linux-sg2042/kernel/locking/test-ww_mutex.c

646 lines
13 KiB
C
Raw Normal View History

/*
* Module-based API test facility for ww_mutexes
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, you can access it online at
* http://www.gnu.org/licenses/gpl-2.0.html.
*/
#include <linux/kernel.h>
#include <linux/completion.h>
#include <linux/delay.h>
#include <linux/kthread.h>
#include <linux/module.h>
#include <linux/random.h>
#include <linux/slab.h>
#include <linux/ww_mutex.h>
locking: Implement an algorithm choice for Wound-Wait mutexes The current Wound-Wait mutex algorithm is actually not Wound-Wait but Wait-Die. Implement also Wound-Wait as a per-ww-class choice. Wound-Wait is, contrary to Wait-Die a preemptive algorithm and is known to generate fewer backoffs. Testing reveals that this is true if the number of simultaneous contending transactions is small. As the number of simultaneous contending threads increases, Wait-Wound becomes inferior to Wait-Die in terms of elapsed time. Possibly due to the larger number of held locks of sleeping transactions. Update documentation and callers. Timings using git://people.freedesktop.org/~thomash/ww_mutex_test tag patch-18-06-15 Each thread runs 100000 batches of lock / unlock 800 ww mutexes randomly chosen out of 100000. Four core Intel x86_64: Algorithm #threads Rollbacks time Wound-Wait 4 ~100 ~17s. Wait-Die 4 ~150000 ~19s. Wound-Wait 16 ~360000 ~109s. Wait-Die 16 ~450000 ~82s. Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Gustavo Padovan <gustavo@padovan.org> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Sean Paul <seanpaul@chromium.org> Cc: David Airlie <airlied@linux.ie> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Kate Stewart <kstewart@linuxfoundation.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: linux-doc@vger.kernel.org Cc: linux-media@vger.kernel.org Cc: linaro-mm-sig@lists.linaro.org Co-authored-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Ingo Molnar <mingo@kernel.org>
2018-06-15 16:17:38 +08:00
static DEFINE_WD_CLASS(ww_class);
struct workqueue_struct *wq;
struct test_mutex {
struct work_struct work;
struct ww_mutex mutex;
struct completion ready, go, done;
unsigned int flags;
};
#define TEST_MTX_SPIN BIT(0)
#define TEST_MTX_TRY BIT(1)
#define TEST_MTX_CTX BIT(2)
#define __TEST_MTX_LAST BIT(3)
static void test_mutex_work(struct work_struct *work)
{
struct test_mutex *mtx = container_of(work, typeof(*mtx), work);
complete(&mtx->ready);
wait_for_completion(&mtx->go);
if (mtx->flags & TEST_MTX_TRY) {
while (!ww_mutex_trylock(&mtx->mutex))
cond_resched();
} else {
ww_mutex_lock(&mtx->mutex, NULL);
}
complete(&mtx->done);
ww_mutex_unlock(&mtx->mutex);
}
static int __test_mutex(unsigned int flags)
{
#define TIMEOUT (HZ / 16)
struct test_mutex mtx;
struct ww_acquire_ctx ctx;
int ret;
ww_mutex_init(&mtx.mutex, &ww_class);
ww_acquire_init(&ctx, &ww_class);
INIT_WORK_ONSTACK(&mtx.work, test_mutex_work);
init_completion(&mtx.ready);
init_completion(&mtx.go);
init_completion(&mtx.done);
mtx.flags = flags;
schedule_work(&mtx.work);
wait_for_completion(&mtx.ready);
ww_mutex_lock(&mtx.mutex, (flags & TEST_MTX_CTX) ? &ctx : NULL);
complete(&mtx.go);
if (flags & TEST_MTX_SPIN) {
unsigned long timeout = jiffies + TIMEOUT;
ret = 0;
do {
