]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoWARNS=2 fixups.
Mark Murray [Sun, 2 Dec 2001 12:09:41 +0000 (12:09 +0000)]
WARNS=2 fixups.

22 years agoAdded OpenBSD 3.0.
Wolfram Schneider [Sun, 2 Dec 2001 11:19:24 +0000 (11:19 +0000)]
Added OpenBSD 3.0.

22 years agoNetBSD 1.4 branch and NetBSD 1.5 branch are still alive.
Wolfram Schneider [Sun, 2 Dec 2001 11:16:30 +0000 (11:16 +0000)]
NetBSD 1.4 branch and NetBSD 1.5 branch are still alive.

Submitted by: jhawk@netbsd.org, itojun@iijlab.net

22 years agoprototype functions and fix some line wrapping
Alfred Perlstein [Sun, 2 Dec 2001 11:10:46 +0000 (11:10 +0000)]
prototype functions and fix some line wrapping

22 years agoInitial support for the newer SiS chipsets, based on docs we finally
Søren Schmidt [Sun, 2 Dec 2001 10:48:52 +0000 (10:48 +0000)]
Initial support for the newer SiS chipsets, based on docs we finally
got from SiS.

This should also close PR 32421 which has patches which seem
to set the timing registers wrongly according to SiS...

22 years agoMake it works for non ASCII compatible encodings too.
Andrey A. Chernov [Sun, 2 Dec 2001 09:15:54 +0000 (09:15 +0000)]
Make it works for non ASCII compatible encodings too.
The only assumption left is that 'A'..'Z' 'a'..'z' both are contiguous

22 years agoEnable build of smbfs. It works now as before on both UP and MP machines.
Boris Popov [Sun, 2 Dec 2001 09:01:35 +0000 (09:01 +0000)]
Enable build of smbfs. It works now as before on both UP and MP machines.

22 years agoCatch up with KSE changes.
Boris Popov [Sun, 2 Dec 2001 08:56:58 +0000 (08:56 +0000)]
Catch up with KSE changes.

Submitted by: Max Khon <fjoe@iclub.nsu.ru>

22 years agoFix a bug with transmitter restart after receiving a 0 window. The
Matthew Dillon [Sun, 2 Dec 2001 08:49:29 +0000 (08:49 +0000)]
Fix a bug with transmitter restart after receiving a 0 window.  The
receiver was not sending an immediate ack with delayed acks turned on
when the input buffer is drained, preventing the transmitter from
restarting immediately.

Propogate the TCP_NODELAY option to accept()ed sockets.  (Helps tbench and
is a good idea anyway).

Some cleanup.  Identify additonal issues in comments.

MFC after: 1 day

22 years agoPull netsmb requester from the pre-KSE world. This update mostly based
Boris Popov [Sun, 2 Dec 2001 08:47:29 +0000 (08:47 +0000)]
Pull netsmb requester from the pre-KSE world. This update mostly based
on the patches submitted by Max Khon <fjoe@iclub.nsu.ru>

22 years agoAdding man page for the em driver.
Prafulla Deuskar [Sun, 2 Dec 2001 07:45:22 +0000 (07:45 +0000)]
Adding man page for the em driver.

Submitted by:Prafulla Deuskar
Reviewed by:Paul Saab
MFC after:1 week

22 years agoThis is the first commit of the Intel gigabit driver for
Prafulla Deuskar [Sun, 2 Dec 2001 07:37:17 +0000 (07:37 +0000)]
This is the first commit of the Intel gigabit driver for
PRO/1000 cards.

Submitted by:Prafulla Deuskar
Reviewed by: Paul Saab
MFC after:1 week

22 years agoRe-enable the ciss driver, now that it should be building again.
Mike Smith [Sun, 2 Dec 2001 06:19:15 +0000 (06:19 +0000)]
Re-enable the ciss driver, now that it should be building again.

22 years agoCatch up with the proc -> d_thread_t change.
Mike Smith [Sun, 2 Dec 2001 06:17:16 +0000 (06:17 +0000)]
Catch up with the proc -> d_thread_t change.

22 years agoForget to remove fsck from src/release/{alpha,pc98}/boot_crunch.conf,
Makoto Matsushita [Sun, 2 Dec 2001 04:58:40 +0000 (04:58 +0000)]
Forget to remove fsck from src/release/{alpha,pc98}/boot_crunch.conf,
which should be done with src/release/i386/boot_crunch.conf rev 1.51
and src/usr.sbin/sysinstall/install.c rev 1.311.

