]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoThis commit was generated by cvs2svn to compensate for changes in r179193,
John Birrell [Thu, 22 May 2008 07:04:10 +0000 (07:04 +0000)]
This commit was generated by cvs2svn to compensate for changes in r179193,
which included commits to RCS files with non-trunk default branches.

16 years agoThis commit was generated by cvs2svn to compensate for changes in r179191,
John Birrell [Thu, 22 May 2008 07:00:07 +0000 (07:00 +0000)]
This commit was generated by cvs2svn to compensate for changes in r179191,
which included commits to RCS files with non-trunk default branches.

16 years agoVendor import (really!) of the OpenSolaris bits for DTrace.
John Birrell [Thu, 22 May 2008 07:00:07 +0000 (07:00 +0000)]
Vendor import (really!) of the OpenSolaris bits for DTrace.

16 years agoCreate the bucket mutexes with MTX_NOWITNESS. There's now a
Marcel Moolenaar [Thu, 22 May 2008 06:27:46 +0000 (06:27 +0000)]
Create the bucket mutexes with MTX_NOWITNESS. There's now a
hard limit of 512 pending mutexes in the witness code and
we can easily have 1 million bucket mutexes initialized before
witness is up and running. Bumping the limit from 512 to 1M
is not really an option here...

16 years agoAdd the DTrace libraries and D library scripts.
John Birrell [Thu, 22 May 2008 04:26:42 +0000 (04:26 +0000)]
Add the DTrace libraries and D library scripts.

16 years agoHook libdwarf and libproc into the build.
John Birrell [Thu, 22 May 2008 04:22:43 +0000 (04:22 +0000)]
Hook libdwarf and libproc into the build.

16 years agoAdd a BSD licensed DWARF library for use by the DTrace clients.
John Birrell [Thu, 22 May 2008 02:14:23 +0000 (02:14 +0000)]
Add a BSD licensed DWARF library for use by the DTrace clients.

The API for this library is deliberately different to the GPL'd
libdwarf to avoid licensing problems.

16 years agoAdd the Makefile for the process library which was missed in the previous
John Birrell [Thu, 22 May 2008 02:10:14 +0000 (02:10 +0000)]
Add the Makefile for the process library which was missed in the previous
commit.

16 years agoAdd a process library with some stubs that the DTrace client needs.
John Birrell [Thu, 22 May 2008 02:09:21 +0000 (02:09 +0000)]
Add a process library with some stubs that the DTrace client needs.
These will be fleshed out as part of the DTrace userland tracing
development.

For now, the kernel tracing part of DTrace requires minimal functionality
for this library.

The API for this library is deliberately different from the libproc in
OpenSolaris due to licensing restrictions.

16 years agoAdd support for the Compact C Type (CTF) conversions throughout FreeBSD's
John Birrell [Thu, 22 May 2008 01:14:43 +0000 (01:14 +0000)]
Add support for the Compact C Type (CTF) conversions throughout FreeBSD's
system makefiles.

Note that the CTF conversion defaults to off. We may choose to change this
default later if DTrace proves popular and people are prepared to wear
the compilation performance impact of compiling with debug symbols all the
time.

Setting NO_CTF in the make args or user environment turns off CTF conversion.
Even if we choose to default CTF generation to on later, we still need
NO_CTF so that the buildworld process can bootstrap the tools without
needlessly generating CTF data for temporary tools.

Setting WITH_CTF in the make args or user environment (and _NOT_ in
/etc/make.conf) is the only way to enable CTF data conversion. Nore that
this can't be implemented the same way that the WITH_ and WITHOUT_ stuff
is implemented throughout the buildworld because the CTF conversion needs
to work when building a simple object without a Makefile, using the
default rules in sys.mk.

Typing 'make test.o' with no makefile and just a source file test.c
should work. Also, typing 'make WITH_CTF=1 test.o without a makefile and
just a source file test.c should work and produce an object with a CTF
elf section. Typing 'make WITH_CTF=1 CFLAGS=-g test.o' without a makefile
and just a source file test.c should produce an object with both a CTF
elf section and the debug elf sections.

In the FreeBSD build where more .mk files are used than just sys.mk
which is included my make by default, the use of DEBUG_FLAGS is the
correct way to enable a debug build. The important thing to note here
is that it is the DEBUG_FLAGS setting that prevents libraries and
programs from being stripped on installation. So, for the addition of
CTF data conversion, setting DEBUG_FLAGS to contain -g, without NO_CTF,
will cause the ctfconvert and ctfmerge build programs to be executed
also with the -g arg so that debug symbols are retained rather than
being removed after the CTF data elf section has been added.

