]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoThis commit was generated by cvs2svn to compensate for changes in r168988,
Gregory Neil Shapiro [Mon, 23 Apr 2007 22:11:27 +0000 (22:11 +0000)]
This commit was generated by cvs2svn to compensate for changes in r168988,
which included commits to RCS files with non-trunk default branches.

17 years agoBring in sendmail.org code from the future 8.14.2 release which restores
Gregory Neil Shapiro [Mon, 23 Apr 2007 22:11:27 +0000 (22:11 +0000)]
Bring in sendmail.org code from the future 8.14.2 release which restores
ABI compatibility to users of the libmilter.so shared library.

17 years agoMostly-cosmetic fixes in low-memory warning messages:
Bruce A. Mah [Mon, 23 Apr 2007 21:52:14 +0000 (21:52 +0000)]
Mostly-cosmetic fixes in low-memory warning messages:

o Fix linewrap issues.

o Fix two typos (s/Recomended/Recommended/ and s/tunning/tuning/)

o Remove a couple of extra instances of the word "of".

o Update names of kmem_size variables.

Approved by: pjd

17 years agoo Fix INP lock leak in the minttl case
Andre Oppermann [Mon, 23 Apr 2007 19:41:47 +0000 (19:41 +0000)]
o Fix INP lock leak in the minttl case
o Remove indirection in the decision of unlocking inp
o Further annotation of locking in tcp_input()

17 years agoFix old locking bugs which were revealed when pseudofs was made MPSAFE.
Dag-Erling Smørgrav [Mon, 23 Apr 2007 19:17:01 +0000 (19:17 +0000)]
Fix old locking bugs which were revealed when pseudofs was made MPSAFE.

Submitted by: tegge

17 years agoRemove obsolete headers. Move kernel specific headers under #ifdefs
Scott Long [Mon, 23 Apr 2007 18:01:44 +0000 (18:01 +0000)]
Remove obsolete headers.  Move kernel specific headers under #ifdefs

17 years agoGlobal xmit stats calculation fix.
Alexander Motin [Mon, 23 Apr 2007 15:25:14 +0000 (15:25 +0000)]
Global xmit stats calculation fix.

Approved by: glebius (mentor)

17 years agoAdd au_to_zonename.3 alias for au_token.3.
Robert Watson [Mon, 23 Apr 2007 14:46:59 +0000 (14:46 +0000)]
Add au_to_zonename.3 alias for au_token.3.

Obtained from: TrustedBSD Project

17 years agoAudit pathnames looked up in swapon(2) and swapoff(2).
Robert Watson [Mon, 23 Apr 2007 14:41:34 +0000 (14:41 +0000)]
Audit pathnames looked up in swapon(2) and swapoff(2).

MFC after: 2 weeks
Obtained from: TrustedBSD Project

17 years agoToo much diff reduction. 'cmd' has to be u_long.
Pawel Jakub Dawidek [Mon, 23 Apr 2007 13:47:49 +0000 (13:47 +0000)]
Too much diff reduction. 'cmd' has to be u_long.

Reported by: delphij

17 years agoRename mac*devfsdirent*() to mac*devfs*() to synchronize with SEDarwin,
Robert Watson [Mon, 23 Apr 2007 13:36:54 +0000 (13:36 +0000)]
Rename mac*devfsdirent*() to mac*devfs*() to synchronize with SEDarwin,
where similar data structures exist to support devfs and the MAC
Framework, but are named differently.

Obtained from: TrustedBSD Project
Sponsored by: SPARTA, Inc.

17 years agoApply variable name normalization to MAC policies: adopt global conventions
Robert Watson [Mon, 23 Apr 2007 13:15:23 +0000 (13:15 +0000)]
Apply variable name normalization to MAC policies: adopt global conventions
for the naming of variables associated with specific data structures.

Obtained from: TrustedBSD Project

17 years agoCorrect typo.
Pawel Jakub Dawidek [Mon, 23 Apr 2007 12:53:00 +0000 (12:53 +0000)]
Correct typo.

17 years agoMake it possible to specify an initial state for the LED.
Poul-Henning Kamp [Mon, 23 Apr 2007 12:42:15 +0000 (12:42 +0000)]
Make it possible to specify an initial state for the LED.

Requested by: Henrik Brix Andersen <henrik@brixandersen.dk>
PR: 112008

17 years agoAdd VLAN capability.
Poul-Henning Kamp [Mon, 23 Apr 2007 12:19:02 +0000 (12:19 +0000)]
Add VLAN capability.

Submitted by: Slawa Olhovchenkov <slw@zxy.spb.ru>

17 years agoImprove reporting in recoverdisk a good deal.
Poul-Henning Kamp [Mon, 23 Apr 2007 12:17:27 +0000 (12:17 +0000)]
Improve reporting in recoverdisk a good deal.

