fatrace autopkgtests assume access to certain CPU registers

[Francis Ginther]

While testing a newer implementation of the ubuntu proposed-migration test infrastructure, we discovered that the fatrace autopkgtest is depending on CPU registers that may not exist on the testbed. The existing infrastructure emulates an Intel core2duo, but the cpu architecture on the new infrastructure is a Xeon E312xx. The Xeon does not have the IDA (Intel Dynamic Acceleration) support which appears to be what supplies MSR 0xe8.

The bug I am reporting is that the test should account for the possibility that the hardware support is not present and do something other then generate an error (possibly a skipped test).

'ida' support can be grepped from 'cat /proc/cpuinfo'.

[log snippet with error messages]
Preparing to take measurements
  unknown op '{'
read_msr cpu0 0xe8 : Input/output error
Measurement will begin in 5 seconds. Please make sure that the
computer is idle, i. e. do not press keys, start or operate programs, and that
programs are not busy with active tasks other than the one you want to examine.
Starting measurement for 60 seconds...
Measurement complete. Generating report...
Traceback (most recent call last):
  File "/usr/sbin/power-usage-report", line 199, in <module>
    main()
  File "/usr/sbin/power-usage-report", line 196, in main
    powertop_report(output['powertop'])
  File "/usr/sbin/power-usage-report", line 173, in powertop_report
    while not lines[i]:
IndexError: list index out of range
adt-run: DBG: testbed command exited with code 1

[full log]
See fatrace-adt-run.log attached

[Martin Pitt (pitti)]

There is no log attached, so I tried to reproduce this locally:

$ qemu-system-x86_64 -cpu help|grep Xeon
x86 SandyBridge Intel Xeon E312xx (Sandy Bridge)

$ adt-run fatrace --- qemu --qemu-options='-cpu SandyBridge' /srv/vm/adt-vivid-amd64-cloud.img

I get the same error.

[Martin Pitt (pitti)]

"grep -qw ida /proc/cpuinfo" seems to be a too big hammer; core2duo and the default qemu processor don't have that either, but powertop works on those. So this needs some more investigation how to effectively skip the test on the right machines only.

[Martin Pitt (pitti)]

I opted for just asking powertop itself :-) http://anonscm.debian.org/cgit/collab-maint/fatrace.git/commit/?id=ff97be7e927

[Launchpad Janitor (janitor)]

This bug was fixed in the package fatrace - 0.10-1

---------------
fatrace (0.10-1) unstable; urgency=medium

  * New upstream release.
  * debian/tests/power-usage-report: Skip test if processor does not support
    powertop. (LP: #1449632)
  * Update Vcs-* links for anonscm.debian.org.
  * Drop obsolete XS-Testsuite: control tag, dpkg now adds this automatically.

 -- Martin Pitt <email address hidden> Wed, 29 Apr 2015 10:45:09 +0200