diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2013-09-12 15:01:38 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2013-09-12 15:01:38 -0700 |
commit | 26935fb06ee88f1188789807687c03041f3c70d9 (patch) | |
tree | 381c487716540b52348d78bee6555f8fa61d77ef /drivers | |
parent | 3cc69b638e11bfda5d013c2b75b60934aa0e88a1 (diff) | |
parent | bf2ba3bc185269eca274b458aac46ba1ad7c1121 (diff) |
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Pull vfs pile 4 from Al Viro:
"list_lru pile, mostly"
This came out of Andrew's pile, Al ended up doing the merge work so that
Andrew didn't have to.
Additionally, a few fixes.
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (42 commits)
super: fix for destroy lrus
list_lru: dynamically adjust node arrays
shrinker: Kill old ->shrink API.
shrinker: convert remaining shrinkers to count/scan API
staging/lustre/libcfs: cleanup linux-mem.h
staging/lustre/ptlrpc: convert to new shrinker API
staging/lustre/obdclass: convert lu_object shrinker to count/scan API
staging/lustre/ldlm: convert to shrinkers to count/scan API
hugepage: convert huge zero page shrinker to new shrinker API
i915: bail out earlier when shrinker cannot acquire mutex
drivers: convert shrinkers to new count/scan API
fs: convert fs shrinkers to new scan/count API
xfs: fix dquot isolation hang
xfs-convert-dquot-cache-lru-to-list_lru-fix
xfs: convert dquot cache lru to list_lru
xfs: rework buffer dispose list tracking
xfs-convert-buftarg-lru-to-generic-code-fix
xfs: convert buftarg LRU to generic code
fs: convert inode and dentry shrinking to be node aware
vmscan: per-node deferred work
...
Diffstat (limited to 'drivers')
-rw-r--r-- | drivers/gpu/drm/i915/i915_dma.c | 4 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/i915_gem.c | 82 | ||||
-rw-r--r-- | drivers/gpu/drm/ttm/ttm_page_alloc.c | 44 | ||||
-rw-r--r-- | drivers/gpu/drm/ttm/ttm_page_alloc_dma.c | 51 | ||||
-rw-r--r-- | drivers/md/bcache/btree.c | 43 | ||||
-rw-r--r-- | drivers/md/bcache/sysfs.c | 2 | ||||
-rw-r--r-- | drivers/md/dm-bufio.c | 64 | ||||
-rw-r--r-- | drivers/staging/android/ashmem.c | 44 | ||||
-rw-r--r-- | drivers/staging/android/lowmemorykiller.c | 43 | ||||
-rw-r--r-- | drivers/staging/lustre/include/linux/libcfs/linux/linux-mem.h | 38 | ||||
-rw-r--r-- | drivers/staging/lustre/lustre/ldlm/ldlm_pool.c | 148 | ||||
-rw-r--r-- | drivers/staging/lustre/lustre/obdclass/lu_object.c | 98 | ||||
-rw-r--r-- | drivers/staging/lustre/lustre/ptlrpc/sec_bulk.c | 76 |
13 files changed, 413 insertions, 324 deletions
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 9b265a4c6a3d..c27a21034a5e 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1676,7 +1676,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) return 0; out_gem_unload: - if (dev_priv->mm.inactive_shrinker.shrink) + if (dev_priv->mm.inactive_shrinker.scan_objects) unregister_shrinker(&dev_priv->mm.inactive_shrinker); if (dev->pdev->msi_enabled) @@ -1715,7 +1715,7 @@ int i915_driver_unload(struct drm_device *dev) i915_teardown_sysfs(dev); - if (dev_priv->mm.inactive_shrinker.shrink) + if (dev_priv->mm.inactive_shrinker.scan_objects) unregister_shrinker(&dev_priv->mm.inactive_shrinker); mutex_lock(&dev->struct_mutex); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index d9e337feef14..8507c6d1e642 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -57,10 +57,12 @@ static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj, struct drm_i915_fence_reg *fence, bool enable); -static int i915_gem_inactive_shrink(struct shrinker *shrinker, - struct shrink_control *sc); +static unsigned long i915_gem_inactive_count(struct shrinker *shrinker, + struct shrink_control *sc); +static unsigned long i915_gem_inactive_scan(struct shrinker *shrinker, + struct shrink_control *sc); static long i915_gem_purge(struct drm_i915_private *dev_priv, long target); -static void i915_gem_shrink_all(struct drm_i915_private *dev_priv); +static long i915_gem_shrink_all(struct drm_i915_private *dev_priv); static void i915_gem_object_truncate(struct drm_i915_gem_object *obj); static bool cpu_cache_is_coherent(struct drm_device *dev, @@ -1769,16 +1771,21 @@ i915_gem_purge(struct drm_i915_private *dev_priv, long target) return __i915_gem_shrink(dev_priv, target, true); } -static void +static long i915_gem_shrink_all(struct drm_i915_private *dev_priv) { struct drm_i915_gem_object *obj, *next; + long freed = 0; i915_gem_evict_everything(dev_priv->dev); list_for_each_entry_safe(obj, next, &dev_priv->mm.unbound_list, - global_list) + global_list) { + if (obj->pages_pin_count == 0) + freed += obj->base.size >> PAGE_SHIFT; i915_gem_object_put_pages(obj); + } + return freed; } static int @@ -4558,7 +4565,8 @@ i915_gem_load(struct drm_device *dev) dev_priv->mm.interruptible = true; - dev_priv->mm.inactive_shrinker.shrink = i915_gem_inactive_shrink; + dev_priv->mm.inactive_shrinker.scan_objects = i915_gem_inactive_scan; + dev_priv->mm.inactive_shrinker.count_objects = i915_gem_inactive_count; dev_priv->mm.inactive_shrinker.seeks = DEFAULT_SEEKS; register_shrinker(&dev_priv->mm.inactive_shrinker); } @@ -4781,8 +4789,8 @@ static bool mutex_is_locked_by(struct mutex *mutex, struct task_struct *task) #endif } -static int -i915_gem_inactive_shrink(struct