Submitted by: Ulrich Spoerlein <uspoerlein@gmail.com>
PR: 111630

17 years agoInsert explicit space between the output fields to prevent them
Yaroslav Tykhiy [Mon, 23 Apr 2007 11:43:22 +0000 (11:43 +0000)]
Insert explicit space between the output fields to prevent them
from running together when a field overflows.

MFC after: 1 week

17 years agoTurn off route header processing for now due to issues pointed out
George V. Neville-Neil [Mon, 23 Apr 2007 09:32:04 +0000 (09:32 +0000)]
Turn off route header processing for now due to issues pointed out
by Philippe Biondi and Arnaud Ebalard.  This is a temporary fix
until more discussion can be had on the exact risks involved in
allowing source routing in IPv6

Submitted by: itojun
Reviewed by: jinmei
MFC after: 1 day

17 years ago1)Make it possible for rpcbind(8) to bind TCP listening socket to an IP
Matteo Riondato [Mon, 23 Apr 2007 07:09:25 +0000 (07:09 +0000)]
1)Make it possible for rpcbind(8) to bind TCP listening socket to an IP
other than INADDR_ANY.

2) Add the -6 option to specify "IPv6 only".

Glanced at by: bms
Requested by: bms [2]
PR: bin/84494 [1]
Approved by: silence from maintainer (~2 weeks) [1]
MFC after: 2 weeks

17 years agoAdd synchronization. Eliminate the acquisition and release of Giant.
Alan Cox [Mon, 23 Apr 2007 06:12:24 +0000 (06:12 +0000)]
Add synchronization.  Eliminate the acquisition and release of Giant.

Reviewed by: tegge

17 years agomake dev.ath.N.ledpin have an immediate effect
Sam Leffler [Mon, 23 Apr 2007 05:57:06 +0000 (05:57 +0000)]
make dev.ath.N.ledpin have an immediate effect

PR: kern/111810
Submitted by: Henrik Brix Andersen <henrik@brixandersen.dk>
MFC after: 1 week

17 years agoo fix a buffer overflow in save_key() that occurs with 104-bit wep
Sam Leffler [Mon, 23 Apr 2007 05:51:18 +0000 (05:51 +0000)]
o fix a buffer overflow in save_key() that occurs with 104-bit wep
o make some variables parameters (frequency of crack, and maximum channel)
o try to spoof mac if association fails

Submitted by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>

17 years agoAdd a reference and lock the target thread when setting its name.
Daniel Eischen [Mon, 23 Apr 2007 03:36:14 +0000 (03:36 +0000)]
Add a reference and lock the target thread when setting its name.

Submitted by: davidxu (via libthr)

17 years agoWhen generating the version map file, order versions oldest
Daniel Eischen [Mon, 23 Apr 2007 03:24:33 +0000 (03:24 +0000)]
When generating the version map file, order versions oldest
first to make it easier for rtld to choose the oldest version
of a symbol.

Sumbitted by: kan

17 years agoMFp4: Reduce diff against vendor code:
Pawel Jakub Dawidek [Mon, 23 Apr 2007 00:52:07 +0000 (00:52 +0000)]
MFp4: Reduce diff against vendor code:
- Move FreeBSD-specific code to zfs_freebsd_*() functions in zfs_vnops.c
  and keep original functions as similar to vendor's code as possible.
- Add various includes back, now that we have them.

17 years agoFixes cut and paste bug using wrong pointer reference.
Randall Stewart [Mon, 23 Apr 2007 00:51:49 +0000 (00:51 +0000)]
Fixes cut and paste bug using wrong pointer reference.

17 years agoAdd back the original behavior of changing the entire directory path at
Nate Lawson [Sun, 22 Apr 2007 22:33:29 +0000 (22:33 +0000)]
Add back the original behavior of changing the entire directory path at
once (CWD a/b/c vs. 3 CWDs).  If an error occurs, we fall back to the default
method of a single CWD per directory element.  Since this is technically
a violation of the basic FTP RFC, this behavior is under a compile-time
option FTP_COMBINE_CWDS and is off by default.  It should work with most
Unix-based FTP daemons and can save latency.

MFC after: 2 weeks

17 years agoFix 'zpool status -v'. To get object number we should use ZFS_DIRENT_OBJ()
Pawel Jakub Dawidek [Sun, 22 Apr 2007 21:18:40 +0000 (21:18 +0000)]
Fix 'zpool status -v'. To get object number we should use ZFS_DIRENT_OBJ()
macro, as za_first_integer field also contains type. This should be fixed in
ZFS itself, but this bug is not visible on Solaris, because there, type is
not stored in za_first_integer. On the other hand it will be visible on
MacOS X.

