[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: cache problems with non-blocking fork code



Steve Woodford writes:
> I've just got around to patching my VMiNT kernel with the non-blocking
> fork code, plus MH's "quick-fix" posted to the list shortly after.

btw there were at least 2 quick-fixes for the fork stuff...
> 
> I seem to be having some cpu cache problems, in that the symptoms are
> identical to those that I had when I neglected to do a cpush() after
> SCSI transfers in VTOS.
> 
> Is there a known problem with the non-blocking fork mods? I've not
> spent heaps of time debugging this yet, since I thought I'd ask first.
> But the symptoms disappear completely when I disable the cpu cache....

 i cant test this here (just a lowly 68000), but have you tried
just putting a cpush() after the quickswap() call(s)?  if that doesn't
fix it then it then your problem can't be a missing cpush... :)
> 
> The symptoms include failure of some programs to run, eg. syslogd on
> startup; 'ls' from the shell prompt lists an empty directory when I
> know the directory is not empty; various other instabilities which go
> away when the cache is disabled. These problems did not occur until I
> patched the non-blocking fork code in.

 but what is definitely broken now is vfork().  at least for me it is...

 ATB,
	Juergen