]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoRemove files that are not part of GCC distribution anymore.
Alexander Kabaev [Sat, 19 May 2007 02:34:18 +0000 (02:34 +0000)]
Remove files that are not part of GCC distribution anymore.

17 years agoUpdate configuration files for GCC 4.2.
Alexander Kabaev [Sat, 19 May 2007 02:30:20 +0000 (02:30 +0000)]
Update configuration files for GCC 4.2.

17 years agoMerge local FreeBSD changes for -mno-align-long-strings.
Alexander Kabaev [Sat, 19 May 2007 02:26:26 +0000 (02:26 +0000)]
Merge local FreeBSD changes for -mno-align-long-strings.

17 years agoMerge local FreeBSD changes.
Alexander Kabaev [Sat, 19 May 2007 02:23:49 +0000 (02:23 +0000)]
Merge local FreeBSD changes.

17 years agoIn FreeBSD -O0 -O1 optimize alignment for time, not size.
Alexander Kabaev [Sat, 19 May 2007 02:22:16 +0000 (02:22 +0000)]
In FreeBSD -O0 -O1 optimize alignment for time, not size.

17 years agoMerge local changes for -mprofiler-epilogue support.
Alexander Kabaev [Sat, 19 May 2007 02:18:38 +0000 (02:18 +0000)]
Merge local changes for -mprofiler-epilogue support.

17 years agoMerge local FreeBSD changes to support -fformat-extensions.
Alexander Kabaev [Sat, 19 May 2007 02:16:45 +0000 (02:16 +0000)]
Merge local FreeBSD changes to support -fformat-extensions.

17 years agoAdd FreeBSD suffix to GCC version string.
Alexander Kabaev [Sat, 19 May 2007 02:14:13 +0000 (02:14 +0000)]
Add FreeBSD suffix to GCC version string.

17 years agoFreeBSD uses unchanged versions of this files.
Alexander Kabaev [Sat, 19 May 2007 02:12:21 +0000 (02:12 +0000)]
FreeBSD uses unchanged versions of this files.

17 years agoFix a mbuf leak where sc_start fails or the protocol is none.
Andrew Thompson [Sat, 19 May 2007 01:40:18 +0000 (01:40 +0000)]
Fix a mbuf leak where sc_start fails or the protocol is none.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r169695,
Alexander Kabaev [Sat, 19 May 2007 01:27:20 +0000 (01:27 +0000)]
This commit was generated by cvs2svn to compensate for changes in r169695,
which included commits to RCS files with non-trunk default branches.

17 years agoGCC 4.2.0 release miscellaneous support libraries.
Alexander Kabaev [Sat, 19 May 2007 01:27:20 +0000 (01:27 +0000)]
GCC 4.2.0 release miscellaneous support libraries.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r169693,
Alexander Kabaev [Sat, 19 May 2007 01:25:45 +0000 (01:25 +0000)]
This commit was generated by cvs2svn to compensate for changes in r169693,
which included commits to RCS files with non-trunk default branches.

17 years agoGCC 4.2.0 release Objective C runtime support code.
Alexander Kabaev [Sat, 19 May 2007 01:25:45 +0000 (01:25 +0000)]
GCC 4.2.0 release Objective C runtime support code.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r169691,
Alexander Kabaev [Sat, 19 May 2007 01:25:07 +0000 (01:25 +0000)]
This commit was generated by cvs2svn to compensate for changes in r169691,
which included commits to RCS files with non-trunk default branches.

17 years agoGCC 4.2.0 release C++ standard library and runtime support code.
Alexander Kabaev [Sat, 19 May 2007 01:25:07 +0000 (01:25 +0000)]
GCC 4.2.0 release C++ standard library and runtime support code.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r169689,
Alexander Kabaev [Sat, 19 May 2007 01:19:51 +0000 (01:19 +0000)]
This commit was generated by cvs2svn to compensate for changes in r169689,
which included commits to RCS files with non-trunk default branches.

17 years agoGCC 4.2.0 release.
Alexander Kabaev [Sat, 19 May 2007 01:19:51 +0000 (01:19 +0000)]
GCC 4.2.0 release.

17 years agoFix locking assert where we should hold the reader lock.
Andrew Thompson [Fri, 18 May 2007 23:38:35 +0000 (23:38 +0000)]
Fix locking assert where we should hold the reader lock.