shrinker *shrinker, struct shrink_control *sc) +static unsigned long +i915_gem_inactive_count(struct shrinker *shrinker, struct shrink_control *sc) { struct drm_i915_private *dev_priv = container_of(shrinker, @@ -4790,45 +4798,35 @@ i915_gem_inactive_shrink(struct shrinker *shrinker, struct shrink_control *sc) mm.inactive_shrinker); struct drm_device *dev = dev_priv->dev; struct drm_i915_gem_object *obj; - int nr_to_scan = sc->nr_to_scan; bool unlock = true; - int cnt; + unsigned long count; if (!mutex_trylock(&dev->struct_mutex)) { if (!mutex_is_locked_by(&dev->struct_mutex, current)) - return 0; + return SHRINK_STOP; if (dev_priv->mm.shrinker_no_lock_stealing) - return 0; + return SHRINK_STOP; unlock = false; } - if (nr_to_scan) { - nr_to_scan -= i915_gem_purge(dev_priv, nr_to_scan); - if (nr_to_scan > 0) - nr_to_scan -= __i915_gem_shrink(dev_priv, nr_to_scan, - false); - if (nr_to_scan > 0) - i915_gem_shrink_all(dev_priv); - } - - cnt = 0; + count = 0; list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) if (obj->pages_pin_count == 0) - cnt += obj->base.size >> PAGE_SHIFT; + count += obj->base.size >> PAGE_SHIFT; list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) { if (obj->active) continue; if (obj->pin_count == 0 && obj->pages_pin_count == 0) - cnt += obj->base.size >> PAGE_SHIFT; + count += obj->base.size >> PAGE_SHIFT; } if (unlock) mutex_unlock(&dev->struct_mutex); - return cnt; + return count; } /* All the new VM stuff */ @@ -4892,6 +4890,40 @@ unsigned long i915_gem_obj_size(struct drm_i915_gem_object *o, return 0; } +static unsigned long +i915_gem_inactive_scan(struct shrinker *shrinker, struct shrink_control *sc) +{ + struct drm_i915_private *dev_priv = + container_of(shrinker, + struct drm_i915_private, + mm.inactive_shrinker); + struct drm_device *dev = dev_priv->dev; + int nr_to_scan = sc->nr_to_scan; + unsigned long freed; + bool unlock = true; + + if (!mutex_trylock(&dev->struct_mutex)) { + if (!mutex_is_locked_by(&dev->struct_mutex, current)) + return 0; + + if (dev_priv->mm.shrinker_no_lock_stealing) + return 0; + + unlock = false; + } + + freed = i915_gem_purge(dev_priv, nr_to_scan); + if (freed < nr_to_scan) + freed += __i915_gem_shrink(dev_priv, nr_to_scan, + false); + if (freed < nr_to_scan) + freed += i915_gem_shrink_all(dev_priv); + + if (unlock) + mutex_unlock(&dev->struct_mutex); + return freed; +} + struct i915_vma *i915_gem_obj_to_vma(struct drm_i915_gem_object *obj, struct i915_address_space *vm) { diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index bd2a3b40cd12..863bef9f9234 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -377,28 +377,26 @@ out: return nr_free; } -/* Get good estimation how many pages are free in pools */ -static int ttm_pool_get_num_unused_pages(void) -{ - unsigned i; - int total = 0; - for (i = 0; i < NUM_POOLS; ++i) - total += _manager->pools[i].npages; - - return total; -} - /** * Callback for mm to request pool to reduce number of page held. + * + * XXX: (dchinner) Deadlock warning! + * + * ttm_page_pool_free() does memory allocation using GFP_KERNEL. that means + * this can deadlock when called a sc->gfp_mask that is not equal to + * GFP_KERNEL. + * + * This code is crying out for a shrinker per pool.... */ -static int ttm_pool_mm_shrink(struct shrinker *shrink, - struct shrink_control *sc) +static unsigned long +ttm_pool_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) { static atomic_t start_pool = ATOMIC_INIT(0); unsigned i; unsigned pool_offset = atomic_add_return(1, &start_pool); struct ttm_page_pool *pool; int shrink_pages = sc->nr_to_scan; + unsigned long freed = 0; pool_offset = pool_offset % NUM_POOLS; /* select start pool in round robin fashion */ @@ -408,14 +406,28 @@ static int ttm_pool_mm_shrink(struct shrinker *shrink, break; pool = &_manager->pools[(i + pool_offset)%NUM_POOLS]; shrink_pages = ttm_page_pool_free(pool, nr_free); + freed += nr_free - shrink_pages; } - /* return estimated number of unused pages in pool */ - return ttm_pool_get_num_unused_pages(); + return freed; +} + + +static unsigned long +ttm_pool_shrink_count(struct shrinker *shrink, struct shrink_control *sc) +{ + unsigned i; + unsigned long count = 0; + + for (i = 0; i < NUM_POOLS; ++i) + count += _manager->pools[i].npages; + + return count; } static void ttm_pool_mm_shrink_init(struct ttm_pool_manager *manager) { - manager->mm_shrink.shrink = &ttm_pool_mm_shrink; + manager->mm_shrink.count_objects = ttm_pool_shrink_count; + manager->mm_shrink.scan_objects = ttm_pool_shrink_scan; manager->mm_shrink.seeks = 1; register_shrinker(&manager->mm_shrink); } diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c index b8b394319b45..7957beeeaf73 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c @@ -918,19 +918,6 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev) } EXPORT_SYMBOL_GPL(ttm_dma_populate); -/* Get good estimation how many pages are free in pools */ -static int ttm_dma_pool_get_num_unused_pages(void) -{ - struct device_pools *p; - unsigned total = 0; - - mutex_lock(&_manager->lock); - list_for_each_entry(p, &_manager->pools, pools) - total += p->pool->npages_free; - mutex_unlock(&_manager->lock); - return total; -} - /* Put all pages in pages list to