]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoThis commit was generated by cvs2svn to compensate for changes in r159399,
marius [Thu, 8 Jun 2006 13:10:51 +0000 (13:10 +0000)]
This commit was generated by cvs2svn to compensate for changes in r159399,
which included commits to RCS files with non-trunk default branches.

18 years ago* elf64-sparc.c (sparc64_elf_adjust_dynamic_symbol): When linking a
marius [Thu, 8 Jun 2006 13:10:51 +0000 (13:10 +0000)]
* elf64-sparc.c (sparc64_elf_adjust_dynamic_symbol): When linking a
non-shared object, do not reserve space in .plt and .rela.plt
for regular symbols neither defined nor referenced in shared objects.

This is a backport of rev. 1.101 (sourceware.org repository) to
Binutils 2.15 which fixes the creation of bogus relocations in the
PLT of Firefox and Thunderbird binaries and which in turn caused
them to segfault in rtld(1). This is committed to the vendor branch
as it doesn't represent a local change but the original vendor fix
is from after elf_link_hash_flags was replaced with bitfields.

PR: sparc64/89486
Approved by: maintainer timeout
Obtained from: NetBSD
MFC after: 1 week

18 years agoinstall_state() should properly initialize 'addr_type' field of newly created
oleg [Thu, 8 Jun 2006 11:27:45 +0000 (11:27 +0000)]
install_state() should properly initialize 'addr_type' field of newly created
flows for O_LIMIT rules.  Otherwise 'ipfw -d show' is unable to display
PARENT rules properly.
(This bug was exposed by ipfw2.c rev.1.90)

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

18 years agoFix following rules: pipe X (tag|altq) Y ...
oleg [Thu, 8 Jun 2006 11:13:23 +0000 (11:13 +0000)]
Fix following rules: pipe X (tag|altq) Y ...

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

18 years agoWhitespace.
glebius [Thu, 8 Jun 2006 10:19:16 +0000 (10:19 +0000)]
Whitespace.

18 years agou_intXX -> uintXX
glebius [Thu, 8 Jun 2006 09:35:02 +0000 (09:35 +0000)]
u_intXX -> uintXX

18 years agoo NIS clients ask for "shells" map not "shells.list".
maxim [Thu, 8 Jun 2006 09:12:07 +0000 (09:12 +0000)]
o NIS clients ask for "shells" map not "shells.list".

PR: kern/86693
Submitted by: Victor Sudakov
MFC after: 1 month

18 years agoFix the last commit.
glebius [Thu, 8 Jun 2006 08:15:21 +0000 (08:15 +0000)]
Fix the last commit.

Submitted by: jhb
Pointy hat to: glebius

18 years agoFix the LOR that occurs when the MAC compiled into the kernel
kib [Thu, 8 Jun 2006 07:55:10 +0000 (07:55 +0000)]
Fix the LOR that occurs when the MAC compiled into the kernel
and vnode is destroyed.

Reviewed by: rwatson
LOR: 189
MFC after: 2 weeks
Approved by: kan (mentor)

18 years agoo Catch up v_flag split and replace VSYSTEM by VV_SYSTEM.
maxim [Thu, 8 Jun 2006 04:39:02 +0000 (04:39 +0000)]
o Catch up v_flag split and replace VSYSTEM by VV_SYSTEM.

Inspired by: jkoshy's commit to vgone.9

18 years agoFix spurious warnings from neighbor discovery when working with IPv6 over
gnn [Thu, 8 Jun 2006 00:31:17 +0000 (00:31 +0000)]
Fix spurious warnings from neighbor discovery when working with IPv6 over
point to point tunnels (gif).

PR: 93220
Submitted by: Jinmei Tatuya
MFC after: 1 week

18 years agoOops it seems I forgot to remove ARM32_NEW_VM_LAYOUT from here.
cognet [Wed, 7 Jun 2006 22:41:14 +0000 (22:41 +0000)]
Oops it seems I forgot to remove ARM32_NEW_VM_LAYOUT from here.

18 years agoo It's lilac-dmc.Berkeley.EDU, not lbl-csam.arpa.
maxim [Wed, 7 Jun 2006 21:37:42 +0000 (21:37 +0000)]
o It's lilac-dmc.Berkeley.EDU, not lbl-csam.arpa.