17 years agoZero scsi_readcapacity allocations so we can really
Matt Jacob [Fri, 18 May 2007 21:58:07 +0000 (21:58 +0000)]
Zero scsi_readcapacity allocations so we can really
tell if there has been data returned (for some broken
residual checking points).

MFC after: 3 days

17 years agoBe more restrictive with segment validity checks in syncache_expand()
Andre Oppermann [Fri, 18 May 2007 21:42:25 +0000 (21:42 +0000)]
Be more restrictive with segment validity checks in syncache_expand()
and log check failures to syslog at LOG_DEBUG level.

Always prefill the sc->sc_ts field to use it in the checks.

17 years agoo Add syslog logging under LOG_DEBUG to various failures caused by
Andre Oppermann [Fri, 18 May 2007 21:13:01 +0000 (21:13 +0000)]
o Add syslog logging under LOG_DEBUG to various failures caused by
  bogus segments
o Add more KASSERT()s
o Update comments

17 years agoRemove unnecessary assignment.
Robert Watson [Fri, 18 May 2007 21:10:08 +0000 (21:10 +0000)]
Remove unnecessary assignment.

CID: 2227
Found with: Coverity Prevent(tm)

17 years agoAdd tcp_log_addrs() function to generate and standardized TCP log line
Andre Oppermann [Fri, 18 May 2007 19:58:37 +0000 (19:58 +0000)]
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.

It is IPv4 and IPv6 aware creates a line in the following format:

 "TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"

A "\n" is not included at the end.  The caller is supposed to add
further information after the standard tcp log header.

The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use.  All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.

Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.

Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.

tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
    void *ip6hdr)

Usage example:

 struct ip *ip;
 char *tcplog;

 if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
    tcplog, __func__);
free(s, M_TCPLOG);
 }

17 years agoFix statistical accounting for bytes and packets during sack retransmits.
John Baldwin [Fri, 18 May 2007 19:56:24 +0000 (19:56 +0000)]
Fix statistical accounting for bytes and packets during sack retransmits.

MFC after: 1 week
Submitted by: mohans

17 years agoIn nfs_down(), if rep can be NULL, which we test for, then we should
Robert Watson [Fri, 18 May 2007 19:34:54 +0000 (19:34 +0000)]
In nfs_down(), if rep can be NULL, which we test for, then we should
lock and unlock conditionally, not just set the flag on it conditionally.
In practice, this bug couldn't manifest, as in the current revision of
the code, no callers pass a NULL rep.

CID: 1416
Found with: Coverity Prevent(tm)

17 years agoThis file was originally added to document the transition of the
Bruce A. Mah [Fri, 18 May 2007 17:17:21 +0000 (17:17 +0000)]
This file was originally added to document the transition of the
release documentation from *.TXT files to the RELNOTESng *.sgml
files.  I'm pretty sure that, five years later, nobody is wondering
where the original *.TXT files went to in our source tree, so
remove this file to declutter the source tree a bit.

17 years agoMove nss_compat symbols from r1.11 to private section of symbol map.
Jonathan Chen [Fri, 18 May 2007 15:41:54 +0000 (15:41 +0000)]
Move nss_compat symbols from r1.11 to private section of symbol map.

Suggested by: deischen

17 years agoFix build with NETGRAPH_MPPC_COMPRESSION but without NETGRAPH_MPPC_ENCRYPTION.
Alexander Motin [Fri, 18 May 2007 15:28:01 +0000 (15:28 +0000)]
Fix build with NETGRAPH_MPPC_COMPRESSION but without NETGRAPH_MPPC_ENCRYPTION.

Approved by: glebius (mentor)

17 years agoHelp ng_fec deal with multicast addresses.
David Malone [Fri, 18 May 2007 15:05:49 +0000 (15:05 +0000)]
Help ng_fec deal with multicast addresses.

While ng_fec called the ioctl to let interfaces in the bundle know
the list of multicast addresses had changed, it never actually
updated that list on the interfaces in the bundle. Consequently,
the multicast filters could be programmed incorrectly.

if_lagg does this correctly, by maintaining a list of addresses
that it has added to interfaces in the bundle. This commit basically
takes the if_lagg code and adds it to ng_fec.

A version of this patch for RELENG_6 has fixed some problems with
IPv6 ND over ng_fec. This is probably the problem in PR 107523.

