| Message ID | 20230807110936.21819-46-zhengqi.arch@bytedance.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | use refcount+RCU method to implement lockless slab shrink | expand |
On Mon, Aug 07, 2023 at 07:09:33PM +0800, Qi Zheng wrote: > The shrinker_rwsem is a global read-write lock in shrinkers subsystem, > which protects most operations such as slab shrink, registration and > unregistration of shrinkers, etc. This can easily cause problems in the > following cases. .... > This commit uses the refcount+RCU method [5] proposed by Dave Chinner > to re-implement the lockless global slab shrink. The memcg slab shrink is > handled in the subsequent patch. .... > --- > include/linux/shrinker.h | 17 ++++++++++ > mm/shrinker.c | 70 +++++++++++++++++++++++++++++----------- > 2 files changed, 68 insertions(+), 19 deletions(-) There's no documentation in the code explaining how the lockless shrinker algorithm works. It's left to the reader to work out how this all goes together.... > diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h > index eb342994675a..f06225f18531 100644 > --- a/include/linux/shrinker.h > +++ b/include/linux/shrinker.h > @@ -4,6 +4,8 @@ > > #include <linux/atomic.h> > #include <linux/types.h> > +#include <linux/refcount.h> > +#include <linux/completion.h> > > #define SHRINKER_UNIT_BITS BITS_PER_LONG > > @@ -87,6 +89,10 @@ struct shrinker { > int seeks; /* seeks to recreate an obj */ > unsigned flags; > > + refcount_t refcount; > + struct completion done; > + struct rcu_head rcu; What does the refcount protect, why do we need the completion, etc? > + > void *private_data; > > /* These are for internal use */ > @@ -120,6 +126,17 @@ struct shrinker *shrinker_alloc(unsigned int flags, const char *fmt, ...); > void shrinker_register(struct shrinker *shrinker); > void shrinker_free(struct shrinker *shrinker); > > +static inline bool shrinker_try_get(struct shrinker *shrinker) > +{ > + return refcount_inc_not_zero(&shrinker->refcount); > +} > + > +static inline void shrinker_put(struct shrinker *shrinker) > +{ > + if (refcount_dec_and_test(&shrinker->refcount)) > + complete(&shrinker->done); > +} > + > #ifdef CONFIG_SHRINKER_DEBUG > extern int __printf(2, 3) shrinker_debugfs_rename(struct shrinker *shrinker, > const char *fmt, ...); > diff --git a/mm/shrinker.c b/mm/shrinker.c > index 1911c06b8af5..d318f5621862 100644 > --- a/mm/shrinker.c > +++ b/mm/shrinker.c > @@ -2,6 +2,7 @@ > #include <linux/memcontrol.h> > #include <linux/rwsem.h> > #include <linux/shrinker.h> > +#include <linux/rculist.h> > #include <trace/events/vmscan.h> > > #include "internal.h" > @@ -577,33 +578,42 @@ unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, > if (!mem_cgroup_disabled() && !mem_cgroup_is_root(memcg)) > return shrink_slab_memcg(gfp_mask, nid, memcg, priority); > > - if (!down_read_trylock(&shrinker_rwsem)) > - goto out; > - > - list_for_each_entry(shrinker, &shrinker_list, list) { > + rcu_read_lock(); > + list_for_each_entry_rcu(shrinker, &shrinker_list, list) { > struct shrink_control sc = { > .gfp_mask = gfp_mask, > .nid = nid, > .memcg = memcg, > }; > > + if (!shrinker_try_get(shrinker)) > + continue; > + > + /* > + * We can safely unlock the RCU lock here since we already > + * hold the refcount of the shrinker. > + */ > + rcu_read_unlock(); > + > ret = do_shrink_slab(&sc, shrinker, priority); > if (ret == SHRINK_EMPTY) > ret = 0; > freed += ret; > + > /* > - * Bail out if someone want to register a new shrinker to > - * prevent the registration from being stalled for long periods > - * by parallel ongoing shrinking. > + * This shrinker may be deleted from shrinker_list and freed > + * after the shrinker_put() below, but this shrinker is still > + * used for the next traversal. So it is necessary to hold the > + * RCU lock first to prevent this shrinker from being freed, > + * which also ensures that the next shrinker that is traversed > + * will not be freed (even if it is deleted from shrinker_list > + * at the same time). > */ This comment really should be at the head of the function, describing the algorithm used within the function itself. i.e. how reference counts are used w.r.t. the rcu_read_lock() usage to guarantee existence of the shrinker and the validity of the list walk. I'm not going to remember all these little details when I look at this code in another 6 months time, and having to work it out from first principles every time I look at the code will waste of a lot of time... -Dave.