Obtained from: NetBSD

18 years ago- style(9) cleanup.
glebius [Wed, 7 Jun 2006 21:03:20 +0000 (21:03 +0000)]
- style(9) cleanup.
- Fix comments and printf()s about allocating jumbo buffers.

18 years agoAdd pmap locking to pmap_fault_fixup().
alc [Wed, 7 Jun 2006 20:54:31 +0000 (20:54 +0000)]
Add pmap locking to pmap_fault_fixup().

Add an assertion to pmap_vac_me_harder().

Tested by: cognet@

18 years agobandaid type coercion for ia64
sam [Wed, 7 Jun 2006 20:37:14 +0000 (20:37 +0000)]
bandaid type coercion for ia64

Submitted by: marcel

18 years agoadd ath & co.
sam [Wed, 7 Jun 2006 18:10:28 +0000 (18:10 +0000)]
add ath & co.

MFC after: 1 month

18 years agoreposition defines so things build on architectures where AH_REGOPS_FUNC
sam [Wed, 7 Jun 2006 18:09:33 +0000 (18:09 +0000)]
reposition defines so things build on architectures where AH_REGOPS_FUNC
is defined

18 years agoremove ath hal options; having them here causes opt_ah.h to be clobbered
sam [Wed, 7 Jun 2006 17:53:15 +0000 (17:53 +0000)]
remove ath hal options; having them here causes opt_ah.h to be clobbered
by config and that breaks builds unless one duplicates the options in the
config file

MFC after: 1 month

18 years agoProperly synchronize access to the pmap in pmap_extract_and_hold().
alc [Wed, 7 Jun 2006 17:14:48 +0000 (17:14 +0000)]
Properly synchronize access to the pmap in pmap_extract_and_hold().

Eliminate an unneeded variable from pmap_extract_and_hold().

Tested by: cognet@

18 years agoUpdate geli_swap_flags, -e is now used to specify the encryption algorithm.
brueffer [Wed, 7 Jun 2006 17:14:27 +0000 (17:14 +0000)]
Update geli_swap_flags, -e is now used to specify the encryption algorithm.

18 years agoCatch up with the rename of symbol VDOOMED to VI_DOOMED in revision
jkoshy [Wed, 7 Jun 2006 16:51:54 +0000 (16:51 +0000)]
Catch up with the rename of symbol VDOOMED to VI_DOOMED in revision
1.198 of "sys/sys/vnode.h".  Remove a cross-reference to a manual
page that never existed.

---------------------------------------------------------------------- CVS:
PR:              Fill this in if a GNATS PR is affected by the
change.  CVS: Submitted by:    Fill this in if someone else sent
in the change.  CVS: Reviewed by:     Fill this in if someone else
reviewed your modification.  CVS: Approved by:     Fill this in if
you needed approval for this commit.  CVS: Obtained from:   Fill
this in if the change is from third party software.  CVS: MFC after:
N [day[s]|week[s]|month[s]] CVS:    Fill in to get MFC notification
later. (days assumed unless specified) CVS:
---------------------------------------------------------------------- CVS:
Enter Log.  Lines beginning with `CVS:' are removed automatically

18 years agoUpdate comments in struct protosw to reflect changing times:
rwatson [Wed, 7 Jun 2006 13:09:04 +0000 (13:09 +0000)]
Update comments in struct protosw to reflect changing times:

- Between 1996 and 1997, wollman eliminated pr_usrreq() and replaced it
  with direct function pointers.  Update comment to reflect these changes.

- In 2003, I added pru_sosetlabel().  Update comment to reflect this
  change.

MFC after: 1 week

18 years agoWhen counting nodes second time, use the same criteria as for
glebius [Wed, 7 Jun 2006 12:42:15 +0000 (12:42 +0000)]
When counting nodes second time, use the same criteria as for
the first time.

PR: kern/98529
Submitted by: Michael Heyman

18 years agoAdd device IDs for Linksys PCMPC200 Cardbus card.
glebius [Wed, 7 Jun 2006 12:26:23 +0000 (12:26 +0000)]
Add device IDs for Linksys PCMPC200 Cardbus card.

