]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoPortability.
Tim Kientzle [Fri, 13 Apr 2007 16:07:25 +0000 (16:07 +0000)]
Portability.

17 years agoThe minimum size of an RFC3442 destination descriptor is five bytes, so
Ed Maste [Fri, 13 Apr 2007 15:07:10 +0000 (15:07 +0000)]
The minimum size of an RFC3442 destination descriptor is five bytes, so
correct test to -ge 5.  Without this change an RFC3442 encoded default
route would be ignored.

Reported by: Cedric Jonas <cedric at decemplex dot net>

17 years agoFix the handling of IPv6 addresses for subject and process BSM audit
Christian S.J. Peron [Fri, 13 Apr 2007 14:55:19 +0000 (14:55 +0000)]
Fix the handling of IPv6 addresses for subject and process BSM audit
tokens. Currently, we do not support the set{get}audit_addr(2) system
calls which allows processes like sshd to set extended or ip6
information for subject tokens.

The approach that was taken was to change the process audit state
slightly to use an extended terminal ID in the kernel. This allows
us to store both IPv4 IPv6 addresses. In the case that an IPv4 address
is in use, we convert the terminal ID from an struct auditinfo_addr to
a struct auditinfo.

If getaudit(2) is called when the subject is bound to an ip6 address,
we return E2BIG.

- Change the internal audit record to store an extended terminal ID
- Introduce ARG_TERMID_ADDR
- Change the kaudit <-> BSM conversion process so that we are using
  the appropriate subject token. If the address associated with the
  subject is IPv4, we use the standard subject32 token. If the subject
  has an IPv6 address associated with them, we use an extended subject32
  token.
- Fix a couple of endian issues where we do a couple of byte swaps when
  we shouldn't be. IP addresses are already in the correct byte order,
  so reading the ip6 address 4 bytes at a time and swapping them results
  in in-correct address data. It should be noted that the same issue was
  found in the openbsm library and it has been changed there too on the
  vendor branch
- Change A_GETPINFO to use the appropriate structures
- Implement A_GETPINFO_ADDR which basically does what A_GETPINFO does,
  but can also handle ip6 addresses
- Adjust get{set}audit(2) syscalls to convert the data
  auditinfo <-> auditinfo_addr
- Fully implement set{get}audit_addr(2)

NOTE: This adds the ability for processes to correctly set extended subject
information. The appropriate userspace utilities still need to be updated.

MFC after: 1 month
Reviewed by: rwatson
Obtained from: TrustedBSD

17 years agoPav and marcus co-mentored me for a few weeks while adamw was busy in school
Michael Johnson [Fri, 13 Apr 2007 14:29:38 +0000 (14:29 +0000)]
Pav and marcus co-mentored me for a few weeks while adamw was busy in school

17 years ago- Create an empty /etc/zfs/exports file when zfs_enable="YES" and we don't
Pawel Jakub Dawidek [Fri, 13 Apr 2007 11:02:06 +0000 (11:02 +0000)]
- Create an empty /etc/zfs/exports file when zfs_enable="YES" and we don't
  NFS-share anything. This way we can safely start mountd with
  /etc/zfs/exports and mountd won't complain.

  Pointed out by: ceri

- Move 'zfs volinit' before 'zfs mount -a' and 'zfs volfini' after
  'zfs unmount -a'.

17 years agomountd(8) was changed to only abort when all given exports files cannot be
Pawel Jakub Dawidek [Fri, 13 Apr 2007 10:29:25 +0000 (10:29 +0000)]
mountd(8) was changed to only abort when all given exports files cannot be
open, so we not longer has to check if /etc/zfs/exports exists.

17 years agoFatal error is only when cannot open any of the given exports files.
Pawel Jakub Dawidek [Fri, 13 Apr 2007 10:25:49 +0000 (10:25 +0000)]
Fatal error is only when cannot open any of the given exports files.

17 years agoFix vnodes starvation caused by DNLC (ZFS name cache):
Pawel Jakub Dawidek [Fri, 13 Apr 2007 08:42:01 +0000 (08:42 +0000)]
Fix vnodes starvation caused by DNLC (ZFS name cache):
- Tune number of namecache entires better (based on desiredvnodes).
- Handle vfs_lowvnodes event by releasing requested number of name cache
  entries, but no less than 5%.

