On Wed, 2008-10-15 at 16:41 -0400, Chris Mason wrote: > On Fri, 2008-10-10 at 23:32 +0530, Aneesh Kumar K.V wrote: > > The range_cyclic writeback mode use the address_space > > writeback_index as the start index for writeback. With > > delayed allocation we were updating writeback_index > > wrongly resulting in highly fragmented file. Number of > > extents reduced from 4000 to 27 for a 3GB file with > > the below patch. > > > > I tested the ext4 patch queue from today on top of 2.6.27, and this > includes Aneesh's latest patches. > > Things are going at disk speed for streaming writes, with the number of > extents generated for a 32GB file down to 27. So, this is definitely an > improvement for ext4. Just FYI, I ran this with compilebench -i 20 --makej and my log is full of these: ext4_da_writepages: jbd2_start: 1024 pages, ino 520417; err -30 Pid: 4072, comm: pdflush Not tainted 2.6.27 #2 Call Trace: [<ffffffffa0048493>] ext4_da_writepages+0x171/0x2d3 [ext4] [<ffffffff802336be>] ? pick_next_task_fair+0x80/0x91 [<ffffffff80228fa8>] ? source_load+0x2a/0x58 [<ffffffff8038e499>] ? __next_cpu+0x19/0x26 [<ffffffff8026748f>] do_writepages+0x28/0x37 [<ffffffff802a6b39>] __writeback_single_inode+0x14f/0x26d [<ffffffff802a6fb7>] generic_sync_sb_inodes+0x1c1/0x2a2 [<ffffffff802a70a1>] sync_sb_inodes+0x9/0xb [<ffffffff802a73dc>] writeback_inodes+0x64/0xad [<ffffffff802675db>] wb_kupdate+0x9a/0x10c [<ffffffff80267fd1>] ? pdflush+0x0/0x1e9 [<ffffffff80267fd1>] ? pdflush+0x0/0x1e9 [<ffffffff8026810e>] pdflush+0x13d/0x1e9 [<ffffffff80267541>] ? wb_kupdate+0x0/0x10c [<ffffffff80248222>] kthread+0x49/0x77 [<ffffffff8020c5e9>] child_rip+0xa/0x11 [<ffffffff802481d9>] ? kthread+0x0/0x77 [<ffffffff8020c5df>] ? child_rip+0x0/0x11 -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