Add DTrace libraries to the list of libnames.

16 years agoGiant is no longer required and can not be held over kproc_exit().
Andrew Thompson [Wed, 21 May 2008 23:04:10 +0000 (23:04 +0000)]
Giant is no longer required and can not be held over kproc_exit().

16 years agoThanks to report from Neil Hoggarth I found a missing UNLOCK in
Jack F Vogel [Wed, 21 May 2008 21:06:17 +0000 (21:06 +0000)]
Thanks to report from Neil Hoggarth I found a missing UNLOCK in
the watchdog code. This delta also incorporates some missing PCI
IDs that got added.

PR 122928 - might be fixed by this, no verification at this point.

16 years ago- sctputil.c - If debug is on, the INPKILL timer can deref a freed value.
Randall Stewart [Wed, 21 May 2008 16:51:21 +0000 (16:51 +0000)]
- sctputil.c - If debug is on, the INPKILL timer can deref a freed value.
               Change so that we save off a type field for display and
               NULL inp just for good measure.

- sctp_output.c - Fix it so in sending to the loopback we use the
                  src address of the inbound INIT. We don't want
                  to do this for non local addresses since otherwise
                  we might be ingressed filtered so we need to use
                  the best src address and list the address sent to.

Obtained from: time bug - Neil Wilson
MFC after: 1 week

16 years agoDon't use LK_DRAIN before calling VOP_FSYNC() in the two further
Robert Watson [Wed, 21 May 2008 13:59:05 +0000 (13:59 +0000)]
Don't use LK_DRAIN before calling VOP_FSYNC() in the two further
panic cases for audit trail failure -- this doesn't contribute
anything, and might arguably be wrong.

MFC after: 1 week
Requested by: attilio

16 years agoDon't use LK_DRAIN before calling VOP_FSYNC() in the panic case for
Robert Watson [Wed, 21 May 2008 13:05:06 +0000 (13:05 +0000)]
Don't use LK_DRAIN before calling VOP_FSYNC() in the panic case for
audit trail failure -- this doesn't contribute anything, and might
arguably be wrong.

MFC after: 1 week
Requested by: attilio

16 years agoImplement the per-open file data for the cdev.
Konstantin Belousov [Wed, 21 May 2008 09:31:44 +0000 (09:31 +0000)]
Implement the per-open file data for the cdev.

The patch does not change the cdevsw KBI. Management of the data is
provided by the functions
int devfs_set_cdevpriv(void *priv, cdevpriv_dtr_t dtr);
int devfs_get_cdevpriv(void **datap);
void devfs_clear_cdevpriv(void);
All of the functions are supposed to be called from the cdevsw method
contexts.

- devfs_set_cdevpriv assigns the priv as private data for the file
  descriptor which is used to initiate currently performed driver
  operation. dtr is the function that will be called when either the
  last refernce to the file goes away, the device is destroyed  or
  devfs_clear_cdevpriv is called.
- devfs_get_cdevpriv is the obvious accessor.
- devfs_clear_cdevpriv allows to clear the private data for the still
  open file.

Implementation keeps the driver-supplied pointers in the struct
cdev_privdata, that is referenced both from the struct file and struct
cdev, and cannot outlive any of the referee.

Man pages will be provided after the KPI stabilizes.

Reviewed by: jhb
Useful suggestions from: jeff, antoine
Debugging help and tested by: pho
MFC after: 1 month

16 years ago+ Use the default cscope names.
David E. O'Brien [Wed, 21 May 2008 07:56:10 +0000 (07:56 +0000)]
+ Use the default cscope names.
+ Enable fast symbol lookup.
+ Make sure files cscoped are really files, and add .S's.
+ Add 'cscope-clean' target.

16 years agoWe can call ia64_flush_dirty() when the corresponding process is
Marcel Moolenaar [Wed, 21 May 2008 05:15:27 +0000 (05:15 +0000)]
We can call ia64_flush_dirty() when the corresponding process is
locked or not. As such, use PROC_LOCKED() to determine which case
it is and lock the process when not.

16 years agoUse the "options " spelling (vs. "options<TAB>") so that commented lines
David E. O'Brien [Wed, 21 May 2008 03:36:53 +0000 (03:36 +0000)]
Use the "options " spelling (vs. "options<TAB>") so that commented lines
line up nicely.

16 years agoAdd Solaris compatibility definitions for things we don't really want in
John Birrell [Wed, 21 May 2008 00:26:38 +0000 (00:26 +0000)]
Add Solaris compatibility definitions for things we don't really want in
FreeBSD headers.

16 years agoAdd Juniper's copyright.
David E. O'Brien [Tue, 20 May 2008 22:58:47 +0000 (22:58 +0000)]
Add Juniper's copyright.