Reported by: simokawa

17 years agoWhen we are running low on vnodes, there is currently no way to ask other
Pawel Jakub Dawidek [Fri, 13 Apr 2007 08:38:48 +0000 (08:38 +0000)]
When we are running low on vnodes, there is currently no way to ask other
subsystems to release some vnodes. Implement backpressure based on
vfs_lowvnodes event (similar to vm_lowmem for memory).

17 years agoo Look for a zfs(1) exports file only if it exists and is readable. If
Mike Makonnen [Fri, 13 Apr 2007 06:42:25 +0000 (06:42 +0000)]
o Look for a zfs(1) exports file only if it exists and is readable. If
  we don't do this and the file doesn't exist mountd(8) will abort.
o The mountd(8) daemon creates a pidfile, so use it.

17 years agoInstall only types.h from sys/rpc/.
Pawel Jakub Dawidek [Fri, 13 Apr 2007 01:39:33 +0000 (01:39 +0000)]
Install only types.h from sys/rpc/.

Requested by: ache
Explained how by: ru

17 years agoMFp4: Synchronize with vendor (mostly 'zfs rename -r').
Pawel Jakub Dawidek [Thu, 12 Apr 2007 23:16:02 +0000 (23:16 +0000)]
MFp4: Synchronize with vendor (mostly 'zfs rename -r').

17 years agoMFp4: Bring back comments.
Pawel Jakub Dawidek [Thu, 12 Apr 2007 23:14:25 +0000 (23:14 +0000)]
MFp4: Bring back comments.

Requested by: jhb

17 years agoFound the right date for davidxu
Julian Elischer [Thu, 12 Apr 2007 20:43:57 +0000 (20:43 +0000)]
Found the right date for davidxu

17 years agoAdd myself and some others according to my dusty memory.
Julian Elischer [Thu, 12 Apr 2007 20:39:55 +0000 (20:39 +0000)]
Add myself and some others according to my dusty memory.
I may have other mentees that I have forgotten...

17 years agoAdd myself and glebius.
Alexander Motin [Thu, 12 Apr 2007 19:14:09 +0000 (19:14 +0000)]
Add myself and glebius.

Approved by: glebius (mentor)

17 years agoIn .error and .warning, prefer command-line variables
Ruslan Ermilov [Thu, 12 Apr 2007 18:14:00 +0000 (18:14 +0000)]
In .error and .warning, prefer command-line variables
to globals, as per documentation.

Nudged by: Jeremie Le Hen

17 years ago-) Correct sdcount for a plex when removing or adding subdisks.
Lukas Ertl [Thu, 12 Apr 2007 17:54:35 +0000 (17:54 +0000)]
-) Correct sdcount for a plex when removing or adding subdisks.
-) Set correct sizes for plexes and volumes a subdisk has been removed.

Submitted by:   Ulf Lilleengen <lulf_AT_freebsd.org>

17 years agoAvoid infinite loop if the device string given for a drive
Lukas Ertl [Thu, 12 Apr 2007 17:40:44 +0000 (17:40 +0000)]
Avoid infinite loop if the device string given for a drive
only consists of "/".

Submitted by:  Ulf Lilleengen <lulf_AT_freebsd.org>

17 years agoMFamd64
Alan Cox [Thu, 12 Apr 2007 17:00:56 +0000 (17:00 +0000)]
MFamd64
 Define PGEX_RSV.

17 years agoAdd myself and my mentor, arved.
Stefan Walter [Thu, 12 Apr 2007 15:45:30 +0000 (15:45 +0000)]
Add myself and my mentor, arved.

17 years agoMerge
Andrey A. Chernov [Thu, 12 Apr 2007 14:45:25 +0000 (14:45 +0000)]
Merge

17 years agoNuke ipblock. A more complete multicast regression test suite is on the way.
Bruce M Simpson [Thu, 12 Apr 2007 11:46:30 +0000 (11:46 +0000)]
Nuke ipblock. A more complete multicast regression test suite is on the way.

17 years agoFix PAE on SMP by enabling EFER_NXE on all APs.
Ruslan Ermilov [Thu, 12 Apr 2007 11:05:24 +0000 (11:05 +0000)]
Fix PAE on SMP by enabling EFER_NXE on all APs.

Reported by: kris
Diagnosed by: alc

17 years agoFix a bug in the description of the "p6-div" event. [1]
Joseph Koshy [Thu, 12 Apr 2007 09:16:54 +0000 (09:16 +0000)]
Fix a bug in the description of the "p6-div" event. [1]

Update the description of the "p6-div" and "p6-mul" events according
to the "Intel(r) 64 and IA-32 Architectures Software Developers
Manual Volume 3B: System Programming Guide, Part 2, November 2006".

