aio_free_ring(ctx);
free_percpu(ctx->cpu);
+ percpu_ref_exit(&ctx->reqs);
+ percpu_ref_exit(&ctx->users);
kmem_cache_free(kioctx_cachep, ctx);
}
err:
mutex_unlock(&ctx->ring_lock);
free_percpu(ctx->cpu);
- free_percpu(ctx->reqs.pcpu_count);
- free_percpu(ctx->users.pcpu_count);
+ percpu_ref_exit(&ctx->reqs);
+ percpu_ref_exit(&ctx->users);
kmem_cache_free(kioctx_cachep, ctx);
pr_debug("error allocating ioctx %d\n", err);
return ERR_PTR(err);
static void put_reqs_available(struct kioctx *ctx, unsigned nr)
{
struct kioctx_cpu *kcpu;
+ unsigned long flags;
preempt_disable();
kcpu = this_cpu_ptr(ctx->cpu);
+ local_irq_save(flags);
kcpu->reqs_available += nr;
+
while (kcpu->reqs_available >= ctx->req_batch * 2) {
kcpu->reqs_available -= ctx->req_batch;
atomic_add(ctx->req_batch, &ctx->reqs_available);
}
+ local_irq_restore(flags);
preempt_enable();
}
{
struct kioctx_cpu *kcpu;
bool ret = false;
+ unsigned long flags;
preempt_disable();
kcpu = this_cpu_ptr(ctx->cpu);
+ local_irq_save(flags);
if (!kcpu->reqs_available) {
int old, avail = atomic_read(&ctx->reqs_available);
ret = true;
kcpu->reqs_available--;
out:
+ local_irq_restore(flags);
preempt_enable();
return ret;
}
/* everything turned out well, dispose of the aiocb. */
kiocb_free(iocb);
+ put_reqs_available(ctx, 1);
/*
* We have to order our ring_info tail store above and test
if (head == tail)
goto out;
+ head %= ctx->nr_events;
+ tail %= ctx->nr_events;
+
while (ret < nr) {
long avail;
struct io_event *ev;
flush_dcache_page(ctx->ring_pages[0]);
pr_debug("%li h%u t%u\n", ret, head, tail);
-
- put_reqs_available(ctx, ret);
out:
mutex_unlock(&ctx->ring_lock);
+ /*
+ * linux/percpu-defs.h - basic definitions for percpu areas
+ *
+ * DO NOT INCLUDE DIRECTLY OUTSIDE PERCPU IMPLEMENTATION PROPER.
+ *
+ * This file is separate from linux/percpu.h to avoid cyclic inclusion
+ * dependency from arch header files. Only to be included from
+ * asm/percpu.h.
+ *
+ * This file includes macros necessary to declare percpu sections and
+ * variables, and definitions of percpu accessors and operations. It
+ * should provide enough percpu features to arch header files even when
+ * they can only include asm/percpu.h to avoid cyclic inclusion dependency.
+ */
+
#ifndef _LINUX_PERCPU_DEFS_H
#define _LINUX_PERCPU_DEFS_H
+ #ifdef CONFIG_SMP
+
+ #ifdef MODULE
+ #define PER_CPU_SHARED_ALIGNED_SECTION ""
+ #define PER_CPU_ALIGNED_SECTION ""
+ #else
+ #define PER_CPU_SHARED_ALIGNED_SECTION "..shared_aligned"
+ #define PER_CPU_ALIGNED_SECTION "..shared_aligned"
+ #endif
+ #define PER_CPU_FIRST_SECTION "..first"
+
+ #else
+
+ #define PER_CPU_SHARED_ALIGNED_SECTION ""
+ #define PER_CPU_ALIGNED_SECTION "..shared_aligned"
+ #define PER_CPU_FIRST_SECTION ""
+
+ #endif
+
/*
* Base implementations of per-CPU variable declarations and definitions, where
* the section in which the variable is to be placed is provided by the
#define __PCPU_DUMMY_ATTRS \
__attribute__((section(".discard"), unused))
- /*
- * Macro which verifies @ptr is a percpu pointer without evaluating
- * @ptr. This is to be used in percpu accessors to verify that the
- * input parameter is a percpu pointer.
- *
- * + 0 is required in order to convert the pointer type from a
- * potential array type to a pointer to a single item of the array.
- */
- #define __verify_pcpu_ptr(ptr) do { \
- const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
- (void)__vpp_verify; \
- } while (0)
-
/*
* s390 and alpha modules require percpu variables to be defined as
* weak to force the compiler to generate GOT based external
* Declaration/definition used for per-CPU variables that must be read mostly.
*/
#define DECLARE_PER_CPU_READ_MOSTLY(type, name) \
- DECLARE_PER_CPU_SECTION(type, name, "..readmostly")
+ DECLARE_PER_CPU_SECTION(type, name, "..read_mostly")
#define DEFINE_PER_CPU_READ_MOSTLY(type, name) \
- DEFINE_PER_CPU_SECTION(type, name, "..readmostly")
+ DEFINE_PER_CPU_SECTION(type, name, "..read_mostly")
/*
* Intermodule exports for per-CPU variables. sparse forgets about
#define EXPORT_PER_CPU_SYMBOL_GPL(var)
#endif
+ /*
+ * Accessors and operations.
