]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoAvoid truncating invalid wide characters that are outside the range of
Tim J. Robbins [Wed, 16 Oct 2002 11:37:38 +0000 (11:37 +0000)]
Avoid truncating invalid wide characters that are outside the range of
'unsigned char'; signal an error instead.

21 years agoCast the first argument to bzero() to `void *' after casting it to
Robert Drehmel [Wed, 16 Oct 2002 11:15:35 +0000 (11:15 +0000)]
Cast the first argument to bzero() to `void *' after casting it to
`uintptr_t' to pass it as the type bzero() expects.

21 years agoCatch up with US version
Udo Erdelhoff [Wed, 16 Oct 2002 10:55:45 +0000 (10:55 +0000)]
Catch up with US version
layout.sgml 1.4 -> 1.5
trouble.sgml 1.5 -> 1.6

21 years agoProperly put macro args in ().
Poul-Henning Kamp [Wed, 16 Oct 2002 10:52:15 +0000 (10:52 +0000)]
Properly put macro args in ().

Spotted by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 10:45:53 +0000 (10:45 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint

21 years agoBe consistent about funtions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 10:42:13 +0000 (10:42 +0000)]
Be consistent about funtions being static.

Spotted by: FlexeLint

21 years agoPut an XXX: comment here to point out a couple of free() issues on
Poul-Henning Kamp [Wed, 16 Oct 2002 10:40:43 +0000 (10:40 +0000)]
Put an XXX: comment here to point out a couple of free() issues on
pnp_read_bytes().

Spotted by: FlexeLint

21 years agoNo need to declare M_ISADEV here.
Poul-Henning Kamp [Wed, 16 Oct 2002 10:38:48 +0000 (10:38 +0000)]
No need to declare M_ISADEV here.

Spotted by: FlexeLint.

21 years agoBe consistent about funtions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 10:16:17 +0000 (10:16 +0000)]
Be consistent about funtions being static.

Spotte by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 10:14:34 +0000 (10:14 +0000)]
Be consistent about functions being static.
Fix misindentation.

Spotted by: DARPA & NAI Labs.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 09:19:17 +0000 (09:19 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 09:14:59 +0000 (09:14 +0000)]
Be consistent about functions being static.
Properly put macro args in ().

Spotted by: FlexeLint.

21 years agoProperly put macro args in ().
Poul-Henning Kamp [Wed, 16 Oct 2002 09:07:30 +0000 (09:07 +0000)]
Properly put macro args in ().

Spotted by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 09:04:52 +0000 (09:04 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint.

21 years agoGet rid of checking for ip sec history. It is true that packets are not
Guido van Rooij [Wed, 16 Oct 2002 09:01:48 +0000 (09:01 +0000)]
Get rid of checking for ip sec history. It is true that packets are not
supposed to be checked by the firewall rules twice. However, because the
various ipsec handlers never call ip_input(), this never happens anyway.

This fixes the situation where a gif tunnel is encrypted with IPsec. In
such a case, after IPsec processing, the unencrypted contents from the
GIF tunnel are fed back to the ipintrq and subsequently handeld by
ip_input(). Yet, since there still is IPSec history attached, the
packets coming out from the gif device are never fed into the filtering
code.
This fix was sent to Itojun, and he pointed towartds
    http://www.netbsd.org/Documentation/network/ipsec/#ipf-interaction.
This patch actually implements what is stated there (specifically:
Packet came from tunnel devices (gif(4) and ipip(4)) will still
go through ipf(4). You may need to identify these packets by
using interface name directive in ipf.conf(5).

Reviewed by: rwatson
MFC after: 3 weeks

21 years agoFIx misindentation.
Poul-Henning Kamp [Wed, 16 Oct 2002 09:00:53 +0000 (09:00 +0000)]
FIx misindentation.

Spotted by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 08:57:14 +0000 (08:57 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 08:48:39 +0000 (08:48 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint.

21 years agoFix misindentation.
Poul-Henning Kamp [Wed, 16 Oct 2002 08:45:34 +0000 (08:45 +0000)]
Fix misindentation.

Spotted by: FlexeLint

21 years agoRename struct softc to struct mn_softc.
Poul-Henning Kamp [Wed, 16 Oct 2002 08:41:38 +0000 (08:41 +0000)]
Rename struct softc to struct mn_softc.