On Mon, Aug 07, 2023 at 07:09:33PM +0800, Qi Zheng wrote: > diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h > index eb342994675a..f06225f18531 100644 > --- a/include/linux/shrinker.h > +++ b/include/linux/shrinker.h > @@ -4,6 +4,8 @@ > > #include <linux/atomic.h> > #include <linux/types.h> > +#include <linux/refcount.h> > +#include <linux/completion.h> > > #define SHRINKER_UNIT_BITS BITS_PER_LONG > > @@ -87,6 +89,10 @@ struct shrinker { > int seeks; /* seeks to recreate an obj */ > unsigned flags; > > + refcount_t refcount; > + struct completion done; > + struct rcu_head rcu; Documentation, please. What does the refcount protect, what does the completion provide, etc. > + > void *private_data; > > /* These are for internal use */ > @@ -120,6 +126,17 @@ struct shrinker *shrinker_alloc(unsigned int flags, const char *fmt, ...); > void shrinker_register(struct shrinker *shrinker); > void shrinker_free(struct shrinker *shrinker); > > +static inline bool shrinker_try_get(struct shrinker *shrinker) > +{ > + return refcount_inc_not_zero(&shrinker->refcount); > +} > + > +static inline void shrinker_put(struct shrinker *shrinker) > +{ > + if (refcount_dec_and_test(&shrinker->refcount)) > + complete(&shrinker->done); > +} > + > #ifdef CONFIG_SHRINKER_DEBUG > extern int __printf(2, 3) shrinker_debugfs_rename(struct shrinker *shrinker, > const char *fmt, ...); > diff --git a/mm/shrinker.c b/mm/shrinker.c > index 1911c06b8af5..d318f5621862 100644 > --- a/mm/shrinker.c > +++ b/mm/shrinker.c > @@ -2,6 +2,7 @@ > #include <linux/memcontrol.h> > #include <linux/rwsem.h> > #include <linux/shrinker.h> > +#include <linux/rculist.h> > #include <trace/events/vmscan.h> > > #include "internal.h" > @@ -577,33 +578,42 @@ unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, > if (!mem_cgroup_disabled() && !mem_cgroup_is_root(memcg)) > return shrink_slab_memcg(gfp_mask, nid, memcg, priority); > > - if (!down_read_trylock(&shrinker_rwsem)) > - goto out; > - > - list_for_each_entry(shrinker, &shrinker_list, list) { > + rcu_read_lock(); > + list_for_each_entry_rcu(shrinker, &shrinker_list, list) { > struct shrink_control sc = { > .gfp_mask = gfp_mask, > .nid = nid, > .memcg = memcg, > }; > > + if (!shrinker_try_get(shrinker)) > + continue; > + > + /* > + * We can safely unlock the RCU lock here since we already > + * hold the refcount of the shrinker. > + */ > + rcu_read_unlock(); > + > ret = do_shrink_slab(&sc, shrinker, priority); > if (ret == SHRINK_EMPTY) > ret = 0; > freed += ret; > + > /* > - * Bail out if someone want to register a new shrinker to > - * prevent the registration from being stalled for long periods > - * by parallel ongoing shrinking. > + * This shrinker may be deleted from shrinker_list and freed > + * after the shrinker_put() below, but this shrinker is still > + * used for the next traversal. So it is necessary to hold the > + * RCU lock first to prevent this shrinker from being freed, > + * which also ensures that the next shrinker that is traversed > + * will not be freed (even if it is deleted from shrinker_list > + * at the same time). > */ This needs to be moved to the head of the function, and document the whole list walk, get, put and completion parts of the algorithm that make it safe. There's more to this than "we hold a reference count", especially the tricky "we might see the shrinker before it is fully initialised" case.... ..... > void shrinker_free(struct shrinker *shrinker) > { > struct dentry *debugfs_entry = NULL; > @@ -686,9 +712,18 @@ void shrinker_free(struct shrinker *shrinker) > if (!shrinker) > return; > > + if (shrinker->flags & SHRINKER_REGISTERED) { > + shrinker_put(shrinker); > + wait_for_completion(&shrinker->done); > + } Needs a comment explaining why we need to wait here... > + > down_write(&shrinker_rwsem); > if (shrinker->flags & SHRINKER_REGISTERED) { > - list_del(&shrinker->list); > + /* > + * Lookups on the shrinker are over and will fail in the future, > + * so we can now remove it from the lists and free it. > + */ .... rather than here after the wait has been done and provided the guarantee that no shrinker is running or will run again... -Dave.