+ */
+ #ifndef __ASSEMBLY__
+
+ /*
+ * __verify_pcpu_ptr() verifies @ptr is a percpu pointer without evaluating
+ * @ptr and is invoked once before a percpu area is accessed by all
+ * accessors and operations. This is performed in the generic part of
+ * percpu and arch overrides don't need to worry about it; however, if an
+ * arch wants to implement an arch-specific percpu accessor or operation,
+ * it may use __verify_pcpu_ptr() to verify the parameters.
+ *
+ * + 0 is required in order to convert the pointer type from a
+ * potential array type to a pointer to a single item of the array.
+ */
+ #define __verify_pcpu_ptr(ptr) \
+ do { \
+ const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
+ (void)__vpp_verify; \
+ } while (0)
+
+ #ifdef CONFIG_SMP
+
+ /*
+ * Add an offset to a pointer but keep the pointer as-is. Use RELOC_HIDE()
+ * to prevent the compiler from making incorrect assumptions about the
+ * pointer value. The weird cast keeps both GCC and sparse happy.
+ */
+ #define SHIFT_PERCPU_PTR(__p, __offset) \
+ RELOC_HIDE((typeof(*(__p)) __kernel __force *)(__p), (__offset))
+
+ #define per_cpu_ptr(ptr, cpu) \
+ ({ \
+ __verify_pcpu_ptr(ptr); \
+ SHIFT_PERCPU_PTR((ptr), per_cpu_offset((cpu))); \
+ })
+
+ #define raw_cpu_ptr(ptr) \
+ ({ \
+ __verify_pcpu_ptr(ptr); \
+ arch_raw_cpu_ptr(ptr); \
+ })
+
+ #ifdef CONFIG_DEBUG_PREEMPT
+ #define this_cpu_ptr(ptr) \
+ ({ \
+ __verify_pcpu_ptr(ptr); \
+ SHIFT_PERCPU_PTR(ptr, my_cpu_offset); \
+ })
+ #else
+ #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr)
+ #endif
+
+ #else /* CONFIG_SMP */
+
+ #define VERIFY_PERCPU_PTR(__p) \
+ ({ \
+ __verify_pcpu_ptr(__p); \
+ (typeof(*(__p)) __kernel __force *)(__p); \
+ })
+
+ #define per_cpu_ptr(ptr, cpu) ({ (void)(cpu); VERIFY_PERCPU_PTR(ptr); })
+ #define raw_cpu_ptr(ptr) per_cpu_ptr(ptr, 0)
+ #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr)
+
+ #endif /* CONFIG_SMP */
+
+ #define per_cpu(var, cpu) (*per_cpu_ptr(&(var), cpu))
+ #define __raw_get_cpu_var(var) (*raw_cpu_ptr(&(var)))
+ #define __get_cpu_var(var) (*this_cpu_ptr(&(var)))
+
+ /* keep until we have removed all uses of __this_cpu_ptr */
+ #define __this_cpu_ptr(ptr) raw_cpu_ptr(ptr)
+
+ /*
+ * Must be an lvalue. Since @var must be a simple identifier,
+ * we force a syntax error here if it isn't.
+ */
+ #define get_cpu_var(var) \
+ (*({ \
+ preempt_disable(); \
+ this_cpu_ptr(&var); \
+ }))
+
+ /*
+ * The weird & is necessary because sparse considers (void)(var) to be
+ * a direct dereference of percpu variable (var).
+ */
+ #define put_cpu_var(var) \
+ do { \
+ (void)&(var); \
+ preempt_enable(); \
+ } while (0)
+
+ #define get_cpu_ptr(var) \
+ ({ \
+ preempt_disable(); \
+ this_cpu_ptr(var); \
+ })
+
+ #define put_cpu_ptr(var) \
+ do { \
+ (void)(var); \
+ preempt_enable(); \
+ } while (0)
+
+ /*
+ * Branching function to split up a function into a set of functions that
+ * are called for different scalar sizes of the objects handled.
