]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoThis commit was generated by cvs2svn to compensate for changes in r109983,
dougb [Tue, 28 Jan 2003 13:09:17 +0000 (13:09 +0000)]
This commit was generated by cvs2svn to compensate for changes in r109983,
which included commits to RCS files with non-trunk default branches.

21 years agoImport of ISC BIND version 8.3.4
dougb [Tue, 28 Jan 2003 13:09:17 +0000 (13:09 +0000)]
Import of ISC BIND version 8.3.4

21 years agoNO_GEOM cleanup: Don't include subr_disklabel.c in sparc64 build
phk [Tue, 28 Jan 2003 13:00:49 +0000 (13:00 +0000)]
NO_GEOM cleanup: Don't include subr_disklabel.c in sparc64 build

21 years agoBand-XXX-aid an easy to provoke panic.
phk [Tue, 28 Jan 2003 12:10:11 +0000 (12:10 +0000)]
Band-XXX-aid an easy to provoke panic.

MFC: 2 weeks

21 years agoNO_GEOM cleanup: Don't pull subr_disk{slice,label,mbr}.c into i386
phk [Tue, 28 Jan 2003 11:09:22 +0000 (11:09 +0000)]
NO_GEOM cleanup:  Don't pull subr_disk{slice,label,mbr}.c into i386
kernels anymore, they are not needed.

21 years agoAdd PCI revision number for 630A and 900B. Enable parity error detection
mbr [Tue, 28 Jan 2003 10:55:38 +0000 (10:55 +0000)]
Add PCI revision number for 630A and 900B. Enable parity error detection
on 900B and 635(A).

Re-add the enhanced PHY access register method again for older chipsets,
they do not seem to work with all old chips.

Reviewed by: phk
MFC after: 7 days

21 years agoPut the right fix in. Instead of deleting the declaration of __FBSDID, we
benno [Tue, 28 Jan 2003 10:54:31 +0000 (10:54 +0000)]
Put the right fix in.  Instead of deleting the declaration of __FBSDID, we
undef it before this definition.

21 years agoAdd more scaffolding to make this compile kernel code.
phk [Tue, 28 Jan 2003 09:48:59 +0000 (09:48 +0000)]
Add more scaffolding to make this compile kernel code.

21 years agoMark some args unused so this compiles in userland.
phk [Tue, 28 Jan 2003 09:48:09 +0000 (09:48 +0000)]
Mark some args unused so this compiles in userland.

21 years agoUse a void * to carry the private data for return-call'ed ioctl requests.
phk [Tue, 28 Jan 2003 09:47:50 +0000 (09:47 +0000)]
Use a void * to carry the private data for return-call'ed ioctl requests.
Amongst other things this avoids a complex workaround in the userland
regression bits.

21 years ago - Allow idle's pctcpu time to be calculated.
jeff [Tue, 28 Jan 2003 09:30:17 +0000 (09:30 +0000)]
 - Allow idle's pctcpu time to be calculated.

21 years ago - Fix the ksq_load calculation. It now reflects the number of entries on the
jeff [Tue, 28 Jan 2003 09:28:20 +0000 (09:28 +0000)]
 - Fix the ksq_load calculation.  It now reflects the number of entries on the
   run queue for each cpu.
 - Introduce kse stealing into the sched_choose() code.  This helps balance
   cpus better in cases where process turnover is high.  This implementation
   is fairly trivial and will likely be only a temporary measure until
   something more sophisticated has been written.

21 years agoDo not allow a cached vnode to be shared among multiple mounts of the same
tjr [Tue, 28 Jan 2003 09:21:42 +0000 (09:21 +0000)]
Do not allow a cached vnode to be shared among multiple mounts of the same
kind of pseudofs-based filesystem. Fixes (at least) one problem where
when procfs is mounted mupltiple times, trying to unmount one will often
cause the wrong one to get unmounted, and other problem where mounting
one procfs on top of another caused the kernel to lock up.