PR: 107523
Tested by: Rob Gallagher <robert.gallagher@heanet.ie>
Obtained from: if_lagg
MFC after: 3 weeks

17 years agoFix a comment.
John Baldwin [Fri, 18 May 2007 15:05:41 +0000 (15:05 +0000)]
Fix a comment.

17 years agoMove lock_profile_object_{init,destroy}() into lock_{init,destroy}().
John Baldwin [Fri, 18 May 2007 15:04:59 +0000 (15:04 +0000)]
Move lock_profile_object_{init,destroy}() into lock_{init,destroy}().

17 years agoAdd an example which shows how mdconfig(8) can be used
Giorgos Keramidas [Fri, 18 May 2007 13:57:17 +0000 (13:57 +0000)]
Add an example which shows how mdconfig(8) can be used
to mount an ISO 9660 CD image file.

PR: 112691
Submitted by: Warren Block, wblock at wonkity.com
MFC after: 3 days

17 years agoSince renaming of vop_lock to _vop_lock, pre- and post-condition
Konstantin Belousov [Fri, 18 May 2007 13:02:13 +0000 (13:02 +0000)]
Since renaming of vop_lock to _vop_lock, pre- and post-condition
function calls are no more generated for vop_lock.
Rename _vop_lock to vop_lock1 to satisfy tools/vnode_if.awk assumption
about vop naming conventions. This restores pre/post-condition calls.

17 years agoAdd -U and -P options that allow the specification of the per-user
Diomidis Spinellis [Fri, 18 May 2007 12:36:10 +0000 (12:36 +0000)]
Add -U and -P options that allow the specification of the per-user
and per-process summary file location.
These make the program more flexible, and also make it possible to write
sane regression tests.

17 years agoFix a logic bug I re-introduced in my patch I sent to Daniel
Ruslan Ermilov [Fri, 18 May 2007 12:25:48 +0000 (12:25 +0000)]
Fix a logic bug I re-introduced in my patch I sent to Daniel
that would cause the selected shared threading library to be
overwritten with its 32-bit version on amd64.

PR: amd64/112509

17 years agoo Implement the stop_boot subroutine [1]. This subroutine can be used by
Mike Makonnen [Fri, 18 May 2007 12:04:41 +0000 (12:04 +0000)]
o Implement the stop_boot subroutine [1]. This subroutine can be used by
  scripts in rc.d to stop rc(8) from booting into multi-user mode when
  a critical or severe error condition is encountered.

o Modify scripts in etc/rc.d that already implemented this functionality
  independently.

o Document it.

[1] - This subroutine was implemented in FreeBSD in rc.d/fsck. I moved it
      to rc.subr(8). Our version differs slightly in that it takes an
      optional argument to stop the boot even if "autoboot" is not set.

Obtained from: NetBSD
MFC after: 2 weeks

17 years ago - define and use VMCNT_{GET,SET,ADD,SUB,PTR} macros for manipulating
Jeff Roberson [Fri, 18 May 2007 07:10:50 +0000 (07:10 +0000)]
 - define and use VMCNT_{GET,SET,ADD,SUB,PTR} macros for manipulating
   vmcnts.  This can be used to abstract away pcpu details but also changes
   to use atomics for all counters now.  This means sched lock is no longer
   responsible for protecting counts in the switch routines.

Contributed by: Attilio Rao <attilio@FreeBSD.org>

17 years ago - Convert turnstiles and sleepqueus to use UMA. This provides a modest
Jeff Roberson [Fri, 18 May 2007 06:32:24 +0000 (06:32 +0000)]
 - Convert turnstiles and sleepqueus to use UMA.  This provides a modest
   speedup and will be more useful after each gains a spinlock in the
   impending thread_lock() commit.
 - Move initialization and asserts into init/fini routines.  fini routines
   are only needed in the INVARIANTS case for now.

Submitted by: Attilio Rao <attilio@FreeBSD.org>
Tested by: kris, jeff

17 years ago__nss_compat_* from net/nss_compat.c are needed by ports/net/nss_ldap, so export...
Jonathan Chen [Fri, 18 May 2007 05:45:28 +0000 (05:45 +0000)]
__nss_compat_* from net/nss_compat.c are needed by ports/net/nss_ldap, so export them.

