]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agoA Wittgenstein quote according to David Ignatius
phk [Wed, 8 Oct 2008 06:39:05 +0000 (06:39 +0000)]
A Wittgenstein quote according to David Ignatius
http://voices.washingtonpost.com/postpartisan/2008/10/the_town_hall_debate_in_praise.html

15 years agoRemove stale comment (and XXX saying so) about why we zero the file
rwatson [Wed, 8 Oct 2008 06:26:51 +0000 (06:26 +0000)]
Remove stale comment (and XXX saying so) about why we zero the file
descriptor pointer in unp_freerights: we can no longer recurse into
unp_gc due to unp_gc being invoked in a deferred way, but it's still
a good idea.

MFC after: 3 days

15 years agoDifferentiate pr_usrreqs for stream and datagram UNIX domain sockets, and
rwatson [Wed, 8 Oct 2008 06:19:49 +0000 (06:19 +0000)]
Differentiate pr_usrreqs for stream and datagram UNIX domain sockets, and
employ soreceive_dgram for the datagram case.

MFC after: 3 months

15 years agoDon't mention lib/compat, it has gone long ago. Use ports/misc/compat*
delphij [Wed, 8 Oct 2008 01:31:00 +0000 (01:31 +0000)]
Don't mention lib/compat, it has gone long ago.  Use ports/misc/compat*
instead.

15 years agoPass bus_get_dma_tag() into the first argument of bus_dma_create_tag()
imp [Tue, 7 Oct 2008 23:55:32 +0000 (23:55 +0000)]
Pass bus_get_dma_tag() into the first argument of bus_dma_create_tag()
rather than NULL.

15 years agoUpdate udpzerobyte to understand that passing 0 as a length to recv(2)
rwatson [Tue, 7 Oct 2008 21:01:23 +0000 (21:01 +0000)]
Update udpzerobyte to understand that passing 0 as a length to recv(2)
will cause it to return 0, not EAGAIN.

Add UNIX domain socket support to udpzerobyte, which suggests this
regression test should be moved to the general sockets test area rather
than netinet.

15 years agoIn soreceive_dgram, when a 0-length buffer is passed into recv(2) and
rwatson [Tue, 7 Oct 2008 20:57:55 +0000 (20:57 +0000)]
In soreceive_dgram, when a 0-length buffer is passed into recv(2) and
no data is ready, return 0 rather than blocking or returning EAGAIN.
This is consistent with the behavior of soreceive_generic (soreceive)
in earlier versions of FreeBSD, and restores this behavior for UDP.

Discussed with: jhb, sam
MFC after: 3 days

15 years agothe 3945 firmware license does not require acknowledging the EULA,
sam [Tue, 7 Oct 2008 19:42:56 +0000 (19:42 +0000)]
the 3945 firmware license does not require acknowledging the EULA,
remove the sysctl ack requirement

Pointed out by: Joerg Sonnenberger
Reviewed by: thompsa
Approved by: core

15 years agoMore diff reductions against ixp425/boot2/boot2.c. This time, we
imp [Tue, 7 Oct 2008 17:44:04 +0000 (17:44 +0000)]
More diff reductions against ixp425/boot2/boot2.c.  This time, we
bring in FIXUP_BOOT_DRV functionality as an #ifdef.  This is not
enabled at this time, and the md5 remains constant with this change.
Apart from the 'accept any partitioning scheme on the device' changes,
this was the biggest delta...

# and yes, we'll merge these into one source file if we can do that in a
# way that makes sense.

Obtained from: sys/boot/arm/ixp425/boot2/boot2.c

15 years ago- Avoid a spurious error when a command line is specified without
jkoshy [Tue, 7 Oct 2008 17:28:52 +0000 (17:28 +0000)]
- Avoid a spurious error when a command line is specified without
  any process scope PMCs.
- Change the -c '*' option to denote non-halted CPUs.

15 years agoDiff reduction with boot/arm/at91/boot2/boot2.c: indent this statement
imp [Tue, 7 Oct 2008 17:27:37 +0000 (17:27 +0000)]
Diff reduction with boot/arm/at91/boot2/boot2.c: indent this statement
correctly.

15 years agoUse bus_get_dma_tag() instead of NULL here. Not really needed for atmel
imp [Tue, 7 Oct 2008 17:23:16 +0000 (17:23 +0000)]
Use bus_get_dma_tag() instead of NULL here.  Not really needed for atmel
at the moment, but it is more correct.

