]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
25 years agoCleanup. No need to install manpages here; no other system module
markm [Sun, 11 Oct 1998 17:23:34 +0000 (17:23 +0000)]
Cleanup. No need to install manpages here; no other system module
does it, and it is done /in toto/ elsewhere.

25 years agoEliminate nested comment warning.
jdp [Sun, 11 Oct 1998 17:14:56 +0000 (17:14 +0000)]
Eliminate nested comment warning.

Submitted by: lh@aus.org

25 years agoFixed removing of obsolete shared libraries:
bde [Sun, 11 Oct 1998 16:33:02 +0000 (16:33 +0000)]
Fixed removing of obsolete shared libraries:
- the directory was wrong if ${SHLIBDIR} != ${LIBDIR}.  It's still wrong
  if the installation of the obsolete library was done before /aout was
  appended to LIBDIR.
- the version would have become wrong when the default in ../Makefile.inc
  is changed from 2.0.
- the comment mostly described moving of libraries to /usr/lib/compat, but
  we don't do that.

25 years agoSync with sys/i386/boot/biosboot/table.c revision 1.17.
kato [Sun, 11 Oct 1998 15:09:36 +0000 (15:09 +0000)]
Sync with sys/i386/boot/biosboot/table.c revision 1.17.

25 years agoSync with sys/i386/boot/biosboot/boot.c revision up to 1.74.
kato [Sun, 11 Oct 1998 15:09:14 +0000 (15:09 +0000)]
Sync with sys/i386/boot/biosboot/boot.c revision up to 1.74.

25 years agoSync with sysi/386/boot/biosboot/boot.h revision 1.24.
kato [Sun, 11 Oct 1998 15:08:50 +0000 (15:08 +0000)]
Sync with sysi/386/boot/biosboot/boot.h revision 1.24.

25 years agoSync with sys/i386/i386/machdep.c revision 1.313.
kato [Sun, 11 Oct 1998 15:04:38 +0000 (15:04 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.313.

25 years agoSync with sys/i386/conf/files.i386 revision 1.206.
kato [Sun, 11 Oct 1998 15:02:49 +0000 (15:02 +0000)]
Sync with sys/i386/conf/files.i386 revision 1.206.

25 years agoSync with sys/i386/conf/Makefile.i386 revision 1.126.
kato [Sun, 11 Oct 1998 15:01:40 +0000 (15:01 +0000)]
Sync with sys/i386/conf/Makefile.i386 revision 1.126.

25 years agoSync with sys/i386/conf/GENERIC revision 1.124.
kato [Sun, 11 Oct 1998 15:00:24 +0000 (15:00 +0000)]
Sync with sys/i386/conf/GENERIC revision 1.124.

25 years agoAvoid the need for calling functions that malloc after a vfork().
peter [Sun, 11 Oct 1998 14:11:51 +0000 (14:11 +0000)]
Avoid the need for calling functions that malloc after a vfork().

25 years agoOverride for ELF.
rnordier [Sun, 11 Oct 1998 12:59:40 +0000 (12:59 +0000)]
Override for ELF.

25 years agoOverride for ELF. Override .s.o suffix rule.
rnordier [Sun, 11 Oct 1998 11:27:48 +0000 (11:27 +0000)]
Override for ELF.  Override .s.o suffix rule.

25 years agoFix a warning that's been bugging me for ages.
peter [Sun, 11 Oct 1998 10:29:49 +0000 (10:29 +0000)]
Fix a warning that's been bugging me for ages.

25 years agoFix a warning
peter [Sun, 11 Oct 1998 10:28:51 +0000 (10:28 +0000)]
Fix a warning

25 years agoProbe all consoles, not just the first. It doesn't make a lot of
peter [Sun, 11 Oct 1998 10:19:11 +0000 (10:19 +0000)]
Probe all consoles, not just the first.  It doesn't make a lot of
difference, but might later on when we implement some sort of multi-head
console mode.  Select a console after probing them all.
Don't strdup a potentially NULL return from getenv().
If we don't select an active console, choose the first regardless.
Call the console init function, at startup time and on a manual change.
The env_setenv() function needs EV_VOLATILE because it's pointing to
data that isn't malloc'ed and will cause a fault if it's freed later.

25 years agoCosmetic: After the autoboot timeout, print a \n.
peter [Sun, 11 Oct 1998 10:10:41 +0000 (10:10 +0000)]
Cosmetic:  After the autoboot timeout, print a \n.

25 years agoOnly call vidc_init() once (unless forced).
peter [Sun, 11 Oct 1998 10:07:52 +0000 (10:07 +0000)]
Only call vidc_init() once (unless forced).
Cosmetic change to the init-time character eater (like, make it increment
the index counter - if there's a problem, it would sit there in an infinite
loop instead of only running 10 times).

25 years agoFix comconsole mode. The int 0x14 read and status commands were reversed.
peter [Sun, 11 Oct 1998 10:05:13 +0000 (10:05 +0000)]
Fix comconsole mode.  The int 0x14 read and status commands were reversed.
Also, make sure we set %dx each time around otherwise the commands
suddenly start trying to work on things like com92 instead of com1.
Make sure comc_init() is only run once.
Cosmetic change to init-time character eater.

25 years agoWarn that one of the DEBUG statements has a v86 recursion bug.
peter [Sun, 11 Oct 1998 10:01:55 +0000 (10:01 +0000)]
Warn that one of the DEBUG statements has a v86 recursion bug.

25 years agoFix comment to match code
peter [Sun, 11 Oct 1998 09:53:30 +0000 (09:53 +0000)]
Fix comment to match code

25 years agoComment about libm and libmd being needed for some libraries.
peter [Sun, 11 Oct 1998 07:41:31 +0000 (07:41 +0000)]
Comment about libm and libmd being needed for some libraries.
Prompted by: bde

Also. Don't build & install legacy lkm's when NOLKM has been set.
Otherwise it gets built in the src tree rather than obj, because the
previous 'make obj' at the start of legacy-build does respect NOLKM.

25 years agoClear out transmit descriptor memory in fxp_attach when it's malloced.
dg [Sun, 11 Oct 1998 06:28:54 +0000 (06:28 +0000)]
Clear out transmit descriptor memory in fxp_attach when it's malloced.
fxp_stop is called as the first thing in fxp_init, and if the tx desc
list has junk in it, the system may panic. This bug showed up as a side
effect of the changes in rev 1.56, but has been in the code since the
beginning.

25 years agolibg++ and libstdc++ also have internal calls to libm. Programs using the
peter [Sun, 11 Oct 1998 04:56:03 +0000 (04:56 +0000)]
libg++ and libstdc++ also have internal calls to libm.  Programs using the
shared versions should not need to add -lm unless the program uses libm
itself.  Strictly speaking, libg++ depends on libstdc++, but libstdc++
has dynamic dependencies on some exception tables and binutils doesn't
seem to like it when a secondary library has undefined references.  It
doesn't seem to care when -lstdc++ is added on the command line to ld
though.  Anyway, the c++ driver adds -lstdc++ explicitly, so that should
be OK.  c++ also adds -lm explicitly too, even though it wouldn't need
to now. [except for statically linked binaries as .a files don't have an
equivalent of automatic internal shared object dependencies.]

25 years agoFix a couple of out-of-bounds array references in mapping between
jdp [Sun, 11 Oct 1998 04:54:16 +0000 (04:54 +0000)]
Fix a couple of out-of-bounds array references in mapping between
Linux and FreeBSD signal numbers.  Also, check signal numbers passed
in from application programs for validity.  Without these checks,
it is trivial to panic the system from a Linux program.

25 years agolibopie also depends on libmd. Programs using libopie shouldn't have to
peter [Sun, 11 Oct 1998 04:45:50 +0000 (04:45 +0000)]
libopie also depends on libmd.  Programs using libopie shouldn't have to
know about libopie's internals in order to use it.

25 years agolibf2c.so.* depends on libm.so. I'm not sure that this matters in this
peter [Sun, 11 Oct 1998 04:44:12 +0000 (04:44 +0000)]
libf2c.so.* depends on libm.so.  I'm not sure that this matters in this
case because I expect the f2c front-end should add the -lm automatically.

