Re: shutdown panics on i386

[Kurt Schreiner <ks%ub.uni-mainz.de@localhost>]

On Sat, May 02, 2009 at 01:04:35AM +0200, David Young wrote:
> On Wed, Apr 29, 2009 at 04:31:54PM +0100, Patrick Welche wrote:
> > On Wed, Apr 22, 2009 at 03:29:11PM -0500, David Young wrote:
> > > On Wed, Apr 22, 2009 at 10:36:47AM +0200, Kurt Schreiner wrote:
> > > > Hi,
> > > > 
> > > > with a kernel build from -current source last updated yesterday I see
> > > > the following while shutting down or rebooting the system (an old Dell
> > > > C610 Notebook):
> > > > 
> > > > ...
> > > > cardsolt0: detached
> > > > wd0: detached
> > > > extend 'iomem' (0x0 - 0xffffffff), flags = 0x3
> > > >         0x0 - 0x9fbff
> > > >         0xa0000 - 0xbffff
> > > >         0xc0000 - 0xc7fff
> > > >         0x100000 - 0x2ffd2fff
> > > >         0x40000000 - 0x40000fff
> > > >         0x40001000 - 0x40001fff
> > > >         0x40002000 - 0x40002fff
> > > >         0x40003000 - 0x40003fff
> > > > extend_free: start 0x40001000, end 0x40000fff
> > > > panic: extend_free: region not found
> > > > fatal braekpoint trap in supervisor mode
> > > > trap type 1 code 0 eip c014fa74 cs 8 eflags 246 cr2 bbb43ef0 ilevel 0
> > > > Stopped in pid 619.1 (halt) at netbsd:braekpoint+0x4: popl      %ebp
> > > > db{0}>
> > > > 
> > > > And no, ther's no typo in the extend_free: start... line! (Even if the
> > > > whole mess is hand copied from the notebook screen ;-)
> > > 
> > > Please send your dmesg and a backtrace.
> > > 
> > > That is indeed a strange extend_free line, thanks for calling
> > > attention to it!
> > 
> > Surprisingly I have a different cardbus slot to Kurt:
> > 
> > cbb0 at pci1 dev 3 function 0: Ricoh 5C476 PCI-CardBus Bridge (rev. 0xac)
> > cbb1 at pci1 dev 3 function 1: Ricoh 5C476 PCI-CardBus Bridge (rev. 0xac)
> > 
> > Solved by adding
> > 
> > options     PCI_ADDR_FIXUP
> > 
> > (before only PCI_BUS_FIXUP was needed...)
> 
> Kurt, Patrick, Matt,
> 
> I figured it out.  Please try the attached patch.
> 
> You shouldn't need PCI_ADDR_FIXUP any more.  It was helpful that you
> told me that it solved the problem, though, thanks!
> 
> Dave
> 
> -- 
> David Young             OJC Technologies
> dyoung%ojctech.com@localhost      Urbana, IL * (217) 278-3933

Content-Description: pccbb.detach
> Index: pccbb.c
> ===================================================================
> RCS file: /cvsroot/src/sys/dev/pci/pccbb.c,v
> retrieving revision 1.185
> diff -p -u -u -p -r1.185 pccbb.c
> --- pccbb.c   2 Apr 2009 00:09:33 -0000       1.185
> +++ pccbb.c   1 May 2009 22:59:20 -0000
> @@ -643,7 +643,7 @@ pccbb_pci_callback(device_t self)
>                   (unsigned long)pci_conf_read(pc,
>                   sc->sc_tag, PCI_SOCKBASE)));
>  #endif
> -             sc->sc_flags |= CBB_MEMHMAPPED;
> +             sc->sc_flags |= CBB_MEMHMAPPED|CBB_SPECMAPPED;
>       }
>  
>       /* clear data structure for child device interrupt handlers */
Yep, that does the trick! ;-) (And has no negative influences on other
systems I run a kernel with this patch on.)

The "new" output from "shutdown -p"-ing the Dell C610:

login: May  2 09:20:56 hc610-1 syslogd[117]: Exiting on signal 15
syncing disks... done
unmounting file systems... done
cardbus2: detached
pcmcia1: detached
cardbus1: detached
pcmcia0: detached
cardbus0: detached
cardslot1: detached
cardslot0: detached
wd0: detached
cbb1: detached
cbb0: detached
audio0: detached
atabus1: detached
atabus0: detached
uhub0: at usb0 (addr 1) disconnected
uhub0: detached
usb0: detached
uhci0: detached
pchb0: detached
attimer0: detached
audio1: detached
acpi0: entering state 5

And here "poweroff" happens as expected...

BTW: I have and had only PCI_BUS_FIXUP defined in the kernel config,
_no_ PCI_ADDR_FIXUP...

Thanks for the fix!

Kurt