From Bugzilla Helper: User-Agent: Mozilla/5.0 Galeon/1.2.5 (X11; Linux i686; U;) Gecko/20020712 Description of problem: In kernel-2.4.18-7.92 the VIA agpgart will reboot my computer instantly. It will do it both compiled in and as a module. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. modprobe agpgart Actual Results: Computer reboots Expected Results: Computer not to reboot and code to load correctly Additional info: Limbo beta2 Athlon 1700+ Abit KT7A-RAID(VIA KT133A chipset) mem=nopentium
linux-2.4.19-final works fine with VIA agpgart compiled in.
linux-2.4.19-final-ac4 works fine with VIA agpgart compiled in. hehe, now for the odd part kernel-2.4.18-7.93 i686 works fine with VIA agpgart as a module, but kernel-2.4.18-7.92 i686 and athlon both reboot on loading VIA agpgart module
ok, this has gone from odd to bizarre kernel-2.4.18-7.93 athlon works fine with VIA agpgart as a module till you actually use it by startin XFree86 kernel-2.4.18-7.93 athlon with VIA agpgart compiled in reboots as soon as it loads the agpgart code kernel-2.4.18-7.93 athlon works fine with VIA agpgart as a module if I remove mem=nopentium from grub.conf
Created attachment 69244 [details] Diff between 2.4.19-final and 2.4.18-7.93 agpgart_be.c
I checked drivers/char/agp/agp.h drivers/char/agp/agpgart_be.c drivers/char/agp/agpgart_fe.c include/linux/agpgart.h include/linux/agp_backend.h Of those files only agpgart_be.c had any differences between 2.4.19-final and 2.4.18-7.93. I have attached the diff above.
Below I have attached a diff between 2.4.19-final and 2.4.18-7.93 arch/i386/kernel/setup.c setup.c contains the setup for the mem=nopentium option
Created attachment 69245 [details] Diff between 2.4.19-final and 2.4.18-7.93 arch/i386/kernel/setup.c
Patch 1490 aka linux-2.4.19-pageattr-rc1-9.patch seems to be responsible for the agpgart_be.c changes and the point of the patch seems to be a final fix for the AMD agp bugs which relates to mem=nopentium This comment in the patch is telling. /* no more special protections in this 2/4MB area - revert to a large page again. */
This should be fixed in current update kernels, right ?