]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoo Do not mangle current session user login name with jail -u|-U.
maxim [Sun, 16 Apr 2006 12:32:04 +0000 (12:32 +0000)]
o Do not mangle current session user login name with jail -u|-U.

PR: bin/94730
Submitted by: Frank Behrens
MFC after: 1 month

18 years agoCorrect assorted grammos and typos.
schweikh [Sun, 16 Apr 2006 11:54:01 +0000 (11:54 +0000)]
Correct assorted grammos and typos.

18 years agoFix indentation for "Ostersonntag".
brueffer [Sun, 16 Apr 2006 10:34:56 +0000 (10:34 +0000)]
Fix indentation for "Ostersonntag".

Obtained from: OpenBSD

18 years agoTake a better stab at making this compile.
scottl [Sat, 15 Apr 2006 18:54:56 +0000 (18:54 +0000)]
Take a better stab at making this compile.

18 years agoCorrect debug: we are sending child bio here, not parent bio.
pjd [Sat, 15 Apr 2006 18:30:42 +0000 (18:30 +0000)]
Correct debug: we are sending child bio here, not parent bio.

MFC after: 1 week

18 years agoTake a stab at making this compile.
scottl [Sat, 15 Apr 2006 18:04:04 +0000 (18:04 +0000)]
Take a stab at making this compile.

18 years agoo Document /dev/fido.
maxim [Sat, 15 Apr 2006 17:52:02 +0000 (17:52 +0000)]
o Document /dev/fido.

PR: docs/85425
Approved by: phk
MFC after: 2 weeks

18 years agoo s/you health/your health/.
maxim [Sat, 15 Apr 2006 17:07:32 +0000 (17:07 +0000)]
o s/you health/your health/.

PR: conf/95810
Submitted by: Gavin Atkinson

18 years ago- make reentrant version of netdb functions glibc style API, and
ume [Sat, 15 Apr 2006 16:20:27 +0000 (16:20 +0000)]
- make reentrant version of netdb functions glibc style API, and
  expose them to outside of libc.
- make netdb functions NSS friendly.

Reviewed by: arch@ and current@ (no objection)

18 years agoOops, untested code was included accidentally.
iwasaki [Sat, 15 Apr 2006 16:10:53 +0000 (16:10 +0000)]
Oops, untested code was included accidentally.
Fixed.

18 years agoDon't export __gdtoa. The only known ports that rely on
deischen [Sat, 15 Apr 2006 15:35:20 +0000 (15:35 +0000)]
Don't export __gdtoa.  The only known ports that rely on
this or the deprecated POSIX functions {e,g,f}cvt() have
newer versions that do not (rely on them).

Requested by: marius

18 years agoo Document SO_NOSIGPIPE, touch .Dd.
maxim [Sat, 15 Apr 2006 13:37:35 +0000 (13:37 +0000)]
o Document SO_NOSIGPIPE, touch .Dd.

PR: docs/78479
Submitted by: Mikko Tyolajarvi
MFC after: 2 weeks

18 years agoImport ACPI Dock Station support. Note that this is still very young.
iwasaki [Sat, 15 Apr 2006 12:31:34 +0000 (12:31 +0000)]
Import ACPI Dock Station support.  Note that this is still very young.
Additional detach implementaions (or maybe improvement) for other
deivce drivers is required.

Reviewed by: njl, imp
MFC after: 1 week

18 years agoo Fix prototype for SYSCALL_MODULE(9): the third argument
maxim [Sat, 15 Apr 2006 12:04:18 +0000 (12:04 +0000)]
o Fix prototype for SYSCALL_MODULE(9): the third argument
is a pointer to struct sysent.  Correct its description.

PR: docs/84790
Submitted by: Dirk Gouders
MFC after: 2 weeks

18 years agoo Fix printf(9) formatting: do not use hardcode "0x" and "#" flags
maxim [Sat, 15 Apr 2006 11:41:40 +0000 (11:41 +0000)]
o Fix printf(9) formatting: do not use hardcode "0x" and "#" flags
simultaneously.  Remove "#' flag to match a style of the rest of
file.