Reported by:  Harald Servat <redcrash at gmail dot com> [1]

17 years agoBump .Dd for r1.5; fix grammatical problem.
Ceri Davies [Thu, 12 Apr 2007 08:39:13 +0000 (08:39 +0000)]
Bump .Dd for r1.5; fix grammatical problem.

17 years agoSpecify the correct way to modify this file, and warn that the
Mike Makonnen [Thu, 12 Apr 2007 08:35:58 +0000 (08:35 +0000)]
Specify the correct way to modify this file, and warn that the
user should not depend on the internal variables documented in
this man page.

MFC After: 2 weeks

17 years agoRestore the ".Sh BUGS" line that was accidentally removed
Ruslan Ermilov [Thu, 12 Apr 2007 08:22:49 +0000 (08:22 +0000)]
Restore the ".Sh BUGS" line that was accidentally removed
in the previous commit.

17 years agoThere are a couple of bugs in rev. 1.27:
Mike Makonnen [Thu, 12 Apr 2007 08:17:56 +0000 (08:17 +0000)]
There are a couple of bugs in rev. 1.27:
1) The man page should describe the code, not the other way around.
2) Internal variables should not be documented or exposed, except in
   controlled circumstances (i.e. - That's what the -C flag is for).
   The variable should have been saved to the config file in save_config().
3) The next available userid doesn't get automatically updated. The
   end-result is the same (user gets added with the correct uid),
   but in an interactive session the default uid doesn't get updated in
   the display.

So,

o Use the uidstart variable instead of uuid (bug #3)
o Actually save the variable to adduser.conf (bug #2)
o (bug #1 to be fixed in an upcomming commit to adduser.conf.5)

MFC After: 2 weeks

17 years agoErrm... I don't see how rev. 1.26 could have possibly worked or been tested.
Mike Makonnen [Thu, 12 Apr 2007 07:38:26 +0000 (07:38 +0000)]
Errm... I don't see how rev. 1.26 could have possibly worked or been tested.
Fix it for real.

Submitted by: Johnny Lee <johnny@bmtk.com>
MFC After: 2 weeks

17 years agorestore sense to get_imm_packet
Kip Macy [Thu, 12 Apr 2007 04:48:54 +0000 (04:48 +0000)]
restore sense to get_imm_packet

MFC after: 3 days

17 years agoMake it easier to support more platforms.
Tim Kientzle [Thu, 12 Apr 2007 04:45:32 +0000 (04:45 +0000)]
Make it easier to support more platforms.
Thanks to: Joerg Sonnenberger for pointing out the need and the technique.

17 years agoNow that libarchive is being built in more environments,
Tim Kientzle [Thu, 12 Apr 2007 04:44:21 +0000 (04:44 +0000)]
Now that libarchive is being built in more environments,
factor out the platform-specific configuration header a bit
more cleanly.
Suggested by: Joerg Sonnenberger

17 years agoMake Lint happier.
Tim Kientzle [Thu, 12 Apr 2007 04:42:57 +0000 (04:42 +0000)]
Make Lint happier.

17 years agoswitch over to per-txq dma tag to facilitate parallelism on TX
Kip Macy [Thu, 12 Apr 2007 04:31:44 +0000 (04:31 +0000)]
switch over to per-txq dma tag to facilitate parallelism on TX

MFC after: 3 days

17 years agoexplicitly check TSO flag
Kip Macy [Thu, 12 Apr 2007 03:33:30 +0000 (03:33 +0000)]
explicitly check TSO flag
don't clear and then set M_PKTHDR, m_gethdr sets it correctly
improve error handling on m_gethdr failure

MFC after: 3 days

17 years agoAdd ETHER_HDR_LEN to hardware accepted mtu
Kip Macy [Thu, 12 Apr 2007 03:07:24 +0000 (03:07 +0000)]
Add ETHER_HDR_LEN to hardware accepted mtu

MFC after: 3 days

17 years ago- Fix compilaton with DUMP_FILELOCK_VERBOSE.
Jun Kuriyama [Thu, 12 Apr 2007 02:07:12 +0000 (02:07 +0000)]
- Fix compilaton with DUMP_FILELOCK_VERBOSE.
- Use consistent "get_lock_matching_unlock" function name in
  debuglog().

