]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoDisable serialize_methods and enable _OSI support by default. The former
Nate Lawson [Tue, 30 Mar 2004 07:35:18 +0000 (07:35 +0000)]
Disable serialize_methods and enable _OSI support by default.  The former
is necessary because some IBMs use recursive methods (pointed out by
Robert Moore from Intel).  The latter was a typo on my part.  It was disabled
by default when it should have been enabled.

20 years agoAdd basic manual pages for isideogram(), isphonogram(), isrune()
Tim J. Robbins [Tue, 30 Mar 2004 07:23:54 +0000 (07:23 +0000)]
Add basic manual pages for isideogram(), isphonogram(), isrune()
and isspecial().

20 years agoTrim cross-references.
Tim J. Robbins [Tue, 30 Mar 2004 07:19:35 +0000 (07:19 +0000)]
Trim cross-references.

20 years agoDocument the isnumber() and ishexnumber() functions, and explain how they
Tim J. Robbins [Tue, 30 Mar 2004 07:02:04 +0000 (07:02 +0000)]
Document the isnumber() and ishexnumber() functions, and explain how they
differ (at least in theory) from isdigit() and isxdigit().

20 years agoFixed a style bug in previous commit (misformatted comment). Fixed
Bruce Evans [Tue, 30 Mar 2004 07:01:56 +0000 (07:01 +0000)]
Fixed a style bug in previous commit (misformatted comment).  Fixed
some nearby bugs (rotted and missing comments).  Use similar wording
for describing broken options.

20 years agoDown to WARNS=2 for a while.
Jun Kuriyama [Tue, 30 Mar 2004 06:16:41 +0000 (06:16 +0000)]
Down to WARNS=2 for a while.

20 years agoCatch all cases where bread() returns an error and a valid *bp, and release
Scott Long [Tue, 30 Mar 2004 05:01:48 +0000 (05:01 +0000)]
Catch all cases where bread() returns an error and a valid *bp, and release
the *bp.

Obtained from: DragonFlyBSD

20 years agoSwitch to using strtoul() for parsing a potential UID or GID, which gets
Garance A Drosehn [Tue, 30 Mar 2004 04:20:33 +0000 (04:20 +0000)]
Switch to using strtoul() for parsing a potential UID or GID, which gets
this to correctly handle UID's and GID's larger than 2147483647.

Noticed by: bde
MFC after: 1 week

20 years agoInitial check-in of the device driver for 3ware's 9000 series
Vinod Kashyap [Tue, 30 Mar 2004 03:46:00 +0000 (03:46 +0000)]
Initial check-in of the device driver for 3ware's 9000 series
PATA/SATA RAID controllers.  This driver is a SIM under CAM, and
so, behaves like a driver for a SCSI controller.

20 years agoPrefer NULL to 0 when testing and assigning pointer values.
Robert Watson [Tue, 30 Mar 2004 02:16:25 +0000 (02:16 +0000)]
Prefer NULL to 0 when testing and assigning pointer values.

20 years agoBruce would really like the prototype for fmt() to be split across lines
Garance A Drosehn [Tue, 30 Mar 2004 02:02:40 +0000 (02:02 +0000)]
Bruce would really like the prototype for fmt() to be split across lines
this way (although I still think it "looks weird"...).

Requested by: bde
MFC after: 1 week

20 years agoMinor style fixes, mostly adding indent-protection on some comment-blocks.
Garance A Drosehn [Tue, 30 Mar 2004 01:59:22 +0000 (01:59 +0000)]
Minor style fixes, mostly adding indent-protection on some comment-blocks.

Noticed by: bde
MFC after: 1 week

20 years agoReplace pscomp() with a cleaner version, mostly written by bde (*).
Garance A Drosehn [Tue, 30 Mar 2004 01:45:23 +0000 (01:45 +0000)]
Replace pscomp() with a cleaner version, mostly written by bde (*).
This corrects a problem of lost-precision for `-r' (sort-by-CPU).  Also,
for sort-by-CPU and sort-by-memory, any processes which have the same
value CPU or MEMORY are now sorted by TTY and then (if needed) by pid.

(* - I just added the NODEV checks, after doing some testing of my own)

Submitted by: bde
MFC after: 1 week

20 years agoMake libdisk WARNS=4 clean.
Jun Kuriyama [Tue, 30 Mar 2004 01:39:00 +0000 (01:39 +0000)]
Make libdisk WARNS=4 clean.

