dragonfly.git
16 years agoAdd a new option -s. With this option pkg_search(1) will display the description
Matthias Schmidt [Sun, 17 Feb 2008 19:35:24 +0000 (19:35 +0000)]
Add a new option -s.  With this option pkg_search(1) will display the description
of a package.  This is very helpful it you want to know more about a package.

% pkg_search -s host-20040812
Host is a very powerful DNS query and testing command-line tool. Major
differences with the ancient version of the program with the same name
[...]

Note:  You need the complete pkgsrc tree to use -s.

While here:
 o Whitespace fixes
 o Remove unused split commands
 o Bump copyright to 2008.

16 years agoDump entries can have the same timestamp, so accept those as well.
Simon Schubert [Sun, 17 Feb 2008 10:23:57 +0000 (10:23 +0000)]
Dump entries can have the same timestamp, so accept those as well.

16 years agoFix mount_nfs to allow hostnames which begin with a digit.
Matthew Dillon [Sat, 16 Feb 2008 20:48:29 +0000 (20:48 +0000)]
Fix mount_nfs to allow hostnames which begin with a digit.

Submitted-by: "Frank W. Josellis" <frank@dynamical-systems.org>
16 years agoFix an issue where the random number generator's random event injector
Matthew Dillon [Sat, 16 Feb 2008 20:40:47 +0000 (20:40 +0000)]
Fix an issue where the random number generator's random event injector
can stall.  An interrupt could occur after rand_thread_signal is set to
0 but before the injector thread deschedules itself, resulting in the
interrupt waking up the thread before it has gone to sleep and stalling
it forever.  Also swap around the variable ordering so the SMP race
that occurs is non-harmful.  Continue to allow the race.

Reported-by: Robin Carey <robin_carey5@yahoo.co.uk>
16 years agoFix two typos
Matthias Schmidt [Sat, 16 Feb 2008 15:53:39 +0000 (15:53 +0000)]
Fix two typos

16 years agoTypo
Sepherosa Ziehau [Fri, 15 Feb 2008 11:48:15 +0000 (11:48 +0000)]
Typo

16 years ago- Try adjusting TX power as much as possible when doing the initial TX power
Sepherosa Ziehau [Fri, 15 Feb 2008 11:15:38 +0000 (11:15 +0000)]
- Try adjusting TX power as much as possible when doing the initial TX power
  calibration.  Also the second TX power calibration happens 1 second after
  the initial TX power calibration.  if streams are pushed immediately after
  NIC is brought up, this makes thing work much better for most of the 4306
  BBP, whose default TX power seems to be far away from the max TX power.
- Add some experimental (commented out) code to force TX power downgrade.
- For parts whose MAC rev >= 5, if PHY error happens, try resetting NIC but
  keep 802.11 state machine untouched.  PHY error usually happens when you
  roam into a signal dark corner, originally even if you walk the NIC out of
  the signal dark corner, TX performance still sucks; after this change, TX
  performance will stay on the right rank.  Heavy beacon missing that is
  enough to trigger reassoc/rescan does not seem to happen even if in this
  signal dark corner.

16 years agoRelease serializer around firmware_load_image() to avoid possible dead
Sepherosa Ziehau [Thu, 14 Feb 2008 12:53:52 +0000 (12:53 +0000)]
Release serializer around firmware_load_image() to avoid possible dead
lock.  Add comment about it.

16 years agoInitialize firmware image tailq
Sepherosa Ziehau [Thu, 14 Feb 2008 12:30:31 +0000 (12:30 +0000)]
Initialize firmware image tailq

16 years agoAdd missing S option to usage()
Matthias Schmidt [Thu, 14 Feb 2008 09:33:24 +0000 (09:33 +0000)]
Add missing S option to usage()

16 years agoFix incomplete stack traces by gdb.
Simon Schubert [Thu, 14 Feb 2008 00:24:24 +0000 (00:24 +0000)]
Fix incomplete stack traces by gdb.

Gdb tries unwinding a stack frame by analyzing the function prologue.
If it can not find the beginning of the function, which happens for
stripped binaries, etc., it will resort to guessing.  It then assumes
that the function is a frame-less function without any local stack
variables.  This of course is wrong for almost all functions.

We work around this problem by assuming a valid stack frame.

Previous versions of gdb were broken the same way, but libbfd would
actually compensate with another bug which would simply report a wrong
function start address -- the address of a preceding, known function.
Because most functions indeed use a proper stack frame, this would
trick gdb into doing the unwinding properly.

16 years agoSet unique mode for the history, so that repeating a command doesn't
Matthias Schmidt [Wed, 13 Feb 2008 15:13:37 +0000 (15:13 +0000)]
Set unique mode for the history, so that repeating a command doesn't
spam the history.

Obtained-from: NetBSD

16 years agoRemove some more leftovers from _ntp and add _sdpd where necessary.
Matthias Schmidt [Wed, 13 Feb 2008 14:45:28 +0000 (14:45 +0000)]
Remove some more leftovers from _ntp and add _sdpd where necessary.

Reviewed-by: swildner@
16 years agoRemove _ntp and add _sdpd.
Matthias Schmidt [Wed, 13 Feb 2008 12:25:12 +0000 (12:25 +0000)]
Remove _ntp and add _sdpd.

16 years agoDon't forget to set internal error message in kvm_nlist().
Simon Schubert [Wed, 13 Feb 2008 00:12:46 +0000 (00:12 +0000)]
Don't forget to set internal error message in kvm_nlist().

Obtained-from:  FreeBSD

16 years agoUnstaticize some variables so ktrdump will find them.
Simon Schubert [Tue, 12 Feb 2008 23:33:23 +0000 (23:33 +0000)]
Unstaticize some variables so ktrdump will find them.

16 years agoDon't require the tsc frequency to be available.
Simon Schubert [Tue, 12 Feb 2008 23:19:24 +0000 (23:19 +0000)]
Don't require the tsc frequency to be available.

16 years agoOops, its index.cgi, not index.php
Matthias Schmidt [Tue, 12 Feb 2008 23:01:22 +0000 (23:01 +0000)]
Oops, its index.cgi, not index.php

16 years agoCorrect two links.
Matthias Schmidt [Tue, 12 Feb 2008 22:52:46 +0000 (22:52 +0000)]
Correct two links.

16 years agoDon't overwrite dma(8)'s config files.
Matthias Schmidt [Tue, 12 Feb 2008 22:10:20 +0000 (22:10 +0000)]
Don't overwrite dma(8)'s config files.

16 years ago- Install bthcid.conf.
Hasso Tepper [Tue, 12 Feb 2008 21:36:19 +0000 (21:36 +0000)]
- Install bthcid.conf.
- Don't overwrite bthcid.conf and hosts.
- Use permissions which make sense.

16 years agoAdd O_SYNC, a SUSv3 alias for O_FSYNC (well, roughly).
Simon Schubert [Tue, 12 Feb 2008 20:00:38 +0000 (20:00 +0000)]
Add O_SYNC, a SUSv3 alias for O_FSYNC (well, roughly).

16 years agoAdjust for 1.12
Matthias Schmidt [Tue, 12 Feb 2008 16:12:54 +0000 (16:12 +0000)]
Adjust for 1.12

16 years agoUpdate to the version 2008020400 which adds IPv6 addresses for six root
Hasso Tepper [Tue, 12 Feb 2008 14:34:34 +0000 (14:34 +0000)]
Update to the version 2008020400 which adds IPv6 addresses for six root
servers.

16 years agoOops, drop head's version back one for head (it was set to the release's
Matthew Dillon [Tue, 12 Feb 2008 02:43:17 +0000 (02:43 +0000)]
Oops, drop head's version back one for head (it was set to the release's
version).

16 years agoRelease Engineering on HEAD.
Matthew Dillon [Tue, 12 Feb 2008 02:26:08 +0000 (02:26 +0000)]
Release Engineering on HEAD.

* Branch 1.12

* Change HEAD to 1.13

16 years agoRelease engineering, Add a slip target to /usr/src/Makefile for 1.12 and
Matthew Dillon [Tue, 12 Feb 2008 02:12:18 +0000 (02:12 +0000)]
Release engineering, Add a slip target to /usr/src/Makefile for 1.12 and
add a cvsup example file for 1.12.

16 years agoBump to 1.11.1 prior to 1.12 branch and update the preview tag.
Matthew Dillon [Tue, 12 Feb 2008 02:09:24 +0000 (02:09 +0000)]
Bump to 1.11.1 prior to 1.12 branch and update the preview tag.

16 years agoAdjust nrelease to a new package set.
Matthew Dillon [Tue, 12 Feb 2008 02:06:12 +0000 (02:06 +0000)]
Adjust nrelease to a new package set.

Rebuild the installer, bootstrap, and other packages from pkgsrc, and
make the resulting binary packages available to the nrelease build.

No modifications to the pkgsrc installer have been made but it now uses
a more up-to-date support gettext.

16 years agoReport the last error code when searching for a tap device.
Simon Schubert [Mon, 11 Feb 2008 22:50:12 +0000 (22:50 +0000)]
Report the last error code when searching for a tap device.

16 years agoFix WARNS6: mark s as unused.
Sascha Wildner [Mon, 11 Feb 2008 20:10:23 +0000 (20:10 +0000)]
Fix WARNS6: mark s as unused.

16 years agoFix WARNS6: values_len is size_t, thus n must be unsigned.
Sascha Wildner [Mon, 11 Feb 2008 20:09:42 +0000 (20:09 +0000)]
Fix WARNS6: values_len is size_t, thus n must be unsigned.

16 years agoClean up remains of the umsm(4) -> ugensa(4) renaming.
Sascha Wildner [Mon, 11 Feb 2008 19:57:30 +0000 (19:57 +0000)]
Clean up remains of the umsm(4) -> ugensa(4) renaming.

16 years agoumsm(4) -> ugensa(4) as it makes much more sense - there is nothing Qualcomm
Hasso Tepper [Mon, 11 Feb 2008 18:13:58 +0000 (18:13 +0000)]
umsm(4) -> ugensa(4) as it makes much more sense - there is nothing Qualcomm
MSM chipset specific in the driver. Manpage is rewritten in progress.

16 years agoMake the Brother HL1240 printer work with ulpt.
Matthew Dillon [Mon, 11 Feb 2008 16:56:53 +0000 (16:56 +0000)]
Make the Brother HL1240 printer work with ulpt.

Submitted-by: Joe Talbott <josepht@cstone.net>
16 years agoFix arprequest serialization.
Nuno Antunes [Mon, 11 Feb 2008 16:42:39 +0000 (16:42 +0000)]
Fix arprequest serialization.

arprequest() calls ifp->if_output() without locally grabbing the
respective serializer, so ASSERT_SERIALIZED at the beginning of the
function.
Grab the serializer at arp_rtrequest() when it calls arprequest().

Reviewed-by: sephe@
16 years agoRemove link to FreeBSD porters handbook and to removed version example.
Matthias Schmidt [Mon, 11 Feb 2008 16:01:11 +0000 (16:01 +0000)]
Remove link to FreeBSD porters handbook and to removed version example.

16 years agoMention our handbook instead of FreeBSDs.
Matthias Schmidt [Mon, 11 Feb 2008 15:59:37 +0000 (15:59 +0000)]
Mention our handbook instead of FreeBSDs.

16 years agoClarify that KTR_VERBOSE is needed for verbose mode.
Sascha Wildner [Mon, 11 Feb 2008 00:27:31 +0000 (00:27 +0000)]
Clarify that KTR_VERBOSE is needed for verbose mode.

16 years agoFix include file name.
Sascha Wildner [Sun, 10 Feb 2008 23:35:30 +0000 (23:35 +0000)]
Fix include file name.

16 years agoPoint to our handbook.
Sascha Wildner [Sun, 10 Feb 2008 22:53:22 +0000 (22:53 +0000)]
Point to our handbook.

16 years agoPoint to our handbook.
Sascha Wildner [Sun, 10 Feb 2008 22:46:02 +0000 (22:46 +0000)]
Point to our handbook.

16 years agoRemove old cvsup examples.
Sascha Wildner [Sun, 10 Feb 2008 22:34:25 +0000 (22:34 +0000)]
Remove old cvsup examples.

16 years agoOops, forgotten part in last commit.
Sascha Wildner [Sun, 10 Feb 2008 22:32:19 +0000 (22:32 +0000)]
Oops, forgotten part in last commit.

16 years agoClean up and update our supfile examples.
Sascha Wildner [Sun, 10 Feb 2008 22:25:36 +0000 (22:25 +0000)]
Clean up and update our supfile examples.

Remove most FreeBSD supfiles and keep FreeBSD-cvs-supfile (renamed to
FreeBSD-supfile, like OpenBSD and NetBSD).

16 years agoHAMMER 28A/many: Translation and write performance optimizations
Matthew Dillon [Sun, 10 Feb 2008 18:58:23 +0000 (18:58 +0000)]
HAMMER 28A/many: Translation and write performance optimizations

* Cache a buffer's virtual (zoneX->zone2) translation to avoid unnecessary
  re-translations when a hot-cached buffer is available.

* Use the sequential heuristic provided by the kernel to detect sequential
  writes and call bawrite() instead bdwrite() in such cases.  This
  prevents the buffer cache from hitting slow fallback cases and
  completely solves the write performance issue.

  Note that the undo fifo is not yet in place and performance will drop
  slightly once it is, but what you see now is basically what you're gonna
  get in the final.

* Performance is now on par with UFS.

16 years agoYield after processing a couple of vnodes to avoid running too long.
Simon Schubert [Sun, 10 Feb 2008 13:45:21 +0000 (13:45 +0000)]
Yield after processing a couple of vnodes to avoid running too long.

This fixes the stuttering of xmms et al when sync is running.

16 years agotcp_output_dispatch() is only used by SMP kernel
Sepherosa Ziehau [Sun, 10 Feb 2008 11:14:25 +0000 (11:14 +0000)]
tcp_output_dispatch() is only used by SMP kernel

Noticed-by: swildner@
16 years agoRemove obsolete header file.
Sascha Wildner [Sun, 10 Feb 2008 11:13:50 +0000 (11:13 +0000)]
Remove obsolete header file.

16 years agoo Convert search string to lowercase to allow case-insensitive matching.
Matthias Schmidt [Sun, 10 Feb 2008 11:01:46 +0000 (11:01 +0000)]
o Convert search string to lowercase to allow case-insensitive matching.
o Fix non-meaningful example in man page.

16 years agoHAMMER 28/many: Implement zoned blockmap
Matthew Dillon [Sun, 10 Feb 2008 09:51:01 +0000 (09:51 +0000)]
HAMMER 28/many: Implement zoned blockmap

* Implement a zoned blockmap.  Separate B-Tree nodes, records, small blocks
  of data, and large blocks of data into their own zones.  Use 8MB large
  blocks, 32-byte blockmap entry structures, and two layers to support
  59 bits (512 petabytes).

* Create a temporary freeblock allocator so the blockmap can be tested.
  It just allocates sequentially and asserts when it hits the end of the
  volume.   This will be replaced with a real freeblock allocator soon.

* Clean up some of the mess I created from the temporary fifo mechanism
  that had been put in-place to test the major rewiring in 27.

* Adjust newfs_hammer.  The 'hammer' utility has not yet been adjusted
  (it can't decode blockmaps yet but will soon).

16 years ago- Turn on IFM_HDX
Sepherosa Ziehau [Sun, 10 Feb 2008 07:29:27 +0000 (07:29 +0000)]
- Turn on IFM_HDX
- If unknown SR values are encountered, return immediately

16 years agoIt's confusing to say that "Command Queueing Supported" just based
Peter Avalos [Sun, 10 Feb 2008 00:32:53 +0000 (00:32 +0000)]
It's confusing to say that "Command Queueing Supported" just based
upon the scsi flag validity field. Instead, just say "Command Queueing Enabled"
when it is- otherwise remain mute.

Obtained-from: FreeBSD

16 years agoMake CAM_NEW_TRAN_CODE default.
Peter Avalos [Sun, 10 Feb 2008 00:01:03 +0000 (00:01 +0000)]
Make CAM_NEW_TRAN_CODE default.

As previously mentioned, this makes a huge performance difference for one
of my disks, and future work depends on this change.

Obtained-from: FreeBSD

16 years agoMerge from vendor branch TNFTP:
Peter Avalos [Sat, 9 Feb 2008 18:35:32 +0000 (18:35 +0000)]
Merge from vendor branch TNFTP:
Sync tnftp with NetBSD:

Rename HAVE_STRUCT_SOCKADDR_SA_LEN to HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
to accurately reflect the structure member being used.

Turn off the alarmtimer before resetting the SIGALRM handler back to SIG_DFL.

16 years agoSync tnftp with NetBSD:
Peter Avalos [Sat, 9 Feb 2008 18:35:32 +0000 (18:35 +0000)]
Sync tnftp with NetBSD:

Rename HAVE_STRUCT_SOCKADDR_SA_LEN to HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
to accurately reflect the structure member being used.

Turn off the alarmtimer before resetting the SIGALRM handler back to SIG_DFL.

16 years agoMPSAFE locking for the ahc/ahd drivers using lockmgr locks.
Peter Avalos [Sat, 9 Feb 2008 18:13:13 +0000 (18:13 +0000)]
MPSAFE locking for the ahc/ahd drivers using lockmgr locks.

Obtained-from: FreeBSD

16 years agoAdd space to correctly align Office and Office phone.
Matthias Schmidt [Sat, 9 Feb 2008 17:12:05 +0000 (17:12 +0000)]
Add space to correctly align Office and Office phone.

Before:
Login            Name                 TTY  Idle  Login  Time   Office  Phone
matthias         Matthias Schmidt    *v0      5  Sat    16:01 C4347   282-1591

After:
Login            Name                 TTY  Idle  Login  Time   Office  Phone
matthias         Matthias Schmidt    *v0      5  Sat    16:01  C4347   282-1591

16 years agoRemove superfluous ,
Matthias Schmidt [Sat, 9 Feb 2008 15:29:26 +0000 (15:29 +0000)]
Remove superfluous ,

16 years agoIn TCP PRU_CONNECT handling, install inp's route with the route entry on the
Sepherosa Ziehau [Sat, 9 Feb 2008 13:38:03 +0000 (13:38 +0000)]
In TCP PRU_CONNECT handling, install inp's route with the route entry on the
correct CPU:
If the TCP connection's target port is not tcp_thread[0]'s port on which TCP
PRU_CONNECT is processed, then
- In tcp_connect(), the route entry installed in inp's route by in_pcbladdr()
  is freed, so the next fix could take effect.
- In tcp_usr_connect(), tcp_output() is dispatched to the connection's target
  port to be called, so the route entry on the connection's target CPU will
  be installed in inp's route.

Approved-by: dillon@
16 years agoAdd readelf(1) in SEE ALSO.
Nicolas Thery [Sat, 9 Feb 2008 11:40:12 +0000 (11:40 +0000)]
Add readelf(1) in SEE ALSO.

16 years agoMove text which is not part of .Fa to the next line.
Sascha Wildner [Sat, 9 Feb 2008 09:45:03 +0000 (09:45 +0000)]
Move text which is not part of .Fa to the next line.

16 years agoMove text which is not part of .Fa to the next line.
Sascha Wildner [Sat, 9 Feb 2008 09:32:05 +0000 (09:32 +0000)]
Move text which is not part of .Fa to the next line.

16 years agoDon't break the line before .An.
Sascha Wildner [Sat, 9 Feb 2008 02:31:35 +0000 (02:31 +0000)]
Don't break the line before .An.

16 years agoRemove redundant sentence.
Sascha Wildner [Sat, 9 Feb 2008 00:28:26 +0000 (00:28 +0000)]
Remove redundant sentence.

16 years agoAdd a mountctl(2) manual page.
Sascha Wildner [Sat, 9 Feb 2008 00:10:29 +0000 (00:10 +0000)]
Add a mountctl(2) manual page.

16 years agoRemove btpin(1), rfcomm_sppd(1) and sdpquery(1) from /usr/sbin.
Hasso Tepper [Fri, 8 Feb 2008 14:26:38 +0000 (14:26 +0000)]
Remove btpin(1), rfcomm_sppd(1) and sdpquery(1) from /usr/sbin.

16 years agoMobe btpin(1), rfcomm_sppd(1) and sdpquery(1) where they really should be -
Hasso Tepper [Fri, 8 Feb 2008 14:06:25 +0000 (14:06 +0000)]
Mobe btpin(1), rfcomm_sppd(1) and sdpquery(1) where they really should be -
into /usr/bin.

16 years agoRemove empty line.
Sascha Wildner [Fri, 8 Feb 2008 10:41:43 +0000 (10:41 +0000)]
Remove empty line.

16 years agoAdd 'virtual scrolling' support to moused(8). When holding down the middle
Matthias Schmidt [Fri, 8 Feb 2008 10:17:40 +0000 (10:17 +0000)]
Add 'virtual scrolling' support to moused(8).  When holding down the middle
mouse button, motions are interpreted as scrolling.
Besides that, the code contains some whitespace and mdoc cleanup.