+ */
+
+ extern void __bad_size_call_parameter(void);
+
+ #ifdef CONFIG_DEBUG_PREEMPT
+ extern void __this_cpu_preempt_check(const char *op);
+ #else
+ static inline void __this_cpu_preempt_check(const char *op) { }
+ #endif
+
+ #define __pcpu_size_call_return(stem, variable) \
+ ({ \
+ typeof(variable) pscr_ret__; \
+ __verify_pcpu_ptr(&(variable)); \
+ switch(sizeof(variable)) { \
+ case 1: pscr_ret__ = stem##1(variable); break; \
+ case 2: pscr_ret__ = stem##2(variable); break; \
+ case 4: pscr_ret__ = stem##4(variable); break; \
+ case 8: pscr_ret__ = stem##8(variable); break; \
+ default: \
+ __bad_size_call_parameter(); break; \
+ } \
+ pscr_ret__; \
+ })
+
+ #define __pcpu_size_call_return2(stem, variable, ...) \
+ ({ \
+ typeof(variable) pscr2_ret__; \
+ __verify_pcpu_ptr(&(variable)); \
+ switch(sizeof(variable)) { \
+ case 1: pscr2_ret__ = stem##1(variable, __VA_ARGS__); break; \
+ case 2: pscr2_ret__ = stem##2(variable, __VA_ARGS__); break; \
+ case 4: pscr2_ret__ = stem##4(variable, __VA_ARGS__); break; \
+ case 8: pscr2_ret__ = stem##8(variable, __VA_ARGS__); break; \
+ default: \
+ __bad_size_call_parameter(); break; \
+ } \
+ pscr2_ret__; \
+ })
+
+ /*
+ * Special handling for cmpxchg_double. cmpxchg_double is passed two
+ * percpu variables. The first has to be aligned to a double word
+ * boundary and the second has to follow directly thereafter.
+ * We enforce this on all architectures even if they don't support
+ * a double cmpxchg instruction, since it's a cheap requirement, and it
+ * avoids breaking the requirement for architectures with the instruction.
+ */
+ #define __pcpu_double_call_return_bool(stem, pcp1, pcp2, ...) \
+ ({ \
+ bool pdcrb_ret__; \
+ __verify_pcpu_ptr(&(pcp1)); \
+ BUILD_BUG_ON(sizeof(pcp1) != sizeof(pcp2)); \
+ VM_BUG_ON((unsigned long)(&(pcp1)) % (2 * sizeof(pcp1))); \
+ VM_BUG_ON((unsigned long)(&(pcp2)) != \
+ (unsigned long)(&(pcp1)) + sizeof(pcp1)); \
+ switch(sizeof(pcp1)) { \
+ case 1: pdcrb_ret__ = stem##1(pcp1, pcp2, __VA_ARGS__); break; \
+ case 2: pdcrb_ret__ = stem##2(pcp1, pcp2, __VA_ARGS__); break; \
+ case 4: pdcrb_ret__ = stem##4(pcp1, pcp2, __VA_ARGS__); break; \
+ case 8: pdcrb_ret__ = stem##8(pcp1, pcp2, __VA_ARGS__); break; \
+ default: \
+ __bad_size_call_parameter(); break; \
+ } \
+ pdcrb_ret__; \
+ })
+
+ #define __pcpu_size_call(stem, variable, ...) \
+ do { \
+ __verify_pcpu_ptr(&(variable)); \
+ switch(sizeof(variable)) { \
+ case 1: stem##1(variable, __VA_ARGS__);break; \
+ case 2: stem##2(variable, __VA_ARGS__);break; \
+ case 4: stem##4(variable, __VA_ARGS__);break; \
+ case 8: stem##8(variable, __VA_ARGS__);break; \
+ default: \
+ __bad_size_call_parameter();break; \
+ } \
+ } while (0)
+
+ /*
+ * this_cpu operations (C) 2008-2013 Christoph Lameter <cl@linux.com>
+ *
+ * Optimized manipulation for memory allocated through the per cpu
+ * allocator or for addresses of per cpu variables.
+ *
+ * These operation guarantee exclusivity of access for other operations
+ * on the *same* processor. The assumption is that per cpu data is only
+ * accessed by a single processor instance (the current one).
+ *
+ * The arch code can provide optimized implementation by defining macros
+ * for certain scalar sizes. F.e. provide this_cpu_add_2() to provide per
+ * cpu atomic operations for 2 byte sized RMW actions. If arch code does
+ * not provide operations for a scalar size then the fallback in the
+ * generic code will be used.
+ *
+ * cmpxchg_double replaces two adjacent scalars at once. The first two
+ * parameters are per cpu variables which have to be of the same size. A
+ * truth value is returned to indicate success or failure (since a double
+ * register result is difficult to handle). There is very limited hardware
+ * support for these operations, so only certain sizes may work.
+ */
+
+ /*
+ * Operations for contexts where we do not want to do any checks for
+ * preemptions. Unless strictly necessary, always use [__]this_cpu_*()
+ * instead.
+ *
+ * If there is no other protection through preempt disable and/or disabling
+ * interupts then one of these RMW operations can show unexpected behavior
+ * because the execution thread was rescheduled on another processor or an
+ * interrupt occurred and the same percpu variable was modified from the
+ * interrupt context.