PR: kern/85477
Submitted by: Oliver Fromme

18 years agoo s/PsuedoRAID/PseudoRAID/.
maxim [Sat, 15 Apr 2006 10:27:41 +0000 (10:27 +0000)]
o s/PsuedoRAID/PseudoRAID/.

PR: kern/94306
Submitted by: Esa Karkkainen

18 years agoo Be pedantic and do fts_close() when done.
maxim [Sat, 15 Apr 2006 09:26:23 +0000 (09:26 +0000)]
o Be pedantic and do fts_close() when done.

PR: bin/95292
Submitted by: Charles Hardin
Obtained from: NetBSD via OpenBSD, PR

18 years agoFix the interrupt handler to do the mandatory PCI flush before looking at
scottl [Sat, 15 Apr 2006 08:13:06 +0000 (08:13 +0000)]
Fix the interrupt handler to do the mandatory PCI flush before looking at
DMA memory.  The could contribute towards missed link state changes under
heavy bus load.

18 years agoMFen:
delphij [Sat, 15 Apr 2006 05:32:12 +0000 (05:32 +0000)]
MFen:
hardware/common/dev.sgml 1.296 -> 1.299
relnotes/common/new.sgml 1.910 -> 1.911

Obtained from: The FreeBSD Simplified Chinese Project

18 years agoMirror IPv4 pcb locking into in6_setsockaddr() and in6_setpeeraddr():
rwatson [Sat, 15 Apr 2006 05:24:23 +0000 (05:24 +0000)]
Mirror IPv4 pcb locking into in6_setsockaddr() and in6_setpeeraddr():
acquire inpcb lock when reading inpcb port+address in order to prevent
races with other threads that may be changing them.

MFC after: 3 months

18 years agoCopy the new CLOCK_ defines from sys/time.h for SUSv3 compatibility.
jb [Sat, 15 Apr 2006 03:08:55 +0000 (03:08 +0000)]
Copy the new CLOCK_ defines from sys/time.h for SUSv3 compatibility.

Approved by: rwatson

18 years agoFix style nit noticed by bde.
imp [Sat, 15 Apr 2006 02:51:53 +0000 (02:51 +0000)]
Fix style nit noticed by bde.

18 years agoMark the thread pointer used during an adaptive spin volatile so that the
jhb [Fri, 14 Apr 2006 19:51:50 +0000 (19:51 +0000)]
Mark the thread pointer used during an adaptive spin volatile so that the
compiler doesn't decide to cache td_state.  Cachine the state would cause
the spinning thread to not notice when the owning thread stopped executing
(if it was preempted for example) which could result in livelock.

18 years agoList pwritev in the NAME section.
ru [Fri, 14 Apr 2006 19:34:07 +0000 (19:34 +0000)]
List pwritev in the NAME section.

18 years agoFix copy-and-pasto in comment.
jkim [Fri, 14 Apr 2006 18:34:57 +0000 (18:34 +0000)]
Fix copy-and-pasto in comment.

18 years agoFix minimal installation check. 'X' was not displayed when it was selected.
jkim [Fri, 14 Apr 2006 18:18:43 +0000 (18:18 +0000)]
Fix minimal installation check.  'X' was not displayed when it was selected.

Reported by: Ian J. Hart <ianjhart at ntlworld dot com>
Pieter de Goeje <pieter at degoeje dot nl>

18 years ago1) Fix hang at the end of line buffer (PR 95715)
ache [Fri, 14 Apr 2006 17:32:27 +0000 (17:32 +0000)]
1) Fix hang at the end of line buffer (PR 95715)
2) Localize

PR:             95715
Submitted by:   Li-Lun Wang <llwang@infor.org> (slightly edited by me)

18 years agoWhitespace fix
matteo [Fri, 14 Apr 2006 17:26:04 +0000 (17:26 +0000)]
Whitespace fix

Pointed out by: Nate Lawson

18 years agoDont poll for ATA_IDLE on a detached channel in suspend.
sos [Fri, 14 Apr 2006 16:25:42 +0000 (16:25 +0000)]
Dont poll for ATA_IDLE on a detached channel in suspend.

