This can also be observed on oracular linux 6.11.0-25.25 in s2025.03.17.
--
You received this bug notification because you are a member of Canonical
Platform QA Team, which is subscribed to ubuntu-kernel-tests.
https://bugs.launchpad.net/bugs/2106564
Title:
powerpc/ptrace:{ptrace-tm-tar, ptrace-tm-spd-tar} fail with PPR
discrepancy on Power9 PowerVM with j/linux
Status in ubuntu-kernel-tests:
New
Bug description:
Failure observed on bionic linux-hwe-5.4 5.4.0-214.234~18.04.1
(2025.03.17) and jammy linux 5.15.0-139.149 (edited) on metal
P9g-LPAR01kernel (Power9 PowerVM).
The failure was consistently observed across three runs (3/3. 100%) on
this machine.
Log output on j/linux:
# selftests: powerpc/ptrace: ptrace-tm-tar
# test: ptrace_tm_tar
# tags: git_version:unknown
# [User Read (Running)] TAR: 40 PPR: 18000000000000 DSCR: 400
# [Ptrace Read (Running)] TAR: 20 PPR: 8000000000000 DSCR: 200
# [Ptrace Read (Checkpointed)] TAR: 10 PPR: 4000000000000 DSCR: 100
# [Ptrace Write (Checkpointed)] TAR: 40 PPR: 10000000000000 DSCR: 400
# failure: ptrace_tm_tar
# selftests: powerpc/ptrace: ptrace-tm-spd-tar
# test: ptrace_tm_spd_tar
# tags: git_version:unknown
# [User Read (Running)] TAR: 40 PPR: 18000000000000 DSCR: 400
# [Ptrace Read (Running)] TAR: 30 PPR: c000000000000 DSCR: 300
# [Ptrace Read (Checkpointed)] TAR: 10 PPR: 4000000000000 DSCR: 100
# [Ptrace Write (Checkpointed)] TAR: 40 PPR: 10000000000000 DSCR: 400
# failure: ptrace_tm_spd_tar
Log output on b/hwe-5.4:
[powerpc/ptrace:ptrace-tm-tar]
# selftests: powerpc/ptrace: ptrace-tm-tar
# test: ptrace_tm_tar
# tags: git_version:unknown
(X1) # [User Read (Running)] TAR: 40 PPR: 4000000000000 DSCR: 400
# [Ptrace Read (Running)] TAR: 20 PPR: 8000000000000 DSCR: 200
# [Ptrace Read (Checkpointed)] TAR: 10 PPR: 4000000000000 DSCR: 100
(X2) # [Ptrace Write (Checkpointed)] TAR: 40 PPR: 10000000000000 DSCR: 400
# failure: ptrace_tm_tar
not ok 1 selftests: powerpc/ptrace: ptrace-tm-tar # exit=1
[powerpc/ptrace:ptrace-tm-spd-tar]
# selftests: powerpc/ptrace: ptrace-tm-spd-tar
# test: ptrace_tm_spd_tar
# tags: git_version:unknown
(X3) # [User Read (Running)] TAR: 40 PPR: 18000000000000 DSCR: 400
# [Ptrace Read (Running)] TAR: 30 PPR: c000000000000 DSCR: 300
# [Ptrace Read (Checkpointed)] TAR: 10 PPR: 4000000000000 DSCR: 100
(X4) # [Ptrace Write (Checkpointed)] TAR: 40 PPR: 10000000000000 DSCR: 400
# failure: ptrace_tm_spd_tar
not ok 1 selftests: powerpc/ptrace: ptrace-tm-spd-tar # exit=1
* At (X3) above, PPR should've been 10000000000000 (Normal) as set at (X4)
as the transaction is aborted, but it was read as 18000000000000 (High).
Neither tm_spd_tar() nor trace_tm_spd_tar() (via ptrace) set this value
(i.e. or 3,3,3) inside or outside the transaction. Given this should be
an LPAR guest from its naming, its hypervisor side might have done
something unexpected.
* At (X1) above, PPR should've been 10000000000000 (Normal) as set at (X2),
but it was read as 4000000000000 (Very low).
Even though (X1) indicates the possibility that only updated CPPR failed to
be read by tm_tar()'s mfspr(SPRN_PPR) for some reason (*notice that TAR and
DSCR values are correctly read by mfspr as shown in the log), it's likely
that this is just a symptom caused by unexpected behavior from the
hypervisor, as observed at (X3).
Ref. kernel doc on POWER9 TM issues:
https://www.kernel.org/doc/html/latest/arch/powerpc/transactional_memory.html#power9
P9g-LPAR01kernel is a new test instance type, so this is likely not a
regression introduced in this SRU cycle.
This test passed on other environments over many SRU cycles.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-kernel-tests/+bug/2106564/+subscriptions
--
Mailing list: https://launchpad.net/~canonical-ubuntu-qa
Post to : canonical-ubuntu-qa@lists.launchpad.net
Unsubscribe : https://launchpad.net/~canonical-ubuntu-qa
More help : https://help.launchpad.net/ListHelp
