6.12-stable review patch. If anyone has any objections, please let me know.
------------------
From: Gao Xiang hsiangkao@linux.alibaba.com
[ Upstream commit 9c91f959626e6d9f460b8906e27c37fca1b6456a ]
Move related helpers into zdata.c as an intermediate step of getting rid of `struct erofs_workgroup`, and rename:
erofs_workgroup_put => z_erofs_put_pcluster erofs_workgroup_get => z_erofs_get_pcluster erofs_try_to_release_workgroup => erofs_try_to_release_pcluster erofs_shrink_workstation => z_erofs_shrink_scan
Reviewed-by: Chao Yu chao@kernel.org Signed-off-by: Gao Xiang hsiangkao@linux.alibaba.com Link: https://lore.kernel.org/r/20241021035323.3280682-2-hsiangkao@linux.alibaba.c... Stable-dep-of: db902986dee4 ("erofs: fix potential return value overflow of z_erofs_shrink_scan()") Signed-off-by: Sasha Levin sashal@kernel.org --- fs/erofs/internal.h | 8 ++-- fs/erofs/zdata.c | 102 ++++++++++++++++++++++++++++++++++++++--- fs/erofs/zutil.c | 107 +++----------------------------------------- 3 files changed, 105 insertions(+), 112 deletions(-)
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h index b9649e3b2dd56..5fbc76b65f5c3 100644 --- a/fs/erofs/internal.h +++ b/fs/erofs/internal.h @@ -452,17 +452,15 @@ static inline void erofs_pagepool_add(struct page **pagepool, struct page *page) void erofs_release_pages(struct page **pagepool);
#ifdef CONFIG_EROFS_FS_ZIP -void erofs_workgroup_put(struct erofs_workgroup *grp); -bool erofs_workgroup_get(struct erofs_workgroup *grp); -void erofs_workgroup_free_rcu(struct erofs_workgroup *grp); +extern atomic_long_t erofs_global_shrink_cnt; void erofs_shrinker_register(struct super_block *sb); void erofs_shrinker_unregister(struct super_block *sb); int __init erofs_init_shrinker(void); void erofs_exit_shrinker(void); int __init z_erofs_init_subsystem(void); void z_erofs_exit_subsystem(void); -int erofs_try_to_free_all_cached_folios(struct erofs_sb_info *sbi, - struct erofs_workgroup *egrp); +unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi, + unsigned long nr_shrink); int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map, int flags); void *z_erofs_get_gbuf(unsigned int requiredpages); diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c index 8c6082fc86b29..d2235f2a0acdc 100644 --- a/fs/erofs/zdata.c +++ b/fs/erofs/zdata.c @@ -587,8 +587,8 @@ static void z_erofs_bind_cache(struct z_erofs_decompress_frontend *fe) }
/* (erofs_shrinker) disconnect cached encoded data with pclusters */ -int erofs_try_to_free_all_cached_folios(struct erofs_sb_info *sbi, - struct erofs_workgroup *grp) +static int erofs_try_to_free_all_cached_folios(struct erofs_sb_info *sbi, + struct erofs_workgroup *grp) { struct z_erofs_pcluster *const pcl = container_of(grp, struct z_erofs_pcluster, obj); @@ -710,6 +710,23 @@ static int z_erofs_attach_page(struct z_erofs_decompress_frontend *fe, return ret; }
+static bool z_erofs_get_pcluster(struct erofs_workgroup *grp) +{ + if (lockref_get_not_zero(&grp->lockref)) + return true; + + spin_lock(&grp->lockref.lock); + if (__lockref_is_dead(&grp->lockref)) { + spin_unlock(&grp->lockref.lock); + return false; + } + + if (!grp->lockref.count++) + atomic_long_dec(&erofs_global_shrink_cnt); + spin_unlock(&grp->lockref.lock); + return true; +} + static int z_erofs_register_pcluster(struct z_erofs_decompress_frontend *fe) { struct erofs_map_blocks *map = &fe->map; @@ -757,7 +774,7 @@ static int z_erofs_register_pcluster(struct z_erofs_decompress_frontend *fe) xa_lock(&sbi->managed_pslots); pre = __xa_cmpxchg(&sbi->managed_pslots, grp->index, NULL, grp, GFP_KERNEL); - if (!pre || xa_is_err(pre) || erofs_workgroup_get(pre)) { + if (!pre || xa_is_err(pre) || z_erofs_get_pcluster(pre)) { xa_unlock(&sbi->managed_pslots); break; } @@ -801,7 +818,7 @@ static int z_erofs_pcluster_begin(struct z_erofs_decompress_frontend *fe) while (1) { rcu_read_lock(); grp = xa_load(&EROFS_SB(sb)->managed_pslots, blknr); - if (!grp || erofs_workgroup_get(grp)) { + if (!grp || z_erofs_get_pcluster(grp)) { DBG_BUGON(grp && blknr != grp->index); rcu_read_unlock(); break; @@ -869,7 +886,7 @@ static void z_erofs_rcu_callback(struct rcu_head *head) struct z_erofs_pcluster, rcu)); }
-void erofs_workgroup_free_rcu(struct erofs_workgroup *grp) +static void erofs_workgroup_free_rcu(struct erofs_workgroup *grp) { struct z_erofs_pcluster *const pcl = container_of(grp, struct z_erofs_pcluster, obj); @@ -877,6 +894,77 @@ void erofs_workgroup_free_rcu(struct erofs_workgroup *grp) call_rcu(&pcl->rcu, z_erofs_rcu_callback); }
+static bool erofs_try_to_release_pcluster(struct erofs_sb_info *sbi, + struct erofs_workgroup *grp) +{ + int free = false; + + spin_lock(&grp->lockref.lock); + if (grp->lockref.count) + goto out; + + /* + * Note that all cached folios should be detached before deleted from + * the XArray. Otherwise some folios could be still attached to the + * orphan old pcluster when the new one is available in the tree. + */ + if (erofs_try_to_free_all_cached_folios(sbi, grp)) + goto out; + + /* + * It's impossible to fail after the pcluster is freezed, but in order + * to avoid some race conditions, add a DBG_BUGON to observe this. + */ + DBG_BUGON(__xa_erase(&sbi->managed_pslots, grp->index) != grp); + + lockref_mark_dead(&grp->lockref); + free = true; +out: + spin_unlock(&grp->lockref.lock); + if (free) { + atomic_long_dec(&erofs_global_shrink_cnt); + erofs_workgroup_free_rcu(grp); + } + return free; +} + +unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi, + unsigned long nr_shrink) +{ + struct erofs_workgroup *grp; + unsigned int freed = 0; + unsigned long index; + + xa_lock(&sbi->managed_pslots); + xa_for_each(&sbi->managed_pslots, index, grp) { + /* try to shrink each valid pcluster */ + if (!erofs_try_to_release_pcluster(sbi, grp)) + continue; + xa_unlock(&sbi->managed_pslots); + + ++freed; + if (!--nr_shrink) + return freed; + xa_lock(&sbi->managed_pslots); + } + xa_unlock(&sbi->managed_pslots); + return freed; +} + +static void z_erofs_put_pcluster(struct z_erofs_pcluster *pcl) +{ + struct erofs_workgroup *grp = &pcl->obj; + + if (lockref_put_or_lock(&grp->lockref)) + return; + + DBG_BUGON(__lockref_is_dead(&grp->lockref)); + if (grp->lockref.count == 1) + atomic_long_inc(&erofs_global_shrink_cnt); + --grp->lockref.count; + spin_unlock(&grp->lockref.lock); +} + static void z_erofs_pcluster_end(struct z_erofs_decompress_frontend *fe) { struct z_erofs_pcluster *pcl = fe->pcl; @@ -895,7 +983,7 @@ static void z_erofs_pcluster_end(struct z_erofs_decompress_frontend *fe) * any longer if the pcluster isn't hosted by ourselves. */ if (fe->mode < Z_EROFS_PCLUSTER_FOLLOWED_NOINPLACE) - erofs_workgroup_put(&pcl->obj); + z_erofs_put_pcluster(pcl);
fe->pcl = NULL; } @@ -1327,7 +1415,7 @@ static int z_erofs_decompress_queue(const struct z_erofs_decompressqueue *io, if (z_erofs_is_inline_pcluster(be.pcl)) z_erofs_free_pcluster(be.pcl); else - erofs_workgroup_put(&be.pcl->obj); + z_erofs_put_pcluster(be.pcl); } return err; } diff --git a/fs/erofs/zutil.c b/fs/erofs/zutil.c index 218b0249a4822..75704f58ecfa9 100644 --- a/fs/erofs/zutil.c +++ b/fs/erofs/zutil.c @@ -2,6 +2,7 @@ /* * Copyright (C) 2018 HUAWEI, Inc. * https://www.huawei.com/ + * Copyright (C) 2024 Alibaba Cloud */ #include "internal.h"
@@ -19,13 +20,12 @@ static unsigned int z_erofs_gbuf_count, z_erofs_gbuf_nrpages, module_param_named(global_buffers, z_erofs_gbuf_count, uint, 0444); module_param_named(reserved_pages, z_erofs_rsv_nrpages, uint, 0444);