18 years agoReduce the Linux ioctl range to what is needed. I didn't know what
ambrisko [Fri, 14 Apr 2006 16:13:28 +0000 (16:13 +0000)]
Reduce the Linux ioctl range to what is needed.  I didn't know what
I was doing when I first set the range up.

18 years agoDrop the kqueue global mutex as soon as we are finished with it rather
jhb [Fri, 14 Apr 2006 14:27:28 +0000 (14:27 +0000)]
Drop the kqueue global mutex as soon as we are finished with it rather
than keeping it locked until we exit the function to optimize the case
where the lock would be dropped and later reacquired.  The optimization
was broken when kevent's were moved from UFS to VFS and the knote list
lock for a vnode kevent became the lockmgr vnode lock.  If one tried
to use a kqueue that contained events for a kqueue fd followed by a vnode,
then the kq global lock would end up being held when the vnode lock was
acquired which could result in sleeping with a mutex held (and subsequent
panics) if the vnode lock was contested.

Reviewed by: jmg
Tested by: ps (on 6.x)
MFC after: 3 days

18 years agoOutput something reasonable for regular and expanded here-documents.
schweikh [Fri, 14 Apr 2006 13:59:03 +0000 (13:59 +0000)]
Output something reasonable for regular and expanded here-documents.
I would have chosen the EOF markers, but they are no longer available
AFAICS, so output "<<HERE" and "<<XHERE" instead.
(NOTE: These changes only affect DEBUG output.)

18 years agoFix markup and some typos.
ru [Fri, 14 Apr 2006 09:20:25 +0000 (09:20 +0000)]
Fix markup and some typos.

18 years agoAdd device ID for Intel Pro/100 VE Network Connection card
matteo [Fri, 14 Apr 2006 07:49:52 +0000 (07:49 +0000)]
Add device ID for Intel Pro/100 VE Network Connection card

PR: kern/95729
Submitted by: Nicky Bulthuis
MFC after: 1 day

18 years agoo Yacc's lineno variable holds a total number of parsed lines.
maxim [Fri, 14 Apr 2006 07:20:42 +0000 (07:20 +0000)]
o Yacc's lineno variable holds a total number of parsed lines.
Reset it to 1 for each devd config file so if the parser finds
a syntax error devd(8) will report a correct line number.

Submitted by: Niki Denev
MFC after: 2 weeks

18 years agoClear TDF_SINTR in sleepq_resume_thread, also sleepq_catch_signal does
davidxu [Thu, 13 Apr 2006 23:29:25 +0000 (23:29 +0000)]
Clear TDF_SINTR in sleepq_resume_thread, also sleepq_catch_signal does
not need to clear it now, this should fix panic when msleep is recursivly
called. Patch is slightly adjusted after review.

Reviewed by: jhb
Tested by: Csaba Henk, csaba-ml at creo.hu
MFC after: 3 days

18 years agoMake CCD be able to read and write Linux software raids.
cracauer [Thu, 13 Apr 2006 20:35:31 +0000 (20:35 +0000)]
Make CCD be able to read and write Linux software raids.

Supported for raid-0 with <n> disks, raid-1 with 2 disks.

Manpages have examples, warnings etc.

Test scripts on
http://www.cons.org/cracauer/ccdconfig-linux/
Reviewed by: alfred

18 years agoRepair ext2fs writes.
cracauer [Thu, 13 Apr 2006 19:37:32 +0000 (19:37 +0000)]
Repair ext2fs writes.

Strong candidate for backport to 6.x.

When allocating new blocks, the search for block group beginnings
would fail with a segfault.  There was a side-effect read access with
an off-by-one errors.  The results were not used in the error case so
the code worked in the past.  But now the FreeBSD kernel has tighter
mappings and the word accessed is not mapped (for me).

The Linux kernel has rewritten most of the allocation strategy by now.
Also, the Linux kernel cleaned up the integration of these files and
it look feasable to wrap the original Linux files in wrapper that
provides their favorite arguments instead of dragging around our own
code.

