]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoU areas are going away, so don't allocate them. It's worrisome that
das [Sat, 20 Nov 2004 02:29:36 +0000 (02:29 +0000)]
U areas are going away, so don't allocate them.  It's worrisome that
mp_machdep.c was using UAREA_PAGES to allocate something that isn't a
U area, and that there seems to be an implicit assumption that the PCB
is just past the end of the kernel stack.

Reviewed by: arch@

19 years agoU areas are going away, so don't allocate one for process 0.
das [Sat, 20 Nov 2004 02:29:25 +0000 (02:29 +0000)]
U areas are going away, so don't allocate one for process 0.

Reviewed by: arch@

19 years agoDisable U area swapping and remove the routines that create, destroy,
das [Sat, 20 Nov 2004 02:29:00 +0000 (02:29 +0000)]
Disable U area swapping and remove the routines that create, destroy,
copy, and swap U areas.

Reviewed by: arch@

19 years agoMalloc p_stats instead of putting it in the U area. We should consider
das [Sat, 20 Nov 2004 02:28:48 +0000 (02:28 +0000)]
Malloc p_stats instead of putting it in the U area.  We should consider
simply embedding it in struct proc.

Reviewed by: arch@

19 years agoIn preparation to remove U areas, don't refer to p_uarea in libkvm.
das [Sat, 20 Nov 2004 02:28:26 +0000 (02:28 +0000)]
In preparation to remove U areas, don't refer to p_uarea in libkvm.

Reviewed by: arch@

19 years agouser.h is included only to get pcb.h, so use the latter directly instead.
das [Sat, 20 Nov 2004 02:28:14 +0000 (02:28 +0000)]
user.h is included only to get pcb.h, so use the latter directly instead.

19 years agoBring signal(9) in sync with -current.
jkoshy [Sat, 20 Nov 2004 02:11:42 +0000 (02:11 +0000)]
Bring signal(9) in sync with -current.

Reviewed by: ru (mdoc changes)

19 years agoIncrease cnt.v_syscall and cnt.v_trap when needed.
cognet [Sat, 20 Nov 2004 00:55:57 +0000 (00:55 +0000)]
Increase cnt.v_syscall and cnt.v_trap when needed.

19 years agoDo not emit a spurious warning when "directory" argument
ru [Sat, 20 Nov 2004 00:41:08 +0000 (00:41 +0000)]
Do not emit a spurious warning when "directory" argument
to "rmdir -p" is absolute, with one or more leading slash.

19 years agoFixed "rmdir -p" that got broken by rev. 1.15.
ru [Sat, 20 Nov 2004 00:12:08 +0000 (00:12 +0000)]
Fixed "rmdir -p" that got broken by rev. 1.15.
(This also fixes "rmdir -v /nonexistent".)

19 years agoRemove unused macro.
glebius [Fri, 19 Nov 2004 22:35:53 +0000 (22:35 +0000)]
Remove unused macro.

Approved by: julian (mentor)

19 years agoDocument NO_NIS variable.
bz [Fri, 19 Nov 2004 22:23:57 +0000 (22:23 +0000)]
Document NO_NIS variable.

Submitted by: Pawel Worach <pawel dot worach at telia dot com>
Approved by: rwatson (mentor)
MFC after: 1 week

19 years agoCancel pending timeout before scheduling a new one. This fixes
glebius [Fri, 19 Nov 2004 20:23:46 +0000 (20:23 +0000)]
Cancel pending timeout before scheduling a new one. This fixes
item leak, I've introduced with previous change.

Approved by: julian (mentor)

19 years agoCorrect typo. Return ENXIO instead of EIO.
emax [Fri, 19 Nov 2004 18:54:54 +0000 (18:54 +0000)]
Correct typo. Return ENXIO instead of EIO.

19 years agoMan page for the HighPoint RocketRAID 182x driver.
obrien [Fri, 19 Nov 2004 18:24:51 +0000 (18:24 +0000)]
Man page for the HighPoint RocketRAID 182x driver.

Forgotten by: scottl

19 years agoRevert part of rev 1.57. The tag boundary is honored by splitting the
scottl [Fri, 19 Nov 2004 17:56:22 +0000 (17:56 +0000)]
Revert part of rev 1.57.  The tag boundary is honored by splitting the
segment, not by bouncing.