Obtained-from: FreeBSD

16 years agoUse per softc debug level instead of a global one.
Sepherosa Ziehau [Fri, 8 Feb 2008 09:42:30 +0000 (09:42 +0000)]
Use per softc debug level instead of a global one.

16 years agoHAMMER 27/many: Major surgery - change allocation model
Matthew Dillon [Fri, 8 Feb 2008 08:31:00 +0000 (08:31 +0000)]
HAMMER 27/many: Major surgery - change allocation model

After getting stuck on the recovery code and highly unoptimal write
performance issues, remove the super-cluster/cluster and radix tree bitmap
infrastructure and replace it with a circular FIFO.

* Nothing is localized yet with this major surgery commit, which means
  radix nodes, hammer records, file data, and undo fifo elements are
  all being written to a single fifo.  These elements will soon get their
  own abstracted fifos (and in particular, the undo elements will get a
  fixed-sized circular fifo and be considered temporary data).

* No sequence numbers or transaction spaces are generated yet.

* Create a 'hammer_off_t' type (64 bits).  This type reserves 4 bits for
  a zone.  Zones which encode volume numbers reserve another 8 bits,
  giving us a 52 bit byte offset able to represent up to 4096 TB per
  volume.  Zones which do not encode volume numbers have 60 bits available
  for an abstracted offset, resulting in a maximum filesystem size of
  2^60 bytes (1 MTB).  Up to 15 zones can be encoded.

  As of this commit only 2 zones are implemented to wrap up existing
  functionality.

* Adjust the B-Tree to use full 64 bit hammer offsets.  Have one global B-Tree
  for the entire filesystem.  The tree is no longer per-cluster.

* Scrap the recovery and spike code.  Scrap the cluster and super-cluster
  code.  Scrap those portions of the B-Tree code that dealt with spikes.
  Scrap those portions of the IO subsystem that dealt with marking a
  cluster open or closed.

* Expand the hammer_modify_*() functions to include a data range and add
  UNDO record generation.  Do not implement buffer ordering dependancies
  yet (ordering issues are going change radically with the FIFO model).

16 years agoacxcontrol(8) is gone.
Sascha Wildner [Thu, 7 Feb 2008 11:44:23 +0000 (11:44 +0000)]
acxcontrol(8) is gone.

16 years agoFix Picasso's birthday.
Sascha Wildner [Thu, 7 Feb 2008 04:47:38 +0000 (04:47 +0000)]
Fix Picasso's birthday.

Taken-from: FreeBSD

16 years agolinker_file_add_dependancy() never fails so change its return type to void and
Nicolas Thery [Wed, 6 Feb 2008 22:37:46 +0000 (22:37 +0000)]
linker_file_add_dependancy() never fails so change its return type to void and
simplify calling sequences.

16 years agoHAMMER Utilities: Add an 'everything' directive to the prune command. This
Matthew Dillon [Wed, 6 Feb 2008 09:00:28 +0000 (09:00 +0000)]
HAMMER Utilities: Add an 'everything' directive to the prune command.  This
directive causes the filesystem's entire history to be destroyed.

16 years agoHAMMER 26/many: Misc features.
Matthew Dillon [Wed, 6 Feb 2008 08:59:28 +0000 (08:59 +0000)]
HAMMER 26/many: Misc features.

* Add an inode flag to lock the transaction id used to update the inode
  on-disk.  This is used for both inode creation and rename to synchronize
  the inode with the related directory entry so historical lookups do not
  get confused.

* Add a capability to prune all deleted records.

* Properly set the file size for softlinks to reflect the length of
  the softlink string (fixes a NFS issue).

* Add additional device checks to the mount code to detect block devices
  which are already in use.

* Implement the noatime mount option.

* Implement a poor-man's fakename for NFS when backing out of a historical
  access.

* Properly connect ".." for a renamed directory to the new parent directory.

16 years agoFix some NFS related bugs which cause the mount point's mnt_refs counter
Matthew Dillon [Wed, 6 Feb 2008 08:53:15 +0000 (08:53 +0000)]
Fix some NFS related bugs which cause the mount point's mnt_refs counter
to leak, preventing clean umounts.  There were several cases where
_cache_drop() was being called instead of cache_drop().