17 years agoFix a case where the multicast addresses were not removed from some ports. The
Andrew Thompson [Thu, 12 Apr 2007 01:58:57 +0000 (01:58 +0000)]
Fix a case where the multicast addresses were not removed from some ports. The
first port to be removed from the trunk would free the multicast list so
subsequent removed ports didnt have their multicast addresses removed.

17 years agoAdd m_last() inline function.
Andre Oppermann [Wed, 11 Apr 2007 23:13:12 +0000 (23:13 +0000)]
Add m_last() inline function.

17 years agoAdd a flag to struct pfs_vdata to mark the vnode as dead (e.g. process-
Dag-Erling Smørgrav [Wed, 11 Apr 2007 22:40:57 +0000 (22:40 +0000)]
Add a flag to struct pfs_vdata to mark the vnode as dead (e.g. process-
specific nodes when the process exits)

Move the vnode-cache-walking loop which was duplicated in pfs_exit() and
pfs_disable() into its own function, pfs_purge(), which looks for vnodes
marked as dead and / or belonging to the specified pfs_node and reclaims
them.  Note that this loop is still extremely inefficient.

Add a comment in pfs_vncache_alloc() explaining why we have to purge the
vnode from the vnode cache before returning, in case anyone should be
tempted to remove the call to cache_purge().

Move the special handling for pfstype_root nodes into pfs_fileno_alloc()
and pfs_fileno_free() (the root node's fileno must always be 2).  This
also fixes a bug where pfs_fileno_free() would reclaim the root node's
fileno, triggering a panic in the unr code, as that fileno was never
allocated from unr to begin with.

When destroying a pfs_node, release its fileno and purge it from the
vnode cache.  I wish we could put off the call to pfs_purge() until
after the entire tree had been destroyed, but then we'd have vnodes
referencing freed pfs nodes.  This probably doesn't matter while we're
still under Giant, but might become an issue later.

When destroying a pseudofs instance, destroy the tree before tearing
down the fileno allocator.

In pfs_mount(), acquire the mountpoint interlock when required.

MFC after: 3 weeks

17 years agoRemove irrelevant comment about T/TCP in BUGS.
Dag-Erling Smørgrav [Wed, 11 Apr 2007 20:32:59 +0000 (20:32 +0000)]
Remove irrelevant comment about T/TCP in BUGS.

Spotted by: ru@

17 years agoRemove -T from getopt() spec and usage string.
Dag-Erling Smørgrav [Wed, 11 Apr 2007 20:32:09 +0000 (20:32 +0000)]
Remove -T from getopt() spec and usage string.

Submitted by: ru@

17 years agoBelatedly bump Dd for my recent changes.
Ed Maste [Wed, 11 Apr 2007 20:31:06 +0000 (20:31 +0000)]
Belatedly bump Dd for my recent changes.

17 years agoBump Dd.
Dag-Erling Smørgrav [Wed, 11 Apr 2007 19:20:42 +0000 (19:20 +0000)]
Bump Dd.

17 years agoRemove T/TCP support, and the -T option which was needed to disable it.
Dag-Erling Smørgrav [Wed, 11 Apr 2007 19:11:54 +0000 (19:11 +0000)]
Remove T/TCP support, and the -T option which was needed to disable it.

MFC after: 3 weeks

17 years agoAdd myself plus a template for Andrew Gallatin for the mentor link.
Bernd Walter [Wed, 11 Apr 2007 17:53:29 +0000 (17:53 +0000)]
Add myself plus a template for Andrew Gallatin for the mentor link.

17 years agoModified release note: Reword sg note to make more sense.
Bruce A. Mah [Wed, 11 Apr 2007 17:45:37 +0000 (17:45 +0000)]
Modified release note:  Reword sg note to make more sense.

Problems pointed out by: brueffer

17 years agoRemove obsolete comment about privileges: SUSER_ALLOWJAIL is no longer set
Robert Watson [Wed, 11 Apr 2007 16:31:02 +0000 (16:31 +0000)]
Remove obsolete comment about privileges: SUSER_ALLOWJAIL is no longer set
in this code.

17 years agoRemove now-obsolete comment regarding mqueue privileges in jail.
Robert Watson [Wed, 11 Apr 2007 16:22:59 +0000 (16:22 +0000)]
Remove now-obsolete comment regarding mqueue privileges in jail.

17 years ago- Add IANA-assigned ports for HP status & services daemon and I/O backend
Stanislav Sedov [Wed, 11 Apr 2007 16:02:04 +0000 (16:02 +0000)]
- Add IANA-assigned ports for HP status & services daemon and I/O backend
  daemon. The FreeBSD port print/hplip currently provides these services.