21 years agoFix comments and one resulting code confusion about the type of the
Poul-Henning Kamp [Wed, 16 Oct 2002 08:04:11 +0000 (08:04 +0000)]
Fix comments and one resulting code confusion about the type of the
"command" argument to VOP_IOCTL.

Spotted by: FlexeLint.

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Wed, 16 Oct 2002 08:00:32 +0000 (08:00 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint

21 years agoA better solution to avoiding variable sized structs in DEVFS.
Poul-Henning Kamp [Wed, 16 Oct 2002 07:51:18 +0000 (07:51 +0000)]
A better solution to avoiding variable sized structs in DEVFS.

21 years ago#include "opt_devfs.h" to protect against variable sized structures.
Poul-Henning Kamp [Wed, 16 Oct 2002 07:16:47 +0000 (07:16 +0000)]
#include "opt_devfs.h" to protect against variable sized structures.

Spotted by: FlexeLint

21 years agoFixed style bug near $FreeBSD$. VCS ids in copyrights should not be
Bruce Evans [Wed, 16 Oct 2002 07:01:10 +0000 (07:01 +0000)]
Fixed style bug near $FreeBSD$.  VCS ids in copyrights should not be
separated by a blank line.

21 years agoAdd support for building sparc64 releases.
Jake Burkholder [Wed, 16 Oct 2002 05:30:56 +0000 (05:30 +0000)]
Add support for building sparc64 releases.

Reviewed by: murray

21 years agos/clear/cleared/ for consistency (sigh)
Robert Watson [Wed, 16 Oct 2002 05:03:40 +0000 (05:03 +0000)]
s/clear/cleared/ for consistency  (sigh)

Reported by: dd

21 years agoAdd a Bugs section and note that fmtcheck() is out of sync with printf();
Tim J. Robbins [Wed, 16 Oct 2002 04:03:02 +0000 (04:03 +0000)]
Add a Bugs section and note that fmtcheck() is out of sync with printf();
it does not recognise any of the conversions or modifiers added in C99.

21 years agoAdd support for the XSI %C and %S formats, which are the same as %lc
Tim J. Robbins [Wed, 16 Oct 2002 03:55:53 +0000 (03:55 +0000)]
Add support for the XSI %C and %S formats, which are the same as %lc
and %ls.

21 years agoNever allow memory mapped I/O in PCI-X mode on controllers that
Justin T. Gibbs [Wed, 16 Oct 2002 02:59:03 +0000 (02:59 +0000)]
Never allow memory mapped I/O in PCI-X mode on controllers that
do not support that configuration.  This should fix problems with
embedded 7902 controllers running in PCI-X mode.

21 years agocorrect PCB locking in broadcast/multicast case that was exposed by change
Sam Leffler [Wed, 16 Oct 2002 02:33:28 +0000 (02:33 +0000)]
correct PCB locking in broadcast/multicast case that was exposed by change
to use udp_append

Reviewed by: hsu

21 years agoLast commit to add FAST_IPSEC support also rearranged udp_input so that
Sam Leffler [Wed, 16 Oct 2002 02:29:01 +0000 (02:29 +0000)]
Last commit to add FAST_IPSEC support also rearranged udp_input so that
all input packets are stored in the socket using udp_append.  This was
done so that all IPsec- and MAC-related checking is now consolidated in
udp_append.  Note that this can lead to an extra packet copy being done
before a packet is discarded because of one of these checks.

21 years agoTie new "Fast IPsec" code into the build. This involves the usual
Sam Leffler [Wed, 16 Oct 2002 02:25:05 +0000 (02:25 +0000)]
Tie new "Fast IPsec" code into the build.  This involves the usual
configuration stuff as well as conditional code in the IPv4 and IPv6
areas.  Everything is conditional on FAST_IPSEC which is mutually
exclusive with IPSEC (KAME IPsec implmentation).

As noted previously, don't use FAST_IPSEC with INET6 at the moment.