Reported by: Barry Pederson <bp@barryp.org>

17 years agoFix st_rdev handling (implement it, actually).
Pawel Jakub Dawidek [Sun, 22 Apr 2007 21:16:15 +0000 (21:16 +0000)]
Fix st_rdev handling (implement it, actually).

Reported by: gj

17 years agoClement has stepped down from portmgr.
Erwin Lansing [Sun, 22 Apr 2007 21:11:46 +0000 (21:11 +0000)]
Clement has stepped down from portmgr.
Thanks for all your work!

Reminded by: brueffer

17 years agoWhen zfs dataset has jailed=on property, it won't be mounted with
Pawel Jakub Dawidek [Sun, 22 Apr 2007 20:55:08 +0000 (20:55 +0000)]
When zfs dataset has jailed=on property, it won't be mounted with
'zfs mount -a' from the main system - this is by design, as mountpoint
may be set to dangerous value. This all means, that such file system
has to be mounted from within a jail. To make it easier, reorganize
rc.d/zfs script so it can be used from within a jail.

17 years agoNormalize variable naming in the MAC Framework by adopting the normal
Robert Watson [Sun, 22 Apr 2007 19:55:56 +0000 (19:55 +0000)]
Normalize variable naming in the MAC Framework by adopting the normal
variable name conventions for arguments passed into the framework --
for example, name network interfaces 'ifp', sockets 'so', mounts 'mp',
mbufs 'm', processes 'p', etc, wherever possible.  Previously there
was significant variation in this regard.

Normalize copyright lists to ranges where sensible.

17 years agoIn the MAC Framework implementation, file systems have two per-mountpoint
Robert Watson [Sun, 22 Apr 2007 16:18:10 +0000 (16:18 +0000)]
In the MAC Framework implementation, file systems have two per-mountpoint
labels: the mount label (label of the mountpoint) and the fs label (label
of the file system).  In practice, policies appear to only ever use one,
and the distinction is not helpful.

Combine mnt_mntlabel and mnt_fslabel into a single mnt_label, and
eliminate extra machinery required to maintain the additional label.
Update policies to reflect removal of extra entry points and label.

Obtained from: TrustedBSD Project
Sponsored by: SPARTA, Inc.

17 years agoRemove the old software bit-banging MII interface, we started using
Poul-Henning Kamp [Sun, 22 Apr 2007 15:58:56 +0000 (15:58 +0000)]
Remove the old software bit-banging MII interface, we started using
the Rhines shiftregisters in four years ago (1.60).

17 years agoRemove further cobwebs: Two layers of pointless substructures.
Poul-Henning Kamp [Sun, 22 Apr 2007 15:48:29 +0000 (15:48 +0000)]
Remove further cobwebs:  Two layers of pointless substructures.

17 years agoRemove MAC Framework access control check entry points made redundant with
Robert Watson [Sun, 22 Apr 2007 15:31:22 +0000 (15:31 +0000)]
Remove MAC Framework access control check entry points made redundant with
the introduction of priv(9) and MAC Framework entry points for privilege
checking/granting.  These entry points exactly aligned with privileges and
provided no additional security context:

- mac_check_sysarch_ioperm()
- mac_check_kld_unload()
- mac_check_settime()
- mac_check_system_nfsd()

Add mpo_priv_check() implementations to Biba and LOMAC policies, which,
for each privilege, determine if they can be granted to processes
considered unprivileged by those two policies.  These mostly, but not
entirely, align with the set of privileges granted in jails.

Obtained from: TrustedBSD Project

17 years agoInitialize the physical next pointer in the tx descriptors when we
Poul-Henning Kamp [Sun, 22 Apr 2007 15:09:03 +0000 (15:09 +0000)]
Initialize the physical next pointer in the tx descriptors when we
initialize instead of in the start routine.

17 years agoMFP4: Enhancements and bug-fixes to pmcstat(8):
Joseph Koshy [Sun, 22 Apr 2007 15:00:39 +0000 (15:00 +0000)]
MFP4: Enhancements and bug-fixes to pmcstat(8):

 - The '-c' option now takes a comma-separated list of CPU
   numbers, or a literal '*' denoting all CPUs in the system.
   Subsequent system PMCs are allocated on the CPUs so specified.

   Change the default behaviour to allocate system PMCs on all CPUs,
   not just CPU 0.

   Update the manual page and add an example of how to use the new
   functionality.

 - Attach PMCs to a (commandline) child process more reliably.  This
   fixes a long standing bug in counting events incurred by short-lived
   processes.

17 years agoDon't rename fields with #define.
Poul-Henning Kamp [Sun, 22 Apr 2007 14:57:05 +0000 (14:57 +0000)]
Don't rename fields with #define.
Collapse two semantically identical structs.
Add missing vr_ prefix.