if (completion_done(&mtx.done)) {
ret = -EINVAL;
break;
}
cond_resched();
} while (time_before(jiffies, timeout));
} else {
ret = wait_for_completion_timeout(&mtx.done, TIMEOUT);
}
ww_mutex_unlock(&mtx.mutex);
ww_acquire_fini(&ctx);
if (ret) {
pr_err("%s(flags=%x): mutual exclusion failure\n",
__func__, flags);
ret = -EINVAL;
}
flush_work(&mtx.work);
destroy_work_on_stack(&mtx.work);
return ret;
#undef TIMEOUT
}
static int test_mutex(void)
{
int ret;
int i;
for (i = 0; i < __TEST_MTX_LAST; i++) {
ret = __test_mutex(i);
if (ret)
return ret;
}
return 0;
}
static int test_aa(void)
{
struct ww_mutex mutex;
struct ww_acquire_ctx ctx;
int ret;
ww_mutex_init(&mutex, &ww_class);
ww_acquire_init(&ctx, &ww_class);
ww_mutex_lock(&mutex, &ctx);
if (ww_mutex_trylock(&mutex)) {
pr_err("%s: trylocked itself!\n", __func__);
ww_mutex_unlock(&mutex);
ret = -EINVAL;
goto out;
}
ret = ww_mutex_lock(&mutex, &ctx);
if (ret != -EALREADY) {
pr_err("%s: missed deadlock for recursing, ret=%d\n",
__func__, ret);
if (!ret)
ww_mutex_unlock(&mutex);
ret = -EINVAL;
goto out;
}
ret = 0;
out:
ww_mutex_unlock(&mutex);
ww_acquire_fini(&ctx);
return ret;
}
struct test_abba {
struct work_struct work;
struct ww_mutex a_mutex;
struct ww_mutex b_mutex;
struct completion a_ready;
struct completion b_ready;
bool resolve;
int result;
};
static void test_abba_work(struct work_struct *work)
{
struct test_abba *abba = container_of(work, typeof(*abba), work);
struct ww_acquire_ctx ctx;
int err;
ww_acquire_init(&ctx, &ww_class);
ww_mutex_lock(&abba->b_mutex, &ctx);
complete(&abba->b_ready);
wait_for_completion(&abba->a_ready);
err = ww_mutex_lock(&abba->a_mutex, &ctx);
if (abba->resolve && err == -EDEADLK) {
ww_mutex_unlock(&abba->b_mutex);
ww_mutex_lock_slow(&abba->a_mutex, &ctx);
err = ww_mutex_lock(&abba->b_mutex, &ctx);
}
if (!err)
ww_mutex_unlock(&abba->a_mutex);
ww_mutex_unlock(&abba->b_mutex);
ww_acquire_fini(&ctx);
abba->result = err;
}
static int test_abba(bool resolve)
{
struct test_abba abba;
struct ww_acquire_ctx ctx;
int err, ret;
ww_mutex_init(&abba.a_mutex, &ww_class);
ww_mutex_init(&abba.b_mutex, &ww_class);
INIT_WORK_ONSTACK(&abba.work, test_abba_work);
init_completion(&abba.a_ready);
init_completion(&abba.b_ready);
abba.resolve = resolve;
schedule_work(&abba.work);
ww_acquire_init(&ctx, &ww_class);
ww_mutex_lock(&abba.a_mutex, &ctx);
complete(&abba.a_ready);
wait_for_completion(&abba.b_ready);
err = ww_mutex_lock(&abba.b_mutex, &ctx);
if (resolve && err == -EDEADLK) {
ww_mutex_unlock(&abba.a_mutex);
ww_mutex_lock_slow(&abba.b_mutex, &ctx);
err = ww_mutex_lock(&abba.a_mutex, &ctx);
}
if (!err)
ww_mutex_unlock(&abba.b_mutex);
ww_mutex_unlock(&abba.a_mutex);
ww_acquire_fini(&ctx);
flush_work(&abba.work);
destroy_work_on_stack(&abba.work);
ret = 0;
if (resolve) {
if (err || abba.result) {
pr_err("%s: failed to resolve ABBA deadlock, A err=%d, B err=%d\n",
__func__, err, abba.result);
ret = -EINVAL;
}
} else {
if (err != -EDEADLK && abba.result != -EDEADLK) {
pr_err("%s: missed ABBA deadlock, A err=%d, B err=%d\n",
__func__, err, abba.result);
ret = -EINVAL;
}
}
return ret;
}
struct test_cycle {
struct work_struct work;
struct ww_mutex a_mutex;
struct ww_mutex *b_mutex;
struct completion *a_signal;
struct completion b_signal;
int result;
};