Reviewed by: KAME, rwatson
Approved by: silence
Supported by: Vernier Networks

21 years agoadd definitions for RIPEMD-160 HMAC and Skipjack encryption algorithms,
Sam Leffler [Wed, 16 Oct 2002 02:18:56 +0000 (02:18 +0000)]
add definitions for RIPEMD-160 HMAC and Skipjack encryption algorithms,
for use by "Fast IPsec"

21 years ago"Fast IPsec": this is an experimental IPsec implementation that is derived
Sam Leffler [Wed, 16 Oct 2002 02:10:08 +0000 (02:10 +0000)]
"Fast IPsec": this is an experimental IPsec implementation that is derived
from the KAME IPsec implementation, but with heavy borrowing and influence
of openbsd.  A key feature of this implementation is that it uses the kernel
crypto framework to do all crypto work so when h/w crypto support is present
IPsec operation is automatically accelerated.  Otherwise the protocol
implementations are rather differet while the SADB and policy management
code is very similar to KAME (for the moment).

Note that this implementation is enabled with a FAST_IPSEC option.  With this
you get all protocols; i.e. there is no FAST_IPSEC_ESP option.

FAST_IPSEC and IPSEC are mutually exclusive; you cannot build both into a
single system.

This software is well tested with IPv4 but should be considered very
experimental (i.e. do not deploy in production environments).  This software
does NOT currently support IPv6.  In fact do not configure FAST_IPSEC and
INET6 in the same system.

Obtained from: KAME + openbsd
Supported by: Vernier Networks

21 years agoMerge from vendor tzdata2002d.
Garrett Wollman [Wed, 16 Oct 2002 01:55:48 +0000 (01:55 +0000)]
Merge from vendor tzdata2002d.

PR: 44004
MFC after: 3 days

21 years agoRe-sort to reflect change of East Timor from TP to TL.
Garrett Wollman [Wed, 16 Oct 2002 01:55:10 +0000 (01:55 +0000)]
Re-sort to reflect change of East Timor from TP to TL.

21 years agoReplace aux mbufs with packet tags:
Sam Leffler [Wed, 16 Oct 2002 01:54:46 +0000 (01:54 +0000)]
Replace aux mbufs with packet tags:

o instead of a list of mbufs use a list of m_tag structures a la openbsd
o for netgraph et. al. extend the stock openbsd m_tag to include a 32-bit
  ABI/module number cookie
o for openbsd compatibility define a well-known cookie MTAG_ABI_COMPAT and
  use this in defining openbsd-compatible m_tag_find and m_tag_get routines
o rewrite KAME use of aux mbufs in terms of packet tags
o eliminate the most heavily used aux mbufs by adding an additional struct
  inpcb parameter to ip_output and ip6_output to allow the IPsec code to
  locate the security policy to apply to outbound packets
o bump __FreeBSD_version so code can be conditionalized
o fixup ipfilter's call to ip_output based on __FreeBSD_version

Reviewed by: julian, luigi (silent), -arch, -net, darren
Approved by: julian, silence from everyone else
Obtained from: openbsd (mostly)
MFC after: 1 month

21 years agoChange locking so that all snapshots on a particular filesystem share
Kirk McKusick [Wed, 16 Oct 2002 00:19:23 +0000 (00:19 +0000)]
Change locking so that all snapshots on a particular filesystem share
a common lock. This change avoids a deadlock between snapshots when
separate requests cause them to deadlock checking each other for a
need to copy blocks that are close enough together that they fall
into the same indirect block. Although I had anticipated a slowdown
from contention for the single lock, my filesystem benchmarks show
no measurable change in throughput on a uniprocessor system with
three active snapshots. I conjecture that this result is because
every copy-on-write fault must check all the active snapshots, so
the process was inherently serial already. This change removes the
last of the deadlocks of which I am aware in snapshots.

Sponsored by: DARPA & NAI Labs.

21 years agomissed additional callout that can run w/o Giant locked
Sam Leffler [Tue, 15 Oct 2002 23:25:56 +0000 (23:25 +0000)]
missed additional callout that can run w/o Giant locked

21 years agoUse kqueue(2) instead of poll(2) to wait for replies and timeouts
Ian Dowse [Tue, 15 Oct 2002 22:28:59 +0000 (22:28 +0000)]
Use kqueue(2) instead of poll(2) to wait for replies and timeouts
in the UDP RPC client code. As a side-effect, this fixes some bugs
that might prevent the RPC call from ever timing out for example
if the server keeps responding with the wrong xid. This could
probably be simplified further by using the EVFILT_TIMER filter.