17 years ago- Disabled responding to NI queries from a global address by default as
JINMEI Tatuya [Thu, 17 May 2007 21:20:24 +0000 (21:20 +0000)]
- Disabled responding to NI queries from a global address by default as
  specified in RFC4620.  A new flag for icmp6_nodeinfo was added to enable the
  feature.
- Also cleaned up the code so that the semantics of the icmp6_nodeinfo
  flags is clearer (i.e., defined specific macro names instead of using
  hard-coded values).

Approved by: gnn (mentor)
MFC after: 1 week

17 years agoUpdate the man page to reflect that certain variables will be unset in
Christian S.J. Peron [Thu, 17 May 2007 19:14:25 +0000 (19:14 +0000)]
Update the man page to reflect that certain variables will be unset in
the case that the program is set-user-ID or set-group-ID. Add missing
annotations for LIBMAP and LIBMAP_DISABLE.

17 years agoIn the event a process is tainted (setuid/setgid binaries), un-set any
Christian S.J. Peron [Thu, 17 May 2007 18:00:27 +0000 (18:00 +0000)]
In the event a process is tainted (setuid/setgid binaries), un-set any
potentially dangerous environment variables all together. It should be
noted that the run-time linker will not honnor these environment variables
if the process is tainted currently. However, once a child of the tainted
process calls setuid(2), it's status as being tainted (as defined by
issetugid(2)) will be removed. This could be problematic because
subsequent activations of the run-time linker could honnor these
dangerous variables.

This is more of an anti foot-shot mechanism, there is nothing I am
aware of in base that does this, however there may be third party
utilities which do, and there is no real negative impact of clearing
these environment variables.

Discussed on: secteam
Reviewed by: cperciva
PR: kern/109836
MFC after: 2 weeks

17 years agoCorrect typo in comment
Christian S.J. Peron [Thu, 17 May 2007 16:03:14 +0000 (16:03 +0000)]
Correct typo in comment

17 years agoFix the construction of the gctl_req that got broken by my
Marcel Moolenaar [Thu, 17 May 2007 15:34:51 +0000 (15:34 +0000)]
Fix the construction of the gctl_req that got broken by my
previous commit and that introduced optional parameters.
Existing classes (like geli(8)) use empty strings by default
and expect the parameter to be passed to the kernel as such.
Also, the default value of a string argument can be NULL.
Fix both cases by making the optional parameter conditional
upon gc_argname being set and making sure to test for NULL
before dereferencing the pointer.

Reported by: brueffer@

17 years agoEliminate a micro-optimization that hasn't had any effect for 15+ years.
Peter Wemm [Thu, 17 May 2007 15:31:14 +0000 (15:31 +0000)]
Eliminate a micro-optimization that hasn't had any effect for 15+ years.

17 years agoFix small copy/paste mistake.
Alexander Motin [Thu, 17 May 2007 13:33:38 +0000 (13:33 +0000)]
Fix small copy/paste mistake.

17 years ago- Fixed 1-2-1 model to not worry about associd in sockopts
Randall Stewart [Thu, 17 May 2007 12:16:24 +0000 (12:16 +0000)]
- Fixed 1-2-1 model to not worry about associd in sockopts
- Fixed RTOinfo for bounding.
- Fixed connect() to return ECONNREFUSED when an ABORT is received.
- Added comments to direct Static Analysis not to look at some things
  it does not understand (comments are /* sa_ignore XXXXX */)
- Bind when colliding was broken, missing not_found = 1 before
  checking to see if the port was in use caused endless bind loop.
- Cookie life needs to be in milliseconds to conform to socket api.
- Cookie life is not supposed to change if its 0, On the assoc
  level set we changed it to 0 opps.
- Two more static analysis issues identified by the cisco
  tool. Null checks needed.
- An issue for sendfile(). Need to validate the correct
  input argument.
- When sending failed due to a no route to host, we leaked
  the mbuf chain failing to call m_freem().
- Fix #ifdef issue for getting hash block len when HAVE_SHA2 is NOT defined
Reviewed by: gnn

17 years agoo Use the --detach option to kdc(8) instead of using the shell
Mike Makonnen [Thu, 17 May 2007 11:33:08 +0000 (11:33 +0000)]
o Use the --detach option to kdc(8) instead of using the shell
  background operator '&'.

  PR: conf/102722

o No need to include $kerberos5_server_flags in $command_args as
  rc.subr(8) will take care of this.

17 years agoThe precmd routine does not need to check whether the command should be
Mike Makonnen [Thu, 17 May 2007 08:57:14 +0000 (08:57 +0000)]
The precmd routine does not need to check whether the command should be
"forced". If some pre-condition is not met, it should fail as it normally
does and rc.subr(8) will make the appropriate decision. Incidentally, the
previous behaviour had a bug where the "force" flag was respected only
when checking rc.conf(5) knobs. The flag was ignored when verifying the
rpcbind(8) dependency.

MFC after: 2 weeks

17 years agoTighten IP address check, prevent '..' from being passed
Xin LI [Thu, 17 May 2007 06:08:42 +0000 (06:08 +0000)]
Tighten IP address check, prevent '..' from being passed
from the check.

Submitted by: Ren Zhen <bg1tpt gmail com>
MFC after: 2 weeks

17 years agoRegen
Daniel Eischen [Thu, 17 May 2007 05:13:05 +0000 (05:13 +0000)]
Regen

17 years agoDon't export a kern.conftxt sysctl, except when INCLUDE_CONF_FILE is
Warner Losh [Thu, 17 May 2007 05:05:12 +0000 (05:05 +0000)]
Don't export a kern.conftxt sysctl, except when INCLUDE_CONF_FILE is
defined.  This restores the old behavior, and eliminates the
dependency on the kernconf.tmpl when INCLUDE_CONFIG_FILE isn't
included in the kernel config.  There were many people in the terminal
room that had almost, but not quite, up-to-date config files that this
helps.  I don't know if this is the result of skew among the cvsup
servers, or some other more subtle problem.  However, this fix should
work for any config of recent vintage (I tested with the latest, and
one before the recent changes, and eye-balled the intermediate
versions).

Reviewed by: the terminal room crew

17 years agoAdd a description for the WITHOUT_SYMVER option.
Daniel Eischen [Thu, 17 May 2007 05:03:24 +0000 (05:03 +0000)]
Add a description for the WITHOUT_SYMVER option.

17 years agoAllow DEFAULT_THREAD_LIB to be set from /etc/src.conf.
Daniel Eischen [Thu, 17 May 2007 04:54:35 +0000 (04:54 +0000)]
Allow DEFAULT_THREAD_LIB to be set from /etc/src.conf.

Submitted by: ru

17 years agoFix some problems that affect multiple file inclusion. Bruce found
Warner Losh [Thu, 17 May 2007 04:53:52 +0000 (04:53 +0000)]
Fix some problems that affect multiple file inclusion.  Bruce found
this bug and submitted these patches to dunstan@.  He sent them to me
to test, and I discovered they were needed for the atmel kernel config
files.  Since we were playing with them in the terminal room after the
developer's summit today, I thought I'd go ahead and commit them to
allow those folks that now have atmel hardware (thanks Andre) a chance
to try it out w/o my help.  Since dunstan@ is asleep right now, risk
stepping on his toes a little by going ahead and committing this
change.

Submitted by: dunstan@, bde@
Tested by: bde@

17 years agoThe HP Diva RMP3 uses BAR 0x14.
Marcel Moolenaar [Thu, 17 May 2007 04:07:19 +0000 (04:07 +0000)]
The HP Diva RMP3 uses BAR 0x14.

17 years agoendrpcent belongs with rpc name service db, not services.
Jonathan Chen [Thu, 17 May 2007 03:34:33 +0000 (03:34 +0000)]
endrpcent belongs with rpc name service db, not services.

MFC after: 1 week

17 years agofix memory leak from reading nsswitch.conf
Jonathan Chen [Thu, 17 May 2007 03:33:23 +0000 (03:33 +0000)]
fix memory leak from reading nsswitch.conf

MFC after: 1 week

17 years agoMake the serdev I/F non-optional. At least 3 drivers depend on it
Marcel Moolenaar [Thu, 17 May 2007 01:29:55 +0000 (01:29 +0000)]
Make the serdev I/F non-optional. At least 3 drivers depend on it
and they can be loaded as a module.

PR: kern/112706

17 years agoFix typo.
Juli Mallett [Thu, 17 May 2007 00:19:56 +0000 (00:19 +0000)]
Fix typo.

17 years agoo) Correct for missing whitespace.
Juli Mallett [Thu, 17 May 2007 00:18:01 +0000 (00:18 +0000)]
o) Correct for missing whitespace.
o) We don't need to check if ifp == stdin to give the filename, since we already
   know that ifn will be "stdin" if it is.

17 years agoCouple of changes, back down on last TSO change, instead make old
Jack F Vogel [Thu, 17 May 2007 00:14:03 +0000 (00:14 +0000)]
Couple of changes, back down on last TSO change, instead make old
adapter list still capable, but only PCI-E adapters are now enabled.
The user can enable older PCI-X or PCI adapters using ifconfig.
Secondly, Arthur Hartwig pointed out my MSI change was not working
correctly, changed to something that now does. Thanks Arthur.
There was also a fundamental bug in the 82575 MSIX code, the MSIX
registers had to be mapped, opps :)

Rubber-stamped by: Pdeuskar

17 years agoFix confusing misindentation of a closing-brace. (It goes with the switch, not
Juli Mallett [Thu, 17 May 2007 00:11:58 +0000 (00:11 +0000)]
Fix confusing misindentation of a closing-brace.  (It goes with the switch, not
with the while.)

17 years agoUnbreak IPv4 kernel build.
Oleg Bulyzhin [Thu, 17 May 2007 00:05:13 +0000 (00:05 +0000)]
Unbreak IPv4 kernel build.

17 years agoDon't completely skip pci_cfg_save() in the PCI nomatch routine if
John Baldwin [Wed, 16 May 2007 23:42:04 +0000 (23:42 +0000)]
Don't completely skip pci_cfg_save() in the PCI nomatch routine if
the power_nodriver tunable is off.  pci_cfg_save() already checks the
tunable internally, and no other callers of pci_cfg_save() check the
tunable.

Reviewed by: imp

17 years agoFix interrupt routing for some cards after the previous fix to clear BARs
John Baldwin [Wed, 16 May 2007 23:40:08 +0000 (23:40 +0000)]
Fix interrupt routing for some cards after the previous fix to clear BARs
in cardbus cards: update the saved copy of the PCIR_INTLINE register in
the PCI ivars in addition to the actual register.

Reviewed by: imp

17 years ago- Added TSO support
David Christensen [Wed, 16 May 2007 23:34:11 +0000 (23:34 +0000)]
- Added TSO support
- Updated firmware to latest release (v3.4.8) to fix TSO + jumbo frame lockup
- Added MSI (hw.bce.msi_enable) and TSO (hw.bce.tso_enable) sysctls
- Fixed kernel panic when MSI is used and module is unloaded
- Added several new debug routines
- Removed slack space for RX/TX chains since it only covers sloppy coding
- Fixed a potential problem when programming jumbo MTU size in hardware
- Various other comment changes

MFC after: 4 weeks

17 years agoBump G_LIB_VERSION to reflect the ABI change.
Marcel Moolenaar [Wed, 16 May 2007 23:32:40 +0000 (23:32 +0000)]
Bump G_LIB_VERSION to reflect the ABI change.

Pointed out by: pjd@

17 years agoDon't enable symbol versioning on ia64 for now. It causes
Marcel Moolenaar [Wed, 16 May 2007 23:24:15 +0000 (23:24 +0000)]
Don't enable symbol versioning on ia64 for now. It causes
symbol lookup failures that later result in null-pointer
dereferences. This needs looking into, but since we're
close to release it's possible that it's not resolved before
that time.

17 years agoWe don't need --export-dynamic for ld-elf.so.1, because it's a
Marcel Moolenaar [Wed, 16 May 2007 23:19:51 +0000 (23:19 +0000)]
We don't need --export-dynamic for ld-elf.so.1, because it's a
shared object.

17 years agoWork around a vendor issue that was causing the builtin malloc to be
Mark Peek [Wed, 16 May 2007 21:22:38 +0000 (21:22 +0000)]
Work around a vendor issue that was causing the builtin malloc to be
used instead of the system malloc.

Submitted by: ume

17 years agoRemove leading spaces before tabs spotted thanks to silby using
Robert Watson [Wed, 16 May 2007 20:46:58 +0000 (20:46 +0000)]
Remove leading spaces before tabs spotted thanks to silby using
kwrite to read ip_input.c.

17 years agoGenerally migrate to ANSI function headers, and remove 'register' use.
Robert Watson [Wed, 16 May 2007 20:41:08 +0000 (20:41 +0000)]
Generally migrate to ANSI function headers, and remove 'register' use.

