Public bug reported:
---Problem Description---
htm capability is switched off by default qemu machine type `pseries-cosmic`.
We need to enable htm [cap-htm=on].
---uname output---
Linux ltc-kvmfvtzz 4.18.0-7-generic #8-Ubuntu SMP Tue Aug 28 18:20:56 UTC 2018
ppc64le ppc64le ppc64le GNU/Linux
Machine Type = ZZ-L
---Steps to Reproduce---
1. Log into P9 Ubuntu 18.10 host, take console of guest `srik_vm1`
2. Inside guest run TM selftests
git clone --depth 1 https://github.com/torvalds/linux.git;cd
linux/tools/testing/selftests/powerpc/;git log --oneline -1;make;make -C tm
run_tests
All TM tests would be skipped
root@ubuntu1810:~# LD_SHOW_AUXV=1 /bin/true |grep htm
root@ubuntu1810:~#
==
When you enable htm feature explicitly tm tests would pass:
selftests: tm: tm-resched-dscr
========================================
test: tm_resched_dscr
tags: git_version:v4.19-rc3-0-g11da3a7f8
Binding to cpu 7
main test running as pid 1554
Check DSCR TM context switch: OK
success: tm_resched_dscr
ok 1..1 selftests: tm: tm-resched-dscr [PASS]
selftests: tm: tm-syscall
========================================
test: tm_syscall
tags: git_version:v4.19-rc3-0-g11da3a7f8
Testing transactional syscalls for 10 seconds...
6886507 active and suspended transactions behaved correctly.
(There were 1476 transaction retries.)
success: tm_syscall
ok 1..2 selftests: tm: tm-syscall [PASS]
selftests: tm: tm-signal-msr-resv
========================================
test: tm_signal_msr_resv
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_signal_msr_resv
ok 1..3 selftests: tm: tm-signal-msr-resv [PASS]
selftests: tm: tm-signal-stack
========================================
test: tm_signal_stack
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_signal_stack
ok 1..4 selftests: tm: tm-signal-stack [PASS]
selftests: tm: tm-vmxcopy
========================================
test: tm_vmxcopy
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_vmxcopy
ok 1..5 selftests: tm: tm-vmxcopy [PASS]
selftests: tm: tm-fork
========================================
test: tm_fork
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_fork
ok 1..6 selftests: tm: tm-fork [PASS]
selftests: tm: tm-tar
========================================
Starting, 10000 loops
test: tm_tar
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_tar
ok 1..7 selftests: tm: tm-tar [PASS]
selftests: tm: tm-tmspr
========================================
test: tm_tmspr
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_tmspr
ok 1..8 selftests: tm: tm-tmspr [PASS]
selftests: tm: tm-vmx-unavail
========================================
test: tm_vmx_unavail_test
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_vmx_unavail_test
ok 1..9 selftests: tm: tm-vmx-unavail [PASS]
selftests: tm: tm-unavailable
========================================
test: tm_unavailable_test
tags: git_version:v4.19-rc3-0-g11da3a7f8
Checking if FP/VEC registers are sane after a FP unavailable exception...
If MSR.FP=0 MSR.VEC=0: FP ok VEC ok
If MSR.FP=1 MSR.VEC=0: FP ok VEC ok
If MSR.FP=0 MSR.VEC=1: FP ok VEC ok
If MSR.FP=1 MSR.VEC=1: FP ok VEC ok
Checking if FP/VEC registers are sane after a VEC unavailable exception...
If MSR.FP=0 MSR.VEC=0: FP ok VEC ok
If MSR.FP=1 MSR.VEC=0: FP ok VEC ok
If MSR.FP=0 MSR.VEC=1: FP ok VEC ok
If MSR.FP=1 MSR.VEC=1: FP ok VEC ok
Checking if FP/VEC registers are sane after a VSX unavailable exception...
