]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoInitialize bp->b_offset before calling VOP_[SPEC]STRATEGY()
phk [Sat, 18 Oct 2003 11:08:04 +0000 (11:08 +0000)]
Initialize bp->b_offset before calling VOP_[SPEC]STRATEGY()

20 years agoInitialize b_offset before calling VOP_STRATEGY/VOP_SPECSTRATEGY.
phk [Sat, 18 Oct 2003 11:06:15 +0000 (11:06 +0000)]
Initialize b_offset before calling VOP_STRATEGY/VOP_SPECSTRATEGY.

Remove various comments of KASSERTS and comments about B_PHYS which
does not apply anymore.

20 years agoInitialize bp->b_offset and remove comment about B_PHYS.
phk [Sat, 18 Oct 2003 11:02:24 +0000 (11:02 +0000)]
Initialize bp->b_offset and remove comment about B_PHYS.

20 years agoThere is no need to muck about with the B_PHYS flag here. We never even
phk [Sat, 18 Oct 2003 11:01:11 +0000 (11:01 +0000)]
There is no need to muck about with the B_PHYS flag here.  We never even
get close to DEV_STRATEGY() which is the only place it is relevant.

20 years agoANSIfy, WARNSify, CONSTify. Bit of style(9)-ify.
markm [Sat, 18 Oct 2003 10:04:16 +0000 (10:04 +0000)]
ANSIfy, WARNSify, CONSTify. Bit of style(9)-ify.

20 years agoRemove a GCC specifig CFLAG. We should be using WARNS=? for this.
markm [Sat, 18 Oct 2003 09:58:39 +0000 (09:58 +0000)]
Remove a GCC specifig CFLAG. We should be using WARNS=? for this.
WARNS=? is not added here at this point, because I've not tested
it on enough platforms, and I don't want to break builds.

20 years agoWe do not get B_PHYS buffers here anymore. /dev/drum is long gone.
phk [Sat, 18 Oct 2003 09:33:13 +0000 (09:33 +0000)]
We do not get B_PHYS buffers here anymore.  /dev/drum is long gone.

20 years agoConvert some if(bla) panic("foo") to KASSERTS to improve grep-ability.
phk [Sat, 18 Oct 2003 09:32:39 +0000 (09:32 +0000)]
Convert some if(bla) panic("foo") to KASSERTS to improve grep-ability.

20 years agoMark as __unused some arguments that are, erm, unused.
markm [Sat, 18 Oct 2003 09:16:01 +0000 (09:16 +0000)]
Mark as __unused some arguments that are, erm, unused.

20 years agoThe size and contents of the DEV_STRATEGY() macro has progressed to
phk [Sat, 18 Oct 2003 09:03:15 +0000 (09:03 +0000)]
The size and contents of the DEV_STRATEGY() macro has progressed to
the point where it being a macro is no longer sensible, and it will
only be more so in days to come.

BIO_STRATEGY() is now only used from DEV_STRATEGY() and should not
be used directly anymore.

Put the contents of both in the new function dev_strategy() and
make DEV_STRATEGY() call that function.

In addition, this allows us to make the rather magic bufdonebio()
helper function static.

This alse saves hunderedandsome bytes of code in a typical kernel.

20 years agoRearrange the deck-chairs while we wait for scottl to GEOMify RF.
phk [Sat, 18 Oct 2003 08:58:35 +0000 (08:58 +0000)]
Rearrange the deck-chairs while we wait for scottl to GEOMify RF.

20 years agoAdd SBP-II target mode driver.
simokawa [Sat, 18 Oct 2003 05:41:31 +0000 (05:41 +0000)]
Add SBP-II target mode driver.

Though this is still incomplete and has some missing features such as
exclusive login and event notification, it may be enough for someone
who wants to play with it.

This driver is supposed to work with firewire(4), targ(4) of CAM(4)
and scsi_target(8) which can be found in /usr/share/example/scsi_target.
This driver doesn't require sbp(4) which implements initiator mode.

Sample configuration:

Kernel: (you can use modules as well)
device firewire
device scbus
device targ
device sbp_targ

After reboot:
# mdconfig -a -t malloc -s 10m
md0
# scsi_target 0:0:0 /dev/md0
(Assuming sbp_targ0 on scbus0)

You should find the 10MB HDD on FreeBSD/MacOS X/WinXP or whatever connected
to the target using FireWire.