Reviewed by: des

21 years agoBang! Bang! Bang! etc etc
phk [Tue, 28 Jan 2003 09:04:19 +0000 (09:04 +0000)]
Bang! Bang! Bang! etc etc

Remove NODEVFS and NO_GEOM from options.

Approved by: trb@

21 years agostyle(9): add parentheses to sizeof even when not strictly required.
des [Tue, 28 Jan 2003 08:04:40 +0000 (08:04 +0000)]
style(9): add parentheses to sizeof even when not strictly required.

MFC after: 3 days

21 years agoBang! Bang! Bang! etc etc.
phk [Tue, 28 Jan 2003 07:15:22 +0000 (07:15 +0000)]
Bang! Bang! Bang! etc etc.

Remove NODEVFS option.

21 years agoA few fixes to rev 1.221
silby [Tue, 28 Jan 2003 03:39:39 +0000 (03:39 +0000)]
A few fixes to rev 1.221

- Honor the previous behavior of maxfragpackets = 0 or -1
- Take a better stab at fragment statistics
- Move / correct a comment

Suggested by: maxim@
MFC after: 7 days

21 years agoMerge pmap_testbit() and pmap_is_modified(). The latter is the only caller
alc [Tue, 28 Jan 2003 03:01:35 +0000 (03:01 +0000)]
Merge pmap_testbit() and pmap_is_modified().  The latter is the only caller
of the former.

21 years agoFix problems with how libufs was used, with regard to mounted/active fs's,
jmallett [Tue, 28 Jan 2003 02:42:01 +0000 (02:42 +0000)]
Fix problems with how libufs was used, with regard to mounted/active fs's,
in the new world order of libufs, where we also do statfs, and add a missing
close.

21 years agoThis is just a forced-commit to note that the previous commit addresses
gad [Tue, 28 Jan 2003 01:35:52 +0000 (01:35 +0000)]
This is just a forced-commit to note that the previous commit addresses
PR 28773 (not 28733!).  I have closed both PR 28773 and PR 46890.

21 years agoChanges so the 'pw' command will allow '$' as the last character in a userid
gad [Tue, 28 Jan 2003 01:21:57 +0000 (01:21 +0000)]
Changes so the 'pw' command will allow '$' as the last character in a userid
or group name (mainly for the benefit of samba).  This pretty much rewrites
he pw_checkname() routine, but should work exactly the same except for the
above change, and that error messages are somewhat more informative.

PR: 28733 46890
Inspired by: example patch written by Terry Lambert
Reviewed by: no objections on freebsd-arch and freebsd-current
MFC plans: no plans, but will do if people want it in stable.

21 years agoFix signed/unsigned comparison warning/error from 'make release'
jwd [Tue, 28 Jan 2003 00:33:53 +0000 (00:33 +0000)]
Fix signed/unsigned comparison warning/error from 'make release'

21 years agohint.acpi.0.disable -> hint.acpi.0.disabled
fjoe [Mon, 27 Jan 2003 23:30:21 +0000 (23:30 +0000)]
hint.acpi.0.disable -> hint.acpi.0.disabled

Submitted by:   Vitaly Markitantov <vm@dics.com.ua>

21 years agoNo longer force COMPAT_FREEBSD4 to be on.
peter [Mon, 27 Jan 2003 23:01:03 +0000 (23:01 +0000)]
No longer force COMPAT_FREEBSD4 to be on.

21 years agoFix signed/unsigned comparison warnings.
mbr [Mon, 27 Jan 2003 22:46:50 +0000 (22:46 +0000)]
Fix signed/unsigned comparison warnings.

Reviewed by: phk
Obtained from: NetBSD

21 years agoFix signed/unsigned comparison warnings. Fix spelling error in
mbr [Mon, 27 Jan 2003 22:45:08 +0000 (22:45 +0000)]
Fix signed/unsigned comparison warnings. Fix spelling error in
comment.