15 years agodon't hardcode cc
sam [Tue, 7 Oct 2008 16:39:17 +0000 (16:39 +0000)]
don't hardcode cc

Submitted by: Andrey Eltsov

15 years agoAdd IPv6 support to zero-size UDP transmit/receive test.
rwatson [Tue, 7 Oct 2008 14:13:59 +0000 (14:13 +0000)]
Add IPv6 support to zero-size UDP transmit/receive test.

15 years agoUse CPUID to see if the current CPU supports long mode when attemping to
jhb [Tue, 7 Oct 2008 14:05:42 +0000 (14:05 +0000)]
Use CPUID to see if the current CPU supports long mode when attemping to
boot an amd64 kernel.  If not, then fail the boot request with an error
message.  Otherwise, the boot attempt will fail with a BTX fault when
trying to read the EFER MSR.

MFC after: 3 days

15 years agoRegression test for the loopback handling of zero-length UDP packets, which
rwatson [Tue, 7 Oct 2008 10:31:55 +0000 (10:31 +0000)]
Regression test for the loopback handling of zero-length UDP packets, which
should be delivered but without payload.

15 years ago- Allow gstat to print values to different kind of outputs.
lulf [Tue, 7 Oct 2008 10:25:27 +0000 (10:25 +0000)]
- Allow gstat to print values to different kind of outputs.
- Introduce batch mode, where gstat will collect the numbers, print them, and
  exit.
- Document batch mode in the gstat man page.

Submitted by: anders

15 years agoRemove temporary debugging KASSERT's introduced to detect protocols
rwatson [Tue, 7 Oct 2008 09:57:03 +0000 (09:57 +0000)]
Remove temporary debugging KASSERT's introduced to detect protocols
improperly invoking sosend(), soreceive(), and sopoll() instead of
attach either specialized or _generic() versions of those functions
to their pru_sosend, pru_soreceive, and pru_sopoll protosw methods.

MFC after: 3 days

15 years agoRewrite sbreserve_locked()'s comment on NULL thread pointers, eliminating
rwatson [Tue, 7 Oct 2008 09:51:39 +0000 (09:51 +0000)]
Rewrite sbreserve_locked()'s comment on NULL thread pointers, eliminating
an XXXRW about the comment being stale.

MFC after: 3 days

15 years agoDon't pass curthread to sbreserve_locked() in tcp_do_segment(), as the
rwatson [Tue, 7 Oct 2008 09:41:07 +0000 (09:41 +0000)]
Don't pass curthread to sbreserve_locked() in tcp_do_segment(), as the
netisr or ithread's socket buffer size limit is not the right limit to
use.  Instead, pass NULL as the other two calls to sbreserve_locked()
in the TCP input path (tcp_mss()) do.

In practice, this is a no-op, as ithreads and the netisr run without a
process limit on socket buffer use, and a NULL thread pointer leads to
not using the process's limit, if any.  However, if tcp_input() is
called in other contexts that do have limits, this may prevent the
incorrect limit from being used.

MFC after: 3 days

15 years agoLock receive socket buffer in soo_stat() rather than commenting that we
rwatson [Tue, 7 Oct 2008 07:10:28 +0000 (07:10 +0000)]
Lock receive socket buffer in soo_stat() rather than commenting that we
should lock it, which may marginally improve the consistency of the
results.  Remove comment.

MFC after: 3 days

15 years agoLimit DMA memory to lower addressable 4GB, without this patch, we
delphij [Mon, 6 Oct 2008 22:34:17 +0000 (22:34 +0000)]
Limit DMA memory to lower addressable 4GB, without this patch, we
can reliably provoke data corruption on systems equipped with a
plenty of memory during high load.

Reported by: gnn via iXsystems
MFC candidate: RELENG_7_1, RELENG_7

15 years agoReference igb(4) i SEE ALSO.
simon [Mon, 6 Oct 2008 21:55:53 +0000 (21:55 +0000)]
Reference igb(4) i SEE ALSO.

MFC after: 1 week

15 years agoAdd igb(4).
simon [Mon, 6 Oct 2008 21:52:11 +0000 (21:52 +0000)]
Add igb(4).

MFC after: 1 week

15 years agoOnly a few of the loader tunables / sysctl variables are documented,
simon [Mon, 6 Oct 2008 21:38:55 +0000 (21:38 +0000)]
Only a few of the loader tunables / sysctl variables are documented,
as I'm not really sure what the rest do.