21 years agoAdd HP ScanJet 2200C
Max Khon [Tue, 15 Oct 2002 21:59:39 +0000 (21:59 +0000)]
Add HP ScanJet 2200C

PR: 41375
Approved by: joe

21 years agoInclude "opt_kbd.h" since certain structures size depend on the value
Poul-Henning Kamp [Tue, 15 Oct 2002 21:35:01 +0000 (21:35 +0000)]
Include "opt_kbd.h" since certain structures size depend on the value
of KBDIO_DEBUG which may be defined in the kernel config (as it is in NOTES).

This kind of bug is a _really_ horribly thing as we end up with one bit
of code thinking a particular structure is 136 bytes and another that it
is only 112 bytes.

Ideally all places would remember to #include the right "opt_foo.h" file,
but I think in practice file containing the variable sized struct should
#include it explicitly as a precaution.

Detected by: FlexeLint

21 years agoReturn an error if the drive reports heads/sectors that do not make sense.
Nate Lawson [Tue, 15 Oct 2002 21:28:50 +0000 (21:28 +0000)]
Return an error if the drive reports heads/sectors that do not make sense.
This fixes a divide by zero in fdisk(8)

Reviewed by: phk

21 years agoPush most UFS ACL behavior behind a check for MNT_ACLS, permitting ACLs
Robert Watson [Tue, 15 Oct 2002 21:28:24 +0000 (21:28 +0000)]
Push most UFS ACL behavior behind a check for MNT_ACLS, permitting ACLs
to be administratively disabled as needed on UFS/UFS2 file systems.  This
also has the effect of preventing the slightly more expensive ACL code
from running on non-ACL file systems, avoiding storage allocation for
ACLs that may be read from disk.  MNT_ACLS may be set at mount-time
using mount -o acls, or implicitly by setting the FS_ACLS flag using
tunefs.  On UFS1, you may also have to configure ACL store.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoBump __FreeBSD_version for the Binutils 2.13 upgrade, which includes the
David E. O'Brien [Tue, 15 Oct 2002 21:26:53 +0000 (21:26 +0000)]
Bump __FreeBSD_version for the Binutils 2.13 upgrade, which includes the
new FreeBSD emulation, vec, and output format.

21 years agoSpell 'set' as 'cleared' where appropriate.
Robert Watson [Tue, 15 Oct 2002 21:23:22 +0000 (21:23 +0000)]
Spell 'set' as 'cleared' where appropriate.

21 years agoCorrect the definitions of SADB_* to be compatible with
Hajimu UMEMOTO [Tue, 15 Oct 2002 20:59:56 +0000 (20:59 +0000)]
Correct the definitions of SADB_* to be compatible with
RFC2407/IANA assignment.  This change breaks binary
compatibility.  So, you need to recompile IPsec related
applications.

21 years agoThe matcd(4) driver is gone, so eliminate the quasi-FAQ entry in the
Bruce A. Mah [Tue, 15 Oct 2002 20:54:41 +0000 (20:54 +0000)]
The matcd(4) driver is gone, so eliminate the quasi-FAQ entry in the
installation guide that refers to it.

21 years agoUpdate "Distribution Format" section for reality; XFree86 is installed
Bruce A. Mah [Tue, 15 Oct 2002 20:53:02 +0000 (20:53 +0000)]
Update "Distribution Format" section for reality; XFree86 is installed
as a regular package, not as a bunch of tarballs in a special
directory on the CDs.

PR: 43977
Submitted by: Brian Candler <B.Candler@pobox.com>

21 years agoDon't show the command line when doing "make lint".
Poul-Henning Kamp [Tue, 15 Oct 2002 20:49:58 +0000 (20:49 +0000)]
Don't show the command line when doing "make lint".

21 years agoBe consistent about marking functions static.
Poul-Henning Kamp [Tue, 15 Oct 2002 20:32:45 +0000 (20:32 +0000)]
Be consistent about marking functions static.

Found by: FlexeLint.

21 years agoIf the FS_MULTILABEL flag is set in a UFS or UFS2 superblock,
Robert Watson [Tue, 15 Oct 2002 20:00:06 +0000 (20:00 +0000)]
If the FS_MULTILABEL flag is set in a UFS or UFS2 superblock,
automatically set MNT_MULTILABEL in the mount flags.

If FS_ACLS is set in a UFS or UFS2 superblock, automatically
set MNT_ACLS in the mount flags.

If either of these flags is set, but the appropriate kernel option
to support the features associated with the flag isn't available,
then print a warning at mount-time.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoPlug a memory-leak.
Poul-Henning Kamp [Tue, 15 Oct 2002 18:58:38 +0000 (18:58 +0000)]
Plug a memory-leak.

