[LU-517] e2fsprogs can't create a packages if lustre mounted on build host.

• Status: Closed
• Project: Lustre
• Type: Bug
• Priority: Major
• Reporter: Alexey Lyashkov
• Assignee: Andreas Dilger
• Resolution: Fixed

Environment

RHEL6

• Story Points: 1
• Severity: 3

Description

i tried to rebuild e2fsprogs from repo and hit a error.

Size of struct ext2_super_block is 1024
LD_LIBRARY_PATH=../../lib DYLD_LIBRARY_PATH=../../lib ./tst_csum
csum0000: UUID 4f25e8cf-e797-4823-befa-a7884baeecdb(241f), grp 0(df47): d3a4=d3a4
csum0001: UUID 4f25e8cf-e797-4823-befa-a7884baeecdb(241f), grp 1(2346): 3ea5=3ea5
csumffff: UUID 4f25e8cf-e797-4823-befa-a7884baeecdb(241f), grp 2(6746): 49a5=49a5
csum_set: UUID 4f25e8cf-e797-4823-befa-a7884baeecdb(241f), grp 0(df47): d3a4=d3a4
new_uuid: UUID 30303030-3030-3030-3030-303030303030(766d), grp 0(76fd): a4ac=a4ac
csum_new: UUID 30303030-3030-3030-3030-303030303030(766d), grp 0(76fd): a4ac=a4ac
csum_blk: UUID 30303030-3030-3030-3030-303030303030(766d), grp 0(76fd): 2e14=2e14
LD_LIBRARY_PATH=../../lib DYLD_LIBRARY_PATH=../../lib ./tst_read_ea
(lustre)/dev/loop0:/mnt/mds1
mkstemp failed: Not a directory
make[1]: *** [check] Error 1
make[1]: Leaving directory `/root/rpmbuild/BUILD/e2fsprogs-1.41.90.wc3/lib/ext2fs'
make: *** [check-recursive] Error 1
error: Bad exit status from /var/tmp/rpm-tmp.eLhD9Q (%check)

that error removed when i stopped a lustre fs which a run on same test env.

Comments

[Andreas Dilger]

This is probably due to hard coding /dev/loop0 into the test script, which is also used if you are testing Lustre with an "llmount.sh" setup.

It seems the e2fsprogs test script should check if /dev/loop device is free before using it.

[Andreas Dilger]

Actually, it looks like the tst_read_ea test is checking for a filesystem with "user_xattr" as a mount option (which the MDS mountpoint matches) and then trying to create a temporary file in /mnt/mds via the VFS and read it back via libext2fs and verify the value. Unfortunately, it is not possible to create files in /mnt/mds, so the test fails.

Normally, a local ext3/4 filesystem should be found before the MDS mountpoint, so this has never previously been a problem. I guess build host has no ext3/4 filesystem locally mounted?

The fix I'm going to commit is that if mkstemp() fails on the mountpoint (which could happen for a number of reasons such as permission problems, etc, just continue looking for another ext3/4 filesystem to run the test on.

http://review.whamcloud.com/1321

[Andreas Dilger]

Fixed in the 1.41.90.wc4 build.

[Build Master (Inactive)]

Integrated in e2fsprogs-master » i686,el6 #54
LU-517 tests: don't fail mkstemp() on Lustre fs

Andreas Dilger : 7178b715eb0ba5d55ee9cf51b75765e5c19ba67c
Files :

• patches/e2fsprogs-attr_cleanup.patch

[Build Master (Inactive)]

Integrated in e2fsprogs-master » x86_64,el6 #54
LU-517 tests: don't fail mkstemp() on Lustre fs

Andreas Dilger : 7178b715eb0ba5d55ee9cf51b75765e5c19ba67c
Files :

• patches/e2fsprogs-attr_cleanup.patch

[Build Master (Inactive)]

Integrated in e2fsprogs-master » x86_64,el5 #54
LU-517 tests: don't fail mkstemp() on Lustre fs

Andreas Dilger : 7178b715eb0ba5d55ee9cf51b75765e5c19ba67c
Files :

• patches/e2fsprogs-attr_cleanup.patch

[Build Master (Inactive)]

Integrated in e2fsprogs-master » i686,el5 #54
LU-517 tests: don't fail mkstemp() on Lustre fs

Andreas Dilger : 7178b715eb0ba5d55ee9cf51b75765e5c19ba67c
Files :

• patches/e2fsprogs-attr_cleanup.patch