[LU-3675] Support for linux 3.10 kernel

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.5.0
• Type: Improvement
• Priority: Minor
• Reporter: James A Simmons
• Assignee: Lai Siyao
• Resolution: Fixed

Description

Tracker for 3.10 kernel support

Issue Links

• is related to: LU-3319 Adapt to 3.10 upstream kernel proc_di...
• is related to: LU-3373 ldiskfs patches for FC19

Comments

[Peter Jones]

Lai

You have been looking at some of these patches already

Peter

[James A Simmons]

Patch http://review.whamcloud.com/#/c/5390 from LU-2800 handles a block device function change now for 3.10 for lloop.c.

[Bob Glossman (Inactive)]

I've noticed an odd failure to autoload kernel modules on 3.10 kernels. I thought this might be related to the CONFIG_MODULE_FORCE_LOAD setting in the kernel config, but setting that on or off seems to make no difference. Causes a command like 'mount -t lustre ...' client mount to fail due to lustre modules not getting loaded.

In the short term for test purposes this isn't a big problem. Just need to do a manual 'modprobe lustre' before trying to do any client mounts. Longer term really need to find the cause & a solution.

[James A Simmons]

With the patches from LU-3319 and http://review.whamcloud.com/#/c/5380 from LU-2800 I just mounted a ZFS server file system.

[root@spoon46 tests]# df
Filesystem 1K-blocks Used Available Use% Mounted on
unionfs 88763392 61317120 22938624 73% /
none 100 0 100 0% /dev/shm
none 2957192 25312 2931880 1% /tmp
nccsfiler3.ccs.ornl.gov:/vol/techint
1015021568 871888352 143133216 86% /autofs/na3_techint
172.30.248.91:/data/shared-root-homedir
57799680 34323456 20540416 63% /root
172.30.248.91:/chexport
46445568 32187392 11898880 74% /chexport
lustre-mdt1/mdt1 157440 3072 152320 2% /tmp/lustre/mds1
lustre-ost1/ost1 157440 1152 154240 1% /tmp/lustre/ost1
lustre-ost2/ost2 157440 1152 154240 1% /tmp/lustre/ost2
172.30.248.48@tcp0:/lustre
314880 2304 308480 1% /lustre/barry
[root@spoon46 tests]# uname -r
3.10.2

[Bob Glossman (Inactive)]

James, excellent. I have also done some successful zfs only server builds recently, but haven't done any functional testing yet.

Any ETA on the few remaining osd-ldiskfs/osd_handler.c build blocking issues?

[James A Simmons]

I have the proc seq_file coding done for osd-ldiskfs but their is a lbug dealing with the osp layer when I attempt to mount a OST. I haven't updated the filldir_t changes for osd-ldisk yet. Even with those issues fixed their are problems building osd-ldiskfs with the fc19 ldiskfs patch Yeng had provided. I say it could be ready by the end of next week.

[Bob Glossman (Inactive)]

some additional fixes to osd-ldiskfs just today. see comments in LU-3373. don't know if those address some of your issues or not.

As far as I know the seq_file and readdir changes are still needed. Looking forward to seeing those. The end of next week sounds good to me. Wasn't looking for a cast in concrete promise, just some idea of when. Your answer is fine.

[James A Simmons]

Yep the patch http://review.whamcloud.com/#/c/8116 from LU-3373 did the trick. I have something that can build but does it work is another question.

[James A Simmons]

The patch from LU-2800 needed to support 3.10 kernels has been landed. Now only patches from LU-3319 are needed. We can close this ticket and focus on 3.11 kernel support now.

[Peter Jones]

ok. Thanks James