| Message ID | 20200603101020.143372-3-anand.jain@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs: minor block group cleanups | expand |
On Wed, Jun 3, 2020 at 11:13 AM Anand Jain <anand.jain@oracle.com> wrote: > > The name 'count' is a very commonly used name. It is often difficult to > review the code to check if there is any leak. So rename it to > 'bg_count', which is unique enough. Hum? Seriously? count to bg_count? It's a member of the block group structure, adding a bg_ prefix adds no value at all, we know the count is related to the block group. I could somewhat understand if you named it 'refcount' instead. Thanks. > > This patch also serves as a preparatory patch to either make sure > btrfs_get_block_group() is used instead of open coded the same or just > open code every where as btrfs_get_block_group() is a one-liner. > > Signed-off-by: Anand Jain <anand.jain@oracle.com> > --- > fs/btrfs/block-group.c | 8 ++++---- > fs/btrfs/block-group.h | 2 +- > fs/btrfs/free-space-cache.c | 4 ++-- > 3 files changed, 7 insertions(+), 7 deletions(-) > > diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c > index 31ca2cfb7e3e..8111f6750063 100644 > --- a/fs/btrfs/block-group.c > +++ b/fs/btrfs/block-group.c > @@ -118,12 +118,12 @@ u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags) > > void btrfs_get_block_group(struct btrfs_block_group *cache) > { > - atomic_inc(&cache->count); > + atomic_inc(&cache->bg_count); > } > > void btrfs_put_block_group(struct btrfs_block_group *cache) > { > - if (atomic_dec_and_test(&cache->count)) { > + if (atomic_dec_and_test(&cache->bg_count)) { > WARN_ON(cache->pinned > 0); > WARN_ON(cache->reserved > 0); > > @@ -1805,7 +1805,7 @@ static struct btrfs_block_group *btrfs_create_block_group_cache( > > cache->discard_index = BTRFS_DISCARD_INDEX_UNUSED; > > - atomic_set(&cache->count, 1); > + atomic_set(&cache->bg_count, 1); > spin_lock_init(&cache->lock); > init_rwsem(&cache->data_rwsem); > INIT_LIST_HEAD(&cache->list); > @@ -3379,7 +3379,7 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) > ASSERT(list_empty(&block_group->dirty_list)); > ASSERT(list_empty(&block_group->io_list)); > ASSERT(list_empty(&block_group->bg_list)); > - ASSERT(atomic_read(&block_group->count) == 1); > + ASSERT(atomic_read(&block_group->bg_count) == 1); > btrfs_put_block_group(block_group); > > spin_lock(&info->block_group_cache_lock); > diff --git a/fs/btrfs/block-group.h b/fs/btrfs/block-group.h > index b6ee70a039c7..f0ef8be08747 100644 > --- a/fs/btrfs/block-group.h > +++ b/fs/btrfs/block-group.h > @@ -115,7 +115,7 @@ struct btrfs_block_group { > struct list_head list; > > /* Usage count */ > - atomic_t count; > + atomic_t bg_count; > > /* > * List of struct btrfs_free_clusters for this block group. > diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c > index 1bf08c855edb..169b1117c1a3 100644 > --- a/fs/btrfs/free-space-cache.c > +++ b/fs/btrfs/free-space-cache.c > @@ -2925,7 +2925,7 @@ void btrfs_return_cluster_to_free_space( > spin_unlock(&cluster->lock); > return; > } > - atomic_inc(&block_group->count); > + atomic_inc(&block_group->bg_count); > spin_unlock(&cluster->lock); > > ctl = block_group->free_space_ctl; > @@ -3355,7 +3355,7 @@ int btrfs_find_space_cluster(struct btrfs_block_group *block_group, > list_del_init(&entry->list); > > if (!ret) { > - atomic_inc(&block_group->count); > + atomic_inc(&block_group->bg_count); > list_add_tail(&cluster->block_group_list, > &block_group->cluster_list); > cluster->block_group = block_group; > -- > 2.25.1 >
On 3.06.20 г. 13:10 ч., Anand Jain wrote: > The name 'count' is a very commonly used name. It is often difficult to > review the code to check if there is any leak. So rename it to > 'bg_count', which is unique enough. > > This patch also serves as a preparatory patch to either make sure > btrfs_get_block_group() is used instead of open coded the same or just > open code every where as btrfs_get_block_group() is a one-liner. > > Signed-off-by: Anand Jain <anand.jain@oracle.com> This introduces unnecessary churn, instead if all uses happen through the btrfs_get_block_group/put apis you can simply grep for those. All other uses are RO. Also bg_count is not very descriptive given this is really a refcount. Open-coding probably would be fine if this type is switched to refcount_t rather than an atomic_t.
On 3/6/20 6:19 pm, Filipe Manana wrote: > On Wed, Jun 3, 2020 at 11:13 AM Anand Jain <anand.jain@oracle.com> wrote: >> >> The name 'count' is a very commonly used name. It is often difficult to >> review the code to check if there is any leak. So rename it to >> 'bg_count', which is unique enough. > > Hum? Seriously? > > count to bg_count? > It's a member of the block group structure, adding a bg_ prefix adds > no value at all, we know the count is related to the block group. > I could somewhat understand if you named it 'refcount' instead. Oh right. Now, bg_count does not make sense to me as well. Something like bg_refcount is better so that it is easy to search where all its been used. IMO. Are we ok with btrfs_block_group::bg_refcount instead ? Thanks, Anand > Thanks.