correct pool to wait for reuse */ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev) { @@ -1002,18 +989,29 @@ EXPORT_SYMBOL_GPL(ttm_dma_unpopulate); /** * Callback for mm to request pool to reduce number of page held. + * + * XXX: (dchinner) Deadlock warning! + * + * ttm_dma_page_pool_free() does GFP_KERNEL memory allocation, and so attention + * needs to be paid to sc->gfp_mask to determine if this can be done or not. + * GFP_KERNEL memory allocation in a GFP_ATOMIC reclaim context woul dbe really + * bad. + * + * I'm getting sadder as I hear more pathetical whimpers about needing per-pool + * shrinkers */ -static int ttm_dma_pool_mm_shrink(struct shrinker *shrink, - struct shrink_control *sc) +static unsigned long +ttm_dma_pool_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) { static atomic_t start_pool = ATOMIC_INIT(0); unsigned idx = 0; unsigned pool_offset = atomic_add_return(1, &start_pool); unsigned shrink_pages = sc->nr_to_scan; struct device_pools *p; + unsigned long freed = 0; if (list_empty(&_manager->pools)) - return 0; + return SHRINK_STOP; mutex_lock(&_manager->lock); pool_offset = pool_offset % _manager->npools; @@ -1029,18 +1027,33 @@ static int ttm_dma_pool_mm_shrink(struct shrinker *shrink, continue; nr_free = shrink_pages; shrink_pages = ttm_dma_page_pool_free(p->pool, nr_free); + freed += nr_free - shrink_pages; + pr_debug("%s: (%s:%d) Asked to shrink %d, have %d more to go\n", p->pool->dev_name, p->pool->name, current->pid, nr_free, shrink_pages); } mutex_unlock(&_manager->lock); - /* return estimated number of unused pages in pool */ - return ttm_dma_pool_get_num_unused_pages(); + return freed; +} + +static unsigned long +ttm_dma_pool_shrink_count(struct shrinker *shrink, struct shrink_control *sc) +{ + struct device_pools *p; + unsigned long count = 0; + + mutex_lock(&_manager->lock); + list_for_each_entry(p, &_manager->pools, pools) + count += p->pool->npages_free; + mutex_unlock(&_manager->lock); + return count; } static void ttm_dma_pool_mm_shrink_init(struct ttm_pool_manager *manager) { - manager->mm_shrink.shrink = &ttm_dma_pool_mm_shrink; + manager->mm_shrink.count_objects = ttm_dma_pool_shrink_count; + manager->mm_shrink.scan_objects = &ttm_dma_pool_shrink_scan; manager->mm_shrink.seeks = 1; register_shrinker(&manager->mm_shrink); } diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c index ee372884c405..f9764e61978b 100644 --- a/drivers/md/bcache/btree.c +++ b/drivers/md/bcache/btree.c @@ -597,24 +597,19 @@ static int mca_reap(struct btree *b, struct closure *cl, unsigned min_order) return 0; } -static int bch_mca_shrink(struct shrinker *shrink, struct shrink_control *sc) +static unsigned long bch_mca_scan(struct shrinker *shrink, + struct shrink_control *sc) { struct cache_set *c = container_of(shrink, struct cache_set, shrink); struct btree *b, *t; unsigned long i, nr = sc->nr_to_scan; + unsigned long freed = 0; if (c->shrinker_disabled) - return 0; + return SHRINK_STOP; if (c->try_harder) - return 0; - - /* - * If nr == 0, we're supposed to return the number of items we have - * cached. Not allowed to return -1. - */ - if (!nr) - return mca_can_free(c) * c->btree_pages; + return SHRINK_STOP; /* Return -1 if we can't do anything right now */ if (sc->gfp_mask & __GFP_WAIT) @@ -634,14 +629,14 @@ static int bch_mca_shrink(struct shrinker *shrink, struct shrink_control *sc) i = 0; list_for_each_entry_safe(b, t, &c->btree_cache_freeable, list) { - if (!nr) + if (freed >= nr) break; if (++i > 3 && !mca_reap(b, NULL, 0)) { mca_data_free(b); rw_unlock(true, b); - --nr; + freed++; } } @@ -652,7 +647,7 @@ static int bch_mca_shrink(struct shrinker *shrink, struct shrink_control *sc) if (list_empty(&c->btree_cache)) goto out; - for (i = 0; nr && i < c->bucket_cache_used; i++) { + for (i = 0; (nr--) && i < c->bucket_cache_used; i++) { b = list_first_entry(&c->btree_cache, struct btree, list); list_rotate_left(&c->btree_cache); @@ -661,14 +656,27 @@ static int bch_mca_shrink(struct shrinker *shrink, struct shrink_control *sc) mca_bucket_free(b); mca_data_free(b); rw_unlock(true, b); - --nr; + freed++; } else b->accessed = 0; } out: - nr = mca_can_free(c) * c->btree_pages; mutex_unlock(&c->bucket_lock); - return nr; + return freed; +} + +static unsigned long bch_mca_count(struct shrinker *shrink, + struct shrink_control *sc) +{ + struct cache_set *c = container_of(shrink, struct cache_set, shrink); + + if (c->shrinker_disabled) + return 0; + + if (c->try_harder) + return 0; + + return mca_can_free(c) * c->btree_pages; } void bch_btree_cache_free(struct cache_set *c) @@ -737,7 +745,8 @@ int bch_btree_cache_alloc(struct cache_set *c) c->verify_data = NULL; #endif - c->shrink.shrink = bch_mca_shrink; + c->shrink.count_objects = bch_mca_count; + c->shrink.scan_objects = bch_mca_scan; c->shrink.seeks = 4; c->shrink.batch = c->btree_pages * 2; register_shrinker(&c->shrink); diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c index 12a2c2846f99..4fe6ab2fbe2e 100644 --- a/drivers/md/bcache/sysfs.c +++ b/drivers/md/bcache/sysfs.c @@ -556,7 +556,7 @@ STORE(__bch_cache_set) struct shrink_control sc; sc.gfp_mask = GFP_KERNEL; sc.nr_to_scan = strtoul_or_return(buf); - c->shrink.shrink(&c->shrink, &sc); + c->shrink.scan_objects(&c->shrink, &sc); } sysfs_strtoul(congested_read_threshold_us, diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c index 5227e079a6e3..173cbb20d104 100644 --- a/drivers/md/dm-bufio.c +++ b/drivers/md/dm-bufio.c @@ -1425,62 +1425,75 @@ static int __cleanup_old_buffer(struct dm_buffer *b, gfp_t gfp, unsigned long max_jiffies) { if (jiffies - b->last_accessed < max_jiffies) - return 1; + return 0; if (!