PR: conf/99593
Submitted by: Anish Mistry <amistry@am-productions.biz>
Approved by: maxim
MFC after: 1 week

17 years ago- Add IANA assigned port for amanda server control over tcp. The
Stanislav Sedov [Wed, 11 Apr 2007 15:58:36 +0000 (15:58 +0000)]
- Add IANA assigned port for amanda server control over tcp. The
  current misc/amanda-server code uses it.

PR: conf/111050
Submitted by: Charles Sprickman<spork@bway.net>
Approved by: maxim
MFC after: 1 week

17 years agoNew release notes: kern.hostuuid, trunk(4), sg (+MFC),
Bruce A. Mah [Wed, 11 Apr 2007 15:37:37 +0000 (15:37 +0000)]
New release notes:  kern.hostuuid, trunk(4), sg (+MFC),
truss(1) uses ptrace(2), uuidgen(1) moved to /bin.

Modified release notes:  ZFS manpage update and pc98 support,
fix markup in KSE note.

17 years agoFix doc.1 build when NOPORTS is defined. This includes
Ruslan Ermilov [Wed, 11 Apr 2007 15:24:58 +0000 (15:24 +0000)]
Fix doc.1 build when NOPORTS is defined.  This includes
the following ports tree changes:

- libtool13 removed.
- pkgconfig moved to pkg-config.
- Default dependency for USE_GHOSTSCRIPT changed from
  ghostscript-gnu to ghostscript-gpl.
- Also added missing docproj dependencies.

Reported by: Dmitriy Kirhlarov
Submitted by: sem, ru
MFC after: 3 days

17 years agoMake "struct tcp_timer" visible only to the kernel, and unbreak world.
Ruslan Ermilov [Wed, 11 Apr 2007 14:08:42 +0000 (14:08 +0000)]
Make "struct tcp_timer" visible only to the kernel, and unbreak world.

17 years agoFix m_freem_vec() to actually traverse the mbuf chain. This avoids
John Baldwin [Wed, 11 Apr 2007 13:47:24 +0000 (13:47 +0000)]
Fix m_freem_vec() to actually traverse the mbuf chain.  This avoids
double free's and an infinite loop.

CID: 1834
Found by: Coverity Prevent (tm)

17 years agoGroup the loop to acquire/release Giant with the WITNESS_SAVE/RESTORE under
John Baldwin [Wed, 11 Apr 2007 13:44:55 +0000 (13:44 +0000)]
Group the loop to acquire/release Giant with the WITNESS_SAVE/RESTORE under
a single conditional.  The two operations are linked, but since the link
is not very direct, Coverity can't see it.  Humans might also miss the
link as well.  So, this isn't fixing any actual bugs, just improving
readability.

CID: 1787 (likely others as well)
Found by: Coverity Prevent (tm)

17 years ago- Sync service names with IANA (http://www.iana.org/assignments/port-numbers).
Stanislav Sedov [Wed, 11 Apr 2007 13:06:05 +0000 (13:06 +0000)]
- Sync service names with IANA (http://www.iana.org/assignments/port-numbers).
  The registration names for 5222(tcp,udp) and 5269(tcp,udp) was changed to
  xmpp-client and xmpp-server correspondingly.

  This inconsistency causes problems to applications developed on other
  systems, as they tries to use port numbers from /etc/services as fallback.

PR: conf/100606
Submitted by: Ralph Meijer <freebsd-gnats2@ralphm.ik.nu>
Approved by: maxim
MFC after: 1 week

17 years agoThis commit was generated by cvs2svn to compensate for changes in r168616,
Ruslan Ermilov [Wed, 11 Apr 2007 11:09:18 +0000 (11:09 +0000)]
This commit was generated by cvs2svn to compensate for changes in r168616,
which included commits to RCS files with non-trunk default branches.

17 years agoUnbreak world build.
Ruslan Ermilov [Wed, 11 Apr 2007 11:09:18 +0000 (11:09 +0000)]
Unbreak world build.

17 years agoChange the TCP timer system from using the callout system five times
Andre Oppermann [Wed, 11 Apr 2007 09:45:16 +0000 (09:45 +0000)]
Change the TCP timer system from using the callout system five times
directly to a merged model where only one callout, the next to fire,
is registered.

Instead of callout_reset(9) and callout_stop(9) the new function
tcp_timer_activate() is used which then internally manages the callout.

