Messages in this thread | | | Date | Tue, 17 Mar 1998 15:46:36 +0100 | From | ralf@uni-kobl ... | Subject | Re: x86 memsize > 1Gb patches. |
| |
On Mon, Mar 16, 1998 at 10:07:05PM +0100, Pavel Machek wrote:
> > > According to Martin Mares: > > > > Functions such as phys_to_virt et. al. are going to be horribly broken when > > > > we compile modules with a different CONFIG_MAX_MEMSIZE > > > > > > Perhaps the link name of such functions should have the value of C_M_M > > > appended. Thus, phys_to_virt would link as phys_to_virt_1024. > > > > What about just not inlining it anymore ;-) > > Bad idea, phys_to_virt(x) translates to (PAGE_OFFSET | x). You do not > want call overhead on single instruction. (Which is probably exactly > as long as call instruction).
Do the calculation using inline assembler, collect the addresses of the add instructions in a special ELF section. Patch the add instructions on startup and free the section.
Ralf
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |