Red Hat Bugzilla – Bug 164500
RHEL3, 2.4.21-32.0.1.ELsmp: Nvidia 7664 complains about 'bad caching on address 0xf2f50000 ...'
Last modified: 2007-11-30 17:07:08 EST
Escalated to Bugzilla from IssueTracker
Ingo, I think this might be related to bug 163239.
Any progress on this? I have a customer who is running into lockups on their
system, and they are reporting that they are seeing the bad cacheing messages.
nVidia's response is to update your kernel :) How likely is this issue related
to bug 163239 - i.e. is it likely that it's fixed in U6?
Neither problem has been addressed in U6 (which is already in beta).
Anything else on this ? The customer is getting pretty upset.
since being assigned this bug yesterday, I have tried to find some hardware to
replicate this issue but so far no success. a search of the inventory system did
not produce any results for Nvidia cards (but I believe inventory is broken as
we must have a system somewhere with at least one Nvidia card). So I still need
to locate some hardware somehow...
do we know what the PCI deviceid is? can we maybe get a "lspci" output?
Created attachment 119669 [details]
lspci output from machine exhibiting problem
As requested, lspci output from a machine exhibiting the issue
Apparently, there is bogus info in comment #17. According to comment #28,
this is an i686-only problem. Fixing "hardware" field to reflect this.
Removing link to IT 75537, which is a different problem against a
different release. Brian, please ignore comment #31.
Hello, Terence. I believe you've got the wrong bit. The discrepency
according to the messages is that bit 9 was not set (0x200), and this
bit is _PAGE_BIT_NX (defined in include/asm-i386/pgtable.h).
Does the hardware that produces the warnings contain cpus with NX support?
(Eric, please propagate this question to <firstname.lastname@example.org> and post
the answer here. You might need to check directly with the customer.)
The problem reported in this bugzilla is due to an incorrect assumption
in NVidia's driver (which is not part of RHEL3) that expects the soft-NX
bit to be set in PTEs on non-NX-capable systems.
Closing as NOTABUG.
SteffenMann, in regards to BZ comment #47, this problem is *not* the
result of a RHEL3 kernel bug, which is why this is closed as NOTABUG.
The warning is the result of a bug in NVidia's driver (which is external).
Jeffrey Layton, I don't know how why your comment #48 about BZ 171129
has landed in BZ 164500 (via IT 75455). If you've got ITs incorrectly
cross-linked, please fix them. Thanks.