Bug 10652

Summary:	alpha: Radeon PCI ROM unavailable through sysfs
Product:	Drivers	Reporter:	Matt Turner (mattst88)
Component:	PCI	Assignee:	Richard Henderson (rth)
Status:	CLOSED OBSOLETE
Severity:	normal	CC:	alan
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	2.6.24	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	kernel config

Description Matt Turner 2008-05-08 15:34:11 UTC

Distribution: Gentoo
Hardware Environment: Alphaserver DS20L, Radeon X1550 PCI

Problem Description:

Access to the VGA BIOS ROM on my Alphaserver DS20L with a Radeon X1550
PCI does not work. The ROM located at /sys/bus/pci/devices/<pciid>/rom
is blank.

Before starting X, I issue the following commands from the said directory.

echo 1 > enable
echo 1 > rom

rom is empty at this point, which causes X to fail to load when
started. See FreeDesktop bug 15699 for X logs using radeon and
radeonhd drivers.

Only after 'priming' the rom file will X (with radeonhd) load correctly. To prime it, I issue 'cat rom > /dev/null'.

Comment 1 Matt Turner 2008-05-08 15:39:20 UTC

Created attachment 16076 [details]
kernel config

Comment 2 Matt Turner 2008-06-11 11:00:33 UTC

I believe this bug is a symptom of the same problem as in bug 10893. Please reassign to the appropriate place.

Comment 3 Jesse Barnes 2008-06-12 12:53:55 UTC

AFAIK you shouldn't need to do the "echo 1 > enable"; the X driver should take care of that for you (though maybe some versions of libpciaccess don't?).  The fact that you have to read it fully before it becomes available is weird though, alpha must handle the resources required by drivers/pci/rom.c differently than other platforms...

Comment 4 Matt Turner 2008-06-12 12:58:08 UTC

> AFAIK you shouldn't need to do the "echo 1 > enable"; the X driver should
> take
> care of that for you (though maybe some versions of libpciaccess don't?). 
> The
> fact that you have to read it fully before it becomes available is weird
> though, alpha must handle the resources required by drivers/pci/rom.c
> differently than other platforms...

Yes, it is very strange. Also, the version of X I'm using doesn't use libpciaccess (per bug 10893).

Oddly enough, when I put a Radeon 9100 in, I don't experience any of these problems and don't have to jump through hoops to make X start up. It only happens on my X1550.