Hi Dave, On 2023/8/8 10:24, Dave Chinner wrote: > On Mon, Aug 07, 2023 at 07:09:33PM +0800, Qi Zheng wrote: >> diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h >> index eb342994675a..f06225f18531 100644 >> --- a/include/linux/shrinker.h >> +++ b/include/linux/shrinker.h >> @@ -4,6 +4,8 @@ >> >> #include <linux/atomic.h> >> #include <linux/types.h> >> +#include <linux/refcount.h> >> +#include <linux/completion.h> >> >> #define SHRINKER_UNIT_BITS BITS_PER_LONG >> >> @@ -87,6 +89,10 @@ struct shrinker { >> int seeks; /* seeks to recreate an obj */ >> unsigned flags; >> >> + refcount_t refcount; >> + struct completion done; >> + struct rcu_head rcu; > > Documentation, please. What does the refcount protect, what does the > completion provide, etc. How about the following: /* * reference count of this shrinker, holding this can guarantee * that the shrinker will not be released. */ refcount_t refcount; /* * Wait for shrinker::refcount to reach 0, that is, no shrinker * is running or will run again. */ struct completion done; > >> + >> void *private_data; >> >> /* These are for internal use */ >> @@ -120,6 +126,17 @@ struct shrinker *shrinker_alloc(unsigned int flags, const char *fmt, ...); >> void shrinker_register(struct shrinker *shrinker); >> void shrinker_free(struct shrinker *shrinker); >> >> +static inline bool shrinker_try_get(struct shrinker *shrinker) >> +{ >> + return refcount_inc_not_zero(&shrinker->refcount); >> +} >> + >> +static inline void shrinker_put(struct shrinker *shrinker) >> +{ >> + if (refcount_dec_and_test(&shrinker->refcount)) >> + complete(&shrinker->done); >> +} >> + >> #ifdef CONFIG_SHRINKER_DEBUG >> extern int __printf(2, 3) shrinker_debugfs_rename(struct shrinker *shrinker, >> const char *fmt, ...); >> diff --git a/mm/shrinker.c b/mm/shrinker.c >> index 1911c06b8af5..d318f5621862 100644 >> --- a/mm/shrinker.c >> +++ b/mm/shrinker.c >> @@ -2,6 +2,7 @@ >> #include <linux/memcontrol.h> >> #include <linux/rwsem.h> >> #include <linux/shrinker.h> >> +#include <linux/rculist.h> >> #include <trace/events/vmscan.h> >> >> #include "internal.h" >> @@ -577,33 +578,42 @@ unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, >> if (!mem_cgroup_disabled() && !mem_cgroup_is_root(memcg)) >> return shrink_slab_memcg(gfp_mask, nid, memcg, priority); >> >> - if (!down_read_trylock(&shrinker_rwsem)) >> - goto out; >> - >> - list_for_each_entry(shrinker, &shrinker_list, list) { >> + rcu_read_lock(); >> + list_for_each_entry_rcu(shrinker, &shrinker_list, list) { >> struct shrink_control sc = { >> .gfp_mask = gfp_mask, >> .nid = nid, >> .memcg = memcg, >> }; >> >> + if (!shrinker_try_get(shrinker)) >> + continue; >> + >> + /* >> + * We can safely unlock the RCU lock here since we already >> + * hold the refcount of the shrinker. >> + */ >> + rcu_read_unlock(); >> + >> ret = do_shrink_slab(&sc, shrinker, priority); >> if (ret == SHRINK_EMPTY) >> ret = 0; >> freed += ret; >> + >> /* >> - * Bail out if someone want to register a new shrinker to >> - * prevent the registration from being stalled for long periods >> - * by parallel ongoing shrinking. >> + * This shrinker may be deleted from shrinker_list and freed >> + * after the shrinker_put() below, but this shrinker is still >> + * used for the next traversal. So it is necessary to hold the >> + * RCU lock first to prevent this shrinker from being freed, >> + * which also ensures that the next shrinker that is traversed >> + * will not be freed (even if it is deleted from shrinker_list >> + * at the same time). >> */ > > This needs to be moved to the head of the function, and document > the whole list walk, get, put and completion parts of the algorithm > that make it safe. There's more to this than "we hold a reference > count", especially the tricky "we might see the shrinker before it > is fully initialised" case.... How about moving these documents to before list_for_each_entry_rcu(), and then go to the head of shrink_slab_memcg() to explain the memcg slab shrink case. > > > ..... >> void shrinker_free(struct shrinker *shrinker) >> { >> struct dentry *debugfs_entry = NULL; >> @@ -686,9 +712,18 @@ void shrinker_free(struct shrinker *shrinker) >> if (!shrinker) >> return; >> >> + if (shrinker->flags & SHRINKER_REGISTERED) { >> + shrinker_put(shrinker); >> + wait_for_completion(&shrinker->done); >> + } > > Needs a comment explaining why we need to wait here... /* * Wait for all lookups of the shrinker to complete, after that, no * shrinker is running or will run again, then we can safely free * the structure where the shrinker is located, such as super_block * etc. */ >> + >> down_write(&shrinker_rwsem); >> if (shrinker->flags & SHRINKER_REGISTERED) { >> - list_del(&shrinker->list); >> + /* >> + * Lookups on the shrinker are over and will fail in the future, >> + * so we can now remove it from the lists and free it. >> + */ > > .... rather than here after the wait has been done and provided the > guarantee that no shrinker is running or will run again... With the above comment, how about simplifying the comment here to the following: /* * Now we can safely remove it from the shrinker_list and free it. */ Thanks, Qi > > -Dave.
diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h index eb342994675a..f06225f18531 100644 --- a/include/linux/shrinker.h +++ b/include/linux/shrinker.h @@ -4,6 +4,8 @@ #include <linux/atomic.h> #include <linux/types.h> +#include <linux/refcount.h> +#include <linux/completion.h> #define SHRINKER_UNIT_BITS BITS_PER_LONG @@ -87,6 +89,10 @@ struct shrinker { int seeks; /* seeks to recreate an obj */ unsigned flags; + refcount_t refcount; + struct completion done; + struct rcu_head rcu; + void *private_data; /* These are for internal use */ @@ -120,6 +126,17 @@ struct shrinker *shrinker_alloc(unsigned int flags, const char *fmt, ...); void shrinker_register(struct shrinker *shrinker); void shrinker_free(struct shrinker *shrinker); +static inline bool shrinker_try_get(struct shrinker *shrinker) +{ + return refcount_inc_not_zero(&shrinker->refcount); +} + +static inline void shrinker_put(struct shrinker *shrinker) +{ + if (refcount_dec_and_test(&shrinker->refcount)) + complete(&shrinker->done); +} + #ifdef CONFIG_SHRINKER_DEBUG extern int __printf(2, 3) shrinker_debugfs_rename(struct shrinker *shrinker, const char *fmt, ...); diff --git a/mm/shrinker.c b/mm/shrinker.c index 1911c06b8af5..d318f5621862 100644 --- a/mm/shrinker.c +++ b/mm/shrinker.c @@ -2,6 +2,7 @@ #include <linux/memcontrol.h> #include <linux/rwsem.h> #include <linux/shrinker.h> +#include <linux/rculist.h> #include <trace/events/vmscan.h> #include "internal.h" @@ -577,33 +578,42 @@ unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, if (!mem_cgroup_disabled() && !mem_cgroup_is_root(memcg)) return shrink_slab_memcg(gfp_mask, nid, memcg, priority); - if (!down_read_trylock(&shrinker_rwsem)) - goto