| Message ID | 1414763374-1744-1-git-send-email-jbacik@fb.com (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
On Fri, Oct 31, 2014 at 09:49:34AM -0400, Josef Bacik wrote: > Our gluster boxes were spending lots of time in statfs because our fs'es are > huge. The problem is statfs loops through all of the block groups looking for > read only block groups, and when you have several terabytes worth of data that > ends up being a lot of block groups. Move the read only block groups onto a > read only list and only proces that list in > btrfs_account_ro_block_groups_free_space to reduce the amount of churn. Thanks, Looks good. Reviewed-by: Liu Bo <bo.li.liu@oracle.com> -liubo > > Signed-off-by: Josef Bacik <jbacik@fb.com> > --- > V1->V2: > -list_for_each_entry was using the wrong ->member name. > > fs/btrfs/ctree.h | 4 ++++ > fs/btrfs/extent-tree.c | 36 +++++++++++++----------------------- > 2 files changed, 17 insertions(+), 23 deletions(-) > > diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h > index d557264e..438f087 100644 > --- a/fs/btrfs/ctree.h > +++ b/fs/btrfs/ctree.h > @@ -1170,6 +1170,7 @@ struct btrfs_space_info { > struct percpu_counter total_bytes_pinned; > > struct list_head list; > + struct list_head ro_bgs; > > struct rw_semaphore groups_sem; > /* for block groups in our same type */ > @@ -1305,6 +1306,9 @@ struct btrfs_block_group_cache { > > /* For delayed block group creation or deletion of empty block groups */ > struct list_head bg_list; > + > + /* For read-only block groups */ > + struct list_head ro_list; > }; > > /* delayed seq elem */ > diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c > index 0d599ba..f51004f 100644 > --- a/fs/btrfs/extent-tree.c > +++ b/fs/btrfs/extent-tree.c > @@ -3518,6 +3518,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, > found->chunk_alloc = 0; > found->flush = 0; > init_waitqueue_head(&found->wait); > + INIT_LIST_HEAD(&found->ro_bgs); > > ret = kobject_init_and_add(&found->kobj, &space_info_ktype, > info->space_info_kobj, "%s", > @@ -8525,6 +8526,7 @@ static int set_block_group_ro(struct btrfs_block_group_cache *cache, int force) > min_allocable_bytes <= sinfo->total_bytes) { > sinfo->bytes_readonly += num_bytes; > cache->ro = 1; > + list_add_tail(&cache->ro_list, &sinfo->ro_bgs); > ret = 0; > } > out: > @@ -8579,15 +8581,20 @@ int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, > > /* > * helper to account the unused space of all the readonly block group in the > - * list. takes mirrors into account. > + * space_info. takes mirrors into account. > */ > -static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list) > +u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo) > { > struct btrfs_block_group_cache *block_group; > u64 free_bytes = 0; > int factor; > > - list_for_each_entry(block_group, groups_list, list) { > + /* It's df, we don't care if it's racey */ > + if (list_empty(&sinfo->ro_bgs)) > + return 0; > + > + spin_lock(&sinfo->lock); > + list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) { > spin_lock(&block_group->lock); > > if (!block_group->ro) { > @@ -8608,26 +8615,6 @@ static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list) > > spin_unlock(&block_group->lock); > } > - > - return free_bytes; > -} > - > -/* > - * helper to account the unused space of all the readonly block group in the > - * space_info. takes mirrors into account. > - */ > -u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo) > -{ > - int i; > - u64 free_bytes = 0; > - > - spin_lock(&sinfo->lock); > - > - for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) > - if (!list_empty(&sinfo->block_groups[i])) > - free_bytes += __btrfs_get_ro_block_group_free_space( > - &sinfo->block_groups[i]); > - > spin_unlock(&sinfo->lock); > > return free_bytes; > @@ -8647,6 +8634,7 @@ void btrfs_set_block_group_rw(struct btrfs_root *root, > cache->bytes_super - btrfs_block_group_used(&cache->item); > sinfo->bytes_readonly -= num_bytes; > cache->ro = 0; > + list_del_init(&cache->ro_list); > spin_unlock(&cache->lock); > spin_unlock(&sinfo->lock); > } > @@ -9016,6 +9004,7 @@ btrfs_create_block_group_cache(struct btrfs_root *root, u64 start, u64 size) > INIT_LIST_HEAD(&cache->list); > INIT_LIST_HEAD(&cache->cluster_list); > INIT_LIST_HEAD(&cache->bg_list); > + INIT_LIST_HEAD(&cache->ro_list); > btrfs_init_free_space_ctl(cache); > > return cache; > @@ -9425,6 +9414,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, > * are still on the list after taking the semaphore > */ > list_del_init(&block_group->list); > + list_del_init(&block_group->ro_list); > if (list_empty(&block_group->space_info->block_groups[index])) { > kobj = block_group->space_info->block_group_kobjs[index]; > block_group->space_info->block_group_kobjs[index] = NULL; > -- > 1.8.3.1 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index d557264e..438f087 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1170,6 +1170,7 @@ struct btrfs_space_info { struct percpu_counter total_bytes_pinned; struct list_head list; + struct list_head ro_bgs; struct rw_semaphore groups_sem; /* for block groups in our same type */ @@ -1305,6 +1306,9 @@ struct btrfs_block_group_cache { /* For delayed block group creation or deletion of empty block groups */ struct list_head bg_list; + + /* For read-only block groups */ + struct list_head ro_list; }; /* delayed seq elem */ diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 0d599ba..f51004f 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3518,6 +3518,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, found->chunk_alloc = 0; found->flush = 0; init_waitqueue_head(&found->wait); + INIT_LIST_HEAD(&found->ro_bgs); ret = kobject_init_and_add(&found->kobj, &space_info_ktype, info->space_info_kobj, "%s", @@ -8525,6 +8526,7 @@ static int set_block_group_ro(struct btrfs_block_group_cache *cache, int force) min_allocable_bytes <= sinfo->total_bytes) { sinfo->bytes_readonly += num_bytes; cache->ro = 1; + list_add_tail(&cache->ro_list, &sinfo->ro_bgs); ret = 0; } out: @@ -8579,15 +8581,20 @@ int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, /* * helper to account the unused space of all the readonly block group in the - * list. takes mirrors into account. + * space_info. takes mirrors into account. */ -static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list) +u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo) { struct btrfs_block_group_cache *block_group; u64 free_bytes = 0; int factor; - list_for_each_entry(block_group, groups_list, list) { + /* It's df, we don't care if it's racey */ + if (list_empty(&sinfo->ro_bgs)) + return 0; + + spin_lock(&sinfo->lock); + list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) { spin_lock(&block_group->lock); if (!block_group->ro) { @@ -8608,26 +8615,6 @@ static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list) spin_unlock(&block_group->lock); } - - return free_bytes; -} - -/* - * helper to account the unused space of all the readonly block group in the - * space_info. takes mirrors into account. - */ -u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo) -{ - int i; - u64 free_bytes = 0; - - spin_lock(&sinfo->lock); - - for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) - if (!list_empty(&sinfo->block_groups[i])) - free_bytes += __btrfs_get_ro_block_group_free_space( - &sinfo->block_groups[i]); - spin_unlock(&sinfo->lock); return free_bytes; @@ -8647,6 +8634,7 @@ void btrfs_set_block_group_rw(struct btrfs_root *root, cache->bytes_super - btrfs_block_group_used(&cache->item); sinfo->bytes_readonly -= num_bytes; cache->ro = 0; + list_del_init(&cache->ro_list); spin_unlock(&cache->lock); spin_unlock(&sinfo->lock); } @@ -9016,6 +9004,7 @@ btrfs_create_block_group_cache(struct btrfs_root *root, u64 start, u64 size) INIT_LIST_HEAD(&cache->list); INIT_LIST_HEAD(&cache->cluster_list); INIT_LIST_HEAD(&cache->bg_list); + INIT_LIST_HEAD(&cache->ro_list); btrfs_init_free_space_ctl(cache); return cache; @@ -9425,6 +9414,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, * are still on the list after taking the semaphore */ list_del_init(&block_group->list); + list_del_init(&block_group->ro_list); if (list_empty(&block_group->space_info->block_groups[index])) { kobj = block_group->space_info->block_group_kobjs[index]; block_group->space_info->block_group_kobjs[index] = NULL;
Our gluster boxes were spending lots of time in statfs because our fs'es are huge. The problem is statfs loops through all of the block groups looking for read only block groups, and when you have several terabytes worth of data that ends up being a lot of block groups. Move the read only block groups onto a read only list and only proces that list in btrfs_account_ro_block_groups_free_space to reduce the amount of churn. Thanks, Signed-off-by: Josef Bacik <jbacik@fb.com> --- V1->V2: -list_for_each_entry was using the wrong ->member name. fs/btrfs/ctree.h | 4 ++++ fs/btrfs/extent-tree.c | 36 +++++++++++++----------------------- 2 files changed, 17 insertions(+), 23 deletions(-)