Messages in this thread |  | | Subject | Re: [PATCH] change name of rep_nop | Date | Tue, 9 Oct 2001 13:15:49 +0100 (BST) | From | Alan Cox <> |
| |
> consider modules as well as the kernel. modutils 2.4.8 added support > for ppc archdata to allow dynamic patching of modules using the ftr > data. There also has to be code in kernel/module.c::module_arch_init() > to take the archdata and do whatever is required.
I am considering the possibility for the x86 ppro fix but firstly I want to be sure the spin_unlock change is necessary. The PCI one is much less of an impact since we don't go around doing it all the time.
> If anybody starts doing dynamic patching, please let me know so I can > handle modutils and module_arch_init().
For testing purposes it won't be needed thankfully - a non nop patched module might just be a tiny bit slower.
Anyway firstly I have to reasonably show PPro errata 66/92 are what is causing the odd weird ppro crash (the evidence so far is the xchg based unlock helps). The WC v UC misordered store one is a proven case, as is the 0x70000000->0x7003FFFF wbinvd hang which turns up on some ppro boxes that never got the bios E820 fixup
Alan - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
|  |