]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years ago<machine/atomic.h> requires <sys/types.h>.
ru [Wed, 9 Oct 2002 20:20:43 +0000 (20:20 +0000)]
<machine/atomic.h> requires <sys/types.h>.

Reviewed by: jake, mike

21 years agoMdocify and fix a load of errors dating back to the dawn of time.
fanf [Wed, 9 Oct 2002 20:00:25 +0000 (20:00 +0000)]
Mdocify and fix a load of errors dating back to the dawn of time.
Don't mention hard limits for factor(6) since it now has bignum support.

Obtained from: NetBSD (mdoc only -- their man page is still mostly wrong)

21 years agoFactorize bignums using the Pollard Rho algorithm. The code comes
fanf [Wed, 9 Oct 2002 19:55:04 +0000 (19:55 +0000)]
Factorize bignums using the Pollard Rho algorithm. The code comes
from NetBSD, altered to retain the FreeBSD -h feature and various
stylistic improvements and avoid regressions. In the absence of
OpenSSL (and therefore bignums) the old code is used.

PR: 43831
Reviewed by: kris, markm, obrien
Obtained from: NetBSD
MFC after: 2 weeks

21 years agoUse __packed.
jhb [Wed, 9 Oct 2002 19:46:09 +0000 (19:46 +0000)]
Use __packed.

21 years agoStyle fixes to: #include ordering; use const and static; ANSI functions;
fanf [Wed, 9 Oct 2002 19:38:55 +0000 (19:38 +0000)]
Style fixes to: #include ordering; use const and static; ANSI functions;
various usage synopses; bogus and/or unnecessary casting; exit values;
use LINE_MAX instead of magic numbers; declare extern variables in a
header; add $FreeBSD$ where missing.

Reviewed by:    markm, obrien
Obtained from:  NetBSD | fanf

21 years ago- Move p_cpulimit to struct proc from struct plimit and protect it with
jhb [Wed, 9 Oct 2002 17:17:24 +0000 (17:17 +0000)]
- Move p_cpulimit to struct proc from struct plimit and protect it with
  sched_lock.  This means that we no longer access p_limit in mi_switch()
  and the p_limit pointer can be protected by the proc lock.
- Remove PRS_ZOMBIE check from CPU limit test in mi_switch().  PRS_ZOMBIE
  processes don't call mi_switch(), and even if they did there is no longer
  the danger of p_limit being NULL (which is what the original zombie check
  was added for).
- When we bump the current processes soft CPU limit in ast(), just bump the
  private p_cpulimit instead of the shared rlimit.  This fixes an XXX for
  some value of fix.  There is still a (probably benign) bug in that this
  code doesn't check that the new soft limit exceeds the hard limit.

Inspired by: bde (2)

21 years agoRemove 'at' hints for npx and apm as both drivers have identify routines
jhb [Wed, 9 Oct 2002 17:00:46 +0000 (17:00 +0000)]
Remove 'at' hints for npx and apm as both drivers have identify routines
that add an instance of themselves.  The npx(4) driver doesn't even check
the npx 'port' hint but hardcodes IO_NPX instead.  The npx(4) driver also
will use isa IRQ 13 (on x86, 8 on pc98) by default if no 'irq' hint is
specified, so we don't need that hint either.

21 years agoFix build of 64 bit platforms.
mux [Wed, 9 Oct 2002 12:19:36 +0000 (12:19 +0000)]
Fix build of 64 bit platforms.

21 years agoMove the ed, fe, CD9660, MSDOSFS and NFSCLIENT into mfsroot floppy instead
nyan [Wed, 9 Oct 2002 12:01:43 +0000 (12:01 +0000)]
Move the ed, fe, CD9660, MSDOSFS and NFSCLIENT into mfsroot floppy instead
of the pcn, sf, ste, tl and tx.

21 years ago- Remove the splash, ppc, ppbus, plip and em devices from small boot floppy.
nyan [Wed, 9 Oct 2002 12:01:14 +0000 (12:01 +0000)]
- Remove the splash, ppc, ppbus, plip and em devices from small boot floppy.
- Remove the splash, ch and sa devices from normal boot floppy.
- The agp device does not exist.

21 years agoFix build on alpha. It used to case an int to a pointer in order to do
peter [Wed, 9 Oct 2002 09:30:57 +0000 (09:30 +0000)]
Fix build on alpha.  It used to case an int to a pointer in order to do
pointer arithmetic.  Detour via uintptr_t, since we're generating an
offset not dereferencing them.

21 years agoAdd definitions for the new international monetary formatting struct lconv
tjr [Wed, 9 Oct 2002 09:25:11 +0000 (09:25 +0000)]
Add definitions for the new international monetary formatting struct lconv
members to English (Australia) as an example.

