decrement the count field without lock.
* So in general, code holding manage_sem or callback_sem can't rely * So in general, code holding manage_mutex or callback_mutex can't rely
* on the count field not changing. However, if the count goes to * on the count field not changing. However, if the count goes to
* zero, then only attach_task(), which holds both semaphores, can * zero, then only attach_task(), which holds both mutexes, can
* increment it again. Because a count of zero means that no tasks * increment it again. Because a count of zero means that no tasks
* are currently attached, therefore there is no way a task attached * are currently attached, therefore there is no way a task attached
* to that cpuset can fork (the other way to increment the count). * to that cpuset can fork (the other way to increment the count).
* So code holding manage_sem or callback_sem can safely assume that * So code holding manage_mutex or callback_mutex can safely assume that
* if the count is zero, it will stay zero. Similarly, if a task * if the count is zero, it will stay zero. Similarly, if a task
* holds manage_sem or callback_sem on a cpuset with zero count, it * holds manage_mutex or callback_mutex on a cpuset with zero count, it
* knows that the cpuset won't be removed, as cpuset_rmdir() needs * knows that the cpuset won't be removed, as cpuset_rmdir() needs
* both of those semaphores. * both of those mutexes.
*
* A possible optimization to improve parallelism would be to make
* callback_sem a R/W semaphore (rwsem), allowing the callback routines
* to proceed in parallel, with read access, until the holder of
* manage_sem needed to take this rwsem for exclusive write access
* and modify some cpusets.
* *
* The cpuset_common_file_write handler for operations that modify * The cpuset_common_file_write handler for operations that modify
* the cpuset hierarchy holds manage_sem across the entire operation, * the cpuset hierarchy holds manage_mutex across the entire operation,
* single threading all such cpuset modifications across the system. * single threading all such cpuset modifications across the system.
* *
* The cpuset_common_file_read() handlers only hold callback_sem across * The cpuset_common_file_read() handlers only hold callback_mutex across
* small pieces of code, such as when reading out possibly multi-word * small pieces of code, such as when reading out possibly multi-word
* cpumasks and nodemasks. * cpumasks and nodemasks.
* *
* The fork and exit callbacks cpuset_fork() and cpuset_exit(), don't * The fork and exit callbacks cpuset_fork() and cpuset_exit(), don't
* (usually) take either semaphore. These are the two most performance * (usually) take either mutex. These are the two most performance
* critical pieces of code here. The exception occurs on cpuset_exit(), * critical pieces of code here. The exception occurs on cpuset_exit(),
* when a task in a notify_on_release cpuset exits. Then manage_sem * when a task in a notify_on_release cpuset exits. Then manage_mutex
* is taken, and if the cpuset count is zero, a usermode call made * is taken, and if the cpuset count is zero, a usermode call made
* to /sbin/cpuset_release_agent with the name of the cpuset (path * to /sbin/cpuset_release_agent with the name of the cpuset (path
* relative to the root of cpuset file system) as the argument. * relative to the root of cpuset file system) as the argument.
@ -242,9 +236,9 @@ static struct super_block *cpuset_sb;
* *
* The need for this exception arises from the action of attach_task(), * The need for this exception arises from the action of attach_task(),
* which overwrites one tasks cpuset pointer with another. It does * which overwrites one tasks cpuset pointer with another. It does
* so using both semaphores, however there are several performance * so using both mutexes, however there are several performance
* critical places that need to reference task->cpuset without the * critical places that need to reference task->cpuset without the
* expense of grabbing a system global semaphore. Therefore except as * expense of grabbing a system global mutex. Therefore except as
* noted below, when dereferencing or, as in attach_task(), modifying * noted below, when dereferencing or, as in attach_task(), modifying
* a tasks cpuset pointer we use task_lock(), which acts on a spinlock * a tasks cpuset pointer we use task_lock(), which acts on a spinlock
* (task->alloc_lock) already in the task_struct routinely used for * (task->alloc_lock) already in the task_struct routinely used for
@ -256,8 +250,8 @@ static struct super_block *cpuset_sb;
* the routine cpuset_update_task_memory_state(). * the routine cpuset_update_task_memory_state().