The list of adaptors was found at
http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&Inst=Yes&DwnldID=15815

MFC after: 1 week

15 years ago- Add 82574 to list of supported controller chips.
simon [Mon, 6 Oct 2008 20:45:08 +0000 (20:45 +0000)]
- Add 82574 to list of supported controller chips.
- Add Intel PRO/1000 PF and Intel PRO/1000 PT adaptors to list of
  supported adaptors.

The list of adaptors was found at
http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&Inst=Yes&DwnldID=10957

MFC after: 1 week

15 years agoAdd very simple regression test for fstat(2) on sockets: make sure it
rwatson [Mon, 6 Oct 2008 19:42:03 +0000 (19:42 +0000)]
Add very simple regression test for fstat(2) on sockets: make sure it
returns success for various socket types.  It's easy to imagine this
being enhanced to validate the returned data, but...

15 years agoAdd a boot loader for ixp425 based boards like the Gateworks Avila
jhay [Mon, 6 Oct 2008 19:38:10 +0000 (19:38 +0000)]
Add a boot loader for ixp425 based boards like the Gateworks Avila
and ADI Pronghorn Metro with Redboot on them.

15 years agoNow that portalfs doesn't directly invoke uipc_connect2(), make it a
rwatson [Mon, 6 Oct 2008 18:43:11 +0000 (18:43 +0000)]
Now that portalfs doesn't directly invoke uipc_connect2(), make it a
static symbol.

MFC after: 3 days

15 years agoUse soconnect2() rather than directly invoking uipc_connect2() to
rwatson [Mon, 6 Oct 2008 18:38:50 +0000 (18:38 +0000)]
Use soconnect2() rather than directly invoking uipc_connect2() to
interconnect two UNIX domain sockets.

MFC after: 3 days

15 years agoDocument that the bce(4) driver works with Dell PowerEdge 1950/2950
simon [Mon, 6 Oct 2008 18:07:07 +0000 (18:07 +0000)]
Document that the bce(4) driver works with Dell PowerEdge 1950/2950
NIC's.

MFC after: 3 days

15 years ago- Add jme(4).
simon [Mon, 6 Oct 2008 17:51:44 +0000 (17:51 +0000)]
- Add jme(4).

Suggested by: yongari
MFC after: 3 days

15 years agoAdd a HARDWARE section. It doesn't contain any information not
simon [Mon, 6 Oct 2008 17:49:24 +0000 (17:49 +0000)]
Add a HARDWARE section.  It doesn't contain any information not
already in the DESCRIPTION section, but this allows for the manual
page to be used in the automatic hardware notes generation.

MFC after: 3 days

15 years agoAdd myself as requested by the committers-guide
lstewart [Mon, 6 Oct 2008 14:20:03 +0000 (14:20 +0000)]
Add myself as requested by the committers-guide

Approved by: gnn (mentor)

15 years agoCorrect a typo.
jkoshy [Mon, 6 Oct 2008 10:34:44 +0000 (10:34 +0000)]
Correct a typo.

15 years agoBring in the trivial differences between this code and John Hay's new
imp [Sun, 5 Oct 2008 23:59:52 +0000 (23:59 +0000)]
Bring in the trivial differences between this code and John Hay's new
code.  Added a copyright for the work I did to this file a couple of
years ago.  Add John's copyright too, since I'm sure I'll be pulling
more into this code.  This also implements a new -n option to not
allow breaking into the boot sequence which was original in the patch
John posted (not in the original i386 code I based this boot2.c on,
only the name is the same).  I haven't checked to see if he did that,
or if it was one of Sam's improvements.

Submitted by: jhay@

15 years agoRemove XMODEM_DL support. It never was complete and only serves to
imp [Sun, 5 Oct 2008 23:39:28 +0000 (23:39 +0000)]
Remove XMODEM_DL support.  It never was complete and only serves to
increase the diffs with other arm boot2 loaders.

15 years agoIndent with 8-space tabs. This reduces the diffs to the newer ixp425
imp [Sun, 5 Oct 2008 23:37:03 +0000 (23:37 +0000)]
Indent with 8-space tabs.  This reduces the diffs to the newer ixp425
boot2 and may make it easier to merge these files in the future...

15 years ago- Add age(4).
simon [Sun, 5 Oct 2008 19:23:33 +0000 (19:23 +0000)]
- Add age(4).