+ */
+ #define raw_cpu_read(pcp) __pcpu_size_call_return(raw_cpu_read_, pcp)
+ #define raw_cpu_write(pcp, val) __pcpu_size_call(raw_cpu_write_, pcp, val)
+ #define raw_cpu_add(pcp, val) __pcpu_size_call(raw_cpu_add_, pcp, val)
+ #define raw_cpu_and(pcp, val) __pcpu_size_call(raw_cpu_and_, pcp, val)
+ #define raw_cpu_or(pcp, val) __pcpu_size_call(raw_cpu_or_, pcp, val)
+ #define raw_cpu_add_return(pcp, val) __pcpu_size_call_return2(raw_cpu_add_return_, pcp, val)
+ #define raw_cpu_xchg(pcp, nval) __pcpu_size_call_return2(raw_cpu_xchg_, pcp, nval)
+ #define raw_cpu_cmpxchg(pcp, oval, nval) \
+ __pcpu_size_call_return2(raw_cpu_cmpxchg_, pcp, oval, nval)
+ #define raw_cpu_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) \
+ __pcpu_double_call_return_bool(raw_cpu_cmpxchg_double_, pcp1, pcp2, oval1, oval2, nval1, nval2)
+
+ #define raw_cpu_sub(pcp, val) raw_cpu_add(pcp, -(val))
+ #define raw_cpu_inc(pcp) raw_cpu_add(pcp, 1)
+ #define raw_cpu_dec(pcp) raw_cpu_sub(pcp, 1)
+ #define raw_cpu_sub_return(pcp, val) raw_cpu_add_return(pcp, -(typeof(pcp))(val))
+ #define raw_cpu_inc_return(pcp) raw_cpu_add_return(pcp, 1)
+ #define raw_cpu_dec_return(pcp) raw_cpu_add_return(pcp, -1)
+
+ /*
+ * Operations for contexts that are safe from preemption/interrupts. These
+ * operations verify that preemption is disabled.
+ */
+ #define __this_cpu_read(pcp) \
+ ({ \
+ __this_cpu_preempt_check("read"); \
+ raw_cpu_read(pcp); \
+ })
+
+ #define __this_cpu_write(pcp, val) \
+ ({ \
+ __this_cpu_preempt_check("write"); \
+ raw_cpu_write(pcp, val); \
+ })
+
+ #define __this_cpu_add(pcp, val) \
+ ({ \
+ __this_cpu_preempt_check("add"); \
+ raw_cpu_add(pcp, val); \
+ })
+
+ #define __this_cpu_and(pcp, val) \
+ ({ \
+ __this_cpu_preempt_check("and"); \
+ raw_cpu_and(pcp, val); \
+ })
+
+ #define __this_cpu_or(pcp, val) \
+ ({ \
+ __this_cpu_preempt_check("or"); \
+ raw_cpu_or(pcp, val); \
+ })
+
+ #define __this_cpu_add_return(pcp, val) \
+ ({ \
+ __this_cpu_preempt_check("add_return"); \
+ raw_cpu_add_return(pcp, val); \
+ })
+
+ #define __this_cpu_xchg(pcp, nval) \
+ ({ \
+ __this_cpu_preempt_check("xchg"); \
+ raw_cpu_xchg(pcp, nval); \
+ })
+
+ #define __this_cpu_cmpxchg(pcp, oval, nval) \
+ ({ \
+ __this_cpu_preempt_check("cmpxchg"); \
+ raw_cpu_cmpxchg(pcp, oval, nval); \
+ })
+
+ #define __this_cpu_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) \
+ ({ __this_cpu_preempt_check("cmpxchg_double"); \
+ raw_cpu_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2); \
+ })
+
+ #define __this_cpu_sub(pcp, val) __this_cpu_add(pcp, -(typeof(pcp))(val))
+ #define __this_cpu_inc(pcp) __this_cpu_add(pcp, 1)
+ #define __this_cpu_dec(pcp) __this_cpu_sub(pcp, 1)
+ #define __this_cpu_sub_return(pcp, val) __this_cpu_add_return(pcp, -(typeof(pcp))(val))
+ #define __this_cpu_inc_return(pcp) __this_cpu_add_return(pcp, 1)
+ #define __this_cpu_dec_return(pcp) __this_cpu_add_return(pcp, -1)
+
+ /*
+ * Operations with implied preemption protection. These operations can be
+ * used without worrying about preemption. Note that interrupts may still
+ * occur while an operation is in progress and if the interrupt modifies
+ * the variable too then RMW actions may not be reliable.