*/ */
static DECLARE_MUTEX(manage_sem); static DEFINE_MUTEX(manage_mutex);
static DECLARE_MUTEX(callback_sem); static DEFINE_MUTEX(callback_mutex);
/* /*
* A couple of forward declarations required, due to cyclic reference loop: * A couple of forward declarations required, due to cyclic reference loop:
@ -432,7 +426,7 @@ static inline struct cftype *__d_cft(struct dentry *dentry)
} }
/* /*
* Call with manage_sem held. Writes path of cpuset into buf. * Call with manage_mutex held. Writes path of cpuset into buf.
* Returns 0 on success, -errno on error. * Returns 0 on success, -errno on error.
*/ */
@ -484,11 +478,11 @@ static int cpuset_path(const struct cpuset *cs, char *buf, int buflen)
* status of the /sbin/cpuset_release_agent task, so no sense holding * status of the /sbin/cpuset_release_agent task, so no sense holding
* our caller up for that. * our caller up for that.
* *
* When we had only one cpuset semaphore, we had to call this * When we had only one cpuset mutex, we had to call this
* without holding it, to avoid deadlock when call_usermodehelper() * without holding it, to avoid deadlock when call_usermodehelper()
* allocated memory. With two locks, we could now call this while * allocated memory. With two locks, we could now call this while
* holding manage_sem, but we still don't, so as to minimize * holding manage_mutex, but we still don't, so as to minimize
* the time manage_sem is held. * the time manage_mutex is held.
*/ */
static void cpuset_release_agent(const char *pathbuf) static void cpuset_release_agent(const char *pathbuf)
@ -520,15 +514,15 @@ static void cpuset_release_agent(const char *pathbuf)
* cs is notify_on_release() and now both the user count is zero and * cs is notify_on_release() and now both the user count is zero and
* the list of children is empty, prepare cpuset path in a kmalloc'd * the list of children is empty, prepare cpuset path in a kmalloc'd
* buffer, to be returned via ppathbuf, so that the caller can invoke * buffer, to be returned via ppathbuf, so that the caller can invoke
* cpuset_release_agent() with it later on, once manage_sem is dropped. * cpuset_release_agent() with it later on, once manage_mutex is dropped.
* Call here with manage_sem held. * Call here with manage_mutex held.
* *
* This check_for_release() routine is responsible for kmalloc'ing * This check_for_release() routine is responsible for kmalloc'ing
* pathbuf. The above cpuset_release_agent() is responsible for * pathbuf. The above cpuset_release_agent() is responsible for
* kfree'ing pathbuf. The caller of these routines is responsible * kfree'ing pathbuf. The caller of these routines is responsible
* for providing a pathbuf pointer, initialized to NULL, then * for providing a pathbuf pointer, initialized to NULL, then
* calling check_for_release() with manage_sem held and the address * calling check_for_release() with manage_mutex held and the address
* of the pathbuf pointer, then dropping manage_sem, then calling * of the pathbuf pointer, then dropping manage_mutex, then calling
* cpuset_release_agent() with pathbuf, as set by check_for_release(). * cpuset_release_agent() with pathbuf, as set by check_for_release().
*/ */
@ -559,7 +553,7 @@ static void check_for_release(struct cpuset *cs, char **ppathbuf)
* One way or another, we guarantee to return some non-empty subset * One way or another, we guarantee to return some non-empty subset
* of cpu_online_map. * of cpu_online_map.
* *
* Call with callback_sem held. * Call with callback_mutex held.
*/ */
static void guarantee_online_cpus(const struct cpuset *cs, cpumask_t *pmask) static void guarantee_online_cpus(const struct cpuset *cs, cpumask_t *pmask)
@ -583,7 +577,7 @@ static void guarantee_online_cpus(const struct cpuset *cs, cpumask_t *pmask)
* One way or another, we guarantee to return some non-empty subset * One way or another, we guarantee to return some non-empty subset
* of node_online_map. * of node_online_map.
* *
* Call with callback_sem held. * Call with callback_mutex held.
*/ */
static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask) static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
@ -608,12 +602,12 @@ static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
* current->cpuset if a task has its memory placement changed. * current->cpuset if a task has its memory placement changed.
* Do not call this routine if in_interrupt(). * Do not call this routine if in_interrupt().