25 years agolibatm uses libmd; add it to the dependency list so that programs using
peter [Sun, 11 Oct 1998 04:41:43 +0000 (04:41 +0000)]
libatm uses libmd; add it to the dependency list so that programs using
libatm are not forced to know the internals of it's implementation.

25 years agoDeal with /usr/lib/compat/aout, and that we've been installing things
peter [Sun, 11 Oct 1998 04:39:56 +0000 (04:39 +0000)]
Deal with /usr/lib/compat/aout, and that we've been installing things
in the wrong places for a while.
Also, the the libtermlib.so -> libtermcap.so manually for elf, otherwise
the hard link follows the symlink and the result looks rather wierd.  The
*.a files are still hard linked under elf as before.

25 years agoDon't build elf libresolv.so*, no elf freebsd binaries have used it ever.
peter [Sun, 11 Oct 1998 04:21:56 +0000 (04:21 +0000)]
Don't build elf libresolv.so*, no elf freebsd binaries have used it ever.
Also, the a.out compat version is supposed to go into /usr/lib/compat/aout.
Try and clean up the mess left behind in /usr/lib/compat.

25 years agoDon't install an elf libgnumalloc.so* in /usr/lib/compat.. No elf binaries
peter [Sun, 11 Oct 1998 04:18:30 +0000 (04:18 +0000)]
Don't install an elf libgnumalloc.so* in /usr/lib/compat.. No elf binaries
have been linked against it.  Try and clean up the leftovers.  Also, put
the a.out libs in /usr/lib/compat/aout since that's where the default
a.out ldconfig compat path points to.

25 years agoDon't install an elf liby.so[.2] in /usr/lib/compat. We have never had
peter [Sun, 11 Oct 1998 04:10:14 +0000 (04:10 +0000)]
Don't install an elf liby.so[.2] in /usr/lib/compat.  We have never had
any elf binaries linked against it.  Try and clean up the mess left over..

25 years agoadd <sys/linker.h>
dima [Sun, 11 Oct 1998 03:53:35 +0000 (03:53 +0000)]
add <sys/linker.h>

25 years agoIf the bus delay is >= 2 seconds, notify the user that we are waiting
gibbs [Sat, 10 Oct 1998 21:10:36 +0000 (21:10 +0000)]
If the bus delay is >= 2 seconds, notify the user that we are waiting
for devices to settle.  This will hopefully allay any 'first installation'
fears that the machine has hung.

25 years agoMove OBJFORMAT hack to the correct line; what can I say, it was very
jkh [Sat, 10 Oct 1998 19:56:59 +0000 (19:56 +0000)]
Move OBJFORMAT hack to the correct line; what can I say, it was very
late. :)

25 years agovfork -> fork. The child calls execl() which calls malloc(), so
bde [Sat, 10 Oct 1998 19:30:45 +0000 (19:30 +0000)]
vfork -> fork.  The child calls execl() which calls malloc(), so
vfork() can't be used.  We could use alloca() in execl() so that
it can be called between vfork() and execve(), but a "portable"
popen() shouldn't depend on this.  Calling execle() instead of
execl() should be fairly safe, since execle() is supposed to be
callable from signal handlers and signal handlers can't call
malloc().  However, execle() is broken.

25 years agoFixed mbuf leak in fxp_stop().
dg [Sat, 10 Oct 1998 19:26:40 +0000 (19:26 +0000)]
Fixed mbuf leak in fxp_stop().

25 years agovfork -> fork. The child calls abort(), which calls stdio.
bde [Sat, 10 Oct 1998 19:21:39 +0000 (19:21 +0000)]
vfork -> fork.  The child calls abort(), which calls stdio.

25 years agovfork -> fork. This home made popen() had the same bug as the library
bde [Sat, 10 Oct 1998 19:18:30 +0000 (19:18 +0000)]
vfork -> fork.  This home made popen() had the same bug as the library
popen(), but worse.  The child calls execvp(), which calls malloc()
a bit more than execl(), and it calls non-library functions that call
malloc() and who-knows-what else (stdio is called in at least some
error cases).

25 years agoFixed gross breakage in previous commit. The malloc sizes for the
bde [Sat, 10 Oct 1998 19:01:47 +0000 (19:01 +0000)]
Fixed gross breakage in previous commit.  The malloc sizes for the
temporary file names were uninitialized if TMPDIR was set and 1 too
small otherwise.