"I think you're right" by: jake

21 years agoUse ; not , as statement separator in PDEBUG() macro.
Poul-Henning Kamp [Tue, 15 Oct 2002 18:56:13 +0000 (18:56 +0000)]
Use ; not , as statement separator in PDEBUG() macro.

Ignoring a NULL dev in device_set_ivars() sounds wrong, KASSERT it to
non-NULL instead.

Do the same for device_get_ivars() for reasons of symmetry, though
it probably would have yielded a panic anyway, this gives more precise
diagnostics.

Absentmindedly nodded OK to by: jhb

21 years agoPlug an infrequent (I think) memory leak.
Poul-Henning Kamp [Tue, 15 Oct 2002 18:51:02 +0000 (18:51 +0000)]
Plug an infrequent (I think) memory leak.

Spotted by: FlexeLint

21 years agoBe consistent about functions being static.
Poul-Henning Kamp [Tue, 15 Oct 2002 18:24:31 +0000 (18:24 +0000)]
Be consistent about functions being static.

Spotted by: FlexeLint

21 years agoConstification ? Yes, out that door, row on the left, one patch each.
Poul-Henning Kamp [Tue, 15 Oct 2002 18:21:53 +0000 (18:21 +0000)]
Constification ?  Yes, out that door, row on the left, one patch each.

Sponsored by: DARPA & NAI Labs

21 years agoTeach tunefs to print the ACL and multilabel flag information when
Robert Watson [Tue, 15 Oct 2002 18:14:05 +0000 (18:14 +0000)]
Teach tunefs to print the ACL and multilabel flag information when
inspecting a superblock.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoArgh. Put back setting of P_ADVLOCK for the F_WRLCK case that was
John Baldwin [Tue, 15 Oct 2002 18:10:13 +0000 (18:10 +0000)]
Argh.  Put back setting of P_ADVLOCK for the F_WRLCK case that was
accidentally lost in the previous revision.

Submitted by: bde
Pointy hat to: jhb

21 years agofix typo in comments (in preparation for MFC)
Max Khon [Tue, 15 Oct 2002 16:40:57 +0000 (16:40 +0000)]
fix typo in comments (in preparation for MFC)

Approved by: deischen

21 years agoAdd missing .Nm. Add FBSDID. fprintf -> warnx. Lowercase first letter inside
Philippe Charnier [Tue, 15 Oct 2002 16:28:21 +0000 (16:28 +0000)]
Add missing .Nm. Add FBSDID. fprintf -> warnx. Lowercase first letter inside
err().

21 years agoCorrect some of the style problems in this file:
Robert Watson [Tue, 15 Oct 2002 15:30:55 +0000 (15:30 +0000)]
Correct some of the style problems in this file:

I introduced a style problem when I sorted 'a' before 'A'; our
preferred order sorts 'A' first.  Correct.

Use .Cm instead of .Ar.

Submitted by: bde

21 years agoAdd information about the SDLT density codes.
Joerg Wunsch [Tue, 15 Oct 2002 12:53:29 +0000 (12:53 +0000)]
Add information about the SDLT density codes.

Submitted by: "Stewart MacLund" <sundie@lunaticfringe.org>
MFC after: 1 week

21 years agoThis commit was generated by cvs2svn to compensate for changes in r105151,
Ruslan Ermilov [Tue, 15 Oct 2002 10:42:07 +0000 (10:42 +0000)]
This commit was generated by cvs2svn to compensate for changes in r105151,
which included commits to RCS files with non-trunk default branches.

21 years agoFrom ChangeLog:
Ruslan Ermilov [Tue, 15 Oct 2002 10:42:07 +0000 (10:42 +0000)]
From ChangeLog:

: 2002-10-14  Werner LEMBERG  <wl@gnu.org>
:
:  * src/roff/troff/node.cc (troff_output_file::put_char_width,
:  troff_output_file::put_char, troff_output_file::fill_color,
:  troff_output_file::glyph_color): Handle case where color pointer
:  is null.

This paniced troff(1) attempting to format the vinum(4) manpage.

21 years agoAdd cross-references to the wide character counterparts of these functions.
Tim J. Robbins [Tue, 15 Oct 2002 10:11:53 +0000 (10:11 +0000)]
Add cross-references to the wide character counterparts of these functions.

