CIFS errors

[master5006]

So, not exactly sure where to put this, but I would like some help!

I'm running two machines, one with TrueNAS and the other with Ubuntu server 20.04 that is connecting to TrueNAS via cifs.

 

[master5006]

and just to clarify everything seems to be working except for every so often, I think when these errors are being thrown, the connection gets slow between the two servers.

 

[anodos]

Can you post hardware details? The error on the TrueNAS-side is from cam and so you're experiencing an issue that is lower-level (not Samba).

 

[master5006]

Ya the truenas server is an R720XD the raid controller is a PERC710 in IT mode and the Ubuntu server is a R620 with a PERC710 not in IT mode. when I get home today I’ll put the processors and make it more detailed.

 

[mav@]

65536 is likely a DFLTPHYS -- old I/O size limit, just not sure from which device it goes. It may be from some USB2 device, where DFLTPHYS is used as a workaround for random buggy hardware. Or it can go from some old driver, just historically not reporting maxio size. 65539 I guess goes from some user-level monitoring tool, though not sure which. Such weird sizes are not used for data transfers. And considering there is ZFS between the disks and SMB, I don't see how one errors can propagate to another. There must be something more.

 

[HoneyBadger]

Possibly helpful bug: https://jira.ixsystems.com/browse/NAS-107018

Let's steal some dtrace mojo from Alexander Motin at iXsystems to see if we can find which is the offending device. Is there possibly a USB boot device?

@master5006 please paste the below code into a shell script on the TrueNAS machine (eg: dtrace.sh) make it executable with chmod +x and run it

Code:

#!/usr/sbin/dtrace -s
# pragma D option quiet
kernel:cam_periph_mapmem:entry
{
        self->cmd = ((union ccb *)arg0)->csio.cdb_io.cdb_bytes[0];
        self->bus = ((union ccb *)arg0)->csio.ccb_h.path_id;
}
kernel:cam_periph_mapmem:return
/ arg1 == 7 /
{
        printf("cam_periph_mapmem => %d at %s pid %d ppid %d cmd %x scbus%d\n",
            arg1, execname, pid, ppid, self->cmd, self->bus);
        self->cmd = 0;
        self->bus = 0;
}

 

[master5006]

It is a USB install of TrueNAS (I'd like it not to be, but for some reason the bios doesn't see the drives while the raid controller is in IT mode until something boots up from a USB drive)

the script seems like it might be running? but doesn't give any output and I have to CTRL+Z to be able to type any commands again.

 

[master5006]

the script finally did put some stuff out

 

[HoneyBadger]

It is a USB install of TrueNAS (I'd like it not to be, but for some reason the bios doesn't see the drives while the raid controller is in IT mode until something boots up from a USB drive)

Hmm. Not sure if the H710 can be flashed with an LSI BIOS/UEFI boot ROM, but that might be what's needed to get it to allow you to boot off an attached device. Could you sticky-tape an SSD inside and hook it to a SATA port? :D

It's running; did you see any console output of the cam_periph_mapmem errors pop up? The dtrace script is supposed to catch those as they happen and output the scbus ID of where it's going. Once we've got there, we dig through your dmesg/boot logs and figure out what device it maps to.

But I'd bet it's the USB boot media. What brand/make/model?

 

[master5006]

I might be able to try to put a SATA SSD in the server, but I don't see any immediate way I can get power, and not sure that the sata on the motherboard would be able to accept a hard drive, as it's listed as CD on the diagram.

The USB is an 8GB SanDisk Cruzer.

 

[anodos]

Regarding the kernel SMB client error on Linux client, you can try to correlate client log with /var/log/samba4/log.smbd contents to see if there's an error reported on our side when that happens.