* *
* Call without callback_sem or task_lock() held. May be called * Call without callback_mutex or task_lock() held. May be called
* with or without manage_sem held. Doesn't need task_lock to guard * with or without manage_mutex held. Doesn't need task_lock to guard
* against another task changing a non-NULL cpuset pointer to NULL, * against another task changing a non-NULL cpuset pointer to NULL,
* as that is only done by a task on itself, and if the current task * as that is only done by a task on itself, and if the current task
* is here, it is not simultaneously in the exit code NULL'ing its * is here, it is not simultaneously in the exit code NULL'ing its
* cpuset pointer. This routine also might acquire callback_sem and * cpuset pointer. This routine also might acquire callback_mutex and
* current->mm->mmap_sem during call. * current->mm->mmap_sem during call.
* *
* Reading current->cpuset->mems_generation doesn't need task_lock * Reading current->cpuset->mems_generation doesn't need task_lock
@ -658,13 +652,13 @@ void cpuset_update_task_memory_state(void)
} }
if (my_cpusets_mem_gen != tsk->cpuset_mems_generation) { if (my_cpusets_mem_gen != tsk->cpuset_mems_generation) {
down(&callback_sem); mutex_lock(&callback_mutex);
task_lock(tsk); task_lock(tsk);
cs = tsk->cpuset; /* Maybe changed when task not locked */ cs = tsk->cpuset; /* Maybe changed when task not locked */
guarantee_online_mems(cs, &tsk->mems_allowed); guarantee_online_mems(cs, &tsk->mems_allowed);
tsk->cpuset_mems_generation = cs->mems_generation; tsk->cpuset_mems_generation = cs->mems_generation;
task_unlock(tsk); task_unlock(tsk);
up(&callback_sem); mutex_unlock(&callback_mutex);
mpol_rebind_task(tsk, &tsk->mems_allowed); mpol_rebind_task(tsk, &tsk->mems_allowed);
} }
} }
@ -674,7 +668,7 @@ void cpuset_update_task_memory_state(void)
* *
* One cpuset is a subset of another if all its allowed CPUs and * One cpuset is a subset of another if all its allowed CPUs and
* Memory Nodes are a subset of the other, and its exclusive flags * Memory Nodes are a subset of the other, and its exclusive flags
* are only set if the other's are set. Call holding manage_sem. * are only set if the other's are set. Call holding manage_mutex.
*/ */
static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q) static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
@ -692,7 +686,7 @@ static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
* If we replaced the flag and mask values of the current cpuset * If we replaced the flag and mask values of the current cpuset
* (cur) with those values in the trial cpuset (trial), would * (cur) with those values in the trial cpuset (trial), would
* our various subset and exclusive rules still be valid? Presumes * our various subset and exclusive rules still be valid? Presumes
* manage_sem held. * manage_mutex held.
* *
* 'cur' is the address of an actual, in-use cpuset. Operations * 'cur' is the address of an actual, in-use cpuset. Operations
* such as list traversal that depend on the actual address of the * such as list traversal that depend on the actual address of the
@ -746,7 +740,7 @@ static int validate_change(const struct cpuset *cur, const struct cpuset *trial)
* exclusive child cpusets * exclusive child cpusets
* Build these two partitions by calling partition_sched_domains * Build these two partitions by calling partition_sched_domains
* *
* Call with manage_sem held. May nest a call to the * Call with manage_mutex held. May nest a call to the
* lock_cpu_hotplug()/unlock_cpu_hotplug() pair. * lock_cpu_hotplug()/unlock_cpu_hotplug() pair.
*/ */
@ -792,7 +786,7 @@ static void update_cpu_domains(struct cpuset *cur)
} }
/* /*
* Call with manage_sem held. May take callback_sem during call. * Call with manage_mutex held. May take callback_mutex during call.
*/ */
static int update_cpumask(struct cpuset *cs, char *buf) static int update_cpumask(struct cpuset *cs, char *buf)
@ -811,9 +805,9 @@ static int update_cpumask(struct cpuset *cs, char *buf)
if (retval < 0) if (retval < 0)
return retval; return retval;
cpus_unchanged = cpus_equal(cs->cpus_allowed, trialcs.cpus_allowed); cpus_unchanged = cpus_equal(cs->cpus_allowed, trialcs.cpus_allowed);
down(&callback_sem); mutex_lock(&callback_mutex);
cs->cpus_allowed = trialcs.cpus_allowed; cs->cpus_allowed = trialcs.cpus_allowed;
up(&callback_sem); mutex_unlock(&callback_mutex);
if (is_cpu_exclusive(cs) && !cpus_unchanged) if (is_cpu_exclusive(cs) && !cpus_unchanged)
update_cpu_domains(cs); update_cpu_domains(cs);
return 0; return 0;
@ -827,7 +821,7 @@ static int update_cpumask(struct cpuset *cs, char *buf)
* the cpuset is marked 'memory_migrate', migrate the tasks * the cpuset is marked 'memory_migrate', migrate the tasks
* pages to the new memory. * pages to the new memory.
* *
* Call with manage_sem held. May take callback_sem during call. * Call with manage_mutex held. May take callback_mutex during call.
* Will take tasklist_lock, scan tasklist for tasks in cpuset cs, * Will take tasklist_lock, scan tasklist for tasks in cpuset cs,
* lock each such tasks mm->mmap_sem, scan its vma's and rebind * lock each such tasks mm->mmap_sem, scan its vma's and rebind
* their mempolicies to the cpusets new mems_allowed. * their mempolicies to the cpusets new mems_allowed.
@ -862,11 +856,11 @@ static int update_nodemask(struct cpuset *cs, char *buf)
if (retval < 0) if (retval < 0)
goto done; goto done;
down(&callback_sem); mutex_lock(&callback_mutex);
cs->mems_allowed = trialcs.mems_allowed; cs->mems_allowed = trialcs.mems_allowed;
atomic_inc(&cpuset_mems_generation); atomic_inc(&cpuset_mems_generation);
cs->mems_generation = atomic_read(&cpuset_mems_generation); cs->mems_generation = atomic_read(&cpuset_mems_generation);
up(&callback_sem); mutex_unlock(&callback_mutex);
set_cpuset_being_rebound(cs); /* causes mpol_copy() rebind */ set_cpuset_being_rebound(cs); /* causes mpol_copy() rebind */
@ -922,7 +916,7 @@ static int update_nodemask(struct cpuset *cs, char *buf)
* tasklist_lock. Forks can happen again now - the mpol_copy() * tasklist_lock. Forks can happen again now - the mpol_copy()
* cpuset_being_rebound check will catch such forks, and rebind * cpuset_being_rebound check will catch such forks, and rebind
* their vma mempolicies too. Because we still hold the global * their vma mempolicies too. Because we still hold the global
* cpuset manage_sem, we know that no other rebind effort will * cpuset manage_mutex, we know that no other rebind effort will
* be contending for the global variable cpuset_being_rebound. * be contending for the global variable cpuset_being_rebound.
* It's ok if we rebind the same mm twice; mpol_rebind_mm() * It's ok if we rebind the same mm twice; mpol_rebind_mm()
* is idempotent. Also migrate pages in each mm to new nodes. * is idempotent. Also migrate pages in each mm to new nodes.
@ -948,7 +942,7 @@ done:
} }
/* /*
* Call with manage_sem held. * Call with manage_mutex held.
*/ */
static int update_memory_pressure_enabled(struct cpuset *cs, char *buf) static int update_memory_pressure_enabled(struct cpuset *cs, char *buf)
@ -967,7 +961,7 @@ static int update_memory_pressure_enabled(struct cpuset *cs, char *buf)
* cs: the cpuset to update * cs: the cpuset to update
* buf: the buffer where we read the 0 or 1 * buf: the buffer where we read the 0 or 1
* *
* Call with manage_sem held. * Call with manage_mutex held.
*/ */
static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs, char *buf) static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs, char *buf)
@ -989,12 +983,12 @@ static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs, char *buf)
return err; return err;
cpu_exclusive_changed = cpu_exclusive_changed =
(is_cpu_exclusive(cs) != is_cpu_exclusive(&trialcs)); (is_cpu_exclusive(cs) != is_cpu_exclusive(&trialcs));
down(&callback_sem); mutex_lock(&callback_mutex);
if (turning_on) if (turning_on)
set_bit(bit, &cs->flags); set_bit(bit, &cs->flags);
else else
clear_bit(bit, &cs->flags); clear_bit(bit, &cs->flags);
up(&callback_sem); mutex_unlock(&callback_mutex);
if (cpu_exclusive_changed) if (cpu_exclusive_changed)
update_cpu_domains(cs); update_cpu_domains(cs);
@ -1104,7 +1098,7 @@ static int fmeter_getrate(struct fmeter *fmp)
* writing the path of the old cpuset in 'ppathbuf' if it needs to be * writing the path of the old cpuset in 'ppathbuf' if it needs to be
* notified on release. * notified on release.