Spotted by: nyan (via ReichaNet)

22 years agoSysinstall cleanups for installation:
Makoto Matsushita [Sun, 2 Dec 2001 04:47:46 +0000 (04:47 +0000)]
Sysinstall cleanups for installation:
        1) Use devfs to mount filesystems.  If mounting devfs is fail,
           fallback to old code.
        2) When fscking filesystems, use 'fsck_ffs' explicitly.  As a
           result, we no longer need 'fsck' the wrapper program.

Reviewed by: jkh

22 years agoo Update copyright.
Chris D. Faulhaber [Sun, 2 Dec 2001 04:27:13 +0000 (04:27 +0000)]
o Update copyright.
o Updates content to reflect the changes to acl.h for
  internal/external acl_t representations.

Approved by: rwatson
Obtained from: TrustedBSD Project

22 years agoAdd functionality and fix bugs so the driver will work with soon-to-be
Scott Long [Sun, 2 Dec 2001 03:47:33 +0000 (03:47 +0000)]
Add functionality and fix bugs so the driver will work with soon-to-be
released management apps.

1.  Implement poll().  This will check for queued aif's so that a
    subsequent ioctl call to retrieve the next aif will not block.
2.  Don't catch signals when sleeping on a fib sent from userland.  This
    causes a race and panic due to the pthread context switcher waking
    up the tsleep at inopportune times.
3.  Fix some whitespace nits.

MFC after: 3 days

22 years agoAdd manpages describing various vfs and vm APIs.
Alfred Perlstein [Sun, 2 Dec 2001 02:13:35 +0000 (02:13 +0000)]
Add manpages describing various vfs and vm APIs.

Submitted by: Chad David <davidc@acns.ab.ca>

22 years agoUse calloc(3) instead of rolling our own.
Chris D. Faulhaber [Sun, 2 Dec 2001 01:52:34 +0000 (01:52 +0000)]
Use calloc(3) instead of rolling our own.

22 years agoSort includes.
Mark Murray [Sun, 2 Dec 2001 00:06:22 +0000 (00:06 +0000)]
Sort includes.

22 years agoStyle fixups.
Mark Murray [Sat, 1 Dec 2001 21:12:04 +0000 (21:12 +0000)]
Style fixups.

Sort function declarations, includes. Make consistent WRT use of _P()
macro (ugh!)

Inspired by: bde

22 years agoWork around world breakage in previous commit. The bug is in
Bruce Evans [Sat, 1 Dec 2001 19:48:59 +0000 (19:48 +0000)]
Work around world breakage in previous commit.  The bug is in
<security/pam_misc.h>.  It declares a function parameter named 'send',
so nothing that includes both <unistd.h> (which declares send(2)) and
<security/pam_misc.h> be compiled with WARNS=2 unless NO_WERRROR is
set.

Fixed order of WARNS.

22 years agoFix namespace pollution by protecting names used as arguments.
Mark Murray [Sat, 1 Dec 2001 19:32:18 +0000 (19:32 +0000)]
Fix namespace pollution by protecting names used as arguments.

Reported by: bde

22 years agoProtect names that are used elsewhere. This fixes WARNS=2 breakage
Mark Murray [Sat, 1 Dec 2001 18:48:36 +0000 (18:48 +0000)]
Protect names that are used elsewhere. This fixes WARNS=2 breakage
in crypto telnet.

22 years agoWARNS=2 fixes.
Mark Murray [Sat, 1 Dec 2001 17:46:46 +0000 (17:46 +0000)]
WARNS=2 fixes.

Reviewed by: bde (a while back)

22 years agoRemove the 'irq' string from the irqN part of the "interrupts" display.
Mark Murray [Sat, 1 Dec 2001 17:41:37 +0000 (17:41 +0000)]
Remove the 'irq' string from the irqN part of the "interrupts" display.
This allows us to see the irq number when device names ate too long.

22 years agoWARNS=2 and s/register//.
Mark Murray [Sat, 1 Dec 2001 17:34:42 +0000 (17:34 +0000)]
WARNS=2 and s/register//.

Set NO_WERROR=true because there are some flex(1) issues that need
to be dealt with.

22 years agoUse tabs where possible.
David E. O'Brien [Sat, 1 Dec 2001 17:14:34 +0000 (17:14 +0000)]
Use tabs where possible.

22 years agoAllow for ip_output() to be called with a NULL route pointer.
Ruslan Ermilov [Sat, 1 Dec 2001 13:48:16 +0000 (13:48 +0000)]
Allow for ip_output() to be called with a NULL route pointer.
This fixes a panic I introduced yesterday in ip_icmp.c,v 1.64.

22 years agoForget to remove kget.c, since kget feature is already gone.
Makoto Matsushita [Sat, 1 Dec 2001 13:15:40 +0000 (13:15 +0000)]
Forget to remove kget.c, since kget feature is already gone.

22 years agoRemove kget() feature, which is removed from 5-current kernel.
Makoto Matsushita [Sat, 1 Dec 2001 13:13:27 +0000 (13:13 +0000)]
Remove kget() feature, which is removed from 5-current kernel.

Since userconfig feature is implemented by tweaking variables (hint.*)
with loader(8), we can put back an equivalent feature.  Maybe the first
step for this is to commit yokota-san's patch (add userconfig command
for loader).

Approved by: jkh

22 years ago(null commit)
Sheldon Hearn [Sat, 1 Dec 2001 13:13:21 +0000 (13:13 +0000)]
(null commit)

The log for the previous revision misattributed the Hart list to
Heckenbach.  Craig Hart doesn't know where I got the name Heckenbach
from. :-)

22 years agoCorrect misattribution of the Hart list, which I had mistakenly
Sheldon Hearn [Sat, 1 Dec 2001 13:08:20 +0000 (13:08 +0000)]
Correct misattribution of the Hart list, which I had mistakenly
attributed to Heckenbach.

22 years agoAdd support for sysconf(_SC_NPROCESSORS_CONF) and sysconf(_SC_NPROCESSORS_ONLN).
Joseph Koshy [Sat, 1 Dec 2001 12:07:42 +0000 (12:07 +0000)]
Add support for sysconf(_SC_NPROCESSORS_CONF) and sysconf(_SC_NPROCESSORS_ONLN).
This API is supported in SVR4.0MP, Solaris, Linux, AIX and Tru64 Unix.

PR: bin/27489
Submitted by: Arun Sharma <arun@sharma-home.net>
Reviewed by: bde (prior version)

22 years agoAdd a missing underscore.
Mike Barcroft [Sat, 1 Dec 2001 03:48:15 +0000 (03:48 +0000)]
Add a missing underscore.

Submitted by: fenner

22 years agoo Stop abusing MD headers with non-MD types.
Mike Barcroft [Sat, 1 Dec 2001 03:43:01 +0000 (03:43 +0000)]
o Stop abusing MD headers with non-MD types.
o Hide nonstandard functions and types in <netinet/in.h> when
  _POSIX_SOURCE is defined.
o Add some missing types (required by POSIX.1-200x) to <netinet/in.h>.
o Restore vendor ID from Rev 1.1 in <netinet/in.h> and make use of new
  __FBSDID() macro.
o Fix some miscellaneous issues in <arpa/inet.h>.
o Correct final argument for the inet_ntop() function (POSIX.1-200x).
o Get rid of the namespace pollution from <sys/types.h> in
  <arpa/inet.h>.

Reviewed by: fenner
Partially submitted by: bde

22 years agovm/vm_kern.c: rate limit (to once per second) diagnostic printf when
Luigi Rizzo [Sat, 1 Dec 2001 00:21:30 +0000 (00:21 +0000)]
vm/vm_kern.c: rate limit (to once per second) diagnostic printf when
you run out of mbuf address space.

kern/subr_mbuf.c: print a warning message when mb_alloc fails, again
rate-limited to at most once per second. This covers other
cases of mbuf allocation failures. Probably it also overlaps the
one handled in vm/vm_kern.c, so maybe the latter should go away.

This warning will let us gradually remove the printf that are scattered
across most network drivers to report mbuf allocation failures.
Those are potentially dangerous, in that they are not rate-limited and
can easily cause systems to panic.

Unless there is disagreement (which does not seem to be the case
judging from the discussion on -net so far), and because this is
sort of a safety bugfix, I plan to commit a similar change to STABLE
during the weekend (it affects kern/uipc_mbuf.c there).

