Re: cx18-0: ioremap failed, perhaps increasing __VMALLOC_RESERVE in page.h

[Andy Walls <awalls@xxxxxxxxx>]

On Thu, 2008-05-15 at 08:57 -0400, Michael Krufky wrote:
> Andy Walls wrote:
> > On Mon, 2008-05-12 at 10:48 -0400, mkrufky@xxxxxxxxxxx wrote:
> >   
> >> Steven Toth wrote:
> >>     
> >>> Steven Toth wrote:
> >>>       
> >>>>>         if (cx->dev)
> >>>>>                 cx18_iounmap(cx);
> >>>>>           
> >>>> This doesn't feel right.
> >>>>         
> >>> Hans / Andy,
> >>>
> >>> Any comments?
> >>>       
> >> For the record, I've tested again with today's tip ( d87638488880 ) -- 
> >> same exact behavior.
> >>
> >> When I load the modules for the first time, everything is fine.
> >>
> >> If I unload the cx18 module, I am unable to load it again, the same 
> >> error is displayed as I posted in my original message.
> >>     
> >
> > Mike,
> >
> > Could you apply the attached patch and run this test
> >
> > 	(precondition: cx18.ko hasn't been loaded once yet)
> > 	# cat /proc/iomem /proc/meminfo > ~/memstats
> > 	# modprobe cx18 debug=3
> > 	# cat /proc/iomem /proc/meminfo >> ~/memstats
> > 	# modprobe -r cx18
> > 	# cat /proc/iomem /proc/meminfo >> ~/memstats
> > 	# modprobe cx18 debug=3
> > 	# cat /proc/iomem /proc/meminfo >> ~/memstats
> >
> > and provide the contents of dmesg (or /var/log/messages) and memstats?
> >
> > The patch will let me see if the contents of cx->enc_mem are bogus on
> > iounmap() and if iounmap() is even being called.
> >
> > I also want to verify that "cx18 encoder" doesn't get removed
> > from /proc/iomem and that "VmallocUsed" doesn't return to it's previous
> > size when the module is unloaded.  That would show that the iounmap()
> > fails.
> >
> > I'd also want to ensure there is no overlap in /proc/iomem with "cx18
> > encoder" and something else.  The kernel should prevent it, but I want
> > to make sure.
> >
> >
> > (Hopefully the patch applies cleanly, the line numbers won't quite match
> > up with the latest hg version.)
> >
> > Regards,
> > Andy
> >
> >   
> >> Regards,
> >>
> >> Mike
> >>
> >>     

> 
> 
> Andy,
> 
> I'm out of town right now, and things will be hectic when I get 
> back....  I most likely won't get to this until the middle of the week, 
> perhaps even next weekend.

That's OK my next week is booked with travel to a few places as well.

FYI, some amplifying information on the problem.  Your original dmesg
log showed an error that comes from

mm/vmalloc.c:__get_vm_area_node()

that can only be reached by two "goto out;" statements.

I'll be reading The last half of Chapter 7 of
_Understanding_the_Linux_Kernel_ while away next week to try and grok
what's going on, since the Chapter still looks applicable to the newer
kernel code.

-Andy

> -Mike
> 

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@xxxxxxxxxx?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list