]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agoAT_DEBUG and AT_BRK were OBE like 10 years ago, so retire them.
imp [Wed, 17 Dec 2008 06:56:58 +0000 (06:56 +0000)]
AT_DEBUG and AT_BRK were OBE like 10 years ago, so retire them.

Reviewed by: peter

15 years agoRemove obsolete AT_DEBUG stuff. It never should have been committed
imp [Wed, 17 Dec 2008 06:11:42 +0000 (06:11 +0000)]
Remove obsolete AT_DEBUG stuff.  It never should have been committed
in the first place, let alone migrated to linux emulation.

Reviewed by: peter, rdivacky

15 years agoFor RTL8168C SPIN2 controllers, make sure to take the controller
yongari [Wed, 17 Dec 2008 06:01:03 +0000 (06:01 +0000)]
For RTL8168C SPIN2 controllers, make sure to take the controller
out of sleep mode prior to accessing to PHY. This should fix device
attach failure seen on these controllers. Also enable the sleep
mode when device is put into sleep state.

PR: kern/123123, kern/123053

15 years agoavoid trying to acquire a shared lock while holding an exclusive lock
kmacy [Wed, 17 Dec 2008 04:33:52 +0000 (04:33 +0000)]
avoid trying to acquire a shared lock while holding an exclusive lock
by making the ifnet lock acquisition exclusive

15 years ago1. Correct capitalization of a closing tag (no effect on output).
murray [Wed, 17 Dec 2008 04:15:38 +0000 (04:15 +0000)]
1. Correct capitalization of a closing tag (no effect on output).
2. Ensure all &arch entities end with ';' (no effect on output).
3. Ensure all &arch lists for drivers/features are comma-separated.

PR: docs/127840 (item 2 only)
MFC after: 1 week

15 years agomerge in 2 buf_ring helper routines for enqueueing and freeing buf_rings
kmacy [Wed, 17 Dec 2008 04:00:43 +0000 (04:00 +0000)]
merge in 2 buf_ring helper routines for enqueueing and freeing buf_rings

15 years ago1. Update the list of Intel chips which have EM64T and can run the amd64
murray [Wed, 17 Dec 2008 03:24:55 +0000 (03:24 +0000)]
1. Update the list of Intel chips which have EM64T and can run the amd64
   port.

2. Increase the known working maximum memory configuration from
   8gb to 32gb.

PR: docs/102148 (1)
Submitted by: Mike Meyer <mwm@mired.org> (1 - partially)
Reviewed by: hrs
Approved by: re (hrs)
MFC after: 1 minute

15 years agoRemove file no longer part of sysinstall.
kensmith [Wed, 17 Dec 2008 02:54:18 +0000 (02:54 +0000)]
Remove file no longer part of sysinstall.

15 years agoAs threatened a while ago, remove support for tape media. Reports vary
kensmith [Wed, 17 Dec 2008 02:52:45 +0000 (02:52 +0000)]
As threatened a while ago, remove support for tape media.  Reports vary
but most people believe it stopped working a while ago.

15 years agodefault to doing lla_lookup with shared afdata lock and returning a
kmacy [Wed, 17 Dec 2008 00:14:28 +0000 (00:14 +0000)]
default to doing lla_lookup with shared afdata lock and returning a
shared lock on the lle - thus restoring parallel performance to
pre-arpv2 level

15 years agoconvert ifnet and afdata locks from mutexes to rwlocks
kmacy [Wed, 17 Dec 2008 00:11:56 +0000 (00:11 +0000)]
convert ifnet and afdata locks from mutexes to rwlocks

15 years ago* Compare pointer with NULL
kmacy [Tue, 16 Dec 2008 23:56:24 +0000 (23:56 +0000)]
* Compare pointer with NULL
* Remove trailing whitespace (added in r186162)
* Reduce indentation by rephrasing test

Submitted by: Christopher Mallon (christoph dot mallon at gmx dot de)

15 years ago1) Fix a deadlock in the VFS:
attilio [Tue, 16 Dec 2008 23:16:10 +0000 (23:16 +0000)]
1) Fix a deadlock in the VFS:
- threadA runs vfs_rel(mp1)
- threadB does unmount the mp1 fs, sets MNTK_UNMOUNT and drop MNT_ILOCK()
- threadA runs vfs_busy(mp1) and, as long as, MNTK_UNMOUNT is set, sleeps
  waiting for threadB to complete the unmount
- threadB, in vfs_mount_destroy(), finds mnt_lock > 0 and sleeps waiting
  for the refcount to expire.

