| Message ID | 1362993658-30147-1-git-send-email-bo.li.liu@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Mar 11, 2013 at 05:20:58PM +0800, Liu Bo wrote: > Using spinning case instead of blocking will result in better concurrency > overall. Do you have numbers to support that? david -- 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
On Mon, Mar 11, 2013 at 06:26:40PM +0100, David Sterba wrote: > On Mon, Mar 11, 2013 at 05:20:58PM +0800, Liu Bo wrote: > > Using spinning case instead of blocking will result in better concurrency > > overall. > > Do you have numbers to support that? > Sorry, I don't, just judging from what leave_spinning is desiged for and the similar usecases, like insert_reserved_file_extents(), which is also involved in endio write worker. thanks, liubo -- 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
On Mon, Mar 11, 2013 at 07:44:04PM -0600, Liu Bo wrote: > On Mon, Mar 11, 2013 at 06:26:40PM +0100, David Sterba wrote: > > On Mon, Mar 11, 2013 at 05:20:58PM +0800, Liu Bo wrote: > > > Using spinning case instead of blocking will result in better concurrency > > > overall. > > > > Do you have numbers to support that? > > > > Sorry, I don't, just judging from what leave_spinning is desiged for and the > similar usecases, like insert_reserved_file_extents(), which is also involved in > endio write worker. Leaving the path spinning is always faster. As long as you're sure the code inside doesn't schedule it is the right choice. -chris -- 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
On Thu, Mar 14, 2013 at 02:38:29PM -0400, Chris Mason wrote: > On Mon, Mar 11, 2013 at 07:44:04PM -0600, Liu Bo wrote: > > On Mon, Mar 11, 2013 at 06:26:40PM +0100, David Sterba wrote: > > > On Mon, Mar 11, 2013 at 05:20:58PM +0800, Liu Bo wrote: > > > > Using spinning case instead of blocking will result in better concurrency > > > > overall. > > > > > > Do you have numbers to support that? > > > > > > > Sorry, I don't, just judging from what leave_spinning is desiged for and the > > similar usecases, like insert_reserved_file_extents(), which is also involved in > > endio write worker. > > Leaving the path spinning is always faster. As long as you're sure the > code inside doesn't schedule it is the right choice. > Yeah, I've taken care of schedule thing :) thanks, liubo > -- 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/inode.c b/fs/btrfs/inode.c index 13ab4de..1f268888 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2312,6 +2312,7 @@ again: key.type = BTRFS_EXTENT_DATA_KEY; key.offset = start; + path->leave_spinning = 1; if (merge) { struct btrfs_file_extent_item *fi; u64 extent_len; @@ -2368,6 +2369,7 @@ again: btrfs_mark_buffer_dirty(leaf); inode_add_bytes(inode, len); + btrfs_release_path(path); ret = btrfs_inc_extent_ref(trans, root, new->bytenr, new->disk_len, 0, @@ -2381,6 +2383,7 @@ again: ret = 1; out_free_path: btrfs_release_path(path); + path->leave_spinning = 0; btrfs_end_transaction(trans, root); out_unlock: unlock_extent_cached(&BTRFS_I(inode)->io_tree, lock_start, lock_end,
Using spinning case instead of blocking will result in better concurrency overall. Signed-off-by: Liu Bo <bo.li.liu@oracle.com> --- fs/btrfs/inode.c | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)