Discussed-with: jlemon, silby and -net

22 years agoMerge the (in)complete ANSIfication work from src/crypto/telnet.
Mark Murray [Fri, 30 Nov 2001 22:37:28 +0000 (22:37 +0000)]
Merge the (in)complete ANSIfication work from src/crypto/telnet.

22 years agoDamn. The previous mega-commit was incomplete WRT ANSIfication. This
Mark Murray [Fri, 30 Nov 2001 22:28:07 +0000 (22:28 +0000)]
Damn. The previous mega-commit was incomplete WRT ANSIfication. This
fixes that.

22 years agoAfter running "make unifdef", commit the result. This code is now
Mark Murray [Fri, 30 Nov 2001 22:09:10 +0000 (22:09 +0000)]
After running "make unifdef", commit the result. This code is now
a complete subset of the crypto (master) code.

22 years agoAfter running a "make unifdef", commit the resultant diffs.
Mark Murray [Fri, 30 Nov 2001 22:03:20 +0000 (22:03 +0000)]
After running a "make unifdef", commit the resultant diffs.

This code is now a complete sunset of the crypto (master) code.

22 years agoo Update sysctl.8 to reflect renaming of various security-related
Robert Watson [Fri, 30 Nov 2001 21:55:22 +0000 (21:55 +0000)]
o Update sysctl.8 to reflect renaming of various security-related
  sysctls, and to introduce new ones.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs

22 years agoInvoke the new 'unifdef:' target, and make this lib a complete subset
Mark Murray [Fri, 30 Nov 2001 21:55:10 +0000 (21:55 +0000)]
Invoke the new 'unifdef:' target, and make this lib a complete subset
of the crypto (master) code.

22 years agoo Introduce kern.security.bsd.unprivileged_read_msgbuf, which allows
Robert Watson [Fri, 30 Nov 2001 21:40:52 +0000 (21:40 +0000)]
o Introduce kern.security.bsd.unprivileged_read_msgbuf, which allows
  the administrator to restrict access to the kernel message buffer.
  It defaults to '1', which permits access, but if set to '0', requires
  that the process making the sysctl() have appropriate privilege.
o Note that for this to be effective, access to this data via system
  logs derived from /dev/klog must also be limited.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs

22 years agoNew release note: send-pr(1) -a (+ MFC).
Bruce A. Mah [Fri, 30 Nov 2001 21:37:25 +0000 (21:37 +0000)]
New release note:  send-pr(1) -a (+ MFC).

MFC noted:  Increased TCP sndspace/rcvspace defaults.

22 years agoDiff-reduce WRT src/secure/*telnet*/Makefile.
Mark Murray [Fri, 30 Nov 2001 21:34:51 +0000 (21:34 +0000)]
Diff-reduce WRT src/secure/*telnet*/Makefile.

Also, add an "unifdef:" target, so that the telnet sources can
be remade from the crypto sources in src/crypto/telnet.

22 years agoThe transmit burst limit for newreno completely breaks TCP's performance
Matthew Dillon [Fri, 30 Nov 2001 21:33:39 +0000 (21:33 +0000)]
The transmit burst limit for newreno completely breaks TCP's performance
if the receive side is using delayed acks.  Temporarily remove it.

MFC after: 0 days

22 years agoo Further sysctl name simplification, generally stripping 'permitted',
Robert Watson [Fri, 30 Nov 2001 21:33:16 +0000 (21:33 +0000)]
o Further sysctl name simplification, generally stripping 'permitted',
  using '_'s more consistently.

Discussed with: bde, jhb
Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs

22 years agoStyle clean-up, and diff-reduce WRT src/secure/*telnet*/Makefile
Mark Murray [Fri, 30 Nov 2001 21:14:44 +0000 (21:14 +0000)]
Style clean-up, and diff-reduce WRT src/secure/*telnet*/Makefile

Lost in this commit - KerberosIV compatability. This will be
re-added later.

22 years agoStyle clean-up, and diff-reduce WRT src/secure/*/telnet/Makefile
Mark Murray [Fri, 30 Nov 2001 21:12:31 +0000 (21:12 +0000)]
Style clean-up, and diff-reduce WRT src/secure/*/telnet/Makefile

22 years agoClean up makefiles, and turn on WARNS=2. Take into account the telnet
Mark Murray [Fri, 30 Nov 2001 21:10:58 +0000 (21:10 +0000)]
Clean up makefiles, and turn on WARNS=2. Take into account the telnet
#if cleanup.