21 years agoEliminate superfluous memory access in memcpy(3)/memmove(3).
Maxim Sobolev [Tue, 15 Oct 2002 09:50:07 +0000 (09:50 +0000)]
Eliminate superfluous memory access in memcpy(3)/memmove(3).

Obtained from: NetBSD
MFC after: 1 week

21 years agoRemove stray comma at the end of the See Also list.
Tim J. Robbins [Tue, 15 Oct 2002 09:49:54 +0000 (09:49 +0000)]
Remove stray comma at the end of the See Also list.

21 years agoFix kernel module loading on ia64. Cross-module function calls
Marcel Moolenaar [Tue, 15 Oct 2002 05:40:07 +0000 (05:40 +0000)]
Fix kernel module loading on ia64. Cross-module function calls
were improperly relocated due to faulty logic in lookup_fdesc()
in elf_machdep.c. The symbol index (symidx) was bogusly used for
load modules other than the one the relocation applied to. This
resulted in bogus bindings and consequently runtime failures.

The fix is to use the symbol index only for the module being
relocated and to use the symbol name for look-ups in the
modules in the dependent list. As such, we need a function to
return the symbol name given the linker file and symbol index.

21 years agoEnable mpt && ISP_TARGET_MODE in isp (for Lint purposes)
Matt Jacob [Tue, 15 Oct 2002 04:38:40 +0000 (04:38 +0000)]
Enable mpt && ISP_TARGET_MODE in isp (for Lint purposes)

21 years agoFix previous commit: Don't cast integral types to pointers to
Marcel Moolenaar [Tue, 15 Oct 2002 01:50:09 +0000 (01:50 +0000)]
Fix previous commit: Don't cast integral types to pointers to
print them with %p. Cast to unsigned long and print with %#lx.

Discussed with: bde

21 years agoRestore pointer that was removed in 1.128. This wasn't a merge-o.
Peter Wemm [Tue, 15 Oct 2002 01:36:45 +0000 (01:36 +0000)]
Restore pointer that was removed in 1.128.  This wasn't a merge-o.

21 years agoSeveral commit logs point here saying "More details on the new
Garance A Drosehn [Tue, 15 Oct 2002 01:00:14 +0000 (01:00 +0000)]
Several commit logs point here saying "More details on the new
'lpc topq' are in the log message for revision 1.2 of lpc/movejobs.c".

However, those details were mistakenly force-committed as revision 1.2
of common_source/matchjobs.c.  This forced-commit is just to add the
expected revision 1.2 here, as a pointer to the actual details.  See
http://cvsweb.freebsd.org/src/usr.sbin/lpr/common_source/matchjobs.c ,
revision 1.2.

21 years agoMove details of key_t to <sys/_types.h>, so that it can be typedef'd
Mike Barcroft [Tue, 15 Oct 2002 00:40:52 +0000 (00:40 +0000)]
Move details of key_t to <sys/_types.h>, so that it can be typedef'd
in more than just <sys/types.h>.

21 years ago- Add a new global mutex 'ppeers_lock' to protect the p_peers list of
John Baldwin [Tue, 15 Oct 2002 00:14:32 +0000 (00:14 +0000)]
- Add a new global mutex 'ppeers_lock' to protect the p_peers list of
  processes forked with RFTHREAD.
- Use a goto to a label for common code when exiting from fork1() in case
  of an error.
- Move the RFTHREAD linkage setup code later in fork since the ppeers_lock
  cannot be locked while holding a proc lock.  Handle the race of a task
  leader exiting and killing its peers while a peer is forking a new child.
  In that case, go ahead and let the peer process proceed normally as the
  parent is about to kill it.  However, the task leader may have already
  gone to sleep to wait for the peers to die, so the new child process may
  not receive a SIGKILL from the task leader.  Rather than try to destruct
  the new child process, just go ahead and send it a SIGKILL directly and
  add it to the p_peers list.  This ensures that the task leader will wait
  until both the peer process doing the fork() and the new child process
  have received their KILL signals and exited.

Discussed with: truckman (earlier versions)

21 years agoRemove the leaderp variable and just access p_leader directly. The
John Baldwin [Tue, 15 Oct 2002 00:03:40 +0000 (00:03 +0000)]
Remove the leaderp variable and just access p_leader directly.  The
p_leader field is not protected by the proc lock but is only set during
fork1() by the parent process and never changes.

