[kernel][ltp] test fails in ioctl_loop05 - LOOP_SET_DIRECT_IO succeeded unexpectedly
openQA test in scenario sle-15-SP2-Online-aarch64-ltp_syscalls_debug_pagealloc@aarch64-virtio fails in
Test suite description¶
maintainer: pvorel, rpalethorpe. A large collection of tests for individual system calls. This is generally considered to be the most important test suit within the LTP. The IPC tests have been filtered out from the syscalls runtest file. They are ran as part of the syscalls_ipc test case.
NOTE: for GRUB_BOOT_NONDEFAULT=1 depends on order in GRUB_PARAM in install_ltp job.
Fails since (at least) Build 136.2
Last good: 128.1 (or more recent)
Always latest result in this scenario: latest
ioctl_loop05.c:36: PASS: lo_flags has LO_FLAGS_DIRECT_IO flag ioctl_loop05.c:40: PASS: /sys/block/loop0/loop/dio = 1 ioctl_loop05.c:38: PASS: lo_flags doesn't have LO_FLAGS_DIRECT_IO flag ioctl_loop05.c:40: PASS: /sys/block/loop0/loop/dio = 0 ioctl_loop05.c:61: PASS: LOOP_SET_DIRECT_IO succeeded ioctl_loop05.c:36: PASS: lo_flags has LO_FLAGS_DIRECT_IO flag ioctl_loop05.c:40: PASS: /sys/block/loop0/loop/dio = 1 ioctl_loop05.c:74: FAIL: LOOP_SET_DIRECT_IO succeeded unexpectedly Failed. Test took approximately 0.693848999000693 seconds Some test output could not be parsed: 7 lines were ignored.
#1 Updated by metan over 1 year ago
There does not seem to be anything wrong with the test, it works fine on different filesystem but fails on Btrfs for some reason.
Looking at the kernel code it may be that inode->i_sb->s_bdev is not set on Btrfs, but I would have to talk to kernel devs about this.
#2 Updated by metan over 1 year ago
Patch to the test send to ML after confirming with jeffm.
#7 Updated by pvorel over 1 year ago
NOTE: it was also broken on other archs, not just aarch64
And 4 days ago Jan Stancek pushed fix for ioctl_loop05 on s390 (not relevant to this)
jlausuch this could be https://bugzilla.suse.com/show_bug.cgi?id=1176983, also QAM jobs dont use daily LTP with latest fixes, if you still think it is a new issue, please open a new progress ticket, but please check if the bug is applicable and it happens on latest ltp build (in QA:Head repo).
For 15-SP1 QU5 we install LTP from https://build.suse.de/project/repository_state/QA:Head/SLE-15-SP1
Are latest LTP builds ending in all versions in QA:Head?