]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoNew release note: Non-root-owned binaries in standard system paths
Bruce A. Mah [Mon, 10 Sep 2001 18:51:13 +0000 (18:51 +0000)]
New release note:  Non-root-owned binaries in standard system paths
now have the schg flag set.

23 years agoRemove a field width specifier that's not doing anything more than
Mike Heffner [Mon, 10 Sep 2001 18:46:07 +0000 (18:46 +0000)]
Remove a field width specifier that's not doing anything more than
what using snprintf() achieves. It was also being used incorrectly.

23 years agoAdd a cast to silence a warning (error because of WARNS=2) on Alpha.
Mike Barcroft [Mon, 10 Sep 2001 17:23:57 +0000 (17:23 +0000)]
Add a cast to silence a warning (error because of WARNS=2) on Alpha.

Discussed with: des

23 years agoNew release notes: Use net/cvsup, rather than net/cvsup-bin and
Bruce A. Mah [Mon, 10 Sep 2001 16:54:50 +0000 (16:54 +0000)]
New release notes:  Use net/cvsup, rather than net/cvsup-bin and
net/cvsupd-bin.  Also note CVSup update to fix S1G bug.

I'm breaking a self-imposed rule of not mentioning ports-collection
updates in release notes.  I'm only doing this because CVSup is a
fairly essential part of the FreeBSD Project infrastructure.  Save
for exceptions such as this, the release notes will continue to
document changes to the base system only.

Suggested by: mike

23 years agoI am no longer the FreeBSD GCC maintainer.
David E. O'Brien [Mon, 10 Sep 2001 16:05:39 +0000 (16:05 +0000)]
I am no longer the FreeBSD GCC maintainer.

23 years agoRename kernel entry point from locorestart to __start.
Doug Rabson [Mon, 10 Sep 2001 15:34:00 +0000 (15:34 +0000)]
Rename kernel entry point from locorestart to __start.

23 years agoDo not overrun entry array when printing output tables.
Ruslan Ermilov [Mon, 10 Sep 2001 15:09:12 +0000 (15:09 +0000)]
Do not overrun entry array when printing output tables.
Cleanup storage allocation for entries.

Obtained from: NetBSD

23 years agoBack out the last change - it was made to the wrong Makefile.
Doug Rabson [Mon, 10 Sep 2001 14:47:34 +0000 (14:47 +0000)]
Back out the last change - it was made to the wrong Makefile.

23 years ago* Make a start on a realistic definition for bootinfo.
Doug Rabson [Mon, 10 Sep 2001 13:40:00 +0000 (13:40 +0000)]
* Make a start on a realistic definition for bootinfo.
* Switch to proc0's stack and backing store before calling ia64_init
  so that we don't rely on the loader's stack at all.
* Change kernel entry point name from locorestart to __start.

23 years agoFix a warning on alpha (real problem) and make pstat -t work as a bonus.
Peter Wemm [Mon, 10 Sep 2001 12:05:47 +0000 (12:05 +0000)]
Fix a warning on alpha (real problem) and make pstat -t work as a bonus.
'struct tty' was out of sync in user and kernel due to dev_t/udev_t
mixups.  This takes advantage of the fact that dev_t changes type in
userland, so it isn't too pretty.

23 years agoDon't call log() without a format string.
Kris Kennaway [Mon, 10 Sep 2001 12:00:47 +0000 (12:00 +0000)]
Don't call log() without a format string.

MFC after: 2 weeks

23 years agoFixed SIGFPE (divide by zero) if column's width exceeds display width (-w).
Ruslan Ermilov [Mon, 10 Sep 2001 11:56:07 +0000 (11:56 +0000)]
Fixed SIGFPE (divide by zero) if column's width exceeds display width (-w).

Obtained from: OpenBSD

23 years agoAdd DAO mode (Disk At Once).
Søren Schmidt [Mon, 10 Sep 2001 11:44:32 +0000 (11:44 +0000)]
Add DAO mode (Disk At Once).
See burncd.8 for how to use it.

23 years agoFixed the -z option handling:
Ruslan Ermilov [Mon, 10 Sep 2001 11:43:40 +0000 (11:43 +0000)]
Fixed the -z option handling:

-Wuninitialized if used without -t.

PR: bin/30467

Null pointer dereference if used with -t.

Maximum column width computation was wrong.

23 years agoAdd kernel side support for DAO burning.
Søren Schmidt [Mon, 10 Sep 2001 11:43:20 +0000 (11:43 +0000)]
Add kernel side support for DAO burning.

23 years agoAdd new ioctl's and rearrange a bit for DAO mode support.
Søren Schmidt [Mon, 10 Sep 2001 11:42:27 +0000 (11:42 +0000)]
Add new ioctl's and rearrange a bit for DAO mode support.