21 years ago- Use __BUS_ACCESSOR() to define the ISA ivar accessor functions instead of
John Baldwin [Tue, 15 Oct 2002 00:02:51 +0000 (00:02 +0000)]
- Use __BUS_ACCESSOR() to define the ISA ivar accessor functions instead of
  homerolling our own version.
- Rename the enum for memsize from ISA_IVAR_MSIZE to ISA_IVAR_MEMSIZE
  since using 'MSIZE' in the macro invocation of ISA_ACCESSOR() conflicts
  with the 'MSIZE' kernel option.  The accessor function is still
  isa_get_msize().

21 years agoThe a.out md_coredump stuff isn't referenced anywhere anymore, and
Peter Wemm [Tue, 15 Oct 2002 00:02:50 +0000 (00:02 +0000)]
The a.out md_coredump stuff isn't referenced anywhere anymore, and
hasn't been filled in for ages..  Nuked.

21 years agoTurn off the premature locking in xl. The driver tries to use the mutexes
Peter Wemm [Mon, 14 Oct 2002 23:58:46 +0000 (23:58 +0000)]
Turn off the premature locking in xl.  The driver tries to use the mutexes
as spl replacements, but you cant sleep while holding mutexes.

This change has been made on many other drivers.

21 years agoWhen reading or writing the extended attributes of a special device
Kirk McKusick [Mon, 14 Oct 2002 23:18:09 +0000 (23:18 +0000)]
When reading or writing the extended attributes of a special device
or fifo in UFS2, the normal ufs_strategy routine needs to be used
rather than the spec_strategy or fifo_strategy routine. Thus the
ffsext_strategy routine is interposed in the ffs_vnops vectors for
special devices and fifo's to pick off this special case. Otherwise
it simply falls through to the usual spec_strategy or fifo_strategy
routine.

Submitted by: Robert Watson <rwatson@FreeBSD.org>
Sponsored by: DARPA & NAI Labs.

21 years agoPut function return types on a line by themselves.
Alfred Perlstein [Mon, 14 Oct 2002 22:31:52 +0000 (22:31 +0000)]
Put function return types on a line by themselves.
Cleanup my earlier de-__P sweep and remove whitespace between function
names and paramters.

21 years agoLINT related fixes (as noticed by bde)
Matt Jacob [Mon, 14 Oct 2002 22:13:51 +0000 (22:13 +0000)]
LINT related fixes (as noticed by bde)

21 years agoRemove a bogus local variable.
Poul-Henning Kamp [Mon, 14 Oct 2002 21:22:23 +0000 (21:22 +0000)]
Remove a bogus local variable.

Sponsored by: DARPA & NAI Labs.

21 years agoRemove a KASSERT I added in 1.73 to catch uninitialized pipes.
Alfred Perlstein [Mon, 14 Oct 2002 21:15:04 +0000 (21:15 +0000)]
Remove a KASSERT I added in 1.73 to catch uninitialized pipes.

It must be removed because it is done without the pipe being locked
via pipelock() and therefore is vulnerable to races with pipespace()
erroneously triggering it by temporarily zero'ing out the structure
backing the pipe.

It looks as if this assertion is not needed because all manipulation
of the data changed by pipespace() _is_ protected by pipelock().

Reported by: kris, mckusick

21 years agoRemove root from the 'guest' group: missed in a previous pass.
Robert Watson [Mon, 14 Oct 2002 20:55:49 +0000 (20:55 +0000)]
Remove root from the 'guest' group: missed in a previous pass.

Spotted by: jhb

21 years agoo Add typedefs for gid_t, key_t, mode_t, and uid_t.
Mike Barcroft [Mon, 14 Oct 2002 20:50:41 +0000 (20:50 +0000)]
o Add typedefs for gid_t, key_t, mode_t, and uid_t.
o Add comment about ipc_perm's members.
o Add some visibility conditionals for non-standard extentions.
o Remove a comment about ftok() not belonging.

21 years agoDid you ever notice how stupid bugs show up much clearer
Julian Elischer [Mon, 14 Oct 2002 20:43:02 +0000 (20:43 +0000)]
Did you ever notice how stupid bugs show up much clearer
when you see them in a commit message?

