linux/kernel/rcu/srcutiny.c
Paul E. McKenney 1f8da406a9 srcu: Prevent redundant __srcu_read_unlock() wakeup
Tiny SRCU readers can appear at task level, but also in interrupt and
softirq handlers.  Because Tiny SRCU is selected only in kernels built
with CONFIG_SMP=n and CONFIG_PREEMPTION=n, it is not possible for a grace
period to start while there is a non-task-level SRCU reader executing.
This means that it does not make sense for __srcu_read_unlock() to awaken
the Tiny SRCU grace period, because that can only happen when the grace
period is waiting for one value of ->srcu_idx and __srcu_read_unlock()
is ending the last reader for some other value of ->srcu_idx.  After all,
any such wakeup will be redundant.

Worse yet, in some cases, such wakeups generate lockdep splats:

	======================================================
	WARNING: possible circular locking dependency detected
	5.15.0-rc1+ #3758 Not tainted
	------------------------------------------------------
	rcu_torture_rea/53 is trying to acquire lock:
	ffffffff9514e6a8 (srcu_ctl.srcu_wq.lock){..-.}-{2:2}, at:
	xa/0x30

	but task is already holding lock:
	ffff95c642479d80 (&p->pi_lock){-.-.}-{2:2}, at:
	_extend+0x370/0x400

	which lock already depends on the new lock.

	the existing dependency chain (in reverse order) is:

	-> #1 (&p->pi_lock){-.-.}-{2:2}:
	       _raw_spin_lock_irqsave+0x2f/0x50
	       try_to_wake_up+0x50/0x580
	       swake_up_locked.part.7+0xe/0x30
	       swake_up_one+0x22/0x30
	       rcutorture_one_extend+0x1b6/0x400
	       rcu_torture_one_read+0x290/0x5d0
	       rcu_torture_timer+0x1a/0x70
	       call_timer_fn+0xa6/0x230
	       run_timer_softirq+0x493/0x4c0
	       __do_softirq+0xc0/0x371
	       irq_exit+0x73/0x90
	       sysvec_apic_timer_interrupt+0x63/0x80
	       asm_sysvec_apic_timer_interrupt+0x12/0x20
	       default_idle+0xb/0x10
	       default_idle_call+0x5e/0x170
	       do_idle+0x18a/0x1f0
	       cpu_startup_entry+0xa/0x10
	       start_kernel+0x678/0x69f
	       secondary_startup_64_no_verify+0xc2/0xcb

	-> #0 (srcu_ctl.srcu_wq.lock){..-.}-{2:2}:
	       __lock_acquire+0x130c/0x2440
	       lock_acquire+0xc2/0x270
	       _raw_spin_lock_irqsave+0x2f/0x50
	       swake_up_one+0xa/0x30
	       rcutorture_one_extend+0x387/0x400
	       rcu_torture_one_read+0x290/0x5d0
	       rcu_torture_reader+0xac/0x200
	       kthread+0x12d/0x150
	       ret_from_fork+0x22/0x30

	other info that might help us debug this:

	 Possible unsafe locking scenario:

	       CPU0                    CPU1
	       ----                    ----
	  lock(&p->pi_lock);
				       lock(srcu_ctl.srcu_wq.lock);
				       lock(&p->pi_lock);
	  lock(srcu_ctl.srcu_wq.lock);

	 *** DEADLOCK ***

	1 lock held by rcu_torture_rea/53:
	 #0: ffff95c642479d80 (&p->pi_lock){-.-.}-{2:2}, at:
	_extend+0x370/0x400

	stack backtrace:
	CPU: 0 PID: 53 Comm: rcu_torture_rea Not tainted 5.15.0-rc1+

	Hardware name: Red Hat KVM/RHEL-AV, BIOS
	e_el8.5.0+746+bbd5d70c 04/01/2014
	Call Trace:
	 check_noncircular+0xfe/0x110
	 ? find_held_lock+0x2d/0x90
	 __lock_acquire+0x130c/0x2440
	 lock_acquire+0xc2/0x270
	 ? swake_up_one+0xa/0x30
	 ? find_held_lock+0x72/0x90
	 _raw_spin_lock_irqsave+0x2f/0x50
	 ? swake_up_one+0xa/0x30
	 swake_up_one+0xa/0x30
	 rcutorture_one_extend+0x387/0x400
	 rcu_torture_one_read+0x290/0x5d0
	 rcu_torture_reader+0xac/0x200
	 ? rcutorture_oom_notify+0xf0/0xf0
	 ? __kthread_parkme+0x61/0x90
	 ? rcu_torture_one_read+0x5d0/0x5d0
	 kthread+0x12d/0x150
	 ? set_kthread_struct+0x40/0x40
	 ret_from_fork+0x22/0x30

This is a false positive because there is only one CPU, and both locks
are raw (non-preemptible) spinlocks.  However, it is worthwhile getting
rid of the redundant wakeup, which has the side effect of breaking
the theoretical deadlock cycle.  This commit therefore eliminates the
redundant wakeups.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
2021-11-30 17:28:16 -08:00