18 years agoSpell synchronous with required silent 'h'.
brooks [Thu, 13 Apr 2006 18:34:14 +0000 (18:34 +0000)]
Spell synchronous with required silent 'h'.

Reported by: ru, ceri
Pointy hat: brooks

18 years agoAdd missing _ to $_punct.
brooks [Thu, 13 Apr 2006 18:27:49 +0000 (18:27 +0000)]
Add missing _ to $_punct.

Submitted by: Dmitry Pryanishnikov <dmitry at atlantis.dp.ua>

18 years agos/soley/solely
schweikh [Thu, 13 Apr 2006 18:19:44 +0000 (18:19 +0000)]
s/soley/solely

18 years agobce(4) does polling too.
brueffer [Thu, 13 Apr 2006 17:44:40 +0000 (17:44 +0000)]
bce(4) does polling too.

MFC after: 3 days

18 years agoAdd commented out define for TSC board's boot0.
imp [Thu, 13 Apr 2006 17:44:05 +0000 (17:44 +0000)]
Add commented out define for TSC board's boot0.

18 years agoMFp4:
imp [Thu, 13 Apr 2006 17:39:34 +0000 (17:39 +0000)]
MFp4:

For 32-bit SDRAM systems, enable D16 to D31 in the PIO controller.
Otherwise they read back as 0xffff.

Shave 8 bytes from the object size by using AT91C_BASE_PIOA directly
and by not assigning PIO_BSR to 0 in the DBGU init.  That's a nop in
two ways (everything defaults to peripheral A, and writing 0 changes
nothing).

18 years agoMerge from p4:
imp [Thu, 13 Apr 2006 17:34:23 +0000 (17:34 +0000)]
Merge from p4:

Many places used #define FOO ((unsigned int) 0x23) where a simpler
#define FOO 0x23u would have sufficed.  This practice is overly
verbose and has the disadvantage that you can't say

#if FOO == BAR
#endif

because the extra "unsigned int" tokens choke cpp's little brain.
Migrate to the latter style to allow use in preprocessor statements.
The two are the same semantically anyway in a C context (at least for
the uses they are put to presently, C gurus can explain to me how they
differ).

18 years agoNeed at91rm9200_lowlevel.h for SDRAM_BASE definition.
imp [Thu, 13 Apr 2006 17:31:02 +0000 (17:31 +0000)]
Need at91rm9200_lowlevel.h for SDRAM_BASE definition.

18 years agoTurn on ithread_destroy() and call it from intr_event_destroy() to tear
jhb [Thu, 13 Apr 2006 17:29:04 +0000 (17:29 +0000)]
Turn on ithread_destroy() and call it from intr_event_destroy() to tear
down an interrupt event's associated thread (if it has one).

18 years agoBring back arm-specific workaround from rev 1.15:
cognet [Thu, 13 Apr 2006 15:10:25 +0000 (15:10 +0000)]
Bring back arm-specific workaround from rev 1.15:
Do not use the IO-mapping to issue the reset on the 82546 on arm. For some
reason, it results in corrupted descriptors.

18 years agoWe need to adjust the bus handle if the resource is SYS_RES_MEMORY, NOT
cognet [Thu, 13 Apr 2006 15:07:59 +0000 (15:07 +0000)]
We need to adjust the bus handle if the resource is SYS_RES_MEMORY, NOT
SYS_RES_IOPORT.

18 years agoDisable/enable fiqs as well as irqs.
cognet [Thu, 13 Apr 2006 14:25:28 +0000 (14:25 +0000)]
Disable/enable fiqs as well as irqs.

18 years agoFix DEVICE_POLLING support (compile-only tested).
ru [Thu, 13 Apr 2006 14:12:26 +0000 (14:12 +0000)]
Fix DEVICE_POLLING support (compile-only tested).

18 years agoRecord the /lib/libgpib.so.1 move to /usr/lib.
ru [Thu, 13 Apr 2006 13:19:10 +0000 (13:19 +0000)]
Record the /lib/libgpib.so.1 move to /usr/lib.