23 years agoMake the `nsops' variable in `semop' unsigned. This prevents an
Dima Dorfman [Mon, 10 Sep 2001 11:36:08 +0000 (11:36 +0000)]
Make the `nsops' variable in `semop' unsigned.  This prevents an
overflow if uap->nsops (which is already unsigned) is over INT_MAX;
consequently, the bounds check below becomes valid.  Previously, if a
value over INT_MAX was passed in uap->nsops, the bounds check wouldn't
catch it, and the value would be used to compute copyin()'s third
argument.

Obtained from: NetBSD

23 years agoFix some signed/unsigned integer confusion, and add bounds checking of
Kris Kennaway [Mon, 10 Sep 2001 11:28:07 +0000 (11:28 +0000)]
Fix some signed/unsigned integer confusion, and add bounds checking of
arguments to some functions.

Obtained from: NetBSD
Reviewed by: peter
MFC after: 2 weeks

23 years agoFix a warning. Dont convert away from const to just re-add it in the
Peter Wemm [Mon, 10 Sep 2001 10:48:19 +0000 (10:48 +0000)]
Fix a warning.  Dont convert away from const to just re-add it in the
wrapped function.

23 years agoFix a warning in bsearch(). Like index() etc, it de-qualifies its
Peter Wemm [Mon, 10 Sep 2001 10:33:46 +0000 (10:33 +0000)]
Fix a warning in bsearch().  Like index() etc, it de-qualifies its
arguments.  Caveat Emptor.  However, the only two consumer of it (cam)
deals with it correctly and preserves the constness.

23 years agomdoc(7) police: markup nits.
Ruslan Ermilov [Mon, 10 Sep 2001 09:08:47 +0000 (09:08 +0000)]
mdoc(7) police: markup nits.

PR: docs/30473
Submitted by: Jonathan Hanna <jhanna@home.com>

23 years agoTypo s/AirLAncer/AirLancer/g
Warner Losh [Mon, 10 Sep 2001 08:14:00 +0000 (08:14 +0000)]
Typo s/AirLAncer/AirLancer/g

23 years agoclarify PCMLM56 status
Warner Losh [Mon, 10 Sep 2001 08:07:46 +0000 (08:07 +0000)]
clarify PCMLM56 status

23 years agoThe EXCard-10-PCMCIA entry was slightly bogus.
Warner Losh [Mon, 10 Sep 2001 08:05:29 +0000 (08:05 +0000)]
The EXCard-10-PCMCIA entry was slightly bogus.

23 years agoFix a warning. l_name is managed by us and is malloc/free'ed.
Peter Wemm [Mon, 10 Sep 2001 07:53:04 +0000 (07:53 +0000)]
Fix a warning.  l_name is managed by us and is malloc/free'ed.
It is the userland declaration of l_name that is inconvenient for us.

23 years agoAdd on UPAGES to ki_rssize since it is there as result of the process
Peter Wemm [Mon, 10 Sep 2001 07:29:32 +0000 (07:29 +0000)]
Add on UPAGES to ki_rssize since it is there as result of the process
and can be swapped out with the process.

23 years agoUPAGES as a constant is gone in KSE. We are going to have to trust
Peter Wemm [Mon, 10 Sep 2001 07:19:26 +0000 (07:19 +0000)]
UPAGES as a constant is gone in KSE.  We are going to have to trust
and/or fix ki_rssize.

23 years agoImplement support for MAXMEM option and hw.physmem environment variable
Doug Rabson [Mon, 10 Sep 2001 07:03:59 +0000 (07:03 +0000)]
Implement support for MAXMEM option and hw.physmem environment variable
which can be used to artificially reduce the memory size of a machine
for debugging (or other) purposes.

23 years agoFix LINT breakage caused by previous commit. The linux_rt_sendsig
Marcel Moolenaar [Mon, 10 Sep 2001 07:00:17 +0000 (07:00 +0000)]
Fix LINT breakage caused by previous commit. The linux_rt_sendsig
and linux_sendsig functions guarded their debugging output with
ldebug(sigreturn). This has been mistaken for a cut-n-paste bug,
and was replaced by ldebug(rt_sendsig) and ldebug(sendsig) resp.
Since the sendsig functions are not syscalls, this brokei any
build that defines DEBUG.

The fix maps both functions to the unused syscall 0 so that they
can be enabled/disabled independently from sigreturn, but not
independently from each other.

23 years agoThis will have to be revised, but allow putting 'makeoptions GCC3=true'
Peter Wemm [Mon, 10 Sep 2001 06:23:33 +0000 (06:23 +0000)]
This will have to be revised, but allow putting 'makeoptions GCC3=true'
in a kernel config file.  This should minimize the tearing-out-hair process
while updating the kernel for gcc-3 compliance.