Fix the deadlock by adding a flag called MNTK_REFEXPIRE which signals the
unmounter is waiting for mnt_ref to expire.
The vfs_busy contenders got awake, fails, and if they retry the
MNTK_REFEXPIRE won't allow them to sleep again.

2) Simplify significantly the code of vfs_mount_destroy() trimming
   unnecessary codes:
   - as long as any reference exited, it is no-more possible to have
     write-op (primarty and secondary) in progress.
   - it is no needed to drop and reacquire the mount lock.
   - filling the structures with dummy values is unuseful as long as
     it is going to be freed.

Tested by: pho, Andrea Barberio <insomniac at slackware dot it>
Discussed with: kib

15 years ago- Simplify handling of the deferring of mbuf transmit until after lle lock drop
kmacy [Tue, 16 Dec 2008 23:06:36 +0000 (23:06 +0000)]
- Simplify handling of the deferring of mbuf transmit until after lle lock drop
- add a couple of comments to clarify intent

15 years agoAlso propagate the if_hwassist value to the parent so that cksum offload works.
thompsa [Tue, 16 Dec 2008 22:16:34 +0000 (22:16 +0000)]
Also propagate the if_hwassist value to the parent so that cksum offload works.

Submitted by: Tom Hicks (thicks_averesys.com)

15 years agoAccording to phk@, VOP_STRATEGY should never, _ever_, return
trasz [Tue, 16 Dec 2008 21:13:11 +0000 (21:13 +0000)]
According to phk@, VOP_STRATEGY should never, _ever_, return
anything other than 0.  Make it so.  This fixes
"panic: VOP_STRATEGY failed bp=0xc320dd90 vp=0xc3b9f648",
encountered when writing to an orphaned filesystem.  Reason
for the panic was the following assert:
KASSERT(i == 0, ("VOP_STRATEGY failed bp=%p vp=%p", bp, bp->b_vp));
at vfs_bio:bufstrategy().

Reviewed by: scottl, phk
Approved by: rwatson (mentor)
Sponsored by: FreeBSD Foundation

15 years agoFix forced mdconfig -du. E.g. the following would previously
trasz [Tue, 16 Dec 2008 20:59:27 +0000 (20:59 +0000)]
Fix forced mdconfig -du.  E.g. the following would previously
result in panic:

mdconfig -af blah.img -o force
mount /dev/md0 /mnt
mdconfig -du 0

Reviewed by: scottl
Approved by: rwatson (mentor)
Sponsored by: FreeBSD Foundation

15 years agoStart to clean up the MIPS elf machine dependent file.
imp [Tue, 16 Dec 2008 20:07:47 +0000 (20:07 +0000)]
Start to clean up the MIPS elf machine dependent file.
o Add support for compiling elf64 for this file (the rest of the changes are
  coming later)
o Fill in some misssing relocation types.  We need to support these in
  elf_machdep.c's relocation routines eventually, but that's future work
  too.

15 years agoMips never had /usr/libexec/ld-elf.so.1, so remove it. We don't have
imp [Tue, 16 Dec 2008 20:04:02 +0000 (20:04 +0000)]
Mips never had /usr/libexec/ld-elf.so.1, so remove it.  We don't have
any binaries that need it.  This was moved before FreeBSD 5.x.

15 years agoUnroll two loops of SHA1Update(). 60 bytes of static memory is not a price.
mav [Tue, 16 Dec 2008 19:15:31 +0000 (19:15 +0000)]
Unroll two loops of SHA1Update(). 60 bytes of static memory is not a price.

15 years agoImplement g_vfs_orphan(). Without it, the filesystem never closes
trasz [Tue, 16 Dec 2008 17:04:52 +0000 (17:04 +0000)]
Implement g_vfs_orphan().  Without it, the filesystem never closes
the device, which means refcount on periph drivers never drops,
which means cam_sim_free() never returns, which results in umass
sleeping there ad infinitum.