Glanced by: jhb

20 years agoShorten some XXXKSE commentry
Peter Wemm [Mon, 29 Mar 2004 22:46:54 +0000 (22:46 +0000)]
Shorten some XXXKSE commentry

20 years agoKill some XXXKSE's. vnlru/syncer are single threaded.
Peter Wemm [Mon, 29 Mar 2004 22:45:33 +0000 (22:45 +0000)]
Kill some XXXKSE's.  vnlru/syncer are single threaded.

20 years agoClean up the stub fake vnode locking implemenations. The main reason this
Peter Wemm [Mon, 29 Mar 2004 22:41:21 +0000 (22:41 +0000)]
Clean up the stub fake vnode locking implemenations.  The main reason this
stuff was here (NFS) was fixed by Alfred in November.  The only remaining
consumer of the stub functions was umapfs, which is horribly horribly
broken.  It has missed out on about the last 5 years worth of maintenence
that was done on nullfs (from which umapfs is derived).  It needs major
work to bring it up to date with the vnode locking protocol.  umapfs really
needs to find a caretaker to bring it into the 21st century.

Functions GC'ed:
vop_noislocked, vop_nolock, vop_nounlock, vop_sharedlock.

20 years agoAdd per-softc locking to if_tun:
Robert Watson [Mon, 29 Mar 2004 22:16:39 +0000 (22:16 +0000)]
Add per-softc locking to if_tun:

- Add tun_mtx to tun_softc.  Annotate what is (and isn't) locked by it.
- Lock down tun_flags, tun_pid.
- In the output path, cache the value of tun_flags so it's consistent
  when processing a particular packet rather than re-reading the field.
- In general, use unlocked reads for debugging.
- Annotate a couple of places where additional unlocked reads may be
  possible.
- Annotate that tun_pid is used as a bug in tunopen().

if_tun is now largely MPSAFE, although questions remain about some of
the cdevsw fields and how they are synchronized.

20 years agoRemove duplicate MLINK.
Tim J. Robbins [Mon, 29 Mar 2004 21:46:52 +0000 (21:46 +0000)]
Remove duplicate MLINK.

20 years agoBring these files closer to style(9) conformance by comparing a
Robert Drehmel [Mon, 29 Mar 2004 21:15:41 +0000 (21:15 +0000)]
Bring these files closer to style(9) conformance by comparing a
dereferenced character pointer to '\0' instead of using the ! operator.

20 years agoEliminate support for FreeBSD 3.x and earlier now that we're in the
Warner Losh [Mon, 29 Mar 2004 21:10:05 +0000 (21:10 +0000)]
Eliminate support for FreeBSD 3.x and earlier now that we're in the
glide path for the 5.x branch.

20 years agoFix the case where $ntpdate_hosts was not specified and /etc/ntp.conf
Dag-Erling Smørgrav [Mon, 29 Mar 2004 20:00:54 +0000 (20:00 +0000)]
Fix the case where $ntpdate_hosts was not specified and /etc/ntp.conf
does not exist.

Submitted by: ru

20 years agoMdoc Police:
Hiten Pandya [Mon, 29 Mar 2004 19:42:21 +0000 (19:42 +0000)]
Mdoc Police:

* Use .Fa instead of .Pa for function arguments.

* Utilize the .Po/.Pc commands for parenthesis
  so the format is not ruined.

Inspired by: pjd's last commit to this file

20 years agoFinish tidying up a couple of leftovers from the KSTACK_PAGES stuff. Some
Peter Wemm [Mon, 29 Mar 2004 19:38:05 +0000 (19:38 +0000)]
Finish tidying up a couple of leftovers from the KSTACK_PAGES stuff.  Some
files still #included the opt_ file.  powerpc hadn't been updated yet.

20 years agoNote that we can pass NULL as 4th argument for copystr() and copyinstr()
Pawel Jakub Dawidek [Mon, 29 Mar 2004 19:23:45 +0000 (19:23 +0000)]
Note that we can pass NULL as 4th argument for copystr() and copyinstr()
functions.

With little mdoc(7) help from: hmp

20 years agoLock down if_tun global variables using a new mutex, tunmtx. As with
Robert Watson [Mon, 29 Mar 2004 18:42:51 +0000 (18:42 +0000)]
Lock down if_tun global variables using a new mutex, tunmtx.  As with
other pseudo-interfaces, break out tear-down of a softc into a
separate tun_destroy() function, and invoke that from the module
unloader.  Hold tunmtx across manipulations of the global softc list.