* *
* Call holding manage_sem. May take callback_sem and task_lock of * Call holding manage_mutex. May take callback_mutex and task_lock of
* the task 'pid' during call. * the task 'pid' during call.
*/ */
@ -1144,13 +1138,13 @@ static int attach_task(struct cpuset *cs, char *pidbuf, char **ppathbuf)
get_task_struct(tsk); get_task_struct(tsk);
} }
down(&callback_sem); mutex_lock(&callback_mutex);
task_lock(tsk); task_lock(tsk);
oldcs = tsk->cpuset; oldcs = tsk->cpuset;
if (!oldcs) { if (!oldcs) {
task_unlock(tsk); task_unlock(tsk);
up(&callback_sem); mutex_unlock(&callback_mutex);
put_task_struct(tsk); put_task_struct(tsk);
return -ESRCH; return -ESRCH;
} }
@ -1164,7 +1158,7 @@ static int attach_task(struct cpuset *cs, char *pidbuf, char **ppathbuf)
from = oldcs->mems_allowed; from = oldcs->mems_allowed;
to = cs->mems_allowed; to = cs->mems_allowed;
up(&callback_sem); mutex_unlock(&callback_mutex);
mm = get_task_mm(tsk); mm = get_task_mm(tsk);
if (mm) { if (mm) {
@ -1221,7 +1215,7 @@ static ssize_t cpuset_common_file_write(struct file *file, const char __user *us
} }
buffer[nbytes] = 0; /* nul-terminate */ buffer[nbytes] = 0; /* nul-terminate */
down(&manage_sem); mutex_lock(&manage_mutex);
if (is_removed(cs)) { if (is_removed(cs)) {
retval = -ENODEV; retval = -ENODEV;
@ -1264,7 +1258,7 @@ static ssize_t cpuset_common_file_write(struct file *file, const char __user *us
if (retval == 0) if (retval == 0)
retval = nbytes; retval = nbytes;
out2: out2:
up(&manage_sem); mutex_unlock(&manage_mutex);
cpuset_release_agent(pathbuf); cpuset_release_agent(pathbuf);
out1: out1:
kfree(buffer); kfree(buffer);
@ -1304,9 +1298,9 @@ static int cpuset_sprintf_cpulist(char *page, struct cpuset *cs)
{ {
cpumask_t mask; cpumask_t mask;
down(&callback_sem); mutex_lock(&callback_mutex);
mask = cs->cpus_allowed; mask = cs->cpus_allowed;
up(&callback_sem); mutex_unlock(&callback_mutex);
return cpulist_scnprintf(page, PAGE_SIZE, mask); return cpulist_scnprintf(page, PAGE_SIZE, mask);
} }
@ -1315,9 +1309,9 @@ static int cpuset_sprintf_memlist(char *page, struct cpuset *cs)
{ {
nodemask_t mask; nodemask_t mask;
down(&callback_sem); mutex_lock(&callback_mutex);
mask = cs->mems_allowed; mask = cs->mems_allowed;
up(&callback_sem); mutex_unlock(&callback_mutex);
return nodelist_scnprintf(page, PAGE_SIZE, mask); return nodelist_scnprintf(page, PAGE_SIZE, mask);
} }
@ -1598,7 +1592,7 @@ static int pid_array_to_buf(char *buf, int sz, pid_t *a, int npids)
* Handle an open on 'tasks' file. Prepare a buffer listing the * Handle an open on 'tasks' file. Prepare a buffer listing the
* process id's of tasks currently attached to the cpuset being opened. * process id's of tasks currently attached to the cpuset being opened.
* *
* Does not require any specific cpuset semaphores, and does not take any. * Does not require any specific cpuset mutexes, and does not take any.
*/ */
static int cpuset_tasks_open(struct inode *unused, struct file *file) static int cpuset_tasks_open(struct inode *unused, struct file *file)
{ {
@ -1754,7 +1748,7 @@ static int cpuset_populate_dir(struct dentry *cs_dentry)
* name: name of the new cpuset. Will be strcpy'ed. * name: name of the new cpuset. Will be strcpy'ed.