Submitted by: pjd
Reviewed by: scottl, pjd
Approved by: rwatson (mentor)
Sponsored by: FreeBSD Foundation

15 years agoA few locking fixes and cleanups to pfil hook registration,
rwatson [Tue, 16 Dec 2008 17:03:22 +0000 (17:03 +0000)]
A few locking fixes and cleanups to pfil hook registration,
unregistration, and execution:

- Add some brackets for clarity and trim a bit of vertical whitespace.
- Remove comments that may not contribute to clarity, such as "Lock"
  before acquiring a lock and "Get memory" before allocating memory.
- During hook registration, don't drop pfil_list_lock between checking
  for a duplicate and registering the hook, as this leaves a race
  condition by failing to enforce the "no duplicate hooks" invariant.
- Don't lock the hook during registration, since it's not yet in use.
- Document assumption that hooks will be quiesced before being
  unregistered.
- Don't write-lock hooks during removal because they are assumed
  quiesced.
- Rename "done" label to "locked_error" to be clear that it's an error
  path on the way out of hook execution.

MFC after: pretty soon

15 years agoFix locking in periph drivers - don't try to unlock periph
trasz [Tue, 16 Dec 2008 17:01:52 +0000 (17:01 +0000)]
Fix locking in periph drivers - don't try to unlock periph
that was already deallocated.

Reviewed by: scottl
Approved by: rwatson (mentor)
Sponsored by: FreeBSD Foundation

15 years agoAdd SIM refcounting. This is slightly different from what DragonFly
trasz [Tue, 16 Dec 2008 16:57:33 +0000 (16:57 +0000)]
Add SIM refcounting.  This is slightly different from what DragonFly
does - in DragonFly, it's cam_sim_release() what actually frees the
SIM; cam_sim_free does nothing more than calling cam_sim_release().
Here, we drain in cam_sim_free, waiting for refcount to drop to zero.
We cannot do the same think DragonFly does, because after cam_sim_free
returns, client would destroy the sim->mtx, and CAM would trip over
an initialized mutex.

Reviewed by: scottl
Approved by: rwatson (mentor)
Sponsored by: FreeBSD Foundation

15 years agoGet rid of dead_sim. There is no way to make it work - any attempt
trasz [Tue, 16 Dec 2008 16:54:51 +0000 (16:54 +0000)]
Get rid of dead_sim.  There is no way to make it work - any attempt
to actually use it would panic on mtx operation, as dead_sim doesn't
have a proper mutex.  Even if it had a properly initialized mutex,
it wouldn't have properly locked and owned one.

Reviewed by: scottl
Approved by: rwatson (mentor)
Sponsored by: FreeBSD Foundation

15 years agoAdd confirmation of the password that got entered.
kensmith [Tue, 16 Dec 2008 16:15:19 +0000 (16:15 +0000)]
Add confirmation of the password that got entered.

PR: 29893 (among others)

15 years agoCall ata_legacy() only once on attach and save it's result. Scanning PCI
mav [Tue, 16 Dec 2008 16:04:40 +0000 (16:04 +0000)]
Call ata_legacy() only once on attach and save it's result. Scanning PCI
configuration registers (which are not going to change) on every interrupt
looks expensive, especially when interrupt is shared. Profiling shows me 3%
of time spent by atapci0 on pure network load due to IRQ sharing with em0.

15 years agoMake it possible to cherry-pick packages out of a directory.
phk [Tue, 16 Dec 2008 15:05:52 +0000 (15:05 +0000)]
Make it possible to cherry-pick packages out of a directory.

15 years agoIPFW's pfil hook/unhook code ignores the return values of pfil_add_hook()
rwatson [Tue, 16 Dec 2008 15:05:35 +0000 (15:05 +0000)]
IPFW's pfil hook/unhook code ignores the return values of pfil_add_hook()
and pfil_remove_hook(), so cast them to (void).

MFC after: pretty soon

15 years agoAvoid 256 integer divisions per rc4_init() call. Replace it with using
mav [Tue, 16 Dec 2008 13:58:37 +0000 (13:58 +0000)]
Avoid 256 integer divisions per rc4_init() call. Replace it with using
separate index variable.

It gives more then double rc4_init() performance increase on tested i386 P4.
It also gives about 15% speedup to PPTP VPN with stateless MPPE encryption
(by ng_mppc) which calls rc4_init() for every packet.

15 years agoipfw doesn't use the radix node head lock to protect the radix tree - remove acquisition
kmacy [Tue, 16 Dec 2008 11:06:30 +0000 (11:06 +0000)]
ipfw doesn't use the radix node head lock to protect the radix tree - remove acquisition

15 years agoBug fixes:
jkoshy [Tue, 16 Dec 2008 11:04:02 +0000 (11:04 +0000)]
Bug fixes:
- Initialize variables before use.
- Remove a KASSERT() that could falsely trigger if there are other sources
  of NMIs in the system.