16 years agoMIPS arch target kgdb(1) support.
David E. O'Brien [Tue, 20 May 2008 22:54:42 +0000 (22:54 +0000)]
MIPS arch target kgdb(1) support.

Obtained from: Juniper Networks.

16 years ago-d is a flag, not an argument. Use .Fl here.
Andrew Thompson [Tue, 20 May 2008 20:47:21 +0000 (20:47 +0000)]
-d is a flag, not an argument. Use .Fl here.

16 years agoAllow VM object creation in ufs_lookup. (If vfs.vmiodirenable is set)
Stephan Uphoff [Tue, 20 May 2008 19:05:43 +0000 (19:05 +0000)]
Allow VM object creation in ufs_lookup. (If vfs.vmiodirenable is set)
Directory IO without a VM object will store data in 'malloced' buffers
severely limiting caching of the data. Without this  change VM objects for
directories are only created on an open() of the directory.
TODO: Inline test if VM object already exists to avoid locking/function call
overhead.

Tested by: kris@
Reviewed by: jeff@
Reported by: David Filo

16 years ago- Adds support for the multi-asconf (From Kozuka-san)
Randall Stewart [Tue, 20 May 2008 13:47:46 +0000 (13:47 +0000)]
- Adds support for the multi-asconf (From Kozuka-san)
- Adds some prepwork (Not all yet) for vimage in particular
  support the delete the sctppcbinfo.xx structs. There is
  still a leak in here if it were to be called plus we stil
  need the regrouping (From Me and Michael Tuexen)
- Adds support for UDP tunneling. For BSD there is no
  socket yet setup so its disabled, but major argument
  changes are in here to emcompass the passing of the port
  number (zero when you don't have a udp tunnel, the default
  for BSD). Will add some hooks in UDP here shortly (discussed
  with Robert) that will allow easy tunneling. (Mainly from
  Peter Lei and Michael Tuexen with some BSD work from me :-D)
- Some ease for windows, evidently leave is reserved by their
  compile move label leave: -> out:

MFC after: 1 week

16 years agoDrop Giant before calling kproc_exit(), so we don't trigger assertion in
Pawel Jakub Dawidek [Tue, 20 May 2008 12:34:30 +0000 (12:34 +0000)]
Drop Giant before calling kproc_exit(), so we don't trigger assertion in
kproc_exit().

Discussed with: imp

16 years agoAdd ID for T61p, which is not IBM really, but well.
Pawel Jakub Dawidek [Tue, 20 May 2008 12:26:45 +0000 (12:26 +0000)]
Add ID for T61p, which is not IBM really, but well.

16 years ago- Change the meaning of -h flag from giving the output in megabytes to
Pawel Jakub Dawidek [Tue, 20 May 2008 12:24:31 +0000 (12:24 +0000)]
- Change the meaning of -h flag from giving the output in megabytes to
  giving the output in a human-readable form. This behaviour is consistent
  with most of system tools.
- Add -m and -g options to give output in megabytes and gigabytes
  respectively.

16 years agoUpdate to reflect reality:
Rui Paulo [Tue, 20 May 2008 12:07:02 +0000 (12:07 +0000)]
Update to reflect reality:
* iasl(8) supports ACPI 3.0b.
* Add new options.

MFC after: 1 week

16 years agoUse 'dowrite' as an argument name instead of too general 'write'.
Pawel Jakub Dawidek [Tue, 20 May 2008 11:45:05 +0000 (11:45 +0000)]
Use 'dowrite' as an argument name instead of too general 'write'.

16 years agoForce commit to note, that the typo in KASSERT() was:
Pawel Jakub Dawidek [Tue, 20 May 2008 11:43:56 +0000 (11:43 +0000)]
Force commit to note, that the typo in KASSERT() was:

PR: kern/116896
Reported by: VANHULLEBUS Yvan <vanhu@netasq.com>

16 years ago- Define changes in sctp.h
Randall Stewart [Tue, 20 May 2008 09:51:36 +0000 (09:51 +0000)]
- Define changes in sctp.h
- Bug in CA that does not get us incrementing the PBA properly which
  made us more conservative.
- comment updated in sctp_input.c
- memsets added before we log
- added arg to hmac id's
MFC after: 2 weeks

16 years agoDocument the azp for controlling audio power
Kevin Lo [Tue, 20 May 2008 06:33:10 +0000 (06:33 +0000)]
Document the azp for controlling audio power

16 years agoAdd support for controlling the audio power
Kevin Lo [Tue, 20 May 2008 06:32:07 +0000 (06:32 +0000)]
Add support for controlling the audio power