NetBSD Rev. 1.9 and 1.7

Reviewed by: phk
Obtained from: NetBSD

21 years agoMake this work in the !INET6 case -- if we mismatch the AF, don't return a
mbr [Mon, 27 Jan 2003 22:43:23 +0000 (22:43 +0000)]
Make this work in the !INET6 case -- if we mismatch the AF, don't return a
bogus (uninitialized) structure. Also, ignore v4 ifa's with no broadcast
address (rather than core dumping).

NetBSD Rev 1.8

Reviewed by: phk
Obtained from: NetBSD

21 years agoCheck pmap_flag before sendto.
mbr [Mon, 27 Jan 2003 22:38:25 +0000 (22:38 +0000)]
Check pmap_flag before sendto.

NetBSD r 1.5

Reviewed by: phk
Obtained from: NetBSD

21 years agoMake sure we don't look before the beginning of the string.
mbr [Mon, 27 Jan 2003 22:36:53 +0000 (22:36 +0000)]
Make sure we don't look before the beginning of the string.

NetBSD Rev 1.5

Reviewed by: phk
Obtained from: NetBSD

21 years agoAdd missing __rpc_fixup_addr. This is needed to make
mbr [Mon, 27 Jan 2003 22:27:55 +0000 (22:27 +0000)]
Add missing __rpc_fixup_addr.  This is needed to make
mount_nfs -T work for scoped addresses.

NetBSD Rev 1.11

Reviewed by: phk
Obtained from: NetBSD

21 years agoFree the correct buffer in error handling.
mbr [Mon, 27 Jan 2003 22:22:59 +0000 (22:22 +0000)]
Free the correct buffer in error handling.
Handle that malloc may return NULL.

NetBSD Rev. 1.8

Reviewed by: phk
Obtained from: NetBSD

21 years agoReset the record lenght and received bytes once a record
mbr [Mon, 27 Jan 2003 22:19:32 +0000 (22:19 +0000)]
Reset the record lenght and received bytes once a record
is finished. This fixes clients doing two RPCs over the
same connection at the same time. Without this fix, we
could end with a reply to old data.

Submitted by: Frank van der Linden <fvdl@netbsd.org>
Reviewed by: rwatson
Obtained from: NetBSD

21 years agoUpdate mount_nfs documentation to reference new sysctl for -c default.
dillon [Mon, 27 Jan 2003 20:12:32 +0000 (20:12 +0000)]
Update mount_nfs documentation to reference new sysctl for -c default.

21 years agoAdd the sabtty children devices as unordered. Use the unit numbr of
jake [Mon, 27 Jan 2003 18:39:09 +0000 (18:39 +0000)]
Add the sabtty children devices as unordered.  Use the unit numbr of
the sabtty device to create its description so that they will be
unique for machines with multiple sab chips.

This fixes a panic on machines with an rsc card.

Tested by: obrien (e250)

21 years agoA few changes for clarity.
trhodes [Mon, 27 Jan 2003 18:16:36 +0000 (18:16 +0000)]
A few changes for clarity.

PR: 47170
Submitted by: Gary W. Swearingen <swear@attbi.com> (original version)

21 years agoDon't try to fseek before the beginning of the file; POSIX requires that
fenner [Mon, 27 Jan 2003 18:16:32 +0000 (18:16 +0000)]
Don't try to fseek before the beginning of the file; POSIX requires that
 this return an error.
This re-enables display of the first entry in /var/account/acct instead
 of the error "lastcomm: /var/account/acct: Invalid argument"

21 years agoUnbreak i386 snapshots by moving the 802.11 support (device wlan)
ru [Mon, 27 Jan 2003 17:54:49 +0000 (17:54 +0000)]
Unbreak i386 snapshots by moving the 802.11 support (device wlan)
out from BOOTMFS to the drivers floppy.

21 years agoDon't dereference null vnode pointer if controling terminal was revoked.
phk [Mon, 27 Jan 2003 16:54:17 +0000 (16:54 +0000)]
Don't dereference null vnode pointer if controling terminal was revoked.