Efficiency tweak:
- When checking PMCs that overflowed, ignore PMCs that were not configured for
  sampling.

15 years agoremove assertion checks for now - ipfw uses its own lock for protecting its radix...
kmacy [Tue, 16 Dec 2008 11:01:36 +0000 (11:01 +0000)]
remove assertion checks for now - ipfw uses its own lock for protecting its radix tree instance

15 years agoRephrase and/or fix some comments in Synaptics touchpad initialization
dumbbell [Tue, 16 Dec 2008 09:51:13 +0000 (09:51 +0000)]
Rephrase and/or fix some comments in Synaptics touchpad initialization
function.

15 years agoFix printing of shared memory limits.
ru [Tue, 16 Dec 2008 09:17:05 +0000 (09:17 +0000)]
Fix printing of shared memory limits.

MFC after: 3 days

15 years agocheck pointers against NULL
kmacy [Tue, 16 Dec 2008 06:01:08 +0000 (06:01 +0000)]
check pointers against NULL

15 years agoDon't count InFramesL2FilterDiscards into Ierr. This value does not represent
delphij [Tue, 16 Dec 2008 05:03:22 +0000 (05:03 +0000)]
Don't count InFramesL2FilterDiscards into Ierr.  This value does not represent
a real packet error but simply indicate that an unexpected unicast or multicast
error was received by the NIC, which was not counted in the past as well.

Reported by: many (on -stable@)
Reviewed by: davidch
MFC after: 3 days

15 years agoTest whether sc->tx_mbuf_map[i], not whether sc->tx_mbuf_map is NULL before
delphij [Tue, 16 Dec 2008 04:59:04 +0000 (04:59 +0000)]
Test whether sc->tx_mbuf_map[i], not whether sc->tx_mbuf_map is NULL before
doing bus_dmamap_sync() since it operates on the former, not the latter.

Reviewed by: davidch

15 years agostyle and spelling fix
kmacy [Tue, 16 Dec 2008 04:41:39 +0000 (04:41 +0000)]
style and spelling fix

15 years agoassert that the radix node head is locked when manipulating the tree
kmacy [Tue, 16 Dec 2008 04:40:43 +0000 (04:40 +0000)]
assert that the radix node head is locked when manipulating the tree

15 years agoHandle VFS_VGET() failing with an error other than EOPNOTSUPP in addition
kensmith [Tue, 16 Dec 2008 04:34:09 +0000 (04:34 +0000)]
Handle VFS_VGET() failing with an error other than EOPNOTSUPP in addition
to failing with that error.

PR: 125149
Submitted by: Jaakko Heinonen (jh <at> saunalahti <dot> fi)
Reviewed by: mohans, kan
MFC after: 3 days

15 years agocheck pointer against NULL
kmacy [Tue, 16 Dec 2008 03:18:59 +0000 (03:18 +0000)]
check pointer against NULL
add new line after declaration for style

15 years agoconvert more pointer validation checks to checking against NULL
kmacy [Tue, 16 Dec 2008 03:12:44 +0000 (03:12 +0000)]
convert more pointer validation checks to checking against NULL

15 years agosimplify locking in find_pfxlist_reachable_router
kmacy [Tue, 16 Dec 2008 03:05:18 +0000 (03:05 +0000)]
simplify locking in find_pfxlist_reachable_router

15 years agodon't unlock lle if it is NULL
kmacy [Tue, 16 Dec 2008 02:48:12 +0000 (02:48 +0000)]
don't unlock lle if it is NULL

15 years agoexplicitly check return of lla_lookup against NULL
kmacy [Tue, 16 Dec 2008 02:47:22 +0000 (02:47 +0000)]
explicitly check return of lla_lookup against NULL

15 years agoadvance tail pointer in nd6_output_lle and check lla_output return against NULL
kmacy [Tue, 16 Dec 2008 02:33:53 +0000 (02:33 +0000)]
advance tail pointer in nd6_output_lle and check lla_output return against NULL

15 years agocheck return from lla_lookup against NULL not zero
kmacy [Tue, 16 Dec 2008 02:30:42 +0000 (02:30 +0000)]
check return from lla_lookup against NULL not zero

15 years agomake sure redirect doesn't return without dropping the lock
kmacy [Tue, 16 Dec 2008 02:06:26 +0000 (02:06 +0000)]
make sure redirect doesn't return without dropping the lock

