Re: version information
On Wed, 2003-02-12 at 18:11, Florian Lohoff wrote:
> On Wed, Feb 12, 2003 at 12:30:10AM +0100, Guido Guenther wrote:
> > Hmm...shouldn't ll/sc be handeld by the emulation in the kernel for R3k?
> > The source tells me someting different. Do we agree that this is a
> > kernel and not a glibc issue? Can someone with a R3k Decstation comment on
> > this?
> The glibc switched from sysmips(MIPS_ATOMIC_SET, ...) to ll/sc in its
> default (even for -mips1) so the upgrade breaks on machines without
> ll/sc emulation.
>
> As we should have ll/sc emulation for all non SMP architectures this
> is definitly a kernel issue.
well, since we found out, that it's the kernel's fault... has anyone
already dared to fix it and provide some patch? :-)
after a quick look at arch/mips/kernel/traps.c (2.4.20) it seems that
it's just a matter of adding a few addition switch-cases to do_cpu(): ?
[..]
bad_cid:
#ifndef CONFIG_CPU_HAS_LLSC
switch (mips_cpu.cputype) {
case CPU_TX3927:
case CPU_TX39XX:
do_ri(regs);
return;
}
#endif
[..]
...and close this bug; if provided with a new kernel image I would be
quite happy to test it, as building on a r3000a @ 20mhz with low memory
isn't quite fun...
> I would guess that a new kernel is needed. Probably we should have a
> check in glibc install to prevent users from marking their systems void.
regards,
--
Herbert Valerio Riedel / Phone: (EUROPE) +43-1-58801-18840
Email: hvr@hvrlab.org / Finger hvr@gnu.org for GnuPG Public Key
GnuPG Key Fingerprint: 7BB9 2D6C D485 CE64 4748 5F65 4981 E064 883F 4142
Reply to: