openbsd
14 years agoWhen removing an announced prefix inherit the metric and ext_tag from the
claudio [Mon, 19 Jul 2010 09:16:30 +0000 (09:16 +0000)]
When removing an announced prefix inherit the metric and ext_tag from the
LSA that is currently in the tree. Pruning with default metrics is bad
since the checksum of the LSA is different and therefor an additional
update cycle is needed to convince the other systems.
OK bluhm@, sthen@

14 years agoadd a "ControlPersist" option that automatically starts a background
djm [Mon, 19 Jul 2010 09:15:12 +0000 (09:15 +0000)]
add a "ControlPersist" option that automatically starts a background
ssh(1) multiplex master when connecting. This connection can stay alive
indefinitely, or can be set to automatically close after a user-specified
duration of inactivity. bz#1330 - patch by dwmw2 AT infradead.org, but
further hacked on by wmertens AT cisco.com, apb AT cequrux.com,
martin-mindrot-bugzilla AT earth.li and myself; "looks ok" markus@

14 years agolsa_refresh() was able to resurect dead LSA resulting in zombie anouncements.
claudio [Mon, 19 Jul 2010 09:11:08 +0000 (09:11 +0000)]
lsa_refresh() was able to resurect dead LSA resulting in zombie anouncements.
Make sure that if the LSA is managed by this router that we force timed-out
LSA to stay timed out which causes them to be pruned in the secound round.
This problem can show up when not exactly the same LSA is pruned because
the checksum is used as tiebreaker resulting in indeterministic selection
of which LSA is considered newer.
OK bluhm@, sthen@

14 years agoDocument the recently introduced select(2) / poll(2) support for video(4).
mglocker [Mon, 19 Jul 2010 08:59:38 +0000 (08:59 +0000)]
Document the recently introduced select(2) / poll(2) support for video(4).

14 years ago* don't round blocksize to multiples of 16; can cause problems when
jakemsr [Mon, 19 Jul 2010 07:57:36 +0000 (07:57 +0000)]
* don't round blocksize to multiples of 16; can cause problems when
sample size or number of channels is not a power of 2.
* use 'param->bps' instead of 'param->precision / NBBY' for the number
of bytes per sample.

ok ratchov

14 years agoremove Xr to self; from Daniel Dickman
jmc [Mon, 19 Jul 2010 07:54:12 +0000 (07:54 +0000)]
remove Xr to self; from Daniel Dickman

14 years agoadd a quirk to allow attaching devices which claim to have a vendor
jakemsr [Mon, 19 Jul 2010 05:50:37 +0000 (05:50 +0000)]
add a quirk to allow attaching devices which claim to have a vendor
defined interface class, but are audio class compliant enough to
be supported by this driver.
the E-MU 0202 USB is such a device.  this gets it attaching, but
needs some forthcoming commits to make it usable.

14 years agosync
jakemsr [Mon, 19 Jul 2010 05:18:52 +0000 (05:18 +0000)]
sync

14 years agoCreative Labs/E-MU 0202 USB
jakemsr [Mon, 19 Jul 2010 05:12:49 +0000 (05:12 +0000)]
Creative Labs/E-MU 0202 USB

14 years agomove audio specific quirks out of usb_quirks and into uaudio
jakemsr [Mon, 19 Jul 2010 05:08:37 +0000 (05:08 +0000)]
move audio specific quirks out of usb_quirks and into uaudio

ok krw

14 years agoRemove unused function.
lum [Mon, 19 Jul 2010 04:41:28 +0000 (04:41 +0000)]
Remove unused function.
ok canacar@

14 years agobz#1797: fix swapped args in upload_dir_internal(), breaking recursive
djm [Mon, 19 Jul 2010 03:16:33 +0000 (03:16 +0000)]
bz#1797: fix swapped args in upload_dir_internal(), breaking recursive
upload depth checks and causing verbose printing of transfers to always
be turned on; patch from imorgan AT nas.nasa.gov

14 years agoUpdate to upstream from ncurses-5.7-20100717. Fixes rs1 in wsvt25 and a
nicm [Mon, 19 Jul 2010 00:28:18 +0000 (00:28 +0000)]
Update to upstream from ncurses-5.7-20100717. Fixes rs1 in wsvt25 and a
few other bits and pieces.

ok millert

14 years agoFix the cproj family to not return garbage on finite arguments.
guenther [Mon, 19 Jul 2010 00:04:07 +0000 (00:04 +0000)]
Fix the cproj family to not return garbage on finite arguments.

ok millert@

14 years agoremove some nasty hacks;
jmc [Sun, 18 Jul 2010 21:45:01 +0000 (21:45 +0000)]
remove some nasty hacks;

14 years agopackages-specs man page is section 7, not 5;
jmc [Sun, 18 Jul 2010 21:22:36 +0000 (21:22 +0000)]
packages-specs man page is section 7, not 5;

14 years agoSupport from vblank events in drm.
oga [Sun, 18 Jul 2010 21:01:06 +0000 (21:01 +0000)]
Support from vblank events in drm.

Previously, if userland wanted to wait on a certain vertical blank, it
had to call an ioctl which slept. Now, they can ask for an even on the
drm fd, which is then read off, and can be poll(4)ed on. For dri2 this
fits better into the workflow since the fd gets added to the xserver
main loop, and replies to the dri2 clients happen upon recieving the
events.

This functionality is only used with xserver 1.8 (and for the intel
driver in our tree, this support is currently #if 0ed out due to bugs
with vblanks on 945 that are still being chased)

matthieu@ ok.

14 years agoClear event status on resume, to avoid going back to sleep/reboot/shutdown
mlarkin [Sun, 18 Jul 2010 19:49:35 +0000 (19:49 +0000)]
Clear event status on resume, to avoid going back to sleep/reboot/shutdown
immediately after resume. Tested by a bunch of people on hackers@ on
various hardware. Fixes at least Toshiba NB200 and Gateway LT3103u.

ok marco

14 years agodocument OPSYS and OPSYS_VER as obsolete, mention OSREV; ok espie@
naddy [Sun, 18 Jul 2010 18:50:38 +0000 (18:50 +0000)]
document OPSYS and OPSYS_VER as obsolete, mention OSREV; ok espie@

14 years agoUse crealf() and cimagf() instead of creal() and cimag() when the
guenther [Sun, 18 Jul 2010 18:42:26 +0000 (18:42 +0000)]
Use crealf() and cimagf() instead of creal() and cimag() when the
argument is a float.  Eliminate a cast rendered superfluous as a result.

ok kettenis@, millert@

14 years agoregen
kettenis [Sun, 18 Jul 2010 17:11:40 +0000 (17:11 +0000)]
regen

14 years agoAnother PLX PCI Express switch.
kettenis [Sun, 18 Jul 2010 17:10:52 +0000 (17:10 +0000)]
Another PLX PCI Express switch.

From Mitja Muzenic.

14 years agoregen
kettenis [Sun, 18 Jul 2010 17:02:52 +0000 (17:02 +0000)]
regen

14 years agoAdd missing Core i3/i5 PCIe bridge. Fix entries for the 5-series/3400 chipset
kettenis [Sun, 18 Jul 2010 17:01:38 +0000 (17:01 +0000)]
Add missing Core i3/i5 PCIe bridge.  Fix entries for the 5-series/3400 chipset
LPC bridge such that they can be used to identify the chipset variant and
add some missing entries.  Add QS57 HD audio.

14 years agoKill inconsequential dead store.
oga [Sun, 18 Jul 2010 15:46:41 +0000 (15:46 +0000)]
Kill inconsequential dead store.

found by Clang static analyzer.

14 years agoRadeon HD4670, from Henry Precheur; thanks!
oga [Sun, 18 Jul 2010 15:42:55 +0000 (15:42 +0000)]
Radeon HD4670, from Henry Precheur; thanks!

14 years agoget the markup right for [-]command
jmc [Sun, 18 Jul 2010 15:37:37 +0000 (15:37 +0000)]
get the markup right for [-]command

14 years agoRadeon HD4200 (needs a newer radeon X driver than in tree though),
oga [Sun, 18 Jul 2010 15:30:48 +0000 (15:30 +0000)]
Radeon HD4200 (needs a newer radeon X driver than in tree though),
tested by stsp@.