16 years ago- Use firmware(9) for acx(4) and revoke acx(4)'s private ioctls
Sepherosa Ziehau [Wed, 6 Feb 2008 08:21:22 +0000 (08:21 +0000)]
- Use firmware(9) for acx(4) and revoke acx(4)'s private ioctls
- Nuke acxcontrol(8)
- Update acx(4) manpage

16 years ago- Add sysctl nodes for statistics (*)
Sepherosa Ziehau [Wed, 6 Feb 2008 05:06:08 +0000 (05:06 +0000)]
- Add sysctl nodes for statistics (*)
- Add sysctl node for channel dwell time when scanning
- Add sysctl node for long retry limit

# (*) this paves way to nuke acxcontrol(8) and use firmware(9) for acx(4)

16 years agoHAMMER 26/many: More NFS support work, rename fixes
Matthew Dillon [Wed, 6 Feb 2008 00:27:27 +0000 (00:27 +0000)]
HAMMER 26/many: More NFS support work, rename fixes

* Add a fakename return argument to vop_nlookupdotdot().  This allows HAMMER
  to supply a fake name for historical (@@<timestamp>) transitions when
  regenerating the namecache topology.  The feature is needed when operating
  as a NFS server.

* Do not mark a vnode VROOT if it is a historical root rather than the
  realtime root.  This fixes confusion the namecache has when dealing
  with the root vnode of a HAMMER filesystem.

* Fix a bug in nrename().  ino_data.parent_obj_id was not being properly
  adjusted when moving directories.

16 years agoImplement NFS support and export control for HAMMER.
Matthew Dillon [Tue, 5 Feb 2008 20:52:01 +0000 (20:52 +0000)]
Implement NFS support and export control for HAMMER.

16 years agoFix a compiler warning.
Matthew Dillon [Tue, 5 Feb 2008 20:50:31 +0000 (20:50 +0000)]
Fix a compiler warning.

16 years ago* Implement a mountctl() op for setting export control on a filesystem.
Matthew Dillon [Tue, 5 Feb 2008 20:49:52 +0000 (20:49 +0000)]
* Implement a mountctl() op for setting export control on a filesystem.

* Adjust mountd to try to use the mountctl() op BEFORE calling a UFS-style
  mount() to set export ops for a filesystem.

* Add a prototype for the mountctl() system call in sys/mountctl.h

* Cleanup WARNS for the mountctl utility.

16 years agoAdd some words about __MAKE_CONF.
Sascha Wildner [Tue, 5 Feb 2008 19:22:04 +0000 (19:22 +0000)]
Add some words about __MAKE_CONF.

16 years agoAlso fix a double free.
Matthias Schmidt [Tue, 5 Feb 2008 14:35:43 +0000 (14:35 +0000)]
Also fix a double free.

Obtained-from: FreeBSD

16 years agoFix memory leaks.
Matthias Schmidt [Tue, 5 Feb 2008 14:25:06 +0000 (14:25 +0000)]
Fix memory leaks.

Obtained-From: OpenBSD

16 years agoo Remove version number
Matthias Schmidt [Tue, 5 Feb 2008 13:00:36 +0000 (13:00 +0000)]
o Remove version number
o Minor style changes

16 years agoSet flow control pause time to 512us _again_ as it was two years ago
Sepherosa Ziehau [Tue, 5 Feb 2008 11:12:27 +0000 (11:12 +0000)]
Set flow control pause time to 512us _again_ as it was two years ago
(if_em.c rev1.42).

Submitted-by: Gary Stanley <gary@velocity-servers.net>
16 years agoReplace the nologin shell script with a tiny c version. Just for the sake
Matthias Schmidt [Tue, 5 Feb 2008 10:58:04 +0000 (10:58 +0000)]
Replace the nologin shell script with a tiny c version.  Just for the sake
of completeness :)

16 years agoHAMMER 25/many: Pruning code
Matthew Dillon [Tue, 5 Feb 2008 07:58:43 +0000 (07:58 +0000)]
HAMMER 25/many: Pruning code

* Add b_tid to struct buf so dirty buffer cache buffers can be tagged with
  a transaction id to try to retain consistency when doing as-of queries
  on files that change size (so the data records have a TID <= the inode
  record).  This is also an issue when a file is created and immediately
  written to.  This may be temporary, a more sophisticated solution is needed.