Suggested by: stass
MFC after: 3 days (including other recent changes)

15 years agoAdd initial version of the nullfs(5) man page
danger [Sun, 5 Oct 2008 18:57:27 +0000 (18:57 +0000)]
Add initial version of the nullfs(5) man page

MFC after: 3 days

15 years ago- Revert to the pre-r183242 behavior of building sade(8) on sparc64.
marius [Sun, 5 Oct 2008 14:15:22 +0000 (14:15 +0000)]
- Revert to the pre-r183242 behavior of building sade(8) on sparc64.
- Fix whitespace.

15 years agoDisable ATAPI DMA as it's once again broken in that it causes data
marius [Sun, 5 Oct 2008 14:00:44 +0000 (14:00 +0000)]
Disable ATAPI DMA as it's once again broken in that it causes data
corruption with the on-board AcerLabs M5229 controllers. While at
it, remove the pointless "nothing to autoload yet." message.

MFC after: 3 days

15 years agoDocument that vr(4) supports Soekris Engineering net5501.
simon [Sun, 5 Oct 2008 11:39:16 +0000 (11:39 +0000)]
Document that vr(4) supports Soekris Engineering net5501.

15 years agoMajor whitespace cleanup to follow FDP style. This was done since
simon [Sun, 5 Oct 2008 11:01:50 +0000 (11:01 +0000)]
Major whitespace cleanup to follow FDP style.  This was done since
markup was getting in the way when doing "real" changes.

The resulting HTML output is unchanged.

15 years agoMove a lists which was inside a <para> to outside the <para>.
simon [Sun, 5 Oct 2008 10:31:46 +0000 (10:31 +0000)]
Move a lists which was inside a <para> to outside the <para>.
This removes some vertical whitespace in the output.

This one was missed in r183620.

15 years ago- Add support for Pp command - this improves output for axe(4).
simon [Sun, 5 Oct 2008 09:45:45 +0000 (09:45 +0000)]
- Add support for Pp command - this improves output for axe(4).
- Add support for .Fx, just in case somebody starts using it.
- Only add the arch information to the first para per manual page.
  This improves output for axe(4), and generally seems to make sense
  (I hope).
- Remove an unneeded line for non-compat mode.

15 years agoMove some lists which were inside a <para> to outside the <para>.
simon [Sun, 5 Oct 2008 09:35:46 +0000 (09:35 +0000)]
Move some lists which were inside a <para> to outside the <para>.
This removes some vertical whitespace in the output.

Whitespace was intentionally not fixed in the SGML for the related
lists, to make the actual change more clear.  This file is in need of
a major whitespace cleanup anyway.

15 years agoRemove reference to hardware devices which does not exist anymore.
simon [Sun, 5 Oct 2008 08:27:56 +0000 (08:27 +0000)]
Remove reference to hardware devices which does not exist anymore.
This fixes the build.

15 years agoRemove reference to hardware devices which does not exist anymore.
simon [Sun, 5 Oct 2008 08:16:38 +0000 (08:16 +0000)]
Remove reference to hardware devices which does not exist anymore.
This fixes the build.

15 years agoo Typo fixes.
maxim [Sun, 5 Oct 2008 05:12:48 +0000 (05:12 +0000)]
o Typo fixes.

PR: docs/127866
Submitted by: Marius Korsmo

15 years agoIf the current thread has the trap bit set (i.e. a debugger had
davidxu [Sun, 5 Oct 2008 02:03:54 +0000 (02:03 +0000)]
If the current thread has the trap bit set (i.e. a debugger had
single stepped the process to the system call), we need to clear
the trap flag from the new frame. Otherwise, the new thread will
receive a (likely unexpected) SIGTRAP when it executes the first
instruction after returning to userland.

15 years agodynamically allocate the task structure in firmware_mountroot: when
sam [Sat, 4 Oct 2008 23:58:02 +0000 (23:58 +0000)]
dynamically allocate the task structure in firmware_mountroot: when
booting from an MFS root (e.g. from an install CD) firmware_mountroot
can be called twice with the second call happening before the task
callback occurs; this results in the task structure contents being
corrupted because it was declared static.

Submitted by: marius (original version)

15 years agoNote recently added ICH10 support
delphij [Sat, 4 Oct 2008 21:31:12 +0000 (21:31 +0000)]
Note recently added ICH10 support

15 years agoCurrent linux_fooaffinity() emulation fails, as the FreeBSD affinity
kib [Sat, 4 Oct 2008 19:23:30 +0000 (19:23 +0000)]
Current linux_fooaffinity() emulation fails, as the FreeBSD affinity
syscalls expect the bitmap size in the range from 32 to 128. Old glibc
always assumed size 1024, while newer glibc searches for approriate
size, starting from 1024 and going up.