17 years agoFurther MAC test policy cleanup and enhancement:
Robert Watson [Sun, 22 Apr 2007 13:29:37 +0000 (13:29 +0000)]
Further MAC test policy cleanup and enhancement:

- Redistribute counter declarations to where they are used, rather than at
  the file header, so it's more clear where we do (and don't) have
  counters.

- Add many more counters, one per policy entry point, so that many
  individual access controls and object life cycle events are tracked.

- Perform counter increments for label destruction explicitly in entry
  point functions rather than in LABEL_DESTROY().

- Use LABEL_INIT() instead of SLOT_SET() directly in label init functions
  to be symmetric with destruction.

- Align counter names more carefully with entry point names.

- More constant and variable name normalization.

Obtained from: TrustedBSD Project

17 years agoRun if_vr(4) through FlexeLint and clean some of the cobwebs found.
Poul-Henning Kamp [Sun, 22 Apr 2007 12:55:36 +0000 (12:55 +0000)]
Run if_vr(4) through FlexeLint and clean some of the cobwebs found.

17 years agoMoves the PCB features and flags from sctp_pcb.h to
Randall Stewart [Sun, 22 Apr 2007 12:12:38 +0000 (12:12 +0000)]
Moves the PCB features and flags from sctp_pcb.h to
sctp.h so that netstat can access and display these
values.

17 years agoPerform overdue clean up mac_test policy:
Robert Watson [Sun, 22 Apr 2007 11:35:15 +0000 (11:35 +0000)]
Perform overdue clean up mac_test policy:

- Add a more detailed comment describing the mac_test policy.

- Add COUNTER_DECL() and COUNTER_INC() macros to declare and manage
  various test counters, reducing the verbosity of the test policy
  quite a bit.

- Add LABEL_CHECK() macro to abbreviate normal validation of labels.
  Unlike the previous check macros, this checks for a NULL label and
  doesn't test NULL labels.  This means that optionally passed labels
  will now be handled automatically, although in the case of optional
  credentials, NULL-checks are still required.

- Add LABEL_DESTROY() macro to abbreviate the handling of label
  validation and tear-down.

- Add LABEL_NOTFREE() macro to abbreviate check for non-free labels.

- Normalize the names of counters, magic values.

- Remove unused policy "enabled" flag.

Obtained from: TrustedBSD Project

17 years ago- Somehow the disable fragment option got lost. We could
Randall Stewart [Sun, 22 Apr 2007 11:06:27 +0000 (11:06 +0000)]
-  Somehow the disable fragment option got lost. We could
   set/clear it but would not do it. Now we will.
-  Moved to latest socket api for extended sndrcv info struct.
-  Moved to support all new levels of fragment interleave.

17 years agoNow that we're MPSAFE, tell namei() to acquire Giant if necessary.
Dag-Erling Smørgrav [Sun, 22 Apr 2007 08:41:52 +0000 (08:41 +0000)]
Now that we're MPSAFE, tell namei() to acquire Giant if necessary.

17 years agoAdd some notes clarifying usage and a couple of known bugs
Kris Kennaway [Sun, 22 Apr 2007 06:20:12 +0000 (06:20 +0000)]
Add some notes clarifying usage and a couple of known bugs

17 years agoTrace my mentor lineage as far back as I can determine.
Tom McLaughlin [Sun, 22 Apr 2007 02:36:08 +0000 (02:36 +0000)]
Trace my mentor lineage as far back as I can determine.

Not really sure how to handle committers who no longer have a ports
commit bit but are still active in other repos.  Maybe a new node
definition? *shrug*

17 years agoFix sorting in previous commit.
Colin Percival [Sun, 22 Apr 2007 00:44:51 +0000 (00:44 +0000)]
Fix sorting in previous commit.

Pointed out by: brueffer
Pointy hat to: cperciva

17 years agoAdd myself. Edwin was my "mentor", even though he released me from
Colin Percival [Sun, 22 Apr 2007 00:12:10 +0000 (00:12 +0000)]
Add myself.  Edwin was my "mentor", even though he released me from
mentorship before approving a single commit.

17 years agomezz was my co-mentor
Tom McLaughlin [Sat, 21 Apr 2007 23:47:09 +0000 (23:47 +0000)]
mezz was my co-mentor

17 years agoAdd xride who was mentored by me and co-mentored by garga
Tom McLaughlin [Sat, 21 Apr 2007 23:44:36 +0000 (23:44 +0000)]
Add xride who was mentored by me and co-mentored by garga

17 years agoBecause there are so many more partitions on pc98 than on wintel (16
Warner Losh [Sat, 21 Apr 2007 22:47:35 +0000 (22:47 +0000)]
Because there are so many more partitions on pc98 than on wintel (16
vs 4), supress all unused partition output unless -v is specified.
This makes operating on a 'typical' disk with one partition less
painful.  The 30 lines needed for the empty partitions no longer
scroll the useful information off the screen.  When the user requests
a specific partition, the unused information is not suppressed.