+ */
+ #define this_cpu_read(pcp) __pcpu_size_call_return(this_cpu_read_, pcp)
+ #define this_cpu_write(pcp, val) __pcpu_size_call(this_cpu_write_, pcp, val)
+ #define this_cpu_add(pcp, val) __pcpu_size_call(this_cpu_add_, pcp, val)
+ #define this_cpu_and(pcp, val) __pcpu_size_call(this_cpu_and_, pcp, val)
+ #define this_cpu_or(pcp, val) __pcpu_size_call(this_cpu_or_, pcp, val)
+ #define this_cpu_add_return(pcp, val) __pcpu_size_call_return2(this_cpu_add_return_, pcp, val)
+ #define this_cpu_xchg(pcp, nval) __pcpu_size_call_return2(this_cpu_xchg_, pcp, nval)
+ #define this_cpu_cmpxchg(pcp, oval, nval) \
+ __pcpu_size_call_return2(this_cpu_cmpxchg_, pcp, oval, nval)
+ #define this_cpu_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) \
+ __pcpu_double_call_return_bool(this_cpu_cmpxchg_double_, pcp1, pcp2, oval1, oval2, nval1, nval2)
+
+ #define this_cpu_sub(pcp, val) this_cpu_add(pcp, -(typeof(pcp))(val))
+ #define this_cpu_inc(pcp) this_cpu_add(pcp, 1)
+ #define this_cpu_dec(pcp) this_cpu_sub(pcp, 1)
+ #define this_cpu_sub_return(pcp, val) this_cpu_add_return(pcp, -(typeof(pcp))(val))
+ #define this_cpu_inc_return(pcp) this_cpu_add_return(pcp, 1)
+ #define this_cpu_dec_return(pcp) this_cpu_add_return(pcp, -1)
+
+ #endif /* __ASSEMBLY__ */
#endif /* _LINUX_PERCPU_DEFS_H */
exit_root_id:
cgroup_exit_root_id(root);
cancel_ref:
- percpu_ref_cancel_init(&root_cgrp->self.refcnt);
+ percpu_ref_exit(&root_cgrp->self.refcnt);
out:
free_cgrp_cset_links(&tmp_links);
return ret;
int flags, const char *unused_dev_name,
void *data)
{
+ struct super_block *pinned_sb = NULL;
+ struct cgroup_subsys *ss;
struct cgroup_root *root;
struct cgroup_sb_opts opts;
struct dentry *dentry;
int ret;
+ int i;
bool new_sb;
/*
goto out_unlock;
}
+ /*
+ * Destruction of cgroup root is asynchronous, so subsystems may
+ * still be dying after the previous unmount. Let's drain the
+ * dying subsystems. We just need to ensure that the ones
+ * unmounted previously finish dying and don't care about new ones
+ * starting. Testing ref liveliness is good enough.
+ */
+ for_each_subsys(ss, i) {
+ if (!(opts.subsys_mask & (1 << i)) ||
+ ss->root == &cgrp_dfl_root)
+ continue;
+
+ if (!percpu_ref_tryget_live(&ss->root->cgrp.self.refcnt)) {
+ mutex_unlock(&cgroup_mutex);
+ msleep(10);
+ ret = restart_syscall();
+ goto out_free;
+ }
+ cgroup_put(&ss->root->cgrp);
+ }
+
for_each_root(root) {
bool name_match = false;
}
/*
- * A root's lifetime is governed by its root cgroup.
- * tryget_live failure indicate that the root is being
- * destroyed. Wait for destruction to complete so that the
- * subsystems are free. We can use wait_queue for the wait
- * but this path is super cold. Let's just sleep for a bit
- * and retry.
+ * We want to reuse @root whose lifetime is governed by its
+ * ->cgrp. Let's check whether @root is alive and keep it
+ * that way. As cgroup_kill_sb() can happen anytime, we
+ * want to block it by pinning the sb so that @root doesn't
+ * get killed before mount is complete.
+ *
+ * With the sb pinned, tryget_live can reliably indicate
+ * whether @root can be reused. If it's being killed,
+ * drain it. We can use wait_queue for the wait but this
+ * path is super cold. Let's just sleep a bit and retry.
*/
- if (!percpu_ref_tryget_live(&root->cgrp.self.refcnt)) {
+ pinned_sb = kernfs_pin_sb(root->kf_root, NULL);
+ if (IS_ERR(pinned_sb) ||
+ !percpu_ref_tryget_live(&root->cgrp.self.refcnt)) {
mutex_unlock(&cgroup_mutex);
+ if (!IS_ERR_OR_NULL(pinned_sb))
+ deactivate_super(pinned_sb);
msleep(10);
ret = restart_syscall();
goto out_free;
CGROUP_SUPER_MAGIC, &new_sb);
if (IS_ERR(dentry) || !new_sb)
cgroup_put(&root->cgrp);
+
+ /*
+ * If @pinned_sb, we're reusing an existing root and holding an
+ * extra ref on its sb. Mount is complete. Put the extra ref.