PR: kern/75582
Submitted by: Gary Palmer

18 years agoNow that we use pmap_mapdev_boostrap(), we can get ride of the got_mmu
cognet [Wed, 7 Jun 2006 11:28:17 +0000 (11:28 +0000)]
Now that we use pmap_mapdev_boostrap(), we can get ride of the got_mmu
hack.

Submitted by: kevlo

18 years agoAdd section describing how ARP load balancing works and its
glebius [Wed, 7 Jun 2006 10:26:51 +0000 (10:26 +0000)]
Add section describing how ARP load balancing works and its
limitations:
  - Can't load balance behind router
  - ARP balancing + pfsync probably won't work

PR: kern/93829
Wording by: scottl

18 years agoReflect latest changes in rc.subr wrt prefix/etcdir variables being set
flz [Wed, 7 Jun 2006 10:20:57 +0000 (10:20 +0000)]
Reflect latest changes in rc.subr wrt prefix/etcdir variables being set
automatically by load_rc_config.

With help from: ru (mdoc-fu)
Reviewed by: brueffer
Approved by: cperciva

18 years agoFix watchdog timeout errors seen on a few systems.
yongari [Wed, 7 Jun 2006 09:05:20 +0000 (09:05 +0000)]
Fix watchdog timeout errors seen on a few systems.
SK-NET GENESIS document says reading SK_ISSR should stop generating
further interrupts(Since we drop a driver lock before invoking
ifp->if_input handler we should disable interrupts in ISR in order
to protect integrity of softc from subsequent interrupts). But it
seems that there is possibility of loosing interrupts between
reading SK_ISSR and determining which interrupts are reported.
To cope with the situation we continuously read SK_ISSR register
until there are no interrupts. However, it seems that the above
work around doesn't fix all cases. To protect watchdog handler
from triggering false alarm add a work around code which try to
reclaim pending Tx descriptors before resetting hardware. This
should fix occasional watchdog timeout errors seen on this driver.

Reported by: Frank Behrens <frank AT pinky dot sax dot de >
Tested by: Frank Behrens <frank AT pinky dot sax dot de >

18 years agoRemove sa1_cache_clean_addr. It isn't needed.
imp [Wed, 7 Jun 2006 05:36:10 +0000 (05:36 +0000)]
Remove sa1_cache_clean_addr.  It isn't needed.

Submitted by: kevlo

18 years agoAdd adding_user.8 to SEE ALSO, note that usernames may contain any character
trhodes [Wed, 7 Jun 2006 04:58:54 +0000 (04:58 +0000)]
Add adding_user.8 to SEE ALSO, note that usernames may contain any character
but not being with a hyphen, similar to adding_user.8.

PR: 35732

18 years agoSimply running ``make world'' will bomb unless you dig up the
trhodes [Wed, 7 Jun 2006 03:33:48 +0000 (03:33 +0000)]
Simply running ``make world'' will bomb unless you dig up the
undocumented HISTORICAL_MAKE_WORLD variable and set it.  Note it
here so the blow up will not really be a surprise to people who
read.

Link the buildingworld chapter of our handbook in the README
while I'm here.

18 years agoUse calloc() instead of zeroing the memory our own.
delphij [Wed, 7 Jun 2006 01:43:26 +0000 (01:43 +0000)]
Use calloc() instead of zeroing the memory our own.

18 years agoAllow to use the old -a option to specify an encryption algorithm to use
pjd [Tue, 6 Jun 2006 22:06:24 +0000 (22:06 +0000)]
Allow to use the old -a option to specify an encryption algorithm to use
(for backward compatibility), but print a warning to inform about the
change.

18 years agoFor network filesystems (NFS, samba, etc.), do not pass "export" nmount()
rodrigc [Tue, 6 Jun 2006 21:56:49 +0000 (21:56 +0000)]
For network filesystems (NFS, samba, etc.), do not pass "export" nmount()
parameter to try to delete an NFS export.  It won't work, and will log
a spurious warning to syslog.