20 years agoIf we change obj_rtld.path after initialising __progname, make sure we
Doug Rabson [Mon, 29 Mar 2004 18:37:37 +0000 (18:37 +0000)]
If we change obj_rtld.path after initialising __progname, make sure we
change __progname to point at the new storage otherwise it ends up
pointing at freed memory which leads to confusing garbled error messages.

20 years agoUse a common return path for filt_soread() and filt_sowrite() to
Robert Watson [Mon, 29 Mar 2004 18:06:15 +0000 (18:06 +0000)]
Use a common return path for filt_soread() and filt_sowrite() to
simplify the impact of locking on these functions.

Submitted by: sam
Sponsored by: FreeBSD Foundation

20 years agoIn sofree(), moving caching of 'head' from 'so->so_head' to later in
Robert Watson [Mon, 29 Mar 2004 17:57:43 +0000 (17:57 +0000)]
In sofree(), moving caching of 'head' from 'so->so_head' to later in
the function once it has been determined to be non-NULL to simplify
locking on an earlier return.

20 years agoIf debug.mpsafenet, initialize UNIX domain socket timeouts as MPSAFE;
Robert Watson [Mon, 29 Mar 2004 17:00:05 +0000 (17:00 +0000)]
If debug.mpsafenet, initialize UNIX domain socket timeouts as MPSAFE;
otherwise, assert Giant in the callouts.

20 years agoo Also check that the mutex type is not less than the minimum allowable value.
Mike Makonnen [Mon, 29 Mar 2004 13:57:55 +0000 (13:57 +0000)]
o Also check that the mutex type is not less than the minimum allowable value.
o Don't check attribute for NULL. It's the callers responsibility.

20 years agoMake the minimum implementation of pthread_kill conform to the
Mike Makonnen [Mon, 29 Mar 2004 13:56:04 +0000 (13:56 +0000)]
Make the minimum implementation of pthread_kill conform to the
functionality spelled out in SUSv3.
o Signal of 0 means do everything except send the signal
o Check that the signal is not invalid
o Check that the target thread is not dead/invalid

20 years agoo Don't explicitly check the thread for NULL. That is the caller's
Mike Makonnen [Mon, 29 Mar 2004 13:53:43 +0000 (13:53 +0000)]
o Don't explicitly check the thread for NULL. That is the caller's
  responsibility.
o If a thread is not joinable, the correct return value is EINVAL.

20 years agoo If a thread is marked as detached AND on the dead threads list
Mike Makonnen [Mon, 29 Mar 2004 13:51:51 +0000 (13:51 +0000)]
o If a thread is marked as detached AND on the dead threads list
  the correct return value is ESRCH.
o Don't check the attribute for NULL. It's the caller's responsibility.
o Make the bitwise comparison explicit.

20 years agoAdd a comment about time stamper.
Yoshihiro Takahashi [Mon, 29 Mar 2004 12:51:46 +0000 (12:51 +0000)]
Add a comment about time stamper.

Submitted by: chi@bd.mbn.or.jp (Chiharu Shibata)

20 years agoIf a condition variable is statically initialized don't return
Mike Makonnen [Mon, 29 Mar 2004 11:24:02 +0000 (11:24 +0000)]
If a condition variable is statically initialized don't return
an error. Return successfully without doing anything.

20 years agoThis commit was generated by cvs2svn to compensate for changes in r127563,
Dag-Erling Smørgrav [Mon, 29 Mar 2004 11:18:25 +0000 (11:18 +0000)]
This commit was generated by cvs2svn to compensate for changes in r127563,
which included commits to RCS files with non-trunk default branches.

20 years agoSynch with NetBSD: avoid "unused parameter" warning.
Dag-Erling Smørgrav [Mon, 29 Mar 2004 11:18:25 +0000 (11:18 +0000)]
Synch with NetBSD: avoid "unused parameter" warning.

20 years agoAdd directories under /usr/share/nls and /usr/local/share/nls
Tim J. Robbins [Mon, 29 Mar 2004 11:07:35 +0000 (11:07 +0000)]
Add directories under /usr/share/nls and /usr/local/share/nls
for the new UTF-8 locales.

Reminded by: ache

20 years agoThe thread suspend function now returns ETIMEDOUT, not EAGAIN.
Mike Makonnen [Mon, 29 Mar 2004 09:35:07 +0000 (09:35 +0000)]
The thread suspend function now returns ETIMEDOUT, not EAGAIN.

20 years agoMake fingerd(8) WARNS2 clean.
Pawel Jakub Dawidek [Mon, 29 Mar 2004 09:29:51 +0000 (09:29 +0000)]
Make fingerd(8) WARNS2 clean.