Tested on: VGN-SZ79TN

16 years agoThis small change will allow this driver in HEAD to build
Jack F Vogel [Mon, 19 May 2008 23:01:11 +0000 (23:01 +0000)]
This small change will allow this driver in HEAD to build
on 6.3 as well as 7 :)

16 years agoDetect if argv[1] is "" and avoid calling malloc(0). Prior to this commit,
Colin Percival [Mon, 19 May 2008 18:38:01 +0000 (18:38 +0000)]
Detect if argv[1] is "" and avoid calling malloc(0).  Prior to this commit,
running 'tar ""' would print 'No memory' instead of the correct error
message, 'Must specify one of -c, -r, -t, -u, -x' if malloc is set to
System V mode (malloc(0) == NULL).

16 years agoThere's no way for entry to possibly be NULL at the end of write_entry
Colin Percival [Mon, 19 May 2008 18:09:26 +0000 (18:09 +0000)]
There's no way for entry to possibly be NULL at the end of write_entry
(in fact, there has never been any way for it to be NULL, going all the
way back to revision 1.1 of this file), so remove the check and
unconditionally free entry.

Found by: Coverity Prevent

16 years agoReturn ARCHIVE_FATAL if we can't allocate memory instead of going ahead and
Colin Percival [Mon, 19 May 2008 18:06:48 +0000 (18:06 +0000)]
Return ARCHIVE_FATAL if we can't allocate memory instead of going ahead and
dereferencing NULL.

Found by: Coverity Prevent

16 years agoRemove -, .Ar adds that automatically.
Andrew Thompson [Mon, 19 May 2008 17:59:09 +0000 (17:59 +0000)]
Remove -, .Ar adds that automatically.

Pointed out by: brueffer

16 years agoBump document date for last change.
Andrew Thompson [Mon, 19 May 2008 17:54:32 +0000 (17:54 +0000)]
Bump document date for last change.

16 years agoFlush stdout after each line to make piping to another command easier.
Andrew Thompson [Mon, 19 May 2008 17:51:00 +0000 (17:51 +0000)]
Flush stdout after each line to make piping to another command easier.

16 years agoAllow wlandebug to set the default debug level which is inherited when vaps are
Andrew Thompson [Mon, 19 May 2008 17:49:57 +0000 (17:49 +0000)]
Allow wlandebug to set the default debug level which is inherited when vaps are
created.

16 years agoAdd -L to usage().
Bruce M Simpson [Mon, 19 May 2008 11:35:11 +0000 (11:35 +0000)]
Add -L to usage().

16 years agoAdd an -L option to ignore loopback Internet sockets.
Bruce M Simpson [Mon, 19 May 2008 11:32:44 +0000 (11:32 +0000)]
Add an -L option to ignore loopback Internet sockets.

MFC after: 2 weeks

16 years agoAdd age(4) to the list of drivers supported by GENERIC kernel.
Pyun YongHyeon [Mon, 19 May 2008 02:30:27 +0000 (02:30 +0000)]
Add age(4) to the list of drivers supported by GENERIC kernel.

16 years agoHook up age(4) to the build.
Pyun YongHyeon [Mon, 19 May 2008 02:21:00 +0000 (02:21 +0000)]
Hook up age(4) to the build.

16 years agoAdd age(4) to the list of supported network interface.
Pyun YongHyeon [Mon, 19 May 2008 02:17:24 +0000 (02:17 +0000)]
Add age(4) to the list of supported network interface.

16 years agoAdd an entry for the age(4) module.
Pyun YongHyeon [Mon, 19 May 2008 02:13:50 +0000 (02:13 +0000)]
Add an entry for the age(4) module.

16 years agoAdd age(4) to the list of drivers use the miibus interface.
Pyun YongHyeon [Mon, 19 May 2008 02:10:39 +0000 (02:10 +0000)]
Add age(4) to the list of drivers use the miibus interface.

16 years agoAdd age(4) to the list of drivers supporting ALTQ.
Pyun YongHyeon [Mon, 19 May 2008 02:06:39 +0000 (02:06 +0000)]
Add age(4) to the list of drivers supporting ALTQ.

16 years agoAdd age(4) to the list of devices supporting full VLAN processing
Pyun YongHyeon [Mon, 19 May 2008 02:03:52 +0000 (02:03 +0000)]
Add age(4) to the list of devices supporting full VLAN processing
in hardware.

16 years agoAdd man page for age(4).
Pyun YongHyeon [Mon, 19 May 2008 02:00:16 +0000 (02:00 +0000)]
Add man page for age(4).
brueffer fixed various nits in the document.