Tested by: Arno J. Klaassen <arno at heho dot snv dot jussieu dot fr>
MFC after: 3 days

18 years agoConvert the last offender, the SA1110 port, to ARM32_NEW_VM_LAYOUT, and
cognet [Tue, 6 Jun 2006 21:06:57 +0000 (21:06 +0000)]
Convert the last offender, the SA1110 port, to ARM32_NEW_VM_LAYOUT, and
completely nuke the !ARM32_NEW_VM_LAYOUT case.

18 years agoRemove a bogus, useless, "i++".
cognet [Tue, 6 Jun 2006 20:47:59 +0000 (20:47 +0000)]
Remove a bogus, useless, "i++".

18 years agoGarbage collect library we did not compile for ages.
kan [Tue, 6 Jun 2006 19:31:45 +0000 (19:31 +0000)]
Garbage collect library we did not compile for ages.

18 years agoClarify and merge two sentences.
brueffer [Tue, 6 Jun 2006 19:03:51 +0000 (19:03 +0000)]
Clarify and merge two sentences.

Discussed with: pjd

18 years agoAlso need to quote values so they can be passed on correctly.
imp [Tue, 6 Jun 2006 18:07:58 +0000 (18:07 +0000)]
Also need to quote values so they can be passed on correctly.

18 years agoAdd partial pmap locking.
alc [Tue, 6 Jun 2006 17:27:53 +0000 (17:27 +0000)]
Add partial pmap locking.

Tested by: cognet@

18 years agoSync to p4:
trhodes [Tue, 6 Jun 2006 17:22:55 +0000 (17:22 +0000)]
Sync to p4:

o Add shutdown KEYWORD;
o Remove PID check;
o Stop auditd with '-t'
o General cleanup.

18 years agoFix an unwanted gcc4 warning.
emax [Tue, 6 Jun 2006 16:58:19 +0000 (16:58 +0000)]
Fix an unwanted gcc4 warning.

Submitted by: delphij
MFC after: 3 days

18 years agoCreate a new target 'buildenvvars'. This target reports the build
imp [Tue, 6 Jun 2006 16:24:54 +0000 (16:24 +0000)]
Create a new target 'buildenvvars'.  This target reports the build
environment for cross building (the same one you'd get interactively
in make buildenv).  This cannot be a simple

        make -f Makefile.inc1 -V WMAKEENV

because in PATH is not set correctly unless one takes a trip through
the Makefile/Makefile.inc1 indirection, the logic of which is too
large to reproduce outside of Makefiles.

18 years ago`flush' is appropriate than `reload'.
ume [Tue, 6 Jun 2006 15:34:50 +0000 (15:34 +0000)]
`flush' is appropriate than `reload'.

Requested by: ceri

18 years agoUse the acpi_event_sleep_button_sleep() function instead of
nyan [Tue, 6 Jun 2006 15:20:06 +0000 (15:20 +0000)]
Use the acpi_event_sleep_button_sleep() function instead of
the acpi_SetSleepState().

Submitted by: njl

18 years agoWhen DIAGNOSTIC is defined, verify if we don't free crypto requests from
pjd [Tue, 6 Jun 2006 15:04:52 +0000 (15:04 +0000)]
When DIAGNOSTIC is defined, verify if we don't free crypto requests from
the crypto queue or from the return queue.

18 years agoForce commit as I forgot to document:
pjd [Tue, 6 Jun 2006 14:50:34 +0000 (14:50 +0000)]
Force commit as I forgot to document:

- Don't initialize crp_olen, it should be set by the driver.
- crp_ilen should contain length of the entier buffer.

18 years ago- Unbreak the build when geli is compiled into the kernel (on as module),
pjd [Tue, 6 Jun 2006 14:48:19 +0000 (14:48 +0000)]
- Unbreak the build when geli is compiled into the kernel (on as module),
  by silencing unfounded compiler warning.

Reported by:

18 years agoMdoc cleanup and some wording improvements.
brueffer [Tue, 6 Jun 2006 14:02:13 +0000 (14:02 +0000)]
Mdoc cleanup and some wording improvements.

18 years agoHandle errors in the same way it is done in safe(4).
pjd [Tue, 6 Jun 2006 13:58:12 +0000 (13:58 +0000)]
Handle errors in the same way it is done in safe(4).