22 years agoForced commit.
Mark Murray [Fri, 30 Nov 2001 21:09:11 +0000 (21:09 +0000)]
Forced commit.

This file was taken from the Heimdal implementation of Kerberos 5, done
by KTH.

22 years agoVery large style makeover.
Mark Murray [Fri, 30 Nov 2001 21:06:38 +0000 (21:06 +0000)]
Very large style makeover.

1) ANSIfy.
2) Clean up ifdefs so that
   a) ones that never/always apply are appropriately either
      fully removed, or just the #if junk is removed.
   b) change #if defined(FOO) for appropiate values of FOO.
      (currently AUTHENTICATION and ENCRYPTION)
3) WARNS=2 fixing
4) GC other unused stuff

This code can now be unifdef(1)ed to make non-crypto telnet.

22 years agoo Move current inhabitants of kern.security to kern.security.bsd, so
Robert Watson [Fri, 30 Nov 2001 20:58:31 +0000 (20:58 +0000)]
o Move current inhabitants of kern.security to kern.security.bsd, so
  that new models can inhabit kern.security.<modelname>.
o While I'm there, shorten somewhat excessive variable names, and clean
  things up a little.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs

22 years agoo Improve consistency of style by fixing tabs around _PATH_AUTHCONF.
Robert Watson [Fri, 30 Nov 2001 16:25:09 +0000 (16:25 +0000)]
o Improve consistency of style by fixing tabs around _PATH_AUTHCONF.

22 years agoo Introduce _PATH_CAPABILITY and _PATH_CAPABILITY_DB to the system
Robert Watson [Fri, 30 Nov 2001 16:24:20 +0000 (16:24 +0000)]
o Introduce _PATH_CAPABILITY and _PATH_CAPABILITY_DB to the system
  path set.  These files will hold per-user capability information.

Obtained from: TrustedBSD Project
Submitted by: tmm

22 years agoAdd a couple of minor changes.
Mitsuru IWASAKI [Fri, 30 Nov 2001 16:06:00 +0000 (16:06 +0000)]
Add a couple of minor changes.
 - set sc->acpi_s4bios to 1 by default for hibernation until
   OS-initiated S4 transition is implemented.
 - change the behavior of acpi_sleep_state_sysctl() if new value is
   the same as old one, do nothing instead of EINVAL.

22 years agoUse 'mkdir -p /.attribute/system' instead of breaking it into
Robert Watson [Fri, 30 Nov 2001 15:32:07 +0000 (15:32 +0000)]
Use 'mkdir -p /.attribute/system' instead of breaking it into
two seperate mkdir targets.

Submitted by: jedgar

22 years agoUse 'mkdir -p /.attribute/system' instead of breaking it into
Robert Watson [Fri, 30 Nov 2001 15:21:20 +0000 (15:21 +0000)]
Use 'mkdir -p /.attribute/system' instead of breaking it into
two seperate mkdir targets.

22 years agoREADME.extattr incorrectly specified sample command lines for
Robert Watson [Fri, 30 Nov 2001 15:15:27 +0000 (15:15 +0000)]
README.extattr incorrectly specified sample command lines for
UFS_EXTATTR_AUTOSTART.  Insert the missing 'initattr' arguments
to extattrctl.

Noticed by: green

22 years agoPay attention to failures to SIOCAIFADDR and SIOCDIFFADDR.
Brian Somers [Fri, 30 Nov 2001 14:01:21 +0000 (14:01 +0000)]
Pay attention to failures to SIOCAIFADDR and SIOCDIFFADDR.

22 years agoDon't provide an RTA_GATEWAY sockaddr when we write RTM_CHANGE messages
Brian Somers [Fri, 30 Nov 2001 14:01:20 +0000 (14:01 +0000)]
Don't provide an RTA_GATEWAY sockaddr when we write RTM_CHANGE messages
to the routing socket.

The local address on a point-to-point interface is not actually a
gateway address - despite it appearing in the second column of
netstat -r's output.  Providing a gateway to an RTM_CHANGE will
currently change the route's interface so that it's using the
specified gateway - not what we want.