23 years agoRip some well duplicated code out of cpu_wait() and cpu_exit() and move
Peter Wemm [Mon, 10 Sep 2001 04:28:58 +0000 (04:28 +0000)]
Rip some well duplicated code out of cpu_wait() and cpu_exit() and move
it to the MI area.  KSE touched cpu_wait() which had the same change
replicated five ways for each platform.  Now it can just do it once.
The only MD parts seemed to be dealing with fpu state cleanup and things
like vm86 cleanup on x86.  The rest was identical.

XXX: ia64 and powerpc did not have cpu_throw(), so I've put a functional
stub in place.

Reviewed by: jake, tmm, dillon

23 years agogcc-3 has objections about the bluetrap6 and bluetrap13 inline asm
Peter Wemm [Mon, 10 Sep 2001 04:22:20 +0000 (04:22 +0000)]
gcc-3 has objections about the bluetrap6 and bluetrap13 inline asm
functions.  Apparently multi-line string asm arguments are deprecated.

23 years agoFix some malformed macro concatenation that gcc-3 has objections about.
Peter Wemm [Mon, 10 Sep 2001 04:20:34 +0000 (04:20 +0000)]
Fix some malformed macro concatenation that gcc-3 has objections about.

23 years agoAdd cs driver for IBM EtherJet card. Doesn't completely work yet, but
Warner Losh [Mon, 10 Sep 2001 04:08:28 +0000 (04:08 +0000)]
Add cs driver for IBM EtherJet card.  Doesn't completely work yet, but
it won't work without it.

23 years ago2Mbps card from Teletronics. Looks like a generic OEM card for the
Warner Losh [Mon, 10 Sep 2001 03:30:47 +0000 (03:30 +0000)]
2Mbps card from Teletronics.  Looks like a generic OEM card for the
awi driver, but the MAC address isn't recognized, so maybe this is
wrong.

23 years agoELSA Air Lancer wireless card. Appears to be a Lucent OEM.
Warner Losh [Mon, 10 Sep 2001 03:18:21 +0000 (03:18 +0000)]
ELSA Air Lancer wireless card.  Appears to be a Lucent OEM.

Submitted by: Eric Masson <e-masson@kisoft-services.com>

23 years agoLots of style(9) related white space fixes. Mostly missing spaces
Brooks Davis [Mon, 10 Sep 2001 02:36:18 +0000 (02:36 +0000)]
Lots of style(9) related white space fixes.  Mostly missing spaces
after if, for, while, switch, '{', and '}' plus a bunch of bogus
whitespace at the end of lines.

23 years agoAdd support for monitor mode. This means that after enabling the
Brooks Davis [Mon, 10 Sep 2001 02:05:10 +0000 (02:05 +0000)]
Add support for monitor mode.  This means that after enabling the
correct mode via ancontrol, you can use bpf to sniff raw 802.11 frames.
Who want's to port AirSnort. ;-)

Submitted by: Doug Ambrisko <ambrisko@ambrisko.com> (author)
David Wolfskill <david@catwhisker.org> (port to current)

23 years agoRemove/comment tokens after #endif (#endif NETATALK)
Peter Wemm [Mon, 10 Sep 2001 01:33:03 +0000 (01:33 +0000)]
Remove/comment tokens after #endif  (#endif NETATALK)

23 years agoFix a minor buglet/typo here that gcc3 complains about.
Peter Wemm [Mon, 10 Sep 2001 01:27:23 +0000 (01:27 +0000)]
Fix a minor buglet/typo here that gcc3 complains about.

23 years ago#endif /* comment */
Peter Wemm [Mon, 10 Sep 2001 01:23:39 +0000 (01:23 +0000)]
#endif /* comment */

23 years agoStop gcc's fixincludes making an exception for this file
Peter Wemm [Mon, 10 Sep 2001 01:15:48 +0000 (01:15 +0000)]
Stop gcc's fixincludes making an exception for this file

23 years agoMihira-san says that this works for him. Linksys EtherFast 10/100 +
Warner Losh [Sun, 9 Sep 2001 23:58:09 +0000 (23:58 +0000)]
Mihira-san says that this works for him.  Linksys EtherFast 10/100 +
56k modem with the fat connector.