20 years agoOops; remove ko_KR.eucKR/LC_MONETARY -> ko_KR.UTF-8/LC_MONETARY symlink.
Tim J. Robbins [Mon, 29 Mar 2004 09:03:14 +0000 (09:03 +0000)]
Oops; remove ko_KR.eucKR/LC_MONETARY -> ko_KR.UTF-8/LC_MONETARY symlink.

20 years agoo Remove more references to SIGTHR
Mike Makonnen [Mon, 29 Mar 2004 05:45:01 +0000 (05:45 +0000)]
o Remove more references to SIGTHR
o Remove clock resolution information left over from libc_r

20 years agoOops. Remove some ';'s in #defines added by a previous update.
Garance A Drosehn [Mon, 29 Mar 2004 03:03:28 +0000 (03:03 +0000)]
Oops.  Remove some ';'s in #defines added by a previous update.

Noticed by: bde

20 years agoConditionally acquire Giant when entering the sockets layer via the
Robert Watson [Mon, 29 Mar 2004 02:21:56 +0000 (02:21 +0000)]
Conditionally acquire Giant when entering the sockets layer via the
socket-specific system calls based on debug.mpsafenet, rather than
acquiring Giant unconditionally.

20 years agoMFi386: Use the BSD madvise() syscall implementation for Linux binary
Bruce M Simpson [Mon, 29 Mar 2004 02:19:43 +0000 (02:19 +0000)]
MFi386: Use the BSD madvise() syscall implementation for Linux binary
emulation, instead of treating it as an unimplemented syscall.

Requested by: marcel

20 years agoAdd missing cprd_flags member to partial resource structure in
Bill Paul [Mon, 29 Mar 2004 02:15:29 +0000 (02:15 +0000)]
Add missing cprd_flags member to partial resource structure in
resource_var.h.

In kern_ndis.c:ndis_convert_res(), fill in the cprd_flags and
cprd_sharedisp fields as best we can.

In if_ndis.c:ndis_setmulti(), don't bother updating the multicast
filter if our multicast address list is empty.

Add some missing updates to ndis_var.h and ntoskrnl_var.h that I
forgot to check in when I added the KeDpc stuff.

20 years agoMFi386: correctly calculate the top-of-stack when a kthread is created
Ken Smith [Mon, 29 Mar 2004 02:01:42 +0000 (02:01 +0000)]
MFi386: correctly calculate the top-of-stack when a kthread is created
with a larger kernel stack. Remove inclusion of opt_kstack_pages.h now
that it's unused.

Note: If anyone's toes got stepped on by me doing this let me know
privately please.

Approved by: rwatson (mentor)

20 years agoConditionally acquire Giant when entering the socket layer via file
Robert Watson [Mon, 29 Mar 2004 01:55:32 +0000 (01:55 +0000)]
Conditionally acquire Giant when entering the socket layer via file
descriptor operations based on debug.mpsafenet, rather than acquiring
Giant unconditionally.

20 years agoAdd a new LC_MONETARY data file for the ko_KR.UTF-8 locale instead of using
Tim J. Robbins [Mon, 29 Mar 2004 01:53:31 +0000 (01:53 +0000)]
Add a new LC_MONETARY data file for the ko_KR.UTF-8 locale instead of using
the ko_KR.eucKR version.

20 years agoWhen validating that the length sum in recvit(), we fail to release
Robert Watson [Mon, 29 Mar 2004 01:37:06 +0000 (01:37 +0000)]
When validating that the length sum in recvit(), we fail to release
Giant on an error.  Add a Giant acquisition.

Reviewed by: sam, bms

20 years agoAdd a new LC_MONETARY data file for the ja_JP.UTF-8 locale instead of using
Tim J. Robbins [Mon, 29 Mar 2004 01:29:13 +0000 (01:29 +0000)]
Add a new LC_MONETARY data file for the ja_JP.UTF-8 locale instead of using
the ja_JP.eucJP version. Although the latter looks like plain ASCII,
it is actually JIS Roman, so it is not correct to treat it as UTF-8.

20 years agoHave this source explicitly include <sys/proc.h>, since it references
Garance A Drosehn [Mon, 29 Mar 2004 01:27:13 +0000 (01:27 +0000)]
Have this source explicitly include <sys/proc.h>, since it references
values such as P_CONTROLT and PS_INMEM.  But this still won't define
PID_MAX for us, since that is hidden inside of '#ifdef _KERNEL'.

