Hi Thorsten,
thanks for collecting this issue and providing relevant context!
On 26/09/2022 08:00, Thorsten Leemhuis wrote:
[adding Greg and Sasha to the recipients, to ensure they see this; CCing Kurt as well, to keep him in the loop]
On 22.09.22 15:44, Trond Myklebust wrote:
On Thu, 2022-09-22 at 13:42 +0200, Paul Menzel wrote:
Am 21.09.22 um 14:44 schrieb Trond Myklebust:
On Wed, 2022-09-21 at 13:42 +0200, Paul Menzel wrote:
Moving from Linux 5.10.113 to 5.15.69, starting Mozilla Thunderbird or Mozilla Firefox with the home on NFS, both programs get killed, and Linux 5.15.69 logs:
[ 3827.604396] BUG: unable to handle page fault for address: 000000001d473c07 [ 3827.611297] #PF: supervisor read access in kernel mode [ 3827.616452] #PF: error_code(0x0000) - not-present page [ 3827.621604] PGD 0 P4D 0 [ 3827.624152] Oops: 0000 [#1] SMP PTI [ 3827.627657] CPU: 0 PID: 2378 Comm: firefox Not tainted 5.15.69.mx64.435 #1 [ 3827.634551] Hardware name: Dell Inc. Precision Tower 3620/0MWYPT, BIOS 2.20.0 12/09/2021[…]
[ 3827.743328] Call Trace: [ 3827.745779] <TASK> [ 3827.747883] nfs_scan_commit+0x76/0xb0 [nfs] [ 3827.752167] __nfs_commit_inode+0x108/0x180 [nfs] [ 3827.756886] nfs_wb_all+0x59/0x110 [nfs] [ 3827.760822] nfs4_inode_return_delegation+0x58/0x90 [nfsv4] [ 3827.766413] nfs4_proc_remove+0x101/0x110 [nfsv4] [ 3827.771130] nfs_unlink+0xf5/0x2d0 [nfs] [ 3827.775065] vfs_unlink+0x10b/0x280 [ 3827.778563] do_unlinkat+0x19e/0x2c0 [ 3827.782158] __x64_sys_unlink+0x3e/0x60 [ 3827.786002] ? __x64_sys_readlink+0x1b/0x30 [ 3827.790192] do_syscall_64+0x40/0x90 [ 3827.793779] entry_SYSCALL_64_after_hwframe+0x61/0xcb
[…]
Does cherry-picking commit 6e176d47160c ("NFSv4: Fixes for nfs4_inode_return_delegation()") into 5.15.69 from the upstream kernel tree fix the problem?
8<--------------------------------------------------- From 6e176d47160cec8bcaa28d9aa06926d72d54237c Mon Sep 17 00:00:00 2001 From: Trond Myklebust trond.myklebust@hammerspace.com Date: Sun, 10 Oct 2021 10:58:12 +0200 Subject: [PATCH] NFSv4: Fixes for nfs4_inode_return_delegation()
[…]
Indeed with that commit, present since v5.16-rc1, we are unable to reproduce the issue, so it seems to be the fix. It looks like there are not a lot of 5.15 NFS users out there. ;-)
I believe this is a dependency that was introduced by the back port of commit e591b298d7ec ("NFS: Save some space in the inode") into 5.15.68. So the reason it wasn't seen is because the change is very recent.
Side note: I wonder if that is causing this problem from Kurt as well: https://lore.kernel.org/all/f6755107-b62c-a388-0ab5-0a6633bf9082@garloff.de/
Looks like it: After confirming that the 5.15.69 kernel worked again fine backing out those last three NFS commits, I reapplied them and cherry-picked commit 6e176d47160c as suggested. The kernel worked flawlessly thus far, so this seems to indeed be a requirement for e591b298d7ec not to cause harm.
FYI Greg and Sasha: please also consider pulling 6e176d47160c ("NFSv4: Fixes for nfs4_inode_return_delegation()") into that stable series.
Greg, I noticed you in the past few days added quite a few patches into the queue for the next 5.15.y release, but this one was not among them afaics. So just to be sure: is that still on your todo list or is more needed to get 6e176d47160c added in time for the next stable -rc?
So by all means, Greg, please put this in the stable queue unless the NFS wizards out there consider it safer to revert e591b298d7ec instead.
Thanks,