Submitted by: sanpei@sanpei.org (MIHIRA yoshiro)
Ignored since Jan 2001 by: imp

23 years agoThe "dirpref" directory layout preference improvements make use of
Ian Dowse [Sun, 9 Sep 2001 23:48:28 +0000 (23:48 +0000)]
The "dirpref" directory layout preference improvements make use of
an array "fs_contigdirs[]" to avoid too many directories getting
created in each cylinder group. The memory required for this and
two other arrays (fs_csp[] and fs_maxcluster[]) is allocated with
a single malloc() call, and divided up afterwards.  However, the
'space' pointer is not advanced correctly, so fs_contigdirs and
fs_maxcluster end up pointing to the same address.

Add the missing code to advance the 'space' pointer, and remove
an unnecessary update of the pointer that follows.

This is likely to fix the "ffs_clusteralloc: map mismatch" panics
that have been reported recently.

Submitted by: Luke Mewburn <lukem@wasabisystems.com>

23 years agoAddtron AWP-100 wireless card.
Warner Losh [Sun, 9 Sep 2001 23:42:26 +0000 (23:42 +0000)]
Addtron AWP-100 wireless card.

Submitted by: Alfred Perlstein <bright@mu.org>

23 years agoEZCard, not EXCard
Warner Losh [Sun, 9 Sep 2001 23:40:28 +0000 (23:40 +0000)]
EZCard, not EXCard

23 years agoSpeculatively commit fix for Surecom EP-427TX PCMCIA adapter. This
Warner Losh [Sun, 9 Sep 2001 23:28:09 +0000 (23:28 +0000)]
Speculatively commit fix for Surecom EP-427TX PCMCIA adapter.  This
appears to be another OEM version of the Netgear FA411.  This is a
guess, since the original didn't include the flags, but this is too
similar to my netgear card...

Submitted by: neal@nelsonnet.org

23 years agoSMC EZ Card 10 PCMCIA
Warner Losh [Sun, 9 Sep 2001 23:13:59 +0000 (23:13 +0000)]
SMC EZ Card 10 PCMCIA

Submitted by: Goncharov D <gda@sani.ru>

23 years agoAdd NE-2000 compatible card sold by addron.
Warner Losh [Sun, 9 Sep 2001 23:10:38 +0000 (23:10 +0000)]
Add NE-2000 compatible card sold by addron.

Submitted by: johs@copyleft.no

23 years agoAdd more standards which we conform to.
Jeroen Ruigrok van der Werven [Sun, 9 Sep 2001 21:09:53 +0000 (21:09 +0000)]
Add more standards which we conform to.
Note our implementation is not thread nor async-cancel safe.
Explicitely note atof() does not check nor report errors.
Note that strtod() should be used instead.

23 years agoAdd NetGEAR FA411 card. This appears to be based on the AX88190 chipset
Warner Losh [Sun, 9 Sep 2001 20:31:26 +0000 (20:31 +0000)]
Add NetGEAR FA411 card.  This appears to be based on the AX88190 chipset
and works with those flags.

23 years agoDetail thread-safe and async-cancel-safe status.
Jeroen Ruigrok van der Werven [Sun, 9 Sep 2001 18:52:00 +0000 (18:52 +0000)]
Detail thread-safe and async-cancel-safe status.
Also add C99 conformity status plus clarification that C99 leaves the
flushing of unwritten data, closure of open streams, and removal of
temporary files to the implementation.

23 years agoWhen booting verbose, print the config space for this device. This
Warner Losh [Sun, 9 Sep 2001 17:42:58 +0000 (17:42 +0000)]
When booting verbose, print the config space for this device.  This
will help debugging problem systems as it will reduce the number of
commands the user needs to type and send me the output of.

23 years agoIf flags were specified for the device, print them in the probe
Warner Losh [Sun, 9 Sep 2001 17:28:02 +0000 (17:28 +0000)]
If flags were specified for the device, print them in the probe
message.  Some scsi devices have rather elaberate flags, and it is
hard to know which ones were specified in pccard.conf when debug
information only includes the dmesg.

23 years agoregenerate this file too (forgotten after last usbdevs update)
Bill Paul [Sun, 9 Sep 2001 17:00:32 +0000 (17:00 +0000)]
regenerate this file too (forgotten after last usbdevs update)

23 years agoFix some problems with the reference/modified tracking which I introduced
Doug Rabson [Sun, 9 Sep 2001 16:36:59 +0000 (16:36 +0000)]
Fix some problems with the reference/modified tracking which I introduced
with the last change to the way the pmap_emulate_reference() works. This
should fix a number of memory corruption problems and also should stop the
mtimes of executables changing all the time.

23 years agoAutomatically load the "snp" module if it isn't already present.
Dima Dorfman [Sun, 9 Sep 2001 14:57:48 +0000 (14:57 +0000)]
Automatically load the "snp" module if it isn't already present.

