]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoFix kernel breakage introduced in the previous commit (redeclatation
sobomax [Tue, 24 Oct 2006 17:24:11 +0000 (17:24 +0000)]
Fix kernel breakage introduced in the previous commit (redeclatation
of the audit functions).

17 years agoMerge res_send.c of BIND9 9.3.3rc2 into main chunk. It includes
ume [Tue, 24 Oct 2006 14:41:43 +0000 (14:41 +0000)]
Merge res_send.c of BIND9 9.3.3rc2 into main chunk.  It includes
following fix:

Retransmission timeouts should be based on which attempt
it is to the nameserver and not the nameserver itself.

Obtained from: ISC
MFC after: 3 days

17 years agoRegenerate.
rwatson [Tue, 24 Oct 2006 13:54:56 +0000 (13:54 +0000)]
Regenerate.

17 years agoHook up audit functions in the freebsd32 compatibility code. It is
rwatson [Tue, 24 Oct 2006 13:49:44 +0000 (13:49 +0000)]
Hook up audit functions in the freebsd32 compatibility code.  It is
believed these likely don't require wrappers.

Reported by: sobomax
MFC after: 3 days

17 years agoReduce our pointy-hat pool capacity by one (i.e. bump manpage date,
keramida [Tue, 24 Oct 2006 13:27:08 +0000 (13:27 +0000)]
Reduce our pointy-hat pool capacity by one (i.e. bump manpage date,
after the last change to this manpage).

Reminded by: ru

17 years agoDocument NO_IGNORE and its interation with FORBIDDEN.
keramida [Tue, 24 Oct 2006 12:30:28 +0000 (12:30 +0000)]
Document NO_IGNORE and its interation with FORBIDDEN.

Triggered by:    a question by Panagiotis Hadgis <hadgis@ceid.upatras.gr>

17 years agoProtect sigqueue_take() call by child process's lock, it fixed a
davidxu [Tue, 24 Oct 2006 12:04:21 +0000 (12:04 +0000)]
Protect sigqueue_take() call by child process's lock, it fixed a
potential race with ptrace 'attach' which changes parent of the
child process.

17 years agoDitch crummy fattime <--> timespec conversion functions
phk [Tue, 24 Oct 2006 11:55:18 +0000 (11:55 +0000)]
Ditch crummy fattime <--> timespec conversion functions

17 years agoDrop crummy fattime to timespec conversion routines.
phk [Tue, 24 Oct 2006 11:43:41 +0000 (11:43 +0000)]
Drop crummy fattime to timespec conversion routines.

Leave a XXX here for anybody able to test.

17 years ago(Forced commit.) Forgot to note that previous change was
ru [Tue, 24 Oct 2006 11:20:29 +0000 (11:20 +0000)]
(Forced commit.)  Forgot to note that previous change was

Submitted by: Stepan A. Baranov

17 years agoRemove the non-standard and undocumented STAILQ_REMOVE_HEAD_UNTIL()
ru [Tue, 24 Oct 2006 11:19:21 +0000 (11:19 +0000)]
Remove the non-standard and undocumented STAILQ_REMOVE_HEAD_UNTIL()
macro now that we no longer use it.

17 years agoSwitch to using STAILQ_REMOVE_HEAD() instead of STAILQ_REMOVE_HEAD_UNTIL().
ru [Tue, 24 Oct 2006 11:17:46 +0000 (11:17 +0000)]
Switch to using STAILQ_REMOVE_HEAD() instead of STAILQ_REMOVE_HEAD_UNTIL().

Submitted by: Stepan A. Baranov

This corresponds to OpenBSD rev. 1.134:

: revision 1.134
: date: 2004/05/04 16:59:31;  author: grange;  state: Exp;  lines: +10 -10
: Remove useless ``elm'' argument from the SIMPLEQ_REMOVE_HEAD macro.
: This matches our SLIST behaviour and NetBSD's SIMPLEQ as well.
:
: ok millert krw deraadt

17 years agoReplace slightly crummy fattime<->timespec conversion functions.
phk [Tue, 24 Oct 2006 11:14:05 +0000 (11:14 +0000)]
Replace slightly crummy fattime<->timespec conversion functions.