21 years agoAdd support for the 6 new C99 struct lconv members dealing with formatting
tjr [Wed, 9 Oct 2002 09:19:28 +0000 (09:19 +0000)]
Add support for the 6 new C99 struct lconv members dealing with formatting
international monetary values: int_p_cs_precedes, int_n_cs_precedes,
int_p_sep_by_space, int_n_sep_by_space, int_p_sign_posn, int_n_sign_posn.
This should not break existing binaries or LC_MONETARY data files.

Reviewed by: ache
MFC after: 1 month

21 years agoChange BUS_SPACE_UNRESTRICTED (~0ul) to plain ~0 when used in the
peter [Wed, 9 Oct 2002 08:54:32 +0000 (08:54 +0000)]
Change BUS_SPACE_UNRESTRICTED (~0ul) to plain ~0 when used in the
'int nsegments' argument to bus_dma_tag_create().  ~0ul does not fit in
an int on machines with 64 bit longs.

21 years agoPrint the OS name in the right place in the document title.
ru [Wed, 9 Oct 2002 08:51:52 +0000 (08:51 +0000)]
Print the OS name in the right place in the document title.

21 years agoChange BUS_SPACE_UNRESTRICTED (~0ul) to plain ~0 when used in the
peter [Wed, 9 Oct 2002 08:50:26 +0000 (08:50 +0000)]
Change BUS_SPACE_UNRESTRICTED (~0ul) to plain ~0 when used in the
'int nsegments' argument to bus_dma_tag_create().  ~0ul does not fit in
an int.

21 years agoMFV:
ru [Wed, 9 Oct 2002 08:48:56 +0000 (08:48 +0000)]
MFV:

- Suppress additional whitespace after the subsection header (.Ss).
  (This affects a tiny number of our manpages.)

- Allow for a layout customization of the OS name in the document
  title (.Dt).  (Needed for Russian and French manpages.)

21 years agoDrop almost 3k from /bin/sync by moving errno to a seperate file
peter [Wed, 9 Oct 2002 08:04:24 +0000 (08:04 +0000)]
Drop almost 3k from /bin/sync by moving errno to a seperate file
to avoid all syscalls pulling in sys_errlst[].

Noted by:  bde

21 years agoDon't kill itself in --fast-read mode if there is no decompressor started.
sobomax [Wed, 9 Oct 2002 07:33:29 +0000 (07:33 +0000)]
Don't kill itself in --fast-read mode if there is no decompressor started.

Reported by: Gordon Tetlow <gordont@gnf.org>
Submitted by: tjr
MFC after: 1 week
(along with other --fast-read changes)

21 years agoWhen creating a snapshot, create a list of initially allocated blocks.
mckusick [Wed, 9 Oct 2002 07:28:35 +0000 (07:28 +0000)]
When creating a snapshot, create a list of initially allocated blocks.
Whenever doing a copy-on-write check, first look in the list of
initially allocated blocks to see if it is there. If so, no further
check is needed. If not, fall through and do the full check. This
change eliminates one of two known deadlocks caused by snapshots.
Handling the second deadlock will be the subject of another check-in.
This change also reduces the cost of the copy-on-write check by
speeding up the verification of frequently checked blocks.

Sponsored by: DARPA & NAI Labs.

21 years agoAdd support g_clone_bio() and g_std_done() to spawn multiple children
phk [Wed, 9 Oct 2002 07:11:59 +0000 (07:11 +0000)]
Add support g_clone_bio() and g_std_done() to spawn multiple children
of a bio and correctly gather status when done.

Sponsored by: DARPA & NAI Labs.

21 years agoAdd a field for tallying the number of spawned bio's a bio has.
phk [Wed, 9 Oct 2002 07:11:03 +0000 (07:11 +0000)]
Add a field for tallying the number of spawned bio's a bio has.

Rearrange and comment some GEOM related fields.

Sponsored by: DARPA & NAI Labs.

21 years agoUpdate documentation for kldload(8)'s ability to load multiple modules.
jmallett [Wed, 9 Oct 2002 06:46:54 +0000 (06:46 +0000)]
Update documentation for kldload(8)'s ability to load multiple modules.

Submitted by: Daniel O'Connor <doconnor@gsoft.com.au>

21 years agoWhen creating a snapshot, create a list of initially allocated blocks.
mckusick [Wed, 9 Oct 2002 06:13:48 +0000 (06:13 +0000)]
When creating a snapshot, create a list of initially allocated blocks.
Whenever doing a copy-on-write check, first look in the list of
initially allocated blocks to see if it is there. If so, no further
check is needed. If not, fall through and do the full check. This
change eliminates one of two known deadlocks caused by snapshots.
Handling the second deadlock will be the subject of another check-in.
This change also reduces the cost of the copy-on-write check by
speeding up the verification of frequently checked blocks.