* mode: mode to set on new inode * mode: mode to set on new inode
* *
* Must be called with the semaphore on the parent inode held * Must be called with the mutex on the parent inode held
*/ */
static long cpuset_create(struct cpuset *parent, const char *name, int mode) static long cpuset_create(struct cpuset *parent, const char *name, int mode)
@ -1766,7 +1760,7 @@ static long cpuset_create(struct cpuset *parent, const char *name, int mode)
if (!cs) if (!cs)
return -ENOMEM; return -ENOMEM;
down(&manage_sem); mutex_lock(&manage_mutex);
cpuset_update_task_memory_state(); cpuset_update_task_memory_state();
cs->flags = 0; cs->flags = 0;
if (notify_on_release(parent)) if (notify_on_release(parent))
@ -1782,28 +1776,28 @@ static long cpuset_create(struct cpuset *parent, const char *name, int mode)
cs->parent = parent; cs->parent = parent;
down(&callback_sem); mutex_lock(&callback_mutex);
list_add(&cs->sibling, &cs->parent->children); list_add(&cs->sibling, &cs->parent->children);
number_of_cpusets++; number_of_cpusets++;
up(&callback_sem); mutex_unlock(&callback_mutex);
err = cpuset_create_dir(cs, name, mode); err = cpuset_create_dir(cs, name, mode);
if (err < 0) if (err < 0)
goto err; goto err;
/* /*
* Release manage_sem before cpuset_populate_dir() because it * Release manage_mutex before cpuset_populate_dir() because it
* will down() this new directory's i_mutex and if we race with * will down() this new directory's i_mutex and if we race with
* another mkdir, we might deadlock. * another mkdir, we might deadlock.
*/ */
up(&manage_sem); mutex_unlock(&manage_mutex);
err = cpuset_populate_dir(cs->dentry); err = cpuset_populate_dir(cs->dentry);
/* If err < 0, we have a half-filled directory - oh well ;) */ /* If err < 0, we have a half-filled directory - oh well ;) */
return 0; return 0;
err: err:
list_del(&cs->sibling); list_del(&cs->sibling);
up(&manage_sem); mutex_unlock(&manage_mutex);
kfree(cs); kfree(cs);
return err; return err;
} }
@ -1825,18 +1819,18 @@ static int cpuset_rmdir(struct inode *unused_dir, struct dentry *dentry)
/* the vfs holds both inode->i_mutex already */ /* the vfs holds both inode->i_mutex already */
down(&manage_sem); mutex_lock(&manage_mutex);
cpuset_update_task_memory_state(); cpuset_update_task_memory_state();
if (atomic_read(&cs->count) > 0) { if (atomic_read(&cs->count) > 0) {
up(&manage_sem); mutex_unlock(&manage_mutex);
return -EBUSY; return -EBUSY;
} }
if (!list_empty(&cs->children)) { if (!list_empty(&cs->children)) {
up(&manage_sem); mutex_unlock(&manage_mutex);
return -EBUSY; return -EBUSY;
} }
parent = cs->parent; parent = cs->parent;
down(&callback_sem); mutex_lock(&callback_mutex);
set_bit(CS_REMOVED, &cs->flags); set_bit(CS_REMOVED, &cs->flags);
if (is_cpu_exclusive(cs)) if (is_cpu_exclusive(cs))
update_cpu_domains(cs); update_cpu_domains(cs);
@ -1848,10 +1842,10 @@ static int cpuset_rmdir(struct inode *unused_dir, struct dentry *dentry)
cpuset_d_remove_dir(d); cpuset_d_remove_dir(d);
dput(d); dput(d);
number_of_cpusets--; number_of_cpusets--;
up(&callback_sem); mutex_unlock(&callback_mutex);
if (list_empty(&parent->children)) if (list_empty(&parent->children))
check_for_release(parent, &pathbuf); check_for_release(parent, &pathbuf);
up(&manage_sem); mutex_unlock(&manage_mutex);
cpuset_release_agent(pathbuf); cpuset_release_agent(pathbuf);
return 0; return 0;
} }
@ -1960,19 +1954,19 @@ void cpuset_fork(struct task_struct *child)
* Description: Detach cpuset from @tsk and release it. * Description: Detach cpuset from @tsk and release it.
* *
* Note that cpusets marked notify_on_release force every task in * Note that cpusets marked notify_on_release force every task in
* them to take the global manage_sem semaphore when exiting. * them to take the global manage_mutex mutex when exiting.
* This could impact scaling on very large systems. Be reluctant to * This could impact scaling on very large systems. Be reluctant to
* use notify_on_release cpusets where very high task exit scaling * use notify_on_release cpusets where very high task exit scaling
* is required on large systems. * is required on large systems.
