🚀 go-pugleaf

This announces the availability of version 1.2.13pl9 of Linux/68k. The
source and a precompiled kernel are in /pub/Linux/LOCAL/680x0 on
ftp.uni-erlangen.de and its mirrors.

Changes against the last version:

 - Enabling sync transfers for all targets at a wd33c93 (argument 0)
   works again. (Matthias.Welwarsky@ppp.th-darmstadt.de)

 - Lots of smaller fixes by Andreas Schwab:

    o The Atari Lance and BioNet drivers now check whether they run on
      an Atari.
    o Serial driver modules can now be loaded by kerneld.
    o On address errors (odd address of instruction), SIGBUS instead of
      SIGSEGV is generated now.
    o Fixed a typo in the /proc/*/fd readdir function: it returned the
      pid instead of the file number.
    o Some compiler generated symbols (memcpy, memset, ...) don't need
      version control in ksyms.c.
    o fs/umsdos/ioctl.h fixed
    o strtok() doesn't use a static variable anymore, this one was
      copied into each object file.
    o Several Atari modules (serial drivers, BioNet) are built only if
      CONFIG_ATARI is set.
    o Removed gcc-extension in <linux/kd.h> for programs compiled with
      -ansi.
    o Removed some unnecessary includings of config.h
    o New file <linux/compile.h>, to avoid version.h changing on every
      compile.
    o New <linux/uio.h> for GNU libc.
    o arch/m68k/atari/ksyms.c renamed to ataksyms.c
    o Oops: in the bus error handler, a "break" was missing after the
      '060 handler.
    o Wrong stop condition for dump_stack() fixed.

 - ATAPI CD-ROM are supported now. This works at least on the Amiga,
   for Atari further tests are needed (byte swap).
   (Geert.Uytterhoeven@cs.kuleuven.ac.be)

 - Reading AFFS directories should work again, and reading multiple
   entries at a time is faster now. (dorchain@cscip.uni-sb.de)

 - Fixed version Printf problem in amiboot/amiga-lilo.
   (Geert.Uytterhoeven@cs.kuleuven.ac.be)

 - In the OPOST speedup of ttys, die OLCUC and O_NONBLOCK flags were
   ignored. (guenther@pool.informatik.rwth-aachen.de)

 - The "old" Pearl 8x8 font is included again, in a VGA-ized version,
   i.e. with IBM-encoding. Its name (for setfont) is "PEARL8x8" and it
   is choosen as default font for Amigas if vertical resolution is
   below 400 lines. (jshiffle@netcom.com & myself)

 - The Amiga built-in serial driver should work as a module now.
   (dorchain@cscip.uni-sb.de)

 - Some changes to symbol tables to make CONFIG_MODVERSIONS work.
   (dorchain@cscip.uni-sb.de)

 - Added log levels to printks in amiflop.c (dorchain@cscip.uni-sb.de)

 - For A1200 IDE, read the status before acknowleding an interrupt.
   This makes some drives work better. (jongk@prac.cs.utwente.nl)

 - For the Atari, kstat.interrupts (int counts by level) is maintained
   again (specially for vmstat :-)
   (schwab@issan.informatik.uni-dortmund.de)

 - Fixed a bug in the shared memory implementation with bad swap
   entries. This should have been the cause for the reported IPC
   problems. (schwab@issan.informatik.uni-dortmund.de)

 - New Zorro board definition (Geert.Uytterhoeven@cs.kuleuven.ac.be)

 - The Atari SLM driver didn't compile due to a missing header (myself)

 - The Amiga and Atari sound drivers were unified into a general DMA
   sound driver for both machines (drivers/sound/dmasound.c). For the
   Amiga, there are several new things and the driver is beta. For
   Atari, everything should work as usual.
   
    o Added support for the A-law and 16-bit little endian audio
      formats.
    o At least playing Sun audio files on Amiga works now, just copy
      them to /dev/audio.
    o Using the system beep may cause weird results on Amiga.
    o Amiga: Stereo and 16-bit (pseudo 14-bit) samples may work, still
      need to verify that.
    o The command line option to configure driver features has been
      renamed from "atasound=" to "dmasound=". (myself)
    o Return correct error numbers on select (not implemented) and bad
      minors. (schwab@issan.informatik.uni-dortmund.de)

   (Geert.Uytterhoeven@cs.kuleuven.ac.be)

   I've just renamed CONFIG_SOUND into CONFIG_DMASOUND to avoid
   conflicts in future with the PC sound drivers...

 - Freshly allocated kernel pointer tables are initialized now. This
   should avoid the spurious "bad pmd" messages.
   (schwab@issan.informatik.uni-dortmund.de, myself)

 - There are now max. 16 kernel root/pointer tables. An extra page is
   allocated if the one from head.S is full. This relaxes the (worst
   case) memory limit of 3*32 = 96 MB to 11*32 = 352 MB (at least the
   Medusa can have 128 MB :-) and also gives more room for
   kernel_map() and vmalloc(). (myself)

 - (Finally) made the '060 support complete so that it should work now:

    o Integrated the MC060IFPSP. There are still things left to do in
      the error handling there, and memory accesses could be further
      optimized. 

    o The system uses now only one supervisor stack pointer by
      default. This is necessary for the '060, since it only has one
      SSP (no MSP/ISP distinction anymore). The fact that there is
      only one stack now theoretically increases chances for kernel
      stack overflows, but in practice it seems that doesn't happen
      (tested on Jes' A3000/040 and my TT030).

      In difference to Jesper's original patch, I left the old code
      for the master stack in place, and surrounded it by "#ifdef
      MASTER_STACK"s. That comes from <asm/segment.h> and is undefined
      by default.

    o Jesper also observed strange effects in memory management that
      obviously page table contents stay dirty in the cache and aren't
      written to memory, though the pages have cache mode
      write-through. His quick fix for this is pushing the cache in
      some places. The CPUSHA060() macros should go away if the real
      reason is found. (BTW, I renamed Jesper's FLUSH060 to
      CPUSHA060(), which is more suggestive... :-) and moved the
      definition to <asm/segment.h> to avoid defining it in several
      sources.)

   (jskov@iesd.auc.dk, with little additions by me)

 - The Falcon can now boot with the kernel in Magnum or FX fast ram.
   That formerly didn't work because head.S made the bad assumption
   that, if the kernel is below 16 MB, it is at phys. addr. 0x0. But
   for the two ram boards mentioned above, the fast ram base address
   is 0x400000. (guenther@pool.informatik.rwth-aachen.de)

 - The command line option names for video ("video=", "atavideo=") and
   floppy ("fd_def_df0=", "ataflop=") have been unified into "video="
   and "floppy=", resp., because they're mutually exclusive between
   the machine types and there is no real need for different names.
   The old names are still valid to avoid breaking things, but they
   will go away in some future (probably 1.3.x).
   (schwab@issan.informatik.uni-dortmund.de)

 - Fixed a minor TCP protocol error
   (guenther@pool.informatik.rwth-aachen.de)

 - The root partition is now mounted read-only by default. The "ro"
   kernel option of course still works, but isn't really needed
   anymore. (schwab@issan.informatik.uni-dortmund.de)

 - (Brief) documentation about the AFFS in Documentation/filesystems/
   supplied by Hans-Joachim Widmaier <hjwidmai@foxboro.com>

Current Amiga bootstrap version: 3.1
Current Amiga-Lilo version     : 0.6
Current Atari bootstrap version: 1.7

The precompiled kernel contains both, Amiga and Atari support, so it
is very big. You probably want to compile your own kernel tailored to
your personal needs.

Roman