19 years agoRevert part of rev 1.56. Tag boundaries are handled by splitting segments,
scottl [Fri, 19 Nov 2004 17:51:29 +0000 (17:51 +0000)]
Revert part of rev 1.56.  Tag boundaries are handled by splitting segments,
not through bouncing.

19 years agoCatch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].
obrien [Fri, 19 Nov 2004 17:31:31 +0000 (17:31 +0000)]
Catch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].

19 years agoCatch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].
obrien [Fri, 19 Nov 2004 17:12:56 +0000 (17:12 +0000)]
Catch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].

19 years agoUse statfs instead of getmntinfo(). This will make the procfs checks
csjp [Fri, 19 Nov 2004 16:11:54 +0000 (16:11 +0000)]
Use statfs instead of getmntinfo(). This will make the procfs checks
play nicer in prisons. It also simplifies things.

Reviewed by: rwatson
Bumped into by: Jilles Tjoelker

19 years ago- Fix typo in portaudit reference.
simon [Fri, 19 Nov 2004 11:44:26 +0000 (11:44 +0000)]
- Fix typo in portaudit reference.
- Bump document date for last commit.

Pointed out by: ceri, ru
Pointy hat to: simon

19 years agoInstall the header for libmagic.
obrien [Fri, 19 Nov 2004 04:04:14 +0000 (04:04 +0000)]
Install the header for libmagic.

PR: 73647
Submitted by: Uranus <uranus@it.muds.net>

19 years agoCatch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].
obrien [Fri, 19 Nov 2004 03:56:47 +0000 (03:56 +0000)]
Catch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].

PR: 73879
Submitted by: Steve Kargl <sgk@troutmask.apl.washington.edu>

19 years agoCatch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].
obrien [Fri, 19 Nov 2004 03:53:36 +0000 (03:53 +0000)]
Catch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].

PR: 73878
Submitted by: Steve Kargl <sgk@troutmask.apl.washington.edu>

19 years agoMake logic more readable.
obrien [Fri, 19 Nov 2004 03:52:41 +0000 (03:52 +0000)]
Make logic more readable.

19 years agoCatch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].
obrien [Fri, 19 Nov 2004 03:51:12 +0000 (03:51 +0000)]
Catch up with PHK's sio(4) cuaa->cuad rework [sys/dev/sio/sio.c rev. 1.456].

19 years agoAdd the the Allied Telesyn AT-2700 familly.
obrien [Fri, 19 Nov 2004 03:47:01 +0000 (03:47 +0000)]
Add the the Allied Telesyn AT-2700 familly.

19 years agoPlace function comment above the right function.
marks [Fri, 19 Nov 2004 00:58:30 +0000 (00:58 +0000)]
Place function comment above the right function.

19 years agoRemove ntp_gettime.c which was a wrapper around sysctlbyname(3).
marks [Thu, 18 Nov 2004 23:59:19 +0000 (23:59 +0000)]
Remove ntp_gettime.c which was a wrapper around sysctlbyname(3).
This is now a native system call.

Reviewed by: imp, phk, njl, peter
Approved by: njl

19 years agoRebuild from compat/freebsd32/syscalls.master:1.43
marks [Thu, 18 Nov 2004 23:56:09 +0000 (23:56 +0000)]
Rebuild from compat/freebsd32/syscalls.master:1.43

Reviewed by: imp, phk, njl, peter
Approved by: njl

19 years ago32-bit FreeBSD ABI compatibility stubs from syscalls.master:1.179
marks [Thu, 18 Nov 2004 23:54:26 +0000 (23:54 +0000)]
32-bit FreeBSD ABI compatibility stubs from syscalls.master:1.179

Reviewed by: imp, phk, njl, peter
Approved by: njl

19 years agoRebuild from syscalls.master:1.179
marks [Thu, 18 Nov 2004 23:52:40 +0000 (23:52 +0000)]
Rebuild from syscalls.master:1.179

Reviewed by: imp, phk, njl, peter
Approved by: njl

19 years agoAdd ntp_gettime(2) system call.
marks [Thu, 18 Nov 2004 23:46:14 +0000 (23:46 +0000)]
Add ntp_gettime(2) system call.

Reviewed by: imp, phk, njl, peter
Approved by: njl

19 years agoAdd system call implementation of ntp_gettime(2).
marks [Thu, 18 Nov 2004 23:44:49 +0000 (23:44 +0000)]
Add system call implementation of ntp_gettime(2).
Moved most of the work to ntp_gettime1(), which is now called by
ntp_gettime() and ntp_sysctl().