Submitted by: "Peter Edwards" <pmedwards@eircom.net>

21 years agoActually extract the second field from a line in ftpchroot(5)
yar [Mon, 27 Jan 2003 15:34:22 +0000 (15:34 +0000)]
Actually extract the second field from a line in ftpchroot(5)
instead of just using the rest of the line behind the first field.

21 years agoAllow more than one separator character between fields in ftpchroot(5).
yar [Mon, 27 Jan 2003 14:41:08 +0000 (14:41 +0000)]
Allow more than one separator character between fields in ftpchroot(5).

21 years agoAdd PCI id for fxp(Intel Pro/100 M Mobile Connection).
sanpei [Mon, 27 Jan 2003 14:31:32 +0000 (14:31 +0000)]
Add PCI id for fxp(Intel Pro/100 M Mobile Connection).

PR: i386/41915
Submitted by: Ross A. Lippert <ripper@eskimo.com>
MFC after: 1 week

21 years agoBack the previous commit out. It didn't actually fix the problem I was
benno [Mon, 27 Jan 2003 12:59:52 +0000 (12:59 +0000)]
Back the previous commit out.  It didn't actually fix the problem I was
seeing and the memory barrier isn't needed with the bridges we're using.

Fix the function style however.

21 years agoPut the Perfection 1660 scanner entry in the right place. (Entries
joe [Mon, 27 Jan 2003 09:43:24 +0000 (09:43 +0000)]
Put the Perfection 1660 scanner entry in the right place.  (Entries
are sorted by product id).

21 years agoUpdate the code that deals with disk enclosures:
sos [Mon, 27 Jan 2003 09:04:29 +0000 (09:04 +0000)]
Update the code that deals with disk enclosures:

Properly handle the newer Promise SuperSwap 1000 enclosures.
Print out what kind of enclosure was found in the probe.
Misc cleanups in the enclosure handling code.

Sponsored by: Advanis Inc.

21 years agoDon't use the kern.disks sysctl to find a list of md devices, use the
phk [Mon, 27 Jan 2003 08:01:08 +0000 (08:01 +0000)]
Don't use the kern.disks sysctl to find a list of md devices, use the
MDIOCLIST ioctl instead.

Spotted by: keramida

21 years agoImplement MDIOCLIST which returns the unit numbers of configured md(4)
phk [Mon, 27 Jan 2003 07:58:18 +0000 (07:58 +0000)]
Implement MDIOCLIST which returns the unit numbers of configured md(4)
devices.

We use the md_pad[] array and if there are more units than its size the
last returned unit number will be -1, but the number of units returned
is correct.

21 years agoAdd an ioctl to return the unit numbers of the configured md devices.
phk [Mon, 27 Jan 2003 07:54:59 +0000 (07:54 +0000)]
Add an ioctl to return the unit numbers of the configured md devices.

21 years agoEnsure that the TTY file descriptor is greater than or equal to 10 so that
tjr [Mon, 27 Jan 2003 07:41:12 +0000 (07:41 +0000)]
Ensure that the TTY file descriptor is greater than or equal to 10 so that
it doesn't interfere with the user's redirections.

PR: 47136
MFC after: 1 week

21 years agoConvert newfs to use libufs. I've tested this on md filesystems, as has
jmallett [Mon, 27 Jan 2003 07:24:32 +0000 (07:24 +0000)]
Convert newfs to use libufs.  I've tested this on md filesystems, as has
keramida, and all seems well.

21 years agoMFp4:
imp [Mon, 27 Jan 2003 05:47:01 +0000 (05:47 +0000)]
MFp4:
u_int*_t -> uint*_t to conform more closely with C99.

21 years agoEnable device zs and device sab by default.
jake [Mon, 27 Jan 2003 05:05:52 +0000 (05:05 +0000)]
Enable device zs and device sab by default.

