From: Vladislav B. <vs...@vl...> - 2006-05-31 16:49:36
|
Nate, I fixed the missed spin_unlock() (committed). But I didn't find the reason of the first "sleeping function called from invalid context" message, looks like something went wrong a while above this message. If you have an easy way to trigger task management commands, it would be great if you checked all of them under load and report us your results. Vlad Nathaniel Clark wrote: > On Wednesday 31 May 2006 11:29, Vladislav Bolkhovitin wrote: > >>Nathaniel Clark wrote: >> >>>Vlad, >>> >>>Have you run into a problem with aborts hanging SCST? How thoroughly > > tested > >>>is the abort handling code? >> >>Unfortunately, there are no special tests for the task management, >>therefore it often quietly get broken. But about 3 months ago, when I >>touched it last time, ABORTs definitely worked. >> >> >>>I've had trouble with this, but never tracked it down, was wondering if > > other > >>>people had run into it. >> >>I have not heard about such problems for at least several months. Maybe, >>because since then the data flow control works so well, so ABORTs became >>VERY rare, ever under extremely high load. >> >>What kind of problem do you have? Are there any messages in the logs? > > > The most recent crash (last night) happened after a target reset (see > attached). > > The code I've got is checked out of sf CVS on 24th of May. > > I'll update and retest, but any leads on where to look would be great. > > - Nate > > >>> - Nathaniel Clark |