18 years agoDon't increase hst_obytes field twice - it is already done at the begining
pjd [Tue, 6 Jun 2006 13:32:26 +0000 (13:32 +0000)]
Don't increase hst_obytes field twice - it is already done at the begining
of the function.
It was wrong anyway, because we also support uio's structures, not only
mbufs.

18 years agoAdd `reload' which invalidates the cache for every user.
ume [Tue, 6 Jun 2006 12:55:58 +0000 (12:55 +0000)]
Add `reload' which invalidates the cache for every user.

18 years ago- Note that the synchronisation interface needs to be up and have
glebius [Tue, 6 Jun 2006 12:35:53 +0000 (12:35 +0000)]
- Note that the synchronisation interface needs to be up and have
  an IP address assigned.
- Add "quick" keyword to pf.conf example.

PR: docs/85209

18 years agoMake ke_rqindex unsigned.
davidxu [Tue, 6 Jun 2006 12:26:17 +0000 (12:26 +0000)]
Make ke_rqindex unsigned.

18 years agoPrint pfsync interface status if either syncpeer or syncdev is configured.
glebius [Tue, 6 Jun 2006 11:21:08 +0000 (11:21 +0000)]
Print pfsync interface status if either syncpeer or syncdev is configured.

18 years agoExtract pointer value for mnt_stat from vp after the NULL check, not
rwatson [Tue, 6 Jun 2006 08:43:27 +0000 (08:43 +0000)]
Extract pointer value for mnt_stat from vp after the NULL check, not
before.

Coverity ID: 134394
Found with: Coverity Prevent (tm)

18 years agoUse NET_CALLOUT_MPSAFE for netgraph callout initializer.
glebius [Tue, 6 Jun 2006 08:05:27 +0000 (08:05 +0000)]
Use NET_CALLOUT_MPSAFE for netgraph callout initializer.

18 years agoThe procedure of raceless switching between polling mode and
glebius [Tue, 6 Jun 2006 08:03:49 +0000 (08:03 +0000)]
The procedure of raceless switching between polling mode and
taskqueued interrupt mode is going to be quite complex. Since
the polling mode is considered legacy feature for em(4) driver,
the decision is made to make polling and new interrupt handler
mutually exclusive, selected at compile time.

If kernel is compiled with DEVICE_POLLING, the fast taskqueued
interrupt handler code is disabled and the em_poll() and legacy
em_intr() functions are enabled. Otherwise, legacy functions
are disabled and only em_intr_fast() code is compiled.

Discussed with: scottl

18 years agoRemove section committed by mistake. It is not yet ready.
pjd [Tue, 6 Jun 2006 07:10:42 +0000 (07:10 +0000)]
Remove section committed by mistake. It is not yet ready.

18 years agoAdd regression tests for geli(8) authentication feature.
pjd [Tue, 6 Jun 2006 06:46:51 +0000 (06:46 +0000)]
Add regression tests for geli(8) authentication feature.

Supported by: Wheel Sp. z o.o. (http://www.wheel.pl)

18 years ago- Update regression tests after recent changes.
pjd [Tue, 6 Jun 2006 06:44:19 +0000 (06:44 +0000)]
- Update regression tests after recent changes.
- Verify various sector sizes.

18 years agoCleanups for mailwrapper(8):
delphij [Tue, 6 Jun 2006 05:01:12 +0000 (05:01 +0000)]
Cleanups for mailwrapper(8):
 - K&R -> ANSI prototype [O]
 - Do not bother to do free right before exit() or execve() [O]
 - Remove some dead code in addarg()
 - Make additional parameters specified in mailer.conf(5)
   actually work and document the fact. [N]
 - Avoid using __progname but instead use getprogname()
   and setprogname() to provide more sensible messages. [O, N]
 - Update $OpenBSD$ and $NetBSD$ to reflect the fact that we
   have sync'ed with their code.
 - WARNS=6

Obtained from/Inspired by: OpenBSD [O], NetBSD [N] (partially)

18 years agoAdd partial pmap locking.
alc [Tue, 6 Jun 2006 04:32:20 +0000 (04:32 +0000)]
Add partial pmap locking.

Eliminate the unused allpmaps list.

