mirror_zfs/module/zfs/zthr.c
Mark Johnston 6e2a59181e
Avoid memory allocations in the ARC eviction thread
When the eviction thread goes to shrink an ARC state, it allocates a set
of marker buffers used to hold its place in the state's sublists.

This can be problematic in low memory conditions, since
1) the allocation can be substantial, as we allocate NCPU markers;
2) on at least FreeBSD, page reclamation can block in
   arc_wait_for_eviction()

In particular, in stress tests it's possible to hit a deadlock on
FreeBSD when the number of free pages is very low, wherein the system is
waiting for the page daemon to reclaim memory, the page daemon is
waiting for the ARC eviction thread to finish, and the ARC eviction
thread is blocked waiting for more memory.

Try to reduce the likelihood of such deadlocks by pre-allocating markers
for the eviction thread at ARC initialization time.  When evicting
buffers from an ARC state, check to see if the current thread is the ARC
eviction thread, and use the pre-allocated markers for that purpose
rather than dynamically allocating them.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes #12985
2022-01-21 10:28:13 -08:00

548 lines
17 KiB
C

/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017, 2020 by Delphix. All rights reserved.
*/
/*
* ZTHR Infrastructure
* ===================
*
* ZTHR threads are used for isolated operations that span multiple txgs
* within a SPA. They generally exist from SPA creation/loading and until
* the SPA is exported/destroyed. The ideal requirements for an operation
* to be modeled with a zthr are the following:
*
* 1] The operation needs to run over multiple txgs.
* 2] There is be a single point of reference in memory or on disk that
* indicates whether the operation should run/is running or has
* stopped.
*
* If the operation satisfies the above then the following rules guarantee
* a certain level of correctness:
*
* 1] Any thread EXCEPT the zthr changes the work indicator from stopped
* to running but not the opposite.
* 2] Only the zthr can change the work indicator from running to stopped
* (e.g. when it is done) but not the opposite.
*
* This way a normal zthr cycle should go like this:
*
* 1] An external thread changes the work indicator from stopped to
* running and wakes up the zthr.
* 2] The zthr wakes up, checks the indicator and starts working.
* 3] When the zthr is done, it changes the indicator to stopped, allowing
* a new cycle to start.
*
* Besides being awakened by other threads, a zthr can be configured
* during creation to wakeup on its own after a specified interval
* [see zthr_create_timer()].
*
* Note: ZTHR threads are NOT a replacement for generic threads! Please
* ensure that they fit your use-case well before using them.
*
* == ZTHR creation
*
* Every zthr needs four inputs to start running:
*
* 1] A user-defined checker function (checkfunc) that decides whether
* the zthr should start working or go to sleep. The function should
* return TRUE when the zthr needs to work or FALSE to let it sleep,
* and should adhere to the following signature:
* boolean_t checkfunc_name(void *args, zthr_t *t);
*
* 2] A user-defined ZTHR function (func) which the zthr executes when
* it is not sleeping. The function should adhere to the following
* signature type:
* void func_name(void *args, zthr_t *t);
*
* 3] A void args pointer that will be passed to checkfunc and func
* implicitly by the infrastructure.
*
* 4] A name for the thread. This string must be valid for the lifetime
* of the zthr.
*
* The reason why the above API needs two different functions,
* instead of one that both checks and does the work, has to do with
* the zthr's internal state lock (zthr_state_lock) and the allowed
* cancellation windows. We want to hold the zthr_state_lock while
* running checkfunc but not while running func. This way the zthr
* can be cancelled while doing work and not while checking for work.
*
* To start a zthr:
* zthr_t *zthr_pointer = zthr_create(checkfunc, func, args,
* pri);
* or
* zthr_t *zthr_pointer = zthr_create_timer(checkfunc, func,
* args, max_sleep, pri);
*
* After that you should be able to wakeup, cancel, and resume the
* zthr from another thread using the zthr_pointer.
*
* NOTE: ZTHR threads could potentially wake up spuriously and the
* user should take this into account when writing a checkfunc.
* [see ZTHR state transitions]
*
* == ZTHR wakeup
*
* ZTHR wakeup should be used when new work is added for the zthr. The
* sleeping zthr will wakeup, see that it has more work to complete
* and proceed. This can be invoked from open or syncing context.
*
* To wakeup a zthr:
* zthr_wakeup(zthr_t *t)
*
* == ZTHR cancellation and resumption
*
* ZTHR threads must be cancelled when their SPA is being exported
* or when they need to be paused so they don't interfere with other
* operations.