23 years agoUse protected variables names in prototypes visible to the userland as
Dima Dorfman [Sun, 9 Sep 2001 14:57:17 +0000 (14:57 +0000)]
Use protected variables names in prototypes visible to the userland as
per style(9) and most other header files in sys/.

23 years agoupdate cs_CZ's MONETARY information to more used format
Alexey Zelkin [Sun, 9 Sep 2001 14:49:13 +0000 (14:49 +0000)]
update cs_CZ's MONETARY information to more used format

Discussed with: Cejka Rudolf <cejkar@dcse.fee.vutbr.cz> (submitter)

23 years agoAdd a missing newline to an error message.
Dima Dorfman [Sun, 9 Sep 2001 14:48:52 +0000 (14:48 +0000)]
Add a missing newline to an error message.

PR: 30436
Submitted by: Joerg Lehners <Lehners@Informatik.Uni-Oldenburg.DE>

23 years agoUse CFLAGS instead of COPTS.
Dima Dorfman [Sun, 9 Sep 2001 14:30:36 +0000 (14:30 +0000)]
Use CFLAGS instead of COPTS.

23 years agoInclude ttymsg.h from ../../usr.bin/wall instead of rolling our own
Dima Dorfman [Sun, 9 Sep 2001 14:30:11 +0000 (14:30 +0000)]
Include ttymsg.h from ../../usr.bin/wall instead of rolling our own
prototype.

23 years ago- Silence warnings: apply `const' generously, mark some variables
Dima Dorfman [Sun, 9 Sep 2001 14:25:02 +0000 (14:25 +0000)]
- Silence warnings: apply `const' generously, mark some variables
  __unused, and change local variables named `sin' (struct
  sockaddr_in) to `sin4'.  (`sin' conflicts with the definition of
  sin(3), which gcc assumes to be defined even if math.h isn't
  included (it's a builtin).  This is probably a bug in gcc.)

- Apply WARNS=1.  WARNS=2 was not used because this program assigns
  string literals to (struct iovec).iov_base for writing, and the only
  clean way to silence -Wwrite-strings in that case would be to
  strdup() and consequently free() those literals, which I considered
  too disruptive.

Reviewed by: bde (partially)

23 years ago- Move the prototype of ttymsg() into ttymsg.h. syslogd and talkd
Dima Dorfman [Sun, 9 Sep 2001 14:23:31 +0000 (14:23 +0000)]
- Move the prototype of ttymsg() into ttymsg.h.  syslogd and talkd
  also use this, and they shouldn't have to have their own prototypes.

- Silence warnings about constness and signedness in ttymsg().  This
  includes changing the return value to a `const char *', and changing
  the types of `left' and `wret' (both byte counts) to ssize_t.

Reviewed by: bde

23 years agoIron out an identity crisis. I'm Irish, not English.
Brian Somers [Sun, 9 Sep 2001 08:33:40 +0000 (08:33 +0000)]
Iron out an identity crisis.  I'm Irish, not English.

23 years agoRun the uustat command as the uucp user, not as root.
Kris Kennaway [Sun, 9 Sep 2001 05:53:01 +0000 (05:53 +0000)]
Run the uustat command as the uucp user, not as root.

23 years agoThe basis for the recent coredump commit had the wrong attribution.
Matthew Dillon [Sun, 9 Sep 2001 05:08:33 +0000 (05:08 +0000)]
The basis for the recent coredump commit had the wrong attribution.
The new attribution is below.

Submitted by: peter, ps

23 years agoMake sure that all non-root-owned binaries in standard system
Matthew Dillon [Sun, 9 Sep 2001 04:54:10 +0000 (04:54 +0000)]
Make sure that all non-root-owned binaries in standard system
paths are chflaged 'schg' to prevent exploit vectors when run
by cron, by a root user, or by a user other then the one owning the
binary.  This applies to most of the uucp binaries, cu, tip, and
man (man was already installed properly).

MFC will occur when approved.

23 years agoAdd support for the Accton USB320-EC Ethernet adapter.
Doug White [Sun, 9 Sep 2001 03:08:30 +0000 (03:08 +0000)]
Add support for the Accton USB320-EC Ethernet adapter.

MFC after: 14

23 years agoRegen.
Doug White [Sun, 9 Sep 2001 03:07:32 +0000 (03:07 +0000)]
Regen.

23 years agoAdd entry for Accton USB320-EC Ethernet adapter.
Doug White [Sun, 9 Sep 2001 03:07:15 +0000 (03:07 +0000)]
Add entry for Accton USB320-EC Ethernet adapter.

MFC after: 14

23 years agoUse a better stripe size in the examples.
Greg Lehey [Sun, 9 Sep 2001 02:23:06 +0000 (02:23 +0000)]
Use a better stripe size in the examples.