15 years agoneed to check that lle is not null before unlock if the break condition is not met
kmacy [Tue, 16 Dec 2008 02:05:11 +0000 (02:05 +0000)]
need to check that lle is not null before unlock if the break condition is not met
also fix the break condition to explicitly check against NULL

15 years agounlock the llentry after use in find_pfxlist_reachable_router
kmacy [Tue, 16 Dec 2008 01:58:30 +0000 (01:58 +0000)]
unlock the llentry after use in find_pfxlist_reachable_router

15 years agoIf possible, try to obtain max_mhz on cpufreq attach instead of first request.
mav [Tue, 16 Dec 2008 01:24:05 +0000 (01:24 +0000)]
If possible, try to obtain max_mhz on cpufreq attach instead of first request.

On HyperThreading CPUs logical cores have same frequency, so setting it
on any core will change the other's one. In most cases first request
to the second core will be the "set" request, done after setting frequency
of the first core. In such case second CPU will obtain throttled frequency
of the first core as it's max_mhz making cpufreq broken due to different
frequency sets.

15 years agoInitialize the variable "router", and apply "static_route" flag
qingli [Tue, 16 Dec 2008 01:21:19 +0000 (01:21 +0000)]
Initialize the variable "router", and apply "static_route" flag
across the entire nd6_cache_lladdr() function.

15 years agoStop treating Xorg as a distribution in the mainline portion of sysinstall
kensmith [Tue, 16 Dec 2008 01:18:10 +0000 (01:18 +0000)]
Stop treating Xorg as a distribution in the mainline portion of sysinstall
and leave it to be handled in the packages section (or post-install
completely) along with all the other packages.

15 years agofix comment
sam [Tue, 16 Dec 2008 01:17:36 +0000 (01:17 +0000)]
fix comment

Submitted by: Daan Vreeken

15 years agounlock and destroy an llentry's lock before freeing
kmacy [Tue, 16 Dec 2008 00:20:49 +0000 (00:20 +0000)]
unlock and destroy an llentry's lock before freeing

Found by: sam

15 years agoadd macro for destroying an llentry's rwlock
kmacy [Tue, 16 Dec 2008 00:20:15 +0000 (00:20 +0000)]
add macro for destroying an llentry's rwlock

15 years agounlock looked up llentrys in defrouter_select
kmacy [Tue, 16 Dec 2008 00:18:04 +0000 (00:18 +0000)]
unlock looked up llentrys in defrouter_select

15 years agofix two use after frees in nd6_cache_lladdr caused by last minute unlock shuffling
kmacy [Tue, 16 Dec 2008 00:16:51 +0000 (00:16 +0000)]
fix two use after frees in nd6_cache_lladdr caused by last minute unlock shuffling

15 years agoAdd some HDMI codec IDs.
mav [Tue, 16 Dec 2008 00:08:51 +0000 (00:08 +0000)]
Add some HDMI codec IDs.

15 years agoAdd VIA VT1708B codec IDs.
mav [Mon, 15 Dec 2008 23:49:09 +0000 (23:49 +0000)]
Add VIA VT1708B codec IDs.

15 years agoFix ucastkey teardown for drivers that plumb clear keys or similar
sam [Mon, 15 Dec 2008 22:50:50 +0000 (22:50 +0000)]
Fix ucastkey teardown for drivers that plumb clear keys or similar
(e.g. ath): we must check the key index and not whether the key
points at a cipher other than "undef".  This looks like it's been
broken for a while.  Might be worth adding an explicit clear cipher
at some point though this would require changes to the usage of
IEEE80211_KEY_UNDEFINED.

PR: 125906

15 years agoremove extraneous ';'
sam [Mon, 15 Dec 2008 22:31:54 +0000 (22:31 +0000)]
remove extraneous ';'

15 years agoBy default assume a 8139 chip if the EEPROM contents prove inconclusive. The
wilko [Mon, 15 Dec 2008 21:54:23 +0000 (21:54 +0000)]
By default assume a 8139 chip if the EEPROM contents prove inconclusive.  The
same LOM hardware with goofed-up EEPROM programming also needed reading the
Ethernet address from the chips registers as the EEPROM did not have a
sensible address programmed.

