On Thu, 16 Jun 2022 at 01:56, NeilBrown <neilb@xxxxxxx> wrote:
>
> On Wed, 15 Jun 2022, Daire Byrne wrote:
> ..
> > However, it is at this point that I started to experience some
> > stability issues with the re-export server that are not present with
> > the vanilla unpatched v5.19-rc2 kernel. In particular the knfsd
> > threads start to lock up with stack traces like this:
> >
> > [ 1234.460696] INFO: task nfsd:5514 blocked for more than 123 seconds.
> > [ 1234.461481] Tainted: G W E 5.19.0-1.dneg.x86_64 #1
> > [ 1234.462289] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
> > disables this message.
> > [ 1234.463227] task:nfsd state:D stack: 0 pid: 5514
> > ppid: 2 flags:0x00004000
> > [ 1234.464212] Call Trace:
> > [ 1234.464677] <TASK>
> > [ 1234.465104] __schedule+0x2a9/0x8a0
> > [ 1234.465663] schedule+0x55/0xc0
> > [ 1234.466183] ? nfs_lookup_revalidate_dentry+0x3a0/0x3a0 [nfs]
> > [ 1234.466995] __nfs_lookup_revalidate+0xdf/0x120 [nfs]
>
> I can see the cause of this - I forget a wakeup. This patch should fix
> it, though I hope to find a better solution.
>
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index 54c2c7adcd56..072130d000c4 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -2483,17 +2483,16 @@ int nfs_unlink(struct inode *dir, struct dentry *dentry)
> if (!(dentry->d_flags & DCACHE_PAR_UPDATE)) {
> /* Must have exclusive lock on parent */
> did_set_par_update = true;
> + lock_acquire_exclusive(&dentry->d_update_map, 0,
> + 0, NULL, _THIS_IP_);
> dentry->d_flags |= DCACHE_PAR_UPDATE;
> }
>
> spin_unlock(&dentry->d_lock);
> error = nfs_safe_remove(dentry);
> nfs_dentry_remove_handle_error(dir, dentry, error);
> - if (did_set_par_update) {
> - spin_lock(&dentry->d_lock);
> - dentry->d_flags &= ~DCACHE_PAR_UPDATE;
> - spin_unlock(&dentry->d_lock);
> - }
> + if (did_set_par_update)
> + d_unlock_update(dentry);
> out:
> trace_nfs_unlink_exit(dir, dentry, error);
> return error;
>
> >
> > So all in all, the performance improvements in the knfsd re-export
> > case is looking great and we have real world use cases that this helps
> > with (batch processing workloads with latencies >10ms). If we can
> > figure out the hanging knfsd threads, then I can test it more heavily.
>
> Hopefully the above patch will allow the more heavy testing to continue.
> In any case, thanks a lot for the testing so far,
Patch applied but unfortunately I'm still getting the same trace, but
this time I also captured a preceding stack for a hung process local
to the reexport server - I wonder if it's happening somewhere in the
VFS changes rather than nfsd which then exports the path?
[ 373.930506] INFO: task XXXX:5072 blocked for more than 122 seconds.