* *
* Don't even think about derefencing 'cs' after the cpuset use count * Don't even think about derefencing 'cs' after the cpuset use count
* goes to zero, except inside a critical section guarded by manage_sem * goes to zero, except inside a critical section guarded by manage_mutex
* or callback_sem. Otherwise a zero cpuset use count is a license to * or callback_mutex. Otherwise a zero cpuset use count is a license to
* any other task to nuke the cpuset immediately, via cpuset_rmdir(). * any other task to nuke the cpuset immediately, via cpuset_rmdir().
* *
* This routine has to take manage_sem, not callback_sem, because * This routine has to take manage_mutex, not callback_mutex, because
* it is holding that semaphore while calling check_for_release(), * it is holding that mutex while calling check_for_release(),
* which calls kmalloc(), so can't be called holding callback__sem(). * which calls kmalloc(), so can't be called holding callback_mutex().
* *
* We don't need to task_lock() this reference to tsk->cpuset, * We don't need to task_lock() this reference to tsk->cpuset,
* because tsk is already marked PF_EXITING, so attach_task() won't * because tsk is already marked PF_EXITING, so attach_task() won't
@ -2022,10 +2016,10 @@ void cpuset_exit(struct task_struct *tsk)
if (notify_on_release(cs)) { if (notify_on_release(cs)) {
char *pathbuf = NULL; char *pathbuf = NULL;
down(&manage_sem); mutex_lock(&manage_mutex);
if (atomic_dec_and_test(&cs->count)) if (atomic_dec_and_test(&cs->count))
check_for_release(cs, &pathbuf); check_for_release(cs, &pathbuf);
up(&manage_sem); mutex_unlock(&manage_mutex);
cpuset_release_agent(pathbuf); cpuset_release_agent(pathbuf);
} else { } else {
atomic_dec(&cs->count); atomic_dec(&cs->count);
@ -2046,11 +2040,11 @@ cpumask_t cpuset_cpus_allowed(struct task_struct *tsk)
{ {
cpumask_t mask; cpumask_t mask;
down(&callback_sem); mutex_lock(&callback_mutex);
task_lock(tsk); task_lock(tsk);
guarantee_online_cpus(tsk->cpuset, &mask); guarantee_online_cpus(tsk->cpuset, &mask);
task_unlock(tsk); task_unlock(tsk);
up(&callback_sem); mutex_unlock(&callback_mutex);
return mask; return mask;
} }
@ -2074,11 +2068,11 @@ nodemask_t cpuset_mems_allowed(struct task_struct *tsk)
{ {
nodemask_t mask; nodemask_t mask;
down(&callback_sem); mutex_lock(&callback_mutex);
task_lock(tsk); task_lock(tsk);
guarantee_online_mems(tsk->cpuset, &mask); guarantee_online_mems(tsk->cpuset, &mask);
task_unlock(tsk); task_unlock(tsk);
up(&callback_sem); mutex_unlock(&callback_mutex);
return mask; return mask;
} }
@ -2104,7 +2098,7 @@ int cpuset_zonelist_valid_mems_allowed(struct zonelist *zl)
/* /*
* nearest_exclusive_ancestor() - Returns the nearest mem_exclusive * nearest_exclusive_ancestor() - Returns the nearest mem_exclusive
* ancestor to the specified cpuset. Call holding callback_sem. * ancestor to the specified cpuset. Call holding callback_mutex.
* If no ancestor is mem_exclusive (an unusual configuration), then * If no ancestor is mem_exclusive (an unusual configuration), then
* returns the root cpuset. * returns the root cpuset.
*/ */
@ -2131,12 +2125,12 @@ static const struct cpuset *nearest_exclusive_ancestor(const struct cpuset *cs)
* GFP_KERNEL allocations are not so marked, so can escape to the * GFP_KERNEL allocations are not so marked, so can escape to the
* nearest mem_exclusive ancestor cpuset. * nearest mem_exclusive ancestor cpuset.
* *
* Scanning up parent cpusets requires callback_sem. The __alloc_pages() * Scanning up parent cpusets requires callback_mutex. The __alloc_pages()
* routine only calls here with __GFP_HARDWALL bit _not_ set if * routine only calls here with __GFP_HARDWALL bit _not_ set if
* it's a GFP_KERNEL allocation, and all nodes in the current tasks * it's a GFP_KERNEL allocation, and all nodes in the current tasks
* mems_allowed came up empty on the first pass over the zonelist. * mems_allowed came up empty on the first pass over the zonelist.