static void test_cycle_work(struct work_struct *work)
{
struct test_cycle *cycle = container_of(work, typeof(*cycle), work);
struct ww_acquire_ctx ctx;
int err;
ww_acquire_init(&ctx, &ww_class);
ww_mutex_lock(&cycle->a_mutex, &ctx);
complete(cycle->a_signal);
wait_for_completion(&cycle->b_signal);
err = ww_mutex_lock(cycle->b_mutex, &ctx);
if (err == -EDEADLK) {
ww_mutex_unlock(&cycle->a_mutex);
ww_mutex_lock_slow(cycle->b_mutex, &ctx);
err = ww_mutex_lock(&cycle->a_mutex, &ctx);
}
if (!err)
ww_mutex_unlock(cycle->b_mutex);
ww_mutex_unlock(&cycle->a_mutex);
ww_acquire_fini(&ctx);
cycle->result = err;
}
static int __test_cycle(unsigned int nthreads)
{
struct test_cycle *cycles;
unsigned int n, last = nthreads - 1;
int ret;
cycles = kmalloc_array(nthreads, sizeof(*cycles), GFP_KERNEL);
if (!cycles)
return -ENOMEM;
for (n = 0; n < nthreads; n++) {
struct test_cycle *cycle = &cycles[n];
ww_mutex_init(&cycle->a_mutex, &ww_class);
if (n == last)
cycle->b_mutex = &cycles[0].a_mutex;
else
cycle->b_mutex = &cycles[n + 1].a_mutex;
if (n == 0)
cycle->a_signal = &cycles[last].b_signal;
else
cycle->a_signal = &cycles[n - 1].b_signal;
init_completion(&cycle->b_signal);
INIT_WORK(&cycle->work, test_cycle_work);
cycle->result = 0;
}
for (n = 0; n < nthreads; n++)
queue_work(wq, &cycles[n].work);
flush_workqueue(wq);
ret = 0;
for (n = 0; n < nthreads; n++) {
struct test_cycle *cycle = &cycles[n];
if (!cycle->result)
continue;
pr_err("cyclic deadlock not resolved, ret[%d/%d] = %d\n",
n, nthreads, cycle->result);
ret = -EINVAL;
break;
}
for (n = 0; n < nthreads; n++)
ww_mutex_destroy(&cycles[n].a_mutex);
kfree(cycles);
return ret;
}
static int test_cycle(unsigned int ncpus)
{
unsigned int n;
int ret;
for (n = 2; n <= ncpus + 1; n++) {
ret = __test_cycle(n);
if (ret)
return ret;
}
return 0;
}
struct stress {
struct work_struct work;
struct ww_mutex *locks;
unsigned long timeout;
int nlocks;
};
static int *get_random_order(int count)
{
int *order;
int n, r, tmp;
mm: treewide: remove GFP_TEMPORARY allocation flag GFP_TEMPORARY was introduced by commit e12ba74d8ff3 ("Group short-lived and reclaimable kernel allocations") along with __GFP_RECLAIMABLE. It's primary motivation was to allow users to tell that an allocation is short lived and so the allocator can try to place such allocations close together and prevent long term fragmentation. As much as this sounds like a reasonable semantic it becomes much less clear when to use the highlevel GFP_TEMPORARY allocation flag. How long is temporary? Can the context holding that memory sleep? Can it take locks? It seems there is no good answer for those questions. The current implementation of GFP_TEMPORARY is basically GFP_KERNEL | __GFP_RECLAIMABLE which in itself is tricky because basically none of the existing caller provide a way to reclaim the allocated memory. So this is rather misleading and hard to evaluate for any benefits. I have checked some random users and none of them has added the flag with a specific justification. I suspect most of them just copied from other existing users and others just thought it might be a good idea to use without any measuring. This suggests that GFP_TEMPORARY just motivates for cargo cult usage without any reasoning. I believe that our gfp flags are quite