14 years agoFix an unaligned memory access.
martinh [Sun, 18 Jul 2010 15:15:40 +0000 (15:15 +0000)]
Fix an unaligned memory access.

14 years agoRepair Gdium support.
miod [Sun, 18 Jul 2010 13:36:13 +0000 (13:36 +0000)]
Repair Gdium support.

14 years agoGive each device on the bus full openings.
mk [Sun, 18 Jul 2010 12:49:10 +0000 (12:49 +0000)]
Give each device on the bus full openings.

Tested by me on an ARC-1210 on both single and multi volume
configurations.

ok dlg

14 years agoAdd support for Winbond/Nuvoton W83627DHG-P.
kettenis [Sun, 18 Jul 2010 12:44:55 +0000 (12:44 +0000)]
Add support for Winbond/Nuvoton W83627DHG-P.

From Mitja Muzenic.

14 years agofix a double "is";
jmc [Sun, 18 Jul 2010 08:03:04 +0000 (08:03 +0000)]
fix a double "is";

14 years ago- grammar fix from Glen Barber (freebsd pr 148701)
jmc [Sun, 18 Jul 2010 07:59:40 +0000 (07:59 +0000)]
- grammar fix from Glen Barber (freebsd pr 148701)
- fix some list issues found by mandoc

14 years agoMove some macros in <msdosfs/denode.h> to outside the #ifdef _KERNEL
guenther [Sat, 17 Jul 2010 19:27:07 +0000 (19:27 +0000)]
Move some macros in <msdosfs/denode.h> to outside the #ifdef _KERNEL
so that libkvm doesn't need to define that, thereby avoiding some warnings.

ok miod@

14 years agoWrap kernel function declarations in #ifdef _KERNEL to eliminate warnings
guenther [Sat, 17 Jul 2010 19:24:58 +0000 (19:24 +0000)]
Wrap kernel function declarations in #ifdef _KERNEL to eliminate warnings
when the header is used by userspace.
ok miod@

14 years agoCorrect the #includes to avoid warnings in <rpc/svc.h>
guenther [Sat, 17 Jul 2010 19:23:10 +0000 (19:23 +0000)]
Correct the #includes to avoid warnings in <rpc/svc.h>
ok miod@

14 years agoCorrect value of DBL_MIN and DBL_MAX to match <sys/limits.h>
miod [Sat, 17 Jul 2010 17:52:43 +0000 (17:52 +0000)]
Correct value of DBL_MIN and DBL_MAX to match <sys/limits.h>

14 years agoadd missing header needed by event_warn()
chl [Sat, 17 Jul 2010 17:16:47 +0000 (17:16 +0000)]
add missing header needed by event_warn()

ok nicm@

14 years agouse the correct block ack structure when building an ADDBA response.
damien [Sat, 17 Jul 2010 16:30:01 +0000 (16:30 +0000)]
use the correct block ack structure when building an ADDBA response.

14 years agoDrop 3rd and 4th clauses from David Young's license.
damien [Sat, 17 Jul 2010 16:25:09 +0000 (16:25 +0000)]
Drop 3rd and 4th clauses from David Young's license.

from NetBSD

14 years agoWhen running in full diff mode, don't wait for user input before
ajacoutot [Sat, 17 Jul 2010 06:47:13 +0000 (06:47 +0000)]
When running in full diff mode, don't wait for user input before
starting the comparison anymore. It does not bring us anything and
allows to chop 30 lines of the script.
While here, remove some empty lines.

14 years agoPlug memory leak.
ray [Sat, 17 Jul 2010 00:00:32 +0000 (00:00 +0000)]
Plug memory leak.

OK nicm

14 years agoSimplify print_status by removing NULL handling.
ray [Fri, 16 Jul 2010 23:27:58 +0000 (23:27 +0000)]
Simplify print_status by removing NULL handling.

OK nicm

14 years agoClose FILEs when fork fails.
ray [Fri, 16 Jul 2010 21:47:02 +0000 (21:47 +0000)]
Close FILEs when fork fails.