For now, use FreeBSD size of cpuset_t for bitmap size parameter and
return EINVAL if length of user space bitmap less than our size of
cpuset_t.

Submitted by: dchagin
MFC after: 1 week
[This requires MFC of the actual linux affinity syscalls]

15 years agoStyle changes: compare pointer to NULL and move a }.
bz [Sat, 4 Oct 2008 17:07:58 +0000 (17:07 +0000)]
Style changes: compare pointer to NULL and move a }.

MFC after: 6 weeks

15 years agoRemove an INP_RUNLOCK() missed in SVN r183606, cvs rev. 1.195 raw_ip.c
bz [Sat, 4 Oct 2008 16:48:09 +0000 (16:48 +0000)]
Remove an INP_RUNLOCK() missed in SVN r183606, cvs rev. 1.195 raw_ip.c
when transitioning from so_cred to inp_cred.

MFC after: 6 weeks

15 years ago- Minor mdoc cleanup.
simon [Sat, 4 Oct 2008 16:21:04 +0000 (16:21 +0000)]
- Minor mdoc cleanup.
- Make HARDWARE section a bit more friendly for the Hardware Notes.

15 years ago- Add ae(4) to the list of devices support full hardware VLAN
stas [Sat, 4 Oct 2008 15:44:22 +0000 (15:44 +0000)]
- Add ae(4) to the list of devices support full hardware VLAN
  tags processing.

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

15 years ago- Add ae(4).
simon [Sat, 4 Oct 2008 15:41:25 +0000 (15:41 +0000)]
- Add ae(4).
- Add 2008 to copyright years.

15 years agoCache so_cred as inp_cred in the inpcb.
bz [Sat, 4 Oct 2008 15:06:34 +0000 (15:06 +0000)]
Cache so_cred as inp_cred in the inpcb.
This means that inp_cred is always there, even after the socket
has gone away. It also means that it is constant for the lifetime
of the inp.
Both facts lead to simpler code and possibly less locking.

Suggested by: rwatson
Reviewed by: rwatson
MFC after: 6 weeks
X-MFC Note: use a inp_pspare for inp_cred

15 years agoDon't explicitly bzer driver softc
rnoland [Sat, 4 Oct 2008 14:51:14 +0000 (14:51 +0000)]
Don't explicitly bzer driver softc

new-bus already handles this for us.

Suggested by jhb@

Approved by: jhb (mentor)

15 years agoUse M_WAITOK when allocating driver memory
rnoland [Sat, 4 Oct 2008 14:48:40 +0000 (14:48 +0000)]
Use M_WAITOK when allocating driver memory

We don't explicity check for error here and M_WAITOK will just put the
process to sleep waiting on resources to become available.

Suggested by jhb@

Approved by: jhb (mentor)

15 years agoChange drm_order() to use ffsl()
rnoland [Sat, 4 Oct 2008 14:45:34 +0000 (14:45 +0000)]
Change drm_order() to use ffsl()

pointed out by jhb@

Approved by: jhb (mentor)

15 years ago- Add manual page for ae(4) driver.
stas [Sat, 4 Oct 2008 14:21:54 +0000 (14:21 +0000)]
- Add manual page for ae(4) driver.

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

15 years ago- Document that 'show alllocks' and 'show locks' are only available
simon [Sat, 4 Oct 2008 14:12:09 +0000 (14:12 +0000)]
- Document that 'show alllocks' and 'show locks' are only available
  with witness.
- Remove references to non-existing witness(9) and replace with
  witness(4) where appropriate.
- Bump document date.

MFC after: 3 days

15 years agoChange the linprocfs <pid>/maps and procfs <pid>/map handlers to use
kib [Sat, 4 Oct 2008 14:08:16 +0000 (14:08 +0000)]
Change the linprocfs <pid>/maps and procfs <pid>/map handlers to use
sbuf instead of doing uiomove. This allows for reads from non-zero
offsets to work.

Patch is forward-ported des@' one, and was adopted to current code
by dchagin@ and me.

Reviewed by: des (linprocfs part)
PR: kern/101453
MFC after: 1 week

15 years agoMinor style(9) fixes for U-Boot API glue.
raj [Sat, 4 Oct 2008 13:19:15 +0000 (13:19 +0000)]
Minor style(9) fixes for U-Boot API glue.