Reviewed by: brueffer

16 years agoHook up age(4) to the build.
Pyun YongHyeon [Mon, 19 May 2008 01:53:47 +0000 (01:53 +0000)]
Hook up age(4) to the build.

16 years agoAdd age(4), a driver for Attansic/Atheros L1 gigabit ethernet
Pyun YongHyeon [Mon, 19 May 2008 01:39:59 +0000 (01:39 +0000)]
Add age(4), a driver for Attansic/Atheros L1 gigabit ethernet
controller. L1 has several threshold/timer registers and they
seem to require careful tuned parameters to get best
performance. Datasheet for L1 is not available to open source
driver writers so age(4) focus on stability and correctness of
basic Tx/Rx operation. ATM the performance of age(4) is far from
optimal which in turn means there are mis-programmed registers or
incorrectly configured registers.
Currently age(4) supports all known hardware assistance including
  - MSI support.
  - TCP Segmentation Offload.
  - Hardware VLAN tag insertion/stripping.
  - TCP/UDP checksum offload.
  - Interrupt moderation.
  - Hardware statistics counter support.
  - Jumbo frame support.
  - WOL support.

L1 gigabit ethernet controller is mainly found on ASUS
motherboards. Note, it seems that there are other variants of
hardware as known as L2(Fast ethernet) and newer gigabit ethernet
(AR81xx) from Atheros. These are not supported by age(4) and
requires a seperate driver. Big thanks to all people who reported
feedback or tested patches.

Tested by: kevlo, bsam, Francois Ranchin < fyr AT fyrou DOT net >
Thomas Nystroem < thn AT saeab DOT se >
Roman Pogosyan < asternetadmin AT gmail DOT com >
Derek Tattersal < dlt AT mebtel DOT net >
Oliver Seitz < karlkiste AT yahoo DOT com >

16 years agoConnect atphy(4) to the build.
Pyun YongHyeon [Mon, 19 May 2008 01:18:02 +0000 (01:18 +0000)]
Connect atphy(4) to the build.

16 years agoAdd Attansic/Atheros F1 PHY driver.
Pyun YongHyeon [Mon, 19 May 2008 01:12:10 +0000 (01:12 +0000)]
Add Attansic/Atheros F1 PHY driver.

16 years ago- Assert that we don't send new provider event for a provider which has
Pawel Jakub Dawidek [Sun, 18 May 2008 22:50:50 +0000 (22:50 +0000)]
- Assert that we don't send new provider event for a provider which has
  G_PF_WITHER flag set.
- Fix typo in assertion condition (sorry, but I forgot who report that).

16 years agoRemove some DTrace hook definitions that are now in dtrace_bsd.h
John Birrell [Sun, 18 May 2008 22:11:47 +0000 (22:11 +0000)]
Remove some DTrace hook definitions that are now in dtrace_bsd.h
which contains all the hook definitions rather than splattering
them all over the header files.

The definitions are only valid when the KDTRACE_HOOKS kernel
option is defined, so other kernel sources have no need to
see them.

16 years agoAdd sysinit levels for DTrace.
John Birrell [Sun, 18 May 2008 22:10:10 +0000 (22:10 +0000)]
Add sysinit levels for DTrace.

16 years agoPlay nice with DDB pager.
Pawel Jakub Dawidek [Sun, 18 May 2008 21:13:10 +0000 (21:13 +0000)]
Play nice with DDB pager.

Educated by: jhb's BSDCan presentation

16 years agoBe more friendly for DDB pager.
Pawel Jakub Dawidek [Sun, 18 May 2008 21:08:12 +0000 (21:08 +0000)]
Be more friendly for DDB pager.

Educated by: jhb's BSDCan presentation

16 years agoAdd the hooks for the extra data that DTrace allocates for struct thread
John Birrell [Sun, 18 May 2008 19:50:43 +0000 (19:50 +0000)]
Add the hooks for the extra data that DTrace allocates for struct thread
and struct proc.

Add a field to struct thread to stash the error variable (or returned
status) from the last syscall so that it is available during a
DTrace probe.

16 years agoAdd a couple of files which depend of the KDTRACE_HOOKS option.
John Birrell [Sun, 18 May 2008 19:47:49 +0000 (19:47 +0000)]
Add a couple of files which depend of the KDTRACE_HOOKS option.

The syscall names are required by KDTRACE_HOOKS too.

And the unzip

16 years agoAdd support for the DTrace struct proc and struct thread extended
John Birrell [Sun, 18 May 2008 19:43:52 +0000 (19:43 +0000)]
Add support for the DTrace struct proc and struct thread extended
data via ctor and dtor event handlers.