Noticed by: bde

20 years agoMFi386: correctly calculate the top-of-stack when a kthread is created
Ken Smith [Mon, 29 Mar 2004 01:24:13 +0000 (01:24 +0000)]
MFi386: correctly calculate the top-of-stack when a kthread is created
with a larger kernel stack. Remove inclusion of opt_kstack_pages.h now
that it's unused.

Reviewed by: marcel
Approved by: rwatson (mentor)

20 years agoSince "kp" is a pointer, I should be comparing against NULL not 0.
Garance A Drosehn [Mon, 29 Mar 2004 01:18:13 +0000 (01:18 +0000)]
Since "kp" is a pointer, I should be comparing against NULL not 0.

Noticed by: bde

20 years agoRemove "-frename-registers" option for sparc64 kernel builds. That
Ken Smith [Mon, 29 Mar 2004 01:15:39 +0000 (01:15 +0000)]
Remove "-frename-registers" option for sparc64 kernel builds.  That
was not present in what I originally tested when checking to see if
the kernel built/ran with the -O2 change.  Recent instability in
sparc64 kernel was tracked to this.  A reproducible kernel stack
traceback followed by hard hang during the call to msleep() at the
point the kernel waits 15 seconds for the SCSI bus to settle crept in
to recent kernel builds and it seems to go away with this patch.

Noticed by: kris
Approved by: rwatson (mentor)

20 years agoVarious style improvements, mostly in comments and indentation.
Garance A Drosehn [Mon, 29 Mar 2004 01:15:27 +0000 (01:15 +0000)]
Various style improvements, mostly in comments and indentation.

Suggested by: bde (well, for most of them)

20 years agoModify BPF descriptor assertions to assert Giant when a BPF descriptor
Robert Watson [Mon, 29 Mar 2004 00:33:39 +0000 (00:33 +0000)]
Modify BPF descriptor assertions to assert Giant when a BPF descriptor
lock is asserted and running non-MPSAFE.

20 years agoConditionally assert Giant in fputsock() based on the value of
Robert Watson [Mon, 29 Mar 2004 00:33:02 +0000 (00:33 +0000)]
Conditionally assert Giant in fputsock() based on the value of
debug.mpsafenet.

20 years agoIn the routines I've been working on, sort the variable declartions
Garance A Drosehn [Mon, 29 Mar 2004 00:25:09 +0000 (00:25 +0000)]
In the routines I've been working on, sort the variable declartions
so that non-pointers are listed after pointer-type variables.

Noticed by: bde

20 years agoSplit two 'fmt' strings so they're easier to read on 80-char windows.
Garance A Drosehn [Mon, 29 Mar 2004 00:16:19 +0000 (00:16 +0000)]
Split two 'fmt' strings so they're easier to read on 80-char windows.

20 years agoSort the declarations of global variables.
Garance A Drosehn [Mon, 29 Mar 2004 00:12:03 +0000 (00:12 +0000)]
Sort the declarations of global variables.

20 years agoSort the routine prototypes.
Garance A Drosehn [Sun, 28 Mar 2004 23:44:29 +0000 (23:44 +0000)]
Sort the routine prototypes.

20 years agoInvert the logic of NET_LOCK_GIANT(), and remove the one reference to it.
Robert Watson [Sun, 28 Mar 2004 23:12:19 +0000 (23:12 +0000)]
Invert the logic of NET_LOCK_GIANT(), and remove the one reference to it.
Previously, Giant would be grabbed at entry to the IP local delivery code
when debug.mpsafenet was set to true, as that implied Giant wouldn't be
grabbed in the driver path.  Now, we will use this primitive to
conditionally grab Giant in the event the entire network stack isn't
running MPSAFE (debug.mpsafenet == 0).

20 years agoRegen from ibcs2_wait() becoming MPSAFE.
Robert Watson [Sun, 28 Mar 2004 22:51:40 +0000 (22:51 +0000)]
Regen from ibcs2_wait() becoming MPSAFE.

Submitted by: Dan Nelson <dnelson@allantgroup.com>

20 years agoibcs2_wait() now MPSAFE.
Robert Watson [Sun, 28 Mar 2004 22:51:01 +0000 (22:51 +0000)]
ibcs2_wait() now MPSAFE.

Submitted by: Dan Nelson <dnelson@allantgroup.com>

20 years agoUse the BSD madvise() syscall implementation for Linux binary emulation,
Bruce M Simpson [Sun, 28 Mar 2004 21:43:27 +0000 (21:43 +0000)]
Use the BSD madvise() syscall implementation for Linux binary emulation,
instead of treating it as an unimplemented syscall. This appears to make
StarOffice 7.0 Linux binaries work according to submitter; also tested
with nvidia driver by submitter.

