>Hm, is there any instance where gcc will open-code a SWP instruction?
No. The only place I know of where it's used is LinuxThreads.
>Or are there any syscall handlers which are likely to use SWP (presuming
>the majority of the kernel mutexes never get paged out, and thus
>ignoring any non-aborting variety)?
I don't think the kernel ever uses SWP to touch anything that's entitled to be
paged, either. The whole of its own address space is permanently mapped, so
it's only accesses to user pages that might fault.
p.
_______________________________________________
http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
