On Mon, Mar 23, 2015 at 5:55 PM, J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote:
> On Sun, Mar 22, 2015 at 03:25:11PM -0400, Trond Myklebust wrote:
>> On Wed, Mar 18, 2015 at 7:06 AM, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
>> > I'm running 4.0-rc against a local xfs server. I have a few commits
>> > reverted to get over the nfsd use after free bugs:
>> >
>> > b4019c0e219bb1301865f8b2efedb4773526ed91
>> > c69899a17ca4836230720e65493942d9582a0424
>> > 425c1d4e5b6d4bd700eb94ad8318bdb05431fdc7
>> > 6b447539aa9aaac0a0215f3e28a0839553210e7e
>> > 39071e6fff7d7e11a5993afd67240ef04a4d05a0
>> > 63f5f796af613898669b23ccfc091ec77de7591c
>> >
>> > but now I'm running into the following hang:
>> >
>> > generic/075 22s ...[ 360.216954] INFO: task fsx:17125 blocked for more than 120 seconds.
>> > [ 360.218762] Not tainted 4.0.0-rc3+ #151
>> > [ 360.220034] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
>> > [ 360.222310] fsx D ffff88007af37a18 0 17125 16912 0x00000000
>> > [ 360.224732] ffff88007af37a18 ffff88007a0e5d90 0000000000000002 ffffffff8249b500
>> > [ 360.227590] ffff88007a0e5550 ffff88007af37ad8 ffff88007af37fd8 0000000000000000
>> > [ 360.230711] 7fffffffffffffff 0000000000000002 ffffffff81e3b5f0 ffff88007af37a38
>> > [ 360.233162] Call Trace:
>> > [ 360.233953] [<ffffffff81e3b5f0>] ? yield+0x30/0x30
>> > [ 360.235368] [<ffffffff81e3b2e2>] schedule+0x32/0x80
>> > [ 360.236719] [<ffffffff81e3ef0c>] schedule_timeout+0x19c/0x210
>> > [ 360.238638] [<ffffffff8112bcd5>] ? ktime_get+0x105/0x140
>> > [ 360.240380] [<ffffffff8108c13e>] ? kvm_clock_read+0x1e/0x20
>> > [ 360.242259] [<ffffffff8108c149>] ? kvm_clock_get_cycles+0x9/0x10
>> > [ 360.244046] [<ffffffff8112bc75>] ? ktime_get+0xa5/0x140
>> > [ 360.245694] [<ffffffff8114d87e>] ? __delayacct_blkio_start+0x1e/0x30
>> > [ 360.247201] [<ffffffff81104ccd>] ? trace_hardirqs_on_caller+0x10d/0x1d0
>> > [ 360.248567] [<ffffffff81e3a638>] io_schedule_timeout+0xa8/0x130
>> > [ 360.249752] [<ffffffff810f9676>] ? prepare_to_wait_exclusive+0x56/0x90
>> > [ 360.251075] [<ffffffff81e3b620>] bit_wait_io+0x30/0x50
>> > [ 360.252222] [<ffffffff81e3b7ce>] __wait_on_bit_lock+0x6e/0xb0
>> > [ 360.253432] [<ffffffff81184302>] ? find_get_entries+0x22/0x160
>> > [ 360.254719] [<ffffffff81181a4c>] ? find_get_entry+0x8c/0xc0
>> > [ 360.255799] [<ffffffff811819c0>] ? find_get_pages_contig+0x1a0/0x1a0
>> > [ 360.257058] [<ffffffff81181c65>] __lock_page+0x95/0xa0
>> > [ 360.258175] [<ffffffff810f9ae0>] ? wake_atomic_t_function+0x30/0x30
>> > [ 360.259369] [<ffffffff81190646>] truncate_inode_pages_range+0x3c6/0x710
>> > [ 360.260627] [<ffffffff81190a10>] truncate_inode_pages+0x10/0x20
>> > [ 360.261849] [<ffffffff81190a66>] truncate_pagecache+0x46/0x70
>> > [ 360.262960] [<ffffffff81357fcf>] nfs_setattr_update_inode+0x12f/0x140
>> > [ 360.264192] [<ffffffff8137832b>] nfs4_proc_setattr+0xbb/0x100
>> > [ 360.265366] [<ffffffff81358162>] nfs_setattr+0xe2/0x200
>> > [ 360.266410] [<ffffffff811e8f90>] notify_change+0x170/0x3f0
>> > [ 360.267447] [<ffffffff811ca581>] do_truncate+0x61/0xa0
>> > [ 360.268476] [<ffffffff811ca924>] do_sys_ftruncate.constprop.18+0x104/0x160
>> > [ 360.269678] [<ffffffff81797b2e>] ? trace_hardirqs_on_thunk+0x3a/0x3f
>> > [ 360.270836] [<ffffffff811ca9a9>] SyS_ftruncate+0x9/0x10
>> > [ 360.271880] [<ffffffff81e40849>] system_call_fastpath+0x12/0x17
>> > [ 360.272940] 2 locks held by fsx/17125:
>> > [ 360.273719] #0: (sb_writers#9){.+.+.+}, at: [<ffffffff811ca8ef>] do_sys_ftruncate.constprop.18+0xcf/0x160
>> > [ 360.276008] #1: (&sb->s_type->i_mutex_key#11){+.+.+.}, at: [<ffffffff811ca573>] do_truncate+0x53/0xa0
>> > [ 385.757079] nfs: server 127.0.0.1 not responding, still trying
>> > [ 445.919739] nfs: server 127.0.0.1 OK
>> >
>> > from which the run eventually recovers after a few minutes
>> >
>>
>> Looks very similar to the problem I reported a few weeks ago. Just
>> disable splice reads by commenting out the RQ_SPLICE_OK line in
>> svc_process_common(), and Bob's your uncle...
>
> Argh, sorry, did I drop that report? I can't find it now.
>
> --b.
I was thinking about this thread:
http://lkml.kernel.org/r/1425237291.24845.13.camel@xxxxxxxxxxxxxxx
--
Trond Myklebust
Linux NFS client maintainer, PrimaryData
trond.myklebust@xxxxxxxxxxxxxxx
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