Submitted by: Matthias Schuendehuette

20 years agoCommit two fixes from OpenBSD's stable branch:
Max Laier [Sun, 28 Mar 2004 20:04:12 +0000 (20:04 +0000)]
Commit two fixes from OpenBSD's stable branch:
- Fix binat for incoming connections when a netblock (not just a single
  address) is used for source in the binat rule. closes PR 3535, reported by
  Karl O.Pinc. ok henning@, cedric@

- Fix a problem related to empty anchor rulesets, which could cause a kernel
  panic.

Approved by: bms(mentor)

20 years agoTell fdisk(8) about our geometry.
Poul-Henning Kamp [Sun, 28 Mar 2004 19:39:08 +0000 (19:39 +0000)]
Tell fdisk(8) about our geometry.

Strictly speaking this is unnecessary, but it allows nanobsd to work
on systems from before the -x and -y arguments to mdconfig(8) worked
for vnode backing.

Submitted by: "Oivind H. Danielsen" <oivind.danielsen@kopek.net>

20 years agoAdd a simple tool that prints out a register offset table to map
Marcel Moolenaar [Sun, 28 Mar 2004 18:01:15 +0000 (18:01 +0000)]
Add a simple tool that prints out a register offset table to map
from the gdb(1) register number to offsets within struct reg and
struct fpreg. The tool is useful only on selected platforms. On
ia64 the registers are all over the place to simplify handling of
them in various situations, but which makes creating or maintaining
such an offset table error prone to do by hand.
Since remote kernel debugging operates on the same register numbers,
it would be a natural choice to use an identical offset table in the
kernel. However, since the kernel does not operate on struct reg nor
struct fpreg in the remote gdb(1) case, such would not make sense.
Whether we want to use this tool to create offsets for use in the
kernel or duplicate the knowledge of which register number maps to
what register is something that will become clear soon.

Note: in order to build cross debuggers, one cannot use target
headers and/or target definitions. That's why offsets need to be
hardcoded in the first place. Unpleasant, but necessary.

20 years agoRemove unused argument.
Pawel Jakub Dawidek [Sun, 28 Mar 2004 15:48:00 +0000 (15:48 +0000)]
Remove unused argument.

20 years agoRemove the garbage collector thread. All resources are freed
Mike Makonnen [Sun, 28 Mar 2004 14:05:28 +0000 (14:05 +0000)]
Remove the garbage collector thread. All resources are freed
in-line. If the exiting thread cannot release a resource, then
the next thread to exit will release it.

20 years agoAdd PC98 supports.
Yoshihiro Takahashi [Sun, 28 Mar 2004 13:42:27 +0000 (13:42 +0000)]
Add PC98 supports.

Submitted by: Watanabe Kazuhiro <CQG00620@nifty.ne.jp> (mostly)

20 years ago- Fix PC98 supports after importing auto selection. [1]
Yoshihiro Takahashi [Sun, 28 Mar 2004 13:40:23 +0000 (13:40 +0000)]
- Fix PC98 supports after importing auto selection. [1]
- Fix 1.44MB floppy drive probe sequence. [2]

Submitted by: Watanabe Kazuhiro <CQG00620@nifty.ne.jp> [1]
chi@bd.mbn.or.jp (Chiharu Shibata) [2]

20 years agoMFi386: revision 1.1136.
Yoshihiro Takahashi [Sun, 28 Mar 2004 12:06:29 +0000 (12:06 +0000)]
MFi386: revision 1.1136.

20 years agoMFi386: revision 1.586.
Yoshihiro Takahashi [Sun, 28 Mar 2004 12:03:39 +0000 (12:03 +0000)]
MFi386: revision 1.586.

20 years agoFix a warning: compare u_long ps_strings to 0 instead of NULL.
Jens Schweikhardt [Sun, 28 Mar 2004 11:50:54 +0000 (11:50 +0000)]
Fix a warning: compare u_long ps_strings to 0 instead of NULL.

20 years agoReset callout if in nfs_timeout and rpcclnt_timeout functions. Timer
Alexander Kabaev [Sun, 28 Mar 2004 05:55:27 +0000 (05:55 +0000)]
Reset callout if in nfs_timeout and rpcclnt_timeout functions. Timer
are supposed to continue firing as long as there is work to do, not
stop after the first invocation.