out; - - list_for_each_entry(shrinker, &shrinker_list, list) { + rcu_read_lock(); + list_for_each_entry_rcu(shrinker, &shrinker_list, list) { struct shrink_control sc = { .gfp_mask = gfp_mask, .nid = nid, .memcg = memcg, }; + if (!shrinker_try_get(shrinker)) + continue; + + /* + * We can safely unlock the RCU lock here since we already + * hold the refcount of the shrinker. + */ + rcu_read_unlock(); + ret = do_shrink_slab(&sc, shrinker, priority); if (ret == SHRINK_EMPTY) ret = 0; freed += ret; + /* - * Bail out if someone want to register a new shrinker to - * prevent the registration from being stalled for long periods - * by parallel ongoing shrinking. + * This shrinker may be deleted from shrinker_list and freed + * after the shrinker_put() below, but this shrinker is still + * used for the next traversal. So it is necessary to hold the + * RCU lock first to prevent this shrinker from being freed, + * which also ensures that the next shrinker that is traversed + * will not be freed (even if it is deleted from shrinker_list + * at the same time). */ - if (rwsem_is_contended(&shrinker_rwsem)) { - freed = freed ? : 1; - break; - } + rcu_read_lock(); + shrinker_put(shrinker); } - up_read(&shrinker_rwsem); -out: + rcu_read_unlock(); cond_resched(); return freed; } @@ -671,13 +681,29 @@ void shrinker_register(struct shrinker *shrinker) } down_write(&shrinker_rwsem); - list_add_tail(&shrinker->list, &shrinker_list); + list_add_tail_rcu(&shrinker->list, &shrinker_list); shrinker->flags |= SHRINKER_REGISTERED; shrinker_debugfs_add(shrinker); up_write(&shrinker_rwsem); + + init_completion(&shrinker->done); + /* + * Now the shrinker is fully set up, take the first reference to it to + * indicate that lookup operations are now allowed to use it via + * shrinker_try_get(). + */ + refcount_set(&shrinker->refcount, 1); } EXPORT_SYMBOL_GPL(shrinker_register); +static void shrinker_free_rcu_cb(struct rcu_head *head) +{ + struct shrinker *shrinker = container_of(head, struct shrinker, rcu); + + kfree(shrinker->nr_deferred); + kfree(shrinker); +} + void shrinker_free(struct shrinker *shrinker) { struct dentry *debugfs_entry = NULL; @@ -686,9 +712,18 @@ void shrinker_free(struct shrinker *shrinker) if (!shrinker) return; + if (shrinker->flags & SHRINKER_REGISTERED) { + shrinker_put(shrinker); + wait_for_completion(&shrinker->done); + } + down_write(&shrinker_rwsem); if (shrinker->flags & SHRINKER_REGISTERED) { - list_del(&shrinker->list); + /* + * Lookups on the shrinker are over and will fail in the future, + * so we can now remove it from the lists and free it. + */ + list_del_rcu(&shrinker->list); debugfs_entry = shrinker_debugfs_detach(shrinker, &debugfs_id); shrinker->flags &= ~SHRINKER_REGISTERED; } else { @@ -702,9 +737,6 @@ void shrinker_free(struct shrinker *shrinker) if (debugfs_entry) shrinker_debugfs_remove(debugfs_entry, debugfs_id); - kfree(shrinker->nr_deferred); - shrinker->nr_deferred = NULL; - - kfree(shrinker); + call_rcu(&shrinker->rcu, shrinker_free_rcu_cb); } EXPORT_SYMBOL_GPL(shrinker_free);
