[F][SRU][PATCH 0/1] selftests/eeh: Skip ahci adapters

Po-Hsu Lin po-hsu.lin at canonical.com
Wed Feb 24 07:14:22 UTC 2021


[Impact]
When trying to run this test on P8 node entei with Focal kernel,
it will try to break 4 devices on Focal, and one of them is using
the AHCI driver which doesn't support error recovery:

  $ sudo ./eeh-basic.sh
  0000:00:00.0, Skipped: bridge
  0001:00:00.0, Skipped: bridge
  0020:00:00.0, Skipped: bridge
  0021:00:00.0, Skipped: bridge
  0021:01:00.0, Skipped: bridge
  0021:02:01.0, Skipped: bridge
  0021:02:08.0, Skipped: bridge
  0021:02:09.0, Skipped: bridge
  0021:02:0a.0, Skipped: bridge
  0021:02:0b.0, Skipped: bridge
  0021:02:0c.0, Skipped: bridge
  0021:0d:00.0, Added
  0021:0e:00.0, Added
  0021:0f:00.0, Skipped: bridge
  0021:10:00.0, Added
  0022:00:00.0, Skipped: bridge
  0022:01:00.0, Added
  Found 4 breakable devices...
  Breaking 0021:0d:00.0...
  0021:0d:00.0, waited 0/60
  0021:0d:00.0, waited 1/60
  0021:0d:00.0, waited 2/60
  0021:0d:00.0, waited 3/60
  0021:0d:00.0, waited 4/60
  0021:0d:00.0, waited 5/60
  0021:0d:00.0, waited 6/60
  0021:0d:00.0, waited 7/60
  0021:0d:00.0, waited 8/60
  0021:0d:00.0, Recovered after 9 seconds
  Breaking 0021:0e:00.0...
  0021:0e:00.0, waited 0/60
  0021:0e:00.0, waited 1/60
  ./eeh-basic.sh: 74: sleep: Input/output error
  0021:0e:00.0, waited 2/60
  ./eeh-basic.sh: 74: sleep: Input/output error
  ....
  ./eeh-basic.sh: 74: sleep: Input/output error
  0021:0e:00.0, waited 59/60
  ./eeh-basic.sh: 74: sleep: Input/output error
  0021:0e:00.0, waited 60/60
  ./eeh-basic.sh: 74: sleep: Input/output error
  0021:0e:00.0, Failed to recover!
  Breaking 0021:10:00.0...
  Skipping 0021:10:00.0, Initial PE state is not ok
  Breaking 0022:01:00.0...
  Skipping 0022:01:00.0, Initial PE state is not ok
  3 devices failed to recover (4 tested)
  ./eeh-basic.sh: 81: lspci: Input/output error
  ./eeh-basic.sh: 81: diff: Input/output error
  ./eeh-basic.sh: 82: rm: Input/output error
  ./eeh-basic.sh: 84: test: 3: unexpected operator

With the driver failed to recover, the system will start acting up.
  $ ls
  ls: command not found

And drop into a read-only state

[Fixes]
* bbe9064f30f06e ("selftests/eeh: Skip ahci adapters")

This is only affecting Focal and it can be cherry-picked.

[Test case]
Run the eeh-basic.sh script in tools/testing/selftests/powerpc/eeh/
on the affected P8 node, the test should pass without any issue.

[Where problems could occur]
This fix is limited to PowerPC testing tool and it's already in
Groovy kernel, it's unlike to cause any issue.

Michael Ellerman (1):
  selftests/eeh: Skip ahci adapters

 tools/testing/selftests/powerpc/eeh/eeh-basic.sh | 5 +++++
 1 file changed, 5 insertions(+)

-- 
2.7.4




More information about the kernel-team mailing list