Patiently explained to me by: ru

22 years agoAdd some DEBUG logging to tell us when interface addresses are being
Brian Somers [Fri, 30 Nov 2001 14:01:18 +0000 (14:01 +0000)]
Add some DEBUG logging to tell us when interface addresses are being
added and removed

22 years agoDuring SIOCAIFADDR, if in_ifinit() fails and we've already added an
Brian Somers [Fri, 30 Nov 2001 14:00:55 +0000 (14:00 +0000)]
During SIOCAIFADDR, if in_ifinit() fails and we've already added an
interface address, blow the address away again before returning the
error.

In in_ifinit(), if we get an error from rtinit() and we've also got
a destination address, return the error rather than masking EEXISTS.
Failing to create a host route when configuring an interface should
be treated as an error.

22 years agoMerge NetBSD's changes from netbsd_strtod.c in preparation of
Alexey Zelkin [Fri, 30 Nov 2001 12:48:30 +0000 (12:48 +0000)]
Merge NetBSD's changes from netbsd_strtod.c in preparation of
removing it from our source tree in order to have one version
of strtod() for all arches. netbsd_strtod.c still left in source
tree until alpha folks make sure that our native strtod() works
as well as NetBSD's one.

Reviewed by: peter, bde (some time ago)

22 years agocpuid bit 30 is 'IA64', for when you're running in i386 mode on an ia64
Peter Wemm [Fri, 30 Nov 2001 11:57:23 +0000 (11:57 +0000)]
cpuid bit 30 is 'IA64', for when you're running in i386 mode on an ia64
cpu.  (This is for either userland apps running in i386 mode on an ia64
OS, or when the cpu is in i386 legacy mode running an i386 OS).

22 years agoMake the error messaging more helpful.
Crist J. Clark [Fri, 30 Nov 2001 11:35:01 +0000 (11:35 +0000)]
Make the error messaging more helpful.

PR: 31483
Approved by: iwasaki, ru
MFC after: 4 days

22 years ago- Make ip_rtaddr() global, and use it to look up the correct source
Ruslan Ermilov [Fri, 30 Nov 2001 10:40:28 +0000 (10:40 +0000)]
- Make ip_rtaddr() global, and use it to look up the correct source
  address in icmp_reflect().
- Two new "struct icmpstat" members: icps_badaddr and icps_noroute.

PR: kern/31575
Obtained from: BSD/OS
MFC after: 1 week

22 years ago* Don't use critical_enter/critical_exit when accessing the VHPT - its
Doug Rabson [Fri, 30 Nov 2001 10:07:54 +0000 (10:07 +0000)]
* Don't use critical_enter/critical_exit when accessing the VHPT - its
  pointless and would be inadequate for SMP systems. We will rely on the
  VM system's locks to serialise this for now.
* Change pmap_remove() so that if the range being removed is larger than
  the number of pages mapped by the pmap, we iterate over the currently
  mapped pages instead of over the virtual address range. This should
  make a difference when removing large virtual address ranges from an
  address space.

22 years agoProperly classify 0x80 - 0xa0 range
Andrey A. Chernov [Fri, 30 Nov 2001 07:19:21 +0000 (07:19 +0000)]
Properly classify 0x80 - 0xa0 range

22 years agoSpelling error: dependant -> dependent
Valentino Vaschetto [Fri, 30 Nov 2001 07:18:23 +0000 (07:18 +0000)]
Spelling error: dependant -> dependent

PR

22 years agoAdd 0x98 to CONTROL section
Andrey A. Chernov [Fri, 30 Nov 2001 07:10:08 +0000 (07:10 +0000)]
Add 0x98 to CONTROL section

22 years agoBack out my 'fix', resid is different for strategy than for write
Alfred Perlstein [Fri, 30 Nov 2001 06:24:34 +0000 (06:24 +0000)]
Back out my 'fix', resid is different for strategy than for write

22 years agoImplement several of the c99 updates to printf(3):
Bill Fenner [Fri, 30 Nov 2001 06:12:15 +0000 (06:12 +0000)]
Implement several of the c99 updates to printf(3):
 - New length modifiers: hh, j, t, z.
 - New flag: '.  Note that %'f is not yet implemented.
 - Use "inf"/"nan" for efg formats, "INF"/"NAN" for EFG formats.
 - Implemented %q in terms of %ll; if "quad_t" is not "long long"
   %q will break.