The single new callout is a mutex callout on inpcb simplifying the
locking a bit.

tcp_timer() is the called function which handles all race conditions
in one place and then dispatches the individual timer functions.

Reviewed by: rwatson (earlier version)

17 years agoBump date to the day of the import.
Christian Brueffer [Wed, 11 Apr 2007 08:55:04 +0000 (08:55 +0000)]
Bump date to the day of the import.

17 years agoCorrect a typo in an event name alias.
Joseph Koshy [Wed, 11 Apr 2007 04:52:09 +0000 (04:52 +0000)]
Correct a typo in an event name alias.

Reported by:  Harald Servat <redcrash at gmail dot com>

17 years agoPut some overly verbose prints under bootverbose. This is on the vendor
Nate Lawson [Wed, 11 Apr 2007 02:03:36 +0000 (02:03 +0000)]
Put some overly verbose prints under bootverbose.  This is on the vendor
branch but we need to work out a different interface with the vendor.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r168609,
Nate Lawson [Wed, 11 Apr 2007 02:03:36 +0000 (02:03 +0000)]
This commit was generated by cvs2svn to compensate for changes in r168609,
which included commits to RCS files with non-trunk default branches.

17 years agoAdd work around for hardware Tx checksum offload bug in Yukon II.
Pyun YongHyeon [Wed, 11 Apr 2007 00:47:29 +0000 (00:47 +0000)]
Add work around for hardware Tx checksum offload bug in Yukon II.
Yukon II generated corrupted TCP checksum for short TCP packets
that's less than 60 bytes in size(e.g. window probe packet, pure ACK
packet etc). Padding the frame with zeros to make the frame minimum
ethernet frame size didn't work at all. Instead of dropping Tx
checksum offload support we calculate TCP checksum with S/W method
when we encounter short TCP frames.
Fortunately it seems that short UDP datagrams appear to be handled
correctly by Yukon II.

While I'm here simplify ethernet/VLAN header size calculation logic.

PR: 111384

17 years agoIf available, take UUID from smbios.system.uuid, if not fall back to
Pawel Jakub Dawidek [Wed, 11 Apr 2007 00:05:25 +0000 (00:05 +0000)]
If available, take UUID from smbios.system.uuid, if not fall back to
software-generated UUID. Store the result in /etc/hostid and use it in
the future. Perform simple UUID format check, as there is a lot of
hardware with broken UUIDs. The check should be improved to also eliminate
fake UUIDs like 00000000-0000-0000-0000-000000000000.

Requested by: many

17 years agoMove rpc/types.h under sys/, as this is used by ZFS kernel module.
Pawel Jakub Dawidek [Tue, 10 Apr 2007 22:10:16 +0000 (22:10 +0000)]
Move rpc/types.h under sys/, as this is used by ZFS kernel module.

Repo-copied by: simon

17 years agostrchr() and strrchr() are already present in the kernel, but with less
Wojciech A. Koszek [Tue, 10 Apr 2007 21:42:12 +0000 (21:42 +0000)]
strchr() and strrchr() are already present in the kernel, but with less
popular names. Hence:

- comment current index() and rindex() functions, as these serve the same
  functionality as, respectively, strchr() and strrchr() from userland;
- add inlined version of strchr() and strrchr(), as we tend to use them more
  often;
- remove str[r]chr() definitions from ZFS code;

Reviewed by: pjd
Approved by: cognet (mentor)

17 years agoRemove trailing '.' for consistency!
Pawel Jakub Dawidek [Tue, 10 Apr 2007 21:40:13 +0000 (21:40 +0000)]
Remove trailing '.' for consistency!

17 years agoWhitespace fixes
Scott Long [Tue, 10 Apr 2007 21:37:37 +0000 (21:37 +0000)]
Whitespace fixes

17 years agoLet brgphy(4) attach for the Broadcom BCM5755 ASIC based chipsets
Marius Strobl [Tue, 10 Apr 2007 20:43:23 +0000 (20:43 +0000)]
Let brgphy(4) attach for the Broadcom BCM5755 ASIC based chipsets
as well.

Obtained from: OpenBSD
MFC after: 1 week

17 years agoOn i386 compile the back-end with EISA support as well as the EISA
Marius Strobl [Tue, 10 Apr 2007 20:33:31 +0000 (20:33 +0000)]
On i386 compile the back-end with EISA support as well as the EISA
front-end if the dpt(4) module is built along with a kernel that
includes eisa(4) or when compiling it stand-alone (logic based on
the corresponding ISA logic in sys/modules/sound/sound/Makefile).
As as side-effect this fixes the stand-alone build of the dpt(4)
module after dpt.h 1.17, dpt_eisa.c 1.22 and dpt_scsi.c 1.55.