This is damage control after a patch that has been committed prematurely.

Tested by: kris

20 years agoFixed s style bug in previous commit (tab lossage). Fixed some nearby
Bruce Evans [Sun, 28 Mar 2004 04:17:36 +0000 (04:17 +0000)]
Fixed s style bug in previous commit (tab lossage).  Fixed some nearby
style bugs (more tab lossage, unclear description of TDF_USTATCLOCK,
and English usage errors).

20 years agoIf <x> is a process id that does not exist, then just print the header
Garance A Drosehn [Sun, 28 Mar 2004 03:00:56 +0000 (03:00 +0000)]
If <x> is a process id that does not exist, then just print the header
(if any) and exit, thus matching the behavior on -stable and other OS's.
My earlier attempt to fix this (v1.65) only seemed to work because of a
lucky random value in nentries (which was not being initialized back
when I tested that earlier patch).

20 years agoFix `-o rtprio' so it prints the correct value.
Garance A Drosehn [Sun, 28 Mar 2004 02:13:31 +0000 (02:13 +0000)]
Fix `-o rtprio' so it prints the correct value.

PR: bin/59417
Submitted by: Jan Willem Knopper
This fix by: bde (in the audit-trail of the PR)

20 years agoAdd a cross-reference to pkill(1), now that pkill is hooked up to the build.
Garance A Drosehn [Sat, 27 Mar 2004 23:45:33 +0000 (23:45 +0000)]
Add a cross-reference to pkill(1), now that pkill is hooked up to the build.

20 years agoThe compatibility specification property is named "compatible", not
Thomas Moestl [Sat, 27 Mar 2004 22:39:47 +0000 (22:39 +0000)]
The compatibility specification property is named "compatible", not
"compat".

Spotted by: Marius Strobl <marius@alchemy.franken.de>

20 years agoIf a non-existent user is given as part of `-U userlist', treat it as
Garance A Drosehn [Sat, 27 Mar 2004 22:14:42 +0000 (22:14 +0000)]
If a non-existent user is given as part of `-U userlist', treat it as
a fatal error instead of a minor warning.  It is possible that a few
users are used to the previous behavior, but I'm claiming it was a bug.

20 years agoExplicitly wrap two long-ish linesi of code, to make them easier to read.
Garance A Drosehn [Sat, 27 Mar 2004 21:59:54 +0000 (21:59 +0000)]
Explicitly wrap two long-ish linesi of code, to make them easier to read.

20 years agoChange the #if-ish logic which is used to add the `-f' option when `ps'
Garance A Drosehn [Sat, 27 Mar 2004 21:56:49 +0000 (21:56 +0000)]
Change the #if-ish logic which is used to add the `-f' option when `ps'
is compiled with LAZY_PS, so that there is only one PS_ARGS string to
modify when changing the option-list.  Also get `-f' to show up in the
usage() statement when compiled with LAZY_PS.

20 years agoMove the 'f' case so it shows up in the right place, alphabetically.
Garance A Drosehn [Sat, 27 Mar 2004 21:40:04 +0000 (21:40 +0000)]
Move the 'f' case so it shows up in the right place, alphabetically.

20 years agoReduce 'td' argument to 'cred' (struct ucred) argument in those functions:
Pawel Jakub Dawidek [Sat, 27 Mar 2004 21:05:46 +0000 (21:05 +0000)]
Reduce 'td' argument to 'cred' (struct ucred) argument in those functions:
- in_pcbbind(),
- in_pcbbind_setup(),
- in_pcbconnect(),
- in_pcbconnect_setup(),
- in6_pcbbind(),
- in6_pcbconnect(),
- in6_pcbsetport().
"It should simplify/clarify things a great deal." --rwatson

Requested by: rwatson
Reviewed by: rwatson, ume

20 years agoRemove unused argument.
Pawel Jakub Dawidek [Sat, 27 Mar 2004 20:41:32 +0000 (20:41 +0000)]
Remove unused argument.

Reviewed by: ume

20 years agoApparently, some atheros drivers want rand(), so implement it (in terms
Bill Paul [Sat, 27 Mar 2004 20:38:43 +0000 (20:38 +0000)]
Apparently, some atheros drivers want rand(), so implement it (in terms
of random()).

Requested by: juli
Bribe offered: tacos

20 years agoRemove unused prototype.
Pawel Jakub Dawidek [Sat, 27 Mar 2004 20:38:02 +0000 (20:38 +0000)]
Remove unused prototype.