(gfp & __GFP_IO)) { if (test_bit(B_READING, &b->state) || test_bit(B_WRITING, &b->state) || test_bit(B_DIRTY, &b->state)) - return 1; + return 0; } if (b->hold_count) - return 1; + return 0; __make_buffer_clean(b); __unlink_buffer(b); __free_buffer_wake(b); - return 0; + return 1; } -static void __scan(struct dm_bufio_client *c, unsigned long nr_to_scan, - struct shrink_control *sc) +static long __scan(struct dm_bufio_client *c, unsigned long nr_to_scan, + gfp_t gfp_mask) { int l; struct dm_buffer *b, *tmp; + long freed = 0; for (l = 0; l < LIST_SIZE; l++) { - list_for_each_entry_safe_reverse(b, tmp, &c->lru[l], lru_list) - if (!__cleanup_old_buffer(b, sc->gfp_mask, 0) && - !--nr_to_scan) - return; + list_for_each_entry_safe_reverse(b, tmp, &c->lru[l], lru_list) { + freed += __cleanup_old_buffer(b, gfp_mask, 0); + if (!--nr_to_scan) + break; + } dm_bufio_cond_resched(); } + return freed; } -static int shrink(struct shrinker *shrinker, struct shrink_control *sc) +static unsigned long +dm_bufio_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) { - struct dm_bufio_client *c = - container_of(shrinker, struct dm_bufio_client, shrinker); - unsigned long r; - unsigned long nr_to_scan = sc->nr_to_scan; + struct dm_bufio_client *c; + unsigned long freed; + c = container_of(shrink, struct dm_bufio_client, shrinker); if (sc->gfp_mask & __GFP_IO) dm_bufio_lock(c); else if (!dm_bufio_trylock(c)) - return !nr_to_scan ? 0 : -1; + return SHRINK_STOP; - if (nr_to_scan) - __scan(c, nr_to_scan, sc); + freed = __scan(c, sc->nr_to_scan, sc->gfp_mask); + dm_bufio_unlock(c); + return freed; +} - r = c->n_buffers[LIST_CLEAN] + c->n_buffers[LIST_DIRTY]; - if (r > INT_MAX) - r = INT_MAX; +static unsigned long +dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc) +{ + struct dm_bufio_client *c; + unsigned long count; - dm_bufio_unlock(c); + c = container_of(shrink, struct dm_bufio_client, shrinker); + if (sc->gfp_mask & __GFP_IO) + dm_bufio_lock(c); + else if (!dm_bufio_trylock(c)) + return 0; - return r; + count = c->n_buffers[LIST_CLEAN] + c->n_buffers[LIST_DIRTY]; + dm_bufio_unlock(c); + return count; } /* @@ -1582,7 +1595,8 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign __cache_size_refresh(); mutex_unlock(&dm_bufio_clients_lock); - c->shrinker.shrink = shrink; + c->shrinker.count_objects = dm_bufio_shrink_count; + c->shrinker.scan_objects = dm_bufio_shrink_scan; c->shrinker.seeks = 1; c->shrinker.batch = 0; register_shrinker(&c->shrinker); @@ -1669,7 +1683,7 @@ static void cleanup_old_buffers(void) struct dm_buffer *b; b = list_entry(c->lru[LIST_CLEAN].prev, struct dm_buffer, lru_list); - if (__cleanup_old_buffer(b, 0, max_age * HZ)) + if (!__cleanup_old_buffer(b, 0, max_age * HZ)) break; dm_bufio_cond_resched(); } diff --git a/drivers/staging/android/ashmem.c b/drivers/staging/android/ashmem.c index 21a3f7250531..8e76ddca0999 100644 --- a/drivers/staging/android/ashmem.c +++ b/drivers/staging/android/ashmem.c @@ -341,27 +341,26 @@ out: /* * ashmem_shrink - our cache shrinker, called from mm/vmscan.c :: shrink_slab * - * 'nr_to_scan' is the number of objects (pages) to prune, or 0 to query how - * many objects (pages) we have in total. + * 'nr_to_scan' is the number of objects to scan for freeing. * * 'gfp_mask' is the mask of the allocation that got us into this mess. * - * Return value is the number of objects (pages) remaining, or -1 if we cannot + * Return value is the number of objects freed or -1 if we cannot * proceed without risk of deadlock (due to gfp_mask). * * We approximate LRU via least-recently-unpinned, jettisoning unpinned partial * chunks of ashmem regions LRU-wise one-at-a-time until we hit 'nr_to_scan' * pages freed. */ -static int ashmem_shrink(struct shrinker *s, struct shrink_control *sc) +static unsigned long +ashmem_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) { struct ashmem_range *range, *next; + unsigned long freed = 0; /* We might recurse into filesystem code, so bail out if necessary */ - if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) - return -1; - if (!sc->nr_to_scan) - return lru_count; + if (!