17 years agoBetter naming of fattime conversion functions, they do convert to timespec
phk [Tue, 24 Oct 2006 10:27:23 +0000 (10:27 +0000)]
Better naming of fattime conversion functions, they do convert to timespec
after all.

Add 'utc' argument to control if fattimestamps are on UTC or local timezone
calendar.

17 years agoDocument net.link.bridge.ipfw_arp for filtering ARP packets with ipfw.
thompsa [Tue, 24 Oct 2006 08:47:39 +0000 (08:47 +0000)]
Document net.link.bridge.ipfw_arp for filtering ARP packets with ipfw.

MFC after: 3 days

17 years agoRefine the checksum hack a little. It appears that the chip can handle UDP
scottl [Tue, 24 Oct 2006 08:24:31 +0000 (08:24 +0000)]
Refine the checksum hack a little.  It appears that the chip can handle UDP
and TCP checksum offloading fine, it only has a problem with IP checksums on
IP fragments..  Barring a fix or workaround available from the hardware, the
real solution would be to have finer grained control in the stack over what
can and cannot be assisted in hardware.

17 years agoIgnore SIGSYS when BSM is compiled in. Otherwise, attempt to invoke su on
sobomax [Tue, 24 Oct 2006 08:18:10 +0000 (08:18 +0000)]
Ignore SIGSYS when BSM is compiled in. Otherwise, attempt to invoke su on
system that don't have audit framefork compiled into kernel or ia32 binary
on amd64 system will result in SIGSYS. There is one place in su.c itself
where it tries to check for errno != ENOSYS, but it has been a nop since su
does not catch SIGSYS anyway. There are few other places in libbsm,
where attempt to invoke audit syscal would result in SIGSYS if no audit
support is present in the kernel, so that the only reliable method for
now is to disable SIGSYS completely in the case when BSM is compiled in.

In the long run, both direct invocation of audit-related syscalls and
libbsm should be made more intellegent to handle the case when BSM is not
compiled into the kernel gracefully.

MFC after: 3 days
           (provided re@ approval)

17 years agoDon't leak files.
imp [Tue, 24 Oct 2006 07:12:31 +0000 (07:12 +0000)]
Don't leak files.
This also eliminates the need for ifp init.

Submitted by: ru@

17 years agoDocument multiple hints lines.
imp [Tue, 24 Oct 2006 00:44:40 +0000 (00:44 +0000)]
Document multiple hints lines.
Tweak machine description to more accurately reflect second argument being
the cpu architecture.  Note correspondence with MACHINE and MACHINE_ARCH.

17 years agoEnd my resistance to jmg's multiple hints files and bring in support
imp [Tue, 24 Oct 2006 00:31:59 +0000 (00:31 +0000)]
End my resistance to jmg's multiple hints files and bring in support
for having multiple hints files generate a correct hints.c (eg, with
all the specified ones catenated together).

17 years agostyle(9) pedantry
imp [Tue, 24 Oct 2006 00:07:39 +0000 (00:07 +0000)]
style(9) pedantry

17 years agoSince multiple hints lines do not work, make doing that a fatal error.
imp [Tue, 24 Oct 2006 00:02:23 +0000 (00:02 +0000)]
Since multiple hints lines do not work, make doing that a fatal error.

17 years agoIf the user did not chose a distribution from the menu, a sensible
erwin [Mon, 23 Oct 2006 20:49:49 +0000 (20:49 +0000)]
If the user did not chose a distribution from the menu, a sensible
default distribution (user) will be installed, however, no kernel
will be installed.  Fix this by installing a sensible default kernel,
SMP/GENERIC, depending on WITH_SMP and number of CPUS found.