Fixed style bugs in previous commit.

Fixed missing checks for malloc failure in previous commit.

Report malloc failure consistently, at least in temp.c.

25 years agoRemoved debugging cruft.
bde [Sat, 10 Oct 1998 18:37:02 +0000 (18:37 +0000)]
Removed debugging cruft.

Broken in: previous commit

25 years agoMake kernel build on alpha again. New kld stuff needs rindex.
gpalmer [Sat, 10 Oct 1998 18:32:18 +0000 (18:32 +0000)]
Make kernel build on alpha again. New kld stuff needs rindex.

25 years agoKill an extraneous dot.
rnordier [Sat, 10 Oct 1998 15:47:16 +0000 (15:47 +0000)]
Kill an extraneous dot.
PR: 8103
Submitted by: Kazuo Horikawa <horikawa@jp.freebsd.org>

25 years agoDon't pretend to support ix86's with 16-bit ints by using longs just
bde [Sat, 10 Oct 1998 14:30:05 +0000 (14:30 +0000)]
Don't pretend to support ix86's with 16-bit ints by using longs just
to ensure 32-bit variables.  Doing so broke i386's with 64-bit longs.

This also fixes printf format errors.

25 years agoFix a typo.
rnordier [Sat, 10 Oct 1998 14:03:03 +0000 (14:03 +0000)]
Fix a typo.
Output another newline before invoking bootstrap.

25 years agoFixed warnings for pointer versus int type mismatches. Addresses must
bde [Sat, 10 Oct 1998 14:02:07 +0000 (14:02 +0000)]
Fixed warnings for pointer versus int type mismatches.  Addresses must
be converted to fixed-sized integers when they are passed across the
binary interface to the kernel.

25 years agoFixed warnings for pointer versus int type mismatches. Addresses must
bde [Sat, 10 Oct 1998 13:51:24 +0000 (13:51 +0000)]
Fixed warnings for pointer versus int type mismatches.  Addresses must
be converted to fixed-sized integers when they are passed across the
binary interface to the kernel.

Didn't fix rotted bits (including not passing dosdev to the kernel and
serious out of dateness when initially committed).

25 years agomp_machdep.c: Set a vector to boot code (PC-98).
kato [Sat, 10 Oct 1998 13:37:16 +0000 (13:37 +0000)]
mp_machdep.c: Set a vector to boot code (PC-98).
locore.s: Tell the bios to warmboot next time (PC-98).

25 years agoRemove the description of EBADF (that's an implementation detail if I
eivind [Sat, 10 Oct 1998 13:31:32 +0000 (13:31 +0000)]
Remove the description of EBADF (that's an implementation detail if I
ever saw one), and move the description of NULL behaviour out to a
'NOTES' section, with an extra note that programs should not rely up
on it.

Kinda-approve-by: bde (by not replying to the mail with the diff)

25 years agoFixed searching for section "1aout" man pages. Section numbers must be
bde [Sat, 10 Oct 1998 13:11:39 +0000 (13:11 +0000)]
Fixed searching for section "1aout" man pages.  Section numbers must be
single digits or in the section list.

25 years agoDon't build or install ${MAN1}. It's just a copy of ${MAN1aout} and
bde [Sat, 10 Oct 1998 12:48:48 +0000 (12:48 +0000)]
Don't build or install ${MAN1}.  It's just a copy of ${MAN1aout} and
installing it clobbers the elf ${MAN1}.  The MAN1 -> MAN1aout changes
actually work now.

Fixed order of MAN* and BINDIR.

25 years agoPC-98 uses different boot code from IBM-PC. So, MACHINE is used
kato [Sat, 10 Oct 1998 11:16:08 +0000 (11:16 +0000)]
PC-98 uses different boot code from IBM-PC.  So, MACHINE is used
instead of MACHINE_ARCH.

Submitted by: Takahashi Yoshihiro <nyan@wyvern.cc.kogakuin.ac.jp>

25 years agoQuick commit to see if I can't fix Robert's i386/boot problem in one line.
jkh [Sat, 10 Oct 1998 10:54:54 +0000 (10:54 +0000)]
Quick commit to see if I can't fix Robert's i386/boot problem in one line.
If this doesn't work, we'll try something else.

