Bug 3063

Summary:	DRI broken for 32-bit applications on x86_64 platform
Product:	Fedora	Reporter:	Porcelain Mouse <porcelain_mouse>
Component:	xorg-x11-drv-catalyst	Assignee:	Nobody is listening <orphaned>
Status:	RESOLVED WONTFIX
Severity:	normal
Priority:	P5
Version:	19
Hardware:	All
OS:	GNU/Linux
namespace:

Description Porcelain Mouse 2013-12-08 01:40:13 CET

I'm having trouble troubleshooting this issue.  AFAIK, all device permissions and libs are install and found by 32-bit versions of programs.  But, the 32-bit programs think DRI isn't working while 64-bit stuff seems to work fine.

This message from 'LIBGL_DEBUG=verbose glxinfo32' seems to occur most immediately before ultimately failing to find DRI:

ukiOpenByBusid: ukiOpenMinor returns -1003
libGL error: open uki failed (Operation not permitted)

strace shows a bit more detail.  The problem is foreshadowed by this error early on:

ioctl(4, 0xc0246400, 0x8be5c78)         = -1 EFAULT (Bad address)

df=4 is /dev/ati/card0.  Both 64 and 32 bit versions open this device successfully.  But the 32-bit version sends it a bad ioctl.

But, things really go bad after re-opening the device a bit later.  Again, both 64 & 32 bit versions open() successfully, but then about 60 ioctl()s are issued and none succeed for the 32-bit version.

...
ioctl(4, 0xc044646e, 0x8c2be18)         = -1 EFAULT (Bad address)
...

Comment 1 Nicolas Chauvet 2013-12-08 10:36:45 CET

You would need xorg-x11-drv-catalyst-lib.i686, but anyway nobody listen this bug as the package is unmaintained in RPMFusion. It is already removed from f20 tree.

Comment 2 Porcelain Mouse 2013-12-08 22:10:50 CET

No, I have that package installed, already.  Could that package be bad?

I read about this package being unmaintained, but I assumed that was temporary.  It would be nice to know a little more.  Is there a place were the future of ATi driver support is being discussed?