Reviewed by:    jhb

17 years agoIt so happens that everything that is stupid is not unconstitutional.
phk [Mon, 23 Oct 2006 13:25:17 +0000 (13:25 +0000)]
It so happens that everything that is stupid is not unconstitutional.
               -- Supreme Court Justice Antonio Scalia

17 years agoMove "device splash" back to MI NOTES and "files", it's MI.
ru [Mon, 23 Oct 2006 13:23:14 +0000 (13:23 +0000)]
Move "device splash" back to MI NOTES and "files", it's MI.

17 years agoMove MI parts of syscons into MI "files".
ru [Mon, 23 Oct 2006 13:05:01 +0000 (13:05 +0000)]
Move MI parts of syscons into MI "files".

17 years agoIf not compiled for debugging, redirect standard input/output/error
ru [Mon, 23 Oct 2006 10:48:19 +0000 (10:48 +0000)]
If not compiled for debugging, redirect standard input/output/error
to /dev/null before becoming a daemon.

17 years agoRevision 1.25 had the ATKBD_DFLT_KEYMAP option turned on and then off:
ru [Mon, 23 Oct 2006 10:05:36 +0000 (10:05 +0000)]
Revision 1.25 had the ATKBD_DFLT_KEYMAP option turned on and then off:

: # Options for atkbd:
: options        ATKBD_DFLT_KEYMAP       # specify the built-in keymap
: makeoptions    ATKBD_DFLT_KEYMAP=jp.106
[...]
: nooption       ATKBD_DFLT_KEYMAP
: nomakeoption   ATKBD_DFLT_KEYMAP

(Previously the option was inherited from MI NOTES.)  So my tool in
rev. 1.26 reduced this to removing all "ATKBD_DFLT_KEYMAP" lines,
leaving the option effectively disabled as it was before, but since
it's actually supported on sparc64, turn it on now.

17 years agoMechanically kill redundant nodevice/nooption/nomakeoption, i.e.,
ru [Mon, 23 Oct 2006 09:45:22 +0000 (09:45 +0000)]
Mechanically kill redundant nodevice/nooption/nomakeoption, i.e.,
those that do not exist in MI NOTES or switched on/off in the MD
NOTES.

17 years agoRe-add screen savers, which I removed by mistake.
des [Mon, 23 Oct 2006 08:38:42 +0000 (08:38 +0000)]
Re-add screen savers, which I removed by mistake.

Noticed by: ru

17 years agoRevert the rev. 1.4, it causes breakage on most arches except i386.
kib [Mon, 23 Oct 2006 07:25:25 +0000 (07:25 +0000)]
Revert the rev. 1.4, it causes breakage on most arches except i386.

Remove the const qualifier from ap argument for __v2printf, that induced
that breakage, and seems to be the real reason for bad code. ap is modified
inside the __v2printf body by va_arg macro.

Pointy hat to: kib
Approved by: pjd (mentor)

17 years agoThe page queues lock is no longer required by vm_page_wakeup().
alc [Mon, 23 Oct 2006 05:27:31 +0000 (05:27 +0000)]
The page queues lock is no longer required by vm_page_wakeup().

17 years agoSynchronize pf.os with OpenBSD.
delphij [Mon, 23 Oct 2006 05:09:44 +0000 (05:09 +0000)]
Synchronize pf.os with OpenBSD.

Obtained from: OpenBSD
MFC After: 3 days
Approved by: mlaier (maintainer)

17 years agoo Eliminate nexus_print_resources(). Use resource_list_print_type()
marcel [Mon, 23 Oct 2006 00:38:58 +0000 (00:38 +0000)]
o  Eliminate nexus_print_resources(). Use resource_list_print_type()
   instead.
o  Eliminate nexus_print_all_resources(). Inline the function body
   in nexus_print_child().

17 years agonet80211 seems to give us more rates than specified via ic_sup_rates under
mlaier [Mon, 23 Oct 2006 00:34:07 +0000 (00:34 +0000)]
net80211 seems to give us more rates than specified via ic_sup_rates under
yet to be investigated circumstances. If that happens truncate to the number
of rates that the firmware supports.