*
* To cancel a zthr:
* zthr_cancel(zthr_pointer);
*
* To resume it:
* zthr_resume(zthr_pointer);
*
* ZTHR cancel and resume should be invoked in open context during the
* lifecycle of the pool as it is imported, exported or destroyed.
*
* A zthr will implicitly check if it has received a cancellation
* signal every time func returns and every time it wakes up [see
* ZTHR state transitions below].
*
* At times, waiting for the zthr's func to finish its job may take
* time. This may be very time-consuming for some operations that
* need to cancel the SPA's zthrs (e.g spa_export). For this scenario
* the user can explicitly make their ZTHR function aware of incoming
* cancellation signals using zthr_iscancelled(). A common pattern for
* that looks like this:
*
* int
* func_name(void *args, zthr_t *t)
* {
* ... <unpack args> ...
* while (!work_done && !zthr_iscancelled(t)) {
* ... <do more work> ...
* }
* }
*
* == ZTHR cleanup
*
* Cancelling a zthr doesn't clean up its metadata (internal locks,
* function pointers to func and checkfunc, etc..). This is because
* we want to keep them around in case we want to resume the execution
* of the zthr later. Similarly for zthrs that exit themselves.
*
* To completely cleanup a zthr, cancel it first to ensure that it
* is not running and then use zthr_destroy().
*
* == ZTHR state transitions
*
* zthr creation
* +
* |
* | woke up
* | +--------------+ sleep
* | | ^
* | | |
* | | | FALSE
* | | |
* v v FALSE +
* cancelled? +---------> checkfunc?
* + ^ +
* | | |
* | | | TRUE
* | | |
* | | func returned v
* | +---------------+ func
* |
* | TRUE
* |
* v
* zthr stopped running
*
* == Implementation of ZTHR requests
*
* ZTHR cancel and resume are requests on a zthr to change its
* internal state. These requests are serialized using the
* zthr_request_lock, while changes in its internal state are
* protected by the zthr_state_lock. A request will first acquire
* the zthr_request_lock and then immediately acquire the
* zthr_state_lock. We do this so that incoming requests are
* serialized using the request lock, while still allowing us
* to use the state lock for thread communication via zthr_cv.
*
* ZTHR wakeup broadcasts to zthr_cv, causing sleeping threads
* to wakeup. It acquires the zthr_state_lock but not the
* zthr_request_lock, so that a wakeup on a zthr in the middle
* of being cancelled will not block.
*/
#include <sys/zfs_context.h>
#include <sys/zthr.h>
struct zthr {
/* running thread doing the work */
kthread_t *zthr_thread;
/* lock protecting internal data & invariants */
kmutex_t zthr_state_lock;
/* mutex that serializes external requests */
kmutex_t zthr_request_lock;
/* notification mechanism for requests */
kcondvar_t zthr_cv;
/* flag set to true if we are canceling the zthr */
boolean_t zthr_cancel;
/* flag set to true if we are waiting for the zthr to finish */
boolean_t zthr_haswaiters;
kcondvar_t zthr_wait_cv;
/*
* maximum amount of time that the zthr is spent sleeping;
* if this is 0, the thread doesn't wake up until it gets
* signaled.
*/
hrtime_t zthr_sleep_timeout;
/* Thread priority */
pri_t zthr_pri;
/* consumer-provided callbacks & data */
zthr_checkfunc_t *zthr_checkfunc;
zthr_func_t *zthr_func;
void *zthr_arg;
const char *zthr_name;
};
static void
zthr_procedure(void *arg)
{
zthr_t *t = arg;
mutex_enter(&t->zthr_state_lock);
ASSERT3P(t->zthr_thread, ==, curthread);
while (!t->zthr_cancel) {
if (t->zthr_checkfunc(t->zthr_arg, t)) {
mutex_exit(&t->zthr_state_lock);
t->zthr_func(t->zthr_arg, t);
mutex_enter(&t->zthr_state_lock);
} else {
if (t->zthr_sleep_timeout == 0) {
cv_wait_idle(&t->zthr_cv, &t->zthr_state_lock);
} else {
(void) cv_timedwait_idle_hires(&t->zthr_cv,
&t->zthr_state_lock, t->zthr_sleep_timeout,
MSEC2NSEC(1), 0);
}
}
if (t->zthr_haswaiters) {
t->zthr_haswaiters = B_FALSE;
cv_broadcast(&t->zthr_wait_cv);
}
}
/*
* Clear out the kernel thread metadata and notify the
* zthr_cancel() thread that we've stopped running.