15 years agoU-Boot API glue improvements:
raj [Sat, 4 Oct 2008 13:10:38 +0000 (13:10 +0000)]
U-Boot API glue improvements:

- extend ub_dev_read() and ub_dev_recv() so that the actual len and
  all error codes can be passed and processed properly; unify behaviour of
  these routines

- introduce syscall general error code (API_ESYSC)

15 years ago- Add cross-references.
jkoshy [Sat, 4 Oct 2008 12:54:25 +0000 (12:54 +0000)]
- Add cross-references.
- Tweak -mdoc use.

15 years agoAdd cross-references.
jkoshy [Sat, 4 Oct 2008 12:53:08 +0000 (12:53 +0000)]
Add cross-references.

15 years agoCross-reference new manual pages.
jkoshy [Sat, 4 Oct 2008 12:44:05 +0000 (12:44 +0000)]
Cross-reference new manual pages.

15 years ago- Cross-reference new manual pages.
jkoshy [Sat, 4 Oct 2008 12:35:02 +0000 (12:35 +0000)]
- Cross-reference new manual pages.
- Spell new PMC class names correctly.

15 years agoAdd manual pages for performance measurement counters present in
jkoshy [Sat, 4 Oct 2008 12:27:49 +0000 (12:27 +0000)]
Add manual pages for performance measurement counters present in
Intel Atom(tm), Core(tm) and Core2(tm) CPUs.

15 years ago- Add ae(4) to loader.conf.
stas [Sat, 4 Oct 2008 11:49:53 +0000 (11:49 +0000)]
- Add ae(4) to loader.conf.

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

15 years ago- Add ae(4) to sysinstall devices list.
stas [Sat, 4 Oct 2008 11:49:26 +0000 (11:49 +0000)]
- Add ae(4) to sysinstall devices list.

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

15 years ago- Spell cam correctly (scbus), this makes it possible to compile hptiop
antoine [Sat, 4 Oct 2008 10:39:31 +0000 (10:39 +0000)]
- Spell cam correctly (scbus), this makes it possible to compile hptiop
in GENERIC and LINT. [1]
- Rename hpt_dbg_level to hpt_iop_dbg_level to avoid multiple definition
of hpt_dbg_level (hptmv also has hpt_dbg_level).

PR: 127551 [1]
Reviewed by: scottl@
MFC after: 1 month

15 years agoFix a typo.
jkoshy [Sat, 4 Oct 2008 07:19:26 +0000 (07:19 +0000)]
Fix a typo.

15 years agosc->ndis_txidx should be cycle between 0 and sc->ndis_maxpkts, not
weongyo [Sat, 4 Oct 2008 04:15:39 +0000 (04:15 +0000)]
sc->ndis_txidx should be cycle between 0 and sc->ndis_maxpkts, not
NDIS_TXPKTS and don't allocate unused extra spaces for sc->ndis_txarray
and sc->ndis_txpool.

PR: kern/127644
Submitted by: Antoine Pelisse <apelisse_at_gmail.com>
MFC after: 1 week

15 years agoFix Vflags abuse in fdescfs. There should be no functional changes.
trasz [Fri, 3 Oct 2008 23:21:14 +0000 (23:21 +0000)]
Fix Vflags abuse in fdescfs.  There should be no functional changes.

Approved by: rwatson (mentor)

15 years agoFix Vflags abuse in cd9660. There should be no functional changes.
trasz [Fri, 3 Oct 2008 23:17:22 +0000 (23:17 +0000)]
Fix Vflags abuse in cd9660.  There should be no functional changes.

Approved by: rwatson (mentor)

15 years agoAbort transfers on all pipes before closing them. This fixes the crash
emax [Fri, 3 Oct 2008 22:40:42 +0000 (22:40 +0000)]
Abort transfers on all pipes before closing them. This fixes the crash
when Bluetooth USB device is pulled out without stopping the stack first.

Submitted by: Vladimir Grebenschikov vova at fbsd dot ru
MFC after: 1 week

15 years agoOops, missed updating a place with with 's/lock1/plock/' when adding
jhb [Fri, 3 Oct 2008 18:13:05 +0000 (18:13 +0000)]
Oops, missed updating a place with with 's/lock1/plock/' when adding
interlock support to WITNESS.  Specifically, the printf listing the
first location when duplicate locks of the same type are acquired.