Found by: Jeremie Le Hen
Obtained from: OpenBSD (w/ changes)
MFC after: 3 days

17 years agoDon't descend into at91 until we sort out the boot loader issues more generally
imp [Mon, 23 Oct 2006 00:05:28 +0000 (00:05 +0000)]
Don't descend into at91 until we sort out the boot loader issues more generally

17 years agoClarify that the input hook requires binary packet data. Instead of
emaste [Sun, 22 Oct 2006 21:33:57 +0000 (21:33 +0000)]
Clarify that the input hook requires binary packet data.  Instead of
referencing an omitted example, include a short (Perl) example that
queues an ICMP Echo message.

17 years agoThe page queues lock is no longer required by vm_page_busy() or
alc [Sun, 22 Oct 2006 21:18:48 +0000 (21:18 +0000)]
The page queues lock is no longer required by vm_page_busy() or
vm_page_wakeup().  Reduce or eliminate its use accordingly.

17 years agoAdd two new functions to convert FAT filesystem format timestamps
phk [Sun, 22 Oct 2006 18:19:08 +0000 (18:19 +0000)]
Add two new functions to convert FAT filesystem format timestamps
to and from struct timespec, to replace the crummy conversion
function which have been copy&pasted into three different
filesystems already.

Apart from general crummyness as indicated by code like:

for (year = 1970;; year++) {
inc = year & 0x03 ? 365 : 366;
if (days < inc)
break;
days -= inc;
}

They also contain specialized crummyness which tries to compensate
for the general crummyness by caching recent conversion results,
with no regard for locking or consistency.

These replacement functions are smaller, O(1) and handle the Y2.1K
leap-year correctly.

Ideally, these functions should live in a module of their own,
which the three offending filesystems would depend on, but the
size is 877 bytes of code (on i386), so that would be false
economy.

17 years agoFix ttyv* device names.
ru [Sun, 22 Oct 2006 18:02:01 +0000 (18:02 +0000)]
Fix ttyv* device names.

PR: docs/104376
Submitted by: Michael Le Barbier
MFC after: 3 days

17 years agoSuppress a spurious warning message when a kernel without INET6 is
hrs [Sun, 22 Oct 2006 17:21:03 +0000 (17:21 +0000)]
Suppress a spurious warning message when a kernel without INET6 is
used.

Spotted by: ru
Reviewed by: ume
MFC after: 3 days

17 years agoFix build: remove (now) unnecessary PG_BUSY check, it's handled by vm object locking.
delphij [Sun, 22 Oct 2006 16:33:43 +0000 (16:33 +0000)]
Fix build: remove (now) unnecessary PG_BUSY check, it's handled by vm object locking.

17 years agoFix some typos and grammar.
ru [Sun, 22 Oct 2006 16:21:55 +0000 (16:21 +0000)]
Fix some typos and grammar.

PR: docs/66770
Submitted by: Michel Lavondès
MFC after: 3 days

17 years agoComplete break-out of sys/sys/mac.h into sys/security/mac/mac_framework.h
rwatson [Sun, 22 Oct 2006 11:52:19 +0000 (11:52 +0000)]
Complete break-out of sys/sys/mac.h into sys/security/mac/mac_framework.h
begun with a repo-copy of mac.h to mac_framework.h.  sys/mac.h now
contains the userspace and user<->kernel API and definitions, with all
in-kernel interfaces moved to mac_framework.h, which is now included
across most of the kernel instead.

This change is the first step in a larger cleanup and sweep of MAC
Framework interfaces in the kernel, and will not be MFC'd.

Obtained from: TrustedBSD Project
Sponsored by: SPARTA

17 years agoReplace PG_BUSY with VPO_BUSY. In other words, changes to the page's
alc [Sun, 22 Oct 2006 04:28:14 +0000 (04:28 +0000)]
Replace PG_BUSY with VPO_BUSY.  In other words, changes to the page's
busy flag, i.e., VPO_BUSY, are now synchronized by the per-vm object
lock instead of the global page queues lock.

