[Top] [All Lists]

Re: [PATCH] [RFC] xfs: lookaside cache for xfs_buf_find

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: [PATCH] [RFC] xfs: lookaside cache for xfs_buf_find
From: Mark Tinguely <tinguely@xxxxxxx>
Date: Mon, 23 Sep 2013 09:17:35 -0500
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <1378690396-15792-1-git-send-email-david@xxxxxxxxxxxxx>
References: <1378690396-15792-1-git-send-email-david@xxxxxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:9.0) Gecko/20120122 Thunderbird/9.0
On 09/08/13 20:33, Dave Chinner wrote:
From: Dave Chinner<dchinner@xxxxxxxxxx>

CPU overhead of buffer lookups dominate most metadata intensive
workloads. The thing is, most such workloads are hitting a
relatively small number of buffers repeatedly, and so caching
recently hit buffers is a good idea.


I think this needs more testing.

I get the following panic in a loop test after a few (3-8) iterations:

 while true
   tar zxpf xfs.tar
   cd xfs
   make modules
   cd ..
   rm -r xfs

BUG: unable to handle kernel paging request at ffff880831c1d218
IP: [<ffffffffa01886c8>] _xfs_buf_find_lookaside+0x98/0xb0 [xfs]
PGD 1c5d067 PUD 85ffe0067 PMD 85fe51067 PTE 8000000831c1d060
Modules linked in: xfs(O) e1000e exportfs libcrc32c ext3 jbd [last unloaded: xfs
CPU: 0 PID: 23423 Comm: tar Tainted: G           O 3.11.0-rc1+ #3
task: ffff880837f087a0 ti: ffff880831c46000 task.ti: ffff880831c46000
RIP: 0010:[<ffffffffa01886c8>] [<ffffffffa01886c8>] _xfs_buf_find_lookaside+0x9
8/0xb0 [xfs]
RSP: 0018:ffff880831c47918  EFLAGS: 00010286
RAX: ffff880831c1d200 RBX: ffff8808372e0000 RCX: 0000000000000003
 RDX: 0000000000000011 RSI: 00000000000009c0 RDI: ffff8808372e0000
 RBP: ffff880831c47938 R08: ffff8808372e0000 R09: ffff8808376e8d80
 R10: 0000000000000010 R11: 00000000000009c0 R12: 00000000000009c0
 R13: 0000000000000010 R14: 0000000000000001 R15: 00000000000009c0
FS: 00007fa4bc51f700(0000) GS:ffff88085bc00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: ffff880831c1d218 CR3: 000000082ed00000 CR4: 00000000000007f0
  ffff880831c47938 ffff880831c47aa8 0000000000000010 ffff880834ab7900
  ffff880831c479b8 ffffffffa018a679 ffff8808372e00c0 ffff88082eed01a0
  0000000000000029 ffff8808372e01f0 0000000000000000 000200015bfe1c68
 Call Trace:
  [<ffffffffa018a679>] _xfs_buf_find+0x159/0x520 [xfs]
  [<ffffffffa018aea0>] xfs_buf_get_map+0x30/0x130 [xfs]
  [<ffffffffa018afc6>] xfs_buf_read_map+0x26/0xa0 [xfs]
  [<ffffffffa01fbf5d>] xfs_trans_read_buf_map+0x16d/0x4c0 [xfs]
  [<ffffffffa01e784c>] xfs_imap_to_bp+0x6c/0x120 [xfs]
  [<ffffffffa01e7975>] xfs_iread+0x75/0x2f0 [xfs]
  [<ffffffff8114eafb>] ? inode_init_always+0xfb/0x1c0
  [<ffffffffa019311a>] xfs_iget_cache_miss+0x5a/0x1e0 [xfs]
  [<ffffffffa01933db>] xfs_iget+0x13b/0x1c0 [xfs]
  [<ffffffffa01dfaad>] xfs_ialloc+0xbd/0x860 [xfs]
  [<ffffffffa01e02e7>] xfs_dir_ialloc+0x97/0x2e0 [xfs]
  [<ffffffffa01a2308>] ? xfs_trans_reserve+0x308/0x310 [xfs]

I got the same panic running xfstest 319 with the patch at:
once it hung on a xfs_buf lock before the panic.

And these are the only tests that I threw at this patch.


<Prev in Thread] Current Thread [Next in Thread>