[uClibc]uClibc/mipsel

[John Rigby]

I had a similar problem to this and it turned out to be mmap.  The 
implementation in libc/sysdeps/linux/common/syscalls.c is wrong for mips.  
This works better: 

_syscall6 (__ptr_t, mmap, __ptr_t, addr, size_t, len, int, prot,
           int, flags, int, fd, __off_t, offset);

This works fine for me (on mips).  I suppose the proper solution is to add a 
_mmap.c to libc/sysdeps/linux/mips.  (like v850 does).

John

On Thursday 14 February 2002 23:08, Geoffrey Espin wrote:
> Erik,
>
> > > Using the latest CVS and simply dumping
> > >     uClibc/libc/sysdeps/linux/mipsel
> > > for
> > >     uClibc/libc/sysdeps/linux/mips
> >
> > Hmm. I already did that...
>
> Damm, I used "cvs up -P -d" and I thought it would've nuked the junk.
> (Oh, yeah, I've made that mistake before.)
>
> > Cool. I spent about two hours before the 0.9.9 release getting
>
> And really, really appreciated.
>
> > So feel free to pass on any changes you have that look proper,
> > and I'll review and include them.
>
> Will do.
>
> > That seems really really wierd.  Sounds like a stack alignment
> > problem.  Taking a look in glibc-2.2.4/sysdeps/mips/elf/start.S
> > I see the glibc folk are doing some stack alignment stuff that I
> > didn't do in libc/sysdeps/linux/mips/crt0.S...  Perhaps we need
> > to do something similar....
>
> That's what I needed, thanks, will start looking tomorrow.
>
> > > Separately: when I run the 'malloc' test, I get:
> > > This looks like VM OOM killer... is that what should happen, or is
> >
> > I guess it does depend on how much memory you have.  I can
> > certainly see how it could exhaust all memory on a low memory
> > system.,,  I expect that dropping MAX_SIZE, N_ALLOCS, and N_PTRS
> > down to much smaller numbers in test/malloc/malloc.c should avoid
>
> Okay... hmmm, would be nice to auto-detect this...
>
> Geoff