On Wed, Jun 3, 2020 at 11:43 AM Anand Jain <anand.jain@oracle.com> wrote: > > > > On 3/6/20 6:19 pm, Filipe Manana wrote: > > On Wed, Jun 3, 2020 at 11:13 AM Anand Jain <anand.jain@oracle.com> wrote: > >> > >> The name 'count' is a very commonly used name. It is often difficult to > >> review the code to check if there is any leak. So rename it to > >> 'bg_count', which is unique enough. > > > > Hum? Seriously? > > > > count to bg_count? > > It's a member of the block group structure, adding a bg_ prefix adds > > no value at all, we know the count is related to the block group. > > I could somewhat understand if you named it 'refcount' instead. > > > Oh right. Now, bg_count does not make sense to me as well. > > Something like bg_refcount is better so that it is easy to search > where all its been used. IMO. > > Are we ok with btrfs_block_group::bg_refcount instead ? Why rename at all? It's pretty obvious it's a reference count. Count/counter is not an unusual name to use for a reference count, it even has get and put helpers... > > Thanks, > Anand > > > > Thanks.
> > Why rename at all? > It's pretty obvious it's a reference count. Count/counter is not an > unusual name to use for a reference count, it even has get and put > helpers... I understand as the member count is in btrfs_block_group so prefix bg_ looks redundant, but prefix bg_ is still useful. With out the prefix, the grep ">count" *.c didn't help straight away to verify where is block_group count used. At two places we missed using the get helper which the patch 3/3 fixed. It's ok to drop this rename patch if you find it's not inline with the rest of the codes.
On Wed, Jun 3, 2020 at 12:38 PM Anand Jain <anandsuveer@gmail.com> wrote: > > > > > > Why rename at all? > > It's pretty obvious it's a reference count. Count/counter is not an > > unusual name to use for a reference count, it even has get and put > > helpers... > > I understand as the member count is in btrfs_block_group so prefix bg_ > looks redundant, but prefix bg_ is still useful. With out the prefix, > the grep ">count" *.c didn't help straight away to verify where is Following that logic, we would add the bg_ prefix to most members of block_group... It has members named 'start', 'length', 'flags', etc. Those are generic too, many other structures have members with exactly those names... Don't take it wrong, but I don't find it useful at all. Thanks. > block_group count used. At two places we missed using the get helper > which the patch 3/3 fixed. > > It's ok to drop this rename patch if you find it's not inline with the > rest of the codes.
diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c index 31ca2cfb7e3e..8111f6750063 100644 --- a/fs/btrfs/block-group.c +++ b/fs/btrfs/block-group.c @@ -118,12 +118,12 @@ u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags) void btrfs_get_block_group(struct btrfs_block_group *cache) { - atomic_inc(&cache->count); + atomic_inc(&cache->bg_count); } void btrfs_put_block_group(struct btrfs_block_group *cache) { - if (atomic_dec_and_test(&cache->count)) { + if (atomic_dec_and_test(&cache->bg_count)) { WARN_ON(cache->pinned > 0); WARN_ON(cache->reserved > 0); @@ -1805,7 +1805,7 @@ static struct btrfs_block_group *btrfs_create_block_group_cache( cache->discard_index = BTRFS_DISCARD_INDEX_UNUSED; - atomic_set(&cache->count, 1); + atomic_set(&cache->bg_count, 1); spin_lock_init(&cache->lock); init_rwsem(&cache->data_rwsem); INIT_LIST_HEAD(&cache->list); @@ -3379,7 +3379,7 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) ASSERT(list_empty(&block_group->dirty_list)); ASSERT(list_empty(&block_group->io_list)); ASSERT(list_empty(&block_group->bg_list)); - ASSERT(atomic_read(&block_group->count) == 1); + ASSERT(atomic_read(&block_group->bg_count) == 1); btrfs_put_block_group(block_group); spin_lock(&info->block_group_cache_lock); diff --git a/fs/btrfs/block-group.h b/fs/btrfs/block-group.h index b6ee70a039c7..f0ef8be08747 100644 --- a/fs/btrfs/block-group.h +++ b/fs/btrfs/block-group.h @@ -115,7 +115,7 @@ struct btrfs_block_group { struct list_head list; /* Usage count */ - atomic_t count; + atomic_t bg_count; /* * List of struct btrfs_free_clusters for this block group. diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 1bf08c855edb..169b1117c1a3 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -2925,7 +2925,7 @@ void btrfs_return_cluster_to_free_space( spin_unlock(&cluster->lock); return; } - atomic_inc(&block_group->count); + atomic_inc(&block_group->bg_count); spin_unlock(&cluster->lock); ctl = block_group->free_space_ctl; @@ -3355,7 +3355,7 @@ int btrfs_find_space_cluster(struct btrfs_block_group *block_group, list_del_init(&entry->list); if (!ret) { - atomic_inc(&block_group->count); + atomic_inc(&block_group->bg_count); list_add_tail(&cluster->block_group_list, &block_group->cluster_list); cluster->block_group = block_group;
The name 'count' is a very commonly used name. It is often difficult to review the code to check if there is any leak. So rename it to 'bg_count', which is unique enough. This patch also serves as a preparatory patch to either make sure btrfs_get_block_group() is used instead of open coded the same or just open code every where as btrfs_get_block_group() is a one-liner. Signed-off-by: Anand Jain <anand.jain@oracle.com> --- fs/btrfs/block-group.c | 8 ++++---- fs/btrfs/block-group.h | 2 +- fs/btrfs/free-space-cache.c | 4 ++-- 3 files changed, 7 insertions(+), 7 deletions(-)