Re: [PATCH] btrfs: add block plug for .writepages

From: Shaohua Li
Date: Tue Nov 01 2011 - 03:14:59 EST


2011/11/1 NamJae Jeon <linkinjeon@xxxxxxxxx>:
> 2011/11/1 Chris Mason <chris.mason@xxxxxxxxxx>:
>> On Tue, Nov 01, 2011 at 12:45:07AM +0900, Namjae Jeon wrote:
>>> Add block plug for btrfs .writepages. Block plug is helpful to reduce block lock contention.
>>>
>>> Signed-off-by: Namjae Jeon <linkinjeon@xxxxxxxxx>
>>> ---
>>>  fs/btrfs/extent_io.c |    4 ++++
>>>  1 files changed, 4 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
>>> index d418164..8f2d6bd 100644
>>> --- a/fs/btrfs/extent_io.c
>>> +++ b/fs/btrfs/extent_io.c
>>> @@ -2388,6 +2388,7 @@ static int extent_write_cache_pages(struct extent_io_tree *tree,
>>>       pgoff_t end;            /* Inclusive */
>>>       int scanned = 0;
>>>       int tag;
>>> +     struct blk_plug plug;
>>>
>>>       pagevec_init(&pvec, 0);
>>>       if (wbc->range_cyclic) {
>>> @@ -2405,6 +2406,8 @@ static int extent_write_cache_pages(struct extent_io_tree *tree,
>>>  retry:
>>>       if (wbc->sync_mode == WB_SYNC_ALL)
>>>               tag_pages_for_writeback(mapping, index, end);
>>> +
>>> +     blk_start_plug(&plug);
>>>       while (!done && !nr_to_write_done && (index <= end) &&
>>>              (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
>>>                       min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
>>> @@ -2468,6 +2471,7 @@ retry:
>>>               pagevec_release(&pvec);
>>>               cond_resched();
>>>       }
>>> +     blk_finish_plug(&plug);
>>>       if (!scanned && !done) {
>>>               /*
>>>                * We hit the last page and there is more work to be done: wrap
>>
>> Thanks for the patch, but we actually do this at a lower level in btrfs.
>> The actual IO can be done to multiple block devices, so we try to plug
>> for each device individually.
>>
>> -chris
>>
>>
> Hi. Chris.
>
> I got your point. pending bio is operated by worker_loop thread.
> But Is block plug needed in extent_readpages ?
.readpages are called in readahead, where we already have block
plug.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/