17 years ago-Fix so getoptinfo() is in conformance with socket api (it
Randall Stewart [Wed, 16 May 2007 20:23:39 +0000 (20:23 +0000)]
-Fix so getoptinfo() is in conformance with socket api (it
 can be used on any read socketopt).
-Clean up of sendmsg call and make it conformant when no syscall
 is available.

17 years ago Resolves symbols map so sctp_xxx() functions are included.
Randall Stewart [Wed, 16 May 2007 20:20:54 +0000 (20:20 +0000)]
 Resolves symbols map so sctp_xxx() functions are included.
PR: 112704
Submitted by: kou@cozmixng.org

17 years agoDon't use a subshell where it isn't needed.
Dag-Erling Smørgrav [Wed, 16 May 2007 20:16:06 +0000 (20:16 +0000)]
Don't use a subshell where it isn't needed.

Noticed by: John E Hein <jhein@timing.com>

17 years agoWrite 0 to all the bars in cardbus_cardbus_setup_regs. This is needed
Warner Losh [Wed, 16 May 2007 20:00:36 +0000 (20:00 +0000)]
Write 0 to all the bars in cardbus_cardbus_setup_regs.  This is needed
because on at least my dc based cards there's garbage in there.  The
recent changes in the resource code appears to have unmasked this
problem...  At least dc now probes/attaches better than it did before.

Also, we no longer need to write to the cfg for the other registers.

17 years agoUpdate the comments on if_alloc(), if_free(), if_free_type(), and
Brooks Davis [Wed, 16 May 2007 19:59:01 +0000 (19:59 +0000)]
Update the comments on if_alloc(), if_free(), if_free_type(), and
if_attach.

Remove a comment about pre-3.0 network drivers from if_attach().

Be a bit more consistant about whitespace near comments.

17 years agoo Don't hold lock over bus_teardown_intr when we're waiting for the
Warner Losh [Wed, 16 May 2007 19:53:49 +0000 (19:53 +0000)]
o Don't hold lock over bus_teardown_intr when we're waiting for the
  thread to die, since that can cause deadlock with the ISR (in theory).
o Minor cleanup
o comment nit

17 years agoBetter message on cbb resource alloc failure
Warner Losh [Wed, 16 May 2007 18:44:56 +0000 (18:44 +0000)]
Better message on cbb resource alloc failure

17 years agoChange PCIM_CIS_ASI_TUPLE to _CONFIG.
Warner Losh [Wed, 16 May 2007 18:42:38 +0000 (18:42 +0000)]
Change PCIM_CIS_ASI_TUPLE to _CONFIG.
Add PCI_MAX_BAR_0
minor style nit.
Add PCIM_CIS_CONFIG_MASK

17 years agoAdd some doxygen docs for pci_cfg_{save,restore}.
Warner Losh [Wed, 16 May 2007 18:41:42 +0000 (18:41 +0000)]
Add some doxygen docs for pci_cfg_{save,restore}.

17 years agoThe struct if_data members ifi_recvquota and ifi_xmitquota have been
Brooks Davis [Wed, 16 May 2007 18:37:37 +0000 (18:37 +0000)]
The struct if_data members ifi_recvquota and ifi_xmitquota have been
unused for ages.  Rename them to ifi_spare_char1 and ifi_spare_char2
respectively to indicate this face.

17 years agoRemove now unused stuff forgotten in the previous commit.
Andre Oppermann [Wed, 16 May 2007 17:55:22 +0000 (17:55 +0000)]
Remove now unused stuff forgotten in the previous commit.

17 years agoRevert config(8) version bump. It brings major pain for people working on
Wojciech A. Koszek [Wed, 16 May 2007 17:23:54 +0000 (17:23 +0000)]
Revert config(8) version bump. It brings major pain for people working on
different versions of FreeBSD source tree.

Old config(8) can now be used unless you want to use INCLUDE_CONFIG_FILE
option.

Approved by: imp
Reviewed by: imp

17 years agoCatch up to driver changes required for the mfip sub-driver.
Scott Long [Wed, 16 May 2007 17:23:36 +0000 (17:23 +0000)]
Catch up to driver changes required for the mfip sub-driver.

17 years agoAdd the 'mfip' sub-driver for gaining SCSI-passthrough access to devices
Scott Long [Wed, 16 May 2007 17:19:47 +0000 (17:19 +0000)]
Add the 'mfip' sub-driver for gaining SCSI-passthrough access to devices
on an MFI controller.