* Fix a bug in the special handling of create_tid for as-of queries
  in btree_search().  An assignment was off by one, causing historical
  queries to not be able to find bits of data here and there.

* Freeze the transaction id for newly created inodes until the initial
  inode record is laid down on disk, so the transaction id matches the
  transaction id of the related directory entry.

* Major work on the pruning code.   When pruning the tree to a particular
  granularity the create_tid and delete_tid of related records must be
  aligned to that granularity in order to avoid creating 'holes' at
  various time points.

  This requires some serious B-Tree manipulation because the right-hand
  boundary may need to be shifted when the create_tid of an existing
  record is forward aligned.  This work is still in progress but it works
  in basic testing.

  Prune the tree in the reverse direction instead of in the forward
  direction.  This keeps the B-Tree consistent when we have to adjust
  the right-hand boundary to accomodate the realignment of create_tid.

16 years agoAdd a conditional so we don't have to drag in everything when a user
Matthew Dillon [Tue, 5 Feb 2008 01:44:04 +0000 (01:44 +0000)]
Add a conditional so we don't have to drag in everything when a user
just wants to use hammer_ioctl.h.

Reported-by: "Simon 'corecode' Schubert" <corecode@fs.ei.tum.de>
16 years agoDon't use special CWARNFLAGS. Our default is good enough.
Sascha Wildner [Mon, 4 Feb 2008 22:22:06 +0000 (22:22 +0000)]
Don't use special CWARNFLAGS. Our default is good enough.

16 years agoifconfig_blah0="authmode wpa" causes ifconfig_getargs() to interpret
Sascha Wildner [Mon, 4 Feb 2008 19:03:06 +0000 (19:03 +0000)]
ifconfig_blah0="authmode wpa" causes ifconfig_getargs() to interpret
"wpa" as an invitation to start wpa_supplicant(8) (and to strip it
from the line) instead of just passing "authmode wpa" to ifconfig(8)
as expected.

Fix this by adding some code to ifconfig_getargs() and wpaif() which
checks if "wpa" is an argument to "authmode".

Submitted-by: Frank W. Josellis <frank@dynamical-systems.org>
16 years agoInstall dma(8) config files with mode 640 as root:mail to prevent ``normal''
Matthias Schmidt [Mon, 4 Feb 2008 13:39:46 +0000 (13:39 +0000)]
Install dma(8) config files with mode 640 as root:mail to prevent ``normal''
users from spying at roots smtp login data.

16 years agoAdd a new config option to dma(8). If a user wants to use plain text SMTP
Matthias Schmidt [Mon, 4 Feb 2008 10:11:41 +0000 (10:11 +0000)]
Add a new config option to dma(8).  If a user wants to use plain text SMTP
login over an insecure connection, he has to set the INSECURE option in
the config file.  Otherwise plain text login is only available over encrypted
connections.

Discussed-with: corecode@

16 years agoo Remove per-user config file support
Matthias Schmidt [Mon, 4 Feb 2008 08:58:54 +0000 (08:58 +0000)]
o Remove per-user config file support
o Remove old-style-connect with gethostbyname() etc.  It was #ifdefed out
  and getaddrinfo() will do the job.

16 years agoHAMMER utilities: Add the 'prune' and 'history' commands.
Matthew Dillon [Mon, 4 Feb 2008 08:34:22 +0000 (08:34 +0000)]
HAMMER utilities:  Add the 'prune' and 'history' commands.

16 years agoHAMMER 25/many: Add an ioctl API for HAMMER.
Matthew Dillon [Mon, 4 Feb 2008 08:33:17 +0000 (08:33 +0000)]
HAMMER 25/many: Add an ioctl API for HAMMER.

* Add HAMMERIOC_PRUNE - a command which will scan a range of inode numbers
  and prune them according to the supplied list.   This is a preliminary
  implementation.

* Add HAMMERIOC_GETHISTORY - a command which scans the history for a
  particular file or a particular file offset within a file and displays
  it.

16 years agoUse correct extension.
Simon Schubert [Mon, 4 Feb 2008 01:38:11 +0000 (01:38 +0000)]
Use correct extension.

Puzzled-gdb-debugging-by: aggelos@