Breakage reported by: n_hibma

17 years agoA fix for the SG_GET_TIMEOUT function slipped into a previous commit by
Scott Long [Tue, 10 Apr 2007 20:03:42 +0000 (20:03 +0000)]
A fix for the SG_GET_TIMEOUT function slipped into a previous commit by
accident.  Remove the text describing the problem as it is no longer
relevant.  Also give real implementations for the GET and SET ioctls.

17 years agoEliminate memory leak from an accidental malloc().
Ed Maste [Tue, 10 Apr 2007 17:37:53 +0000 (17:37 +0000)]
Eliminate memory leak from an accidental malloc().

17 years agoAdd UFS_GJOURNAL options to the GENERIC kernel.
Pawel Jakub Dawidek [Tue, 10 Apr 2007 16:49:41 +0000 (16:49 +0000)]
Add UFS_GJOURNAL options to the GENERIC kernel.

Approved by: re (kensmith)

17 years agoAdd a pfsync_syncpeer option to /etc/defaults/rc.conf and rc.conf(5),
Giorgos Keramidas [Tue, 10 Apr 2007 16:42:14 +0000 (16:42 +0000)]
Add a pfsync_syncpeer option to /etc/defaults/rc.conf and rc.conf(5),
which can be used to turn off multicast pfsync support, and enable
the transmission of directed PFSYNC (IP protocol: 240) packets to
a specific "sync peer" host.

PR: conf/111225
Submitted by: Bas van Beek <bas@tobin.nl>
Approved by: mtm, mlaier
MFC after: 2 weeks

17 years agoUpdate comment regarding how we check privilege on FreeBSD: we now use
Robert Watson [Tue, 10 Apr 2007 16:09:00 +0000 (16:09 +0000)]
Update comment regarding how we check privilege on FreeBSD: we now use
priv_check().

17 years agoAllow PRIV_NETINET_REUSEPORT in jail.
Robert Watson [Tue, 10 Apr 2007 15:59:49 +0000 (15:59 +0000)]
Allow PRIV_NETINET_REUSEPORT in jail.

17 years agoAdd a new privilege, PRIV_NETINET_REUSEPORT, which will replace superuser
Robert Watson [Tue, 10 Apr 2007 15:58:38 +0000 (15:58 +0000)]
Add a new privilege, PRIV_NETINET_REUSEPORT, which will replace superuser
checks to see whether bind() can reuse a port/address combination while
it's already in use (for some definition of use).

17 years agoRemove unnecessary suser() check in the sysctl to set up ath_hal
Robert Watson [Tue, 10 Apr 2007 15:48:45 +0000 (15:48 +0000)]
Remove unnecessary suser() check in the sysctl to set up ath_hal
logging: the sysctl framework will already have checked for privilege
if a sysctl value is being set.

Discussed a long time ago with: sam

17 years agoDo allow POSIX mqueue unlink privilege inside a jail, as we all all
Robert Watson [Tue, 10 Apr 2007 15:40:27 +0000 (15:40 +0000)]
Do allow POSIX mqueue unlink privilege inside a jail, as we all all
other POSIX mqueue privileges inside a jail.

17 years agoMinor style cleanups (mostly removal of trailing whitespaces).
Pawel Jakub Dawidek [Tue, 10 Apr 2007 15:29:37 +0000 (15:29 +0000)]
Minor style cleanups (mostly removal of trailing whitespaces).

17 years agoCorrect typos.
Pawel Jakub Dawidek [Tue, 10 Apr 2007 15:22:40 +0000 (15:22 +0000)]
Correct typos.

17 years agoMFp4: Add missing kmem_size() definition.
Pawel Jakub Dawidek [Tue, 10 Apr 2007 13:26:45 +0000 (13:26 +0000)]
MFp4: Add missing kmem_size() definition.

17 years agoMFp4: Allow to set zfs_recover via vfs.zfs.recover from /boot/loader.conf.
Pawel Jakub Dawidek [Tue, 10 Apr 2007 12:54:19 +0000 (12:54 +0000)]
MFp4: Allow to set zfs_recover via vfs.zfs.recover from /boot/loader.conf.

17 years agoMFp4: Hide under '#ifdef _KERNEL' only what's really needed.
Pawel Jakub Dawidek [Tue, 10 Apr 2007 12:52:14 +0000 (12:52 +0000)]
MFp4: Hide under '#ifdef _KERNEL' only what's really needed.