+ */
+ if (pinned_sb) {
+ WARN_ON(new_sb);
+ deactivate_super(pinned_sb);
+ }
+
return dentry;
}
rcu_read_lock();
css_for_each_child(child, css) {
- if (css->flags & CSS_ONLINE) {
+ if (child->flags & CSS_ONLINE) {
ret = true;
break;
}
container_of(work, struct cgroup_subsys_state, destroy_work);
struct cgroup *cgrp = css->cgroup;
+ percpu_ref_exit(&css->refcnt);
+
if (css->ss) {
/* css free path */
if (css->parent)
err_free_id:
cgroup_idr_remove(&ss->css_idr, css->id);
err_free_percpu_ref:
- percpu_ref_cancel_init(&css->refcnt);
+ percpu_ref_exit(&css->refcnt);
err_free_css:
call_rcu(&css->rcu_head, css_free_rcu_fn);
return err;
out_free_id:
cgroup_idr_remove(&root->cgroup_idr, cgrp->id);
out_cancel_ref:
- percpu_ref_cancel_init(&cgrp->self.refcnt);
+ percpu_ref_exit(&cgrp->self.refcnt);
out_free_cgrp:
kfree(cgrp);
out_unlock:
static struct kmem_cache *pwq_cache;
-static int wq_numa_tbl_len; /* highest possible NUMA node id + 1 */
static cpumask_var_t *wq_numa_possible_cpumask;
/* possible CPUs of each node */
int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
int nr_busy = pool->nr_workers - nr_idle;
- /*
- * nr_idle and idle_list may disagree if idle rebinding is in
- * progress. Never return %true if idle_list is empty.
- */
- if (list_empty(&pool->idle_list))
- return false;
-
return nr_idle > 2 && (nr_idle - 2) * MAX_IDLE_WORKERS_RATIO >= nr_busy;
}
pool = worker->pool;
/* this can only happen on the local cpu */
- if (WARN_ON_ONCE(cpu != raw_smp_processor_id()))
+ if (WARN_ON_ONCE(cpu != raw_smp_processor_id() || pool->cpu != cpu))
return NULL;
/*
* worker_set_flags - set worker flags and adjust nr_running accordingly
* @worker: self
* @flags: flags to set
- * @wakeup: wakeup an idle worker if necessary
*
- * Set @flags in @worker->flags and adjust nr_running accordingly. If
- * nr_running becomes zero and @wakeup is %true, an idle worker is
- * woken up.
+ * Set @flags in @worker->flags and adjust nr_running accordingly.
*
* CONTEXT:
* spin_lock_irq(pool->lock)
*/
-static inline void worker_set_flags(struct worker *worker, unsigned int flags,
- bool wakeup)
+static inline void worker_set_flags(struct worker *worker, unsigned int flags)
{
struct worker_pool *pool = worker->pool;
WARN_ON_ONCE(worker->task != current);
- /*
- * If transitioning into NOT_RUNNING, adjust nr_running and
- * wake up an idle worker as necessary if requested by
- * @wakeup.
- */
+ /* If transitioning into NOT_RUNNING, adjust nr_running. */
if ((flags & WORKER_NOT_RUNNING) &&
!(worker->flags & WORKER_NOT_RUNNING)) {
- if (wakeup) {
- if (atomic_dec_and_test(&pool->nr_running) &&
- !list_empty(&pool->worklist))
- wake_up_worker(pool);
- } else
- atomic_dec(&pool->nr_running);
+ atomic_dec(&pool->nr_running);
}
worker->flags |= flags;
pwq_activate_delayed_work(work);
list_del_init(&work->entry);
- pwq_dec_nr_in_flight(get_work_pwq(work), get_work_color(work));
+ pwq_dec_nr_in_flight(pwq, get_work_color(work));
/* work->data points to pwq iff queued, point to pool */
set_work_pool_and_keep_pending(work, pool->id);
(worker->hentry.next || worker->hentry.pprev)))
return;
- /* can't use worker_set_flags(), also called from start_worker() */
+ /* can't use worker_set_flags(), also called from create_worker() */
worker->flags |= WORKER_IDLE;
pool->nr_idle++;
worker->last_active = jiffies;
list_del_init(&worker->entry);
}
-static struct worker *alloc_worker(void)
+static struct worker *alloc_worker(int node)
{
struct worker *worker;
- worker = kzalloc(sizeof(*worker), GFP_KERNEL);
+ worker = kzalloc_node(sizeof(*worker), GFP_KERNEL, node);
if (worker) {
INIT_LIST_HEAD(&worker->entry);
INIT_LIST_HEAD(&worker->scheduled);
detach_completion = pool->detach_completion;
mutex_unlock(&pool->attach_mutex);
+ /* clear leftover flags without pool->lock after it is detached */
+ worker->flags &= ~(WORKER_UNBOUND | WORKER_REBOUND);
+
if (detach_completion)
complete(detach_completion);
}
* create_worker - create a new workqueue worker
* @pool: pool the new worker will belong to
*
- * Create a new worker which is attached to @pool. The new worker must be
- * started by start_worker().
+ * Create and start a new worker which is attached to @pool.
*
* CONTEXT:
* Might sleep. Does GFP_KERNEL allocations.
if (id < 0)
goto fail;
- worker = alloc_worker();
+ worker = alloc_worker(pool->node);
if (!worker)
goto fail;
/* successful, attach the worker to the pool */
worker_attach_to_pool(worker, pool);
+ /* start the newly created worker */
+ spin_lock_irq(&pool->lock);
+ worker->pool->nr_workers++;
+ worker_enter_idle(worker);
+ wake_up_process(worker->task);
+ spin_unlock_irq(&pool->lock);
+
return worker;
fail:
return NULL;
}
-/**
- * start_worker - start a newly created worker
- * @worker: worker to start
- *
- * Make the pool aware of @worker and start it.