21 years agoAdd a note about the ofwcons device name change.
jake [Mon, 27 Jan 2003 04:53:30 +0000 (04:53 +0000)]
Add a note about the ofwcons device name change.

21 years agoTeach sysinstall about the em(4) device.
trhodes [Mon, 27 Jan 2003 04:51:46 +0000 (04:51 +0000)]
Teach sysinstall about the em(4) device.

PR: 46439
Submitted by: Dan Lukes <dan@obluda.cz>
Approved by: re (murray)
Tested? yes
MFC: 1 day

21 years agoChange ofwcons to use the output-device property from the firmware for the
jake [Mon, 27 Jan 2003 04:42:17 +0000 (04:42 +0000)]
Change ofwcons to use the output-device property from the firmware for the
name of the device that it creates.  Update /etc/ttys accordingly.

An alias is created for the old name so that old /etc/ttys will continue to
work, but due to aliases being implemented as symlinks in devfs you cannot
login as root when using the alias device.

Discussed with: grehan

21 years agoBack out some changes that snuck in with the last commit.
benno [Mon, 27 Jan 2003 04:32:10 +0000 (04:32 +0000)]
Back out some changes that snuck in with the last commit.

Pointy hat to: benno

21 years agoFlesh out bus_dmamap_sync.
benno [Mon, 27 Jan 2003 04:27:01 +0000 (04:27 +0000)]
Flesh out bus_dmamap_sync.

21 years agoUse td->td_sticks, not td->td_kse->ke_sticks.
benno [Mon, 27 Jan 2003 04:26:09 +0000 (04:26 +0000)]
Use td->td_sticks, not td->td_kse->ke_sticks.

Forgotten by: davidxu

21 years agoRemove a duplicate definition of the __FBSDID macro.
benno [Mon, 27 Jan 2003 04:24:57 +0000 (04:24 +0000)]
Remove a duplicate definition of the __FBSDID macro.

21 years agoNow return NULLified struct in case of empty config file
ache [Mon, 27 Jan 2003 03:39:33 +0000 (03:39 +0000)]
Now return NULLified struct in case of empty config file
(previous variant return NULL pointer for both empty file case and error case,
so caller can't sense error properly).

It not affect existen programs because property_find() now returns NULL
for both NULL pointer and NULLified struct.

21 years agoHandle read errors
ache [Mon, 27 Jan 2003 03:14:04 +0000 (03:14 +0000)]
Handle read errors

21 years agoNote that -c is no longer supported. This was requested by a few users and
trhodes [Mon, 27 Jan 2003 02:05:59 +0000 (02:05 +0000)]
Note that -c is no longer supported.  This was requested by a few users and
we even have a PR about it.

PR: 47388
Discussed with: bmah
Requested by: many

21 years agoIf we don't know where the sblock is (e.g. filling out a blank disk), then
jmallett [Mon, 27 Jan 2003 01:57:15 +0000 (01:57 +0000)]
If we don't know where the sblock is (e.g. filling out a blank disk), then
get it from the fs structure.  Really libufs should have interfaces to generate
both what we export, and what we import, based on eachother, and this should
be full of redundant code to make sure everything is right...  But really, we
don't even deal with checksums, so plenty of room to improve.

21 years agoSimplify vm_object_page_remove(): The object's memq is now ordered. The
alc [Mon, 27 Jan 2003 01:12:35 +0000 (01:12 +0000)]
Simplify vm_object_page_remove(): The object's memq is now ordered.  The
two cases that existed before for performance optimization purposes can
be reduced to one.

21 years agoUnbreak SMP cases for these architectures.
julian [Mon, 27 Jan 2003 00:00:06 +0000 (00:00 +0000)]
Unbreak SMP cases for these architectures.
statclock_process() changed arguments.
note: it may be worth checking if curkse is needed on these architectures..
(and if so, why?)

21 years agoFix standard kse breakage of non-x86 platforms.
jake [Sun, 26 Jan 2003 23:52:45 +0000 (23:52 +0000)]
Fix standard kse breakage of non-x86 platforms.