Manpage is not finished yet.

20 years ago- Call sbp_reset_start() for mgm timeout.
simokawa [Sat, 18 Oct 2003 04:59:47 +0000 (04:59 +0000)]
- Call sbp_reset_start() for mgm timeout.
- Change type of target->luns to allocate an array of LUNs dynamically.
This allows targets to change their number of LUNs after each bus reset.
- Serialize ORB POINTER command for each LUN.
- Improve debug messages.

20 years ago- Enable 16byte commands.
simokawa [Sat, 18 Oct 2003 04:54:08 +0000 (04:54 +0000)]
- Enable 16byte commands.
- Fix printf warnings on 64bit architectures.
- Accept 'k', 'm' and etc. for -s option.

Reviewed by: njl

20 years agoWrap db_active check in #ifdef DDB, as db_active is not defined ifndef
rwatson [Sat, 18 Oct 2003 02:23:57 +0000 (02:23 +0000)]
Wrap db_active check in #ifdef DDB, as db_active is not defined ifndef
DDB.

20 years agoAdd a new cn_flags fields to struct consdev, the low-level console
rwatson [Sat, 18 Oct 2003 02:13:39 +0000 (02:13 +0000)]
Add a new cn_flags fields to struct consdev, the low-level console
definition structure.  Define one flag, CN_FLAG_NODEBUG, which
indicates the console driver cannot be used in the context of the
debugger.  This may be used, for example, if the console device
interacts with kernel services that cannot be used from the
debugger context, such as the network stack.  These drivers are
skipped over for calls to cn_checkc() and cn_putc(), and the
calling function simply moves on to the next available console.

20 years agofix station mode breakage (repeat after me: "test every change, no
sam [Fri, 17 Oct 2003 23:59:11 +0000 (23:59 +0000)]
fix station mode breakage (repeat after me: "test every change, no
matter how small...")

20 years agoadd statistics for all failures and/or abnormal events; still need
sam [Fri, 17 Oct 2003 23:15:30 +0000 (23:15 +0000)]
add statistics for all failures and/or abnormal events; still need
to add per-node statistics

20 years agofixup debug msgs
sam [Fri, 17 Oct 2003 22:15:26 +0000 (22:15 +0000)]
fixup debug msgs

20 years agoinclude FH/DS parameters element in probe response frames
sam [Fri, 17 Oct 2003 22:09:20 +0000 (22:09 +0000)]
include FH/DS parameters element in probe response frames

20 years agoo consolidate rx filter calculations in one place
sam [Fri, 17 Oct 2003 21:58:39 +0000 (21:58 +0000)]
o consolidate rx filter calculations in one place
o enable beacon reception when operating in adhoc mode
  so the 802.11 layer can use them to create nodes for peers

20 years agocorrect comment
sam [Fri, 17 Oct 2003 21:56:42 +0000 (21:56 +0000)]
correct comment

20 years agoindicate device receives all management frames
sam [Fri, 17 Oct 2003 21:55:53 +0000 (21:55 +0000)]
indicate device receives all management frames

20 years agoo add capability to indicate if device receives all management frames
sam [Fri, 17 Oct 2003 21:54:59 +0000 (21:54 +0000)]
o add capability to indicate if device receives all management frames
o use recv mgmt capability to decide if outbound frames should be
  discarded if no node table entry is present

20 years agoFix a couple of bugs with AIF handling:
scottl [Fri, 17 Oct 2003 21:44:06 +0000 (21:44 +0000)]
Fix a couple of bugs with AIF handling:
 - Correct the logic for the AIF array index pointers so that correct slot is
   always looked at.
 - Copy the full FIB payload size when copying AIF's, not just the first 64
   bytes.

Thanks to Mirapoint, Inc, for pointing these problems out and offering a
solution.

20 years agoparameterize locking to improve portability and possible
sam [Fri, 17 Oct 2003 21:41:52 +0000 (21:41 +0000)]
parameterize locking to improve portability and possible
change to different locking strategies

20 years ago - Remove the correct thread from the run queue in setrunqueue(). This
jeff [Fri, 17 Oct 2003 20:53:04 +0000 (20:53 +0000)]
 - Remove the correct thread from the run queue in setrunqueue().  This
   fixes ULE + KSE.

20 years agouse ND_IFINFO().
ume [Fri, 17 Oct 2003 20:20:14 +0000 (20:20 +0000)]
use ND_IFINFO().

20 years agoMake the regression test run also with obj directories.
phk [Fri, 17 Oct 2003 19:52:07 +0000 (19:52 +0000)]
Make the regression test run also with obj directories.

20 years agonuke duplicate function and unused function.
ume [Fri, 17 Oct 2003 17:50:09 +0000 (17:50 +0000)]
nuke duplicate function and unused function.

Obtained from: KAME

20 years agorevert wrongly dropped null check by previous commit.
ume [Fri, 17 Oct 2003 17:34:31 +0000 (17:34 +0000)]
revert wrongly dropped null check by previous commit.

20 years ago - fix to UID test description, non-zero -> zero
kensmith [Fri, 17 Oct 2003 17:03:38 +0000 (17:03 +0000)]
- fix to UID test description, non-zero -> zero

PR: docs/57799
Reviewed by: des
Approved by: blackend (mentor)

20 years agoModified release note:
hrs [Fri, 17 Oct 2003 16:47:42 +0000 (16:47 +0000)]
Modified release note:
- Use multicast(4) and pim(4) hyperlink.
- Fix a wrong description in rfcomm_sppd(1) and
  rfcomm_pppd(8) item.

20 years agoForced commit for attribution. Last commit was
peter [Fri, 17 Oct 2003 16:31:00 +0000 (16:31 +0000)]
Forced commit for attribution.  Last commit was

Submitted by: deischen

20 years agoUpdate context code for my last ABI breakage of mcontext. I'm worried
peter [Fri, 17 Oct 2003 16:30:09 +0000 (16:30 +0000)]
Update context code for my last ABI breakage of mcontext.  I'm worried
about the fpu code here.  It should be using fxsave/fxrstor instead of
saving/restoring the control word.  The SSE registers are used a lot in
gcc generated code on amd64.  I'm not sure how this all fits together
though.

20 years agodon't print ndp -i result if the interface is not
ume [Fri, 17 Oct 2003 16:17:47 +0000 (16:17 +0000)]
don't print ndp -i result if the interface is not
initialized

Obtained from: KAME

20 years ago- add dom_if{attach,detach} framework.
ume [Fri, 17 Oct 2003 15:46:31 +0000 (15:46 +0000)]
- add dom_if{attach,detach} framework.
- transition to use ifp->if_afdata.

Obtained from: KAME

20 years agoExplicitly specify an alignment for abitag. Without it, gcc specifies a
peter [Fri, 17 Oct 2003 15:43:13 +0000 (15:43 +0000)]
Explicitly specify an alignment for abitag.  Without it, gcc specifies a
section alignnment of 16 bytes for amd64 and this breaks file(1).
Before:
./cp: ELF 64-bit LSB executable, AMD x86-64, version 1 (FreeBSD), for \
  FreeBSD 127.7.9, statically linked, stripped
after:    ^^^^^^^
./ls: ELF 64-bit LSB executable, AMD x86-64, version 1 (FreeBSD), for \
  FreeBSD 5.0.1, dynamically linked (uses shared libs), stripped

The reason for this is that the NOTE sections are not contiguous
internally.  If the note section has an alignment of 16, then anything
that looks for the data is supposed to round up the payload start to
the next multiple of the alignment.  But FreeBSD/amd64 broke because the
structure is declared as a single structure, not a (header,payload) group,
where the payload had an explicit alignment roundup.

The alternative is to change things like file(1) to ignore the ELF payload
alignment rules for the PT_NOTE section only for FreeBSD.

20 years agoAdd multicast(4) and pim(4) manual pages and hook up to the build.
bmah [Fri, 17 Oct 2003 15:12:01 +0000 (15:12 +0000)]
Add multicast(4) and pim(4) manual pages and hook up to the build.

Submitted by: Pavlin Radoslavov <pavlin@icir.org>
Reviewed by: hsu, bmah
MFC after: 2 weeks

20 years agoWhen expunging unlinked files from a snapshot, skip over holes in the
mckusick [Fri, 17 Oct 2003 13:57:58 +0000 (13:57 +0000)]
When expunging unlinked files from a snapshot, skip over holes in the
file rather than panicing with "indiracct: botched params".

