pkey01 from ubuntu_ltp_syscalls failed on B-aws / B-aws-fips / X-4.15 AWS / B-oracle with "Test killed by SIGBUS!"

[Po-Hsu Lin]

This issue was only spotted on AWS instance "i3en.24xlarge" with B-aws-fips kernel (aws-fips : 4.15.0-2002.2), test interrupted with:
    tst_test.c:1163: BROK: Test killed by SIGBUS!

Test log:
 tag=pkey01 stime=1570753169 dur=0 exit=exited stat=2 core=no cu=0 cs=0
 startup='Fri Oct 11 00:19:29 2019'
 tst_test.c:1118: INFO: Timeout per run is 0h 05m 00s
 pkey01.c:164: INFO: Set PKEY_DISABLE_ACCESS on (MAP_ANONYMOUS|MAP_PRIVATE) buffer
 pkey01.c:187: PASS: Child ended by SIGSEGV as expected
 pkey01.c:191: INFO: Remove PKEY_DISABLE_ACCESS from the buffer
 pkey01.c:197: PASS: Read buffer success, buffer[0] = 0
 pkey01.c:164: INFO: Set PKEY_DISABLE_ACCESS on (MAP_ANONYMOUS|MAP_SHARED) buffer
 pkey01.c:187: PASS: Child ended by SIGSEGV as expected
 pkey01.c:191: INFO: Remove PKEY_DISABLE_ACCESS from the buffer
 pkey01.c:197: PASS: Read buffer success, buffer[0] = 0
 pkey01.c:164: INFO: Set PKEY_DISABLE_ACCESS on (MAP_ANONYMOUS|MAP_PRIVATE|MAP_HUGETLB) buffer
 pkey01.c:187: PASS: Child ended by SIGSEGV as expected
 pkey01.c:191: INFO: Remove PKEY_DISABLE_ACCESS from the buffer
 tst_test.c:1163: BROK: Test killed by SIGBUS!

 Summary:
 passed 5
 failed 0
 skipped 0
 warnings 0

Probably a hiccup. As this test has passed on other instances like i3en.2xlarge, c5n.xlarge, c5.large, t3.medium, r5.large

[Po-Hsu Lin (cypressyew)]

Can be found on X-4.15 AWS (4.15.0-1056.58~16.04.1-aws) on some instances
* c5.metal
* i3en.24xlarge
* r5.metal

[Po-Hsu Lin (cypressyew)]

Found on 4.15.0-1033.36~16.04.1-oracle with BM.Standard2.52
    tst_test.c:1163: BROK: Test killed by SIGBUS!

[Po-Hsu Lin (cypressyew)]

Still can be seen on B-aws 4.15.0-1074.78
Only spotted on bare-metal nodes.

[Po-Hsu Lin (cypressyew)]

This can be found on Bionic Oracle 4.15.0-1054.58 with instance BM.Standard2.52 only (passed with VM.DenseIO2.8 and VM.Standard2.1)

[Po-Hsu Lin (cypressyew)]

Didn't see this with Oracle 4.15.0-1055.59 on BM.Standard2.52, VM.DenseIO2.8 and VM.Standard2.1

[Po-Hsu Lin (cypressyew)]

Still valid on 4.15.0-1059.65 - oracle
BM.Standard2.52 only

[Po-Hsu Lin (cypressyew)]

This is caused by move_pages12 test (bug 1831043)
If you just run these two tests on BM.Standard2.52 with 4.15.0-1031-oracle:
$ cat /home/ubuntu/syscalls
#fallocate06 fallocate06
#ioctl_sg01 ioctl_sg01
#fanotify09 fanotify09
move_pages12 move_pages12
pkey01 pkey01
$ sudo /opt/ltp/runltp -f /home/ubuntu/syscalls

You will see this test fail with "Test killed by SIGBUS!"
But if you run just this job, it will pass without any issue.

[Po-Hsu Lin (cypressyew)]

Spotted on X-Oracle 4.15.0-1081.89~16.04.1 BM.Standard2.52
Manually tested to double confirm this is caused by move_pages12 mentioned in my previous comment #7

However this issue does not exist with B-oracle 4.15.0-1081.89, even with node BM.Standard2.52 failing with the move_pages12 test.