Patch developed by: pyun@
Funky hardware on loan:  www.id-it.nl

MFC after: 2 weeks

15 years agoAnother step assimilating IPv[46] PCB code - directly use
bz [Mon, 15 Dec 2008 21:50:54 +0000 (21:50 +0000)]
Another step assimilating IPv[46] PCB code - directly use
the inpcb names rather than the following IPv6 compat macros:
in6pcb,in6p_sp, in6p_ip6_nxt,in6p_flowinfo,in6p_vflag,
in6p_flags,in6p_socket,in6p_lport,in6p_fport,in6p_ppcb and
sotoin6pcb().

Apart from removing duplicate code in netipsec, this is a pure
whitespace, not a functional change.

Discussed with: rwatson
Reviewed by: rwatson (version before review requested changes)
MFC after: 4 weeks (set the timer and see then)

15 years agoDon't try reading the SXP_PINS_DIFF on the 10160 and 12160 SCSI
marius [Mon, 15 Dec 2008 21:42:38 +0000 (21:42 +0000)]
Don't try reading the SXP_PINS_DIFF on the 10160 and 12160 SCSI
controllers. Reading this register, for which there are indications
that it doesn't really exist, returns 0 on at least some 12160
and doing so on Sun Fire V880 causes a data access error exception.

Reported and tested by: Beat Gaetzi
Approved by: mjacob
Obtained from: OpenBSD (modulo setting isp_lvdmode)

15 years agoUnbreak ofwdump build by moving the pcell_t definition to after the kernel
nwhitehorn [Mon, 15 Dec 2008 18:20:59 +0000 (18:20 +0000)]
Unbreak ofwdump build by moving the pcell_t definition to after the kernel
types headers, and inside the _KERNEL ifdef.

Pointy hat to: me

15 years agoChange Dell's gen2 catch-all from Dell PERC H700/H800 to Dell PERC Gen2
ambrisko [Mon, 15 Dec 2008 17:11:40 +0000 (17:11 +0000)]
Change Dell's gen2 catch-all from Dell PERC H700/H800 to Dell PERC Gen2
and bump the driver version from 2 to 3 and fix the related style problem.

Suggested by: LSI

15 years agoDon't tell the user to remove the media until its actually possible for
kensmith [Mon, 15 Dec 2008 15:46:56 +0000 (15:46 +0000)]
Don't tell the user to remove the media until its actually possible for
them to remove the media.  CDs need to be unmounted before they can be
removed.

15 years agoAdapt parts of the sparc64 Open Firmware bus enumeration code (in particular,
nwhitehorn [Mon, 15 Dec 2008 15:31:10 +0000 (15:31 +0000)]
Adapt parts of the sparc64 Open Firmware bus enumeration code (in particular,
the code for parsing interrupt maps) to PowerPC and reflect their new MI
status by moving them to the shared dev/ofw directory.

This commit also modifies the OFW PCI enumeration procedure on PowerPC to
allow the bus to find non-firmware-enumerated devices that Apple likes to add,
and adds some useful Open Firmware properties (compat and name) to the pnpinfo
string of children on OFW SBus, EBus, PCI, and MacIO links. Because of the
change to PCI enumeration on PowerPC, X has started working again on PPC
machines with Grackle hostbridges.

Reviewed by: marius
Obtained from: sparc64

15 years ago- Disambiguate a few panic messages.
jkoshy [Mon, 15 Dec 2008 14:41:55 +0000 (14:41 +0000)]
- Disambiguate a few panic messages.
- Style fixes: wrap long lines, parenthesize return values.

15 years agoElaborate on the hardware features fxp(4) supports.
brueffer [Mon, 15 Dec 2008 14:07:07 +0000 (14:07 +0000)]
Elaborate on the hardware features fxp(4) supports.

Submitted by: yongari
MFC after: 1 week

15 years agoAdd a symlink from enc.4 to if_enc.4.
bz [Mon, 15 Dec 2008 13:19:11 +0000 (13:19 +0000)]
Add a symlink from enc.4 to if_enc.4.

Reviewed by: brueffer
MFC after: 4 weeks

15 years agoAdd support for HTTP 1.1 If-Modified-Since behavior.
murray [Mon, 15 Dec 2008 08:27:44 +0000 (08:27 +0000)]
Add support for HTTP 1.1 If-Modified-Since behavior.

fetch(1) accepts a new argument -i <file> that if specified will cause
the file to be downloaded only if it is more recent than the mtime of
<file>.

libfetch(3) accepts the mtime in the url structure and a flag to
indicate when this behavior is desired.