Also add the partition name to the -s output.

Initialize the partition name to 'FreeBSD' when -I is specified.

17 years agoAllow MAC policy modules to control access to audit configuration system
Robert Watson [Sat, 21 Apr 2007 22:08:48 +0000 (22:08 +0000)]
Allow MAC policy modules to control access to audit configuration system
calls.  Add MAC Framework entry points and MAC policy entry points for
audit(), auditctl(), auditon(), setaudit(), aud setauid().

MAC Framework entry points are only added for audit system calls where
additional argument context may be useful for policy decision-making; other
audit system calls without arguments may be controlled via the priv(9)
entry points.

Update various policy modules to implement audit-related checks, and in
some cases, other missing system-related checks.

Obtained from: TrustedBSD Project
Sponsored by: SPARTA, Inc.

17 years agoTeach netinet6 to use PRIV_NETINET_REUSEPORT.
Robert Watson [Sat, 21 Apr 2007 18:14:04 +0000 (18:14 +0000)]
Teach netinet6 to use PRIV_NETINET_REUSEPORT.

17 years agoAttempt to rationalize NFS privileges:
Robert Watson [Sat, 21 Apr 2007 18:11:19 +0000 (18:11 +0000)]
Attempt to rationalize NFS privileges:

- Replace PRIV_NFSD with PRIV_NFS_DAEMON, add PRIV_NFS_LOCKD.

- Use PRIV_NFS_DAEMON in the NFS server.

- In the NFS client, move the privilege check from nfslockdans(), which
  occurs every time a write is performed on /dev/nfslock, and instead do it
  in nfslock_open() just once.  This allows us to avoid checking the saved
  uid for root, and just use the effective on open.  Use PRIV_NFS_LOCKD.

17 years agoModify TLB invalidation handling.
Stephan Uphoff [Sat, 21 Apr 2007 14:17:30 +0000 (14:17 +0000)]
Modify TLB invalidation handling.

Reviewed by: alc@, peter@
MFC after: 1 week

17 years agoImprove sharenfs option handling, so it is possible to give hosts list.
Pawel Jakub Dawidek [Sat, 21 Apr 2007 13:17:23 +0000 (13:17 +0000)]
Improve sharenfs option handling, so it is possible to give hosts list.
Before the change the command above:

# zfs set sharenfs=freefall.freebsd.org,69.147.83.54 tank/foo

was translated to:

/tank/foo -freefall.freebsd.org -69.147.83.54

instead of:

/tank/foo freefall.freebsd.org 69.147.83.54

This commit corrects this.

17 years agoCorrect a sanity check.
Joseph Koshy [Sat, 21 Apr 2007 12:04:03 +0000 (12:04 +0000)]
Correct a sanity check.

17 years agoMFp4:
Pawel Jakub Dawidek [Sat, 21 Apr 2007 12:02:57 +0000 (12:02 +0000)]
MFp4:

@118370 Correct typo.

@118371 Integrate changes from vendor.

@118491 Show backtrace on unexpected code paths.

@118494 Integrate changes from vendor.

@118504 Fix sendfile(2). I had two ways of fixing it:
1. Fixing sendfile(2) itself to use VOP_GETPAGES() instead of
   hacking around with vn_rdwr(UIO_NOCOPY), which was suggested
   by ups.
2. Modify ZFS behaviour to handle this special case.

Although 1 is more correct, I've choosen 2, because hack from 1
have a side-effect of beeing faster - it reads ahead MAXBSIZE
bytes instead of reading page by page. This is not easy to implement
with VOP_GETPAGES(), at least not for me in this very moment.

Reported by: Andrey V. Elsukov <bu7cher@yandex.ru>

@118525 Reorganize the code to reduce diff.

@118526 This code path is expected. It is simply when file is opened with
O_FSYNC flag.

Reported by: kris
Reported by: Michal Suszko <dry@dry.pl>

17 years agoRegression tests for recent changes to inet6_rth_* family of functions
Mike Makonnen [Sat, 21 Apr 2007 11:23:33 +0000 (11:23 +0000)]
Regression tests for recent changes to inet6_rth_* family of functions
regarding RFC3542 compliance.

17 years agoUpdate release notes:
Hiroki Sato [Sat, 21 Apr 2007 03:45:18 +0000 (03:45 +0000)]
Update release notes:
- hw.pci.do_powerstate split into hw.pci.do_power_nodriver
and hw.pci.do_power_resume.

