calmsacibis995
139 lines
4.7 KiB
Plaintext
139 lines
4.7 KiB
Plaintext
|
|
|
|
-- 12/5/94
|
|
|
|
Add EnterInteractiveMode, Halt, PowerDown, Restart, Reboot somewhere
|
|
to the IPMH code (finit????) so that the libsk fake versions don't get
|
|
incorporated into the prom. (12/4/94) I've defined stubs for the above
|
|
routines.
|
|
|
|
|
|
tpsc.c PREEMPT_.. depends on IPxx. I've temporarily make IPMH equivalent to
|
|
IP22.
|
|
|
|
nvram.c contains IPxx specific compilations. However, the ENV model
|
|
(offset and length) is inappropriate for IPMH. A separate env.c is in
|
|
libmsk for now so I've temporarily caused IPMH-nvram.c to generate a null
|
|
.o file. The differences between the old and the new may be too great to
|
|
just interpolate changes into nvram.c. The makefile presently builds
|
|
EVERESTnvram.c. I could make IPMHnvram.c to handle the problem.
|
|
|
|
passwd_cmd.c 1) contains IPxx specific compilations; 2) calls the
|
|
nvram routines directly and is sensitive to the physical implemenation
|
|
of the nvram storage. I've temporarily modified passwd_cmd.c to generate
|
|
a null object for IPMH.
|
|
|
|
-- 12/8/94
|
|
|
|
arcs/lib/Makefile clobber fails in libsk/graphis because no graphics
|
|
option is defined. I've TEMPORARILY defined (IPMH_X)SUBDIRS=none.
|
|
|
|
|
|
** It looks like the 64 bit trees have cleaned up sbd.h enough that I'll switch
|
|
back to using sbd.h instead of r4k.h.
|
|
|
|
endian_cmd.c references a bunch of SER_xxx symbols that come from
|
|
IPxxnvram.h and EEROM_xx symbols that come from mc.h. Both .h's get
|
|
sucked in via cpu.h so I've provided temporary definitions in IPMH.h
|
|
to satisify endian_cmd. It may be that endian_cmd.c needs to be changed
|
|
or not included.
|
|
|
|
stand/arcslib/cfg needs a IPxxconf.cf for some reason. Temporarily created
|
|
a symlink to the one in IPMH.
|
|
|
|
|
|
--12/9/94
|
|
|
|
In mp_show_fault, which prints part of the registers on an exception,
|
|
the pointer to the register area is varc_reg_t* rp (32bit)
|
|
bit). Everywhere else its k_machreg_t* rp (64bit). This means that
|
|
some of the code indexes into the register array as 64 bit values and
|
|
some as 32 bit values, the result being that part of the values
|
|
fetched are right, part are wrong. I believe that the code that views
|
|
the registers as varc_reg_t is wrong as the gpda code allocates using
|
|
k_machreg_t.
|
|
|
|
|
|
2/16/95
|
|
|
|
Significant clean up of the tree in preparation to checking it into
|
|
the pulpwood tree.
|
|
|
|
|
|
2/17/95
|
|
|
|
|
|
Filled in a lot of the stubs. Note, r4kcache.s doesn't simulate but does
|
|
build. In stubs.c there's a conditional section (#ifndef CACHEOPS) that
|
|
stubs out all the r4kcache routines. For a real platform this must be disable.
|
|
|
|
|
|
|
|
|
|
Resolved
|
|
==============================================================================
|
|
|
|
|
|
-- 12/8/94
|
|
|
|
pc_kbtable.c has been revised and the change isn't reflected in the
|
|
latest ROOT. I edited pckm.h in $ROOT/usr/include/sys/pckm.h.
|
|
Presumably it will catch up in a subsequent ROOT update.
|
|
|
|
The init_malloc code is sensitive to dmabuf_linemask. This is set in
|
|
config_cache so the stub for config_cache must do this.
|
|
|
|
The menu code requires the validate_pw code for most of the menu
|
|
items. With passwd_cmd stubbed, validate_pw must return 1.
|
|
|
|
The uprom code is such that if you issue the exit command, either
|
|
explictly or implicitly, code calls EnterInteractiveMode. That
|
|
recurses in the uprom code because in uprom, the stack isn't reset.
|
|
No problem unless you do it alot.
|
|
|
|
|
|
-- 12/12/94
|
|
|
|
I've added some exception simulation to SIM.c. SIGSEGV, SIGBUS and
|
|
SIGILL are caught and "handled".
|
|
|
|
|
|
-- 12/13/94
|
|
|
|
I've moved the simulator startup back to "start" which is in csu.s
|
|
which is in the sloader. That code does cache ops so I added
|
|
sillHandler which basically ignores cache ops. That code also fiddles
|
|
with the stack so I've added instructions that check for the code
|
|
being in the USEG and skip over the stack operations and the stack
|
|
memory validation if in USEG.
|
|
|
|
I've added code that model the coprocessor registers as simple ram
|
|
registers. There's no side effects for now.
|
|
|
|
--12/21/94
|
|
|
|
Env/FLASH is pretty screwy. I'm moving everything into flash.c and env.c,
|
|
platform specific modules. Everything is a lot cleaner. Remains to be seen
|
|
if this code stays platform specific.
|
|
|
|
-- 12/7/94
|
|
|
|
$ROOT/usr/include/sys/immu.h is referenced in libsk and depends on a
|
|
conditional compilation depending on IPXX. I've got a temporary hack
|
|
in that equates IPMH with IP22.
|
|
***** NOTE: temporary immu.h that is IP32 aware is in arcs/include/sys
|
|
Temporary hack removed 5/22/95.
|
|
|
|
|
|
-- 12/8/94
|
|
dump_cmd.c depends on K0_RAMBASE. K0_RAMBASE comes from mc.h which is
|
|
included in IP22.h which is included in sys/cpu.h. I'v added K0_RAMBASE
|
|
to IPMH.h
|
|
K0_RAMBASE is now defined in IP32.h
|
|
|
|
scsi.c references $HOME/usr/include/sys/cpu.h which doesn't define
|
|
IPMH. cpu.h is expected to define scuzzy_t for scsi.c. I've put a temporary
|
|
mod to cpu.h in arcs/include/sys and IPMH/Makefile headers installs IPMH.h
|
|
in arcs/include/sys.
|
|
|