Pointy hat to: davidxu

21 years agoUse kg_numupcalls to see if we are closing a thread group,
davidxu [Sun, 26 Jan 2003 23:39:33 +0000 (23:39 +0000)]
Use kg_numupcalls to see if we are closing a thread group,
not kg_kses which is not changed when a group is still working.

21 years agoVery minor EOL whitespace diff-reducer.
markm [Sun, 26 Jan 2003 23:34:47 +0000 (23:34 +0000)]
Very minor EOL whitespace diff-reducer.

21 years agoMissed a bit of cleanup.
markm [Sun, 26 Jan 2003 23:29:22 +0000 (23:29 +0000)]
Missed a bit of cleanup.

21 years agofix warnings
alfred [Sun, 26 Jan 2003 23:25:00 +0000 (23:25 +0000)]
fix warnings

21 years agomake these more useful for lint(1). Minor diff-reductions while I'm
markm [Sun, 26 Jan 2003 23:14:47 +0000 (23:14 +0000)]
make these more useful for lint(1). Minor diff-reductions while I'm
about it.

21 years agoFix namespace pollution introduced in previous commit.
mbr [Sun, 26 Jan 2003 23:01:49 +0000 (23:01 +0000)]
Fix namespace pollution introduced in previous commit.

Reviewed by: phk

21 years agoRemove the get_term() function. It pretty much can't work for
markm [Sun, 26 Jan 2003 23:01:36 +0000 (23:01 +0000)]
Remove the get_term() function. It pretty much can't work for
FreeBSD, and makes ugly diffs with the other crt1.c's. Leave
behind a comment (words supplied by Thomas Moestl) that explain
the issue.

OK'ed by: tmm

21 years agoFix a typo and reword a bit the description of the new SCHED_xxx stuff.
keramida [Sun, 26 Jan 2003 22:26:46 +0000 (22:26 +0000)]
Fix a typo and reword a bit the description of the new SCHED_xxx stuff.

Reviewed by: jeff

21 years agoIf we fail to open O_RDWR, try opening O_RDONLY, if we use ioctls
phk [Sun, 26 Jan 2003 21:55:43 +0000 (21:55 +0000)]
If we fail to open O_RDWR, try opening O_RDONLY, if we use ioctls
to fiddle the disk we can get away with it.

Try to use DIOCBSDBB to write boot code.

21 years agoImplement DIOCBSDBB ioctl which overwrites first BBSIZE bytes of BSD
phk [Sun, 26 Jan 2003 21:54:36 +0000 (21:54 +0000)]
Implement DIOCBSDBB ioctl which overwrites first BBSIZE bytes of BSD
labeled disk.

This is complicated by the fact that BBSIZE is greater than the
PAGE_SIZE limit ioctl inflicts on arguments which are automatically
copied in.

As long as we don't need access to userland memory (copyin/out) we
can deal with the ioctl using g_callme() which executes it from the
GEOM event thread.

Once we need copyin/out, we need to return the bio with EDIRIOCTL
in order to make geom_dev call us back in the original process context
where copyin will work.

Unfortunately, that results in us getting called with Giant, so
we have to DROP_GIANT/PICKUP_GIANT around the code where we diddle
GEOMs internals.

Sometimes you just can't win...

... But it does make geom_bsd.c an almost complete example of the
GEOM beastiarium.

21 years agoAdd ioctl which overwrites the first BBSIZE bytes of a BSD labeled disk.
phk [Sun, 26 Jan 2003 21:49:44 +0000 (21:49 +0000)]
Add ioctl which overwrites the first BBSIZE bytes of a BSD labeled disk.

21 years agoFix KSE related patch.
julian [Sun, 26 Jan 2003 21:32:08 +0000 (21:32 +0000)]
Fix KSE related patch.
Make it compile for the SMP case..
statclock_process() has changed prototypes.