The size of the extra data is allocated opaquely and this file
contains a function which the dtrace module can call to check
that the kernel supports at least the amount of data that it needs.

This file is optionally compiled into nthe kernel if the KDTRACE_HOOKS
kernel option is defined.

16 years agoAdd kernel support for the Statically Defined Trace provider.
John Birrell [Sun, 18 May 2008 19:32:36 +0000 (19:32 +0000)]
Add kernel support for the Statically Defined Trace provider.

This is BSD licensed code written specifically for FreeBSD.

It initialises using SYSINIT so that the SDT provider, probe and
argument description linkage is done whenever a module is loaded,
regardless of whether the DTrace modules are loaded or not.

This file is optionally compiled into the kernel if the KDTRACE_HOOKS
option is defined.

16 years agoAdd two kernel options:
John Birrell [Sun, 18 May 2008 19:28:51 +0000 (19:28 +0000)]
Add two kernel options:

- KDTRACE_HOOKS for the shim layer of hooks which separate BSD licensed
                code from CDDL code.
- DDB_CTF       for the code that parses the CTF (compact C type format)
                data for use by the DTrace Function Boundary Trace
                provider and (possibly) ddb if we plan to do that.

16 years agodevctl_process_running(): Check for devsoftc.inuse == 1 instead of
Rui Paulo [Sun, 18 May 2008 13:55:51 +0000 (13:55 +0000)]
devctl_process_running(): Check for devsoftc.inuse == 1 instead of
devsoftc.async_proc != NULL because the latter might not be true
sometimes.
This way /etc/rc.suspend gets executed.

Reviwed by: njl
Submitted by: Mitsuru IWASAKI <iwasaki at jp.FreeBSD.org>
Tested also by: Andreas Wetzel <mickey242 at gmx.net>
MFC after: 1 week

16 years agoRemove the unknown device that is breaking the tinderbox build.
John Birrell [Sun, 18 May 2008 11:08:26 +0000 (11:08 +0000)]
Remove the unknown device that is breaking the tinderbox build.

16 years agoAdd myself.
Rui Paulo [Sun, 18 May 2008 11:05:41 +0000 (11:05 +0000)]
Add myself.

16 years agoSimple testing program for *at family of syscalls.
Roman Divacky [Sun, 18 May 2008 08:34:04 +0000 (08:34 +0000)]
Simple testing program for *at family of syscalls.

Approved by: kib (mentor)

16 years agoAdd SIGINFO (and for portability to SIGINFO-lacking systems, SIGUSR1)
Colin Percival [Sun, 18 May 2008 06:24:47 +0000 (06:24 +0000)]
Add SIGINFO (and for portability to SIGINFO-lacking systems, SIGUSR1)
handling to bsdtar.  When writing archives (including copying via the
@archive directive) a line is output to stderr indicating what is being
done (adding or copying), the path, and how far through the file we are;
extracting currently does not report progress within each file, but
this is likely to happen eventually.

Discussed with: kientzle
Obtained from: tarsnap

16 years agoRetire pmap_addr_hint(). It is no longer used.
Alan Cox [Sun, 18 May 2008 04:16:57 +0000 (04:16 +0000)]
Retire pmap_addr_hint().  It is no longer used.

16 years agoAdd support for the Epson 4800 scanner.
Remko Lodder [Sun, 18 May 2008 03:07:40 +0000 (03:07 +0000)]
Add support for the Epson 4800 scanner.

PR: 118391
Submitted by: "Pedro F. Giffuni" <giffunip at tutopia dot com>
Approved by: imp (mentor, implicit)
MFC after: 3 days
Committed at: BSDCan 2008

16 years agoMove the wait for a default route to rc.d/routing. Once we test for
Brooks Davis [Sun, 18 May 2008 02:57:54 +0000 (02:57 +0000)]
Move the wait for a default route to rc.d/routing.  Once we test for
non-dhcp interfaces to negotiate/associate this will make more sense.

This also correctly gets run after both devd and netif are run so it has
a chance of working.

16 years agoResort the if_ti driver to match the PCI Network cards instead of placing
Remko Lodder [Sat, 17 May 2008 23:50:00 +0000 (23:50 +0000)]
Resort the if_ti driver to match the PCI Network cards instead of placing
it under the mii devices list.