Submitted by: Mark Santcroos <marks@ripe.net>

20 years agoSimplify count_dev()
phk [Fri, 17 Oct 2003 11:56:48 +0000 (11:56 +0000)]
Simplify count_dev()

20 years ago- NI_WITHSCOPEID was deprecated
ume [Fri, 17 Oct 2003 11:46:40 +0000 (11:46 +0000)]
- NI_WITHSCOPEID was deprecated
- use strlcpy() and snprintf()
- use err()
- style

Obtained from: KAME

20 years ago- style
ume [Fri, 17 Oct 2003 11:43:44 +0000 (11:43 +0000)]
- style
- rename variable
- use strlcpy
- const'fy

Obtained from: KAME

20 years agoUses wait4() instead of wait3() when waiting for a child process to exit.
seanc [Fri, 17 Oct 2003 07:53:30 +0000 (07:53 +0000)]
Uses wait4() instead of wait3() when waiting for a child process to exit.

Reviewed by: yar

20 years ago1. Catch up with the move of the declaration of panic() from <sys/param.h>
bde [Fri, 17 Oct 2003 06:25:37 +0000 (06:25 +0000)]
1. Catch up with the move of the declaration of panic() from <sys/param.h>
   back to <sys/systm.h>.
2. Remove hard sentence breaks.

Reminded by: Stefan Farfeleder <stefan@fafoe.narf.at> (1)

20 years ago - Synchronize access to a vm page's valid field using the containing
alc [Fri, 17 Oct 2003 05:07:17 +0000 (05:07 +0000)]
 - Synchronize access to a vm page's valid field using the containing
   vm object's lock.
 - Release the vm object and vm page queues locks around vput().

20 years agoHalt the cpu on amd64 as well. For some strange reason, this makes
peter [Fri, 17 Oct 2003 03:49:03 +0000 (03:49 +0000)]
Halt the cpu on amd64 as well.  For some strange reason, this makes
a fair bit of difference to the power consumption and lets my cpu cool
down enough for the temperature sensitive fan controller to completely
stop the cpu fan at times.

20 years agoImplement cpu_idle() on ia64. We put the processor in a lightweight
marcel [Fri, 17 Oct 2003 02:24:59 +0000 (02:24 +0000)]
Implement cpu_idle() on ia64. We put the processor in a lightweight
halt state that minimizes power consumption while still preserving
cache and TLB coherency. Halting the processor is not conditional at
this time. Tested with UP and SMP kernels.

20 years ago - The kse may be null in sched_pctcpu().
jeff [Thu, 16 Oct 2003 21:13:14 +0000 (21:13 +0000)]
 - The kse may be null in sched_pctcpu().

Reported by: kris

20 years ago - Only kse_reassign() in the !running case.
jeff [Thu, 16 Oct 2003 20:32:57 +0000 (20:32 +0000)]
 - Only kse_reassign() in the !running case.

Reported by: kris

20 years ago - Call sched_add() with the correct argument on SMP.
jeff [Thu, 16 Oct 2003 20:06:19 +0000 (20:06 +0000)]
 - Call sched_add() with the correct argument on SMP.

Reported by: Valentin Chopov <valentin@valcho.net>

20 years agofix horribly botched MFp4 merge
sam [Thu, 16 Oct 2003 19:55:28 +0000 (19:55 +0000)]
fix horribly botched MFp4 merge

20 years agopfil hooks can modify packet contents so check if the destination
sam [Thu, 16 Oct 2003 18:57:45 +0000 (18:57 +0000)]
pfil hooks can modify packet contents so check if the destination
address has been changed when PFIL_HOOKS is enabled and, if it has,
arrange for the proper action by ip*_forward.

Submitted by: Pyun YongHyeon
Supported by: FreeBSD Foundation

20 years agoNew release note:
hrs [Thu, 16 Oct 2003 17:16:24 +0000 (17:16 +0000)]
New release note:
Bluetooth protocol stack update.

Based on report by: emax

Modified release note:
s|challenge/|challenge/response| in SA-03:15 item.

20 years agopfil hooks can modify packet contents so check if the destination
sam [Thu, 16 Oct 2003 16:25:25 +0000 (16:25 +0000)]
pfil hooks can modify packet contents so check if the destination
address has been changed when PFIL_HOOKS is enabled and, if it has,
arrange for the proper action by ip*_forward.