Reviewed by: imp, phk, njl, peter
Approved by: njl

19 years agoAdd a couple of summary paragraphs at the head of mbuf_tags(9) to
rwatson [Thu, 18 Nov 2004 23:01:07 +0000 (23:01 +0000)]
Add a couple of summary paragraphs at the head of mbuf_tags(9) to
indicate how m_tags might be used, and what their high level properties
are.  This may help developers using tags get a more clear picture
before delving into the technical details of ABI cookies and API calls.

Suggested by: Rene de Vries <rene at canyon dot xs4all dot nl>

19 years agoAdd some useful target mode diagnostics for incoming commands
gibbs [Thu, 18 Nov 2004 20:33:43 +0000 (20:33 +0000)]
Add some useful target mode diagnostics for incoming commands
under the AHC_SHOW_TQIN debug flag.

19 years agoRevert to basing all timeout/timer values in ms rather than us. The switch
gibbs [Thu, 18 Nov 2004 20:22:31 +0000 (20:22 +0000)]
Revert to basing all timeout/timer values in ms rather than us.  The switch
to us was to help out the Linux port, but really just invited overflow.
In fact, the request sense timer was overflowing prior to this change making
it much shorter than intended.

aic_osm_lib.h:
Be more careful about overflow in all timer/timeout primitives.

19 years ago- Document the DISABLE_VULNERABILITIES knob.
simon [Thu, 18 Nov 2004 19:40:00 +0000 (19:40 +0000)]
- Document the DISABLE_VULNERABILITIES knob.
- Don't say this manual page is too long, it's not that long.
- s/man page/manual page/g

MFC after: 3 days

19 years agoCheck in updated bthidd(8). This is still work in progress.
emax [Thu, 18 Nov 2004 18:05:15 +0000 (18:05 +0000)]
Check in updated bthidd(8). This is still work in progress.

19 years agoAdd dropped implementation of ioctl for fifos.
phk [Thu, 18 Nov 2004 17:18:11 +0000 (17:18 +0000)]
Add dropped implementation of ioctl for fifos.

19 years agoOk, first blunder: ioctls are not entirely unused on vnodes anymore :-)
phk [Thu, 18 Nov 2004 17:15:04 +0000 (17:15 +0000)]
Ok, first blunder:  ioctls are not entirely unused on vnodes anymore :-)

Add dropped call to VOP_IOCTL().

19 years ago-Add a note that currently two syntax styles for label element declaration
csjp [Thu, 18 Nov 2004 15:24:47 +0000 (15:24 +0000)]
-Add a note that currently two syntax styles for label element declaration
 is supported.
-Document the new more preferred syntax
-Add examples for the new syntax
-Add a note that the old syntax will be deprecated in the future.

Reviewed by: rwatson

19 years agoPass path to filesystem when mounting root
phk [Thu, 18 Nov 2004 14:31:24 +0000 (14:31 +0000)]
Pass path to filesystem when mounting root

19 years agoMake chrootdir global and use it in log messages
yar [Thu, 18 Nov 2004 14:15:32 +0000 (14:15 +0000)]
Make chrootdir global and use it in log messages
regarding restricted users.

MFC after: 2 weeks

19 years agoNitpicking on style(9) and whitespace.
yar [Thu, 18 Nov 2004 13:55:15 +0000 (13:55 +0000)]
Nitpicking on style(9) and whitespace.

Tested with: md5(1)

19 years ago- Since divert protocol is not connection oriented, remove SS_ISCONNECTED flag
glebius [Thu, 18 Nov 2004 13:49:18 +0000 (13:49 +0000)]
- Since divert protocol is not connection oriented, remove SS_ISCONNECTED flag
  from divert sockets.
- Remove div_disconnect() method, since it shouldn't be called now.
- Remove div_abort() method. It was never called directly, since protocol
  doesn't have listen queue. It was called only from div_disconnect(),
  which is removed now.

Reviewed by: rwatson, maxim
Approved by: julian (mentor)
MT5 after: 1 week
MT4 after: 1 month

19 years agoUse __FBSDID.
yar [Thu, 18 Nov 2004 13:46:29 +0000 (13:46 +0000)]
Use __FBSDID.