Sponsored by: DARPA & NAI Labs.

21 years agoThe appropriate units for disk block addresses are always DEV_BSIZE,
mckusick [Wed, 9 Oct 2002 04:01:23 +0000 (04:01 +0000)]
The appropriate units for disk block addresses are always DEV_BSIZE,
even when the underlying device has a larger sector size. Therefore,
the filesystem code should not (and with this patch does not) try to
use the underlying sector size when doing disk block address calculations.

This patch fixes problems in -current when using the swap-based
memory-disk device (mdconfig -a -t swap ...). This bugfix is not
relevant to -stable as -stable does not have the memory-disk device.

Sponsored by: DARPA & NAI Labs.

21 years agoConvert make(1) to use ANSI style function declarations. Variable
jmallett [Wed, 9 Oct 2002 03:42:10 +0000 (03:42 +0000)]
Convert make(1) to use ANSI style function declarations.  Variable
documentation already adequatedly existed in the description in most
cases.  Where it did not, it was added.  If no documentation existed
beforehand, then none was added.  Some unused dummies for use in the
traversal functions were marked as __unused during the conversion.
Occasionally, local style fixes were applied to lines already being
modified or influenced.

Now make(1) should always build with WARNS=3.

21 years agoRound out the facilty for a 'bound' thread to loan out its KSE
julian [Wed, 9 Oct 2002 02:33:36 +0000 (02:33 +0000)]
Round out the facilty for a 'bound' thread to loan out its KSE
in specific situations. The owner thread must be blocked, and the
borrower can not proceed back to user space with the borrowed KSE.
The borrower will return the KSE on the next context switch where
teh owner wants it back. This removes a lot of possible
race conditions and deadlocks. It is consceivable that the
borrower should inherit the priority of the owner too.
that's another discussion and would be simple to do.

Also, as part of this, the "preallocatd spare thread" is attached to the
thread doing a syscall rather than the KSE. This removes the need to lock
the scheduler when we want to access it, as it's now "at hand".

DDB now shows a lot mor info for threaded proceses though it may need
some optimisation to squeeze it all back into 80 chars again.
(possible JKH project)

Upcalls are now "bound" threads, but "KSE Lending" now means that
other completing syscalls can be completed using that KSE before the upcall
finally makes it back to the UTS. (getting threads OUT OF THE KERNEL is
one of the highest priorities in the KSE system.) The upcall when it happens
will present all the completed syscalls to the KSE for selection.

21 years agoUse __FBSDID in the lst.lib files, to match to main make(1) files. It is
jmallett [Wed, 9 Oct 2002 02:00:22 +0000 (02:00 +0000)]
Use __FBSDID in the lst.lib files, to match to main make(1) files.  It is
defined to __RCSID when bootstrapping, by the Makefile, but this change is
for consistency

21 years agoRemove extern declarations from functions and source files where they would
jmallett [Wed, 9 Oct 2002 01:56:02 +0000 (01:56 +0000)]
Remove extern declarations from functions and source files where they would
happily fit into headers.

21 years agoRemove unused local-locals, where upper-level locals may safely be used.
jmallett [Wed, 9 Oct 2002 01:51:00 +0000 (01:51 +0000)]
Remove unused local-locals, where upper-level locals may safely be used.

21 years agoAs back-out requests must be followed, back out the removal of wargames(6)
jmallett [Wed, 9 Oct 2002 01:46:37 +0000 (01:46 +0000)]
As back-out requests must be followed, back out the removal of wargames(6)
at mike@'s request.

21 years agoAdd note that there is a kernel-imposed limit on the number of threads
archie [Tue, 8 Oct 2002 22:42:42 +0000 (22:42 +0000)]
Add note that there is a kernel-imposed limit on the number of threads
in a KSE group that may be simultaneously blocked in the kernel.

21 years agoReference the correct local variable in all parts of a for loop, so we
jmallett [Tue, 8 Oct 2002 21:29:04 +0000 (21:29 +0000)]
Reference the correct local variable in all parts of a for loop, so we
operate on the correct data (properly).

Tracked down by: Dan Nelson

21 years ago - Remove LK_INTERLOCK from the vn_lock() in ffs_snapshot().
jeff [Tue, 8 Oct 2002 21:00:52 +0000 (21:00 +0000)]
 - Remove LK_INTERLOCK from the vn_lock() in ffs_snapshot().