17 years agoEliminate unnecessary PG_BUSY tests.
alc [Sun, 22 Oct 2006 04:18:01 +0000 (04:18 +0000)]
Eliminate unnecessary PG_BUSY tests.

17 years agoInitialize T1 to silent gcc warning.
kevlo [Sun, 22 Oct 2006 02:19:33 +0000 (02:19 +0000)]
Initialize T1 to silent gcc warning.

Approved by: cognet

17 years agoUse macro TAILQ_FOREACH_SAFE instead of expanding it.
davidxu [Sun, 22 Oct 2006 00:09:41 +0000 (00:09 +0000)]
Use macro TAILQ_FOREACH_SAFE instead of expanding it.

17 years agoSince revision 1.333 of kern_sig.c no longer uses P_WEXIT, the change
davidxu [Sat, 21 Oct 2006 23:59:15 +0000 (23:59 +0000)]
Since revision 1.333 of kern_sig.c no longer uses P_WEXIT, the change
opened a race window which can cause memory leak in signal queue.
Here we free memory for signal queue when process state is set to
PRS_ZOMBIE.

17 years ago- Document the default disk block usage unit.
ru [Sat, 21 Oct 2006 23:57:38 +0000 (23:57 +0000)]
- Document the default disk block usage unit.
- Add option -h, human-readable output.

PR: docs/53732

17 years agoMFp4: default to not loading the fpga
imp [Sat, 21 Oct 2006 22:51:21 +0000 (22:51 +0000)]
MFp4: default to not loading the fpga

17 years agoMFp4: Update to smaller code footprint.
imp [Sat, 21 Oct 2006 22:44:26 +0000 (22:44 +0000)]
MFp4: Update to smaller code footprint.

17 years agoMFp4: Move to smaller code footprint.
imp [Sat, 21 Oct 2006 22:43:39 +0000 (22:43 +0000)]
MFp4: Move to smaller code footprint.

17 years agoMFp4: Move to smaller code. This was somehow forgotten before.
imp [Sat, 21 Oct 2006 22:43:07 +0000 (22:43 +0000)]
MFp4: Move to smaller code.  This was somehow forgotten before.

17 years agoEliminate unnecessary PG_BUSY tests. They originally served a purpose
alc [Sat, 21 Oct 2006 21:02:04 +0000 (21:02 +0000)]
Eliminate unnecessary PG_BUSY tests.  They originally served a purpose
that is now handled by vm object locking.

17 years agoThere seems to be a problem with txcsum and UDP. Turn it off until it is
scottl [Sat, 21 Oct 2006 20:41:08 +0000 (20:41 +0000)]
There seems to be a problem with txcsum and UDP.  Turn it off until it is
understood better.  This fixes timeout problems with NFS.

17 years agoMinor markup fix: A comma should be seperated by space from macro
simon [Sat, 21 Oct 2006 20:36:50 +0000 (20:36 +0000)]
Minor markup fix: A comma should be seperated by space from macro
argument.  This resulted in "pwritev,()" instead of the intended
"pwritev()," in the output.

MFC after: 3 days

17 years agoMake it clear that section 8 also documents system management
ru [Sat, 21 Oct 2006 20:35:49 +0000 (20:35 +0000)]
Make it clear that section 8 also documents system management
procedures, not necessarily commands.

PR: docs/35011
Obtained from: NetBSD (partly)

17 years agoComment out the first paragraph; while it tells the true (we're using
ru [Sat, 21 Oct 2006 20:09:51 +0000 (20:09 +0000)]
Comment out the first paragraph; while it tells the true (we're using
the old BSD config(8) utility in FreeBSD), it does not look good.

PR: docs/100328

17 years agoSmall fixes.
ru [Sat, 21 Oct 2006 18:05:55 +0000 (18:05 +0000)]
Small fixes.