273 lines
7.5 KiB
C

// SPDX-License-Identifier: GPL-2.0+
/*
* Sleepable Read-Copy Update mechanism for mutual exclusion,
* tiny version for non-preemptible single-CPU use.
*
* Copyright (C) IBM Corporation, 2017
*
* Author: Paul McKenney <paulmck@linux.ibm.com>
*/
#include <linux/export.h>
#include <linux/mutex.h>
#include <linux/preempt.h>
#include <linux/rcupdate_wait.h>
#include <linux/sched.h>
#include <linux/delay.h>
#include <linux/srcu.h>
#include <linux/rcu_node_tree.h>
#include "rcu_segcblist.h"
#include "rcu.h"
int rcu_scheduler_active __read_mostly;
static LIST_HEAD(srcu_boot_list);
static bool srcu_init_done;
static int init_srcu_struct_fields(struct srcu_struct *ssp)
{
ssp->srcu_lock_nesting[0] = 0;
ssp->srcu_lock_nesting[1] = 0;
init_swait_queue_head(&ssp->srcu_wq);
ssp->srcu_cb_head = NULL;
ssp->srcu_cb_tail = &ssp->srcu_cb_head;
ssp->srcu_gp_running = false;
ssp->srcu_gp_waiting = false;
ssp->srcu_idx = 0;
ssp->srcu_idx_max = 0;
INIT_WORK(&ssp->srcu_work, srcu_drive_gp);
INIT_LIST_HEAD(&ssp->srcu_work.entry);
return 0;
}
#ifdef CONFIG_DEBUG_LOCK_ALLOC
int __init_srcu_struct(struct srcu_struct *ssp, const char *name,
struct lock_class_key *key)
{
/* Don't re-initialize a lock while it is held. */
debug_check_no_locks_freed((void *)ssp, sizeof(*ssp));
lockdep_init_map(&ssp->dep_map, name, key, 0);
return init_srcu_struct_fields(ssp);
}
EXPORT_SYMBOL_GPL(__init_srcu_struct);
#else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
/*
* init_srcu_struct - initialize a sleep-RCU structure
* @ssp: structure to initialize.
*
* Must invoke this on a given srcu_struct before passing that srcu_struct
* to any other function. Each srcu_struct represents a separate domain
* of SRCU protection.
*/
int init_srcu_struct(struct srcu_struct *ssp)
{
return init_srcu_struct_fields(ssp);
}
EXPORT_SYMBOL_GPL(init_srcu_struct);
#endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
/*
* cleanup_srcu_struct - deconstruct a sleep-RCU structure
* @ssp: structure to clean up.
*
* Must invoke this after you are finished using a given srcu_struct that
* was initialized via init_srcu_struct(), else you leak memory.
*/
void cleanup_srcu_struct(struct srcu_struct *ssp)
{
WARN_ON(ssp->srcu_lock_nesting[0] || ssp->srcu_lock_nesting[1]);
flush_work(&ssp->srcu_work);
WARN_ON(ssp->srcu_gp_running);
WARN_ON(ssp->srcu_gp_waiting);
WARN_ON(ssp->srcu_cb_head);
WARN_ON(&ssp->srcu_cb_head != ssp->srcu_cb_tail);
WARN_ON(ssp->srcu_idx != ssp->srcu_idx_max);
WARN_ON(ssp->srcu_idx & 0x1);
}
EXPORT_SYMBOL_GPL(cleanup_srcu_struct);
/*
* Removes the count for the old reader from the appropriate element of
* the srcu_struct.
*/
void __srcu_read_unlock(struct srcu_struct *ssp, int idx)
{
int newval = READ_ONCE(ssp->srcu_lock_nesting[idx]) - 1;
WRITE_ONCE(ssp->srcu_lock_nesting[idx], newval);
if (!newval && READ_ONCE(ssp->srcu_gp_waiting) && in_task())
swake_up_one(&ssp->srcu_wq);
}
EXPORT_SYMBOL_GPL(__srcu_read_unlock);
/*
* Workqueue handler to drive one grace period and invoke any callbacks
* that become ready as a result. Single-CPU and !PREEMPTION operation
* means that we get away with murder on synchronization. ;-)
*/
void srcu_drive_gp(struct work_struct *wp)
{
int idx;
struct rcu_head *lh;
struct rcu_head *rhp;
struct srcu_struct *ssp;
ssp = container_of(wp, struct srcu_struct, srcu_work);
if (ssp->srcu_gp_running || USHORT_CMP_GE(ssp->srcu_idx, READ_ONCE(ssp->srcu_idx_max)))
return; /* Already running or nothing to do. */
/* Remove recently arrived callbacks and wait for readers. */
WRITE_ONCE(ssp->srcu_gp_running, true);
local_irq_disable();
lh = ssp->srcu_cb_head;
ssp->srcu_cb_head = NULL;
ssp->srcu_cb_tail = &ssp->srcu_cb_head;
local_irq_enable();
idx = (ssp->srcu_idx & 0x2) / 2;
WRITE_ONCE(ssp->srcu_idx, ssp->srcu_idx + 1);
WRITE_ONCE(ssp->srcu_gp_waiting, true); /* srcu_read_unlock() wakes! */
swait_event_exclusive(ssp->srcu_wq, !READ_ONCE(ssp->srcu_lock_nesting[idx]));
WRITE_ONCE(ssp->srcu_gp_waiting, false); /* srcu_read_unlock() cheap. */
WRITE_ONCE(ssp->srcu_idx, ssp->srcu_idx + 1);
/* Invoke the callbacks we removed above. */
while (lh) {
rhp = lh;
lh = lh->next;
local_bh_disable();
rhp->func(rhp);
local_bh_enable();
}
/*
* Enable rescheduling, and if there are more callbacks,
* reschedule ourselves. This can race with a call_srcu()
* at interrupt level, but the ->srcu_gp_running checks will
* straighten that out.
*/
WRITE_ONCE(ssp->srcu_gp_running, false);
if (USHORT_CMP_LT(ssp->srcu_idx, READ_ONCE(ssp->srcu_idx_max)))
schedule_work(&ssp->srcu_work);
}
EXPORT_SYMBOL_GPL(srcu_drive_gp);
static void srcu_gp_start_if_needed(struct srcu_struct *ssp)
{
unsigned short cookie;
cookie = get_state_synchronize_srcu(ssp);
if (USHORT_CMP_GE(READ_ONCE(ssp->srcu_idx_max), cookie))
return;
WRITE_ONCE(ssp->srcu_idx_max, cookie);
if (!READ_ONCE(ssp->srcu_gp_running)) {
if (likely(srcu_init_done))
schedule_work(&ssp->srcu_work);
else if (list_empty(&ssp->srcu_work.entry))
list_add(&ssp->srcu_work.entry, &srcu_boot_list);
}
}
/*
* Enqueue an SRCU callback on the specified srcu_struct structure,
* initiating grace-period processing if it is not already running.
*/
void call_srcu(struct srcu_struct *ssp, struct rcu_head *rhp,
rcu_callback_t func)
{
unsigned long flags;
rhp->func = func;
rhp->next = NULL;
local_irq_save(flags);
*ssp->srcu_cb_tail = rhp;
ssp->srcu_cb_tail = &rhp->next;
local_irq_restore(flags);
srcu_gp_start_if_needed(ssp);
}
EXPORT_SYMBOL_GPL(call_srcu);
/*
* synchronize_srcu - wait for prior SRCU read-side critical-section completion
*/
void synchronize_srcu(struct srcu_struct *ssp)
{
struct rcu_synchronize rs;
init_rcu_head_on_stack(&rs.head);
init_completion(&rs.completion);
call_srcu(ssp, &rs.head, wakeme_after_rcu);
wait_for_completion(&rs.completion);
destroy_rcu_head_on_stack(&rs.head);
}
EXPORT_SYMBOL_GPL(synchronize_srcu);
/*
* get_state_synchronize_srcu - Provide an end-of-grace-period cookie
*/
unsigned long get_state_synchronize_srcu(struct srcu_struct *ssp)
{
unsigned long ret;
barrier();
ret = (READ_ONCE(ssp->srcu_idx) + 3) & ~0x1;
barrier();
return ret & USHRT_MAX;
}
EXPORT_SYMBOL_GPL(get_state_synchronize_srcu);
/*
* start_poll_synchronize_srcu - Provide cookie and start grace period
*
* The difference between this and get_state_synchronize_srcu() is that
* this function ensures that the poll_state_synchronize_srcu() will
* eventually return the value true.
*/
unsigned long start_poll_synchronize_srcu(struct srcu_struct *ssp)
{
unsigned long ret = get_state_synchronize_srcu(ssp);
srcu_gp_start_if_needed(ssp);
return ret;
}
EXPORT_SYMBOL_GPL(start_poll_synchronize_srcu);
/*
* poll_state_synchronize_srcu - Has cookie's grace period ended?
*/
bool poll_state_synchronize_srcu(struct srcu_struct *ssp, unsigned long cookie)
{
bool ret = USHORT_CMP_GE(READ_ONCE(ssp->srcu_idx), cookie);
barrier();
return ret;
}
EXPORT_SYMBOL_GPL(poll_state_synchronize_srcu);
/* Lockdep diagnostics. */
void __init rcu_scheduler_starting(void)
{
rcu_scheduler_active = RCU_SCHEDULER_RUNNING;
}
/*
* Queue work for srcu_struct structures with early boot callbacks.
* The work won't actually execute until the workqueue initialization
* phase that takes place after the scheduler starts.
*/
void __init srcu_init(void)
{
struct srcu_struct *ssp;
srcu_init_done = true;
while (!list_empty(&srcu_boot_list)) {
ssp = list_first_entry(&srcu_boot_list,
struct srcu_struct, srcu_work.entry);
list_del_init(&ssp->srcu_work.entry);
schedule_work(&ssp->srcu_work);
}
}