[ 373.931410] Tainted: G W E 5.19.0-3.dneg.x86_64 #1
[ 373.932313] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
[ 373.933442] task:XXXX state:D stack: 0 pid: 5072 ppid:
1 flags:0x00000000
[ 373.934639] Call Trace:
[ 373.935007] <TASK>
[ 373.935306] __schedule+0x2a9/0x8a0
[ 373.935844] schedule+0x55/0xc0
[ 373.936294] ? nfs_lookup_revalidate_dentry+0x3a0/0x3a0 [nfs]
[ 373.937137] __nfs_lookup_revalidate+0xdf/0x120 [nfs]
[ 373.937875] ? put_prev_task_stop+0x170/0x170
[ 373.938525] nfs_lookup_revalidate+0x15/0x20 [nfs]
[ 373.939226] lookup_fast+0xda/0x150
[ 373.939756] path_openat+0x12a/0x1090
[ 373.940293] ? __filemap_fdatawrite_range+0x54/0x70
[ 373.941100] do_filp_open+0xb2/0x120
[ 373.941635] ? hashlen_string+0xd0/0xd0
[ 373.942190] ? _raw_spin_unlock+0xe/0x30
[ 373.942766] do_sys_openat2+0x245/0x320
[ 373.943305] do_sys_open+0x46/0x80
[ 373.943839] __x64_sys_open+0x21/0x30
[ 373.944428] do_syscall_64+0x3b/0x90
[ 373.944979] entry_SYSCALL_64_after_hwframe+0x46/0xb0
[ 373.945688] RIP: 0033:0x7fcd80ceeeb0
[ 373.946226] RSP: 002b:00007fff90fd8298 EFLAGS: 00000246 ORIG_RAX:
0000000000000002
[ 373.947330] RAX: ffffffffffffffda RBX: 00007fcd81d6e981 RCX: 00007fcd80ceeeb0
[ 373.947333] RDX: 00000000000001b6 RSI: 0000000000000000 RDI: 00007fff90fd8360
[ 373.947334] RBP: 00007fff90fd82f0 R08: 00007fcd81d6e986 R09: 0000000000000000
[ 373.947335] R10: 0000000000000024 R11: 0000000000000246 R12: 0000000000cd6110
[ 373.947337] R13: 0000000000000008 R14: 00007fff90fd8360 R15: 00007fff90fdb580
[ 373.947339] </TASK>
[ 373.947421] INFO: task nfsd:5696 blocked for more than 122 seconds.
[ 373.947423] Tainted: G W E 5.19.0-3.dneg.x86_64 #1
[ 373.947424] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
[ 373.947425] task:nfsd state:D stack: 0 pid: 5696
ppid: 2 flags:0x00004000
[ 373.947428] Call Trace:
[ 373.947429] <TASK>
[ 373.947430] __schedule+0x2a9/0x8a0
[ 373.947434] schedule+0x55/0xc0
[ 373.947436] ? nfs_lookup_revalidate_dentry+0x3a0/0x3a0 [nfs]
[ 373.947451] __nfs_lookup_revalidate+0xdf/0x120 [nfs]
[ 373.947464] ? put_prev_task_stop+0x170/0x170
[ 373.947466] nfs_lookup_revalidate+0x15/0x20 [nfs]
[ 373.947478] lookup_dcache+0x5a/0x80
[ 373.947481] lookup_one_unlocked+0x59/0xa0
[ 373.947484] lookup_one_len_unlocked+0x1d/0x20
[ 373.947487] nfsd_lookup_dentry+0x190/0x470 [nfsd]
[ 373.947509] nfsd_lookup+0x88/0x1b0 [nfsd]
[ 373.947522] nfsd3_proc_lookup+0xb4/0x100 [nfsd]
[ 373.947537] nfsd_dispatch+0x161/0x290 [nfsd]
[ 373.947551] svc_process_common+0x48a/0x620 [sunrpc]
[ 373.947589] ? nfsd_svc+0x330/0x330 [nfsd]
[ 373.947602] ? nfsd_shutdown_threads+0xa0/0xa0 [nfsd]
[ 373.947621] svc_process+0xbc/0xf0 [sunrpc]
[ 373.951088] nfsd+0xda/0x190 [nfsd]
[ 373.951136] kthread+0xf0/0x120
[ 373.951138] ? kthread_complete_and_exit+0x20/0x20
[ 373.951140] ret_from_fork+0x22/0x30
[ 373.951149] </TASK>
I double checked that the patch had been applied and I hadn't made a
mistake with installation.
I could perhaps try running with just the VFS patches to see if I can
still reproduce the "local" VFS hang without the nfsd patches? Your
previous VFS only patchset was stable for me.
Daire