Pointy hat to: me
Found by: green

21 years agoRemove the wargames(6) novelty [and not the fun kind]:
jmallett [Tue, 8 Oct 2002 20:19:26 +0000 (20:19 +0000)]
Remove the wargames(6) novelty [and not the fun kind]:
1) It never was very funny.
2) It was a N line shell script where N is less than
   X, where X is the number of lines of the Berkeley
   copyright and license.
3) I rewrote it in C with the intention of making it
   more interactive, ala the movie, but never did.
4) Because it really is possible for at least some
   parts of a distribution to shrink between releases.

X-No-MFC: This game will remain in RELENG_4 and previous.
X-Bikesheds-To: markm@
Approved by: The #bsdcode cabal.
Obtained from: Professor Falken

21 years agoWe're missing none of these games, they just aren't here.
jmallett [Tue, 8 Oct 2002 20:15:30 +0000 (20:15 +0000)]
We're missing none of these games, they just aren't here.

21 years agoSpell 'Temperature' correctly.
jhb [Tue, 8 Oct 2002 19:41:12 +0000 (19:41 +0000)]
Spell 'Temperature' correctly.

21 years agoFix for the panic when using a Promise TX2.
sos [Tue, 8 Oct 2002 18:25:10 +0000 (18:25 +0000)]
Fix for the panic when using a Promise TX2.

The problem is that the code does a check for the granparent of
the Promise chip, if this is a bridge of the right type, we have
a TX4 on our hands, and need to handle that ones "issues".
Now the grandparent check cause subtle bugs in the newbus system,
mainly that pci_get_devid doesn't return an error value.
This patch works around the issue by using BUS_READ_IVAR() instead.

21 years agoRevert MEM_USR back to 0xa000 for BTX clients. Instead, adjust boot2
jhb [Tue, 8 Oct 2002 18:19:02 +0000 (18:19 +0000)]
Revert MEM_USR back to 0xa000 for BTX clients.  Instead, adjust boot2
to run at 0xc000 by changing its virtual start address from 0x1000 to
0x2000.

Tested by: phk

21 years agoAdd MLINKS to kse.2.
archie [Tue, 8 Oct 2002 17:43:40 +0000 (17:43 +0000)]
Add MLINKS to kse.2.

21 years agoFix the location of the length bytes in the 12-byte read/write CDB
ken [Tue, 8 Oct 2002 17:12:44 +0000 (17:12 +0000)]
Fix the location of the length bytes in the 12-byte read/write CDB
structure.  This has been broken since 1998, but probably hasn't been
noticed because it takes a read/write of 64K blocks (32MB with 512 byte
blocks) to trigger using the 12 byte read/write CDB in scsi_read_write().

Submitted by: "Moore, Eric Dean" <emoore@lsil.com>
MFC after: 3 days

21 years agoNow that ufsread.c doesn't do 64bit divide remainder operations,
phk [Tue, 8 Oct 2002 15:49:35 +0000 (15:49 +0000)]
Now that ufsread.c doesn't do 64bit divide remainder operations,
don't bother with libkern.

Sponsored by: DARPA & NAI Labs.

21 years agoSave a couple of bytes by not returning ints nobody care about.
phk [Tue, 8 Oct 2002 15:48:43 +0000 (15:48 +0000)]
Save a couple of bytes by not returning ints nobody care about.

Sponsored by: DARPA & NAI labs

21 years agoIt seems that the only problem with UFS2 booting on i386 is the 64bit
phk [Tue, 8 Oct 2002 15:46:45 +0000 (15:46 +0000)]
It seems that the only problem with UFS2 booting on i386 is the 64bit
divide/remainder calls.  For reasons not resolved, compiling the
relevant routines from libkern into boot2 results in stack corruption.

Do the simple thing: Don't use 64bit divide/remainder operations.

Sponsored by: DARPA & NAI Labs

21 years agoDon't panic for a bad ivar request, just return ENOENT.
jhb [Tue, 8 Oct 2002 15:44:35 +0000 (15:44 +0000)]
Don't panic for a bad ivar request, just return ENOENT.

21 years agoMFi386: revision 1.67.
nyan [Tue, 8 Oct 2002 12:54:56 +0000 (12:54 +0000)]
MFi386: revision 1.67.

21 years agoMFi386: revisions 1.539, 1.540 and 1.541.
nyan [Tue, 8 Oct 2002 12:53:26 +0000 (12:53 +0000)]
MFi386: revisions 1.539, 1.540 and 1.541.

21 years agoFix to support pc98.
nyan [Tue, 8 Oct 2002 12:13:19 +0000 (12:13 +0000)]
Fix to support pc98.

21 years agoReinstate rev 1.36 with an important line that got missed. Note this
green [Tue, 8 Oct 2002 12:09:16 +0000 (12:09 +0000)]
Reinstate rev 1.36 with an important line that got missed.  Note this
also improves the "random undocumented offsets into various memory
spaces" a little bit.

21 years agoDo not strip CTL* escapes from redirection filenames in argstr(); they
tjr [Tue, 8 Oct 2002 11:22:49 +0000 (11:22 +0000)]
Do not strip CTL* escapes from redirection filenames in argstr(); they
are later stripped with rmescapes() in expandarg(). If the filename has
already been unescaped, doing it again in rmescapes() can walk off the
end of the string, leading to memory corruption and eventually SIGSEGV.

Noticed by: kris

21 years agoRemove unused TYPE_WD and TYPE_WFD.
phk [Tue, 8 Oct 2002 10:17:16 +0000 (10:17 +0000)]
Remove unused TYPE_WD and TYPE_WFD.

21 years agoCorrectly calculate dmadat: We need to take the address of _end, it's
phk [Tue, 8 Oct 2002 10:15:42 +0000 (10:15 +0000)]
Correctly calculate dmadat:  We need to take the address of _end, it's
contents is irrelevant and likely to be zero;

This doesn't change the resultant value, but it does save a couple of bytes
because &_end is constant.

21 years agoMake it possible to kldload(8) many modules.
jmallett [Tue, 8 Oct 2002 09:57:03 +0000 (09:57 +0000)]
Make it possible to kldload(8) many modules.

MFC after: 2 months
Sponsored by: Bright Path Solutions
Reviewed by: tjr

21 years agoUnbreak boot2 by backing out rev 1.36 to Makefile, which does not
phk [Tue, 8 Oct 2002 07:52:28 +0000 (07:52 +0000)]
Unbreak boot2 by backing out rev 1.36 to Makefile, which does not
work as advertised:

bang# pwd
/bang/src/sys/boot/i386/boot2
bang# make clean >& /dev/null
bang# cvs -q update -r 1.35 Makefile >& /dev/null
bang# make >& /dev/null
bang# cat /usr/obj/`pwd`/boot2.h
#define XREADORG 0x725
bang# cvs -q update -r 1.36 Makefile > & /dev/null
bang# make clean > & /dev/null
bang# make > & /dev/null
bang# cat /usr/obj/`pwd`/boot2.h
#define XREADORG 0x25
bang#

21 years agoAllow for multiple #includes of this file.
phk [Tue, 8 Oct 2002 07:24:19 +0000 (07:24 +0000)]
Allow for multiple #includes of this file.

21 years agoSave four bytes by shortening a string two chars.
phk [Tue, 8 Oct 2002 07:06:24 +0000 (07:06 +0000)]
Save four bytes by shortening a string two chars.

Sponsored by: DARPA & NAI Labs.

21 years agoFor now, don't wait for drives to stop returning EBUSY. There is too
phk [Tue, 8 Oct 2002 07:03:58 +0000 (07:03 +0000)]
For now, don't wait for drives to stop returning EBUSY.  There is too
much broken harware around it seems.

Sponsored by: DARPA & NAI Labs.

21 years agoFreeBSD 4.7 enters the BSD family tree sometime on October 8, 2002.
murray [Tue, 8 Oct 2002 05:25:58 +0000 (05:25 +0000)]
FreeBSD 4.7 enters the BSD family tree sometime on October 8, 2002.

21 years agoTeach pkg_add -r how to find 4.7-RELEASE packages with an appropriate
bmah [Tue, 8 Oct 2002 05:15:21 +0000 (05:15 +0000)]
Teach pkg_add -r how to find 4.7-RELEASE packages with an appropriate
osreldate.

(Actually, due to differences in package compression formats, I'm
not sure that a -CURRENT pkg_add -r will do the right thing in
this case, once it finds them.)

21 years agoWe now use XFree86-4 packages rather than a special XFree86-3
murray [Tue, 8 Oct 2002 05:06:05 +0000 (05:06 +0000)]
We now use XFree86-4 packages rather than a special XFree86-3
distribution, so this script is no longer needed.

21 years agoTreat the pathptrn field as a real pattern with the aid of fnmatch().
dd [Tue, 8 Oct 2002 04:21:54 +0000 (04:21 +0000)]
Treat the pathptrn field as a real pattern with the aid of fnmatch().

21 years agoImport the libc fnmatch() into the kernel. This will be used by,
dd [Tue, 8 Oct 2002 04:15:55 +0000 (04:15 +0000)]
Import the libc fnmatch() into the kernel.  This will be used by,
among other things, the DEVFS rule subsystem to match nodes against a
path pattern supplied by the user.

fnmatch.c was repo-copied from src/lib/libc/gen/fnmatch.c, and the
only changes to it are those necessary to make it compile in the
kernel.  The relevant parts of fnmatch.h were imported into libkern.h.