(sc->gfp_mask & __GFP_FS)) + return SHRINK_STOP; mutex_lock(&ashmem_mutex); list_for_each_entry_safe(range, next, &ashmem_lru_list, lru) { @@ -374,17 +373,32 @@ static int ashmem_shrink(struct shrinker *s, struct shrink_control *sc) range->purged = ASHMEM_WAS_PURGED; lru_del(range); - sc->nr_to_scan -= range_size(range); - if (sc->nr_to_scan <= 0) + freed += range_size(range); + if (--sc->nr_to_scan <= 0) break; } mutex_unlock(&ashmem_mutex); + return freed; +} +static unsigned long +ashmem_shrink_count(struct shrinker *shrink, struct shrink_control *sc) +{ + /* + * note that lru_count is count of pages on the lru, not a count of + * objects on the list. This means the scan function needs to return the + * number of pages freed, not the number of objects scanned. + */ return lru_count; } static struct shrinker ashmem_shrinker = { - .shrink = ashmem_shrink, + .count_objects = ashmem_shrink_count, + .scan_objects = ashmem_shrink_scan, + /* + * XXX (dchinner): I wish people would comment on why they need on + * significant changes to the default value here + */ .seeks = DEFAULT_SEEKS * 4, }; @@ -690,11 +704,11 @@ static long ashmem_ioctl(struct file *file, unsigned int cmd, unsigned long arg) if (capable(CAP_SYS_ADMIN)) { struct shrink_control sc = { .gfp_mask = GFP_KERNEL, - .nr_to_scan = 0, + .nr_to_scan = LONG_MAX, }; - ret = ashmem_shrink(&ashmem_shrinker, &sc); - sc.nr_to_scan = ret; - ashmem_shrink(&ashmem_shrinker, &sc); + + nodes_setall(sc.nodes_to_scan); + ashmem_shrink_scan(&ashmem_shrinker, &sc); } break; } diff --git a/drivers/staging/android/lowmemorykiller.c b/drivers/staging/android/lowmemorykiller.c index fe74494868ef..6f094b37f1f1 100644 --- a/drivers/staging/android/lowmemorykiller.c +++ b/drivers/staging/android/lowmemorykiller.c @@ -66,11 +66,20 @@ static unsigned long lowmem_deathpending_timeout; pr_info(x); \ } while (0) -static int lowmem_shrink(struct shrinker *s, struct shrink_control *sc) +static unsigned long lowmem_count(struct shrinker *s, + struct shrink_control *sc) +{ + return global_page_state(NR_ACTIVE_ANON) + + global_page_state(NR_ACTIVE_FILE) + + global_page_state(NR_INACTIVE_ANON) + + global_page_state(NR_INACTIVE_FILE); +} + +static unsigned long lowmem_scan(struct shrinker *s, struct shrink_control *sc) { struct task_struct *tsk; struct task_struct *selected = NULL; - int rem = 0; + unsigned long rem = 0; int tasksize; int i; short min_score_adj = OOM_SCORE_ADJ_MAX + 1; @@ -92,19 +101,17 @@ static int lowmem_shrink(struct shrinker *s, struct shrink_control *sc) break; } } - if (sc->nr_to_scan > 0) - lowmem_print(3, "lowmem_shrink %lu, %x, ofree %d %d, ma %hd\n", - sc->nr_to_scan, sc->gfp_mask, other_free, - other_file, min_score_adj); - rem = global_page_state(NR_ACTIVE_ANON) + - global_page_state(NR_ACTIVE_FILE) + - global_page_state(NR_INACTIVE_ANON) + - global_page_state(NR_INACTIVE_FILE); - if (sc->nr_to_scan <= 0 || min_score_adj == OOM_SCORE_ADJ_MAX + 1) { - lowmem_print(5, "lowmem_shrink %lu, %x, return %d\n", - sc->nr_to_scan, sc->gfp_mask, rem); - return rem; + + lowmem_print(3, "lowmem_scan %lu, %x, ofree %d %d, ma %hd\n", + sc->nr_to_scan, sc->gfp_mask, other_free, + other_file, min_score_adj); + + if (min_score_adj == OOM_SCORE_ADJ_MAX + 1) { + lowmem_print(5, "lowmem_scan %lu, %x, return 0\n", + sc->nr_to_scan, sc->gfp_mask); + return 0; } + selected_oom_score_adj = min_score_adj; rcu_read_lock(); @@ -154,16 +161,18 @@ static int lowmem_shrink(struct shrinker *s, struct shrink_control *sc) lowmem_deathpending_timeout = jiffies + HZ; send_sig(SIGKILL, selected, 0); set_tsk_thread_flag(selected, TIF_MEMDIE); - rem -= selected_tasksize; + rem += selected_tasksize; } - lowmem_print(4, "lowmem_shrink %lu, %x, return %d\n", + + lowmem_print(4, "lowmem_scan %lu, %x, return %lu\n", sc->nr_to_scan, sc->gfp_mask, rem); rcu_read_unlock(); return rem; } static struct shrinker lowmem_shrinker = { - .shrink = lowmem_shrink, + .scan_objects = lowmem_scan, + .count_objects = lowmem_count, .seeks = DEFAULT_SEEKS * 16 }; diff --git a/drivers/staging/lustre/include/linux/libcfs/linux/linux-mem.h b/drivers/staging/lustre/include/linux/libcfs/linux/linux-mem.h index 63efb7b456c6..2af15d41e77a 100644 --- a/drivers/staging/lustre/include/linux/libcfs/linux/linux-mem.h +++ b/drivers/staging/lustre/include/linux/libcfs/linux/linux-mem.h @@ -79,42 +79,4 @@ do { __oldfs = get_fs(); set_fs(get_ds());} while(0) #define MMSPACE_CLOSE set_fs(__oldfs) -/* - * Shrinker - */ - -# define SHRINKER_ARGS(sc, nr_to_scan, gfp_mask) \ - struct shrinker *shrinker, \ - struct shrink_control *sc -# define shrink_param(sc, var) ((sc)->var) - -typedef int (*shrinker_t)(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask)); - -static inline -struct shrinker *set_shrinker(int seek, shrinker_t func) -{ - struct shrinker *s; - - s = kmalloc(sizeof(*s), GFP_KERNEL); - if (s == NULL) - return (NULL); - - s->shrink = func; - s->seeks = seek; - - register_shrinker(s); - - return s; -} - -static inline -void remove_shrinker(struct shrinker *shrinker) -{ - if (shrinker == NULL) - return; - - unregister_shrinker(shrinker); - kfree(shrinker); -} - #endif /* __LINUX_CFS_MEM_H__ */ diff --git a/drivers/staging/lustre/lustre/ldlm/ldlm_pool.c b/drivers/staging/lustre/lustre/ldlm/ldlm_pool.c index 454027d68d54..0025ee6356da 100644 --- a/drivers/staging/lustre/lustre/ldlm/ldlm_pool.c +++ b/drivers/staging/lustre/lustre/ldlm/ldlm_pool.c @@ -521,7 +521,7 @@ static int ldlm_cli_pool_shrink(struct ldlm_pool *pl, int nr, unsigned int gfp_mask) { struct ldlm_namespace *ns; - int canceled = 0, unused; + int unused; ns = ldlm_pl2ns(pl); @@ -540,14 +540,10 @@ static int ldlm_cli_pool_shrink(struct ldlm_pool *pl, unused = ns->ns_nr_unused; spin_unlock(&ns->ns_lock); - if (nr) { - canceled = ldlm_cancel_lru(ns, nr, LCF_ASYNC, - LDLM_CANCEL_SHRINK); - } - /* - * Return the number of potentially reclaimable locks. - */ - return ((unused - canceled) / 