Reviewed by: ume

20 years agoRevise the direct or optimized case to use uiomove_fromphys() by the reader
Alan Cox [Sat, 27 Mar 2004 19:50:23 +0000 (19:50 +0000)]
Revise the direct or optimized case to use uiomove_fromphys() by the reader
instead of ephemeral mappings using pmap_qenter() by the writer.  The
writer is still, however, responsible for wiring the pages, just not
mapping them.  Consequently, the allocation of KVA for the direct case is
unnecessary.  Remove it and the sysctls limiting it, i.e.,
kern.ipc.maxpipekvawired and kern.ipc.amountpipekvawired.  The number
of temporarily wired pages is still, however, limited by
kern.ipc.maxpipekva.

Note: On platforms lacking a direct virtual-to-physical mapping,
uiomove_fromphys() uses sf_bufs to cache ephemeral mappings.  Thus,
the number of available sf_bufs can influence the performance of pipes
on platforms such i386.  Surprisingly, I saw the greatest gain from this
change on such a machine: lmbench's pipe bandwidth result increased from
~1050MB/s to ~1850MB/s on my 2.4GHz, 400MHz FSB P4 Xeon.

20 years agoUpdate the date on the man page, since this the previous change added
Garance A Drosehn [Sat, 27 Mar 2004 18:26:40 +0000 (18:26 +0000)]
Update the date on the man page, since this the previous change added
a few new options.

20 years agoSupport more POSIX/SUSv3 options:
Garance A Drosehn [Sat, 27 Mar 2004 18:22:17 +0000 (18:22 +0000)]
Support more POSIX/SUSv3 options:

- Change `-p' to allow a list of process IDs, and `-t' to allow a list
  of terminal names, instead of only a single value for each.
- Add the `-A' option of SUSv3, which is exactly the same as `-ax'.
- Add the `-G gidlist' (group id).
- Allow any of these "selector options" to be specified multiple times,
  and have `ps' keep adding to a given list -- instead of replacing the
  previously-specified values.
- Fix interactions between selector-options, so that: "If any are
  specified, ... ps shall select the processes represented by the
  inclusive OR of all the selection-criteria options." (from SUSv3)
- Add a `-X' option, which is the reverse of the `-x' option.

- various minor improvements in parsing and error handling.

This does not get us to match POSIX/SUSv3, but it gets us closer.  The
`-g pgidlist', `-R ruserlist' and `-s sidlist' options mentioned in
freebsd-standards are still under debate, so they skipped for now.
It should be true that this introduces no user-visible incompatible
changes, except to support "new stuff" that was not supported before.

20 years agoChange the type of the various CPU masks to cpumask_t. Note that as
Marcel Moolenaar [Sat, 27 Mar 2004 18:21:24 +0000 (18:21 +0000)]
Change the type of the various CPU masks to cpumask_t. Note that as
long as there are still explicit uses of int, whether in types or
in function names (such as atomic_set_int() in sched_ule.c), we can
not change cpumask_t to be anything other than u_int. See also the
commit log for sys/sys/types.h, revision 1.84.

20 years agoMFi386: correctly calculate the top-of-stack when a kthread is created
Marcel Moolenaar [Sat, 27 Mar 2004 17:44:25 +0000 (17:44 +0000)]
MFi386: correctly calculate the top-of-stack when a kthread is created
with a larger kernel stack. Remove inclusion of opt_kstack_pages.h now
that it's unused.

20 years agoAdd ACPI PnP string. This affects devinfo(8) output with -v option.
Takanori Watanabe [Sat, 27 Mar 2004 16:26:00 +0000 (16:26 +0000)]
Add ACPI PnP string. This affects devinfo(8) output with -v option.

20 years agoFix typo in the device id for the new cards.
Scott Long [Sat, 27 Mar 2004 15:56:34 +0000 (15:56 +0000)]
Fix typo in the device id for the new cards.

20 years agoo Since we're not using signals for thread synchronization anymore,
Mike Makonnen [Sat, 27 Mar 2004 15:05:28 +0000 (15:05 +0000)]
o Since we're not using signals for thread synchronization anymore,
  sigprocmask no longer needs to be wrapped.
o raise(3) is applied to the calling thread in a threaded program.
o In the sigaction wrapper reference the correct structure.
o Don't treat SIGTHR especially anymore (infact it won't exist in
  a little while).

20 years agoStop using signals for synchronizing threads. The performance penalty
Mike Makonnen [Sat, 27 Mar 2004 14:39:21 +0000 (14:39 +0000)]
Stop using signals for synchronizing threads. The performance penalty
was too much.