This is a warning telling us that a shared memory backed file was read and it could not be read entirely. This would be triggered by an attempt of reading a file in /tmp/ for example. shmem_file_operations has a pointer for .read_iter() which points to "shmem_file_read_iter", the base for this stack trace.
There is a warning for slowpath, what usually tells us that kernel had to do a slower path for allocating memory. Since this test is testing async reads, it might be the case that the data wasn't ready after sometime because kernel was slowly trying to claim pages or something similar.
@Naresh,
Would you mind re-running this test ? So we can check if this stack trace happens again ?
I will re-test with 10 iterations and let you know.
56 times the test run and all times it PASS. no kernel crash or segmentation fault. https://lkft.validation.linaro.org/scheduler/job/349773#L642
Yep, that is why I asked for a re-run. This was a false positive due to high memory pressure, most likely, caused by previous tests in the same run. Best guess is that kernel was trying to acquire mem pages for the buffer copy (userland <-> kernel), for the vfs_read() of a tmpfs filesystem, and it took more time than test was designed to handle.
I expect this type of issues to be reduced when we move away from QEMU arm32/64 guests on x86_64 and run KVM guests on real arm for the tests.