Supported by: FreeBSD Foundation
Submitted by: Pyun YongHyeon

20 years agoDrop dummynet lock when calling back into the network stack to deliver
sam [Thu, 16 Oct 2003 16:21:25 +0000 (16:21 +0000)]
Drop dummynet lock when calling back into the network stack to deliver
packets.  This eliminates a LOR with Giant that caused outbound pipes
to fail.

Supported by: FreeBSD Foundation

20 years agoCorrect handling of cloning loop avoidance: rtalloc1 may return a null
sam [Thu, 16 Oct 2003 16:17:17 +0000 (16:17 +0000)]
Correct handling of cloning loop avoidance: rtalloc1 may return a null
pointer in which case we should not do the unlock.

Supported by: FreeBSD Foundatin

20 years agoo correct handling of a frame that has too many segments to fit in the
sam [Thu, 16 Oct 2003 16:13:11 +0000 (16:13 +0000)]
o correct handling of a frame that has too many segments to fit in the
  tx descriptor array
o while here fix a whitespace nit

Obtained from: NetBSD

20 years agoA comment about spam and cockroaches.
fanf [Thu, 16 Oct 2003 15:25:19 +0000 (15:25 +0000)]
A comment about spam and cockroaches.

Suggested by: dwmalone

20 years agoAF_LINK sockaddr has to be attached to ifp->if_addrlist until the
ume [Thu, 16 Oct 2003 13:38:29 +0000 (13:38 +0000)]
AF_LINK sockaddr has to be attached to ifp->if_addrlist until the
end, as many of the code assumes that TAILQ_FIRST(ifp->if_addrlist)
is non-null.

Submitted by: itojun

20 years agoChanges for new SMP-safe kobj method dispatch algorithm.
dfr [Thu, 16 Oct 2003 13:29:26 +0000 (13:29 +0000)]
Changes for new SMP-safe kobj method dispatch algorithm.

20 years agoDon't forget to load %es with the kernel data segment selector in
bde [Thu, 16 Oct 2003 10:44:24 +0000 (10:44 +0000)]
Don't forget to load %es with the kernel data segment selector in
Xcpustop().  %es is used in at least the call to savectx() when savectx()
calls bcopy(), so not loading it was fatal if a stop IPI interrupts
user mode.

This reduces bugs starting and stopping CPUs for debuggers.  CPUs are
stopped mainly in kdb_trap() and cpu_reset().  At reset time there is
a good chance that all the CPUs are in the kernel, so the bug was
probably harmless then.

20 years ago - Fix a minor problem with my last commit, we don't want to return from
jeff [Thu, 16 Oct 2003 10:04:54 +0000 (10:04 +0000)]
 - Fix a minor problem with my last commit, we don't want to return from
   sched_switch if the thread is running, we want to fall through and pick
   a new thread because we have been preempted.

20 years ago - mtx_ownedby() was unpopular and is no longer needed. Remove it.
jeff [Thu, 16 Oct 2003 09:48:09 +0000 (09:48 +0000)]
 - mtx_ownedby() was unpopular and is no longer needed.  Remove it.

20 years agoAdd a workaround for the fact that the priv field was removed from
dfr [Thu, 16 Oct 2003 09:18:35 +0000 (09:18 +0000)]
Add a workaround for the fact that the priv field was removed from
struct driver. We were the last user of that field (and we are scheduled
for demolition) so there wasn't much point in keeping it.

20 years ago* Add multiple inheritance to kobj. Each class can have zero or more base
dfr [Thu, 16 Oct 2003 09:16:28 +0000 (09:16 +0000)]
* Add multiple inheritance to kobj. Each class can have zero or more base
  classes and if a method is not found in a given class, its base classes
  are searched (in the order they were declared). This search is recursive,
  i.e. a method may be define in a base class of a base class.
* Change the kobj method lookup algorithm to one which is SMP-safe. This
  relies only on the constraint that an observer of a sequence of writes
  of pointer-sized values will see exactly one of those values, not a
  mixture of two or more values. This assumption holds for all processors
  which FreeBSD supports.
* Add locking to kobj class initialisation.
* Add a simpler form of 'inheritance' for devclasses. Each devclass can
  have a parent devclass. Searches for drivers continue up the chain of
  devclasses until either a matching driver is found or a devclass is
  reached which has no parent. This can allow, for instance, pci drivers
  to match cardbus devices (assuming that cardbus declares pci as its
  parent devclass).
* Increment __FreeBSD_version.