17 years agoBetter handle the -F case:
marcel [Sat, 21 Oct 2006 18:05:06 +0000 (18:05 +0000)]
Better handle the -F case:
o  When stat(2) fails (i.e. the file has been moved) there's no new
   file with the same name yet, so keep showing the file that's open.
   This yields the same behaviour as -f, for which we don't stat(2).
o  When a new file with the same name has been created (i.e stat(2)
   succeeds but the inode or device numbers differ from the opened
   file), show any new lines in the opened file (i.e. the old or
   rotated file) before reopening the new file.

These changes fix the observed behaviour that tail(1) doesn't show
the very last lines of the rotated (log) files.

PR: bin/101979
Tested by: Jos Backus <jos@catnook.com>
MFC after: 2 months

17 years agoBump document date for previous revision.
ru [Sat, 21 Oct 2006 18:04:42 +0000 (18:04 +0000)]
Bump document date for previous revision.

17 years agoSpelling.
ru [Sat, 21 Oct 2006 18:04:20 +0000 (18:04 +0000)]
Spelling.

17 years agoAvoid a potential hard sentence break.
ru [Sat, 21 Oct 2006 18:03:53 +0000 (18:03 +0000)]
Avoid a potential hard sentence break.

17 years agoBump document date for the previous change.
ru [Sat, 21 Oct 2006 17:39:35 +0000 (17:39 +0000)]
Bump document date for the previous change.

17 years agoTurn on gdbserver for the arch that supports it..
jmg [Sat, 21 Oct 2006 17:27:36 +0000 (17:27 +0000)]
Turn on gdbserver for the arch that supports it..

17 years agoRevise the markup, sort sections, fix some grammar bugs.
ru [Sat, 21 Oct 2006 17:15:11 +0000 (17:15 +0000)]
Revise the markup, sort sections, fix some grammar bugs.
Not fixed: the description of media-type is completely bogus.

17 years agoBump document date for the previous change.
ru [Sat, 21 Oct 2006 16:08:21 +0000 (16:08 +0000)]
Bump document date for the previous change.

17 years agoFix .Dd arguments.
ru [Sat, 21 Oct 2006 16:02:42 +0000 (16:02 +0000)]
Fix .Dd arguments.

17 years agoo Fix typo.
maxim [Sat, 21 Oct 2006 15:12:36 +0000 (15:12 +0000)]
o Fix typo.

Obtained from: DragonFlyBSD
MFC after: 1 week

17 years agoRecord the pccardc death.
ru [Sat, 21 Oct 2006 14:19:52 +0000 (14:19 +0000)]
Record the pccardc death.

17 years agoMFCs noted: libpcap 0.9.4, tcpdump 3.9.4.
bmah [Sat, 21 Oct 2006 14:06:34 +0000 (14:06 +0000)]
MFCs noted:  libpcap 0.9.4, tcpdump 3.9.4.

17 years agoMFi386: 1.13: Fix booting with ps2 keyboards.
ru [Sat, 21 Oct 2006 12:52:46 +0000 (12:52 +0000)]
MFi386: 1.13: Fix booting with ps2 keyboards.

17 years agoWorkaround for (what seems to be) compiler error for gcc 3.4.6. On
kib [Sat, 21 Oct 2006 11:49:07 +0000 (11:49 +0000)]
Workaround for (what seems to be) compiler error for gcc 3.4.6. On
i386 with default optimization level (-O2), va_list pointer ap in the
__v2printf function is advanced before the use. That cause argument
shift and garbage instead last argument in printf-family when xprintf is
activated.

The nsswitch is easy victim of the bug.

Reviewed by: kan
Approved by: kan (mentor)
MFC after: 1 week

17 years agoSet LC_ALL=C in order to avoid problems with character ranges and
cperciva [Sat, 21 Oct 2006 11:04:34 +0000 (11:04 +0000)]
Set LC_ALL=C in order to avoid problems with character ranges and
sorting.

