Comment 7 for bug 1258231

Revision history for this message
David Anderson (davea42) wrote : Re: NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context

12 core machine running 13.10, Saucy Salamander.
Never saw these problems before 13.10 .

3.11.0-14-generic #21-Ubuntu SMP X86_64
319.32 nvidia gets kernel error
319.60 nvidia does not (so far) get kernel error.

26 November 9AM: software&updates->additional drivers->
       switched to the nvidia 319.60 'proprietary' driver (from
       319.32 'proprietary tested'). Rebooted.

Turned on Seti@home gpu processingand seti@home did a number of GPU
runs without difficulty. That is 10 days, no problem.

5 December 4PM: as a test, switched back to the 319.32 nvidia
       driver. Rebooted.

7 December 4:16AM: Got kernel warnings, the first is.
       /var/log/syslog:Dec 7 04:16:34 Dseti3 kernel: [129416.201395] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
      4800 lines like that. Including a few 'CPU#0 stuck'.

       So that is just 36 hours till a problem arose.
       10AM: Switched back to 319.60 driver. Rebooted.
       Resuming seti@home

This test may mean nothing, but 319.60 does let me do useful GPU work, so far.