*/
t->zthr_thread = NULL;
t->zthr_cancel = B_FALSE;
cv_broadcast(&t->zthr_cv);
mutex_exit(&t->zthr_state_lock);
thread_exit();
}
zthr_t *
zthr_create(const char *zthr_name, zthr_checkfunc_t *checkfunc,
zthr_func_t *func, void *arg, pri_t pri)
{
return (zthr_create_timer(zthr_name, checkfunc,
func, arg, (hrtime_t)0, pri));
}
/*
* Create a zthr with specified maximum sleep time. If the time
* in sleeping state exceeds max_sleep, a wakeup(do the check and
* start working if required) will be triggered.
*/
zthr_t *
zthr_create_timer(const char *zthr_name, zthr_checkfunc_t *checkfunc,
zthr_func_t *func, void *arg, hrtime_t max_sleep, pri_t pri)
{
zthr_t *t = kmem_zalloc(sizeof (*t), KM_SLEEP);
mutex_init(&t->zthr_state_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&t->zthr_request_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&t->zthr_cv, NULL, CV_DEFAULT, NULL);
cv_init(&t->zthr_wait_cv, NULL, CV_DEFAULT, NULL);
mutex_enter(&t->zthr_state_lock);
t->zthr_checkfunc = checkfunc;
t->zthr_func = func;
t->zthr_arg = arg;
t->zthr_sleep_timeout = max_sleep;
t->zthr_name = zthr_name;
t->zthr_pri = pri;
t->zthr_thread = thread_create_named(zthr_name, NULL, 0,
zthr_procedure, t, 0, &p0, TS_RUN, pri);
mutex_exit(&t->zthr_state_lock);
return (t);
}
void
zthr_destroy(zthr_t *t)
{
ASSERT(!MUTEX_HELD(&t->zthr_state_lock));
ASSERT(!MUTEX_HELD(&t->zthr_request_lock));
VERIFY3P(t->zthr_thread, ==, NULL);
mutex_destroy(&t->zthr_request_lock);
mutex_destroy(&t->zthr_state_lock);
cv_destroy(&t->zthr_cv);
cv_destroy(&t->zthr_wait_cv);
kmem_free(t, sizeof (*t));
}
/*
* Wake up the zthr if it is sleeping. If the thread has been cancelled
* or is in the process of being cancelled, this is a no-op.
*/
void
zthr_wakeup(zthr_t *t)
{
mutex_enter(&t->zthr_state_lock);
/*
* There are 5 states that we can find the zthr when issuing
* this broadcast:
*
* [1] The common case of the thread being asleep, at which
* point the broadcast will wake it up.
* [2] The thread has been cancelled. Waking up a cancelled
* thread is a no-op. Any work that is still left to be
* done should be handled the next time the thread is
* resumed.
* [3] The thread is doing work and is already up, so this
* is basically a no-op.
* [4] The thread was just created/resumed, in which case the
* behavior is similar to [3].
* [5] The thread is in the middle of being cancelled, which
* will be a no-op.
*/
cv_broadcast(&t->zthr_cv);
mutex_exit(&t->zthr_state_lock);
}
/*
* Sends a cancel request to the zthr and blocks until the zthr is
* cancelled. If the zthr is not running (e.g. has been cancelled
* already), this is a no-op. Note that this function should not be
* called from syncing context as it could deadlock with the zthr_func.
*/
void
zthr_cancel(zthr_t *t)
{
mutex_enter(&t->zthr_request_lock);
mutex_enter(&t->zthr_state_lock);
/*
* Since we are holding the zthr_state_lock at this point
* we can find the state in one of the following 4 states:
*
* [1] The thread has already been cancelled, therefore
* there is nothing for us to do.
* [2] The thread is sleeping so we set the flag, broadcast
* the CV and wait for it to exit.
* [3] The thread is doing work, in which case we just set
* the flag and wait for it to finish.
* [4] The thread was just created/resumed, in which case
* the behavior is similar to [3].
*
* Since requests are serialized, by the time that we get
* control back we expect that the zthr is cancelled and
* not running anymore.
*/
if (t->zthr_thread != NULL) {
t->zthr_cancel = B_TRUE;
/* broadcast in case the zthr is sleeping */
cv_broadcast(&t->zthr_cv);
while (t->zthr_thread != NULL)
cv_wait(&t->zthr_cv, &t->zthr_state_lock);
ASSERT(!t->zthr_cancel);
}
mutex_exit(&t->zthr_state_lock);
mutex_exit(&t->zthr_request_lock);
}
/*
* Sends a resume request to the supplied zthr. If the zthr is already
* running this is a no-op. Note that this function should not be
* called from syncing context as it could deadlock with the zthr_func.