Approved by: -arch

21 years agoMake kldunload cbb work:
imp [Tue, 8 Oct 2002 03:53:52 +0000 (03:53 +0000)]
Make kldunload cbb work:
o Implement the thread killing interlock as described by jhb in arch@
  while talking to markm.
o Hold Giant around cbb_insert()/cbb_remove().  Deep in the belly of
  the vm code we panic if we don't hold this when we activate the memory
  for reading the CIS.
o If we had to do the kludge alloc, then do a kludge free.

21 years agoAdd device driver for Belkin F5U103 and compatible USB-to-serial adapters.
kan [Tue, 8 Oct 2002 03:09:57 +0000 (03:09 +0000)]
Add device driver for Belkin F5U103 and compatible USB-to-serial adapters.

Reviewed by: n_hibma
Approved by: obrien

21 years agoRegen after usbdevs rev. 1.102.
kan [Tue, 8 Oct 2002 03:04:14 +0000 (03:04 +0000)]
Regen after usbdevs rev. 1.102.

Reviewed by: n_hibma
Approved by: obrien

21 years agoAdd entries for:
kan [Tue, 8 Oct 2002 03:00:37 +0000 (03:00 +0000)]
Add entries for:
Belkin F5U103
Belkin F5U120-PC Hub
GoHubs GoCOM232
Peracom single port
USB-to-serial adapters.

Reviewed by: n_hibma
Approved by: obrien

21 years agoRe-add the code which maps POSIX standard library names into the ones
tjr [Tue, 8 Oct 2002 02:19:54 +0000 (02:19 +0000)]
Re-add the code which maps POSIX standard library names into the ones
FreeBSD uses; f.e. -lpthread -> -pthread, -lxnet -> nothing.

21 years agoIntroducing /dev/devctl. This device reports events in the
imp [Mon, 7 Oct 2002 23:17:44 +0000 (23:17 +0000)]
Introducing /dev/devctl.  This device reports events in the
configuration device hierarchy.  Device arrival, departure and not
matched are presently reported.  This will be the basis for devd, which
I still need to polish a little more before I commit it.  If you don't
use /dev/devctl, it will be a noop.

21 years agoTwo minor bugfixes:
imp [Mon, 7 Oct 2002 23:15:40 +0000 (23:15 +0000)]
Two minor bugfixes:
o Allow the bus_debug variable to be set via the bus.debug tunable.
o Return pnpinfo and location info via the devinfo interface to userland.
  devinfo(8) needs to be updated to print it.

21 years agoMFp4:
imp [Mon, 7 Oct 2002 23:11:29 +0000 (23:11 +0000)]
MFp4:
o Better resume code.  Move the comments around.  Force the socket state to
  be querried.  Ack the interrupts properly.
o Intercept the interrupt requests and keep a list of interrupts to service
  ourselves.  When the card attaches, set its OK bit.  When we get a card
  status change interrupt for that card, clear the OK bit.  Don't call the
  ISR if the OK bit is cleared.  Iwasaki-san and yamamoto-san have both
  sent me patches that fix the same problem this fixes, but at the pccard
  level.
o Try to get the signalling of the thread to actually die.  This might not be
  100% right, but it is less wrong than before.
o Add a SIC next to a TI type that looks like it could be wrong, but isn't.

21 years agoo Add a nomatch routine so we whine when we don't know what to attach to
imp [Mon, 7 Oct 2002 23:05:33 +0000 (23:05 +0000)]
o Add a nomatch routine so we whine when we don't know what to attach to
  the card.
o Add comments about how we're doing the CIS activation.
o Add location and pnp info functions.
o Add better code to hopefully deal with ata cards better (and other drivers
  that allocate resources that we didn't preallocate from the CIS).  OLDCARD
  used to allow it, but NEWCARD was pickier.  I'm not 100% sure this works,
  but it doesn't break anything.

21 years agoMerge changes from NetBSD through version 1.17 of this file. These
imp [Mon, 7 Oct 2002 23:03:17 +0000 (23:03 +0000)]
Merge changes from NetBSD through version 1.17 of this file.  These
give us slightly better error checking than before and interpret what
default bits mean better.  See the NetBSD CVS tree for the authors of
these changes (revs 1.10 .. 1.17).

21 years agoo Add routines to return the location and pnpinfo for this card.
imp [Mon, 7 Oct 2002 23:00:51 +0000 (23:00 +0000)]
o Add routines to return the location and pnpinfo for this card.
  Note, we return the PCI pnp info, but in fact that's wrong to do
  since that data is not defined for CardBus cards.  CardBus says that
  these registers are undefined and one should use the CIS to do
  device matching.  To date, all CardBus cards have had these
  registered defined, no doubt because they are using common silicon
  to produce both the PCI cards and the CardBus cards.  However, it isn't
  any worse than the rest of the system, so just note it in passing and
  move on.
o Also sort prototypes while I'm here.