This preserves the driver API entirely except for one minor feature used
by the ISA compatibility shims. A workaround for ISA compatibility will
be committed separately. The kobj and newbus ABI has changed - all modules
must be recompiled.

20 years ago - Collapse sched_switchin() and sched_switchout() into sched_switch(). Now
jeff [Thu, 16 Oct 2003 08:53:46 +0000 (08:53 +0000)]
 - Collapse sched_switchin() and sched_switchout() into sched_switch().  Now
   mi_switch() calls sched_switch() which calls cpu_switch().  This is
   actually one less function call than it had been.

20 years ago - Update the sched api. sched_{add,rem,clock,pctcpu} now all accept a td
jeff [Thu, 16 Oct 2003 08:39:15 +0000 (08:39 +0000)]
 - Update the sched api.  sched_{add,rem,clock,pctcpu} now all accept a td
   argument rather than a kse.

20 years ago - The non iterative algorithm for interact_update was broken due to
jeff [Thu, 16 Oct 2003 08:17:43 +0000 (08:17 +0000)]
 - The non iterative algorithm for interact_update was broken due to
   rounding errors.  This was the source of the majority of the
   interactivity problems.  Reintroduce the old algorithm and its XXX.
 - Up the interactivity threshold to 30.  It really could stand to be even
   a tiny bit higher.
 - Let the sleep and run time accumulate up to 5 seconds of history rather
   than two.  This helps stop XFree86 from becoming non-interactive during
   bursts of activity.

20 years agoBump FreeBSD version for changed layout of cdevsw
phk [Thu, 16 Oct 2003 07:58:37 +0000 (07:58 +0000)]
Bump FreeBSD version for changed layout of cdevsw

20 years agoHandle realloc() failure correctly.
tjr [Thu, 16 Oct 2003 07:07:20 +0000 (07:07 +0000)]
Handle realloc() failure correctly.

20 years agoMalloc buckets of size 128 have been having their 64-byte offset
mckusick [Thu, 16 Oct 2003 02:00:12 +0000 (02:00 +0000)]
Malloc buckets of size 128 have been having their 64-byte offset
trashed after being freed. This has caused several panics including
kern/42277 related to soft updates. Jim Kuhn tracked the problem
down to ipfw limit rule processing.  In the expiry of dynamic rules,
it is possible for an O_LIMIT_PARENT rule to be removed when it still
has live children.  When the children eventually do expire, a pointer
to the (long gone) parent is dereferenced and a count decremented.
Since this memory can, and is, allocated for other purposes (in the
case of kern/42277 an inodedep structure), chaos ensues. The offset
in question in inodedep is the offset of the 16 bit count field in
the ipfw2 ipfw_dyn_rule.

Submitted by: Jim Kuhn <jkuhn@sandvine.com>
Reviewed by: "Evgueni V. Gavrilov" <aquatique@rusunix.org>
Reviewed by: Ben Pfountz <netprince@vt.edu>
MFC after: 1 week

20 years agoQuiet warnings about mis-matched pointer assignment.
obrien [Wed, 15 Oct 2003 20:05:57 +0000 (20:05 +0000)]
Quiet warnings about mis-matched pointer assignment.

Tested on: PowerPC

20 years agoIntroduce a new optional memberfunction for cdevsw, fdopen() which
phk [Wed, 15 Oct 2003 20:00:59 +0000 (20:00 +0000)]
Introduce a new optional memberfunction for cdevsw, fdopen() which
passes the fdidx from VOP_OPEN down.

This is for all I know the final API for this functionality, but
the locking semantics for messing with the filedescriptor from
the device driver are not settled at this time.

20 years agoPrint the correct logical drive lun number returned by the controller.
ps [Wed, 15 Oct 2003 18:52:44 +0000 (18:52 +0000)]
Print the correct logical drive lun number returned by the controller.