25 years agoPC-98 doesn't have CMOS ram.
kato [Sat, 10 Oct 1998 10:36:12 +0000 (10:36 +0000)]
PC-98 doesn't have CMOS ram.

25 years agoPR: bin/8250
thepish [Sat, 10 Oct 1998 09:58:20 +0000 (09:58 +0000)]
PR: bin/8250
protect against buffer overruns in mail temporary files.

25 years agoAdd missing newline to debugging output.
jkh [Sat, 10 Oct 1998 09:43:44 +0000 (09:43 +0000)]
Add missing newline to debugging output.

25 years agoPC-98 doesn't have CMOS ram.
kato [Sat, 10 Oct 1998 09:38:02 +0000 (09:38 +0000)]
PC-98 doesn't have CMOS ram.

25 years agoAdd GENERIC rev 1.124 changes
obrien [Sat, 10 Oct 1998 09:25:12 +0000 (09:25 +0000)]
Add GENERIC rev 1.124 changes

25 years agoDon't chain makes with &&; apparently this is evil for parallelism.
jkh [Sat, 10 Oct 1998 08:21:18 +0000 (08:21 +0000)]
Don't chain makes with &&; apparently this is evil for parallelism.
According to: bde

25 years agoAdjust for new MFS root code.
jkh [Sat, 10 Oct 1998 08:15:38 +0000 (08:15 +0000)]
Adjust for new MFS root code.

25 years agoAdd entries for MFS which are consistent with the others, now that
jkh [Sat, 10 Oct 1998 08:13:58 +0000 (08:13 +0000)]
Add entries for MFS which are consistent with the others, now that
Peter has made this more selectable.

25 years agoAllow more flexible use of MFS root.
jkh [Sat, 10 Oct 1998 08:12:24 +0000 (08:12 +0000)]
Allow more flexible use of MFS root.
Submitted by: peter

25 years agoAttempt to work around the page fault in tulip_txput(). I've been running
peter [Sat, 10 Oct 1998 02:44:53 +0000 (02:44 +0000)]
Attempt to work around the page fault in tulip_txput().  I've been running
this myself for ages, but wasn't able to get any feedback from the people
that I sent it to for testing.

Guy Helmer <ghelmer@scl.ameslab.gov> has given it a shot (before getting on
a plane, thanks!) and it appears to stop his reproducable page fault panic
in the testing he was able to do.

25 years agoDisplay module type as well as module name when we find one preloaded.
peter [Sat, 10 Oct 1998 02:29:07 +0000 (02:29 +0000)]
Display module type as well as module name when we find one preloaded.

25 years agorm *.so and *.ko in clean as well.
peter [Sat, 10 Oct 1998 01:00:56 +0000 (01:00 +0000)]
rm *.so and *.ko in clean as well.

25 years agoFix conficts in probe:
imp [Sat, 10 Oct 1998 00:44:12 +0000 (00:44 +0000)]
Fix conficts in probe:
o For bt and aha only probe the one I/O range if a specific I/O is specified
  in the config file.
o Don't even try to probe I/O ranges that have been seen already.
o If we conflict with an IRQ or DRQ, then fail the probe.

Requested by: bde, gibbs
Approved by: jkh

25 years agoUse Mike Smith's linker module search path code.
peter [Sat, 10 Oct 1998 00:07:53 +0000 (00:07 +0000)]
Use Mike Smith's linker module search path code.
Implement preloading in a fairly MI way, assuming the information is
prepared.
DDB interface helpers..  Provide some support for db_kld.c so that we
don't have to export too much detail.
Debugging and cosmetic nits left in from development..
The other half of the containing file hack so modules can associate
themselves with their "file".

25 years agoKeep track of the containing file for modules. This is a bit of a hack,
peter [Sat, 10 Oct 1998 00:03:07 +0000 (00:03 +0000)]
Keep track of the containing file for modules.  This is a bit of a hack,
but I can't think of another (relatively) easy way of getting the info
since the boot-time initialization is not done immediately after "loading".
XXX module_register() gained an extra arg.  This might break the alpha
compile, if so, just add a zero to get the old behavior.