Pointed out by: pluknet at gmail.com

17 years agoAdd a missing link: if_edsc.4 -> edsc.4 .
Yaroslav Tykhiy [Sat, 21 Apr 2007 03:00:19 +0000 (03:00 +0000)]
Add a missing link: if_edsc.4 -> edsc.4 .
Network interface manpages should have such links.

17 years agoDon't forget to bump document date after changing the content.
Yaroslav Tykhiy [Sat, 21 Apr 2007 01:22:51 +0000 (01:22 +0000)]
Don't forget to bump document date after changing the content.

17 years agoChange the semantics of -i (in-place editing) so that it treats
Yaroslav Tykhiy [Sat, 21 Apr 2007 01:21:36 +0000 (01:21 +0000)]
Change the semantics of -i (in-place editing) so that it treats
each file independently from other files.  The new semantics are
desired in the most of practical cases, e.g.: delete lines 5-9
from each file.

Keep the previous semantics of -i under a new option, -I, which
uses a single continuous address space covering all files to edit
in-place -- they are too cool to just drop them.

Add regression tests for -i and -I.

Approved by: dds
Compared with: GNU sed
Discussed on: -hackers
MFC after: 2 weeks

17 years agoAdd support for specifying a minimal size for vm.kmem_size in the loader via
Stephane E. Potvin [Sat, 21 Apr 2007 01:14:48 +0000 (01:14 +0000)]
Add support for specifying a minimal size for vm.kmem_size in the loader via
vm.kmem_size_min. Useful when using ZFS to make sure that vm.kmem size will
be at least 256mb (for example) without forcing a particular value via vm.kmem_size.

Approved by: njl (mentor)
Reviewed by: alc

17 years agoBah, sorting alphabetically is hard.
Christian Brueffer [Sat, 21 Apr 2007 00:25:05 +0000 (00:25 +0000)]
Bah, sorting alphabetically is hard.

17 years agoAdd markus.
Christian Brueffer [Sat, 21 Apr 2007 00:18:37 +0000 (00:18 +0000)]
Add markus.

17 years agoRemove useless FILES section.
Christian Brueffer [Sat, 21 Apr 2007 00:05:35 +0000 (00:05 +0000)]
Remove useless FILES section.

MFC after: 3 days

17 years agoSpring cleaning: Remove worm(4) manpage, the driver was removed pre-3.0.
Christian Brueffer [Fri, 20 Apr 2007 23:47:30 +0000 (23:47 +0000)]
Spring cleaning: Remove worm(4) manpage, the driver was removed pre-3.0.

17 years agoDon't reinvent vm_page_grab().
Pawel Jakub Dawidek [Fri, 20 Apr 2007 19:49:20 +0000 (19:49 +0000)]
Don't reinvent vm_page_grab().

Reviewed by: ups

17 years agoTest sending 0 bytes.
Pawel Jakub Dawidek [Fri, 20 Apr 2007 19:01:42 +0000 (19:01 +0000)]
Test sending 0 bytes.

17 years agoFix length calculation.
Pawel Jakub Dawidek [Fri, 20 Apr 2007 19:00:43 +0000 (19:00 +0000)]
Fix length calculation.

17 years agoNew release notes:
Christian Brueffer [Fri, 20 Apr 2007 16:12:31 +0000 (16:12 +0000)]
New release notes:
- IPLware 3.33 support for pc98
- CAM MPSAFE
- ahc(4) and ahd(4) MPSAFE
- pseudofs(9) and consumers MPSAFE
- OpenBSM 1.0 alpha 14
- lastcomm -X flag
- ftpd(8) RFC2389 and RFC2640 support

Modified release notes:
- ncurses was updated from version 5.2-20020615

While here, moved the lagg(4) and XFS entries to the correct places.

17 years agoXref linsysfs(5).
Christian Brueffer [Fri, 20 Apr 2007 15:38:06 +0000 (15:38 +0000)]
Xref linsysfs(5).

MFC after: 3 days

17 years agoFix a memory leak in the uname/gname lookup cache.
Tim Kientzle [Fri, 20 Apr 2007 15:32:13 +0000 (15:32 +0000)]
Fix a memory leak in the uname/gname lookup cache.
Thanks to: VMiklos

17 years agoo Remove unncessary TOF_SIGLEN flag from struct tcpopt
Andre Oppermann [Fri, 20 Apr 2007 15:28:01 +0000 (15:28 +0000)]
o Remove unncessary TOF_SIGLEN flag from struct tcpopt
o Correctly set to->to_signature in tcp_dooptions()
o Update comments

17 years agoAdd more KASSERT's.
Andre Oppermann [Fri, 20 Apr 2007 15:21:29 +0000 (15:21 +0000)]
Add more KASSERT's.