- *
- * CONTEXT:
- * spin_lock_irq(pool->lock).
- */
-static void start_worker(struct worker *worker)
-{
- worker->pool->nr_workers++;
- worker_enter_idle(worker);
- wake_up_process(worker->task);
-}
-
-/**
- * create_and_start_worker - create and start a worker for a pool
- * @pool: the target pool
- *
- * Grab the managership of @pool and create and start a new worker for it.
- *
- * Return: 0 on success. A negative error code otherwise.
- */
-static int create_and_start_worker(struct worker_pool *pool)
-{
- struct worker *worker;
-
- worker = create_worker(pool);
- if (worker) {
- spin_lock_irq(&pool->lock);
- start_worker(worker);
- spin_unlock_irq(&pool->lock);
- }
-
- return worker ? 0 : -ENOMEM;
-}
-
/**
* destroy_worker - destroy a workqueue worker
* @worker: worker to be destroyed
mod_timer(&pool->mayday_timer, jiffies + MAYDAY_INITIAL_TIMEOUT);
while (true) {
- struct worker *worker;
-
- worker = create_worker(pool);
- if (worker) {
- del_timer_sync(&pool->mayday_timer);
- spin_lock_irq(&pool->lock);
- start_worker(worker);
- if (WARN_ON_ONCE(need_to_create_worker(pool)))
- goto restart;
- return true;
- }
-
- if (!need_to_create_worker(pool))
+ if (create_worker(pool) || !need_to_create_worker(pool))
break;
- __set_current_state(TASK_INTERRUPTIBLE);
- schedule_timeout(CREATE_COOLDOWN);
+ schedule_timeout_interruptible(CREATE_COOLDOWN);
if (!need_to_create_worker(pool))
break;
del_timer_sync(&pool->mayday_timer);
spin_lock_irq(&pool->lock);
+ /*
+ * This is necessary even after a new worker was just successfully
+ * created as @pool->lock was dropped and the new worker might have
+ * already become busy.
+ */
if (need_to_create_worker(pool))
goto restart;
return true;
lockdep_copy_map(&lockdep_map, &work->lockdep_map);
#endif
+ /* ensure we're on the correct CPU */
WARN_ON_ONCE(!(pool->flags & POOL_DISASSOCIATED) &&
raw_smp_processor_id() != pool->cpu);
list_del_init(&work->entry);
/*
- * CPU intensive works don't participate in concurrency
- * management. They're the scheduler's responsibility.
+ * CPU intensive works don't participate in concurrency management.
+ * They're the scheduler's responsibility. This takes @worker out
+ * of concurrency management and the next code block will chain
+ * execution of the pending work items.
*/
if (unlikely(cpu_intensive))
- worker_set_flags(worker, WORKER_CPU_INTENSIVE, true);
+ worker_set_flags(worker, WORKER_CPU_INTENSIVE);
/*
- * Unbound pool isn't concurrency managed and work items should be
- * executed ASAP. Wake up another worker if necessary.
+ * Wake up another worker if necessary. The condition is always
+ * false for normal per-cpu workers since nr_running would always
+ * be >= 1 at this point. This is used to chain execution of the
+ * pending work items for WORKER_NOT_RUNNING workers such as the
+ * UNBOUND and CPU_INTENSIVE ones.
*/
- if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
+ if (need_more_worker(pool))
wake_up_worker(pool);
/*
}
} while (keep_working(pool));
- worker_set_flags(worker, WORKER_PREP, false);
+ worker_set_flags(worker, WORKER_PREP);
sleep:
/*
* pool->lock is held and there's no work to process and no need to
move_linked_works(work, scheduled, &n);
process_scheduled_works(rescuer);
- spin_unlock_irq(&pool->lock);
-
- worker_detach_from_pool(rescuer, pool);
-
- spin_lock_irq(&pool->lock);
/*
* Put the reference grabbed by send_mayday(). @pool won't
- * go away while we're holding its lock.
+ * go away while we're still attached to it.
*/
put_pwq(pwq);
/*
- * Leave this pool. If keep_working() is %true, notify a
+ * Leave this pool. If need_more_worker() is %true, notify a
* regular worker; otherwise, we end up with 0 concurrency
* and stalling the execution.