The shrinker_rwsem is a global read-write lock in shrinkers subsystem, which protects most operations such as slab shrink, registration and unregistration of shrinkers, etc. This can easily cause problems in the following cases. 1) When the memory pressure is high and there are many filesystems mounted or unmounted at the same time, slab shrink will be affected (down_read_trylock() failed). Such as the real workload mentioned by Kirill Tkhai: ``` One of the real workloads from my experience is start of an overcommitted node containing many starting containers after node crash (or many resuming containers after reboot for kernel update). In these cases memory pressure is huge, and the node goes round in long reclaim. ``` 2) If a shrinker is blocked (such as the case mentioned in [1]) and a writer comes in (such as mount a fs), then this writer will be blocked and cause all subsequent shrinker-related operations to be blocked. Even if there is no competitor when shrinking slab, there may still be a problem. The down_read_trylock() may become a perf hotspot with frequent calls to shrink_slab(). Because of the poor multicore scalability of atomic operations, this can lead to a significant drop in IPC (instructions per cycle). We used to implement the lockless slab shrink with SRCU [2], but then kernel test robot reported -88.8% regression in stress-ng.ramfs.ops_per_sec test case [3], so we reverted it [4]. This commit uses the refcount+RCU method [5] proposed by Dave Chinner to re-implement the lockless global slab shrink. The memcg slab shrink is handled in the subsequent patch. For now, all shrinker instances are converted to dynamically allocated and will be freed by call_rcu(). So we can use rcu_read_{lock,unlock}() to ensure that the shrinker instance is valid. And the shrinker instance will not be run again after unregistration. So the structure that records the pointer of shrinker instance can be safely freed without waiting for the RCU read-side critical section. In this way, while we implement the lockless slab shrink, we don't need to be blocked in unregister_shrinker(). The following are the test results: stress-ng --timeout 60 --times --verify --metrics-brief --ramfs 9 & 1) Before applying this patchset: setting to a 60 second run per stressor dispatching hogs: 9 ramfs stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s (secs) (secs) (secs) (real time) (usr+sys time) ramfs 735238 60.00 12.37 363.70 12253.05 1955.08 for a 60.01s run time: 1440.27s available CPU time 12.36s user time ( 0.86%) 363.70s system time ( 25.25%) 376.06s total time ( 26.11%) load average: 10.79 4.47 1.69 passed: 9: ramfs (9) failed: 0 skipped: 0 successful run completed in 60.01s (1 min, 0.01 secs) 2) After applying this patchset: setting to a 60 second run per stressor dispatching hogs: 9 ramfs stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s (secs) (secs) (secs) (real time) (usr+sys time) ramfs 746698 60.00 12.45 376.16 12444.02 1921.47 for a 60.01s run time: 1440.28s available CPU time 12.44s user time ( 0.86%) 376.16s system time ( 26.12%) 388.60s total time ( 26.98%) load average: 9.01 3.85 1.49 passed: 9: ramfs (9) failed: 0 skipped: 0 successful run completed in 60.01s (1 min, 0.01 secs) We can see that the ops/s has hardly changed. [1]. https://lore.kernel.org/lkml/20191129214541.3110-1-ptikhomirov@virtuozzo.com/ [2]. https://lore.kernel.org/lkml/20230313112819.38938-1-zhengqi.arch@bytedance.com/ [3]. https://lore.kernel.org/lkml/202305230837.db2c233f-yujie.liu@intel.com/ [4]. https://lore.kernel.org/all/20230609081518.3039120-1-qi.zheng@linux.dev/ [5]. https://lore.kernel.org/lkml/ZIJhou1d55d4H1s0@dread.disaster.area/ Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com> --- include/linux/shrinker.h | 17 ++++++++++ mm/shrinker.c | 70 +++++++++++++++++++++++++++++----------- 2 files changed, 68 insertions(+), 19 deletions(-)