17 years agoMinor typo fix, noticed while I was going through *_pager.c files.
Giorgos Keramidas [Tue, 10 Apr 2007 12:34:51 +0000 (12:34 +0000)]
Minor typo fix, noticed while I was going through *_pager.c files.

17 years agoExpand the description of the host's UUID.
Ruslan Ermilov [Tue, 10 Apr 2007 11:19:52 +0000 (11:19 +0000)]
Expand the description of the host's UUID.

OK'ed by: pjd

17 years agoSwap "underflow"/"overflow" in the table header.
Ruslan Ermilov [Tue, 10 Apr 2007 11:17:00 +0000 (11:17 +0000)]
Swap "underflow"/"overflow" in the table header.

Submitted by: Ricardo Nabinger Sanchez
MFC after: 3 days

17 years agoFix the NAMEI zone leak when snapshot was successfully created.
Konstantin Belousov [Tue, 10 Apr 2007 09:31:42 +0000 (09:31 +0000)]
Fix the NAMEI zone leak when snapshot was successfully created.

Reported and tested by: Peter Holm
MFC after: 2 weeks

17 years agoRecalculate the NEWBLOCK flag for pagedep structure after the softdep
Konstantin Belousov [Tue, 10 Apr 2007 09:30:41 +0000 (09:30 +0000)]
Recalculate the NEWBLOCK flag for pagedep structure after the softdep
lock is dropped, since pagedep may be already processed and deallocated.

Found and tested by: kris
MFC after: 2 weeks

17 years agoWhen LK_NOWAIT is passed as argument to process_worklist_item(), this
Konstantin Belousov [Tue, 10 Apr 2007 09:28:17 +0000 (09:28 +0000)]
When LK_NOWAIT is passed as argument to process_worklist_item(), this
does not prevent handle_workitem_remove() from recursing into a blocking
version. Add the dirrem to worklist instead of processing it now if this
is the case.

Reported and tested by: kris
Submitted by: tegge
MFC after: 2 weeks

17 years agoFix an uninitialized variable warning.
Andrew Thompson [Tue, 10 Apr 2007 08:02:33 +0000 (08:02 +0000)]
Fix an uninitialized variable warning.

17 years agoUpdate to match revision 1.23 of touch.c.
Greg Lehey [Tue, 10 Apr 2007 07:24:47 +0000 (07:24 +0000)]
Update to match revision 1.23 of touch.c.
Clarify some details.

17 years agoUsage: print base name of program.
Greg Lehey [Tue, 10 Apr 2007 07:22:30 +0000 (07:22 +0000)]
Usage: print base name of program.

-A flag: respect the -a and -m flags.
 imply the -c flag.

Requested in principle by: brian

17 years agoMake use of ptrace(2) instead of procfs in truss(1), eliminating
Xin LI [Tue, 10 Apr 2007 04:03:34 +0000 (04:03 +0000)]
Make use of ptrace(2) instead of procfs in truss(1), eliminating
yet another need of an available /proc/ mount.

Tested with: make universe
Submitted by: howardsu
Reviewed by: alfred

17 years agoBump document date for new trunk commands.
Andrew Thompson [Tue, 10 Apr 2007 03:35:42 +0000 (03:35 +0000)]
Bump document date for new trunk commands.

17 years agoFix build, trunk is a device not an option.
Andrew Thompson [Tue, 10 Apr 2007 03:09:38 +0000 (03:09 +0000)]
Fix build, trunk is a device not an option.

17 years agoTry to stabilize ZFS with regard to memory consumption:
Pawel Jakub Dawidek [Tue, 10 Apr 2007 02:35:57 +0000 (02:35 +0000)]
Try to stabilize ZFS with regard to memory consumption:
- Allow to shrink ARC down to 16MB (instead of 64MB).
- Set arc_max to 1/2 of kmem_map by default.
- Start freeing things earlier when low memory situation is detected.
- Serialize execution of arc_lowmem().

I decided to setup minimum ZFS memory requirements to 512MB of RAM and 256MB of
kmem_map size. If there is less RAM or kmem_map, a warning will be printed.
World is cruel, be no better. In other words: modern file system requires
modern hardware:)

From ZFS administration guide:

"Currently the minimum amount of memory recommended to install a Solaris
 system is 512 Mbytes. However, for good ZFS performance, at least one
 Gbyte or more of memory is recommended."