21 years agoDon't abuse the fact that -a == ~a + 1. Signed vs unsigned checkers
imp [Mon, 7 Oct 2002 22:58:24 +0000 (22:58 +0000)]
Don't abuse the fact that -a == ~a + 1.  Signed vs unsigned checkers
complain when a is a unsigned type.  So instead use the latter here
and be on our way.

Spotted by: flexlint by way of phk

21 years agoAdd extra set of parens around the barbit macro to make it clear what
imp [Mon, 7 Oct 2002 22:57:07 +0000 (22:57 +0000)]
Add extra set of parens around the barbit macro to make it clear what
we're intending to shift.

Spotted by: flexlint by way of phk (should fix about 40 messages)

21 years agoCorrect a bug in adding 0x700 to a number.
green [Mon, 7 Oct 2002 22:21:16 +0000 (22:21 +0000)]
Correct a bug in adding 0x700 to a number.

21 years agoConditionalize the number of sectors loaded by boot1.s on UFS1/UFS12.
phk [Mon, 7 Oct 2002 21:36:06 +0000 (21:36 +0000)]
Conditionalize the number of sectors loaded by boot1.s on UFS1/UFS12.

Conditionalize the "XX bytes left" checks reference on UFS1/UFS12.

Conditionally build the necessary 64bit math for boot2 if UFS12.

Sponsored by: DARPA & NAI Labs.

21 years agoRemove duplicate uninstalled aio.h header.
mike [Mon, 7 Oct 2002 21:18:30 +0000 (21:18 +0000)]
Remove duplicate uninstalled aio.h header.

21 years agoDocument behaviour change in reboot(8) introduced in reboot.c rev. 1.17:
thomas [Mon, 7 Oct 2002 21:18:20 +0000 (21:18 +0000)]
Document behaviour change in reboot(8) introduced in reboot.c rev. 1.17:
when using '-p' with reboot, and the power down action failds, reboot
the system normally. The behaviour of 'halt -p' and of shutdown(8) is
unchanged.

Approved by: roberto

21 years agoChange the comment character from # to // in boot1.s and run
phk [Mon, 7 Oct 2002 20:56:09 +0000 (20:56 +0000)]
Change the comment character from # to // in boot1.s and run
it through CPP so we can conditionalized things.

Sponsored by: DARPA & NAI Labs

21 years agoAdd back a fdrop() call at the end of kern_open() that got lost in
iedowse [Mon, 7 Oct 2002 20:49:22 +0000 (20:49 +0000)]
Add back a fdrop() call at the end of kern_open() that got lost in
revision 1.218. This bug caused a "struct file" reference to be
leaked if VOP_ADVLOCK(), vn_start_write(), or mac_check_vnode_write()
failed during the open operation.

PR: kern/43739
Reported by: Arne Woerner <woerner@mediabase-gmbh.de>

21 years agoTrack openbsd changes that don't affect us yet (PK supoprt is
sam [Mon, 7 Oct 2002 20:02:34 +0000 (20:02 +0000)]
Track openbsd changes that don't affect us yet (PK supoprt is
currently disabled):

o Don't use constants for the output parameter, use the iparam count as a
  pointer to the first result location.
o Fix bits vs bytes counting problems.
o Split out the hardware and software normalization versions of modexp.
o Enable hardware normalization for chips that support it.
o On reset, disable hardware normalization for 582x and make sure the
  chip is in little endian mode.
o Since sw normalization is now the only option, simplify normalization
  handling.

Also fix RNG harvesting: disabling PK support (for the moment) had disabled
the MCR2 interrupt; consider both KEY support and RNG support when deciding
whether or not to enable it.

Obtained from: openbsd

21 years agoCorrectly adjust for moved start address.
phk [Mon, 7 Oct 2002 19:12:36 +0000 (19:12 +0000)]
Correctly adjust for moved start address.

It seems that the existence of a "depend" target in src/sys/boot is not
to be taken as an indication that it actually does what one would expect,
at least it clearly threw my testing off.