18 years ago- Let make(1) track dependencies and regen netid map when needed.
ru [Thu, 13 Apr 2006 13:03:35 +0000 (13:03 +0000)]
- Let make(1) track dependencies and regen netid map when needed.
- Add missing dependency for netid.byname.

18 years agoAdd missing library dependencies.
ru [Thu, 13 Apr 2006 12:49:24 +0000 (12:49 +0000)]
Add missing library dependencies.

18 years agoFix Mother's/Father's day in most case.
flz [Thu, 13 Apr 2006 12:14:44 +0000 (12:14 +0000)]
Fix Mother's/Father's day in most case.

PR: misc/79415
Submitted by: thierry
Approved by: cperciva (mentor)
MFC after: 1 week

18 years agoRe-apply changes to make this compile with -Wundef:
ru [Thu, 13 Apr 2006 11:16:59 +0000 (11:16 +0000)]
Re-apply changes to make this compile with -Wundef:

/usr/src/sys/modules/drm/radeon/../../../dev/drm/radeon_state.c:2976:5: "BITS_PER_LONG" is not defined

(Previous revision has trashed my changes.)

18 years agoRegen.
ru [Thu, 13 Apr 2006 10:38:43 +0000 (10:38 +0000)]
Regen.

18 years agoImplications are auto-generated.
ru [Thu, 13 Apr 2006 10:37:29 +0000 (10:37 +0000)]
Implications are auto-generated.

18 years agoUse ps true power instead of tr/tail which aren't available early enough.
flz [Thu, 13 Apr 2006 08:30:43 +0000 (08:30 +0000)]
Use ps true power instead of tr/tail which aren't available early enough.

PR: conf/95654
Submitted by: Rong-En Fan <rafan@infor.org>
Noticed/Fix by: many people on freebsd-current@
Approved by: cperciva (mentor)

18 years agoApply style.Makefile(5) here.
ru [Thu, 13 Apr 2006 08:12:31 +0000 (08:12 +0000)]
Apply style.Makefile(5) here.

18 years agoInstall the shared library under /usr/lib; it's not used by anything
ru [Thu, 13 Apr 2006 08:08:21 +0000 (08:08 +0000)]
Install the shared library under /usr/lib; it's not used by anything
in /bin or /sbin.

Not objected to by: phk

18 years agoPlace shared libpcap library under /lib, currently needed by
ru [Thu, 13 Apr 2006 07:25:02 +0000 (07:25 +0000)]
Place shared libpcap library under /lib, currently needed by
/sbin/ipf and /sbin/pflogd.

OK'ed by: mlaier

18 years agoCommit the various network interface configutation updates I've been
brooks [Thu, 13 Apr 2006 06:50:46 +0000 (06:50 +0000)]
Commit the various network interface configutation updates I've been
working on.
  1) Make it possible to configure interfaces with certain characters in
     their names that aren't valid in shell variables.  Currently supported
     characters are ".-/+".  They are converted into '_' characters.
  2) Replace nearly all eval statements in network.subr with a new
     function get_if_var which substitues an interface name (after the
     translations above) for "IF" in a variable name.
  3) Fix list_net_interfaces() in the nodhcp case.
  4) Allow the administrator to specify if dhclient should be started
     when /etc/rc.d/netif configures the interface or only by devd.
     This can be set on both a per interface and system wide basis.

PR: conf/88974 [1,2], conf/92433 [1,2]

18 years agoAdd a note regarding bumping the freebsd*mc files for mergemaster's sake.
gshapiro [Thu, 13 Apr 2006 04:12:38 +0000 (04:12 +0000)]
Add a note regarding bumping the freebsd*mc files for mergemaster's sake.

18 years agoForced commit to bump the $FreeBSD$ version so mergemaster merges
gshapiro [Thu, 13 Apr 2006 04:08:22 +0000 (04:08 +0000)]
Forced commit to bump the $FreeBSD$ version so mergemaster merges
/etc/mail/freebsd.cf and /etc/mail/freebsd.submit.cf.

PR: conf/95634
Submitted by: Rostislav Krasny
Approved by: re (scottl)
Obtained from: keramida

18 years agoInclude opt_pmap.h for PMAP_SHPGPERPROC.
alc [Thu, 13 Apr 2006 03:31:48 +0000 (03:31 +0000)]
Include opt_pmap.h for PMAP_SHPGPERPROC.