19 years agos/priveleged/privileged/
glebius [Thu, 18 Nov 2004 13:37:33 +0000 (13:37 +0000)]
s/priveleged/privileged/

19 years agoFix a minor comment typo: priveleged = privileged.
keramida [Thu, 18 Nov 2004 12:20:52 +0000 (12:20 +0000)]
Fix a minor comment typo: priveleged = privileged.

19 years agoFix buffer overflow. This is FreeBSD-SA-04:16.fetch.
cperciva [Thu, 18 Nov 2004 12:01:30 +0000 (12:01 +0000)]
Fix buffer overflow.  This is FreeBSD-SA-04:16.fetch.

Approved by: des

19 years agoDon't say, "file: permission denied," if the operation
yar [Thu, 18 Nov 2004 11:50:01 +0000 (11:50 +0000)]
Don't say, "file: permission denied," if the operation
is disabled entirely.

19 years agoUse uniform punctuation, capitalization, and language style
yar [Thu, 18 Nov 2004 11:45:13 +0000 (11:45 +0000)]
Use uniform punctuation, capitalization, and language style
in server messages wherever this doesn't contradict to a particular
message format.

19 years agoApply __printflike() to the appendf() prototype so the compiler
yar [Thu, 18 Nov 2004 11:32:35 +0000 (11:32 +0000)]
Apply __printflike() to the appendf() prototype so the compiler
can detect format errors.

19 years agoFix perror_reply() vs. reply() usage.
yar [Thu, 18 Nov 2004 11:27:31 +0000 (11:27 +0000)]
Fix perror_reply() vs. reply() usage.

19 years ago'\n' needs not to appear in reply() strings.
yar [Thu, 18 Nov 2004 11:07:00 +0000 (11:07 +0000)]
'\n' needs not to appear in reply() strings.

19 years agoLog pathname arguments to ftp commands as the user specified them;
yar [Thu, 18 Nov 2004 10:02:28 +0000 (10:02 +0000)]
Log pathname arguments to ftp commands as the user specified them;
add the working directory pathname to the log message if any of
such arguments isn't absolute.  This has advantage over the old
way of logging that an admin can see what users are actually trying
to do, and where.  The old code was also not too robust when it
came to a chrooted session and an absolute pathname.

Pointed out by: Nick Leuta
MFC after:      2 weeks

19 years agoUse S_ISDIR() macro instead of a hand-rolled test.
yar [Thu, 18 Nov 2004 09:26:58 +0000 (09:26 +0000)]
Use S_ISDIR() macro instead of a hand-rolled test.

19 years ago - Eliminate the acquisition and release of the bqlock in bremfree() by
jeff [Thu, 18 Nov 2004 08:44:09 +0000 (08:44 +0000)]
 - Eliminate the acquisition and release of the bqlock in bremfree() by
   setting the B_REMFREE flag in the buf.  This is done to prevent lock order
   reversals with code that must call bremfree() with a local lock held.
   This also reduces overhead by removing two lock operations per buf for
   fsync() and similar.
 - Check for the B_REMFREE flag in brelse() and bqrelse() after the bqlock
   has been acquired so that we may remove ourself from the free-list.
 - Provide a bremfreef() function to immediately remove a buf from a
   free-list for use only by NFS.  This is done because the nfsclient code
   overloads the b_freelist queue for its own async. io queue.
 - Simplify the numfreebuffers accounting by removing a switch statement
   that executed the same code in every possible case.
 - getnewbuf() can encounter locked bufs on free-lists once Giant is removed.
   Remove a panic associated with this condition and delay asserts that
   inspect the buf until after it is locked.

Reviewed by: phk
Sponsored by: Isilon Systems, Inc.

19 years agoMake more sense out of vop_stdcreatevobject()
phk [Thu, 18 Nov 2004 07:55:05 +0000 (07:55 +0000)]
Make more sense out of vop_stdcreatevobject()

19 years agoCorrect another location where inquiry length as calculated from the
gibbs [Thu, 18 Nov 2004 07:24:29 +0000 (07:24 +0000)]
Correct another location where inquiry length as calculated from the
inquiry additional length field is off by one.

MFC: 3 days

19 years agothis should have been deleted a while ago
imp [Thu, 18 Nov 2004 04:30:17 +0000 (04:30 +0000)]
this should have been deleted a while ago

19 years agoCorrect an off-by-one in the calculation of the inquiry buffer length
gibbs [Thu, 18 Nov 2004 04:14:32 +0000 (04:14 +0000)]
Correct an off-by-one in the calculation of the inquiry buffer length
to request from devices during the "long inquiry" portion of our probe.
This same bug was fixed in the 4.x stream a few years ago, but the fix
was never propogated to -current.

This fix is slightly different than in -stable:

 o Use offsetof() instead of a hard coded constant so as the make
   the code more self-explainatory.

 o Round odd long inquiry lengths up so as to avoid tickling ignore
   wide residue bugs in broken parallel SCSI devices running with a
   wide transfer negotiation.

MFC: 3 days

19 years agoCross-reference signal(9).
jkoshy [Thu, 18 Nov 2004 02:06:17 +0000 (02:06 +0000)]
Cross-reference signal(9).

19 years agogetcwd() won't leave a error string in the buffer, unlike getwd().
yar [Thu, 18 Nov 2004 00:14:09 +0000 (00:14 +0000)]
getcwd() won't leave a error string in the buffer, unlike getwd().

19 years agoMove KERNPHYSADDR and KERNVIRTADDR from std.i80321 to std.iq31244.
cognet [Thu, 18 Nov 2004 00:05:09 +0000 (00:05 +0000)]
Move KERNPHYSADDR and KERNVIRTADDR from std.i80321 to std.iq31244.
Use the correct KERNPHYSADDR.

19 years agoFixed transition from SHARED=symlinks to SHARED=copies.
ru [Wed, 17 Nov 2004 23:48:17 +0000 (23:48 +0000)]
Fixed transition from SHARED=symlinks to SHARED=copies.

19 years agoCorrect URL of the programming manual.
brueffer [Wed, 17 Nov 2004 23:36:11 +0000 (23:36 +0000)]
Correct URL of the programming manual.

19 years agoAdd missing /* DEVICE_POLLING */
brueffer [Wed, 17 Nov 2004 23:33:56 +0000 (23:33 +0000)]
Add missing /* DEVICE_POLLING */

19 years agoFix host route addition for more than one address to a loopback interface
mlaier [Wed, 17 Nov 2004 23:14:03 +0000 (23:14 +0000)]
Fix host route addition for more than one address to a loopback interface
after allowing more than one address with the same prefix.

Reported by: Vladimir Grebenschikov <vova NO fbsd SPAM ru>
Submitted by: ru (also NetBSD rev. 1.83)
Pointyhat to: mlaier

19 years agoUpdate release notes to note that FreeBSD/i386 requires at least an 80486
jhb [Wed, 17 Nov 2004 23:03:17 +0000 (23:03 +0000)]
Update release notes to note that FreeBSD/i386 requires at least an 80486
and also requires a floating point unit.

Reviewed by: hrs, scottl

19 years ago- list D-Link DFE-528TX as supported
brueffer [Wed, 17 Nov 2004 22:51:46 +0000 (22:51 +0000)]
- list D-Link DFE-528TX as supported
- correct whitespace in a few entries
- remove some doubt from the D-Link DFE-538TX entry
- bump .Dd

MFC after: 3 days

19 years agoUse POSIX functions instead of legacy ones:
yar [Wed, 17 Nov 2004 22:43:46 +0000 (22:43 +0000)]
Use POSIX functions instead of legacy ones:
getwd() -> getcwd()
wait3() -> waitpid()

19 years agos/3COM/3Com/
brueffer [Wed, 17 Nov 2004 21:39:40 +0000 (21:39 +0000)]
s/3COM/3Com/

MFC after: 3 days

19 years agoList the LevelOne GNC-0105T (8169S) as supported and bump .Dd
brueffer [Wed, 17 Nov 2004 21:39:06 +0000 (21:39 +0000)]
List the LevelOne GNC-0105T (8169S) as supported and bump .Dd

MFC after: 3 days

19 years agoonly clear the IFF_OACTIVE flag when we have a chance of being able to
jmg [Wed, 17 Nov 2004 21:35:22 +0000 (21:35 +0000)]
only clear the IFF_OACTIVE flag when we have a chance of being able to
queue a packet to the hardware... instead of when the hardware queue is
empty..

don't initalize cur_tx now that it doesn't need to be...

Pointed out by: bde

19 years agoRemove the whole uart_cpu_identify() stuff again. Now that it's no longer
marius [Wed, 17 Nov 2004 20:01:43 +0000 (20:01 +0000)]
Remove the whole uart_cpu_identify() stuff again. Now that it's no longer
used on sparc64 they are only stubs on all architectures and it doesn't
look like if we would need it in the near future again.

