Re: PowerMac G5 Quad 2.5 panics at radeondrm

> Date: Mon, 19 Dec 2016 11:06:31 -0800
> From: Bryan Vyhmeister <[email protected]>
> 
> On Thu, Dec 08, 2016 at 09:08:19PM +0100, Mark Kettenis wrote:
> > > Date: Thu, 8 Dec 2016 11:54:55 -0800
> > > From: Bryan Vyhmeister <[email protected]>
> > > 
> > > I just installed OpenBSD-current on my PowerMac11,2 which is the very
> > > last generation of PowerMac G5 systems. My particular one is the quad
> > > 2.5GHz model.
> > > 
> > > As a sidenote, I tried installing an Intel 535 SSD and an OCZ Vertex 4
> > > SSD and neither was detected by the SATA controller. I had read about
> > > that being an issue on some SSDs and that Samsung SSDs did not have this
> > > issue with older and slower SATA chipsets. I tried a Samsung 850 Pro and
> > > it worked fine.
> > > 
> > > I was able to install just fine with the installer but when I rebooted
> > > with bsd.mp, I get a panic right after cpu and memory detection due to
> > > radeondrm. A boot -c and disable radeondrm allows the system to boot
> > > fine but I would like to get this Radeon X1950 that I bought for this
> > > purpose working since in theory I could have a decent Xorg experience.
> > > Here is the panic transcribed by hand:
> > > 
> > > ...snip...
> > > memc0 at mainbus0: u4 rev 0x42
> > > kiic0 at memc0 offset 0xf8001000
> > > iic0 at kiic0
> > > lmtemp0 at iic0 addr 0x4a: ds1775
> > > maxds0 at iic0 addr 0x4b: ds1631
> > > maxtmp0 at iic0 addr 0x4c: max6690
> > > maxtmp1 at iic0 addr 0x4e: max6690
> > > "dart" at memc0 offset 0xf8033000 not configured
> > > openpic0 at memc0 offset 0xf8040000: version 0x4614 feature 7b0302 BE
> > > mpcpcibr0 at mainbus0 pci: u4-pcie
> > > pci0 at mpcpcibr0 bus 0
> > > radeondrm0 at pci0 dev 0 function 0 "ATI Radeon X1950" rev 0x00
> > > drm0 at radeondrm0
> > > radeondrm0: irq 3
> > > kern dsi on addr ffffffff iar 57c218
> > > panic: trap type 300 at 57c218 (pci_conf_write+0x1c) lr 590f50
> > > Stopped at      Debugger+0x10:  lwz r0,36(41)
> > >    TID    PID    UID     PRFLAGS     PFLAGS  CPU  COMMAND
> > > *    0      0      0     0x10000      0x200    0  swapper
> > > Debugger() at Debugger+0xc
> > > End of kernel: 0x904700
> > > end trace frame: 0x904700, count: 14
> > > 
> > > 
> > > I did not grab the dmesg from booting with radeondrm disabled (I can
> > > grab it later today if need be) but the dmesg from bsd.rd is at the
> > > bottom before I installed the Samsung SSD. Any idea what might be
> > > causing the panic?
> > 
> > Does the diff below help?
> > 
> > Index: radeon_kms.c
> > ===================================================================
> > RCS file: /cvs/src/sys/dev/pci/drm/radeon/radeon_kms.c,v
> > retrieving revision 1.48
> > diff -u -p -r1.48 radeon_kms.c
> > --- radeon_kms.c	8 Apr 2016 08:27:53 -0000	1.48
> > +++ radeon_kms.c	8 Dec 2016 20:07:37 -0000
> > @@ -478,7 +478,7 @@ radeondrm_attach_kms(struct device *pare
> >  	int			 is_agp;
> >  	pcireg_t		 type;
> >  	uint8_t			 iobar;
> > -#if !defined(__sparc64__)
> > +#if !defined(__sparc64__) && !defined(__macppc__)
> >  	pcireg_t		 addr, mask;
> >  	int			 s;
> >  #endif
> > @@ -550,7 +550,7 @@ radeondrm_attach_kms(struct device *pare
> >  		return;
> >  	}
> >  
> > -#if !defined(__sparc64__)
> > +#if !defined(__sparc64__) && !defined(__macppc__)
> >  	/*
> >  	 * Make sure we have a base address for the ROM such that we
> >  	 * can map it later.
> > 
> 
> Do you have any further ideas on how to go about troubleshooting this?

Try to determine what pci_conf_write() call is failing, and with what
arguments it is being called.  Note that these calls may hide as
pci_write_config_dword(), pci_write_config_word() or
pci_write_config_byte() calls.