OK nicm

14 years agoExit strip(1) with an error, if an objfile could not be read.
mpf [Fri, 16 Jul 2010 21:44:06 +0000 (21:44 +0000)]
Exit strip(1) with an error, if an objfile could not be read.
Input and OK millert@

14 years agorevisison -> revision
ray [Fri, 16 Jul 2010 19:58:22 +0000 (19:58 +0000)]
revisison -> revision

14 years agoCombine two diff_output lines into one, from diff.
ray [Fri, 16 Jul 2010 17:53:20 +0000 (17:53 +0000)]
Combine two diff_output lines into one, from diff.

OK stsp xsa nicm

14 years agoMove the description of load averages/top line to a better place.
lum [Fri, 16 Jul 2010 16:27:02 +0000 (16:27 +0000)]
Move the description of load averages/top line to a better place.
Suggestions and ok canacar@

14 years agofix up the mini synopses in the subsections; fix from ingo
jmc [Fri, 16 Jul 2010 16:10:20 +0000 (16:10 +0000)]
fix up the mini synopses in the subsections; fix from ingo

14 years agomore timing paranoia - compare all parts of the expected decrypted
djm [Fri, 16 Jul 2010 14:07:35 +0000 (14:07 +0000)]
more timing paranoia - compare all parts of the expected decrypted
data before returning. AFAIK not exploitable in the SSH protocol.
"groovy" deraadt@

14 years agodiff exits 2 on error, not 1.
ray [Fri, 16 Jul 2010 08:06:00 +0000 (08:06 +0000)]
diff exits 2 on error, not 1.

OK millert

14 years agoWe never create or destroy pmaps from interrupt context, so wrapping the
kettenis [Fri, 16 Jul 2010 06:22:31 +0000 (06:22 +0000)]
We never create or destroy pmaps from interrupt context, so wrapping the
associated pool calls in splvm()/splx() is unnecessary and confusing.

ok deraadt@, drahn@

14 years agoMake the terminal checking capability of systat better.
lum [Fri, 16 Jul 2010 05:22:48 +0000 (05:22 +0000)]
Make the terminal checking capability of systat better.
Checks taken from top(1) screen.c/init_termcap()
ok canacar@ sthen@ nicm@

14 years agoavoid bogus compiler warning
djm [Fri, 16 Jul 2010 04:45:30 +0000 (04:45 +0000)]
avoid bogus compiler warning

14 years agogc sc_lock. it is unused. ioctl commands can run concurrently now that
dlg [Fri, 16 Jul 2010 01:23:11 +0000 (01:23 +0000)]
gc sc_lock. it is unused. ioctl commands can run concurrently now that
iopools handle sleeping for ccbs.

14 years agoText ending in a full stop, exclamation mark or question mark
schwarze [Fri, 16 Jul 2010 00:34:33 +0000 (00:34 +0000)]
Text ending in a full stop, exclamation mark or question mark
should not flag the end of a sentence if:

1) The punctuation is followed by closing delimiters
and not preceded by alphanumeric characters, like in
"There is no full stop (.) in this sentence"

or

2) The punctuation is a child of a macro
and not preceded by alphanumeric characters, like in
"There is no full stop
.Pq \&.
in this sentence"

jmc@ and sobrado@ like this

14 years agoAfter .Sm on, spacing ought to restart right away, before the next token,
schwarze [Fri, 16 Jul 2010 00:03:29 +0000 (00:03 +0000)]
After .Sm on, spacing ought to restart right away, before the next token,
and not with a delay, after the next token.  But be careful not to cause
leading white space at the beginning of a line or column.
Improves chmod(1), ksh(1), tar(1), ps(1) and probably many more.
ok kristaps@ jmc@ sobrado@

14 years agowhen mpii(4) gets an async event (ie, sas topology change or
dlg [Thu, 15 Jul 2010 23:52:32 +0000 (23:52 +0000)]
when mpii(4) gets an async event (ie, sas topology change or
integrated raid has changed state) that needs an acknowledgement,
it currently goes scsi_io_get. because events are handled at interrupt
time it cannot wait for ccbs, so if the ccb pool is empty, it wont
get a ccb to ack the event with.