-static atomic_long_t erofs_global_shrink_cnt; /* for all mounted instances */ -/* protected by 'erofs_sb_list_lock' */ -static unsigned int shrinker_run_no; +atomic_long_t erofs_global_shrink_cnt; /* for all mounted instances */
-/* protects the mounted 'erofs_sb_list' */ +/* protects `erofs_sb_list_lock` and the mounted `erofs_sb_list` */ static DEFINE_SPINLOCK(erofs_sb_list_lock); static LIST_HEAD(erofs_sb_list); +static unsigned int shrinker_run_no; static struct shrinker *erofs_shrinker_info;
static unsigned int z_erofs_gbuf_id(void) @@ -214,97 +214,6 @@ void erofs_release_pages(struct page **pagepool) } }
-bool erofs_workgroup_get(struct erofs_workgroup *grp) -{ - if (lockref_get_not_zero(&grp->lockref)) - return true; - - spin_lock(&grp->lockref.lock); - if (__lockref_is_dead(&grp->lockref)) { - spin_unlock(&grp->lockref.lock); - return false; - } - - if (!grp->lockref.count++) - atomic_long_dec(&erofs_global_shrink_cnt); - spin_unlock(&grp->lockref.lock); - return true; -} - -static void __erofs_workgroup_free(struct erofs_workgroup *grp) -{ - atomic_long_dec(&erofs_global_shrink_cnt); - erofs_workgroup_free_rcu(grp); -} - -void erofs_workgroup_put(struct erofs_workgroup *grp) -{ - if (lockref_put_or_lock(&grp->lockref)) - return; - - DBG_BUGON(__lockref_is_dead(&grp->lockref)); - if (grp->lockref.count == 1) - atomic_long_inc(&erofs_global_shrink_cnt); - --grp->lockref.count; - spin_unlock(&grp->lockref.lock); -} - -static bool erofs_try_to_release_workgroup(struct erofs_sb_info *sbi, - struct erofs_workgroup *grp) -{ - int free = false; - - spin_lock(&grp->lockref.lock); - if (grp->lockref.count) - goto out; - - /* - * Note that all cached pages should be detached before deleted from - * the XArray. Otherwise some cached pages could be still attached to - * the orphan old workgroup when the new one is available in the tree. - */ - if (erofs_try_to_free_all_cached_folios(sbi, grp)) - goto out; - - /* - * It's impossible to fail after the workgroup is freezed, - * however in order to avoid some race conditions, add a - * DBG_BUGON to observe this in advance. - */ - DBG_BUGON(__xa_erase(&sbi->managed_pslots, grp->index) != grp); - - lockref_mark_dead(&grp->lockref); - free = true; -out: - spin_unlock(&grp->lockref.lock); - if (free) - __erofs_workgroup_free(grp); - return free; -} - -static unsigned long erofs_shrink_workstation(struct erofs_sb_info *sbi, - unsigned long nr_shrink) -{ - struct erofs_workgroup *grp; - unsigned int freed = 0; - unsigned long index; - - xa_lock(&sbi->managed_pslots); - xa_for_each(&sbi->managed_pslots, index, grp) { - /* try to shrink each valid workgroup */ - if (!erofs_try_to_release_workgroup(sbi, grp)) - continue; - xa_unlock(&sbi->managed_pslots); - - ++freed; - if (!--nr_shrink) - return freed; - xa_lock(&sbi->managed_pslots); - } - xa_unlock(&sbi->managed_pslots); - return freed; -} - void erofs_shrinker_register(struct super_block *sb) { struct erofs_sb_info *sbi = EROFS_SB(sb); @@ -321,8 +230,8 @@ void erofs_shrinker_unregister(struct super_block *sb) struct erofs_sb_info *const sbi = EROFS_SB(sb);
mutex_lock(&sbi->umount_mutex); - /* clean up all remaining workgroups in memory */ - erofs_shrink_workstation(sbi, ~0UL); + /* clean up all remaining pclusters in memory */ + z_erofs_shrink_scan(sbi, ~0UL);
spin_lock(&erofs_sb_list_lock); list_del(&sbi->list); @@ -370,9 +279,7 @@ static unsigned long erofs_shrink_scan(struct shrinker *shrink,
spin_unlock(&erofs_sb_list_lock); sbi->shrinker_run_no = run_no; - - freed += erofs_shrink_workstation(sbi, nr - freed); - + freed += z_erofs_shrink_scan(sbi, nr - freed); spin_lock(&erofs_sb_list_lock); /* Get the next list element before we move this one */ p = p->next;