[PROXMOX 5.1] [lpfc] HBA Emulex LPe12000 error

[fabian]

Hello Fabian,

I did some more research, the last working kernel from the Ubuntu Xenial repository is
linux-image-4.10.0-42-generic_4.10.0-42.46~16.04.1_amd64.deb

The next available image linux-image-4.11.0-14-generic_4.11.0-14.20~16.04.1_amd64.deb is not working, I also tried several Ubuntu 4.13 and mainline kernels, none of them is working.

The first working kernel I found is mainline 4.14.1.

I think this commit is the fix for the problem github.com/torvalds/linux/commit/2877cbffb79ed121a6bcc5edbe629d3aba36cd29

Unfortunately the scsi subsystem was heavily refactored from 4.10 to 4.14 so I am unable to isolate a single patch to fix/reproduce the problem.

Are there any plans to upgrade to a newer kernel soon? I am still evaluating the PVE on my lab but need to bring those boxes in production within the next two weeks, so I need to decide weather to go with 5.0, try to patch myself or move to another solution...

Oliver

I'll see about building with that commit included and letting you test. just to confirm, could you test 4.14-rc1 mainline (or was that a typo and this is the one you actually tested?)?

 

[oliwel]

Hi Fabian,

I just finished my tests - I applied the above mentioned two-line commit on top of current pve-kernel HEAD (1da60899) and it works for me. So perhaps you want to include the commit as patch for the next kernel release making me happy (and buying a subscription for my boxes)

The kernel I tried with success yesterday was kernel.ubuntu.com/~kernel-ppa/mainline/v4.14.1/

Oliver

 

[fabian]

Hi Fabian,

I just finished my tests - I applied the above mentioned two-line commit on top of current pve-kernel HEAD (1da60899) and it works for me. So perhaps you want to include the commit as patch for the next kernel release making me happy (and buying a subscription for my boxes)

The kernel I tried with success yesterday was kernel.ubuntu.com/~kernel-ppa/mainline/v4.14.1/

Oliver

queued, but will take a bit to hit the repos because of other bigger upgrades

 

[oliwel]

Is "a bit" counted in days, weeks or month

 

[komspr]

I compiled a module with this patch https://github.com/torvalds/linux/commit/2877cbffb79ed121a6bcc5edbe629d3aba36cd29 and it works on 4.13.13-5-pve kernel.

 

[fabian]

Is "a bit" counted in days, weeks or month

preview (pvetest) should be available early next week at the latest.

 

[fabian]

preview (pvetest) should be available early next week at the latest.

latest 4.13.13-6-pve kernel contains the commit in question, available in the pvetest repository. note that this kernel also got recompiled with RETPOLINE support for Spectre mitigation, so the changeset is unusually large and you might get bitten by other issues.

 

[komspr]

I confirm that 4.13.13-6-pve kernel has working lpfc module. It works as expected.

 

[micol]

Does LPe12000 and LPe12002 works on kernels higher 4.13? On lastest 4.15.18-4-pve?

 

[micol]

up

 

[micol]

Up

 

[t.lamprecht]

Up

Stop bumping threads continuously!

Does LPe12000 and LPe12002 works on kernels higher 4.13? On lastest 4.15.18-4-pve?

There's currently no report why it shouldn't work.
The respective fixes from the old problem are still in there. So, why don't you just try it? You need to migrate all important guests off the test node either way, and the 4.13 kernel stays installed - so if it really fails you can quickly reboot into the 4.13 kernel and have a working systems again, and report here that there's something off again...