*/
void
zthr_resume(zthr_t *t)
{
mutex_enter(&t->zthr_request_lock);
mutex_enter(&t->zthr_state_lock);
ASSERT3P(&t->zthr_checkfunc, !=, NULL);
ASSERT3P(&t->zthr_func, !=, NULL);
ASSERT(!t->zthr_cancel);
ASSERT(!t->zthr_haswaiters);
/*
* There are 4 states that we find the zthr in at this point
* given the locks that we hold:
*
* [1] The zthr was cancelled, so we spawn a new thread for
* the zthr (common case).
* [2] The zthr is running at which point this is a no-op.
* [3] The zthr is sleeping at which point this is a no-op.
* [4] The zthr was just spawned at which point this is a
* no-op.
*/
if (t->zthr_thread == NULL) {
t->zthr_thread = thread_create_named(t->zthr_name, NULL, 0,
zthr_procedure, t, 0, &p0, TS_RUN, t->zthr_pri);
}
mutex_exit(&t->zthr_state_lock);
mutex_exit(&t->zthr_request_lock);
}
/*
* This function is intended to be used by the zthr itself
* (specifically the zthr_func callback provided) to check
* if another thread has signaled it to stop running before
* doing some expensive operation.
*
* returns TRUE if we are in the middle of trying to cancel
* this thread.
*
* returns FALSE otherwise.
*/
boolean_t
zthr_iscancelled(zthr_t *t)
{
ASSERT3P(t->zthr_thread, ==, curthread);
/*
* The majority of the functions here grab zthr_request_lock
* first and then zthr_state_lock. This function only grabs
* the zthr_state_lock. That is because this function should
* only be called from the zthr_func to check if someone has
* issued a zthr_cancel() on the thread. If there is a zthr_cancel()
* happening concurrently, attempting to grab the request lock
* here would result in a deadlock.
*
* By grabbing only the zthr_state_lock this function is allowed
* to run concurrently with a zthr_cancel() request.
*/
mutex_enter(&t->zthr_state_lock);
boolean_t cancelled = t->zthr_cancel;
mutex_exit(&t->zthr_state_lock);
return (cancelled);
}
boolean_t
zthr_iscurthread(zthr_t *t)
{
return (t->zthr_thread == curthread);
}
/*
* Wait for the zthr to finish its current function. Similar to
* zthr_iscancelled, you can use zthr_has_waiters to have the zthr_func end
* early. Unlike zthr_cancel, the thread is not destroyed. If the zthr was
* sleeping or cancelled, return immediately.
*/
void
zthr_wait_cycle_done(zthr_t *t)
{
mutex_enter(&t->zthr_state_lock);
/*
* Since we are holding the zthr_state_lock at this point
* we can find the state in one of the following 5 states:
*
* [1] The thread has already cancelled, therefore
* there is nothing for us to do.
* [2] The thread is sleeping so we set the flag, broadcast
* the CV and wait for it to exit.
* [3] The thread is doing work, in which case we just set
* the flag and wait for it to finish.
* [4] The thread was just created/resumed, in which case
* the behavior is similar to [3].
* [5] The thread is the middle of being cancelled, which is
* similar to [3]. We'll wait for the cancel, which is
* waiting for the zthr func.
*
* Since requests are serialized, by the time that we get
* control back we expect that the zthr has completed it's
* zthr_func.
*/
if (t->zthr_thread != NULL) {
t->zthr_haswaiters = B_TRUE;
/* broadcast in case the zthr is sleeping */
cv_broadcast(&t->zthr_cv);
while ((t->zthr_haswaiters) && (t->zthr_thread != NULL))
cv_wait(&t->zthr_wait_cv, &t->zthr_state_lock);
ASSERT(!t->zthr_haswaiters);
}
mutex_exit(&t->zthr_state_lock);
}
/*
* This function is intended to be used by the zthr itself
* to check if another thread is waiting on it to finish
*
* returns TRUE if we have been asked to finish.
*
* returns FALSE otherwise.
*/
boolean_t
zthr_has_waiters(zthr_t *t)
{
ASSERT3P(t->zthr_thread, ==, curthread);
mutex_enter(&t->zthr_state_lock);
/*
* Similarly to zthr_iscancelled(), we only grab the
* zthr_state_lock so that the zthr itself can use this
* to check for the request.
*/
boolean_t has_waiters = t->zthr_haswaiters;
mutex_exit(&t->zthr_state_lock);
return (has_waiters);
}