17 years agoo Remove unused and redundant TCP option definitions
Andre Oppermann [Fri, 20 Apr 2007 15:08:09 +0000 (15:08 +0000)]
o Remove unused and redundant TCP option definitions
o Replace usage of MAX_TCPOPTLEN with the correctly constructed and
  derived MAX_TCPOPTLEN

17 years agoRemove bogus check for accept queue length and associated failure handling
Andre Oppermann [Fri, 20 Apr 2007 14:34:54 +0000 (14:34 +0000)]
Remove bogus check for accept queue length and associated failure handling
from the incoming SYN handling section of tcp_input().

Enforcement of the accept queue limits is done by sonewconn() after the
3WHS is completed.  It is not necessary to have an earlier check before a
connection request enters the SYN cache awaiting the full handshake.  It
rather limits the effectiveness of the syncache by preventing legit and
illegit connections from entering it and having them shaken out before we
hit the real limit which may have vanished by then.

Change return value of syncache_add() to void.  No status communication
is required.

17 years agoSimplifly syncache_expand() and clarify its semantics. Zero is returned
Andre Oppermann [Fri, 20 Apr 2007 13:51:34 +0000 (13:51 +0000)]
Simplifly syncache_expand() and clarify its semantics.  Zero is returned
when the ACK is invalid and doesn't belong to any registered connection,
either in syncache or through SYN cookies.  True but a NULL struct socket
is returned when the 3WHS completed but the socket could not be created
due to insufficient resources or limits reached.

For both cases an RST is sent back in tcp_input().

A logic error leading to a panic is fixed where syncache_expand() would
free the mbuf on socket allocation failure but tcp_input() later supplies
it to tcp_dropwithreset() to issue a RST to the peer.

Reported by: kris (the panic)

17 years agoOnly update TCP timestamp on SYN duplication if it is present on
Andre Oppermann [Fri, 20 Apr 2007 13:36:48 +0000 (13:36 +0000)]
Only update TCP timestamp on SYN duplication if it is present on
current SYN in syncache_add().  Otherwise disable timestamps.

17 years agoo Plug memory leak in syncache_add() on MAC label allocation failure.
Andre Oppermann [Fri, 20 Apr 2007 13:30:08 +0000 (13:30 +0000)]
o Plug memory leak in syncache_add() on MAC label allocation failure.
o Simplify code flow with 'done' goto label.
o Remove mbuf argument from syncache_respond().  It doesn't make use
  of it.

17 years agoRemove references to S/Key and list OPIE.
Tom Rhodes [Fri, 20 Apr 2007 09:08:20 +0000 (09:08 +0000)]
Remove references to S/Key and list OPIE.

17 years agoAdded m_tag_copy_chain() call to copy original outgoing packet tags to all of
Alexander Motin [Fri, 20 Apr 2007 08:44:40 +0000 (08:44 +0000)]
Added m_tag_copy_chain() call to copy original outgoing packet tags to all of
it's fragments.

Reviewed by: archie
Approved by: glebius (mentor)

17 years agoOptimized packet distribution plan for the equal links case. Do not
Alexander Motin [Fri, 20 Apr 2007 08:42:08 +0000 (08:42 +0000)]
Optimized packet distribution plan for the equal links case. Do not
split packet on fragments smaller then MP_MIN_FRAG_LEN to reduce total
overhead.

Reviewed by: archie
Approved by: glebius (mentor)

17 years ago- Changed sequence numbers processing to avoid incorrect timeout waiting
Alexander Motin [Fri, 20 Apr 2007 08:38:18 +0000 (08:38 +0000)]
- Changed sequence numbers processing to avoid incorrect timeout waiting
when one of links is inactive and have stale sequence number. To avoid
this sequence numbers of all links are getting updated on every
successful packet reassembling.
- ng_ppp_bump_mseq function created to simplify code.
- ng_ppp_frag_drop function separated from ng_ppp_frag_process to
simplify code.

Reviewed by: archie
Approved by: glebius (mentor)

17 years ago- Fixed mistakes in latency and xmitBytes calculation math
Alexander Motin [Fri, 20 Apr 2007 08:22:57 +0000 (08:22 +0000)]
- Fixed mistakes in latency and xmitBytes calculation math
which lead to ineffective multilink packet distribution plans.
- Changed bytesInQueue calculation math to have more precise information
about links utilization.
- Taken rough account of the link overhead. Better way to do it could be to
get exact overhead from user-level, but I have not done it to keep
binary compatibility.