PR: kern/123147
Submitted by: gavin
Approved by: imp (mentor, implicit)
MFC after: 3 days

16 years agoIn order to map device memory using superpages, mmap(2) must find a
Alan Cox [Sat, 17 May 2008 19:32:48 +0000 (19:32 +0000)]
In order to map device memory using superpages, mmap(2) must find a
superpage-aligned virtual address for the mapping.  Revision 1.65
implemented an overly simplistic and generally ineffectual method for
finding a superpage-aligned virtual address.  Specifically, it rounds
the virtual address corresponding to the end of the data segment up to
the next superpage-aligned virtual address.  If this virtual address
is unallocated, then the device will be mapped using superpages.
Unfortunately, in modern times, where applications like the X server
dynamically load much of their code, this virtual address is already
allocated.  In such cases, mmap(2) simply uses the first available
virtual address, which is not necessarily superpage aligned.

This revision changes mmap(2) to use a more robust method,
specifically, the VMFS_ALIGNED_SPACE option that is now implemented by
vm_map_find().

16 years agoPreset a device object's alignment ("pg_color") based upon the
Alan Cox [Sat, 17 May 2008 16:26:34 +0000 (16:26 +0000)]
Preset a device object's alignment ("pg_color") based upon the
physical address of the device's memory.  This enables
pmap_align_superpage() to propose a virtual address for mapping the
device memory that permits the use of superpage mappings.

16 years agoAdd support for specifying which INDEX files to build via portsnap.conf.
Colin Percival [Sat, 17 May 2008 16:26:27 +0000 (16:26 +0000)]
Add support for specifying which INDEX files to build via portsnap.conf.

Requested by: brooks
Reminded by: brooks, about halfway through his BSDCan talk

16 years agoAdd --keep-newer-files option (as in GNU tar: When in -x mode, ignore
Colin Percival [Sat, 17 May 2008 15:55:29 +0000 (15:55 +0000)]
Add --keep-newer-files option (as in GNU tar: When in -x mode, ignore
files if the existing file is newer than the archive entry).

Currently if any files are ignored, bsdtar will exit with a non-zero
exit status; this is likely to change in the future, but requires some
API changes in libarchive.

Discussed with: kientzle
Obtained from: tarsnap

16 years agoRemove last bits of OS adaptation code from the IPSec code.
George V. Neville-Neil [Sat, 17 May 2008 04:00:11 +0000 (04:00 +0000)]
Remove last bits of OS adaptation code from the IPSec code.

Reviewed By: bz

16 years agoThe if_check() function performed three actions:
Brooks Davis [Sat, 17 May 2008 03:38:13 +0000 (03:38 +0000)]
The if_check() function performed three actions:
 - verified that the ifp->if_snd.ifq_mtx was initalized for
   all attached interfaces.  This was pointless because it was
   initalized for all interfaces in if_attach() so I've removed it.
 - Checked that ifp->if_snd.ifq_maxlen is initalized and set it to
   ifqmaxlen if unset.  This makes more sense in if_attach() so
   I moved it there.
 - The first call of if_slowtimo().  Delete if_check() and call
   if_slowtimo() directly from the SYSINIT().

16 years agoAdd the DTrace kernel module makefiles.
John Birrell [Sat, 17 May 2008 02:31:19 +0000 (02:31 +0000)]
Add the DTrace kernel module makefiles.

16 years agoAdd the DTrace shim layer definitions.
John Birrell [Sat, 17 May 2008 02:16:58 +0000 (02:16 +0000)]
Add the DTrace shim layer definitions.

All shim hooks are defined here. This is the interface between BSD
code in FreeBSD and CDDL code from OpenSolaris.

The hooks defined here are pre-processed out from the source files
when the KDTRACE_HOOKS kernel option isn't defined.

16 years agoAdd the statically defined tracing header.
John Birrell [Sat, 17 May 2008 02:14:19 +0000 (02:14 +0000)]
Add the statically defined tracing header.

Note that this implementation differs from the one in OpenSolaris, so
it is BSD licensed and can be included anywhere.

The kernel definitions defined here are dependent on the kernel option
KDTRACE_HOOKS so that macros added to the sources are pre-processed
out completely when the DTrace kernel hooks aren't compiled in.

16 years agoAdd the DTrace test makefile for the test suite distributed in OpenSolaris.
John Birrell [Sat, 17 May 2008 02:09:48 +0000 (02:09 +0000)]
Add the DTrace test makefile for the test suite distributed in OpenSolaris.

16 years agoFix and speedup timestamp calculations which is roughly based on the patch in
Markus Brueffer [Fri, 16 May 2008 22:31:17 +0000 (22:31 +0000)]
Fix and speedup timestamp calculations which is roughly based on the patch in
the mentioned PR:

- bounds check time->month as it is used as an array index
- fix usage of time->month as array index (month is 1-12)
- fix calculation based on time->day (day is 1-31)
- fix the speedup code as it doesn't calculate correct timestamps before
  the year 2000 and reduce the number of calculation in the year-by-year code
- speedup month calculations by replacing the array content with cumulative
  values
- add microseconds calculation
- fix an endian problem

PR: kern/97786
Submitted by: Andriy Gapon <avg@topspin.kiev.ua>
Reviewed by: scottl (earlier version)
Approved by: emax (mentor)
MFC after: 1 week

16 years agoRetire some stale alpha references.
John Baldwin [Fri, 16 May 2008 20:09:29 +0000 (20:09 +0000)]
Retire some stale alpha references.