*/
- if (keep_working(pool))
+ if (need_more_worker(pool))
wake_up_worker(pool);
rescuer->pool = NULL;
- spin_unlock(&pool->lock);
- spin_lock(&wq_mayday_lock);
+ spin_unlock_irq(&pool->lock);
+
+ worker_detach_from_pool(rescuer, pool);
+
+ spin_lock_irq(&wq_mayday_lock);
}
spin_unlock_irq(&wq_mayday_lock);
}
}
+ dev_set_uevent_suppress(&wq_dev->dev, false);
kobject_uevent(&wq_dev->dev.kobj, KOBJ_ADD);
return 0;
}
return;
/* sanity checks */
- if (WARN_ON(!(pool->flags & POOL_DISASSOCIATED)) ||
+ if (WARN_ON(!(pool->cpu < 0)) ||
WARN_ON(!list_empty(&pool->worklist)))
return;
hash_for_each_possible(unbound_pool_hash, pool, hash_node, hash) {
if (wqattrs_equal(pool->attrs, attrs)) {
pool->refcnt++;
- goto out_unlock;
+ return pool;
}
}
goto fail;
/* create and start the initial worker */
- if (create_and_start_worker(pool) < 0)
+ if (!create_worker(pool))
goto fail;
/* install */
hash_add(unbound_pool_hash, &pool->hash_node, hash);
-out_unlock:
+
return pool;
fail:
if (pool)
if (WARN_ON_ONCE(!(wq->flags & WQ_UNBOUND)))
return;
- /*
- * Unlink @pwq. Synchronization against wq->mutex isn't strictly
- * necessary on release but do it anyway. It's easier to verify
- * and consistent with the linking path.
- */
mutex_lock(&wq->mutex);
list_del_rcu(&pwq->pwqs_node);
is_last = list_empty(&wq->pwqs);
if (!list_empty(&pwq->pwqs_node))
return;
- /*
- * Set the matching work_color. This is synchronized with
- * wq->mutex to avoid confusing flush_workqueue().
- */
+ /* set the matching work_color */
pwq->work_color = wq->work_color;
/* sync max_active to the current setting */
if (WARN_ON((wq->flags & __WQ_ORDERED) && !list_empty(&wq->pwqs)))
return -EINVAL;
- pwq_tbl = kzalloc(wq_numa_tbl_len * sizeof(pwq_tbl[0]), GFP_KERNEL);
+ pwq_tbl = kzalloc(nr_node_ids * sizeof(pwq_tbl[0]), GFP_KERNEL);
new_attrs = alloc_workqueue_attrs(GFP_KERNEL);
tmp_attrs = alloc_workqueue_attrs(GFP_KERNEL);
if (!pwq_tbl || !new_attrs || !tmp_attrs)
/* allocate wq and format name */
if (flags & WQ_UNBOUND)
- tbl_size = wq_numa_tbl_len * sizeof(wq->numa_pwq_tbl[0]);
+ tbl_size = nr_node_ids * sizeof(wq->numa_pwq_tbl[0]);
wq = kzalloc(sizeof(*wq) + tbl_size, GFP_KERNEL);
if (!wq)
if (flags & WQ_MEM_RECLAIM) {
struct worker *rescuer;
- rescuer = alloc_worker();
+ rescuer = alloc_worker(NUMA_NO_NODE);
if (!rescuer)
goto err_destroy;
struct worker *worker;
for_each_cpu_worker_pool(pool, cpu) {
- WARN_ON_ONCE(cpu != smp_processor_id());
-
mutex_lock(&pool->attach_mutex);
spin_lock_irq(&pool->lock);
for_each_cpu_worker_pool(pool, cpu) {
if (pool->nr_workers)
continue;
- if (create_and_start_worker(pool) < 0)
+ if (!create_worker(pool))
return NOTIFY_BAD;
}
break;
for_each_pool(pool, pi) {
mutex_lock(&pool->attach_mutex);
- if (pool->cpu == cpu) {
+ if (pool->cpu == cpu)
rebind_workers(pool);
- } else if (pool->cpu < 0) {
+ else if (pool->cpu < 0)
restore_unbound_workers_cpumask(pool, cpu);
- }
mutex_unlock(&pool->attach_mutex);
}
cpumask_var_t *tbl;
int node, cpu;
- /* determine NUMA pwq table len - highest node id + 1 */
- for_each_node(node)
- wq_numa_tbl_len = max(wq_numa_tbl_len, node + 1);
-
if (num_possible_nodes() <= 1)
return;
* available. Build one from cpu_to_node() which should have been
* fully initialized by now.
*/
- tbl = kzalloc(wq_numa_tbl_len * sizeof(tbl[0]), GFP_KERNEL);
+ tbl = kzalloc(nr_node_ids * sizeof(tbl[0]), GFP_KERNEL);
BUG_ON(!tbl);
for_each_node(node)
- BUG_ON(!alloc_cpumask_var_node(&tbl[node], GFP_KERNEL,
+ BUG_ON(!zalloc_cpumask_var_node(&tbl[node], GFP_KERNEL,
node_online(node) ? node : NUMA_NO_NODE));
for_each_possible_cpu(cpu) {
for_each_cpu_worker_pool(pool, cpu) {
pool->flags &= ~POOL_DISASSOCIATED;
- BUG_ON(create_and_start_worker(pool) < 0);
+ BUG_ON(!create_worker(pool));
}
}