PR: 94509

18 years agoUnexpand TAILQ_FIRST(foo) == NULL to TAILQ_EMPTY.
delphij [Thu, 13 Apr 2006 03:09:34 +0000 (03:09 +0000)]
Unexpand TAILQ_FIRST(foo) == NULL to TAILQ_EMPTY.

Ok'ed by: davidxu

18 years agoCommon register definition for AT91RM9200
imp [Wed, 12 Apr 2006 21:23:36 +0000 (21:23 +0000)]
Common register definition for AT91RM9200

18 years agoRecovery boot loader for the AT91 family of processors. Download it
imp [Wed, 12 Apr 2006 21:22:44 +0000 (21:22 +0000)]
Recovery boot loader for the AT91 family of processors.  Download it
via xmodem to the DBGU port when the AT91 comes up in recovery mode.
The recovery loader will then load your program via xmodem into SDRAM
at 1MB which can do its things.  It needs to be tweaked to the
specific board one is using, but it fits in < 1kB (all of Atmel's ARM
products have at least 8kb of SRAM that I can tell, so this should
work for them all).

Parts of this code were provided by Kwikbyte with copyright
specifically disclaimed.  I heavily modified it to act as a recovery
loader (before it was a bootstrap loader) and to optimize for size
(before I started the size was closer to 8k).

Bootstrap loaders for SPI and IIC to follow.

18 years agolibc_r is no longer provided, and on alpha and sparc64, libthr
ru [Wed, 12 Apr 2006 19:52:34 +0000 (19:52 +0000)]
libc_r is no longer provided, and on alpha and sparc64, libthr
is (sym)linked to libpthread.  Account for this change and
check for MK_LIBTHR instead of MK_LIBC_R where appropriate.

18 years agoFix a symlink.
ru [Wed, 12 Apr 2006 19:51:14 +0000 (19:51 +0000)]
Fix a symlink.

18 years agoInstall shared libpthread library into /lib; needed by some
ru [Wed, 12 Apr 2006 19:42:20 +0000 (19:42 +0000)]
Install shared libpthread library into /lib; needed by some
/sbin programs.

18 years agoDocument the bluetooth/, rc.d/, security/ and ssh/ etc subdirectories.
brueffer [Wed, 12 Apr 2006 18:41:25 +0000 (18:41 +0000)]
Document the bluetooth/, rc.d/, security/ and ssh/ etc subdirectories.

Prodded by: markus
MFC after: 3 days

18 years agoFix the PHY support for the 5780. I mis-merged this from my sources.
ps [Wed, 12 Apr 2006 18:34:00 +0000 (18:34 +0000)]
Fix the PHY support for the 5780.  I mis-merged this from my sources.

18 years agoadapt PacketGetAdapterNames to 0.4.8 calling convention
sam [Wed, 12 Apr 2006 17:12:04 +0000 (17:12 +0000)]
adapt PacketGetAdapterNames to 0.4.8 calling convention

PR: bin/94735
Submitted by: Scot Hetzel <swhetzel@gmail.com>
Requested by: scottl

18 years agoPass BIO_GETATTR requests down.
pjd [Wed, 12 Apr 2006 12:18:44 +0000 (12:18 +0000)]
Pass BIO_GETATTR requests down.

MFC after: 1 week

18 years agoRemove unused prototypes.
pjd [Wed, 12 Apr 2006 12:17:29 +0000 (12:17 +0000)]
Remove unused prototypes.

18 years agoOn padlock initialization, allocate memory with M_WAITOK.
pjd [Wed, 12 Apr 2006 12:13:34 +0000 (12:13 +0000)]
On padlock initialization, allocate memory with M_WAITOK.

18 years agoCorrect an obvious typo.
pjd [Wed, 12 Apr 2006 12:12:44 +0000 (12:12 +0000)]
Correct an obvious typo.