16 years agoAdd support for the Nikon D300 camera
Remko Lodder [Fri, 16 May 2008 19:47:15 +0000 (19:47 +0000)]
Add support for the Nikon D300 camera

PR: usb/118741
Submitted by: Yuri <yuri at tsoft dot com>
Approved by: imp (mentor, implicit)
MFC after: 3 days

16 years agoThis is driver version 1.4.4 of the Intel ixgbe driver.
Jack F Vogel [Fri, 16 May 2008 18:46:30 +0000 (18:46 +0000)]
This is driver version 1.4.4 of the Intel ixgbe driver.
  -It has new hardware support
  -It uses a new method of TX cleanup called Head Write Back
  -It includes the provisional generic TCP LRO feature contributed
   by Myricom and made general purpose by me. This should move into
   the stack upon approval but for this driver drop its in here.
  -Also bug fixes and etc...

MFC in a week if no serious issues arise.

16 years agoAttempt to improve convergence of POSIX semaphore code with style(9).
Robert Watson [Fri, 16 May 2008 18:10:07 +0000 (18:10 +0000)]
Attempt to improve convergence of POSIX semaphore code with style(9).

MFC after: 3 days

16 years agoTeach truss about 32-bit FreeBSD and Linux binaries on amd64. Some
John Baldwin [Fri, 16 May 2008 15:34:06 +0000 (15:34 +0000)]
Teach truss about 32-bit FreeBSD and Linux binaries on amd64.  Some
additional work is needed to handle ABI-specific syscall argument parsing,
but this gets the basic tracing working.

MFC after: 1 week

16 years agoFix RID calculation. The RID is really the BAR for PCI cards,
Marcel Moolenaar [Fri, 16 May 2008 14:57:48 +0000 (14:57 +0000)]
Fix RID calculation. The RID is really the BAR for PCI cards,
so the index needs to be translated into an offset. While we
did add the offset (0x10), we forgot to account for the width.

Tested by: Thomas Vogt
MFC after: 3 days

16 years agoRemoved unused assembly offsets for structures digging.
Attilio Rao [Fri, 16 May 2008 13:23:47 +0000 (13:23 +0000)]
Removed unused assembly offsets for structures digging.

16 years ago- The names of the months and the days are in lowercase according to the
Gabor Kovesdan [Fri, 16 May 2008 12:50:15 +0000 (12:50 +0000)]
- The names of the months and the days are in lowercase according to the
  Hungarian spelling, change them accordingly

Requested by: Janos Mohacsi <mohacsi@niif.hu>

16 years agoRemove a check that didn't allow > 12 byte CDB's to be issued to
Paul Saab [Fri, 16 May 2008 08:27:02 +0000 (08:27 +0000)]
Remove a check that didn't allow > 12 byte CDB's to be issued to
ciss.  This should allow volumes > 2TB to work.

Reported by: Emil Mikulic

16 years agoDocument BOOTP_BLOCKSIZE.
Benno Rice [Fri, 16 May 2008 06:50:40 +0000 (06:50 +0000)]
Document BOOTP_BLOCKSIZE.

16 years agoAllow the block size used when booting over NFS to be overridden. It defaults
Benno Rice [Fri, 16 May 2008 06:27:03 +0000 (06:27 +0000)]
Allow the block size used when booting over NFS to be overridden.  It defaults
to 8192 bytes which is the size currently used.

16 years agoClarify that "ante meridiem" and "post meridiem" mean the same thing
Greg Lehey [Fri, 16 May 2008 04:33:04 +0000 (04:33 +0000)]
Clarify that "ante meridiem" and "post meridiem" mean the same thing
as the more commonly used "a.m." and "p.m.".

Tripped over by:  Callum Gibson.

MFC after:  2 weeks

16 years agoThe beacon miss notification must run without locks held has it calls back into
Andrew Thompson [Fri, 16 May 2008 04:15:54 +0000 (04:15 +0000)]
The beacon miss notification must run without locks held has it calls back into
wpi_raw_xmit();

16 years agoSpelling and capitalization fixes.
Sean Farley [Fri, 16 May 2008 03:13:36 +0000 (03:13 +0000)]
Spelling and capitalization fixes.

MFC after: 3 days