this cuts the code over to using iohandlers to ack these events.
when an ack is required, it shoves the rcb onto a queue and schedules
an iohandler to be run whenever a ccb becomes available. the majority
of the time the ccb will be there and the reply will happen
immediately. however, if the hba is busy we might be deferred until
some other command completes and gives its ccb back to the pool.

because it looks like we can now receive many events at a time
(rather than just one at a time like in mpi), im using a queue to
store the events in. the alternative was an iohandler per rcb, but
that struck me as a bit too heavy.

mikeb is really impressed
no objection from krw@

14 years agodocument .Sm; prodded by kristaps@
schwarze [Thu, 15 Jul 2010 23:46:58 +0000 (23:46 +0000)]
document .Sm; prodded by kristaps@

14 years agosome escapes i missed;
jmc [Thu, 15 Jul 2010 21:54:20 +0000 (21:54 +0000)]
some escapes i missed;

14 years agocvs and rcs use diff_output() to print everything since it is not
ray [Thu, 15 Jul 2010 21:32:37 +0000 (21:32 +0000)]
cvs and rcs use diff_output() to print everything since it is not
always to stdout. A large number of differences are due to this.
This diff reduces many more differences between diff and cvs/rcs.
Personally I think it's kinda ugly. =(

"Looks fine" millert

14 years agorepair incorrect block nesting, which screwed up indentation;
schwarze [Thu, 15 Jul 2010 21:20:38 +0000 (21:20 +0000)]
repair incorrect block nesting, which screwed up indentation;
problem reported and fix OK by jmc@

14 years agoMore delimiters that need quoting inside macros, hunted down by jmc@,
schwarze [Thu, 15 Jul 2010 20:51:38 +0000 (20:51 +0000)]
More delimiters that need quoting inside macros, hunted down by jmc@,
who asked me to commit because he is just running out of the door.

14 years agoAccording to Linux ath9k, the AR9280 1.0 was never sold.
damien [Thu, 15 Jul 2010 20:37:38 +0000 (20:37 +0000)]
According to Linux ath9k, the AR9280 1.0 was never sold.
Since this is the only chip revision that requires split TKIP MIC keys,
remove code that deal with that.

14 years agoremove dead code (AR9286 1.1 does not really exist.)
damien [Thu, 15 Jul 2010 20:25:27 +0000 (20:25 +0000)]
remove dead code (AR9286 1.1 does not really exist.)

14 years agoWhen the first argument or arguments of a macro are opening delimiters
schwarze [Thu, 15 Jul 2010 20:04:35 +0000 (20:04 +0000)]
When the first argument or arguments of a macro are opening delimiters
(parentheses and/or square brackets), both modern groff and mandoc first
output those leading delimiters as plain text, then start the macro scope
after these opening delimiters.  This is similar to printing trailing
punctuation and trailing closing delimiters on a macro line outside and
after the macro scope.  For example, ".Sq ( text )" is "(`text')",
not "`(text)'".  Thus, we now need to quote leading opening delimiters
when we want them inside the macro scope.
These are the cases in src/bin.

"makes sense" jmc@

14 years agosync ar5008_bb_load_noisefloor() with ar9003_bb_load_noisefloor().
damien [Thu, 15 Jul 2010 19:38:40 +0000 (19:38 +0000)]
sync ar5008_bb_load_noisefloor() with ar9003_bb_load_noisefloor().
we must not restore the noisefloor values if load failed on AR5008
family too.

14 years agofix PDADC value for target index (AR5008 family only.)
damien [Thu, 15 Jul 2010 19:33:34 +0000 (19:33 +0000)]
fix PDADC value for target index (AR5008 family only.)

14 years agoOn AR9285, the antCtrlCommon ROM field may have the upper 16 bits set.
damien [Thu, 15 Jul 2010 19:29:00 +0000 (19:29 +0000)]
On AR9285, the antCtrlCommon ROM field may have the upper 16 bits set.
Remove an incorrect mask to fix antenna diversity on some chips.

