[LU-2850] Client support for upstream 3.8 kernel

• Type: Improvement
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.5.0, Lustre 2.6.0
• Affects Version/s: Lustre 2.5.0
• Labels: emc patch
• Assignee: Yang Sheng
• Reporter: Peng Tao

Description

Update client code to support upstream 3.8 kernel.

Issue Links

• is related to: LU-2686 Kernel update for 3.7.2-201.fc18

Comments

[James A Simmons]

The only thing is I have modify http://review.whamcloud.com/#change,5655 due to conflicts with the patch from LU-2059 and LU-2800 autoconf patch. Needs updating as well. At this point it is hard to say which will land first.

[Alexey Shvetsov (Inactive)]

Yep. I run it on small production cluster with lustre/zfs 2.4rc2
servers are 3.8.6 clients are both 3.8.6 and 3.9.4

[James A Simmons]

With the above listed patches and patches from LU-3079 I get this.

LNet: HW CPU cores: 8, npartitions: 2
alg: No test for crc32 (crc32-table)
alg: No test for adler32 (adler32-zlib)
LNet: Added LNI 10.37.248.7@o2ib1 [8/2560/0/180]
Lustre: Lustre: Build Version: 2.4.50-g2f41b68-CHANGED-3.9.4
Lustre: Layout lock feature supported.
Lustre: Mounted lustre-client

It appears to work. We have Lustre 3.9.4 clients. With this I can start my work on ldiskfs for 3.9.4.

[James A Simmons]

The new recipe for this is against master is:

http://review.whamcloud.com/#change,5973 - LU-2686
http://review.whamcloud.com/#change,5974
http://review.whamcloud.com/#change,5763 - LU-2987
http://review.whamcloud.com/#change,5613 - LU-2850
http://review.whamcloud.com/#change,5655

I have tested with the currently support RHEL6.4 kernel. I'm working on getting a 3.9.4 kernel going to test against soon.

[James A Simmons]

Peter can you label this as a 2.5 item

[Peter Jones]

Alexey

2.4.0 is frozen now. This work will land in later releases

Peter

[Alexey Shvetsov (Inactive)]

Are there plans to land all this patches before 2.4.0 release? Or they will be postponed to 2.4.x or 2.5.x?

[Alexey Shvetsov (Inactive)]

BTW for 3.9 kernel support. Patches may be simplyfyed if stuff for ancient kernels (<2.6.32) will be cleaned up. See LU-2800

[Yang Sheng]

I tend to waiting patches landed and then rebase lu-3079.

[James A Simmons]

I noticed that the 5655 patch is not dependent on any other patches. This is fine but I noticed the patches in LU-3079 don't apply cleanly due to the other patches here. Should we just want until a few patches land then rebase the patches in LU-3079 again or make 5655 dependent on the rest of the series in this ticket?

[Peng Tao]

James, please upload it. It helps avoiding dup efforts. Thanks.

Yang, sorry for not including you in the first place. Will add you as inspector in all my future kernel support patches.