PR: bin/104505
MFC after: 3 days

17 years agoThe vm.zone sysctl has gone; zone stats are now supplied by libmemstat(3).
ru [Sat, 21 Oct 2006 10:40:12 +0000 (10:40 +0000)]
The vm.zone sysctl has gone; zone stats are now supplied by libmemstat(3).

17 years ago- Add comma after REQUESTS field missed in previous commit.
ru [Sat, 21 Oct 2006 10:16:26 +0000 (10:16 +0000)]
- Add comma after REQUESTS field missed in previous commit.
- Widen some columns; make width of header columns less cryptic.

17 years agoFix error in rev. 1.68. The intention was to break out the switch(){},
glebius [Sat, 21 Oct 2006 09:44:02 +0000 (09:44 +0000)]
Fix error in rev. 1.68. The intention was to break out the switch(){},
but actually exited from the for(){} loop. This fixes the PPPIOCSCOMPRESS
ioctl.

PR: kern/101333
Submitted by: Igor Popov <igorpopov newmail.ru>

17 years agoPrint the number of allocation failures in UMA zones.
glebius [Sat, 21 Oct 2006 09:08:35 +0000 (09:08 +0000)]
Print the number of allocation failures in UMA zones.

PR: kern/102940

17 years agoFake the link count until we have no choice but to load data from the
trhodes [Sat, 21 Oct 2006 08:17:17 +0000 (08:17 +0000)]
Fake the link count until we have no choice but to load data from the
MFT.

PR: 86965
Submitted by: Lowell Gilbert <lgfbsd@be-well.ilk.org>

17 years agoBe more like the BGE driver. Ensure that at least 16 TX descriptors are
scottl [Sat, 21 Oct 2006 07:54:39 +0000 (07:54 +0000)]
Be more like the BGE driver.  Ensure that at least 16 TX descriptors are
kept unused in the ring.  This check should probably be moved up to
bce_start_locked at some point, as it'll make the loop up there slightly
more efficient, and will eliminate a costly set of busdma operations when
the ring is full.  But this works for now.

This makes all of my UDP torture tests work.  I'll cautiously say that
it might even work for other users now.  Feedback is appreciated.

17 years agostyle(9) cleanup.
kevlo [Sat, 21 Oct 2006 04:25:00 +0000 (04:25 +0000)]
style(9) cleanup.

Approved by: cognet

17 years agoFix whitespace
scottl [Sat, 21 Oct 2006 04:13:20 +0000 (04:13 +0000)]
Fix whitespace

17 years agoOk I am an idiot. On 32 bits big-endian systems, it is needed to handle the
cognet [Sat, 21 Oct 2006 00:46:56 +0000 (00:46 +0000)]
Ok I am an idiot. On 32 bits big-endian systems, it is needed to handle the
syscalls using __syscall but only actually returning 32bits, such as mmap(),
specially : they set the return value in td->td_retval[0], but the userland
functions will expect this in r1, and not in r0 as it is normally done, as it
is the LSB. So add a special case for all these syscalls (all except lseek,
which truly returns 64bits).

Many thanks to Peter Grehan for his patience while explaining me the issue.

17 years agorevert last change.. premature.. need to wait until if_ethersubr.c
julian [Sat, 21 Oct 2006 00:16:31 +0000 (00:16 +0000)]
revert last change.. premature.. need to wait until if_ethersubr.c
uses pfil to get to ipfw.

17 years agoUse __QUAD_LOWWORD for __syscall, to always use the good word, whatever the
cognet [Fri, 20 Oct 2006 22:40:31 +0000 (22:40 +0000)]
Use __QUAD_LOWWORD for __syscall, to always use the good word, whatever the
endianness is.

17 years agofix tab indentation for CP and RV...
jmg [Fri, 20 Oct 2006 21:28:11 +0000 (21:28 +0000)]
fix tab indentation for CP and RV...

If the length is zero, catch this early, instead of making dflen go negative
and letting bad things happen...  We also check to see if RV (checksum) is
0, and handle that has a checksum failure...

