Re: linux 5.13 kernel regression for NFS server

[Chuck Lever III]

Hi Justin-

> On Jun 29, 2021, at 6:39 AM, Justin Piszcz <jpiszcz@xxxxxxxxxxxxxxx> wrote:
> 
> Hello,
> 
> ISSUE: Can no longer mount NFS server(5.13) share.
> WORKAROUND: Reverting back to the 5.13-rc7 kernel for the NFS server, all
> clients can then mount the share.
> 
> Kernel 5.13 vs. 5.13-rc7 for x86_64
> 
> With Kernel 5.13-rc7:
> 
> ## mount is successful (from multiple clients, phone, tv, other Linux
> client)
> # mount 192.168.1.2:/r1 /r1
> #
> 
> With Kernel 5.13 (hangs from multiple clients, phone, tv, other Linux
> client)
> 
> ## mount hangs:
> # mount 192.168.1.2:/r1 /r1
> 
> (strace output where it is hanging)
> ...
> clone(child_stack=NULL,
> flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD,
> child_tidptr=0x7f2285e303d0) = 1741
> wait4(1741, 0x7ffeaab6ce08, 0, NULL)    = ? ERESTARTSYS (To be restarted if
> SA_RESTART is set)
> --- SIGWINCH {si_signo=SIGWINCH, si_code=SI_KERNEL} ---
> wait4(1741,
> 
> No user changes in the kernel .config other than the comment for the version
> text:
> 
> --- config-20210628.1624867695  2021-06-28 04:08:15.152781940 -0400
> +++ config-20210628.1624867962  2021-06-28 04:12:42.591981706 -0400
> @@ -1,6 +1,6 @@
> #
> # Automatically generated file; DO NOT EDIT.
> -# Linux/x86 5.13.0-rc7 Kernel Configuration
> +# Linux/x86 5.13.0 Kernel Configuration
> #
> CONFIG_CC_VERSION_TEXT="gcc (Debian 10.2.1-6) 10.2.1 20210110"
> CONFIG_CC_IS_GCC=y

It's likely this regression is due to a last minute change to
alloc_pages_bulk_array() done just before v5.13. See:

https://lore.kernel.org/linux-nfs/20210629081432.GE3840@xxxxxxxxxxxxxxxxxxx/T/#t

for details.

--
Chuck Lever