Re: glibc developers refuse to support user land fake FS syscalls

Alexander S A Kjeldaas (Alexander.Kjeldaas@fast.no)
Mon, 5 Jul 1999 15:51:32 +0200


On Mon, Jul 05, 1999 at 02:04:22PM +0100, Alan Cox wrote:
> > I know this won't help the current situation, but for the future, how
> > about changing the way syscalls are invoked. Instead of using int
> > 0x80, do a jump to an absolute location which in turn does the int
> > 0x80. The page that includes the instructions to do system-calls
>
> This seems like making everyone may for your peculiar needs. People who
> need this degree of weirdness can use ptrace hooks, add a personality to
> the kernel or build a new glibc with their jump/tlb/cache miss penalty;
>

ptrace is basically an upcall. A personality might or might not be
something that should be used. It is possible to support both types
of syscalls at the same time.

Performance-wise I don't think you'll loose much. The
performance-degradation due to TLB or cache-misses are probably very
low. It would use some slots in a jump-cache which might have some
effect.

An optimization would be to use the same mechanism, but have the
default "invocation page" exported by the kernel binary patch the
invocation point to do a int 0x80 directly.

astor

-- 
 Alexander Kjeldaas, Fast Search & Transfer, Trondheim, Norway

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/