Properly handle checksum failures by not processing read-write VPD data,
and removing all the found read-only data...

Tested by: oleg (dflen going negative)

17 years agoMove some variables to a more likely place
julian [Fri, 20 Oct 2006 19:32:08 +0000 (19:32 +0000)]
Move some variables to a more likely place
and remove "temporary" stuff that is not needed any more.

17 years agoVoltaire again.
phk [Fri, 20 Oct 2006 18:12:25 +0000 (18:12 +0000)]
Voltaire again.

17 years agoThere's no fpga.c, so omit it.
imp [Fri, 20 Oct 2006 16:57:30 +0000 (16:57 +0000)]
There's no fpga.c, so omit it.

17 years agoRemove the check that prevented signals from being delivered to exiting
jhb [Fri, 20 Oct 2006 16:19:21 +0000 (16:19 +0000)]
Remove the check that prevented signals from being delivered to exiting
processes.  It was originally added back when support for Linux threads
(and thus shared sigacts objects) was added, but no one knows why.  My
guess is that at some point during the Linux threads patches, the sigacts
object was torn down during exit1(), so this check was added to prevent
a panic for that race.  However, the stuff that was actually committed to
the tree doesn't teardown sigacts until wait() making the above race moot.
Re-allowing signals here lets one interrupt a NFS request during process
teardown (such as closing descriptors) on an interruptible mount.

Requested by: kib (long time ago)
MFC after: 1 week

17 years agoremove libc_r remanent.
davidxu [Fri, 20 Oct 2006 14:15:42 +0000 (14:15 +0000)]
remove libc_r remanent.

17 years agoo '-s' flag was killed in rev. 1.75. Clean getopt(3).
maxim [Fri, 20 Oct 2006 13:10:27 +0000 (13:10 +0000)]
o '-s' flag was killed in rev. 1.75.  Clean getopt(3).

PR: bin/104616
Submitted by: Oliver Fromme
MFC after: 1 week

17 years ago- MFi386: Remove 'device io'.
nyan [Fri, 20 Oct 2006 11:48:46 +0000 (11:48 +0000)]
- MFi386: Remove 'device io'.
- Remove duplicate options.
- 'nomakeoption ATKBD_DFLT_KEYMAP' is not needed anymore.

17 years agoThere's no need to special-case lseek for arm/big-endian.
cognet [Fri, 20 Oct 2006 11:00:03 +0000 (11:00 +0000)]
There's no need to special-case lseek for arm/big-endian.

17 years agoFix a recent regression regarding valid signals.
netchild [Fri, 20 Oct 2006 10:09:40 +0000 (10:09 +0000)]
Fix a recent regression regarding valid signals.

Submitted by: rdivacky

17 years agoMove more MD devices and options out of MI NOTES.
des [Fri, 20 Oct 2006 09:52:27 +0000 (09:52 +0000)]
Move more MD devices and options out of MI NOTES.

17 years agoDon't show debug registers in "show registers". Special registers should
bde [Fri, 20 Oct 2006 09:44:21 +0000 (09:44 +0000)]
Don't show debug registers in "show registers".  Special registers should
be displayed specially, and debug registers are among of the least
interesting special registers (far behind %cr3).  The debug registers
are still accessible as variables and displayed in another bogus place
("show watches").

17 years agoMFp4:
imp [Fri, 20 Oct 2006 09:12:05 +0000 (09:12 +0000)]
MFp4:

Massive update.  The highlights:
o dramatically cut memory usage by writing better, less intertwingled
  code.
o implement booting off mmc/sd cards (sd only tested one at the moment)
o start to split out board specific stuff for boot2.

17 years agoThe VGA_DEBUG option only exists on {amd64,i386,ia64}.
des [Fri, 20 Oct 2006 08:56:26 +0000 (08:56 +0000)]
The VGA_DEBUG option only exists on {amd64,i386,ia64}.
Also remove 'device io' from amd64 NOTES; DEFAULTS takes care of it.