Apologies to: jhb

21 years agoo split crypto_proc into two threads: one for processing requests and one
sam [Mon, 7 Oct 2002 18:46:38 +0000 (18:46 +0000)]
o split crypto_proc into two threads: one for processing requests and one
  for processing callbacks.  This closes race conditions caused by locking
  too many things with a single mutex.
o reclaim crypto requests under certain (impossible) failure conditions

21 years agoremove CIOGSSESSION (get software session); it was added only for testing
sam [Mon, 7 Oct 2002 18:37:31 +0000 (18:37 +0000)]
remove CIOGSSESSION (get software session); it was added only for testing

21 years agoRemove unneeded <sys/_posix.h> include.
mike [Mon, 7 Oct 2002 17:31:02 +0000 (17:31 +0000)]
Remove unneeded <sys/_posix.h> include.

21 years agoAdded '#include <sys/diskpc98.h>'.
nyan [Mon, 7 Oct 2002 15:26:10 +0000 (15:26 +0000)]
Added '#include <sys/diskpc98.h>'.

Submitted by: kawanobe@st.rim.or.jp (Kawanobe Koh)

21 years agoCorrectly compensate for both offset and unoffset on-disk BSD disklabels.
phk [Mon, 7 Oct 2002 13:11:28 +0000 (13:11 +0000)]
Correctly compensate for both offset and unoffset on-disk BSD disklabels.

Sponsored by: DARPA & NAI Labs.

21 years agoMerged from sys/isa/fd.c revisions 1.224 and 1.241.
nyan [Mon, 7 Oct 2002 13:05:45 +0000 (13:05 +0000)]
Merged from sys/isa/fd.c revisions 1.224 and 1.241.

21 years agoMove MEM_USR a page upwards to make space for larger UFS1 boot2.
phk [Mon, 7 Oct 2002 12:15:11 +0000 (12:15 +0000)]
Move MEM_USR a page upwards to make space for larger UFS1 boot2.

Load 4 sectors more than we used to.  This is harmless overhead for
the UFS1_ONLY case, but sufficient for boot2(UFS1+2).

Sponsored by: DARPA & NAI Labs

21 years agoDon't use DPCYL and DPSECT macros. These are not needed.
nyan [Mon, 7 Oct 2002 10:04:07 +0000 (10:04 +0000)]
Don't use DPCYL and DPSECT macros. These are not needed.

Submitted by: kawanobe@st.rim.or.jp (Kawanobe Koh)

21 years agoCheck the size of the dos_partition structure.
nyan [Mon, 7 Oct 2002 10:02:11 +0000 (10:02 +0000)]
Check the size of the dos_partition structure.

Submitted by: kawanobe@st.rim.or.jp (Kawanobe Koh)

21 years agoAdd the c99(1) utility, which is nothing more than a synonym for
tjr [Mon, 7 Oct 2002 09:37:55 +0000 (09:37 +0000)]
Add the c99(1) utility, which is nothing more than a synonym for
cc -std=iso9899:1999 -pedantic, and is required by SUSv3.

PR; 36087

21 years agoAdded bzip2 variants to the NAME section.
ru [Mon, 7 Oct 2002 09:08:07 +0000 (09:08 +0000)]
Added bzip2 variants to the NAME section.
Added missing break.
Remove the statement that -J requires zlib(3).

21 years agoUps, forgot to tell cvs commit about this file.
phk [Mon, 7 Oct 2002 08:14:10 +0000 (08:14 +0000)]
Ups, forgot to tell cvs commit about this file.
Move UFS1_ONLY to Makefiles instead of common/ufsread.c

Sponsored by: DARPA & NAI Labs

21 years agoMove the definition of UFS1_ONLY into the Makefiles where it belongs.
phk [Mon, 7 Oct 2002 08:09:20 +0000 (08:09 +0000)]
Move the definition of UFS1_ONLY into the Makefiles where it belongs.

Sponsored by: DARPA & NAI Labs.

21 years agoCorrectly convert to appropriate blocksize.
phk [Mon, 7 Oct 2002 07:19:14 +0000 (07:19 +0000)]
Correctly convert to appropriate blocksize.

note to self:
Never check block-size sensitive patches on Sun compatible CD drives.

Sponsored by: DARPA & NAI Labs.

21 years agoMFp4:
imp [Mon, 7 Oct 2002 07:18:32 +0000 (07:18 +0000)]
MFp4:
o Return the right string for CIS4_STR
o kill obsolete comment about not doing something that we now do.

21 years agoCorrectly deal with non-DEVBSIZE drives.
phk [Mon, 7 Oct 2002 07:15:37 +0000 (07:15 +0000)]
Correctly deal with non-DEVBSIZE drives.
Allow BIO_DELETE through too.

This fixes swap-backed md(4) devices.

Sponsored by:   DARPA & NAI Labs.

21 years agoAdd wrappers around the newly created bus_child_pnpinfo_str and
imp [Mon, 7 Oct 2002 07:08:00 +0000 (07:08 +0000)]
Add wrappers around the newly created bus_child_pnpinfo_str and
bus_child_location_str.