25 years agoUpdates for alignment rounding. XXX this is highly machine dependent and
peter [Fri, 9 Oct 1998 23:59:01 +0000 (23:59 +0000)]
Updates for alignment rounding.  XXX this is highly machine dependent and
should probably be moved to i386/i386/link_machdep.c (and the same for the
alpha).
Implement "deleting" a preloaded module by destroying it's tags. This is a
hack.  We cannot reuse the data, it's been destroyed by relocation,
statically initialized variables have been modified, etc.  Note that to
reclaim the load space is going to be more machine-dependent work.
Implement a relocate hook for machdep.c to call so that the physical
addresses get converted to the equivalent KVM addresses.

25 years agoFully implement KLD and preloading.
peter [Fri, 9 Oct 1998 23:55:31 +0000 (23:55 +0000)]
Fully implement KLD and preloading.
- seperate unload for preloaded linker objects.
- Don't build a kernel object if running as an a.out kernel.
- extract the real kernel name rather than hardwiring "kernel" for kldstat.
  (sysctl kern.bootfile getst the full name via bootinfo)
- use real addresses on the kernel "module" rather than fictitious ones.
- preloaded module support
- search module path for file modules.
- symbols are checked to see if they are in the right containing file
  before using their indexes into string tables.  This is to help ddb
  since it only supplies a pointer to an opaque symbol and there is no
  telling which file/object/module/whatever it came from.

25 years agoUpdates for KLD backends.
peter [Fri, 9 Oct 1998 23:49:28 +0000 (23:49 +0000)]
Updates for KLD backends.
 - symbol_values checks that the symbol is indeed belonging to the
   correct symbol and string table pairs before looking up. (since there
   could be many pairs, and KLD/DDB need to find out).
 - different ops for files versus preload modules - the unload mechanism
   is different.  (a preloaded module has to be deleted on unload since
   the in-core image is tainted by relocation and variables used)
 - Do not build an a.out kernel module if we're running on an elf
   kernel. :-)  Note that it should theoretically be possible to
   mix a.out and elf KLD modules providing -mno-underscores was used
   to compile it, or some other symbol conversion takes place.
 - Support preload modules (even though /boot/loader doesn't yet)
 - Search the module path when loading files.

25 years agoImplement merging SYSINIT's from preloaded KLD modules. This means we
peter [Fri, 9 Oct 1998 23:42:47 +0000 (23:42 +0000)]
Implement merging SYSINIT's from preloaded KLD modules.  This means we
check off SYSINIT entries as they are run, and when more arrive, we re-sort
and restart (skipping the already-run entries).
This can *only* be done after KMEM (and malloc) is up and running - this is
fine because KLD is the only consumer of this and it's done after that.
The nice thing about this is that the SYSINIT's within preloaded KLD modules
are executed in their natural order.  It should be possible to register
devices for the probes which follow, etc.  (soon.. several key things
prevent this, such as use of linker sets for things like pci devices).

25 years agoMODINFO_ADDR has real addresses now, remove the manual relocation based
peter [Fri, 9 Oct 1998 23:37:37 +0000 (23:37 +0000)]
MODINFO_ADDR has real addresses now, remove the manual relocation based
on cpu type.

25 years agoRelocate the preload module info from machdep specifically rather than
peter [Fri, 9 Oct 1998 23:36:26 +0000 (23:36 +0000)]
Relocate the preload module info from machdep specifically rather than
trying to do it in locore.  We also walk through the module table
and relocate any MODINFO_ADDR pointers so that they become KVM relative
rather than physical addresses.  This means that hacks for adding
0xf0000000 in places like MFS go away.

25 years agoCall some helper routines to be supplied by kern_linker.c in order to
peter [Fri, 9 Oct 1998 23:34:09 +0000 (23:34 +0000)]
Call some helper routines to be supplied by kern_linker.c in order to
get to all the symbol tables for all modules, not just the core kernel
symbol table.  Yes, DDB can see KLD module symbols with this, both by
lookup and in tracebacks.  No more references to _end from tracebacks
within an LKM. :-)

25 years agoAt the moment, the alpha tentatively uses the ddb elf code rather than KLD,
peter [Fri, 9 Oct 1998 23:32:03 +0000 (23:32 +0000)]
At the moment, the alpha tentatively uses the ddb elf code rather than KLD,
because the alpha boot loader hasn't been converted yet, and because
it needs the full symbol tables with local symbols in order to make sense
of stack tracebacks.  KLD will implement this (using full sybmol table
rather than the globals only) shortly.

25 years agoTurn off the new /sys/boot stuff (except boot0) unless OBJFORMAT
rnordier [Fri, 9 Oct 1998 23:30:16 +0000 (23:30 +0000)]
Turn off the new /sys/boot stuff (except boot0) unless OBJFORMAT
is elf.  (The BTX client must be ELF, though it is packaged as a.out
for compatibility.)

25 years agoWe don't compile this on an elf kernel (and explicitly not on alpha in
peter [Fri, 9 Oct 1998 23:29:44 +0000 (23:29 +0000)]
We don't compile this on an elf kernel (and explicitly not on alpha in
case it's possible to compile in something like ECOFF)
The three db_xxx.c symbol interfaces are "standard" because config isn't
flexible enough without forcing the user to know about it.

25 years agoRemove DDB_ELF_SYMBOLS and DDB_KLD_SYMBOLS options. KLD will be the sole
peter [Fri, 9 Oct 1998 23:26:26 +0000 (23:26 +0000)]
Remove DDB_ELF_SYMBOLS and DDB_KLD_SYMBOLS options.  KLD will be the sole
interface to symbols in an elf kernel.

25 years agoAttempt to at least align MODINFO_* blocks. IMHO, decoding this stuff
peter [Fri, 9 Oct 1998 23:24:55 +0000 (23:24 +0000)]
Attempt to at least align MODINFO_* blocks.  IMHO, decoding this stuff
should be MD code since one day we'll have to recover pages from deleted
preload data.  MI code can't be expected to know how to deal with pmap
internals, assuming it gets done via pmap that is. :-)

25 years agoTurn symbol table info passing back on, although it's probably not
peter [Fri, 9 Oct 1998 23:22:30 +0000 (23:22 +0000)]
Turn symbol table info passing back on, although it's probably not
particularly useful in this form.  KLD can deal with it much better.
setenv kernelname moved earlier.

25 years agoRemove kernelname setenv, the common code does it.
peter [Fri, 9 Oct 1998 23:20:47 +0000 (23:20 +0000)]
Remove kernelname setenv, the common code does it.
Use the metadata for symtab pointers, mainly to keep it common with
elf_freebsd.c.

25 years agoImplement preloading for elf modules
peter [Fri, 9 Oct 1998 23:18:43 +0000 (23:18 +0000)]
Implement preloading for elf modules
 - get dependency info from PT_DYNAMIC's DT_NEEDED tags.
 - store MODINFOMD_DYNAMIC for the kernel's later use
setenv kernelname when we have it
Fix firstaddr/lastaddr calculation (duh! :-)
Explicitly skip string table with section names in it.

25 years agoPreload support for a.out KLD not implemented yet. (almost)
peter [Fri, 9 Oct 1998 23:15:39 +0000 (23:15 +0000)]
Preload support for a.out KLD not implemented yet. (almost)
KLD modules are *not* PIC.  (Shared libs are pic to avoid relocations
causing copy-on-write, that's irrelevant here).
setenv kernelname when we load it.
Use MODINFO_SSYM/ESYM for each symbol section when (if) there are
more than one being loaded.

25 years agoUse a consistant module search path (same as kernel will be).
peter [Fri, 9 Oct 1998 23:12:34 +0000 (23:12 +0000)]
Use a consistant module search path (same as kernel will be).
Use new dependency mechanism.

25 years agoloader <-> kernel preload module interface constants moved to common file.
peter [Fri, 9 Oct 1998 23:11:05 +0000 (23:11 +0000)]
loader <-> kernel preload module interface constants moved to common file.
Remove Mike's explicit data structures for dependency info.  This is
done via DT_NEEDED etc in the dynamic section for now.  This may need
to be revisited later on.

25 years agoelf_machdep.c and rindex.c are now standard
peter [Fri, 9 Oct 1998 23:08:14 +0000 (23:08 +0000)]
elf_machdep.c and rindex.c are now standard

25 years agoDebug typo fixes.
peter [Fri, 9 Oct 1998 23:07:27 +0000 (23:07 +0000)]
Debug typo fixes.
Change symbol_values return type to int, on the chance that we've
been given the wrong symbol table.  Symbols have a string index that
must be relative to the correct string table.
Add prototypes for better kld support for ddb.

25 years agoA hook for storing the origin of a module, this is important when
peter [Fri, 9 Oct 1998 23:05:45 +0000 (23:05 +0000)]
A hook for storing the origin of a module, this is important when
it's being registered in sorted sysinit sequence and not in load order.

25 years agoAdd SI_SUB_KLD
peter [Fri, 9 Oct 1998 23:03:27 +0000 (23:03 +0000)]
Add SI_SUB_KLD
First part of support for merging SYSINIT sets.

This, and the following KLD commits have been OK'ed by jkh and msmith
based on my assertion that it works here (barring merge errors :-).

25 years agoClean up the boot targets a bit.
jkh [Fri, 9 Oct 1998 22:22:34 +0000 (22:22 +0000)]
Clean up the boot targets a bit.
Suggested by: jdp

25 years agoCall dpt_intr from our timeout routine to clear any pending commands before
gibbs [Fri, 9 Oct 1998 21:42:19 +0000 (21:42 +0000)]
Call dpt_intr from our timeout routine to clear any pending commands before
performing actual timeout processing.

Modify a few printf statements.

Submitted by: Simon Shapiro <shimon@simon-shapiro.org>

25 years agoOnly pull 16 bits of residual information from completing queues. This
gibbs [Fri, 9 Oct 1998 21:40:50 +0000 (21:40 +0000)]
Only pull 16 bits of residual information from completing queues.  This
is a work-around from an LRAM access bug on the 940UA.  In a future
microcode revision, the high 16bits of residual information will be moved
to a safe location and we'll return to 32bit residuals.  Since we only
allow 64KB I/O, 16bits is enough.

25 years agoHook up ahapoll so that dumps, synchronize cache commands or any other
gibbs [Fri, 9 Oct 1998 21:38:36 +0000 (21:38 +0000)]
Hook up ahapoll so that dumps, synchronize cache commands or any other
command that comes in through xpt_polled_action works correctly.

25 years agoDon't try to initialise the environment out of a sysinit, it's handled
msmith [Fri, 9 Oct 1998 21:21:34 +0000 (21:21 +0000)]
Don't try to initialise the environment out of a sysinit, it's handled
in MD code instead.

25 years agoNull commit.. CVS aborted on freefall last time (reaonly file).
peter [Fri, 9 Oct 1998 20:38:03 +0000 (20:38 +0000)]
Null commit.. CVS aborted on freefall last time (reaonly file).

An elf_reloc() function for the i386.  Based on alpha/alpha/elf_machdep.c
and rtld-elf/i386/reloc.c.

25 years agoAn elf_reloc() function for the i386. Based on alpha/alpha/elf_machdep.c
peter [Fri, 9 Oct 1998 20:35:45 +0000 (20:35 +0000)]
An elf_reloc() function for the i386.  Based on alpha/alpha/elf_machdep.c
and rtld-elf/i386/reloc.c.

25 years agoAdd JKH's auth.conf parser to turn on/off Kerberos in userland
markm [Fri, 9 Oct 1998 20:14:48 +0000 (20:14 +0000)]
Add JKH's auth.conf parser to turn on/off Kerberos in userland

25 years agoFix some bugs in pthread scheduler:
dt [Fri, 9 Oct 1998 19:01:30 +0000 (19:01 +0000)]
Fix some bugs in pthread scheduler:
  make pthread_yield() more reliable,
  threads always (I hope) preempted at least every 0.1 sec, as intended.

PR: bin/7744
Submitted by: "Richard Seaman, Jr." <dick@tar.com>

25 years agoPass termination information via softc flags to the core driver. The
gibbs [Fri, 9 Oct 1998 17:42:28 +0000 (17:42 +0000)]
Pass termination information via softc flags to the core driver.  The
EISA probe missed out on this change in attach->core API.