* So only GFP_KERNEL allocations, if all nodes in the cpuset are * So only GFP_KERNEL allocations, if all nodes in the cpuset are
* short of memory, might require taking the callback_sem semaphore. * short of memory, might require taking the callback_mutex mutex.
* *
* The first loop over the zonelist in mm/page_alloc.c:__alloc_pages() * The first loop over the zonelist in mm/page_alloc.c:__alloc_pages()
* calls here with __GFP_HARDWALL always set in gfp_mask, enforcing * calls here with __GFP_HARDWALL always set in gfp_mask, enforcing
@ -2171,31 +2165,31 @@ int __cpuset_zone_allowed(struct zone *z, gfp_t gfp_mask)
return 1; return 1;
/* Not hardwall and node outside mems_allowed: scan up cpusets */ /* Not hardwall and node outside mems_allowed: scan up cpusets */
down(&callback_sem); mutex_lock(&callback_mutex);
task_lock(current); task_lock(current);
cs = nearest_exclusive_ancestor(current->cpuset); cs = nearest_exclusive_ancestor(current->cpuset);
task_unlock(current); task_unlock(current);
allowed = node_isset(node, cs->mems_allowed); allowed = node_isset(node, cs->mems_allowed);
up(&callback_sem); mutex_unlock(&callback_mutex);
return allowed; return allowed;
} }
/** /**
* cpuset_lock - lock out any changes to cpuset structures * cpuset_lock - lock out any changes to cpuset structures
* *
* The out of memory (oom) code needs to lock down cpusets * The out of memory (oom) code needs to mutex_lock cpusets
* from being changed while it scans the tasklist looking for a * from being changed while it scans the tasklist looking for a
* task in an overlapping cpuset. Expose callback_sem via this * task in an overlapping cpuset. Expose callback_mutex via this
* cpuset_lock() routine, so the oom code can lock it, before * cpuset_lock() routine, so the oom code can lock it, before
* locking the task list. The tasklist_lock is a spinlock, so * locking the task list. The tasklist_lock is a spinlock, so
* must be taken inside callback_sem. * must be taken inside callback_mutex.
*/ */
void cpuset_lock(void) void cpuset_lock(void)
{ {
down(&callback_sem); mutex_lock(&callback_mutex);
} }
/** /**
@ -2206,7 +2200,7 @@ void cpuset_lock(void)
void cpuset_unlock(void) void cpuset_unlock(void)
{ {
up(&callback_sem); mutex_unlock(&callback_mutex);
} }
/** /**
@ -2218,7 +2212,7 @@ void cpuset_unlock(void)
* determine if task @p's memory usage might impact the memory * determine if task @p's memory usage might impact the memory
* available to the current task. * available to the current task.
* *
* Call while holding callback_sem. * Call while holding callback_mutex.
**/ **/
int cpuset_excl_nodes_overlap(const struct task_struct *p) int cpuset_excl_nodes_overlap(const struct task_struct *p)
@ -2289,7 +2283,7 @@ void __cpuset_memory_pressure_bump(void)
* - Used for /proc/<pid>/cpuset. * - Used for /proc/<pid>/cpuset.
* - No need to task_lock(tsk) on this tsk->cpuset reference, as it * - No need to task_lock(tsk) on this tsk->cpuset reference, as it
* doesn't really matter if tsk->cpuset changes after we read it, * doesn't really matter if tsk->cpuset changes after we read it,
* and we take manage_sem, keeping attach_task() from changing it * and we take manage_mutex, keeping attach_task() from changing it
* anyway. * anyway.
*/ */
@ -2305,7 +2299,7 @@ static int proc_cpuset_show(struct seq_file *m, void *v)
return -ENOMEM; return -ENOMEM;
tsk = m->private; tsk = m->private;
down(&manage_sem); mutex_lock(&manage_mutex);
cs = tsk->cpuset; cs = tsk->cpuset;
if (!cs) { if (!cs) {
retval = -EINVAL; retval = -EINVAL;
@ -2318,7 +2312,7 @@ static int proc_cpuset_show(struct seq_file *m, void *v)
seq_puts(m, buf); seq_puts(m, buf);
seq_putc(m, '\n'); seq_putc(m, '\n');
out: out:
up(&manage_sem); mutex_unlock(&manage_mutex);
kfree(buf); kfree(buf);
return retval; return retval;
} }