Four more s390 patches.

["Martin Schwidefsky" <schwidefsky at de dot ibm dot com>]

Hi,
here are four more patches for the s390 backend:

1) Performance tests have showed that the current mvc loop is good
for small block but the old mvcle implementation has been better for
large blocks.

2001-07-12  Martin Schwidefsky  <schwidefsky@de.ibm.com>

     * sysdeps/s390/s390-32/bcopy.S: Use mvcle for big blocks
     (> 64K) and a mvc loop for small blocks.
     * sysdeps/s390/s390-32/memcpy.S: Likewise.
     * sysdeps/s390/s390-64/bcopy.S: Likewise.
     * sysdeps/s390/s390-64/memcpy.S: Likewise.

(See attached file: bcopy-s390.diff)

2) A POSIX test complained about the names in the __jmp_buf structure.

2001-07-12  Martin Schwidefsky  <schwidefsky@de.ibm.com>

     * sysdeps/s390/s390-32/bits/setjmp.h: Add leading underscores
     to the entries in the __jmp_buf structure.
     * sysdeps/s390/s390-64/bits/setjmp.h: Likewise.

(See attached file: jmpbuf-s390.diff)

3) The fallback to the old style mmap is broken because the compare
for ENOSYS is gone against the wrong register.

2001-07-12  Martin Schwidefsky  <schwidefsky@de.ibm.com>

     * sysdeps/unix/sysv/linux/s390/s390-32/mmap64.S: Correct the
     test for ENOSYS.

(See attached file: mmap64-s390.diff)

4) There is a mismatch between the kernel definition of struct sigaction
and the glibc definition.

2001-07-12  Martin Schwidefsky  <schwidefsky@de.ibm.com>

     * sysdeps/unix/sysv/linux/s390/s390-64/bits/sigaction.h: Reorder
     fields in the sigaction struct to match the definition in the
     kernel.

(See attached file: sigaction-s390.diff)

blue skies,
   Martin

Linux/390 Design & Development, IBM Deutschland Entwicklung GmbH
Schönaicherstr. 220, D-71032 Böblingen, Telefon: 49 - (0)7031 - 16-2247
E-Mail: schwidefsky@de.ibm.com

=?iso-8859-1?Q?bcopy-s390.diff?=

=?iso-8859-1?Q?jmpbuf-s390.diff?=

=?iso-8859-1?Q?mmap64-s390.diff?=

=?iso-8859-1?Q?sigaction-s390.diff?=