20 years agopurge extraneous ';'s
sam [Wed, 15 Oct 2003 18:19:28 +0000 (18:19 +0000)]
purge extraneous ';'s

Supported by: FreeBSD Foundation
Noticed by: bde

20 years agoAdd Allied Telesis SIC-AT boards support.
shiba [Wed, 15 Oct 2003 17:22:15 +0000 (17:22 +0000)]
Add Allied Telesis SIC-AT boards support.
Discussed in from [FreeBSD-tech-jp 3396] to [FreeBSD-tech-jp 3407]
at FreeBSD-tech-jp@jp.freebsd.org.

NOTE: We must put ed_probe_SIC() function into if_ed_isa.c because
      this is a bus dependent code. But the ed driver code is not
      separated explicitly whether it is bus dependent or independent
      now.

Refer to: http://plaza17.mbn.or.jp/~chi/myprog/FreeBSD/sicat.html
Submitted by: chi@bd.mbn.or.jp (Chiharu Shibata)

20 years agoCatch up with the English version and refine translation.
hrs [Wed, 15 Oct 2003 17:08:33 +0000 (17:08 +0000)]
Catch up with the English version and refine translation.

1.638 -> 1.640 relnotes/common/new.sgml

20 years agoRefine translation.
rushani [Wed, 15 Oct 2003 15:42:31 +0000 (15:42 +0000)]
Refine translation.

Submitted by: Hiroo Ono <hiroo@oikumene.gcd.org>,
Hiroyuki CHIBA <clin@imasy.org>
References: [doc-jp-work 724, 726]

20 years agoModified release note:
rushani [Wed, 15 Oct 2003 12:27:52 +0000 (12:27 +0000)]
Modified release note:
Note MFC of sk(4) support for SK-9521 V2.0 and 3COM 3C940.

20 years agoFix typo.
tg [Wed, 15 Oct 2003 10:25:31 +0000 (10:25 +0000)]
Fix typo.

20 years agoNew release note:
hrs [Wed, 15 Oct 2003 10:04:07 +0000 (10:04 +0000)]
New release note:
- xe(4) driver update
- tzdata2003d update

20 years agoMake sure to pull in all relevant crypto for GEOM_BDE
phk [Wed, 15 Oct 2003 08:53:04 +0000 (08:53 +0000)]
Make sure to pull in all relevant crypto for GEOM_BDE

20 years agoFix typos; reformat some entries.
tg [Wed, 15 Oct 2003 08:18:59 +0000 (08:18 +0000)]
Fix typos; reformat some entries.

20 years ago - vm_fault_copy_entry() should not assume that the source object contains
alc [Wed, 15 Oct 2003 08:00:45 +0000 (08:00 +0000)]
 - vm_fault_copy_entry() should not assume that the source object contains
   every page.  If the source entry was read-only, one or more wired pages
   could be in backing objects.
 - vm_fault_copy_entry() should not set the PG_WRITEABLE flag on the page
   unless the destination entry is, in fact, writeable.

20 years ago - If our user_pri doesn't match our actual priority our priority has been
jeff [Wed, 15 Oct 2003 07:47:06 +0000 (07:47 +0000)]
 - If our user_pri doesn't match our actual priority our priority has been
   elevated either due to priority propagation or because we're in the
   kernel in either case, put us on the current queue so that we dont
   stop others from using important resources.  At some point the priority
   elevations from sleeping in the kernel should go away.
 - Remove an optimization in sched_userret().  Before we would only set
   NEEDRESCHED if there was something of a higher priority available.  This
   is a trivial optimization and it breaks priority propagation because it
   doesn't take threads which we may be blocking into account.  Notice that
   the thread which is blocking others gets up to one tick of cpu time before
   we honor this NEEDRESCHED in sched_clock().

20 years agoAdd a wrapper for a function that takes and releases the adapter
deischen [Wed, 15 Oct 2003 05:34:41 +0000 (05:34 +0000)]
Add a wrapper for a function that takes and releases the adapter
lock around a call to the original function.  Make the timeout
function in callout_reset() use the wrapped function to avoid a
lock assertion panic.

Reviewed by: sam
Reported by: cgiordano@ids.net

20 years agoLet the world figure out exactly how superannuated I am.
linimon [Wed, 15 Oct 2003 03:31:06 +0000 (03:31 +0000)]
Let the world figure out exactly how superannuated I am.