If MSR.FP=0 MSR.VEC=0: FP ok VEC ok
If MSR.FP=1 MSR.VEC=0: FP ok VEC ok
If MSR.FP=0 MSR.VEC=1: FP ok VEC ok
If MSR.FP=1 MSR.VEC=1: FP ok VEC ok
result: success
success: tm_unavailable_test
ok 1..10 selftests: tm: tm-unavailable [PASS]
selftests: tm: tm-trap
========================================
test: tm_trap_test
tags: git_version:v4.19-rc3-0-g11da3a7f8
Little-Endian machine detected. Checking if endianness flips inadvertently on
trap in TM... no.
success: tm_trap_test
ok 1..11 selftests: tm: tm-trap [PASS]
selftests: tm: tm-signal-context-chk-gpr
========================================
test: tm_signal_context_chk_gpr
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_signal_context_chk_gpr
ok 1..12 selftests: tm: tm-signal-context-chk-gpr [PASS]
selftests: tm: tm-signal-context-chk-fpu
========================================
test: tm_signal_context_chk_fpu
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_signal_context_chk_fpu
ok 1..13 selftests: tm: tm-signal-context-chk-fpu [PASS]
selftests: tm: tm-signal-context-chk-vmx
========================================
test: tm_signal_context_chk_vmx
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_signal_context_chk_vmx
ok 1..14 selftests: tm: tm-signal-context-chk-vmx [PASS]
selftests: tm: tm-signal-context-chk-vsx
========================================
test: tm_signal_context_chk_vsx
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_signal_context_chk_vsx
ok 1..15 selftests: tm: tm-signal-context-chk-vsx [PASS]
selftests: tm: tm-sigreturn
========================================
test: tm_sigreturn
tags: git_version:v4.19-rc3-0-g11da3a7f8
success: tm_sigreturn
ok 1..16 selftests: tm: tm-sigreturn [PASS]
make: Leaving directory '/root/linux/tools/testing/selftests/powerpc/tm'
root@ubuntu1810:~/linux/tools/testing/selftests/powerpc# uname -r
4.17.0-9-generic
== ==
After enabling:
root@ubuntu1810:~# LD_SHOW_AUXV=1 /bin/true |grep htm
AT_HWCAP2: htm-nosc vcrypto tar isel ebb dscr htm arch_2_07
== Comment: #7 - Murilo Opsfelder Araujo <muri...@br.ibm.com> - 2018-09-12
09:15:37 ==
HTM cap is disabled by default in upstream QEMU pseries-2.12 onwards. IIRC,
pseries-cosmic is an alias to pseries-2.12 so it also has HTM=off by default.
I'm not sure if Cosmic should diverge from upstream that much. If
HTM=on is set for Cosmic, future releases of Ubuntu will likely has
HTM=off when rebasing to newer QEMU code.
Besides that, a migration from a pseries-2.12 HTM=off to a pseries-2.12
HTM=on could cause different behaviours.
== Leonardo Augusto Guimaraes Garcia <lagar...@br.ibm.com>
> I'm not sure if Cosmic should diverge from upstream that much. If HTM=on is
> set for Cosmic, future releases of Ubuntu will likely has HTM=off when
> rebasing to newer QEMU code.
This is really a decision for Canonical IMO. We should mirror the bug
and let them decide.
I am in favor of having HTM=on. But it needs to be made clear to
Canonical that this will imply carrying a patch to enable it as long as
HTM=off remains the default upstream.
** Affects: qemu (Ubuntu)
Importance: Undecided
Assignee: Ubuntu on IBM Power Systems Bug Triage (ubuntu-power-triage)
Status: New
** Tags: architecture-ppc64le bugnameltc-171275 severity-critical
targetmilestone-inin1810
** Tags added: architecture-ppc64le bugnameltc-171275 severity-critical
targetmilestone-inin1810
** Changed in: ubuntu
Assignee: (unassigned) => Ubuntu on IBM Power Systems Bug Triage
(ubuntu-power-triage)
** Package changed: ubuntu => qemu (Ubuntu)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1792438
Title:
[Ubuntu 18.10] Need to enable TM in pseries-cosmic
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1792438/+subscriptions
--
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