14 years agosync initialization values for AR9160 and AR9280 with ath9k.
damien [Thu, 15 Jul 2010 19:24:04 +0000 (19:24 +0000)]
sync initialization values for AR9160 and AR9280 with ath9k.

see http://marc.info/?l=linux-wireless&m=127802223525594&w=2 for the
list of changes/fixes.

14 years agoRemove initialization values for chips that never made it into production:
damien [Thu, 15 Jul 2010 19:07:43 +0000 (19:07 +0000)]
Remove initialization values for chips that never made it into production:
- AR9280 1.0
- AR9285 1.0 and 1.1
- AR9287 1.0

Remove checks that are now useless since we know we will never see those
chips, and rename some functions.

14 years agoRemove unused arguments from check(). From cvs/rcs.
ray [Thu, 15 Jul 2010 18:31:33 +0000 (18:31 +0000)]
Remove unused arguments from check(). From cvs/rcs.

OK millert

14 years agoDie immediately if pread fails. It's a fatal error so treat it as
ray [Thu, 15 Jul 2010 18:23:50 +0000 (18:23 +0000)]
Die immediately if pread fails. It's a fatal error so treat it as
such. Besides, we weren't handling the NULL being returned.

From diff. Minor nit by nicm.

OK xsa stsp nicm

14 years agoUse warnx instead of warn since errno wasn't set. From diff.
ray [Thu, 15 Jul 2010 18:19:18 +0000 (18:19 +0000)]
Use warnx instead of warn since errno wasn't set. From diff.

OK xsa stsp nicm

14 years agosync the elf types and their comments; ok millert
jmc [Thu, 15 Jul 2010 18:15:31 +0000 (18:15 +0000)]
sync the elf types and their comments; ok millert

14 years agodocument the missing "note section";
jmc [Thu, 15 Jul 2010 16:48:41 +0000 (16:48 +0000)]
document the missing "note section";
reported by Michael Kerrisk, freebsd pr 147995

big thanks to miod and millert for filling in the blanks

14 years agoknock out MetaGeek Wi-Spy 2.4i from attaching to uhid.
sthen [Thu, 15 Jul 2010 15:45:48 +0000 (15:45 +0000)]
knock out MetaGeek Wi-Spy 2.4i from attaching to uhid.

14 years agosync
sthen [Thu, 15 Jul 2010 15:45:15 +0000 (15:45 +0000)]
sync

14 years agoadd MetaGeek Wi-Spy 2.4i
sthen [Thu, 15 Jul 2010 15:44:55 +0000 (15:44 +0000)]
add MetaGeek Wi-Spy 2.4i

14 years agostore openssl error output during verification, and log it if it didn't work.
espie [Thu, 15 Jul 2010 13:38:30 +0000 (13:38 +0000)]
store openssl error output during verification, and log it if it didn't work.

14 years agoCorrect function in panic string. From Christophe Fillot
claudio [Thu, 15 Jul 2010 13:30:45 +0000 (13:30 +0000)]
Correct function in panic string. From Christophe Fillot

14 years agoSync with diff/diffreg.c r1.75:
ray [Thu, 15 Jul 2010 11:10:23 +0000 (11:10 +0000)]
Sync with diff/diffreg.c r1.75:
Return -1 on error as advertised.
Handle case when only one stream failed.

OK xsa, stsp

14 years agowhitespace
espie [Thu, 15 Jul 2010 10:37:32 +0000 (10:37 +0000)]
whitespace

14 years agom_getptr(m, 0, ...) may return an mbuf different from m -- if m has no
claudio [Thu, 15 Jul 2010 09:45:09 +0000 (09:45 +0000)]
m_getptr(m, 0, ...) may return an mbuf different from m -- if m has no
data in it. m_getptr() hops over empty buffers and points to the first
allocated data byte. Because of this the m_dup_pkthdr() call done by
m_copym0() can panic because not the first mbuf is passed.
Found the hard way by myself, diff by blambert@ commiitting for him since
he is not around. Tested and OK myself

14 years ago* Comment poll support in conf.h
mglocker [Thu, 15 Jul 2010 04:46:33 +0000 (04:46 +0000)]
* Comment poll support in conf.h
* Comment why we call the video interrupt in the mmap interface now in
uvideo.c