Still to do:
 - %C, %S, %lc, %ls (wide character support)
 - %'f (thousands in integer portion of %f)
 - %a/%A (exact hex representation of floating-point numbers)

Garrett Wollman wrote the first version of the vfprintf.c update;
Mike Barcroft wrote the first version of the printf.3 changes.

22 years agoMake the same fix for writes to RAW objects.
Alfred Perlstein [Fri, 30 Nov 2001 05:59:47 +0000 (05:59 +0000)]
Make the same fix for writes to RAW objects.

22 years agowrite should return the number of bytes written, not 0 on success.
Alfred Perlstein [Fri, 30 Nov 2001 05:54:30 +0000 (05:54 +0000)]
write should return the number of bytes written, not 0 on success.

Submitted by: Jonathan Mini <mini@haikugeek.com>
PR: kern/32350

22 years agoClarify isblank range
Andrey A. Chernov [Fri, 30 Nov 2001 05:39:08 +0000 (05:39 +0000)]
Clarify isblank range

22 years agoStart every data line with keyword
Andrey A. Chernov [Fri, 30 Nov 2001 05:24:09 +0000 (05:24 +0000)]
Start every data line with keyword

22 years agoCleanup: cosmetique, standards conformance, BLANK/SPACE/GRAPH relation
Andrey A. Chernov [Fri, 30 Nov 2001 05:05:53 +0000 (05:05 +0000)]
Cleanup: cosmetique, standards conformance, BLANK/SPACE/GRAPH relation

22 years agoClarify valid isspace() range
Andrey A. Chernov [Fri, 30 Nov 2001 02:01:32 +0000 (02:01 +0000)]
Clarify valid isspace() range

22 years agoClean DIGIT/XDIGIT sections
Andrey A. Chernov [Fri, 30 Nov 2001 01:22:01 +0000 (01:22 +0000)]
Clean DIGIT/XDIGIT sections

22 years agokillall operates on all processes owned by the real uid, not the
Brian S. Dean [Fri, 30 Nov 2001 00:30:28 +0000 (00:30 +0000)]
killall operates on all processes owned by the real uid, not the
effective uid.

MFC after: 3 days

22 years agoPer jlemon request, reintroduce some printf() when an
Luigi Rizzo [Thu, 29 Nov 2001 23:47:47 +0000 (23:47 +0000)]
Per jlemon request, reintroduce some printf() when an
mbuf allocation fails, and fix (i hope) a couple of style bugs.

I believe these printf() are extremely dangerous because now they can
occur on every incoming packet and are not rate limited. They were
meant to warn the sysadmin about lack of resources, but now they
can become a nice way to panic your system under load.

Other drivers (e.g. the fxp driver) have nothing like this.

There is a pending discussion on putting this kind of warnings
elsewhere, and I hope we can fix this soon.

22 years agoReword pciconf(8) -v release to be a little more useful (now that I
Bruce A. Mah [Thu, 29 Nov 2001 23:41:14 +0000 (23:41 +0000)]
Reword pciconf(8) -v release to be a little more useful (now that I
understand what it does), move it to the userland section, and note MFC.

22 years agoNew release notes: ciss driver, sbni driver.
Bruce A. Mah [Thu, 29 Nov 2001 23:24:41 +0000 (23:24 +0000)]
New release notes:  ciss driver, sbni driver.

(Hardware list updated accordingly.)

22 years agoFor i386 architecture, remove an expensive m_devget() (and the
Luigi Rizzo [Thu, 29 Nov 2001 22:46:48 +0000 (22:46 +0000)]
For i386 architecture, remove an expensive m_devget() (and the
underlying unaligned bcopy) on incoming packets that are already
available (albeit unaligned) in a buffer.
The performance improvement varies, depending on CPU and memory
speed, but can be quite large especially on slow CPUs. I have seen
over 50% increase on forwarding speed on the sis driver for the
486/133 (embedded systems), which does exactly the same thing.

The behaviour is controlled by a sysctl variable, hw.dc_quick which
defaults to 1. Set it to 0 to restore the old behaviour.

After running a few experiments (in userland, though) I am convinced
that doing the m_devget() is detrimental to performance in almost
all cases.

Even if your CPU has degraded performance with misaligned data,
the bcopy() in the driver has the same overhead due to misaligment
as the one that you save in the uiomove(), plus you do one extra
copy and pollute the cache.