21 years agoCorrect visibility conditionals for memccpy(), strdup(), and
Mike Barcroft [Mon, 14 Oct 2002 20:38:40 +0000 (20:38 +0000)]
Correct visibility conditionals for memccpy(), strdup(), and
strtok_r().  Merge conditional prototypes into the regular block.

21 years agoTidy up the scheduler's code for changing the priority of a thread.
Julian Elischer [Mon, 14 Oct 2002 20:34:31 +0000 (20:34 +0000)]
Tidy up the scheduler's code for changing the priority of a thread.
Logically pretty much a NOP.

21 years agoRemove old useless debugging code
Julian Elischer [Mon, 14 Oct 2002 20:31:54 +0000 (20:31 +0000)]
Remove old useless debugging code

21 years agoupdate to better reflect reality:
Sam Leffler [Mon, 14 Oct 2002 20:23:41 +0000 (20:23 +0000)]
update to better reflect reality:

o describe additional argument in driver callbacks
o describe flow-control mechanism for processing crypto requests
o remove old cruft
o remove openbsd-specific cruft
o fixup some references
o yada yada ...

21 years agoMoved geom class initialization to SI_SUB_DRIVERS from SI_SUB_PSEUDO.
Jake Burkholder [Mon, 14 Oct 2002 20:20:17 +0000 (20:20 +0000)]
Moved geom class initialization to SI_SUB_DRIVERS from SI_SUB_PSEUDO.
This fixes mounting root from md(4) which calls disk_create() early.

21 years agoFix two memory leaks in error conditions involving the UFS ACL code:
Robert Watson [Mon, 14 Oct 2002 19:55:49 +0000 (19:55 +0000)]
Fix two memory leaks in error conditions involving the UFS ACL code:
if failures occur, make sure that we release both the default ACL
and access ACL storage during new object creation.

Spotted by: phk and his pet flexelint
Sponsored by: DARPA, Network Associates Laboratories

21 years agoWhen scanning the freelist looking for candidate vnodes to recycle,
Kirk McKusick [Mon, 14 Oct 2002 19:54:39 +0000 (19:54 +0000)]
When scanning the freelist looking for candidate vnodes to recycle,
be sure to exit the loop with vp == NULL if no candidates are found.
Formerly, this bug would cause the last vnode inspected to be used,
even if it was not available. The result was a panic "vn_finished_write:
neg cnt".

Sponsored by: DARPA & NAI Labs.

21 years agoIntroduce -a [enable|disable] and -l [enable|disable] flags to the tunefs
Robert Watson [Mon, 14 Oct 2002 19:52:12 +0000 (19:52 +0000)]
Introduce -a [enable|disable] and -l [enable|disable] flags to the tunefs
command, permitting it to set FS_ACLS and FS_MULTILABEL administrative
flags on UFS file systems.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoUnconditionally reset vp->v_vnlock back to the default in the
Kirk McKusick [Mon, 14 Oct 2002 19:44:51 +0000 (19:44 +0000)]
Unconditionally reset vp->v_vnlock back to the default in the
vclean() function (e.g., vp->v_vnlock = &vp->v_lock) rather
than requiring filesystems that use alternate locks to do so
in their vop_reclaim functions. This change is a further cleanup
of the vop_stdlock interface.

Submitted by: Poul-Henning Kamp <phk@critter.freebsd.dk>
Sponsored by: DARPA & NAI Labs.

21 years agoTeach mount(8) about MNT_ACLS for the purposes of mount options and
Robert Watson [Mon, 14 Oct 2002 19:40:00 +0000 (19:40 +0000)]
Teach mount(8) about MNT_ACLS for the purposes of mount options and
mount option printing.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoAdd a knob to turn on and off the CMPXCHG instruction on > i386 IA32 systems.
Jim Pirzyk [Mon, 14 Oct 2002 19:33:12 +0000 (19:33 +0000)]
Add a knob to turn on and off the CMPXCHG instruction on > i386 IA32 systems.
This is most beneficial for vmware client os installs.

Reviewed by: jmallet, iedowse, tlambert2@mindspring.com
MFC After: never, -STABLE does not currently use this instruction

21 years agoAllow firewire, sbp (SCSI over firewire), and fwe (non-standard raw
John Baldwin [Mon, 14 Oct 2002 19:21:44 +0000 (19:21 +0000)]
Allow firewire, sbp (SCSI over firewire), and fwe (non-standard raw
ethernet over firewire) to be statically compiled into the kernel as
devices.