Ok'ed by: marcel

19 years agosync comment on IFF_OACTIVE with reality.. IFF_OACTIVE is set when the
jmg [Wed, 17 Nov 2004 18:32:44 +0000 (18:32 +0000)]
sync comment on IFF_OACTIVE with reality..  IFF_OACTIVE is set when the
hardware cannot take anymore packets, and so will supress the calling of
the device's if_start method...

Submitted by: bde

19 years agoHook up mac_sysv_{msg,sem,shm}.c to the build when compiling with MAC
rwatson [Wed, 17 Nov 2004 17:33:59 +0000 (17:33 +0000)]
Hook up mac_sysv_{msg,sem,shm}.c to the build when compiling with MAC
support.

19 years agoAdd a front-end for the `rtc' device which is a MC146818 compatible
marius [Wed, 17 Nov 2004 16:41:42 +0000 (16:41 +0000)]
Add a front-end for the `rtc' device which is a MC146818 compatible
clock found on the ISA bus (some USIIe, USIIi and USIIIi models) and
EBus (USIII models) instead of a MK48Txx clock.

Testet by: Matthew T. Lager" <freebsd@trinetworks.com> on Sun Fire V100,
Xavier Beaudouin <kiwi@oav.net> on Netra X1 (initial version)

19 years agoAdd a driver back end for MC146818 and compatible clocks based on the
marius [Wed, 17 Nov 2004 16:37:25 +0000 (16:37 +0000)]
Add a driver back end for MC146818 and compatible clocks based on the
respective NetBSD driver for use with the genclock interface.
It's first use will be on sparc64 but it was also tested on alpha with
a preliminary patch to switch alpha to use the genclock code together
with this driver instead of the respective code in alpha/alpha/clock.c
and the rather MD mcclock(4). Using it on i386 and amd64 won't be that
hard but some changes/extensions to improve the genclock code in general
should be done first, e.g. add locking and make it easier to access the
NVRAM usually coupled with RTCs.

19 years agoForce commit to denote that this file was repo-copied from
marius [Wed, 17 Nov 2004 14:46:13 +0000 (14:46 +0000)]
Force commit to denote that this file was repo-copied from
dev/dec/mc146818reg.h

Repo-copy done by: peter

19 years agoo sparc64/isa/isa.c:
marius [Wed, 17 Nov 2004 14:44:10 +0000 (14:44 +0000)]
o sparc64/isa/isa.c:
  - The claim in the commit log of rev. 1.11 of dev/uart/uart_cpu_sparc64.c
    etc. that UARTs are the only relevant ISA devices on sparc64 turned out
    to be false. While there are sparc64 models where UARTs are the only
    devices on the ISA bus there are in fact also low-cost models where all
    devices traditionally found on the EBus are hooked up to the ISA bus.
    There are also models that use a mix between EBus and ISA devices with
    things like an AT keyboard controller and other rather interesting
    devices that we might want to support in the futute hook up to the ISA
    bus.
    In order to not need to add sparc64 specific device_identify methods to
    all of the respective ISA drivers and also not add OFW specific code to
    the common ISA code make the sparc64 ISA bus code fake up PnP devices so
    most ISA drivers probe their devices without further changes.
    Unfortunately Sun doesn't adhere to the ISA bindings defined in IEEE
    1275-1994 for the properties of most of the ISA devices which would
    allow to obtain the vendor and logical IDs from their properties. So we
    we just use a simple table which maps the name properties to PnP IDs.
    This could be done in a more sophisticated way but I courrently don't
    see the need for this. [1]
  - Add the children with fully mapped and specified resources (in the OFW
    sense) similar to what is done in the EBus code for the IRQ resources
    of the children as adjusting the resources and the resource list entries
    respectively in isa_alloc_resource() as done perviously causes trouble
    with drivers which use rman_get_start(), pass-through or allocate and
    release resources multiple times, etc.
    Adjusting the resources might be better off in a bus_activate_resource
    method but the common ISA code currently doesn't allow for an
    isa_activate_resource(). [2]
    With this change:
    - ppbus(4) and lpt(4) attach and work (modulo ECP mode, which requires
      real ISADMA code but it currently only consists of stubs on sparc64).
    - atkbdc(4) and atkbdc(4) attach, no further testing done.
    - fdc(4) itself attaches but causes a hang while attaching fd0 also
      when is DMA disabled, further work in fdc(4) is required here as e.g.
      fd0 uses the address of fd1 on sparc64 (not sure if sparc64 supports
      more than one floppy drive at all).
    All of these drivers previously caused panics in the sparc64 ISA code.
  - Minor changes, e.g. use __FBSDID, remove a dupe word in a comment and
    declare one global variable which isn't used outside of isa.c static.
o dev/uart/uart_cpu_sparc64.c and modules/uart/Makefile:
  - Remove the code for registering the UARTs on the ISA bus from the
    sparc64 uart_cpu_identify() again and rely on probing them via PnP.