Noted by: Sean Eric Fagan <sef@kithrup.com>

23 years agoRemove a bogus cast and lockdown users(1) with WARNS?=2
Mike Barcroft [Sun, 9 Sep 2001 00:40:04 +0000 (00:40 +0000)]
Remove a bogus cast and lockdown users(1) with WARNS?=2

Submitted by: David Hill <david@phobia.ms>
Reviewed by: -audit
MFC after: 1 week

23 years agoAdd Dennis Ritchie.
Greg Lehey [Sun, 9 Sep 2001 00:36:20 +0000 (00:36 +0000)]
Add Dennis Ritchie.

23 years agoCorrect intended fix to my foul-up.
Marcel Moolenaar [Sat, 8 Sep 2001 23:26:40 +0000 (23:26 +0000)]
Correct intended fix to my foul-up.

linux_uid16.c is in sys/compat/linux not in sys/i386/linux.

23 years agoStole unicode translation table from mount_msdos. Add kernel code
Semen Ustimenko [Sat, 8 Sep 2001 23:03:52 +0000 (23:03 +0000)]
Stole unicode translation table from mount_msdos. Add kernel code
to support this translation.

MFC after: 2 weeks

23 years agoCatch up with common usage:
Marcel Moolenaar [Sat, 8 Sep 2001 22:59:30 +0000 (22:59 +0000)]
Catch up with common usage:

o  Define ELFMAG and SELFMAG as an alternative to test the ELF magic.
   ELFMAG is the magic; SELFMAG is the size of the magic. Use with
   strncmp() or memcmp().
o  Define SHT_NUM as the number of predefined sections.

Note: style(9) deviation for consistency.

23 years agoFix opening particular file's attributes (as described in man page).
Semen Ustimenko [Sat, 8 Sep 2001 22:59:12 +0000 (22:59 +0000)]
Fix opening particular file's attributes (as described in man page).
This is useful for debug purposes.

MFC after: 2 weeks

23 years agoReference devvp on ntnode creation and dereference on removal. Previous
Semen Ustimenko [Sat, 8 Sep 2001 22:57:03 +0000 (22:57 +0000)]
Reference devvp on ntnode creation and dereference on removal. Previous
code  lead to page faults becouse i_devvp went zero after VOP_RECLAIM, but
ntnode was reused (not reclaimed).

MFC after: 2 weeks

23 years agoFix errors and warnings when compiling with NTFS_DEBUG > 1
Semen Ustimenko [Sat, 8 Sep 2001 22:53:27 +0000 (22:53 +0000)]
Fix errors and warnings when compiling with NTFS_DEBUG > 1

MFC after: 2 weeks

23 years agoAdd linux_sysctl.c, and linux_uid16.c to the x86 platforms.
Peter Wemm [Sat, 8 Sep 2001 22:35:28 +0000 (22:35 +0000)]
Add linux_sysctl.c, and linux_uid16.c to the x86 platforms.

23 years agoMissing part of dillon's coredump commit. cpu_coredump() was still
Peter Wemm [Sat, 8 Sep 2001 22:18:58 +0000 (22:18 +0000)]
Missing part of dillon's coredump commit. cpu_coredump() was still
passing IO_NODELOCKED to vn_rdwr(), this would cause operations on the
unlocked core vnode and softupdates nastiness if an a.out binary cored.

23 years agoThis brings in a Yahoo coredump patch from Paul, with additional mods by
Matthew Dillon [Sat, 8 Sep 2001 20:02:33 +0000 (20:02 +0000)]
This brings in a Yahoo coredump patch from Paul, with additional mods by
me (addition of vn_rdwr_inchunks).  The problem Yahoo is solving is that
if you have large process images core dumping, or you have a large number of
forked processes all core dumping at the same time, the original coredump code
would leave the vnode locked throughout.  This can cause the directory vnode
to get locked up, which can cause the parent directory vnode to get locked
up, and so on all the way to the root node, locking the entire machine up
for extremely long periods of time.

This patch solves the problem in two ways.  First it uses an advisory
non-blocking lock to abort multiple processes trying to core to the same
file.  Second (my contribution) it chunks up the writes and uses bwillwrite()
to avoid holding the vnode locked while blocking in the buffer cache.