100) * sysctl_vfs_cache_pressure; + if (nr == 0) + return (unused / 100) * sysctl_vfs_cache_pressure; + else + return ldlm_cancel_lru(ns, nr, LCF_ASYNC, LDLM_CANCEL_SHRINK); } struct ldlm_pool_ops ldlm_srv_pool_ops = { @@ -601,9 +597,10 @@ int ldlm_pool_recalc(struct ldlm_pool *pl) return recalc_interval_sec; } -/** +/* * Pool shrink wrapper. Will call either client or server pool recalc callback - * depending what pool \a pl is used. + * depending what pool pl is used. When nr == 0, just return the number of + * freeable locks. Otherwise, return the number of canceled locks. */ int ldlm_pool_shrink(struct ldlm_pool *pl, int nr, unsigned int gfp_mask) @@ -1017,29 +1014,24 @@ static int ldlm_pool_granted(struct ldlm_pool *pl) } static struct ptlrpc_thread *ldlm_pools_thread; -static struct shrinker *ldlm_pools_srv_shrinker; -static struct shrinker *ldlm_pools_cli_shrinker; static struct completion ldlm_pools_comp; /* - * Cancel \a nr locks from all namespaces (if possible). Returns number of - * cached locks after shrink is finished. All namespaces are asked to - * cancel approximately equal amount of locks to keep balancing. + * count locks from all namespaces (if possible). Returns number of + * cached locks. */ -static int ldlm_pools_shrink(ldlm_side_t client, int nr, - unsigned int gfp_mask) +static unsigned long ldlm_pools_count(ldlm_side_t client, unsigned int gfp_mask) { - int total = 0, cached = 0, nr_ns; + int total = 0, nr_ns; struct ldlm_namespace *ns; struct ldlm_namespace *ns_old = NULL; /* loop detection */ void *cookie; - if (client == LDLM_NAMESPACE_CLIENT && nr != 0 && - !(gfp_mask & __GFP_FS)) - return -1; + if (client == LDLM_NAMESPACE_CLIENT && !(gfp_mask & __GFP_FS)) + return 0; - CDEBUG(D_DLMTRACE, "Request to shrink %d %s locks from all pools\n", - nr, client == LDLM_NAMESPACE_CLIENT ? "client" : "server"); + CDEBUG(D_DLMTRACE, "Request to count %s locks from all pools\n", + client == LDLM_NAMESPACE_CLIENT ? "client" : "server"); cookie = cl_env_reenter(); @@ -1047,8 +1039,7 @@ static int ldlm_pools_shrink(ldlm_side_t client, int nr, * Find out how many resources we may release. */ for (nr_ns = ldlm_namespace_nr_read(client); - nr_ns > 0; nr_ns--) - { + nr_ns > 0; nr_ns--) { mutex_lock(ldlm_namespace_lock(client)); if (list_empty(ldlm_namespace_list(client))) { mutex_unlock(ldlm_namespace_lock(client)); @@ -1078,17 +1069,27 @@ static int ldlm_pools_shrink(ldlm_side_t client, int nr, ldlm_namespace_put(ns); } - if (nr == 0 || total == 0) { - cl_env_reexit(cookie); - return total; - } + cl_env_reexit(cookie); + return total; +} + +static unsigned long ldlm_pools_scan(ldlm_side_t client, int nr, unsigned int gfp_mask) +{ + unsigned long freed = 0; + int tmp, nr_ns; + struct ldlm_namespace *ns; + void *cookie; + + if (client == LDLM_NAMESPACE_CLIENT && !(gfp_mask & __GFP_FS)) + return -1; + + cookie = cl_env_reenter(); /* - * Shrink at least ldlm_namespace_nr(client) namespaces. + * Shrink at least ldlm_namespace_nr_read(client) namespaces. */ - for (nr_ns = ldlm_namespace_nr_read(client) - nr_ns; - nr_ns > 0; nr_ns--) - { + for (tmp = nr_ns = ldlm_namespace_nr_read(client); + tmp > 0; tmp--) { int cancel, nr_locks; /* @@ -1097,12 +1098,6 @@ static int ldlm_pools_shrink(ldlm_side_t client, int nr, mutex_lock(ldlm_namespace_lock(client)); if (list_empty(ldlm_namespace_list(client))) { mutex_unlock(ldlm_namespace_lock(client)); - /* - * If list is empty, we can't return any @cached > 0, - * that probably would cause needless shrinker - * call. - */ - cached = 0; break; } ns = ldlm_namespace_first_locked(client); @@ -1111,29 +1106,42 @@ static int ldlm_pools_shrink(ldlm_side_t client, int nr, mutex_unlock(ldlm_namespace_lock(client)); nr_locks = ldlm_pool_granted(&ns->ns_pool); - cancel = 1 + nr_locks * nr / total; - ldlm_pool_shrink(&ns->ns_pool, cancel, gfp_mask); - cached += ldlm_pool_granted(&ns->ns_pool); + /* + * We use to shrink propotionally but with new shrinker API, + * we lost the total number of freeable locks. + */ + cancel = 1 + min_t(int, nr_locks, nr / nr_ns); + freed += ldlm_pool_shrink(&ns->ns_pool, cancel, gfp_mask); ldlm_namespace_put(ns); } cl_env_reexit(cookie); - /* we only decrease the SLV in server pools shrinker, return -1 to - * kernel to avoid needless loop. LU-1128 */ - return (client == LDLM_NAMESPACE_SERVER) ? -1 : cached; + /* + * we only decrease the SLV in server pools shrinker, return + * SHRINK_STOP to kernel to avoid needless loop. LU-1128 + */ + return (client == LDLM_NAMESPACE_SERVER) ? SHRINK_STOP : freed; +} + +static unsigned long ldlm_pools_srv_count(struct shrinker *s, struct shrink_control *sc) +{ + return ldlm_pools_count(LDLM_NAMESPACE_SERVER, sc->gfp_mask); } -static int ldlm_pools_srv_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask)) +static unsigned long ldlm_pools_srv_scan(struct shrinker *s, struct shrink_control *sc) { - return ldlm_pools_shrink(LDLM_NAMESPACE_SERVER, - shrink_param(sc, nr_to_scan), - shrink_param(sc, gfp_mask)); + return ldlm_pools_scan(LDLM_NAMESPACE_SERVER, sc->nr_to_scan, + sc->gfp_mask); } -static int ldlm_pools_cli_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask)) +static unsigned long ldlm_pools_cli_count(struct shrinker *s, struct shrink_control *sc) { - return ldlm_pools_shrink(LDLM_NAMESPACE_CLIENT, - shrink_param(sc, nr_to_scan), - shrink_param(sc, gfp_mask)); + return ldlm_pools_count(LDLM_NAMESPACE_CLIENT, sc->gfp_mask); +} + +static unsigned long ldlm_pools_cli_scan(struct shrinker *s, struct shrink_control *sc) +{ + return ldlm_pools_scan(LDLM_NAMESPACE_CLIENT, sc->nr_to_scan, + sc->gfp_mask); } int ldlm_pools_recalc(ldlm_side_t client) @@ -1216,7 +1224,7 @@ int ldlm_pools_recalc(ldlm_side_t client) } /* - * Recalc at least ldlm_namespace_nr(client) namespaces. + * Recalc at least ldlm_namespace_nr_read(client) namespaces. */ for (nr = ldlm_namespace_nr_read(client); nr > 0; nr--) { int skip; @@ -1383,18 +1391,26 @@ static void ldlm_pools_thread_stop(void) ldlm_pools_thread = NULL; } +static struct shrinker ldlm_pools_srv_shrinker = { + .count_objects = ldlm_pools_srv_count, + .scan_objects = ldlm_pools_srv_scan, + .seeks = DEFAULT_SEEKS, +}; + +static struct shrinker ldlm_pools_cli_shrinker = { + .count_objects = ldlm_pools_cli_count, + .scan_objects = ldlm_pools_cli_scan, + .seeks = DEFAULT_SEEKS, +}; + int ldlm_pools_init(void) { int rc; rc = ldlm_pools_thread_start(); if (rc == 0) { - ldlm_pools_srv_shrinker = - set_shrinker(DEFAULT_SEEKS, - ldlm_pools_srv_shrink); - ldlm_pools_cli_shrinker = - set_shrinker(DEFAULT_SEEKS, - ldlm_pools_cli_shrink); + register_shrinker(&ldlm_pools_srv_shrinker); + register_shrinker(&ldlm_pools_cli_shrinker); } return rc; } @@ -1402,14 +1418,8 @@ EXPORT_SYMBOL(ldlm_pools_init); void ldlm_pools_fini(void) { - if (ldlm_pools_srv_shrinker != NULL) { - remove_shrinker(ldlm_pools_srv_shrinker); - ldlm_pools_srv_shrinker = NULL; - } - if (ldlm_pools_cli_shrinker != NULL) { - remove_shrinker(ldlm_pools_cli_shrinker); - ldlm_pools_cli_shrinker = NULL; - } + unregister_shrinker(&ldlm_pools_srv_shrinker); + unregister_shrinker(&ldlm_pools_cli_shrinker); ldlm_pools_thread_stop(); } EXPORT_SYMBOL(ldlm_pools_fini); diff --git a/drivers/staging/lustre/lustre/obdclass/lu_object.c b/drivers/staging/lustre/lustre/obdclass/lu_object.c index c29ac1c2defd..3a3d5bc5a628 100644 --- a/drivers/staging/lustre/lustre/obdclass/lu_object.c +++ b/drivers/staging/lustre/lustre/obdclass/lu_object.c @@ -1779,7 +1779,6 @@ int lu_env_refill_by_tags(struct lu_env *env, __u32 ctags, } EXPORT_SYMBOL(lu_env_refill_by_tags); -static struct shrinker *lu_site_shrinker = NULL; typedef struct lu_site_stats{ unsigned lss_populated; @@ -1835,61 +1834,68 @@ static void lu_site_stats_get(cfs_hash_t *hs, * objects without taking the lu_sites_guard lock, but this is not * possible in the current implementation. */ -static int lu_cache_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask)) +static unsigned long lu_cache_shrink_count(struct shrinker *sk, + struct shrink_control *sc) { lu_site_stats_t stats; struct lu_site *s; struct lu_site *tmp; - int cached = 0; - int remain = shrink_param(sc, nr_to_scan); - LIST_HEAD(splice); - - if (!(shrink_param(sc, gfp_mask) & __GFP_FS)) { - if (remain != 0) - return -1; - else - /* We must not take the lu_sites_guard lock when - * __GFP_FS is *not* set because of the deadlock - * possibility detailed above. Additionally, - * since we cannot determine the number of - * objects in the cache without taking this - * lock, we're in a particularly tough spot. As - * a result, we'll just lie and say our cache is - * empty. This _should_ be ok, as we can't - * reclaim objects when __GFP_FS is *not* set - * anyways. - */ - return 0; - } + unsigned long cached = 0; - CDEBUG(D_INODE, "Shrink %d objects\n", remain); + if (!(sc->gfp_mask & __GFP_FS)) + return 0; mutex_lock(&lu_sites_guard); list_for_each_entry_safe(s, tmp, &lu_sites, ls_linkage) { - if (shrink_param(sc, nr_to_scan) != 0) { - remain = lu_site_purge(&lu_shrink_env, s, remain); - /* - * Move just shrunk site to the tail of site list to - * assure shrinking fairness. - */ - list_move_tail(&s->ls_linkage, &splice); - } - memset(&stats, 0, sizeof(stats)); lu_site_stats_get(s->ls_obj_hash, &stats, 0); cached += stats.lss_total - stats.lss_busy; - if (shrink_param(sc, nr_to_scan) && remain <= 0) - break; } - list_splice(&splice, lu_sites.prev); mutex_unlock(&lu_sites_guard); cached = (cached / 100) * sysctl_vfs_cache_pressure; - if (shrink_param(sc, nr_to_scan) == 0) - CDEBUG(D_INODE, "%d objects cached\n", cached); + CDEBUG(D_INODE, "%ld objects cached\n", cached); return cached; } +static unsigned long lu_cache_shrink_scan(struct shrinker *sk, + struct shrink_control *sc) +{ + struct lu_site *s; + struct lu_site *tmp; + unsigned long remain = sc->nr_to_scan, freed = 0; + LIST_HEAD(splice); + + if (!