Public bug reported:

## Summary

After upgrading from kernel 6.17.0-1019-aws to 7.0.0-1009-aws, the NFS
client can no longer list (readdir/opendir) any subdirectory on an
NFSv4.1 mount. All subdirectory listings return "Remote I/O error"
(EREMOTEIO). File reads, stat(), and file creation in the root directory
continue to work. Rolling back to 6.17.0-1019-aws immediately resolves
the issue.

## Affected version
- linux-image-7.0.0-1009-aws 7.0.0-1009.9~24.04.1
- Ubuntu 24.04.4 LTS (Resolute HWE kernel)
- Architecture: amd64 (AWS EC2)

## Last known working version
- linux-image-6.17.0-1019-aws 6.17.0-1019.19~24.04.1

## NFS server
- NFS-Ganesha V7.3 (inside Longhorn share-manager v1.11.2 container)
- NFSv4.1, TCP, VFS FSAL
- Backing filesystem: XFS on iSCSI block device

## NFS client mount options
nfs4 
(rw,relatime,vers=4.1,rsize=1048576,wsize=1048576,namlen=255,softerr,softreval,noresvport,fatal_neterrors=none,proto=tcp,timeo=600,retrans=5,sec=sys,local_lock=none)

## Steps to reproduce
1. Boot with kernel 7.0.0-1009-aws
2. Mount an NFSv4.1 export from NFS-Ganesha (Longhorn RWX volume share-manager)
3. Create or access any subdirectory on the mount

## Observed behavior
```
# Root directory works
$ ls /mnt/nfs/
bootstrap_api_key  configs  mnesia  testfile

# stat on subdirectory works
$ stat /mnt/nfs/configs/
  File: /mnt/nfs/configs/
  Size: 4096       Blocks: 8          IO Block: 1048576 directory

# Reading a file inside a subdirectory works
$ cat /mnt/nfs/configs/app.config
[{ekka, ...   # file content returned successfully

# But listing ANY subdirectory fails
$ ls /mnt/nfs/configs/
ls: cannot open directory '/mnt/nfs/configs/': Remote I/O error

$ ls /mnt/nfs/mnesia/
ls: cannot open directory '/mnt/nfs/mnesia/': Remote I/O error

# Even a freshly created directory fails
$ mkdir /mnt/nfs/testdir    # succeeds
$ ls /mnt/nfs/testdir/
ls: cannot access '/mnt/nfs/testdir/..': Remote I/O error
total 0
drwxr-xr-x 2 root root 6 Jul 30 14:06 .
d????????? ? ?    ?    ?            ? ..
```

## Expected behavior
ls on subdirectories should return directory contents, as it does with kernel 
6.17.0-1019-aws.

## Key observations
- opendir() / getdents64() on subdirectories returns EREMOTEIO
- stat() on subdirectories works (NFS GETATTR succeeds)
- open() + read() on files inside subdirectories works (NFS LOOKUP + OPEN + 
READ succeeds)
- mkdir() and touch in the root directory work (NFS CREATE succeeds)
- The NFS server (NFS-Ganesha) is unaffected — local access from the server pod 
works perfectly
- The underlying block device has no I/O errors
- No errors in dmesg related to NFS
- Only affects NFSv4.1 mounts (RWX volumes); iSCSI-direct mounts (RWO volumes) 
are unaffected

## Impact
This regression breaks any workload using Kubernetes ReadWriteMany (RWX) 
PersistentVolumeClaims backed by Longhorn, which uses NFS-Ganesha for RWX 
access. In our case, EMQX (MQTT broker) enters CrashLoopBackOff because its 
RocksDB/mnesia storage requires directory listing to initialize.

## Workaround
Roll back to kernel 6.17.0-1019-aws and hold the 7.0 package:
```
sudo sed -i 's/^GRUB_DEFAULT=.*/GRUB_DEFAULT="Advanced options for 
Ubuntu>Ubuntu, with Linux 6.17.0-1019-aws"/' /etc/default/grub
sudo update-grub
sudo apt-mark hold linux-image-7.0.0-1009-aws linux-image-aws
sudo reboot
```

## Environment details
- 3-node Kubernetes cluster (kubeadm, server v1.35.6 / kubelet v1.35.2)
- Longhorn v1.11.2 (storage)
- NFS-Ganesha inside longhornio/longhorn-share-manager:v1.11.2
- kured auto-rebooted nodes into the new kernel between 02:00-06:00 UTC on 
2026-07-29/30
- Two independent clusters (opa-lab and opa-dev) both affected simultaneously 
after the kernel upgrade

** Affects: linux-aws (Ubuntu)
     Importance: Undecided
         Status: New


** Tags: kernel-bug-exists-upstream regression-release

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2162365

Title:
  linux-aws 7.0.0-1009: NFS readdir on NFSv4.1 subdirectories returns
  EREMOTEIO

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux-aws/+bug/2162365/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to