20 years agoThe KERN_PROC_PROC sysctl took 4 args in 5.0-REL and 5.1-REL. We need to
peter [Wed, 15 Oct 2003 03:11:46 +0000 (03:11 +0000)]
The KERN_PROC_PROC sysctl took 4 args in 5.0-REL and 5.1-REL.  We need to
accept this for a bit longer.  Requiring the new order of 3 args only
was not very helpful.

20 years agoPull the tier-2 card one last time and break the get/setcontext and
peter [Wed, 15 Oct 2003 02:04:52 +0000 (02:04 +0000)]
Pull the tier-2 card one last time and break the get/setcontext and
sigreturn() ABI and the signal context on the stack.

Make the trapframe (and its shadows in the ucontext and sigframe etc)
8 bytes larger in order to preserve 16 byte stack alignment for the
following C code calls.  I could have done some padding after the
trapframe was saved, but some of the C code still expects an argument of
'struct trapframe'.  Anyway, this gives me a spare field that can be used
to store things like 'partial trapframe' status or something else in
the future.

The runtime impact is fairly small, *except* for threaded apps and things
that decode contexts and the signal stack (eg: cvsup binary).  Signal
delivery isn't too badly affected because the kernel generates the
sigframe that sigreturn uses after the handler has been called.

The size of mcontext_t and struct sigframe hasn't changed.  Only
the last few fields (sc_eip etc) got moved a little and I eliminated
a spare field.  mc_len/sc_len did change location though so the
sanity checks there will still trap it.

20 years agoGet some more data if we hit the pmap_enter() thing.
peter [Wed, 15 Oct 2003 00:45:35 +0000 (00:45 +0000)]
Get some more data if we hit the pmap_enter() thing.

20 years agoDetect errors during 'cat' page creation better. This fixes bin/5591.
jkoshy [Wed, 15 Oct 2003 00:36:44 +0000 (00:36 +0000)]
Detect errors during 'cat' page creation better.  This fixes bin/5591.

Reviewed by: ru

20 years agoo convert mutex calls to #defines for portability, etc.
sam [Tue, 14 Oct 2003 22:51:45 +0000 (22:51 +0000)]
o convert mutex calls to #defines for portability, etc.
o destroy mutex's on detach (was missing)

20 years agoMajor update to xe driver:
rsm [Tue, 14 Oct 2003 22:51:35 +0000 (22:51 +0000)]
Major update to xe driver:

- Make multicast work
- Fix (some of) the watchdog timeouts after card reset
- Add support for CE2, CEM28 and CEM33 cards
- General code cleanup

Any card that worked previously should still work, as well as a lot that
didn't.

The driver is not yet style(9) compliant; those changes are forthcoming,
once the functional changes are done.

PR: kern/50644
Reviewed by: imp
Approved by: imp

20 years agoMFV: tzdata2003d
wollman [Tue, 14 Oct 2003 21:03:22 +0000 (21:03 +0000)]
MFV: tzdata2003d

20 years agoTurn malloc+snprintf into asnprintf.
markm [Tue, 14 Oct 2003 20:42:13 +0000 (20:42 +0000)]
Turn malloc+snprintf into asnprintf.

Submitted by: David Hill <hill at phobia dot em ess>

20 years agoAdd the "TB ---" prefix to error and warning messages so that early
des [Tue, 14 Oct 2003 20:01:58 +0000 (20:01 +0000)]
Add the "TB ---" prefix to error and warning messages so that early
failures (e.g. failure to lock the sandbox) don't result in empty failure
reports.

20 years agoLock ip forwarding route cache. While we're at it, remove the global
sam [Tue, 14 Oct 2003 19:19:12 +0000 (19:19 +0000)]
Lock ip forwarding route cache.  While we're at it, remove the global
variable ipforward_rt by introducing an ip_forward_cacheinval() call
to use to invalidate the cache.

Supported by: FreeBSD Foundation

20 years agoMFp4: correct locking issues in nd6_lookup
sam [Tue, 14 Oct 2003 18:49:08 +0000 (18:49 +0000)]
MFp4: correct locking issues in nd6_lookup

Supported by: FreeBSD Foundation

20 years agoremove dangling ';'s` that were harmless
sam [Tue, 14 Oct 2003 18:45:50 +0000 (18:45 +0000)]
remove dangling ';'s` that were harmless

Supported by: FreeBSD Foundation