Re: [PATCH, RFC] vm: Add an tuning knob for vm.max_writeback_pages

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Aug 31 2009, Jens Axboe wrote:
> > I traced the calls to ext4_da_writepages() using ftrace, and found this:
> > 
> >       flush-8:16-1829  [001]    23.416351: ext4_da_writepages: dev sdb ino 12 nr_t_write 32759 pages_skipped 0 range_start 0 range_end 0 nonblocking 0 for_kupdate 0 for_reclaim 0 for_writepages 1 range_cyclic 1
> >       flush-8:16-1829  [000]    25.939354: ext4_da_writepages: dev sdb ino 12 nr_t_write 32768 pages_skipped 0 range_start 0 range_end 0 nonblocking 0 for_kupdate 0 for_reclaim 0 for_writepages 1 range_cyclic 1
> >       flush-8:16-1829  [000]    25.939486: ext4_da_writepages: dev sdb ino 13 nr_t_write 32759 pages_skipped 0 range_start 134180864 range_end 9223372036854775807 nonblocking 0 for_kupdate 0 for_reclaim 0 for_writepages 1 range_cyclic 1
> >       flush-8:16-1829  [000]    27.055687: ext4_da_writepages: dev sdb ino 12 nr_t_write 32768 pages_skipped 0 range_start 0 range_end 0 nonblocking 0 for_kupdate 0 for_reclaim 0 for_writepages 1 range_cyclic 1
> >       flush-8:16-1829  [000]    27.055691: ext4_da_writepages: dev sdb ino 13 nr_t_write 32768 pages_skipped 0 range_start 0 range_end 0 nonblocking 0 for_kupdate 0 for_reclaim 0 for_writepages 1 range_cyclic 1
> >       flush-8:16-1829  [000]    27.878708: ext4_da_writepages: dev sdb ino 13 nr_t_write 32768 pages_skipped 0 range_start 0 range_end 0 nonblocking 0 for_kupdate 0 for_reclaim 0 for_writepages 1 range_cyclic 1
> > 
> > The *first* time the per-bdi code called writepages on the second file
> > (test2, inode #13), range_start was 134180864 (which, curiously
> > enough, is 4096*32759, which was the value of nr_to_write passed to
> > ext4_da_writepages).  Given that the inode only had 32768 pages, the
> > fact that apparently *some* codepath called ext4_da_writepages
> > starting at logical block 32759, with nr_to_write set to 32759, seems
> > very curious indeed.  That doesn't seem right at all.  It's late, so I
> > won't try to trace it down now; plus which it's your code so I figure
> > you can probably figure it out faster....
> 
> Interesting, needs checking up on. I've prepared a v14 patchset today,
> perhaps (if you have time), you can see if it reproduces there? I'm
> running some performance tests today, but will make a note to look into
> this after that.

It's because ext4 writepages sets ->range_start and wb_writeback() is
range cyclic, then the next iteration will have the previous end point
as the starting point. Looks like we need to clear ->range_start in
wb_writeback(), the better place is probably to do that in
fs/fs-writeback.c:generic_sync_wb_inodes() right after the
writeback_single_inode() call. This, btw, should be no different than
the current code, weird/correct or not :-)

-- 
Jens Axboe

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]
  Powered by Linux