21 years agoCorrect typo.
murray [Sun, 26 Jan 2003 21:14:56 +0000 (21:14 +0000)]
Correct typo.

Submitted by: Andreas Kohn <andreas.kohn@gmx.net> (via -STABLE)

21 years agoregen
alfred [Sun, 26 Jan 2003 20:10:35 +0000 (20:10 +0000)]
regen

21 years agoAdd const qualifier to data argument for msgsnd.
alfred [Sun, 26 Jan 2003 20:09:34 +0000 (20:09 +0000)]
Add const qualifier to data argument for msgsnd.

PR: standards/45274
Submitted by: Craig Rodrigues <rodrigc@attbi.com>

21 years agoPut DIOCWLABEL under #ifdef NO_GEOM
phk [Sun, 26 Jan 2003 19:43:18 +0000 (19:43 +0000)]
Put DIOCWLABEL under #ifdef NO_GEOM

21 years agoExtend the format of /etc/ftpchroot so an alternative chroot
yar [Sun, 26 Jan 2003 19:02:56 +0000 (19:02 +0000)]
Extend the format of /etc/ftpchroot so an alternative chroot
directory can be specified for a user or a group.

Add the manpage ftpchroot(5) since the file's format has grown
complex enough.

PR: bin/45327
Portions submitted by: Hideki SAKAMOTO <sakamoto@hlla.is.tsukuba.ac.jp>
MFC after: 1 week

21 years ago- Split db.desc.cmd into desc.control and desc.reqcount.
simokawa [Sun, 26 Jan 2003 18:38:06 +0000 (18:38 +0000)]
- Split db.desc.cmd into desc.control and desc.reqcount.
- remove unncessary descriptor updates.

21 years agoIA64 still needs to be able to run on MBR+BSD combination, so put disklabel
phk [Sun, 26 Jan 2003 18:33:54 +0000 (18:33 +0000)]
IA64 still needs to be able to run on MBR+BSD combination, so put disklabel
back.

Pointed out by: peter

21 years ago- Improve IT/IR DMA queue management.
simokawa [Sun, 26 Jan 2003 15:39:04 +0000 (15:39 +0000)]
- Improve IT/IR DMA queue management.
- Improve debug message for mbuf handling.
- Wait 1 sec for DMA stop in fwohci_i{t,r}x_disable() before freeing buffers.

21 years agoDrop the silly notion that i386 has two bootstrap files now that sys/boot
phk [Sun, 26 Jan 2003 14:35:53 +0000 (14:35 +0000)]
Drop the silly notion that i386 has two bootstrap files now that sys/boot
creates a single file named just "boot".

Apart from the fact that the option "-s" is now gone and that "-b" should
be pointed at /boot/boot instead of /boot/boot1, this patch should be
a no-op.

21 years agoLink /boot/boot1 to the name /boot/boot to avoid per-arch naming of the
phk [Sun, 26 Jan 2003 14:32:53 +0000 (14:32 +0000)]
Link /boot/boot1 to the name /boot/boot to avoid per-arch naming of the
bootstrap code for disklabel using architectures.

21 years agoRemove #if checks for NUMBOOT==0, it's not relevant for any architecture
phk [Sun, 26 Jan 2003 13:50:11 +0000 (13:50 +0000)]
Remove #if checks for NUMBOOT==0, it's not relevant for any architecture
and if we get such an architecture, we can just avoid using the relevant
options.

21 years agoBuild a file "boot" which consists of boot1 and boot2 concatenated.
phk [Sun, 26 Jan 2003 13:33:57 +0000 (13:33 +0000)]
Build a file "boot" which consists of boot1 and boot2 concatenated.

There is little if any reason to treat the two components separately
and it will simplify disklabel(8) and libdisk if we didn't.

21 years agoThis file is no longer needed.
phk [Sun, 26 Jan 2003 13:08:49 +0000 (13:08 +0000)]
This file is no longer needed.