Reviewed by: archie
Approved by: glebius (mentor)

17 years agoNanoBSD modifications:
Adrian Chadd [Fri, 20 Apr 2007 07:21:09 +0000 (07:21 +0000)]
NanoBSD modifications:

* Break out the boot0 loader selection into a variable - NANO_BOOTLOADER -
  so people like me with VGA consoles can override the default (which is
  to use boot0sio)
* Put the boot0 configuration options in NANO_BOOT0CFG in case you want
  to override the defaults.
* Modify nanobsd.8 to reflect the changes and hint the console default is
  serial.

MFC after: 2 weeks

17 years agoImprove logging when -dm is specified: if the node is considered
Max Khon [Fri, 20 Apr 2007 06:33:25 +0000 (06:33 +0000)]
Improve logging when -dm is specified: if the node is considered
out-of-date print not only "modified before source" message
but also the path of youngest source.

17 years agoWhen remaking makefiles check that mtime has actually changed.
Max Khon [Fri, 20 Apr 2007 06:25:45 +0000 (06:25 +0000)]
When remaking makefiles check that mtime has actually changed.

This fixes infinite restart in the following case:

Makefile: foo

foo: bar
do-something

Unlike GNU make, BSD make considers "Makefile" node as remade even
if "foo" is up-to-date and was not actually rebuilt.
GNU make does not consider nodes without commands as remade if child nodes
were not actually rebuilt.

Most probably, more proper fix would be to bring BSD make behaviour in-line
with GNU make but this would be more intrusive change.

17 years agoSchedule the ithread on the same cpu as the interrupt
Kip Macy [Fri, 20 Apr 2007 05:45:46 +0000 (05:45 +0000)]
Schedule the ithread on the same cpu as the interrupt

Tested by: kmacy
Submitted by: jeffr

17 years agoFree cluster if we fail to create the dmamap.
Kip Macy [Fri, 20 Apr 2007 05:16:42 +0000 (05:16 +0000)]
Free cluster if we fail to create the dmamap.

Fixes CID 1829
Found by: Coverity

17 years agoInclude a note in the sendmail 8.14.1 release note regarding the libmilter
Gregory Neil Shapiro [Fri, 20 Apr 2007 05:13:17 +0000 (05:13 +0000)]
Include a note in the sendmail 8.14.1 release note regarding the libmilter
update.

Requested by: re (bmah)

17 years agoEliminate CID 1842 by comparing against (type != EXT_MBUF) => refcnt != NULL
Kip Macy [Fri, 20 Apr 2007 05:12:54 +0000 (05:12 +0000)]
Eliminate CID 1842 by comparing against (type != EXT_MBUF) => refcnt != NULL

17 years agoFix memory leak in m_collapse (CID 1843)
Kip Macy [Fri, 20 Apr 2007 05:06:02 +0000 (05:06 +0000)]
Fix memory leak in m_collapse (CID 1843)

Found by: Coverity
Submitted by: jhb

17 years agoAdd ofw bus methods to the ppc nexus driver. This will be used in future
Peter Grehan [Fri, 20 Apr 2007 03:24:59 +0000 (03:24 +0000)]
Add ofw bus methods to the ppc nexus driver. This will be used in future
EFIKA platform support.

PR: 111522
Submitted by: Andrew Turner, andrew at fubar geek nz

17 years agoIn some cases, like whenever devfs file times are zero, the fix(aa) will not
Tom Rhodes [Fri, 20 Apr 2007 01:47:05 +0000 (01:47 +0000)]
In some cases, like whenever devfs file times are zero, the fix(aa) will not
be applied to dev entries.  This leaves us with file times like "Jan 1 1970."
Work around this problem by replacing the tv_sec == 0 check with a
<= 3600 check.  It's doubtful anyone will be booting within an hour of the
Epoch, let alone care about a few seconds worth of nonzero timestamps.  It's
a hackish work around, but it does work and I have not experienced any
negatives in my testing.

Discussed with: bde
"Ok with me: phk

17 years agoUnbreak module / driver attach breakage. Both snd_envy24 and snd_envy24ht
Ariff Abdullah [Fri, 20 Apr 2007 01:28:51 +0000 (01:28 +0000)]
Unbreak module / driver attach breakage. Both snd_envy24 and snd_envy24ht
mistakenly rely on wrong snd_spicds version.

17 years agoRetire the spl() markers. Add in some minor missed locking as a result.
Scott Long [Thu, 19 Apr 2007 23:34:51 +0000 (23:34 +0000)]
Retire the spl() markers.  Add in some minor missed locking as a result.

17 years agoInline cam_periph_lock|unlock to make debugging easier. Use more
Scott Long [Thu, 19 Apr 2007 22:46:26 +0000 (22:46 +0000)]
Inline cam_periph_lock|unlock to make debugging easier.  Use more
CAM_SIM_LOCK() more uniformly.