Submitted by: ps
Reviewed by: dillon
MFC after: 2 weeks

23 years agoRound of cleanups and enhancements. These include (in random order):
Marcel Moolenaar [Sat, 8 Sep 2001 19:07:04 +0000 (19:07 +0000)]
Round of cleanups and enhancements. These include (in random order):

o  Introduce private types for use in linux syscalls for two reasons:
   1. establish type independence for ease in porting and,
   2. provide a visual queue as to which syscalls have proper
      prototypes to further cleanup the i386/alpha split.
   Linuxulator types are prefixed by 'l_'. void and char have not
   been "virtualized".

o  Provide dummy functions for all syscalls and remove dummy functions
   or implementations of truely obsolete syscalls.

o  Sanitize the shm*, sem* and msg* syscalls.

o  Make a first attempt to implement the linux_sysctl syscall. At this
   time it only returns one MIB (KERN_VERSION), but most importantly,
   it tells us when we need to add additional sysctls :-)

o  Bump the kenel version up to 2.4.2 (this is not the same as the
   KERN_VERSION MIB, BTW).

o  Implement new syscalls, of which most are specific to i386. Our
   syscall table is now up to date with Linux 2.4.2. Some highlights:
   -  Implement the 32-bit uid_t and gid_t bases syscalls.
   -  Implement a couple of 64-bit file size/offset bases syscalls.

o  Fix or improve numerous syscalls and prototypes.

o  Reduce style(9) violations while I'm here. Especially indentation
   inconsistencies within the same file are addressed. Re-indenting
   did not obfuscate actual changes to the extend that it could not
   be combined.

NOTE: I spend some time testing these changes and found that if there
      were regressions, they were not caused by these changes AFAICT.
      It was observed that installing a RH 7.1 runtime environment
      did make matters worse. Hangs and/or reboots have been observed
      with and without these changes, so when it failed to make life
      better in cases it doesn't look like it made it worse.

23 years agoo Introduce private types for use in linux syscalls for two reasons:
Marcel Moolenaar [Sat, 8 Sep 2001 18:49:31 +0000 (18:49 +0000)]
o  Introduce private types for use in linux syscalls for two reasons:
   1. establish type independence for ease in porting and,
   2. provide a visual queue as to which syscalls have proper
      prototypes to further cleanup the i386/alpha split.
   Linuxulator types are prefixed by 'l_'. void and char have not
   been "virtualized".

o  Provide dummy functions for all unimplemented syscalls, except for
   the osf1 syscalls. This can only be done if the osfulator
   implements at least all syscalls used by the linuxulator. Remove
   dummy functions for syscalls that are now truely unimplemented.

o  Set the syscall namespace as follows: Mark a syscall as OSF1 if
   the Linux kernel has prefixed the syscall with 'osf_' and has
   provided special implementations for it. Otherwise mark the
   syscall as LINUX by default. Some of the LINUX syscalls remain
   marked as BSD or POSIX.

o  Rename syscalls so they match the names used in the Linux kernel.
   Also, provide more accurate prototypes. This generally improves
   cross-referencing and reduces head-scratching.

o  Fix the (g|s)etresuid syscalls. They mapped to (g|s)etresgid.

o  Sanitize the the shm*, sem* and msg* syscalls. Their prototypes
   were dictated by the way these syscalls were used in the i386
   code. That has been fixed. NOTE: linux_semctl now passes it's
   'arg' parameter by value and not by reference.

o  Fix prototype of linux_utime. It takes a struct timeval, not a
   struct utimbuf.

o  Fix the linux_sysfs syscall. It's index is not 255, but 254.

o  Implement the following syscalls:
     linux_sysctl

o  Add the following new syscalls:
     (g|s)etresgid
     linux_pivot_root (dummy)
     linux_mincore (dummy)
     linux_pciconfig_iobase (dummy)
     linux_getdents64

23 years agoo Introduce private types for use in linux syscalls for two reasons:
Marcel Moolenaar [Sat, 8 Sep 2001 18:48:40 +0000 (18:48 +0000)]
o  Introduce private types for use in linux syscalls for two reasons:
   1. establish type independence for ease in porting and,
   2. provide a visual queue as to which syscalls have proper
      prototypes to further cleanup the i386/alpha split.
   Linuxulator types are prefixed by 'l_'. void and char have not
   been "virtualized".

o  Remove dummy functions for syscalls that are now truely
   unimplemented.

o  Rename syscalls so they match the names used in the Linux kernel.
   Also, provide more accurate prototypes. This generally improves
   cross-referencing and reduces head-scratching.

o  Provide seperate implementations for the 16-bit uid_t and gid_t
   based syscalls as Linux used to have. The new 32-bit uid_t and
   gid_t based syscalls now map to their FreeBSD equivalents.

o  Fix the linux_ipc syscall so that it doesn't force the shm*, sem*
   and msg* syscalls to have the same syscall. The prototypes for
   these syscalls now match the those used on Alpha. While here,
   add the same kludge for MSGRCV as is present in the Linux kernel.

o  Implement the following syscalls:
     linux_stat64, linux_lstat64 and linux_fstat64
     linux_sysctl

o  Added syscalls numbered 198 - 221. This include:
     - the 32-bit uid_t and gid_t bases syscalls
     - 64-bit file offset/size based syscalls

23 years agoDon't call tsleep from AcpiOsStall(), call DELAY() always instead.
Mitsuru IWASAKI [Sat, 8 Sep 2001 17:03:26 +0000 (17:03 +0000)]
Don't call tsleep from AcpiOsStall(), call DELAY() always instead.
Process switching during calling AcpiOsStall() caused fatal trap 12 at
sleeping/wakeup on some machines.

