cvs commit: src/sys/i386/bios apm.c

[John Baldwin]

On Wednesday 14 November 2007 05:01:32 am Kris Kennaway wrote:
> Julian Elischer wrote:
> > julian      2007-11-14 05:43:55 UTC
> > 
> >   FreeBSD src repository
> > 
> >   Modified files:
> >     sys/i386/bios        apm.c 
> >   Log:
> >   Apply the same sort of locking done in
> >      sys/dev/acpica/acpi.c rev 1.196 a while ago:
> >   
> >   Grab Giant around calls to DEVICE_SUSPEND/RESUME in
> >   acpi_SetSleepState().
> >   If we are resuming non-MPSAFE drivers, they need Giant held for them.
> >   This may fix some obscure suspend/resume problems.  It has fixed keyrate
> >   setting problems that were triggered by cardbus (MPSAFE) changing the
> >   ordering for syscons resume (non-MPSAFE).  Also, add some asserts that
> >   Giant is held in our suspend/resume and shutdown methods.
> >   
> >   Submitted by: Marko Zec
> >   
> >   Revision  Changes    Path
> >   1.149     +10 -0     src/sys/i386/bios/apm.c
> 
> Why are we adding new unconditional giant acquisitions to the tree? 
> Devices indicate whether or not they are mpsafe, why can't this be made 
> conditional?

Actually, we have no way in new-bus to mark new-bus operations like bus 
enumeration, etc. MPSAFE, and much of it is not (e.g. 
attach/detach/suspend/resume, etc. functions).  Additionally, making this 
conditional would require a scheme where you acquire Giant when you walk down 
the tree into a non-MPSAFE driver which would add complication to many (if 
not all) bus drivers.  Considering the rarity of suspend/resume (not a 
critical-path) the overhead of adding all that may not be worth it.  At some 
point when we do have locking for all device driver operations this can be 
removed, but it won't really hurt anything to leave it as it is for now.

-- 
John Baldwin