17 years agoMove TIME_WAIT related functions and timer handling from files
Andre Oppermann [Wed, 16 May 2007 17:14:25 +0000 (17:14 +0000)]
Move TIME_WAIT related functions and timer handling from files
other than repo copied tcp_subr.c into tcp_timewait.c#1.284:

 tcp_input.c#1.350 tcp_timewait() -> tcp_twcheck()

 tcp_timer.c#1.92 tcp_timer_2msl_reset() -> tcp_tw_2msl_reset()
 tcp_timer.c#1.92 tcp_timer_2msl_stop() -> tcp_tw_2msl_stop()
 tcp_timer.c#1.92 tcp_timer_2msl_tw() -> tcp_tw_2msl_scan()

This is a mechanical move with appropriate renames and making
them static if used only locally.

The tcp_tw_2msl_scan() cleanup function is still run from the
tcp_slowtimo() in tcp_timer.c.

17 years agoRelease the bus reference that is acquired when doing a CAMIOCOMMAND
Scott Long [Wed, 16 May 2007 16:57:21 +0000 (16:57 +0000)]
Release the bus reference that is acquired when doing a CAMIOCOMMAND
ioctl.  This should make SIM removal a bit more robust.

17 years agoRemove an assertion in xpt_print_path() that makes debugging code much more
Scott Long [Wed, 16 May 2007 16:55:50 +0000 (16:55 +0000)]
Remove an assertion in xpt_print_path() that makes debugging code much more
difficult.

17 years agoAdd a helper function for registering async callbacks. Besides
Scott Long [Wed, 16 May 2007 16:54:23 +0000 (16:54 +0000)]
Add a helper function for registering async callbacks.  Besides
eliminating a lot of duplicated code, this also fixes a locking edge case.

17 years agoHandle !INCLUDE_CONFIG_FILE entirely in the kernel. This should make some
Wojciech A. Koszek [Wed, 16 May 2007 16:08:04 +0000 (16:08 +0000)]
Handle !INCLUDE_CONFIG_FILE entirely in the kernel. This should make some
developers happy, since it will let them to use old config(8) with newer
kernels.

Reviewed by: imp
Approved by: imp

17 years agoAdd some missing '${BASEDIR}/'s. Prior to this commit, FreeBSD Update
Colin Percival [Wed, 16 May 2007 15:20:51 +0000 (15:20 +0000)]
Add some missing '${BASEDIR}/'s.  Prior to this commit, FreeBSD Update
operating with the "-b basedir" option would not correctly update files
which had flags set or were hardlinked.

Submitted by: Karsten Schmidt
Pointy hat to: cperciva
MFC after: 1 week

17 years agoStyle cleanup.
Alexander Motin [Wed, 16 May 2007 12:11:09 +0000 (12:11 +0000)]
Style cleanup.

Approved by: glebius (mentor)

17 years agoText reformatting.
Alexander Motin [Wed, 16 May 2007 12:00:22 +0000 (12:00 +0000)]
Text reformatting.

Approved by: glebius (mentor)

17 years agoWhen verifying the IPv4 UDP checksum, don't overwrite the checksum
David Malone [Wed, 16 May 2007 09:12:16 +0000 (09:12 +0000)]
When verifying the IPv4 UDP checksum, don't overwrite the checksum
value in the mbuf with the result of the calculation. Previously,
if we chose to return an ICMP message, the quoted UDP checksum bytes
would be different to what was sent.

PR: 112471
Submitted by: Matthew Luckie <mluckie@cs.waikato.ac.nz>
MFC after: 3 weeks

17 years agoExpose all of {check,delete}-old{,-dirs,-files,-libs}.
Dag-Erling Smørgrav [Wed, 16 May 2007 08:46:35 +0000 (08:46 +0000)]
Expose all of {check,delete}-old{,-dirs,-files,-libs}.

MFC after: 3 weeks

17 years agoRemove stray comment which broke delete-old-dirs.
Dag-Erling Smørgrav [Wed, 16 May 2007 08:39:46 +0000 (08:39 +0000)]
Remove stray comment which broke delete-old-dirs.

17 years agoPartially revert the check-old / delete-old modifications to clarify the
Dag-Erling Smørgrav [Wed, 16 May 2007 08:37:40 +0000 (08:37 +0000)]
Partially revert the check-old / delete-old modifications to clarify the
logic.

Apply similar modifications to {check,delete}-old-dirs, which I had
overlooked.