14 years agoadd two new members to structs audio_encoding and audio_prinfo.
jakemsr [Thu, 15 Jul 2010 03:43:11 +0000 (03:43 +0000)]
add two new members to structs audio_encoding and audio_prinfo.
for both structs, the new members are 'bps' and 'msb', which
describe the number of bytes per sample and data alignment in the
sample, respectively.  drivers must properly set these fields in
the 'query_encoding', 'set_parameters' and 'get_default_params'
hardware interface methods.

discussed with ratchov, deraadt

14 years agolimit the pools from 14 bits down. We cannot use PAGE_SIZE because it
deraadt [Thu, 15 Jul 2010 03:20:47 +0000 (03:20 +0000)]
limit the pools from 14 bits down.  We cannot use PAGE_SIZE because it
is a variable on sparc.  This should be revisited... after the arguments
for pagesize vs 4K complete :)

14 years agothe uvm_km_putpage is calling into tangly uvm guts again on not pmap direct.
tedu [Thu, 15 Jul 2010 00:14:17 +0000 (00:14 +0000)]
the uvm_km_putpage is calling into tangly uvm guts again on not pmap direct.
go back to something more like the previous design, and have the thread do
the heavy lifting.  solves vmmaplk panics.
ok deraadt oga thib
[and even simple diffs are hard to get perfect. help from mdempsky and deraadt]

14 years agosync
sthen [Wed, 14 Jul 2010 23:54:14 +0000 (23:54 +0000)]
sync

14 years agosync from mirrors.dat
sthen [Wed, 14 Jul 2010 23:47:54 +0000 (23:47 +0000)]
sync from mirrors.dat

14 years agopass the correct af if not specified.
dhill [Wed, 14 Jul 2010 23:44:41 +0000 (23:44 +0000)]
pass the correct af if not specified.

fixes printing active connections twice.

ok claudio@

14 years agoReturn -1 on error as advertised.
ray [Wed, 14 Jul 2010 22:15:57 +0000 (22:15 +0000)]
Return -1 on error as advertised.
Handle case when only one stream failed.

OK otto, millert

14 years agoimplement poll() for video(4)
jakemsr [Wed, 14 Jul 2010 21:24:33 +0000 (21:24 +0000)]
implement poll() for video(4)
ok mglocker

14 years agoThe append side of a socketbuffer is not allowed to call sbflush().
claudio [Wed, 14 Jul 2010 20:56:35 +0000 (20:56 +0000)]
The append side of a socketbuffer is not allowed to call sbflush().
Learned the hard way. dlg@ confirmed that it is save to just remove them,
the desync will still work but the reader needs to dequeue all packets first.

14 years agoIn urndis_decap(), delay the memcpy() for as long as possible. There is
mk [Wed, 14 Jul 2010 20:44:17 +0000 (20:44 +0000)]
In urndis_decap(), delay the memcpy() for as long as possible.  There is
no point in copying before having done the final sanity check (we copied
before the last one).

ok armani

14 years agourndis_newbuf() is only called with NULL as it's third argument so just
mk [Wed, 14 Jul 2010 20:21:55 +0000 (20:21 +0000)]
urndis_newbuf() is only called with NULL as it's third argument so just
get rid of it.

ok armani

14 years agoNewer 8168x and 810x chipsets have a different way of stopping the chipset.
naddy [Wed, 14 Jul 2010 19:24:27 +0000 (19:24 +0000)]
Newer 8168x and 810x chipsets have a different way of stopping the chipset.
From FreeBSD via Brad.

14 years agoMake pane/window wrapping more logical (so with 10 windows, +10 from
nicm [Wed, 14 Jul 2010 18:37:49 +0000 (18:37 +0000)]
Make pane/window wrapping more logical (so with 10 windows, +10 from
window 5 stays in the same place), and tidy the code. From Tiago Cunha.

14 years agoAnother hand knitted offsetof() that needs to die. This time steal the
claudio [Wed, 14 Jul 2010 17:09:13 +0000 (17:09 +0000)]
Another hand knitted offsetof() that needs to die. This time steal the
mask2prefixlen6 version from bgpd.