complex already and especially those with highlevel semantic should be clearly defined to prevent from confusion and abuse. Therefore I propose dropping GFP_TEMPORARY and replace all existing users to simply use GFP_KERNEL. Please note that SLAB users with shrinkers will still get __GFP_RECLAIMABLE heuristic and so they will be placed properly for memory fragmentation prevention. I can see reasons we might want some gfp flag to reflect shorterm allocations but I propose starting from a clear semantic definition and only then add users with proper justification. This was been brought up before LSF this year by Matthew [1] and it turned out that GFP_TEMPORARY really doesn't have a clear semantic. It seems to be a heuristic without any measured advantage for most (if not all) its current users. The follow up discussion has revealed that opinions on what might be temporary allocation differ a lot between developers. So rather than trying to tweak existing users into a semantic which they haven't expected I propose to simply remove the flag and start from scratch if we really need a semantic for short term allocations. [1] http://lkml.kernel.org/r/20170118054945.GD18349@bombadil.infradead.org [akpm@linux-foundation.org: fix typo] [akpm@linux-foundation.org: coding-style fixes] [sfr@canb.auug.org.au: drm/i915: fix up] Link: http://lkml.kernel.org/r/20170816144703.378d4f4d@canb.auug.org.au Link: http://lkml.kernel.org/r/20170728091904.14627-1-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Matthew Wilcox <willy@infradead.org> Cc: Neil Brown <neilb@suse.de> Cc: "Theodore Ts'o" <tytso@mit.edu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-14 07:28:29 +08:00
order = kmalloc_array(count, sizeof(*order), GFP_KERNEL);
if (!order)
return order;
for (n = 0; n < count; n++)
order[n] = n;
for (n = count - 1; n > 1; n--) {
r = get_random_int() % (n + 1);
if (r != n) {
tmp = order[n];
order[n] = order[r];
order[r] = tmp;
}
}
return order;
}
static void dummy_load(struct stress *stress)
{
usleep_range(1000, 2000);
}
static void stress_inorder_work(struct work_struct *work)
{
struct stress *stress = container_of(work, typeof(*stress), work);
const int nlocks = stress->nlocks;
struct ww_mutex *locks = stress->locks;
struct ww_acquire_ctx ctx;
int *order;
order = get_random_order(nlocks);
if (!order)
return;
do {
int contended = -1;
int n, err;
ww_acquire_init(&ctx, &ww_class);
retry:
err = 0;
for (n = 0; n < nlocks; n++) {
if (n == contended)
continue;
err = ww_mutex_lock(&locks[order[n]], &ctx);
if (err < 0)
break;
}
if (!err)
dummy_load(stress);
if (contended > n)
ww_mutex_unlock(&locks[order[contended]]);
contended = n;
while (n--)
ww_mutex_unlock(&locks[order[n]]);
if (err == -EDEADLK) {
ww_mutex_lock_slow(&locks[order[contended]], &ctx);
goto retry;
}
if (err) {
pr_err_once("stress (%s) failed with %d\n",
__func__, err);
break;
}
ww_acquire_fini(&ctx);
} while (!time_after(jiffies, stress->timeout));
kfree(order);
kfree(stress);
}
struct reorder_lock {
struct list_head link;
struct ww_mutex *lock;
};
static void stress_reorder_work(struct work_struct *work)
{
struct stress *stress = container_of(work, typeof(*stress), work);
LIST_HEAD(locks);
struct ww_acquire_ctx ctx;
struct reorder_lock *ll, *ln;
int *order;
int n, err;