Original idea by: tmm [1]
No objections by: tmm [1], [2]

19 years agoDon't bother exiting storming mode once a second to see if it has gone
jhb [Wed, 17 Nov 2004 14:39:41 +0000 (14:39 +0000)]
Don't bother exiting storming mode once a second to see if it has gone
away, instead only exit storming mode when an interrupt stops firing long
enough for the ithread to exit the loop and go back to sleep.

Tested by: macrus (cruder version)

19 years agoImplement MAC entry points relating to System V IPC, calling into the
rwatson [Wed, 17 Nov 2004 13:14:24 +0000 (13:14 +0000)]
Implement MAC entry points relating to System V IPC, calling into the
MAC policies to perform object life cycle operations and access
control checks.

Submitted by: Dandekar Hrishikesh <rishi_dandekar at sbcglobal dot net>
Obtained from: TrustedBSD Project
Sponsored by: DARPA, SPAWAR, McAfee Research

19 years agoIn the kernel-only portionss of System V IPC objects (messages,
rwatson [Wed, 17 Nov 2004 13:12:06 +0000 (13:12 +0000)]
In the kernel-only portionss of System V IPC objects (messages,
message queues, shared memory segments, and semaphores), add a struct
label pointer, which will hold the MAC labels for the objects.  As a
result of recent work to separate kernel and user space ABIs, this
should not break the ABI for applications using System V IPC, but will
require a rebuild of the ipcs monitoring tool.

Submitted by: Dandekar Hrishikesh <rishi_dandekar at sbcglobal dot net>
Obtained from: TrustedBSD Project
Sponsored by: DARPA, SPAWAR, McAfee Research

19 years agoDefine new MAC framework and policy entry points for System V IPC
rwatson [Wed, 17 Nov 2004 13:10:16 +0000 (13:10 +0000)]
Define new MAC framework and policy entry points for System V IPC
objects and operations:

- System V IPC message, message queue, semaphore, and shared memory
  segment init, destroy, cleanup, create operations.

- System V IPC message, message queue, seamphore, and shared memory
  segment access control entry points, including rights to attach,
  destroy, and manipulate these IPC objects.

Submitted by: Dandekar Hrishikesh <rishi_dandekar at sbcglobal dot net>
Obtained from: TrustedBSD Project
Sponsored by: DARPA, SPAWAR, McAfee Research

19 years agoShorten the description of the mk48txx driver to just say that it supports
marius [Wed, 17 Nov 2004 12:57:24 +0000 (12:57 +0000)]
Shorten the description of the mk48txx driver to just say that it supports
MK48Txx clocks, there are now to many models supported to list them all here.

19 years agoo Sync with the NetBSD mk48txx driver (the result simplyfies some changes
marius [Wed, 17 Nov 2004 12:54:12 +0000 (12:54 +0000)]
o Sync with the NetBSD mk48txx driver (the result simplyfies some changes
  I have in mind for the genclock interface):
  - Recognize the MK48T18 as well (differs from the MK48T08 only in
    packaging options and voltages).
  - Allow MD code to provide functions for reading/writing NVRAM/RTC
    locations.
    If passed NULL, the old behaviour using bus_space_{read,write}_1() is
    used. Otherwise, all access to the chip goes via the MD functions.
    This is necessary for mvmeppc boards where the mk48txx NVRAM/RTC is
    not directly addressable.
  - Cleanup MI mk48txx(4) todclock driver:
    - Prepare mk48txxvar.h and leave only register definitions in
      mk48txxreg.h.
    - Define struct mk48txx_softc as usual devices and allocate necessary
      members in it.
    - Change mk48txx_attach() to only take a device_t.
o While converting the sparc64 eeprom driver to the above changes:
  - Remove some dead code and stale comments.
  - Use the NVRAM size provided by the mk48txx driver instead of hardcoding
    it as suggested by a comment.
  - Add a comment about why it doesn't make much sense to read the hostid
    directly from the NVRAM except for displaying it when attaching.
  - Don't print the hostid if it reads all zero because it's stored
    elsewhere.

