On Sun, 29 Mar 2015, Walter Daugherity wrote:
> No, 5.7-current snapshot still does not fix this bug.  I have tried 
> 2015-Mar-18 and 2015-Mar-26 snapshots, and both still crash under 
> network logging load (e.g., with pppd kdebug and debug enabled, log on 
> to a remote computer and run the "top" command and the system crashes).

Since you've updated across a significant change in the involved code, 
please provide a trace from a panic with the updated build.  (The trace 
has certainly changed, as pool_do_put() doesn't exist anymore, so the 
problem code path involved in the panic has changed.)


...
> The problem may be in syslogd or in the file system,  <...>

No, it absolutely cannot be a bug in syslogd, and from the trace is not a 
bug in the file system, but rather--as kettenis@ explained--the failure to 
block interrupts as necessary in some critical section.



Philip Guenther

Reply via email to