Reported by: pho

15 years agoresync to git master
rnoland [Fri, 3 Oct 2008 16:59:11 +0000 (16:59 +0000)]
resync to git master

This reverts a private patch which is causing issues with many Intel chipsets.
I will review that patch and see what we need to do to fix it up later, but
for the time being, we will just get these chips working again.

This update contains a lot of code cleanup and is post gem merge
(no, we don't have gem support).  It should prove much easier to read the
code now.  A lot of thanks goes to vehemens for that work.  I have adapted
the code to use cdevpriv for tracking per open file data.  That alleviates
the old ugly hack that we used to try and accomplish the task and helped to
clean up the open / close behavior a good bit.  This also replaces the hack
that was put in place a year or so ago to prevent radeons from locking up
with AIGLX enabled.  I have had a couple of radeon testers report that it
still works as expected, though I no longer have radeon hardware to test with
myself.  Other various fixes from the linux crew and Intel, many of
which are muddled in with the gem merge.

Approved by: jhb (mentor)
Obtained from: mesa/drm git master
MFC after: 2 weeks

15 years agoFurther minor cleanups to UNIX domain sockets:
rwatson [Fri, 3 Oct 2008 13:01:56 +0000 (13:01 +0000)]
Further minor cleanups to UNIX domain sockets:

- Staticize and locally prototype functions uipc_ctloutput(), unp_dispose(),
  unp_init(), and unp_externalize(), none of which have been required
  outside of uipc_usrreq.c since uipc_proto.c was removed.
- Remove stale prototype for uipc_usrreq(), which has not existed in the
  code since 1997
- Forward declare and staticize uipc_usrreqs structure in uipc_usrreq.c and
  not un.h.
- Comment on why uipc_connect2() is still non-static -- it is used directly
  by fifofs.
- Remove stale comments, tidy up whitespace.

MFC after: 3 days (where applicable)

15 years agoImplement IPv4 source address selection for unbound sockets.
bz [Fri, 3 Oct 2008 12:21:21 +0000 (12:21 +0000)]
Implement IPv4 source address selection for unbound sockets.

For the jail case we are already looping over the interface addresses
before falling back to the only IP address of a jail in case of no
match. This is in preparation for the upcoming multi-IPv4/v6/no-IP
jail patch this change was developed with initially.

This also changes the semantics of selecting the IP for processes within
a jail as it now uses the same logic as outside the jail (with additional
checks) but no longer is on a mutually exclusive code path.

Benchmarks had shown no difference at 95.0% confidence for neither the
plain nor the jail case (even with the additional overhead).  See:
http://lists.freebsd.org/pipermail/freebsd-net/2008-September/019531.html

Inpsired by a patch from: Yahoo! (partially)
Tested by: latest multi-IP jail patch users (implictly)
Discussed with: rwatson (general things around this)
Reviewed by: mostly silence (feedback from bms)
Help with benchmarking from: kris
MFC after: 2 months

15 years agoITE IT8213F support.
brueffer [Fri, 3 Oct 2008 11:37:33 +0000 (11:37 +0000)]
ITE IT8213F support.

15 years ago- Add driver for Attansic L2 FastEthernet controller found on
stas [Fri, 3 Oct 2008 10:31:31 +0000 (10:31 +0000)]
- Add driver for Attansic L2 FastEthernet controller found on
  Asus EeePC and some Asus mainboards.

Reviewed by: yongari, rpaulo, jhb
Tested by: many
Approved by: kib (mentor)
MFC after: 1 week

15 years agoBackout svn r183528.
bz [Fri, 3 Oct 2008 10:08:36 +0000 (10:08 +0000)]
Backout svn r183528.

SRCDIR is seeded from `pwd` which not only means src/sys/ but
also src/include/ (and possibly src/usr.sbin/amd/include/ ?).
Trying to build world resulted in

  ===> include (includes)
  cd /usr/src/include; make buildincludes; make installincludes
  creating osreldate.h from newvers.sh
  cd: can't cd to /usr/src/include/sys
  *** Error code 2

as there is apparently no src/include/sys.

There are multiple possible solutions ranging from seeding SRCDIR from
the environment to adding more substitution patterns.

Reported by: sam, bz
Proper solution to be implemented and tested by: peter

15 years agoSmall cleanups to openpty().
ed [Fri, 3 Oct 2008 09:42:50 +0000 (09:42 +0000)]
Small cleanups to openpty().