23 years agoIf the local file does not exist, or is a regular file, and we're not trying
Dag-Erling Smørgrav [Sat, 8 Sep 2001 15:17:15 +0000 (15:17 +0000)]
If the local file does not exist, or is a regular file, and we're not trying
to resume a transfer, download the requested document into a temporary file
which we later rename.  This avoids leaving half-completed files around in
case of a crash (it'll still leave a half-completed file, but with a  hope-
fully non-conflicting name), and should reduce the need for human inter-
vention on ports-building machines.

The temporary file name for "foo/bar" is constructed by invoking mkstemps()
with the pattern "foo/.fetch.XXXXXX.bar"

Requested by: obrien

23 years agoAdd missing entry to memory type name table and adjust field widths.
Doug Rabson [Sat, 8 Sep 2001 12:32:12 +0000 (12:32 +0000)]
Add missing entry to memory type name table and adjust field widths.

23 years agoAdd a command 'memmap' to print out the EFI memory map.
Doug Rabson [Sat, 8 Sep 2001 12:21:37 +0000 (12:21 +0000)]
Add a command 'memmap' to print out the EFI memory map.

23 years agoConvert the CMedia driver to using a device mutex and INTR_MPSAFE.
Brian Feldman [Sat, 8 Sep 2001 05:20:52 +0000 (05:20 +0000)]
Convert the CMedia driver to using a device mutex and INTR_MPSAFE.

23 years agoWrap array accesses in macros, which also happen to be lvalues:
Marcel Moolenaar [Sat, 8 Sep 2001 04:20:45 +0000 (04:20 +0000)]
Wrap array accesses in macros, which also happen to be lvalues:

        ifnet_addrs[i - 1]  -> ifaddr_byindex(i)
        ifindex2ifnet[i]    -> ifnet_byindex(i)

This is intended to ease the conversion to SMPng.

Forgotten by: jlemon

23 years agoFixes problem with pkg_add -r when adding multiple pkgs.
Bill Swingle [Fri, 7 Sep 2001 22:32:37 +0000 (22:32 +0000)]
Fixes problem with pkg_add -r when adding multiple pkgs.

Reviewed by: chris

23 years ago1) If __SAPP stream is not seekable, remove __SAPP flag on first call instead
Andrey A. Chernov [Fri, 7 Sep 2001 17:16:02 +0000 (17:16 +0000)]
1) If __SAPP stream is not seekable, remove __SAPP flag on first call instead
of repeating unsuccessful lseek call on each write (original stdio bug).

2) Save errno accross _sseek call in _swrite to not touch it in case write
success (original stdio bug).

3) Add _sseek error checking back, but only for __SOPT mode now.

23 years agoDriver supports CS4280 rather than CS428x.
Orion Hodson [Fri, 7 Sep 2001 16:44:13 +0000 (16:44 +0000)]
Driver supports CS4280 rather than CS428x.

MFC after: 10 days

23 years agoBack out (with prejudice) the last WARNS=2 fix. I cannot understand
Mark Murray [Fri, 7 Sep 2001 16:20:38 +0000 (16:20 +0000)]
Back out (with prejudice) the last WARNS=2 fix. I cannot understand
its failure mode, and will revisit it later.

23 years agoAdd a bit of background to the ACPI release note.
Bruce A. Mah [Fri, 7 Sep 2001 15:30:18 +0000 (15:30 +0000)]
Add a bit of background to the ACPI release note.

23 years agoYay!
Ruslan Ermilov [Fri, 7 Sep 2001 15:18:45 +0000 (15:18 +0000)]
Yay!
Make this manpage readable and match the reality.

23 years agoMove to using .In instead of .Fd #include <> for include mark-up.
Jeroen Ruigrok van der Werven [Fri, 7 Sep 2001 14:46:36 +0000 (14:46 +0000)]
Move to using .In instead of .Fd #include <> for include mark-up.

Inspired by comment from: dd

23 years agoreconnect strfmon.c to build list
Alexey Zelkin [Fri, 7 Sep 2001 13:03:16 +0000 (13:03 +0000)]
reconnect strfmon.c to build list