PR: bin/87841
Submitted by: Jukka A. Ukkonen <jau@iki.fi> (partially)
Reviewed by: des, ru
MFC after: 3 weeks

15 years agore-add accidentally deleted line
kmacy [Mon, 15 Dec 2008 07:51:59 +0000 (07:51 +0000)]
re-add accidentally deleted line

15 years agoAdd arpv2 management code
kmacy [Mon, 15 Dec 2008 06:53:09 +0000 (06:53 +0000)]
Add arpv2 management code

15 years agoThis main goals of this project are:
qingli [Mon, 15 Dec 2008 06:10:57 +0000 (06:10 +0000)]
This main goals of this project are:
1. separating L2 tables (ARP, NDP) from the L3 routing tables
2. removing as much locking dependencies among these layers as
   possible to allow for some parallelism in the search operations
3. simplify the logic in the routing code,

The most notable end result is the obsolescent of the route
cloning (RTF_CLONING) concept, which translated into code reduction
in both IPv4 ARP and IPv6 NDP related modules, and size reduction in
struct rtentry{}. The change in design obsoletes the semantics of
RTF_CLONING, RTF_WASCLONE and RTF_LLINFO routing flags. The userland
applications such as "arp" and "ndp" have been modified to reflect
those changes. The output from "netstat -r" shows only the routing
entries.

Quite a few developers have contributed to this project in the
past: Glebius Smirnoff, Luigi Rizzo, Alessandro Cerri, and
Andre Oppermann. And most recently:

- Kip Macy revised the locking code completely, thus completing
  the last piece of the puzzle, Kip has also been conducting
  active functional testing
- Sam Leffler has helped me improving/refactoring the code, and
  provided valuable reviews
- Julian Elischer setup the perforce tree for me and has helped
  me maintaining that branch before the svn conversion

15 years agoSpell "substitution" correctly.
kientzle [Mon, 15 Dec 2008 06:00:25 +0000 (06:00 +0000)]
Spell "substitution" correctly.

PR: bin/128616
Submitted by: Jaakko Heinonen
Pointy hat: Tim Kientzle
MFC after: 30 days

15 years agoback out unintended change
sam [Mon, 15 Dec 2008 01:44:23 +0000 (01:44 +0000)]
back out unintended change

15 years agoFix definition of IEEE80211_CHAN_MAX; it was defined as 255 but
sam [Mon, 15 Dec 2008 01:26:33 +0000 (01:26 +0000)]
Fix definition of IEEE80211_CHAN_MAX; it was defined as 255 but
really was meant to be 256.  Adjust usage accordingly and replace
bogus usage of this value in checking IEEE channel #'s.

NB: this causes an ABI change; ifconfig must be recompiled

15 years agoo fix DEBUG sku definition
sam [Mon, 15 Dec 2008 01:18:11 +0000 (01:18 +0000)]
o fix DEBUG sku definition
o add definitions for 900MHz radios: Ubiquiti SR9 and XR9, and ZComax GZ-901

15 years agomove channel parsing to a getchannel routine so it can be reused to check
sam [Mon, 15 Dec 2008 01:15:15 +0000 (01:15 +0000)]
move channel parsing to a getchannel routine so it can be reused to check
the channel argument supplied to chanswitch

15 years agofix handling of sku codes like country codes
sam [Mon, 15 Dec 2008 01:10:52 +0000 (01:10 +0000)]
fix handling of sku codes like country codes

15 years agofix handling of unknown country codes; atoi doesn't return -1
sam [Mon, 15 Dec 2008 01:10:08 +0000 (01:10 +0000)]
fix handling of unknown country codes; atoi doesn't return -1
for an invalid string as I thought; so use strtol instead

15 years ago0 is a potential ISO CC; use new NO_COUNTRY #define to identify
sam [Mon, 15 Dec 2008 01:09:01 +0000 (01:09 +0000)]
0 is a potential ISO CC; use new NO_COUNTRY #define to identify
when the CC is not set.  Note NO_COUNTRY is set to 0xffff for now
(must be 16 bits as ieee80211_regdomain struct defines sku's and
cc's as uint16_t which may need fixing).

