Bug 61095

Summary: XFree86 w/ AGP radeon can hang system on ES45
Product: [Retired] Red Hat Linux Reporter: Jeff Wiedemeier <jeff.wiedemeier>
Component: kernelAssignee: Bill Nottingham <notting>
Status: CLOSED CURRENTRELEASE QA Contact: David Lawrence <dkl>
Severity: medium Docs Contact:
Priority: medium    
Version: 7.3CC: bhu, rvokal
Target Milestone: ---   
Target Release: ---   
Hardware: alpha   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2002-03-15 23:19:45 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
acquire page_table_lock around page remap none

Description Jeff Wiedemeier 2002-03-13 15:37:32 UTC 
Description of Problem:

Using XFree86 with 'modprobe agpgart' patch, starting XFree86 on an AGP
AlphaServer ES45 can hang the system. 

Version-Release number of selected component (if applicable):


How Reproducible:

sometimes

Steps to Reproduce:
1. start x
2. if it works, reboot and start x again
3. 

Actual Results:

some of the time the system hangs, other times X starts with DRI correctly

Expected Results:

X start with DRI always

Additional Information:

Haven't been able to reproduce the problem with this line in /etc/modules.conf:

pre-install radeon (/sbin/modprobe agpgart; /bin/true)

which is the way that I was loading the agpgart driver without the 'modprobe
agpgart' patch
Comment 1 Phil Copeland 2002-03-13 20:27:46 UTC 
I'm guessing here ...

Kudzu needs to detect 1) AGP exists, 2) a radion card exists
shovel in the above line

Phil
=--=
Comment 2 Bill Nottingham 2002-03-13 21:40:17 UTC 
That's just a hack. The normal way X works (i.e., to load the agpgart module)
should 'just work'; it indicates a bug somewhere in the kernel drivers.
Comment 3 Jeff Wiedemeier 2002-03-15 21:11:12 UTC 
my bad.

I found the problem (pmd/pte_alloc without page_table_lock held).
I will attach a kernel patch this afternoon.
Comment 4 Jeff Wiedemeier 2002-03-15 21:31:30 UTC 
Created attachment 48705 [details]
acquire page_table_lock around page remap
Comment 5 Phil Copeland 2002-03-15 22:28:56 UTC 
Humm,...
partially successful.
core_titan.c no longer has titan_remap_area_pte()
nor titan_remap_area_pmd()

Wanna give that another shot?

I've made the current srpm available on alpha7.crl.dec.com:~bryce (look for
patch126 in the spec file)
(kernel-2.4.9-32.1.src.rpm)

ie
rpm -ivh kernel-2.4.9-32.1.src.rpm
cd /usr/src/redhat/SPECS
rpm -bp kernel-2.4.spec
cd /usr/src/redhat/BUILD/kernel-2.4/linux

That at least will give you a 'native' src tree to work against.

Phil
=--=
Comment 6 Jeff Wiedemeier 2002-03-15 23:19:40 UTC 
the _remap functions are added to core_titan in the main titan patch (133), so 
this one needs to be applied after that.
Comment 7 Beth Uptagrafft 2002-04-03 22:28:24 UTC 
This will be fixed in the next release.