сifs client unable to mount SMB share without nosharesock mount option if current TCP connection dead

[dudniksergei]

What happened: I tested this manually from the smb-csi driver pod. The CIFS client is unable to mount the SMB share without the 'nosharesock' mount option. It attempts to reuse an established connection to the Samba share. However, if this connection is dead, the CIFS client fails to mount the SMB share. Adding the 'nosharesock' mount option creates a new TCP connection, allowing the SMB share to be mounted successfully

I tested it from csi-node pod

1. We have 5 TCP connections to SAMBA server
   sh-4.4# ss -na| grep 10.81.X.X
   tcp TIME-WAIT 0 0 10.82.72.60:55422 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:963 10.81.X.X:2049

tcp ESTAB 0 0 10.82.72.60:52514 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:43158 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:59186 10.81.X.X:445

2. Mount without nosharesock option failed because current TCP connectios are dead ( for some reason, for example network firewall ).

sh-4.4# mount --verbose -t cifs -o file_mode=0770,dir_mode=0770,username=user,vers=3.0,password="XXXXXXXX" //10.81.X.X/share /var/mnt1
mount error(5): Input/output error
Refer to the mount.cifs(8) manual page (e.g. man mount.cif

3. mount CIFS share with nosharesock option. It works because new connection was created

sh-4.4# mount --verbose -t cifs -o nosharesock,file_mode=0770,dir_mode=0770,username=user,vers=3.0,password="XXXXXXXX" //10.81.X.X/share /var/mnt1
mount.cifs kernel mount options: ip=10.81.X.X,unc=//10.81.X.X/share,nosharesock,file_mode=0770,dir_mode=0770,vers=3.0,user=user,pass=********

4. we have new TCP connection ( source port 51776)
   sh-4.4# ss -na| grep 10.81.X.X

tcp TIME-WAIT 0 0 10.82.72.60:55422 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:963 10.81.X.X:2049

tcp ESTAB 0 0 10.82.72.60:52514 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:43158 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:51776 10.81.X.X:445

tcp ESTAB 0 0 10.82.72.60:59186 10.81.X.X:445

What you expected to happen:

cifs.client creates additional TCP connection if the current conection is dead.

How to reproduce it:

This can be reproduced manually with the CIFS.Client package and a Samba server. Create two shares in the SMB server. Mount the first share, check TCP connection state from the client, then block this connection (using a firewall, for example). Afterward, try to mount the second share.

Anything else we need to know?:
This is behavior cifs.client.

Environment:

• CSI Driver version: v0.7.0
• Kubernetes version (use kubectl version): 1.25
• OS (e.g. from /etc/os-release): CoreOS 4.12
• Kernel (e.g. uname -a): 4.14.0-372.76.1.el8_6.x86_64
• Install tools:
• Others: cifs.client version 6.8

[andyzhangx]

I think this is what nosharesock mount option is used for: don't use existing connection, always create a new connection in the smb mount.

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle rotten
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten