[LU-812] Support for Linux 3.0 kernels

• Type: Improvement
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.4.0
• Assignee: Yang Sheng
• Reporter: Cory Spitz

Description

Cray is interested in bringing Linux 3.0 support to clients and servers. We'll focus on clients first since that will obviously be easiest. However, Lustre won't even configure in a Linux 3.0 environment. Initially, we plan on submitting patches to the build process going. Eventually, we'll work our way up to the server patch series. Of course, the BKL was removed and we'll be looking for suggestions on how to work around that and other major changes. I presume that RHEL will go to Linux 3.x eventually, but Cray will initially be focusing on SLES SP2. Hopefully we can find a way to make the patch sets similar.

We'd like to at least have 3.0 clients be considered upstream for Lustre 2.3.

Issue Links

• is duplicated by: LU-3518 liblustreapi.a not forward compatible between 2.1.5 and 2.3.0 (maybe others)
• is related to: LU-14877 Lock inversion between inode mutex and layout mutex
• is related to: LU-2196 Build shared version of liblustreapi
• is related to: LU-2148 Kernel update for latest FC18 kernel
• is related to: LU-1718 fix Lustre NFS re-export for 3.0+ kernels
• Trackbacks: Lustre Community Development in Progress Features are being developed for future Lustre releases both at Whamcloud and by other organizations in the Lustre community. These will be eligible for inclusion in future Lustre releases as per our processes

Comments

[James A Simmons]

LU-1718 is still waiting to be landed to 2.1 branch.

[Peter Jones]

I think that it was just because LU-1718 was defined as a subtask - it works now that has been resolved

[Yang Sheng]

Stranger, I cannot close this ticket. I remember ever experienced this case. But forget how to deal with it.

[James A Simmons]

Last patch has been merged. We can close this ticket especially since we have LU-1337 and LU-1994 to handle further newer kernel development.

[James A Simmons]

No LU-812 will not be back ported to 2.1. The reason for the 2.1 patch is the api for vfs mount is the same in kernels from before 2.6.18 and also after 3.0.

[Jay Lan (Inactive)]

James, does that mean sles11sp2 client patches will be ported to b2_1 branch? Last I checked, there are 12 LU-812 patches checked in to master and b2_3.

[Peng Tao]

one more patch is added:

http://review.whamcloud.com/4091 llite: 3.0+ kernel fsync should call write

w/o it, I always fail sanity 63b.

[James A Simmons]

I take that back. We have a patch at http://review.whamcloud.com/#change,3661 for 2.1.3 branch.

[James A Simmons]

Patches have landed for Lustre 2.3 and Lustre 2.4 We can close this ticket now.

[Iurii Golovach (Inactive)]

Hi Xuezhao Liu

Please, add next code to your patch (http://review.whamcloud.com/3779 llite: O_LOV_DELAY_CREATE conflicts with FMODE_NONOTIFY) to pass sanity.sh tests.

diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
index 88ab9aa..90055e2 100644
— a/lustre/tests/sanity.sh
+++ b/lustre/tests/sanity.sh
@@ -1624,7 +1624,7 @@ test_33b()

{ rm -fr $DIR/d33 mkdir -p $DIR/d33 chown $RUNAS_ID $DIR/d33 - $RUNAS $OPENFILE -f 1286739555 $DIR/d33/f33 || true + $RUNAS $OPENFILE -f 21354413143 $DIR/d33/f33 || true }

run_test 33b "test open file with malformed flags (No panic)"

And one more question: we (in Xyratex) already met with this issue and designed the same patch you did. Please, let us know what do you think about the comment from Andreas and what are your plans about them (if you plan to integrate them what are the possible time for this action)?

Thank you.