Re: [GIT PULL] Please Pull NFS Client Updates for Linux 7.0

[Mike Snitzer <snitzer@kernel.org>]

On Wed, Feb 18, 2026 at 11:24:42AM -0500, Anna Schumaker wrote:
> 
> 
> On Mon, Feb 16, 2026, at 10:31 AM, Niklas Cassel wrote:
> > On Thu, Feb 12, 2026 at 05:06:25PM -0500, Anna Schumaker wrote:
> >> Hi Linus,
> >> 
> >> The following changes since commit 24d479d26b25bce5faea3ddd9fa8f3a6c3129ea7:
> >> 
> >>   Linux 6.19-rc6 (2026-01-18 15:42:45 -0800)
> >> 
> >> are available in the Git repository at:
> >> 
> >>   git://git.linux-nfs.org/projects/anna/linux-nfs.git tags/nfs-for-7.0-1
> >> 
> >> for you to fetch changes up to dd2fdc3504592d85e549c523b054898a036a6afe:
> >> 
> >>   SUNRPC: fix gss_auth kref leak in gss_alloc_msg error path (2026-02-09 16:39:50 -0500)
> >> 
> >> ----------------------------------------------------------------
> >> NFS Client Updates for Linux 7.0
> >> 
> >> New Features:
> >>  * Use an LRU list for returning unused delegations
> >>  * Introduce a KConfig option to disable NFS v4.0 and make NFS v4.1 the default
> >> 
> >> Bugfixes:
> >>  * NFS/localio: Handle short writes by retrying
> >>  * NFS/localio: prevent direct reclaim recursion into NFS via nfs_writepages
> >>  * NFS/localio: use GFP_NOIO and non-memreclaim workqueue in nfs_local_commit
> >>  * NFS/localio: remove -EAGAIN handling in nfs_local_doio()
> >>  * pNFS: fix a missing wake up while waiting on NFS_LAYOUT_DRAIN
> >>  * fs/nfs: Fix a readdir slow-start regression
> >>  * SUNRPC: fix gss_auth kref leak in gss_alloc_msg error path
> >> 
> >> Other Cleanups and Improvements:
> >>   * A few other NFS/localio cleanups
> >>   * Various other delegation handling cleanups from Christoph
> >>   * Unify security_inode_listsecurity() calls
> >>   * Improvements to NFSv4 lease handling
> >>   * Clean up SUNRPC *_debug fields when CONFIG_SUNRPC_DEBUG is not set
> >> 
> >> Thanks,
> >> Anna
> >
> > Hello Anna,
> >
> > This seems to break my setup:
> >
> > [  103.402033] VFS: Unable to mount root fs via NFS.
> > [  103.402476] devtmpfs: mounted
> > [  103.406171] Freeing unused kernel memory: 4352K
> > [  103.406633] Run /sbin/init as init process
> > [  103.406993]   with arguments:
> > [  103.407253]     /sbin/init
> > [  103.407491]   with environment:
> > [  103.407767]     HOME=/
> > [  103.407976]     TERM=linux
> > [  103.408228] Run /etc/init as init process
> > [  103.408580]   with arguments:
> > [  103.408841]     /etc/init
> > [  103.409071]   with environment:
> > [  103.409346]     HOME=/
> > [  103.409553]     TERM=linux
> > [  103.409802] Run /bin/init as init process
> > [  103.410153]   with arguments:
> > [  103.410414]     /bin/init
> > [  103.410644]   with environment:
> > [  103.410920]     HOME=/
> > [  103.411128]     TERM=linux
> > [  103.411367] Run /bin/sh as init process
> > [  103.411703]   with arguments:
> > [  103.411963]     /bin/sh
> > [  103.412179]   with environment:
> > [  103.412454]     HOME=/
> > [  103.412662]     TERM=linux
> > [  103.412904] Kernel panic - not syncing: No working init found.  Try 
> > passing init= option to kernel. See Linux 
> > Documentation/admin-guide/init.rst for guidance.
> >
> >
> > Doing a git bisect results in:
> >
> > commit 4e0269352534715915aae3fb84dd4d3bb3ff3738
> > Author: Anna Schumaker <anna.schumaker@oracle.com>
> > Date:   Fri Nov 21 14:53:50 2025 -0500
> >
> >     NFS: Add a way to disable NFS v4.0 via KConfig
> >
> >     I introduce NFS4_MIN_MINOR_VERSION as a parallel to
> >     NFS4_MAX_MINOR_VERSION to check if NFS v4.0 has been compiled in and
> >     return an appropriate error if not.
> >
> >
> >
> > Before commit above my config has:
> > CONFIG_NFS_FS=y
> > CONFIG_NFS_V4=y
> > CONFIG_NFS_V4_1=y
> > CONFIG_NFS_V4_2=y
> >
> > after commit above (with make olddefconfig), my config has:
> > CONFIG_NFS_FS=y
> > CONFIG_NFS_V4=y
> > # CONFIG_NFS_V4_0 is not set
> > CONFIG_NFS_V4_1=y
> > CONFIG_NFS_V4_2=y
> >
> >
> >
> > My kernel command line has:
> > nfsroot=192.168.1.10:/srv/nfs/rootfs,nfsvers=4
> >
> >
> > I notice that I am probably stupid who has the above, as I assumed that
> > it meant use the best NFS 4.x version supported by the kernel.
> >
> >
> > If I change it to:
> > nfsroot=192.168.1.10:/srv/nfs/rootfs,nfsvers=4.2
> >
> > The config:
> > CONFIG_NFS_FS=y
> > CONFIG_NFS_V4=y
> > # CONFIG_NFS_V4_0 is not set
> > CONFIG_NFS_V4_1=y
> > CONFIG_NFS_V4_2=y
> >
> > successfully mounts my rootfs.
> >
> >
> > So AFAICT, it seems that nfsvers=4 actually means nfsvers=4.0
> >
> > I am probably not the only person to who has "nfsvers=4" on the kernel
> > command line.
> >
> >
> >
> > Possible solutions I can see:
> >
> > If the intention is for commit 4e0269352534 ("NFS: Add a way to disable
> > NFS v4.0 via KConfig") to allow people to disable NFS 4.0, we could
> > change the new "config NFS_V4_0" to be either "default y".
> 
> This is what I intended, actually. I don't know how the missing "default y"
> slipped through the cracks for this long. Thanks for pointing it out, I'll
> get this fixed up!
> 
> Anna
> 
> >
> > or
> >
> > Perhaps we could modify "nfsvers=4" to actually mean "use whatever highest
> > NFS 4.x version supported by the kernel", rather than use NFS 4.0 and only
> > 4.0.

Another related problem I just encountered, when doing a bisect that
pivoted from 7.0-rc1 back down to 6.19, is that "make olddefconfig"
using the 7.0-rc1's .config as starting point against 6.19 kernel
results in:

# CONFIG_NFS_V4_1 is not set
# CONFIG_NFS_V4_2 is not set

CONFIG_PNFS_FLEXFILE_LAYOUT=m doesn't exist, etc.  So this change
seriously breaks NFS bisect-ability.

Feels like the new CONFIG_NFS_V4_0 control should be standalone and
CONFIG_NFS_V4_1 shouldn't be removed. But the devil is in the details
I'm sure...

Mike