MFC after: 1 week

18 years agoRemove an inappropriate comment which is put into generated
ume [Wed, 12 Apr 2006 12:01:53 +0000 (12:01 +0000)]
Remove an inappropriate comment which is put into generated
host.conf.  Someone may want to edit it later.

Requested by: Rostislav Krasny <rosti.bsd__at__gmail.com>

18 years agoUse ANSI C function protypes and declarations for if_arcsubr.
rwatson [Wed, 12 Apr 2006 07:44:31 +0000 (07:44 +0000)]
Use ANSI C function protypes and declarations for if_arcsubr.

MFC after: 1 month

18 years agoRetire pmap_track_modified(). We no longer need it because we do not
alc [Wed, 12 Apr 2006 04:22:52 +0000 (04:22 +0000)]
Retire pmap_track_modified().  We no longer need it because we do not
create managed mappings within the clean submap.  To prevent regressions,
add assertions blocking the creation of managed mappings within the clean
submap.

Reviewed by: tegge

18 years agoAssert the inpcb lock in udp6_output(), as we dereference various
rwatson [Wed, 12 Apr 2006 03:34:22 +0000 (03:34 +0000)]
Assert the inpcb lock in udp6_output(), as we dereference various
fields.

MFC after: 3 months

18 years agoAdd comment to udp6_input() that locking is missing from multicast
rwatson [Wed, 12 Apr 2006 03:32:54 +0000 (03:32 +0000)]
Add comment to udp6_input() that locking is missing from multicast
UDPv6 delivery.

Lock the inpcb of the UDP connection being delivered to before
processing IPSEC policy and other delivery activities.

MFC after: 3 months

18 years agoAdd udbinfo locking in udp6_input() to protect lookups of the inpcb
rwatson [Wed, 12 Apr 2006 03:23:56 +0000 (03:23 +0000)]
Add udbinfo locking in udp6_input() to protect lookups of the inpcb
lists during UDPv6 receipt.

MFC after: 3 months

18 years agoDon't use spl around call to in_pcballoc() in IPv6 raw socket support;
rwatson [Wed, 12 Apr 2006 03:07:22 +0000 (03:07 +0000)]
Don't use spl around call to in_pcballoc() in IPv6 raw socket support;
all necessary synchronization appears present.

MFC after: 3 months

18 years agoRemove one remaining use of spl in the IPv6 fragmentation code, as
rwatson [Wed, 12 Apr 2006 03:06:20 +0000 (03:06 +0000)]
Remove one remaining use of spl in the IPv6 fragmentation code, as
this code appears properly locked.

MFC after: 3 months

18 years agoAdd missing locking to udp6_getcred(), remove spl use.
rwatson [Wed, 12 Apr 2006 03:03:47 +0000 (03:03 +0000)]
Add missing locking to udp6_getcred(), remove spl use.

MFC after: 3 months

18 years agoRemove spl use from IPv6 inpcb code.
rwatson [Wed, 12 Apr 2006 02:52:14 +0000 (02:52 +0000)]
Remove spl use from IPv6 inpcb code.

In various inpcb methods for IPv6 sockets, don't check of so_pcb is NULL,
assert it isn't.

MFC after: 3 months

18 years agoMake this compile without INVARIANTS.
cognet [Tue, 11 Apr 2006 23:15:47 +0000 (23:15 +0000)]
Make this compile without INVARIANTS.

18 years agouse pwrite to always write at the begining of the file.. If multiple calls
jmg [Tue, 11 Apr 2006 23:10:02 +0000 (23:10 +0000)]
use pwrite to always write at the begining of the file..  If multiple calls
to pidfile_write happen, the pidfile will have nul characters prepended
due to the cached file descriptor offset...

Reviewed by: scottl
MFC after: 3 days

18 years agoIf we end up with a CTIO completing with CAM_REQUEUE_REQ,
mjacob [Tue, 11 Apr 2006 21:36:43 +0000 (21:36 +0000)]
If we end up with a CTIO completing with CAM_REQUEUE_REQ,
be obliging and just redo the request instead of squawking
and dying.