Tested by: cognet@

18 years agoCorrect a typo in the previous revision.
alc [Tue, 6 Jun 2006 02:02:10 +0000 (02:02 +0000)]
Correct a typo in the previous revision.

18 years agoAdd a stub for pmap_enter_object().
alc [Tue, 6 Jun 2006 02:00:08 +0000 (02:00 +0000)]
Add a stub for pmap_enter_object().

18 years agoMake VERBOSE_INIT_ARM compile by fixing various printf formats, and add it
cognet [Tue, 6 Jun 2006 01:14:12 +0000 (01:14 +0000)]
Make VERBOSE_INIT_ARM compile by fixing various printf formats, and add it
as an option.

Submitted by:   Max N. Boyarov <m.boyarov at bsd dot by>

18 years agovm_page_alloc_contig() can sleep, so don't even think about using it
cognet [Mon, 5 Jun 2006 23:42:47 +0000 (23:42 +0000)]
vm_page_alloc_contig() can sleep, so don't even think about using it
in the M_NOWAIT case.

18 years agoReally fix the typo this time: it should be sc->sc_drvbpf to be verified,
avatar [Mon, 5 Jun 2006 23:06:03 +0000 (23:06 +0000)]
Really fix the typo this time: it should be sc->sc_drvbpf to be verified,
not ic->ic_drvbpf.

18 years agoFixing a typo in rev1.196.
avatar [Mon, 5 Jun 2006 22:59:14 +0000 (22:59 +0000)]
Fixing a typo in rev1.196.

18 years agoRemove use of Giant around vn_open() in audit trail setup.
rwatson [Mon, 5 Jun 2006 22:36:12 +0000 (22:36 +0000)]
Remove use of Giant around vn_open() in audit trail setup.

Submitted by: jhb, wsalamon
Obtained from: TrustedBSD Project

18 years agoDo some source && comment cleanup.
mjacob [Mon, 5 Jun 2006 22:25:49 +0000 (22:25 +0000)]
Do some source && comment cleanup.

Clean out the abortive start to homegrown, per-mpt,
Domain Validation. This should really be done at a
higher level.

Use the PIM_SEQSCAN flag for U320- this seems to correct
cases of being unable to consistently negotiate U320 in
the cases where I'd seen this before.

Between this and other recent checkins, this driver is
pretty close to being ready for MFC.

Reviewed by: scottl, ken, scsi@
MFC after: 1 week

18 years agoAdd PIM_SEQSCAN for HBA misc flags and code that understands
mjacob [Mon, 5 Jun 2006 22:22:14 +0000 (22:22 +0000)]
Add PIM_SEQSCAN for HBA misc flags and code that understands
what to do with it.

This forces us to scan targets sequentially, not in parallel.
The reason we might want to do this is that SPI negotiation
might not work right at the SIM level if we try to do it
in parallel. We *could* fix this for each SIM where this is
broken, but it's a lot harder to do that when we can simply
ask CAM to probe sequentially.

If PIM_SEQSCAN is not set (default), the original behaviour for
probing is unchanged.

LUN probing is still done in parallel for each target in either
case.

While we're at it, clean up some resource leakage for error
cases.

Reviewed by: ken, scott, scsi@
MFC after: 1 week

18 years agoDocument geli(8) data authentication.
pjd [Mon, 5 Jun 2006 21:43:51 +0000 (21:43 +0000)]
Document geli(8) data authentication.

Supported by: Wheel Sp. z o.o. (http://www.wheel.pl)

18 years agoConnect new GELI files to the build.
pjd [Mon, 5 Jun 2006 21:42:26 +0000 (21:42 +0000)]
Connect new GELI files to the build.

Supported by: Wheel Sp. z o.o. (http://www.wheel.pl)

18 years agoUserland bits of geli(8) data authentication.
pjd [Mon, 5 Jun 2006 21:40:54 +0000 (21:40 +0000)]
Userland bits of geli(8) data authentication.
Now, encryption algorithm is given using '-e' option, not '-a'.
The '-a' option is now used to specify authentication algorithm.