15 years agoo distinguish between adhoc and ahdemo modes
sam [Mon, 15 Dec 2008 01:06:49 +0000 (01:06 +0000)]
o distinguish between adhoc and ahdemo modes
o do not require 1/2 and 1/4 rate channels be present in the
  calibration list when doing a gsm regulatory change; the
  existing 900MHz cards are not self-identifying so there is
  no way (using the calibration channel list) to check

15 years agoImprove regdomain.xml parser:
sam [Mon, 15 Dec 2008 01:00:18 +0000 (01:00 +0000)]
Improve regdomain.xml parser:
o store XML_Parser in the state block so we can report line numbers for errors
o complain about netband w/o mode
o complain about unknown modes
o complain about band w/o enclosing netband
o complain about duplicate freqband
o complain about unknown channel flags
o complain about band w/o freqband's
o complain about band w/o maxpower
o complain about country w/o ISO cc
o complain about country w/o regdomain reference

15 years agoReplace adhoc checks in ieee80211_start with a per-node flag that
sam [Mon, 15 Dec 2008 00:47:24 +0000 (00:47 +0000)]
Replace adhoc checks in ieee80211_start with a per-node flag that
indicates if an association id is required before outbound traffic
is permitted.  This cleans up the previous change that broke mcast
traffic "to the stack" in ap mode as a side effect.

Reviewed by: sephe, thompsa, weongyo

15 years agofix ini setup
sam [Mon, 15 Dec 2008 00:29:26 +0000 (00:29 +0000)]
fix ini setup

Submitted by: Jiri Fojtasek <jiri.fojtasek@hlohovec.net>

15 years agoMake the tp pointer available for debugging.
phk [Sun, 14 Dec 2008 23:24:00 +0000 (23:24 +0000)]
Make the tp pointer available for debugging.

15 years agoAdd __packed to the part sub-structure to compensate for
marcel [Sun, 14 Dec 2008 23:12:36 +0000 (23:12 +0000)]
Add __packed to the part sub-structure to compensate for
the 32-bit default structure alignment on ARM.

15 years agoAdd an ENVIRONMENT section to note that the RSH variable can be set to
murray [Sun, 14 Dec 2008 22:48:48 +0000 (22:48 +0000)]
Add an ENVIRONMENT section to note that the RSH variable can be set to
override the use of the rsh(1) program in rcmd(3).

PR: docs/51133
Obtained from: NetBSD, partially
MFC after: 3 weeks

15 years agoCombine ath rate control modules with ath after bringing ath_hal
sam [Sun, 14 Dec 2008 22:26:57 +0000 (22:26 +0000)]
Combine ath rate control modules with ath after bringing ath_hal
src into the tree.  The old split was balanced on module dependencies
and symbol exposure that no longer exists.  Users that want a module
setup with rate control algorithm other than sample must override
ATH_RATE in the ath module Makefile.

Reviewed by: imp

15 years agoTo avoid one doubtless netgraph SMP scalability limitation point, switch
mav [Sun, 14 Dec 2008 20:15:30 +0000 (20:15 +0000)]
To avoid one doubtless netgraph SMP scalability limitation point, switch
node queues processing from single swi:net thread to several specialized
threads.

Reviewed by: julian
Tested with: Netperf Cluster

15 years agoBump the FTDI receive buffer size to 256 in order to improve throughput.
phk [Sun, 14 Dec 2008 20:05:03 +0000 (20:05 +0000)]
Bump the FTDI receive buffer size to 256 in order to improve throughput.

15 years agoMove the code that injects received characters into the tty system into
phk [Sun, 14 Dec 2008 20:03:46 +0000 (20:03 +0000)]
Move the code that injects received characters into the tty system into
a separate public function ucomrxchars(), to avoid requirement of
simple metadata prefixing on the USB data stream.

15 years agoMerge the resolver part of BIND 9.4.3 into HEAD.
ume [Sun, 14 Dec 2008 19:39:53 +0000 (19:39 +0000)]
Merge the resolver part of BIND 9.4.3 into HEAD.
It includes the following fix:

  2426.   [bug]           libbind: inet_net_pton() can sometimes return the
                          wrong value if excessively large netmasks are
                          supplied. [RT #18512]

Reported by: Maksymilian Arciemowicz <cxib__at__securityreason.com>

15 years agoprop change.
ume [Sun, 14 Dec 2008 19:35:42 +0000 (19:35 +0000)]
prop change.

15 years agodelete svn:keywords.
ume [Sun, 14 Dec 2008 19:23:19 +0000 (19:23 +0000)]
delete svn:keywords.

15 years agoprop change
ume [Sun, 14 Dec 2008 18:53:37 +0000 (18:53 +0000)]
prop change