18 years agoMerge in timeout into A20-enable routine from cdboot/boot1.
sobomax [Tue, 11 Apr 2006 20:53:49 +0000 (20:53 +0000)]
Merge in timeout into A20-enable routine from cdboot/boot1.

MFC after: 1 day

18 years agoDrop the gateA20() function in the loader as it is unused. All the other
jhb [Tue, 11 Apr 2006 20:11:30 +0000 (20:11 +0000)]
Drop the gateA20() function in the loader as it is unused.  All the other
boot loaders that load the loader already handle A20.  In fact, they are
required to do so in order to setup the environment that btxldr expects.

18 years agoMinor whitespace tweak.
jhb [Tue, 11 Apr 2006 20:09:42 +0000 (20:09 +0000)]
Minor whitespace tweak.

18 years ago- Move _Qp_* and __dtoul from FBSDprivate to the list of FBSD_1.0 symbols
marius [Tue, 11 Apr 2006 19:43:36 +0000 (19:43 +0000)]
- Move _Qp_* and __dtoul from FBSDprivate to the list of FBSD_1.0 symbols
  as well as add __sparc_utrap_install to FBSD_1.0; these are required by
  the SCD libc 64 psABI and thus meant to be officially exported symbols.
- Remove the __fpu_* entries as well as the __sigtramp entry altogether as
  these are internal to the libc FPU emulation and the signal trampoline
  initialization in sigaction(2) respectively and thus don't need to be
  externally visible.
- Add __sparc_utrap_setup to the list of FBSDprivate symbols as it's used
  in src/lib/csu/sparc64/crt1.c to initialize the libc FPU emulation (I
  think alternatively src/lib/csu/sparc64/crt1.c could be changed to use
  __sparc_utrap_install instead, at the expense of increasing the size of
  executables a bit).
- Add an entry for the vfork symbol to the FBSD_1 list and entries for it's
  associated symbols generated by the RSYSCALL() macro to the FBSDprivate
  list. There's some magic in place that automatically generates code for
  vfork() if there's no explicit MD code for it so it might make sense to
  move these symbols from the MD symbol map files to a MI one.

The last two changes make the libc symbol versioning useable on sparc64.

Ok'ed by: deischen

18 years agoBe sure to wakeup the crypto thread when new request was queued.
pjd [Tue, 11 Apr 2006 18:01:04 +0000 (18:01 +0000)]
Be sure to wakeup the crypto thread when new request was queued.
This should fix a hang when starting cryptokeytest (and more).

MFC after: 1 month

18 years agoTweak comment.
jhb [Tue, 11 Apr 2006 17:36:08 +0000 (17:36 +0000)]
Tweak comment.

18 years agoUse the proper condition to determine that we matched an filename.
jhb [Tue, 11 Apr 2006 17:26:54 +0000 (17:26 +0000)]
Use the proper condition to determine that we matched an filename.
Otherwise, we could match on a filename that had the wrong last character
(such as /boot/loaded instead of /boot/loader).

PR: kern/95625
Submitted by: Oliver Fromme <olli@secnetix.de>
MFC after: 1 month

18 years agoA large set of changes:
mjacob [Tue, 11 Apr 2006 16:47:30 +0000 (16:47 +0000)]
A large set of changes:

+ Add boatloads of KASSERTs and *really* check out more locking
issues (to catch recursions when we actually go to real locking
in CAM soon). The KASSERTs also caught lots of other issues like
using commands that were put back on free lists, etc.

+ Target mode: role setting is derived directly from port capabilities.
There is no need to set a role any more. Some target mode resources
are allocated early on (ELS), but target command buffer allocation
is deferred until the first lun enable.

+ Fix some breakages I introduced with target mode in that some commands
are *repeating* commands. That is, the reply shows up but the command
isn't really done (we don't free it). We still need to take it off the
pending list because when we resubmit it, bad things then  happen.

+ Fix more of the way that timed out commands and bus reset is done. The
actual TMF response code was being ignored.

+ For SPI, honor BIOS settings. This doesn't quite fix the problems we've
seen where we can't seem to (re)negotiate U320 on all drives but avoids
it instead by letting us honor the BIOS settings. I'm sure this is not
quite right and will have to change again soon.