- Pass O_NOCTTY to posix_openpt(2). This makes the implementation work
  consistently on implementations that make the PTY the controlling TTY
  by default.

- Call unlockpt() before opening the slave device. POSIX mentions that
  de slave device should only be opened after grantpt() and unlockpt()
  have been called.

- Replace some redundant code by a label.

In theory we could remove a lot of code from openpty() on FreeBSD
-CURRENT, because grantpt(), unlockpt() and revoke() are not needed in
our implementation. We'd better keep them there. This makes the code
still work with older FreeBSD releases and even makes it work on other
non-BSD operating systems.

I've compiled openpty() on Linux. You only need to remove the revoke()
call, because revoke() on Linux always returns -1. Apart from that, it
seems to work like it should.

Reviewed by: jhb

15 years agoAdd hint about hw.snd.default_unit sysctl.
mav [Fri, 3 Oct 2008 09:19:07 +0000 (09:19 +0000)]
Add hint about hw.snd.default_unit sysctl.

15 years agoRemove or update several stale comments.
rwatson [Fri, 3 Oct 2008 09:01:55 +0000 (09:01 +0000)]
Remove or update several stale comments.

A bit of whitespace/style cleanup.

Update copyright.

MFC after: 3 days (applicable changes)

15 years agoUpdate ng_tty for MPSAFE TTY.
thompsa [Fri, 3 Oct 2008 05:14:54 +0000 (05:14 +0000)]
Update ng_tty for MPSAFE TTY.

This changes from a line discipline to the tty_hooks mechanism. Data will come
in directly via rint_bypass and sent to the peer node in a single mbuf.

As line disciplines are no longer used a new netgraph command called
NGM_TTY_SET_TTY is used to attach the tty. This takes a pointer to to the open
file descriptor of the tty and registers the tty hooks. When the tty disappears
the node will shutdown.

Thanks to: ed
Sponsored by: Hobnob, Inc

15 years agoAdd hardware MAC statistics support. Register information was
yongari [Fri, 3 Oct 2008 03:58:16 +0000 (03:58 +0000)]
Add hardware MAC statistics support. Register information was
obtained from Linux forcedeth driver.
While I'm here move creating a sysctl node for process_limit to
function nfe_sysctl_node().

Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >

15 years agoFix bug in LRO on T304 whereby a packet could be sent to the wrong interface's ifp.
kmacy [Fri, 3 Oct 2008 00:50:26 +0000 (00:50 +0000)]
Fix bug in LRO on T304 whereby a packet could be sent to the wrong interface's ifp.

Submitted by: Chelsio Inc.
MFC after: 1 day

15 years agoReturn BUS_PROBE_GENERIC rather BUS_PROBE_DEFAULT for generic CBB we match.
imp [Thu, 2 Oct 2008 22:50:11 +0000 (22:50 +0000)]
Return BUS_PROBE_GENERIC rather BUS_PROBE_DEFAULT for generic CBB we match.

Reviewed by: jhb@

15 years agoAllow static hints to override kenv.
bms [Thu, 2 Oct 2008 22:31:30 +0000 (22:31 +0000)]
Allow static hints to override kenv.

15 years agoOops, initialize sections and sections_end to NULL.
jhb [Thu, 2 Oct 2008 20:42:10 +0000 (20:42 +0000)]
Oops, initialize sections and sections_end to NULL.

Submitted by: Navdeep Parhar
MFC after: 1 week

15 years agoCorrectly handle Intel g33 chips and add support for g45 chips
rnoland [Thu, 2 Oct 2008 20:29:45 +0000 (20:29 +0000)]
Correctly handle Intel g33 chips and add support for g45 chips

g33 based chips use a different method of identifying the gtt size.
g45 based chips gtt is located in a different area of stolen memory.

Approved by: jhb (mentor)
MFC after: 2 weeks

15 years agoadd duplicate cmd entries for vlan and vlandev that are marked !clone
sam [Thu, 2 Oct 2008 20:03:41 +0000 (20:03 +0000)]
add duplicate cmd entries for vlan and vlandev that are marked !clone
so they can be used when not doing a create operation

Reviewed by: ed

15 years agoAdd ICH10 PCI id's.
sos [Thu, 2 Oct 2008 19:18:28 +0000 (19:18 +0000)]
Add ICH10 PCI id's.
Submitted by: Xin Li

Fix the number of PATA ports on newer ICHX chips, they have just 1 port not 2.