| Message ID | 1391720873-4961-1-git-send-email-jbacik@fb.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
On 02/06/2014 04:07 PM, Josef Bacik wrote: > While trying to reproduce a delayed ref problem I noticed the box kept falling > over using all 80gb of my ram with btrfs_inode's and btrfs_delayed_node's. > Turns out this is because we only throttle delayed inode updates in > btrfs_dirty_inode, which doesn't actually get called that often, especially when > all you are doing is creating a bunch of files. So balance delayed inode > updates everytime we create a new inode. With this patch we no longer use up > all of our ram with delayed inode updates. Thanks, > > Signed-off-by: Josef Bacik <jbacik@fb.com> > --- > fs/btrfs/extent-tree.c | 1 + > fs/btrfs/inode.c | 4 ++++ > 2 files changed, 5 insertions(+) > > diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c > index 9c9ecc9..1d9d9ce 100644 > --- a/fs/btrfs/extent-tree.c > +++ b/fs/btrfs/extent-tree.c > @@ -2660,6 +2660,7 @@ int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans, > atomic_read(&trans->transaction->delayed_refs.num_entries); > u64 avg_runtime; > > + return 1; > smp_mb(); Do git add for the file I want, do git commit -a anyway. Josef -- 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/extent-tree.c b/fs/btrfs/extent-tree.c index 9c9ecc9..1d9d9ce 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2660,6 +2660,7 @@ int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans, atomic_read(&trans->transaction->delayed_refs.num_entries); u64 avg_runtime; + return 1; smp_mb(); avg_runtime = fs_info->avg_delayed_ref_runtime; if (num_entries * avg_runtime >= NSEC_PER_SEC) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1af34d0..d44ff9f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -5795,6 +5795,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, } out_unlock: btrfs_end_transaction(trans, root); + btrfs_balance_delayed_items(root); btrfs_btree_balance_dirty(root); if (drop_inode) { inode_dec_link_count(inode); @@ -5868,6 +5869,7 @@ out_unlock: inode_dec_link_count(inode); iput(inode); } + btrfs_balance_delayed_items(root); btrfs_btree_balance_dirty(root); return err; } @@ -5926,6 +5928,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, } btrfs_end_transaction(trans, root); + btrfs_balance_delayed_items(root); fail: if (drop_inode) { inode_dec_link_count(inode); @@ -5992,6 +5995,7 @@ out_fail: btrfs_end_transaction(trans, root); if (drop_on_err) iput(inode); + btrfs_balance_delayed_items(root); btrfs_btree_balance_dirty(root); return err; }
While trying to reproduce a delayed ref problem I noticed the box kept falling over using all 80gb of my ram with btrfs_inode's and btrfs_delayed_node's. Turns out this is because we only throttle delayed inode updates in btrfs_dirty_inode, which doesn't actually get called that often, especially when all you are doing is creating a bunch of files. So balance delayed inode updates everytime we create a new inode. With this patch we no longer use up all of our ram with delayed inode updates. Thanks, Signed-off-by: Josef Bacik <jbacik@fb.com> --- fs/btrfs/extent-tree.c | 1 + fs/btrfs/inode.c | 4 ++++ 2 files changed, 5 insertions(+)