order = get_random_order(stress->nlocks);
if (!order)
return;
for (n = 0; n < stress->nlocks; n++) {
ll = kmalloc(sizeof(*ll), GFP_KERNEL);
if (!ll)
goto out;
ll->lock = &stress->locks[order[n]];
list_add(&ll->link, &locks);
}
kfree(order);
order = NULL;
do {
ww_acquire_init(&ctx, &ww_class);
list_for_each_entry(ll, &locks, link) {
err = ww_mutex_lock(ll->lock, &ctx);
if (!err)
continue;
ln = ll;
list_for_each_entry_continue_reverse(ln, &locks, link)
ww_mutex_unlock(ln->lock);
if (err != -EDEADLK) {
pr_err_once("stress (%s) failed with %d\n",
__func__, err);
break;
}
ww_mutex_lock_slow(ll->lock, &ctx);
list_move(&ll->link, &locks); /* restarts iteration */
}
dummy_load(stress);
list_for_each_entry(ll, &locks, link)
ww_mutex_unlock(ll->lock);
ww_acquire_fini(&ctx);
} while (!time_after(jiffies, stress->timeout));
out:
list_for_each_entry_safe(ll, ln, &locks, link)
kfree(ll);
kfree(order);
kfree(stress);
}
static void stress_one_work(struct work_struct *work)
{
struct stress *stress = container_of(work, typeof(*stress), work);
const int nlocks = stress->nlocks;
struct ww_mutex *lock = stress->locks + (get_random_int() % nlocks);
int err;
do {
err = ww_mutex_lock(lock, NULL);
if (!err) {
dummy_load(stress);
ww_mutex_unlock(lock);
} else {
pr_err_once("stress (%s) failed with %d\n",
__func__, err);
break;
}
} while (!time_after(jiffies, stress->timeout));
kfree(stress);
}
#define STRESS_INORDER BIT(0)
#define STRESS_REORDER BIT(1)
#define STRESS_ONE BIT(2)
#define STRESS_ALL (STRESS_INORDER | STRESS_REORDER | STRESS_ONE)
static int stress(int nlocks, int nthreads, unsigned int flags)
{
struct ww_mutex *locks;
int n;
locks = kmalloc_array(nlocks, sizeof(*locks), GFP_KERNEL);
if (!locks)
return -ENOMEM;
for (n = 0; n < nlocks; n++)
ww_mutex_init(&locks[n], &ww_class);
for (n = 0; nthreads; n++) {
struct stress *stress;
void (*fn)(struct work_struct *work);
fn = NULL;
switch (n & 3) {
case 0:
if (flags & STRESS_INORDER)
fn = stress_inorder_work;
break;
case 1:
if (flags & STRESS_REORDER)
fn = stress_reorder_work;
break;
case 2:
if (flags & STRESS_ONE)
fn = stress_one_work;
break;
}
if (!fn)
continue;
stress = kmalloc(sizeof(*stress), GFP_KERNEL);
if (!stress)
break;
INIT_WORK(&stress->work, fn);
stress->locks = locks;
stress->nlocks = nlocks;
stress->timeout = jiffies + 2*HZ;
queue_work(wq, &stress->work);
nthreads--;
}
flush_workqueue(wq);
for (n = 0; n < nlocks; n++)
ww_mutex_destroy(&locks[n]);
kfree(locks);
return 0;
}
static int __init test_ww_mutex_init(void)
{
int ncpus = num_online_cpus();
int ret;
wq = alloc_workqueue("test-ww_mutex", WQ_UNBOUND, 0);
if (!wq)
return -ENOMEM;
ret = test_mutex();
if (ret)
return ret;
ret = test_aa();
if (ret)
return ret;
ret = test_abba(false);
if (ret)
return ret;
ret = test_abba(true);
if (ret)
return ret;
ret = test_cycle(ncpus);
if (ret)
return ret;
ret = stress(16, 2*ncpus, STRESS_INORDER);
if (ret)
return ret;
ret = stress(16, 2*ncpus, STRESS_REORDER);
if (ret)
return ret;
ret = stress(4095, hweight32(STRESS_ALL)*ncpus, STRESS_ALL);
if (ret)
return ret;
return 0;
}
static void __exit test_ww_mutex_exit(void)
{
destroy_workqueue(wq);
}
module_init(test_ww_mutex_init);
module_exit(test_ww_mutex_exit);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Intel Corporation");