Spotted by: tjr

21 years agoOffer better advice in #error.
phk [Sun, 26 Jan 2003 12:39:46 +0000 (12:39 +0000)]
Offer better advice in #error.

Remove yet a usage message about -N/-W

21 years agoSpell libufs correctly in DPADD.
ru [Sun, 26 Jan 2003 12:34:36 +0000 (12:34 +0000)]
Spell libufs correctly in DPADD.

Reviewed by: juli

21 years agoRemove ia64 and powerpc #ifdefs: this is i386/pc98/alpha only code.
phk [Sun, 26 Jan 2003 12:22:26 +0000 (12:22 +0000)]
Remove ia64 and powerpc #ifdefs: this is i386/pc98/alpha only code.

21 years agoMake disklabel(8) MD for i386/pc98 and alpha only.
phk [Sun, 26 Jan 2003 12:19:36 +0000 (12:19 +0000)]
Make disklabel(8) MD for i386/pc98 and alpha only.

21 years agoRemove part of my stateful locale patch that slipped into the previous rev.
tjr [Sun, 26 Jan 2003 11:45:54 +0000 (11:45 +0000)]
Remove part of my stateful locale patch that slipped into the previous rev.

21 years agoRemove #ifdef __sparc64__ code. GEOM uses native SUN labels on this arch,
phk [Sun, 26 Jan 2003 11:43:50 +0000 (11:43 +0000)]
Remove #ifdef __sparc64__ code.  GEOM uses native SUN labels on this arch,
so no hacks are needed.

21 years agoFix typo in last commit.
phk [Sun, 26 Jan 2003 11:41:37 +0000 (11:41 +0000)]
Fix typo in last commit.
Rearrange bits in writelabel() a bit for improved readability.

21 years agoMove UPCALL related data structure out of kse, introduce a new
davidxu [Sun, 26 Jan 2003 11:41:35 +0000 (11:41 +0000)]
Move UPCALL related data structure out of kse, introduce a new
data structure called kse_upcall to manage UPCALL. All KSE binding
and loaning code are gone.

A thread owns an upcall can collect all completed syscall contexts in
its ksegrp, turn itself into UPCALL mode, and takes those contexts back
to userland. Any thread without upcall structure has to export their
contexts and exit at user boundary.

Any thread running in user mode owns an upcall structure, when it enters
kernel, if the kse mailbox's current thread pointer is not NULL, then
when the thread is blocked in kernel, a new UPCALL thread is created and
the upcall structure is transfered to the new UPCALL thread. if the kse
mailbox's current thread pointer is NULL, then when a thread is blocked
in kernel, no UPCALL thread will be created.

Each upcall always has an owner thread. Userland can remove an upcall by
calling kse_exit, when all upcalls in ksegrp are removed, the group is
atomatically shutdown. An upcall owner thread also exits when process is
in exiting state. when an owner thread exits, the upcall it owns is also
removed.

KSE is a pure scheduler entity. it represents a virtual cpu. when a thread
is running, it always has a KSE associated with it. scheduler is free to
assign a KSE to thread according thread priority, if thread priority is changed,
KSE can be moved from one thread to another.

When a ksegrp is created, there is always N KSEs created in the group. the
N is the number of physical cpu in the current system. This makes it is
possible that even an userland UTS is single CPU safe, threads in kernel still
can execute on different cpu in parallel. Userland calls kse_create to add more
upcall structures into ksegrp to increase concurrent in userland itself, kernel
is not restricted by number of upcalls userland provides.

The code hasn't been tested under SMP by author due to lack of hardware.

Reviewed by: julian

21 years agoPut use of DIOCWLABEL under #ifdef NO_GEOM
phk [Sun, 26 Jan 2003 11:28:56 +0000 (11:28 +0000)]
Put use of DIOCWLABEL under #ifdef NO_GEOM

21 years agoRemember to remove -N and -W from usage.
phk [Sun, 26 Jan 2003 11:28:34 +0000 (11:28 +0000)]
Remember to remove -N and -W from usage.