But more often than not, you do not even have to touch the payload,
e.g. when you are forwarding packets, and even in the often-cited
case of NFS, you often end up passing a pointer to the payload to
the disk controller.

In any case, you can play with the sysctl variable to toggle between
the two behaviours, and see if it makes a difference.

MFC-after: 3 days

22 years agoFix pam_ssh by adding an IPv4or6 (evidently, this was broken by my last
Brian Feldman [Thu, 29 Nov 2001 21:16:11 +0000 (21:16 +0000)]
Fix pam_ssh by adding an IPv4or6 (evidently, this was broken by my last
OpenSSH import) declaration and strdup(3)ing a value which is later
free(3)d, rather than letting the system try to free it invalidly.

22 years agoBring release note on /dev/net into line with reality.
Bruce A. Mah [Thu, 29 Nov 2001 19:00:16 +0000 (19:00 +0000)]
Bring release note on /dev/net into line with reality.

22 years agoFold ANDREW_LOCKD into -current.
Alfred Perlstein [Thu, 29 Nov 2001 17:36:45 +0000 (17:36 +0000)]
Fold ANDREW_LOCKD into -current.

22 years agoNote that stty is a utility and not... err... a program....
Brian Feldman [Thu, 29 Nov 2001 15:46:54 +0000 (15:46 +0000)]
Note that stty is a utility and not... err... a program....

Submitted by: ru

22 years agofix DIAGNOSTIC panic caused by proc/thread typo
Andrew Gallatin [Thu, 29 Nov 2001 15:33:46 +0000 (15:33 +0000)]
fix DIAGNOSTIC panic caused by proc/thread typo

tested by: Martijn Pronk <martijn@smartie.xs4all.nl>

22 years agoClarify that is[x]digit() class is the same in any locale
Andrey A. Chernov [Thu, 29 Nov 2001 15:23:46 +0000 (15:23 +0000)]
Clarify that is[x]digit() class is the same in any locale

22 years agoFix a bug about CIS string comparison. Pccardd should be able to distinguish
Mitsuru IWASAKI [Thu, 29 Nov 2001 14:33:57 +0000 (14:33 +0000)]
Fix a bug about CIS string comparison.  Pccardd should be able to distinguish
card "MELCO" "LPC2-T"
and
card "MELCO" "LPC2-TX"
by this fix.

Reported by: Kitagawa Shoichi <sk@xstar.kiu.ac.jp>,
NINOMIYA Hideyuki <nin@shikoku.ne.jp>
MFC after: 1 week

22 years agoDuh. Back out most of my previous commit. It was mostly covered in the
Crist J. Clark [Thu, 29 Nov 2001 09:44:33 +0000 (09:44 +0000)]
Duh. Back out most of my previous commit. It was mostly covered in the
STANDARDS section of the page. Add one remark there about inet_pton(3)
only understanding decimal values (in contrast to inet_aton(3) and
friends who are happy with 0ac.020.25 for 172.16.0.25).

Caught by: ru
MFC after: 2 days

22 years agoFixed memory leak in dialog_gauge(3).
Ruslan Ermilov [Thu, 29 Nov 2001 07:42:13 +0000 (07:42 +0000)]
Fixed memory leak in dialog_gauge(3).

PR: gnu/32260
Submitted by: Igor Pokrovsky <tiamat@telegraph.spb.ru>
MFC after: 3 days

22 years agoOops, actually fix the namespace pollution for atoll() as the previous
Bruce Evans [Thu, 29 Nov 2001 07:08:56 +0000 (07:08 +0000)]
Oops, actually fix the namespace pollution for atoll() as the previous
commit claimed to do.

22 years agoStty is a program, not a function.
Brian Feldman [Thu, 29 Nov 2001 03:26:42 +0000 (03:26 +0000)]
Stty is a program, not a function.

22 years agoBump p1003.1-90 to p1003.1-96.
Brian Feldman [Thu, 29 Nov 2001 03:21:16 +0000 (03:21 +0000)]
Bump p1003.1-90 to p1003.1-96.

22 years agorename() can't guarantee "to" always exists if it didn't exist in the first
Gregory Neil Shapiro [Thu, 29 Nov 2001 03:17:12 +0000 (03:17 +0000)]
rename() can't guarantee "to" always exists if it didn't exist in the first
place -- for example, rename("existing", "newfile"); on a read-only file
system.

Reviewed by: green
MFC after: 3 days