19 years agoCatch up with version 600000 of config(8), quotes around device names
marius [Wed, 17 Nov 2004 12:07:14 +0000 (12:07 +0000)]
Catch up with version 600000 of config(8), quotes around device names
containing digits are no longer required.

19 years agoKill more unneeded casts found.
yar [Wed, 17 Nov 2004 11:52:41 +0000 (11:52 +0000)]
Kill more unneeded casts found.

Noticed by: Nick Leuta <skynick -at- mail.sc.ru> (some of them)

19 years agoEliminate the define for POSIX and build with Posix behaviour.
harti [Wed, 17 Nov 2004 11:32:46 +0000 (11:32 +0000)]
Eliminate the define for POSIX and build with Posix behaviour.
Our make has been build with POSIX enabled from the first day
and the ifdef'ed out code served no purpose.

19 years agoDon't set the BUS_DMA_ALLOCNOW flag for the parent tag or the tags that are
scottl [Wed, 17 Nov 2004 11:27:30 +0000 (11:27 +0000)]
Don't set the BUS_DMA_ALLOCNOW flag for the parent tag or the tags that are
used for static memory allocations.

Discussed with: sos

19 years agoInitialize lcap and pwd to NULL. This allows a WARNS=6 clean build,
delphij [Wed, 17 Nov 2004 10:01:48 +0000 (10:01 +0000)]
Initialize lcap and pwd to NULL.  This allows a WARNS=6 clean build,
hence bump it to 6.

Note that the last commit message was not quite accurate.  While the
assumption exists in the code, it's not possible to have an
uninitialized p there because if lflag is set when username is NULL
then execution would be terminated earlier.

19 years agoThe code path in main() dealing with lflag assumes that p was
delphij [Wed, 17 Nov 2004 09:52:10 +0000 (09:52 +0000)]
The code path in main() dealing with lflag assumes that p was
initialized with NULL, while it is not.  So let's initialize
it.

19 years agoPush Giant down through ioctl.
phk [Wed, 17 Nov 2004 09:09:55 +0000 (09:09 +0000)]
Push Giant down through ioctl.

Don't grab Giant in the upper syscall/wrapper code

NET_LOCK_GIANT in the socket code (sockets/fifos).

mtx_lock(&Giant) in the vnode code.

mtx_lock(&Giant) in the opencrypto code.  (This may actually not be
needed, but better safe than sorry).

Devfs grabs Giant if the driver is marked as needing Giant.

19 years agoPush Giant down through select and poll.
phk [Wed, 17 Nov 2004 08:01:10 +0000 (08:01 +0000)]
Push Giant down through select and poll.

Don't grab Giant in the upper syscall/wrapper code

NET_LOCK_GIANT in the socket code (sockets/fifos).

mtx_lock(&Giant) in the vnode code.

Devfs grabs Giant if the driver is marked as needing Giant.

19 years agoSplit the FILEDESC_LOCK in two variants.
phk [Wed, 17 Nov 2004 07:39:58 +0000 (07:39 +0000)]
Split the FILEDESC_LOCK in two variants.

FILEDESC_LOCK_FAST will just grab the interlocking mutex and hold
it.  This should be used for simple modifications of a field.

FILEDESC_LOCK holds a (homegrown) sleepable lock which should be used
where sleeping is required.

The homegrown lock will probably be replaced with a generic type of lock
once we have found out how that should look.

Help and reviews by: rwatson

19 years agoMake vnode bypass for fifos (read, write, poll) mandatory.
phk [Wed, 17 Nov 2004 07:30:02 +0000 (07:30 +0000)]
Make vnode bypass for fifos (read, write, poll) mandatory.

19 years agoMake vnode bypass for devices mandatory.
phk [Wed, 17 Nov 2004 07:18:49 +0000 (07:18 +0000)]
Make vnode bypass for devices mandatory.

19 years agoMake interrupt coalescing work on big endian systems.
yongari [Wed, 17 Nov 2004 04:25:10 +0000 (04:25 +0000)]
Make interrupt coalescing work on big endian systems.
Also change struct ucode.length to be in number of elements (u_int32_t)
to help endian handling.

MFC after: 2 weeks