(sc->gfp_mask & __GFP_FS)) + /* We must not take the lu_sites_guard lock when + * __GFP_FS is *not* set because of the deadlock + * possibility detailed above. Additionally, + * since we cannot determine the number of + * objects in the cache without taking this + * lock, we're in a particularly tough spot. As + * a result, we'll just lie and say our cache is + * empty. This _should_ be ok, as we can't + * reclaim objects when __GFP_FS is *not* set + * anyways. + */ + return SHRINK_STOP; + + mutex_lock(&lu_sites_guard); + list_for_each_entry_safe(s, tmp, &lu_sites, ls_linkage) { + freed = lu_site_purge(&lu_shrink_env, s, remain); + remain -= freed; + /* + * Move just shrunk site to the tail of site list to + * assure shrinking fairness. + */ + list_move_tail(&s->ls_linkage, &splice); + } + list_splice(&splice, lu_sites.prev); + mutex_unlock(&lu_sites_guard); + + return sc->nr_to_scan - remain; +} + /* * Debugging stuff. */ @@ -1913,6 +1919,12 @@ int lu_printk_printer(const struct lu_env *env, return 0; } +static struct shrinker lu_site_shrinker = { + .count_objects = lu_cache_shrink_count, + .scan_objects = lu_cache_shrink_scan, + .seeks = DEFAULT_SEEKS, +}; + /** * Initialization of global lu_* data. */ @@ -1947,9 +1959,7 @@ int lu_global_init(void) * inode, one for ea. Unfortunately setting this high value results in * lu_object/inode cache consuming all the memory. */ - lu_site_shrinker = set_shrinker(DEFAULT_SEEKS, lu_cache_shrink); - if (lu_site_shrinker == NULL) - return -ENOMEM; + register_shrinker(&lu_site_shrinker); return result; } @@ -1959,11 +1969,7 @@ int lu_global_init(void) */ void lu_global_fini(void) { - if (lu_site_shrinker != NULL) { - remove_shrinker(lu_site_shrinker); - lu_site_shrinker = NULL; - } - + unregister_shrinker(&lu_site_shrinker); lu_context_key_degister(&lu_global_key); /* diff --git a/drivers/staging/lustre/lustre/ptlrpc/sec_bulk.c b/drivers/staging/lustre/lustre/ptlrpc/sec_bulk.c index 9013745ab105..e90c8fb7da6a 100644 --- a/drivers/staging/lustre/lustre/ptlrpc/sec_bulk.c +++ b/drivers/staging/lustre/lustre/ptlrpc/sec_bulk.c @@ -121,13 +121,6 @@ static struct ptlrpc_enc_page_pool { } page_pools; /* - * memory shrinker - */ -const int pools_shrinker_seeks = DEFAULT_SEEKS; -static struct shrinker *pools_shrinker = NULL; - - -/* * /proc/fs/lustre/sptlrpc/encrypt_page_pools */ int sptlrpc_proc_enc_pool_seq_show(struct seq_file *m, void *v) @@ -226,30 +219,46 @@ static void enc_pools_release_free_pages(long npages) } /* - * could be called frequently for query (@nr_to_scan == 0). * we try to keep at least PTLRPC_MAX_BRW_PAGES pages in the pool. */ -static int enc_pools_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask)) +static unsigned long enc_pools_shrink_count(struct shrinker *s, + struct shrink_control *sc) { - if (unlikely(shrink_param(sc, nr_to_scan) != 0)) { + /* + * if no pool access for a long time, we consider it's fully idle. + * a little race here is fine. + */ + if (unlikely(cfs_time_current_sec() - page_pools.epp_last_access > + CACHE_QUIESCENT_PERIOD)) { spin_lock(&page_pools.epp_lock); - shrink_param(sc, nr_to_scan) = min_t(unsigned long, - shrink_param(sc, nr_to_scan), - page_pools.epp_free_pages - - PTLRPC_MAX_BRW_PAGES); - if (shrink_param(sc, nr_to_scan) > 0) { - enc_pools_release_free_pages(shrink_param(sc, - nr_to_scan)); - CDEBUG(D_SEC, "released %ld pages, %ld left\n", - (long)shrink_param(sc, nr_to_scan), - page_pools.epp_free_pages); - - page_pools.epp_st_shrinks++; - page_pools.epp_last_shrink = cfs_time_current_sec(); - } + page_pools.epp_idle_idx = IDLE_IDX_MAX; spin_unlock(&page_pools.epp_lock); } + LASSERT(page_pools.epp_idle_idx <= IDLE_IDX_MAX); + return max((int)page_pools.epp_free_pages - PTLRPC_MAX_BRW_PAGES, 0) * + (IDLE_IDX_MAX - page_pools.epp_idle_idx) / IDLE_IDX_MAX; +} + +/* + * we try to keep at least PTLRPC_MAX_BRW_PAGES pages in the pool. + */ +static unsigned long enc_pools_shrink_scan(struct shrinker *s, + struct shrink_control *sc) +{ + spin_lock(&page_pools.epp_lock); + sc->nr_to_scan = min_t(unsigned long, sc->nr_to_scan, + page_pools.epp_free_pages - PTLRPC_MAX_BRW_PAGES); + if (sc->nr_to_scan > 0) { + enc_pools_release_free_pages(sc->nr_to_scan); + CDEBUG(D_SEC, "released %ld pages, %ld left\n", + (long)sc->nr_to_scan, page_pools.epp_free_pages); + + page_pools.epp_st_shrinks++; + page_pools.epp_last_shrink = cfs_time_current_sec(); + } + spin_unlock(&page_pools.epp_lock); + /* * if no pool access for a long time, we consider it's fully idle. * a little race here is fine. @@ -262,8 +271,7 @@ static int enc_pools_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask)) } LASSERT(page_pools.epp_idle_idx <= IDLE_IDX_MAX); - return max((int)page_pools.epp_free_pages - PTLRPC_MAX_BRW_PAGES, 0) * - (IDLE_IDX_MAX - page_pools.epp_idle_idx) / IDLE_IDX_MAX; + return sc->nr_to_scan; } static inline @@ -699,6 +707,12 @@ static inline void enc_pools_free(void) sizeof(*page_pools.epp_pools)); } +static struct shrinker pools_shrinker = { + .count_objects = enc_pools_shrink_count, + .scan_objects = enc_pools_shrink_scan, + .seeks = DEFAULT_SEEKS, +}; + int sptlrpc_enc_pool_init(void) { /* @@ -736,12 +750,7 @@ int sptlrpc_enc_pool_init(void) if (page_pools.epp_pools == NULL) return -ENOMEM; - pools_shrinker = set_shrinker(pools_shrinker_seeks, - enc_pools_shrink); - if (pools_shrinker == NULL) { - enc_pools_free(); - return -ENOMEM; - } + register_shrinker(&pools_shrinker); return 0; } @@ -750,11 +759,10 @@ void sptlrpc_enc_pool_fini(void) { unsigned long cleaned, npools; - LASSERT(pools_shrinker); LASSERT(page_pools.epp_pools); LASSERT(page_pools.epp_total_pages == page_pools.epp_free_pages); - remove_shrinker(pools_shrinker); + unregister_shrinker(&pools_shrinker); npools = npages_to_npools(page_pools.epp_total_pages); cleaned = enc_pools_cleanup(page_pools.epp_pools, npools); |