Risk of boot-pool overrun. Samba spamming kernel log.

[UnexpectedDevelopment]

I'm running:

• TrueNAS-SCALE-23.10.0.1
• Six-core Xeon E5-2620
• 64 GB RAM
• 1 x RAIDZ3 | 12 wide | 14.55 TiB
• Broadcom / LSI SAS3008 PCI-Express Fusion-MPT SAS-3 (raid mode turned off)
• Intel Corporation Ethernet Controller 10-Gigabit X540-AT2 (two)

On my TrueNAS server, I have mounted a share from some old Windows file server with:

Code:

mount -t cifs -o vers=1.0,username=administrator //192.168.xxx.xxx/V /mnt/v

Yes, this old Windows server ONLY provides smb vers 1.0. I'm moving the media off of it so I can trash it.

I'm using rsync running on the TrueNAS box to move the media off. As soon as the transfers start the kernel log gets flooded with

Code:

Dec 18 14:56:36 picasso kernel: CIFS: __readahead_batch() returned 2/15

It floods at a rate of about 50 lines per second. If I don't manually zero the system log every few hours, it will fill up the OS disk.

I have seen this reported in a few places around the net, but no one is offering a solution. Has anyone seen this and have a fix?

 

[anodos]

I'm running:

• TrueNAS-SCALE-23.10.0.1
• Six-core Xeon E5-2620
• 64 GB RAM
• 1 x RAIDZ3 | 12 wide | 14.55 TiB
• Broadcom / LSI SAS3008 PCI-Express Fusion-MPT SAS-3 (raid mode turned off)
• Intel Corporation Ethernet Controller 10-Gigabit X540-AT2 (two)

On my TrueNAS server, I have mounted a share from some old Windows file server with:

Code:

mount -t cifs -o vers=1.0,username=administrator //192.168.xxx.xxx/V /mnt/v

Yes, this old Windows server ONLY provides smb vers 1.0. I'm moving the media off of it so I can trash it.

I'm using rsync running on the TrueNAS box to move the media off. As soon as the transfers start the kernel log gets flooded with

Code:

Dec 18 14:56:36 picasso kernel: CIFS: __readahead_batch() returned 2/15

It floods at a rate of about 50 lines per second. If I don't manually zero the system log every few hours, it will fill up the OS disk.

I have seen this reported in a few places around the net, but no one is offering a solution. Has anyone seen this and have a fix?

You can set up a syslog-ng filter to drop these messages, but that said, for migration you're probably better off mounting TrueNAS from old Windows server and peforming robocopy so that you guarantee metadata like alternate data streams is preserved.

 

[UnexpectedDevelopment]

You can set up a syslog-ng filter to drop these messages, but that said, for migration you're probably better off mounting TrueNAS from old Windows server and peforming robocopy so that you guarantee metadata like alternate data streams is preserved.

I thought about that, but it still floods the hardware console which makes the console unusable and it **IS** taking up noticeable CPU time. What you are proposing is just masking the issue. Is there a way to tell the CIFS kernel module to just not report this? Or better yet, use some tweaking parameter that satisfies what is being warned about?

 

[anodos]

Perhaps try turning off caching (cache=none) on your client mount and see if errors go away. The printk generating the message is at warning level, so you can probably adjust the kernel printk level to omit warning (4), of course warning implies it should probably be dealt with. Your rsync job may also strip MacOS related file metadata (if you have any mac clients). This may be undesired (which is why I suggested robocopy as an alernative).

BTW, do note that this is not samba spamming your log (title of this post). This is related to the Linux kernel SMB client.

 

[UnexpectedDevelopment]

Sorry... I never came back to this. Adding cache=none DID address this issue. Thanks!