[LU-1646] mounting lustre oops on older kernels

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.1.2
• Type: Bug
• Priority: Minor
• Reporter: [e-mail]
• Assignee: Bob Glossman (Inactive)
• Resolution: Duplicate

Environment

sles10sp4

• Severity: 3

Description

in lustre 2.1 and up, the lustre_fill_super() was changed to expect the anonymous pointer to be of the type struct lustre_mount_data2. however, the ifdef for lustre_get_sb() for kernels < 2.6.18 was not updated. the attached patch seems to fix things but you might run into trouble when nfs exporting from these old kernels since lmd_mnt cannot be filled in for these older kernels.

Issue Links

• is related to: LU-1718 fix Lustre NFS re-export for 3.0+ ker...

Comments

[James A Simmons]

Which versions of Lustre are you using or plan to use? Some patches are in the works that affect you.

[[e-mail]]

2.1 for the foreseeable future.

[Peter Jones]

Hi Chas

SLES10 is not officially supported for 2.1.x (see http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix). There is some interest in adding this support. If you would like to get your patch landed into the Lustre tree could you please upload it into gerrit so that it can be tested and reviewed? Details on how to do this are on the Whamcloud wiki

Thanks

Peter

[chas williams - CONTRACTOR]

this was fixed via LU-812 in commit 2eba197319886c4f2903fd937aebe69eea35290e. do i need to do anything for this to get backported to the 2.1 and 2.2 releases?

[Peter Jones]

Bob

Could you please look into this change and see whether it would be practical to backport it to b2_1?

Thanks

Peter

[James A Simmons]

More than that patch would need to be back ported. See LU-1718 as well.

[Bob Glossman (Inactive)]

The change in http://review.whamcloud.com/#change,1951 can be back ported, but with side effects. LU-1718 mentioned by James in the previous comment tracks the disable of reexport of lustre fs to NFS due to this change. Doing the back port would disable reexport to NFS for old kernels just as it is for 3.0+ kernels. This would remain true until a fix for LU-1718 is completed and back ported.

[Bob Glossman (Inactive)]

if we proceed with a back port we will also need LU-1718 back ported to reenable NFS export

[Bob Glossman (Inactive)]

chas,
Is it OK if we disable NFS export in 2.1.x for really old kernels? Can you possibly work around that restriction by using 1.8.x with such old kernels, at least until we have a fix from LU-1718 available?

[chas williams - CONTRACTOR]

since you cant even mount lustre on really old kernels, i dont think anyone is using the nfs export feature. so this seems fine to me.

[Bob Glossman (Inactive)]

back port to b2_1
http://review.whamcloud.com/#change,3661

[James A Simmons]

Also besides patch http://review.whamcloud.com/#change,3661 from LU-812 you will need LU-1718 for NFS export which has the patch at http://review.whamcloud.com/#change,3840.

[James A Simmons]

Chas can you test both patches. Also if you can please try a NFS export on one of your clients with lustre.

[James A Simmons]

The patch from LU-812 that solves this problem has been landed to b2_1. This ticket can be closed.

[Peter Jones]

ok thanks James!