Supported by: Wheel Sp. z o.o. (http://www.wheel.pl)

18 years agoImplement data integrity verification (data authentication) for geli(8).
pjd [Mon, 5 Jun 2006 21:38:54 +0000 (21:38 +0000)]
Implement data integrity verification (data authentication) for geli(8).

Supported by: Wheel Sp. z o.o. (http://www.wheel.pl)

18 years agoMake kern.geom.eli.overwrites sysctl a tunable as well.
pjd [Mon, 5 Jun 2006 21:25:19 +0000 (21:25 +0000)]
Make kern.geom.eli.overwrites sysctl a tunable as well.

18 years agoAssuming the interface has an address of x.x.x.195, a mask of
qingli [Mon, 5 Jun 2006 21:20:21 +0000 (21:20 +0000)]
Assuming the interface has an address of x.x.x.195, a mask of
255.255.255.0, and a default route with gateway x.x.x.1. Now if
the address mask is changed to something more specific, e.g.,
255.255.255.128, then after the mask change the default gateway
is no longer reachable.

Since the default route is still present in the routing table,
when the output code tries to resolve the address of the default
gateway in function rt_check(), again, the default route will be
returned by rtalloc1(). Because the lock is currently held on the
rtentry structure, one more attempt to hold the lock will trigger
a crash due to "lock recursed on non-recursive mutex ..."

This is a general problem. The fix checks for the above condition
so that an existing route entry is not mistaken for a new cloned
route. Approriately, an ENETUNREACH error is returned back to the
caller

Approved by: andre

18 years agoAdd g_duplicate_bio() function which does the same thing what g_clone_bio()
pjd [Mon, 5 Jun 2006 21:13:22 +0000 (21:13 +0000)]
Add g_duplicate_bio() function which does the same thing what g_clone_bio()
is doing, but g_duplicate_bio() allocates new bio with M_WAITOK flag.

18 years agoIntroduce the function pmap_enter_object(). It maps a sequence of resident
alc [Mon, 5 Jun 2006 20:35:27 +0000 (20:35 +0000)]
Introduce the function pmap_enter_object().  It maps a sequence of resident
pages from the same object.  Use it in vm_map_pmap_enter() to reduce the
locking overhead of premapping objects.

Reviewed by: tegge@

18 years agoSome mdoc and wording improvements.
brueffer [Mon, 5 Jun 2006 20:29:31 +0000 (20:29 +0000)]
Some mdoc and wording improvements.

18 years agoFix kernel panic in rt2661_tx_intr() if no frames has been sent.
fjoe [Mon, 5 Jun 2006 20:06:29 +0000 (20:06 +0000)]
Fix kernel panic in rt2661_tx_intr() if no frames has been sent.

Obtained from: OpenBSD (sys/dev/ic/rt2661.c rev. 1.15)

18 years agoo Add missed comma.
maxim [Mon, 5 Jun 2006 19:55:44 +0000 (19:55 +0000)]
o Add missed comma.

18 years agoJust a very quick update to get *close* to reality.
mjacob [Mon, 5 Jun 2006 19:46:55 +0000 (19:46 +0000)]
Just a very quick update to get *close* to reality.

18 years ago - ANSIfy.
delphij [Mon, 5 Jun 2006 18:22:13 +0000 (18:22 +0000)]
 - ANSIfy.
 - Remove two unnecessary casts.

These changes would help gcc4 compile.

18 years agoFix cut-n-pasteo: use the i386 version #define for i386 dumps, not the amd64 one.
emaste [Mon, 5 Jun 2006 18:21:29 +0000 (18:21 +0000)]
Fix cut-n-pasteo: use the i386 version #define for i386 dumps, not the amd64 one.

18 years agomy(4) provides support now as well.
brueffer [Mon, 5 Jun 2006 18:02:36 +0000 (18:02 +0000)]
my(4) provides support now as well.

18 years agoAdd altq(4) support.
brueffer [Mon, 5 Jun 2006 17:59:46 +0000 (17:59 +0000)]
Add altq(4) support.

Reviewed by: mlaier
Approved by: rwatson (mentor)
MFC after: 2 weeks

18 years agomove hal bus+tag externalization to the bus glue code where it belongs;
sam [Mon, 5 Jun 2006 17:51:20 +0000 (17:51 +0000)]
move hal bus+tag externalization to the bus glue code where it belongs;
this is a noop on all current freebsd architectures

MFC after: 1 month

18 years ago- Declare the PnP map const.
marius [Mon, 5 Jun 2006 17:48:54 +0000 (17:48 +0000)]
- Declare the PnP map const.
- Add devices found in V210 to the PnP map.
- Don't leak memory if we didn't find a match for a node in the PnP map.

MFC after: 2 weeks

18 years agoresolve merge conflicts
sam [Mon, 5 Jun 2006 17:46:41 +0000 (17:46 +0000)]
resolve merge conflicts

MFC after: 1 month

18 years agoThis commit was generated by cvs2svn to compensate for changes in r159285,
sam [Mon, 5 Jun 2006 17:43:51 +0000 (17:43 +0000)]
This commit was generated by cvs2svn to compensate for changes in r159285,
which included commits to RCS files with non-trunk default branches.

18 years agoVersion 0.9.17.2:
sam [Mon, 5 Jun 2006 17:43:51 +0000 (17:43 +0000)]
Version 0.9.17.2:

o define HAL_SOFTC, HAL_BUS_TAG, and HAL_BUS_HANDLE to be machine
  independent; this fixes portability issues with bsd systems
o add ah_disable api for turning off operation of both MAC and PHY
o add ah_getAntennaSwitch and ah_setAntennaSwitch api's for better control
  of antenna usage and diversity
o add ah_setAckCTSRate and ah_setAckCTSRate for controlling tx rate of
  h/w generated frames
o add ah_setBeaconTimers api for simpler setting of the beacon timer registers
o remove ah_waitForBeaconDone api
o add HAL_TXDESC_DURENA flag to enable h/w duration setting in tx descriptor
o correct documentation of min/max tx power units (.5 dBm)
o switch arm, mips, and powerpc builds to use functions for register
  read/write operations
o fix sparc builds to not reference %g2 and %g3 registers
o add public builds for SoC's

MFC after: 1 month

18 years agoUpon further review, DES prefers this change over that in revision 1.13
ghelmer [Mon, 5 Jun 2006 16:41:27 +0000 (16:41 +0000)]
Upon further review, DES prefers this change over that in revision 1.13
to resolve the directory access problem for processes with P_SUGID flag
set.

Suggested by: des

18 years agoFix minor typos.
joel [Mon, 5 Jun 2006 16:31:57 +0000 (16:31 +0000)]
Fix minor typos.

18 years ago- Document that padlock(4) pretends to accelerate HMAC algorithms.
pjd [Mon, 5 Jun 2006 16:24:31 +0000 (16:24 +0000)]
- Document that padlock(4) pretends to accelerate HMAC algorithms.
- Remove "device cryptodev" as it is not needed for compiling padlock(4)
  into the kernel. Actually it is not advisable, because padlock
  instructions can be used directly from userland, so passing the work
  through the kernel is a bad idea.

18 years ago- Pretend to accelerate various HMAC algorithms, so padlock(4) can be used
pjd [Mon, 5 Jun 2006 16:22:04 +0000 (16:22 +0000)]
- Pretend to accelerate various HMAC algorithms, so padlock(4) can be used
  with fast_ipsec(4) and geli(8) authentication (comming soon).
  If consumer requests only for HMAC algorithm (without encryption), return
  EINVAL.
- Add support for the CRD_F_KEY_EXPLICIT flag, for both encryption and
  authentication.

18 years agoWhen generating BSM tokens for mkfifo(), include mode argument.
rwatson [Mon, 5 Jun 2006 16:14:49 +0000 (16:14 +0000)]
When generating BSM tokens for mkfifo(), include mode argument.

Submitted by: wsalamon
Obtained from: TrustedBSD Project

18 years agoWhen generating the process token, need to check whether the
rwatson [Mon, 5 Jun 2006 16:12:00 +0000 (16:12 +0000)]
When generating the process token, need to check whether the
process was sucessfully audited.  Otherwise, generate the PID
token. This change covers the pid < 0 cases, and pid lookup
failure cases.

Submitted by: wsalamon
Obtained from: TrustedBSD Project