]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoThe got_siginfo = 0 should have been got_sigalarm=0 to match the other
imp [Sun, 10 Oct 2004 06:37:56 +0000 (06:37 +0000)]
The got_siginfo = 0 should have been got_sigalarm=0 to match the other
passes.

Submitted by: Dworkin Muller

19 years agoDon't release the slot twice.. sched_rem() has already done it.
julian [Sun, 10 Oct 2004 05:19:22 +0000 (05:19 +0000)]
Don't release the slot twice.. sched_rem() has already done it.

Submitted by: stephan uphoff (ups at tree dot com)
MFC after: 3 days

19 years agoRemove duplicate line.
julian [Sun, 10 Oct 2004 05:07:43 +0000 (05:07 +0000)]
Remove duplicate line.

19 years agoConvert to newbus. (chances are we could now move this to dev/pbio
imp [Sun, 10 Oct 2004 03:26:20 +0000 (03:26 +0000)]
Convert to newbus.  (chances are we could now move this to dev/pbio
since I believe it is now MI, but that hasn't been done yet).

Reviewed by: dds

19 years agoModify entropy harvesting locking strategy:
rwatson [Sat, 9 Oct 2004 22:04:13 +0000 (22:04 +0000)]
Modify entropy harvesting locking strategy:

- Trade off granularity to reduce overhead, since the current model
  doesn't appear to reduce contention substantially: move to a single
  harvest mutex protecting harvesting queues, rather than one mutex
  per source plus a mutex for the free list.

- Reduce mutex operations in a harvesting event to 2 from 4, and
  maintain lockless read to avoid mutex operations if the queue is
  full.

- When reaping harvested entries from the queue, move all entries from
  the queue at once, and when done with them, insert them all into a
  thread-local queue for processing; then insert them all into the
  empty fifo at once.  This reduces O(4n) mutex operations to O(2)
  mutex operations per wakeup.

In the future, we may want to look at re-introducing granularity,
although perhaps at the granularity of the source rather than the
source class; both the new and old strategies would cause contention
between different instances of the same source (i.e., multiple
network interfaces).

Reviewed by: markm

19 years agoAdd a simple C-based TCP connection generator, which generates and
rwatson [Sat, 9 Oct 2004 20:58:28 +0000 (20:58 +0000)]
Add a simple C-based TCP connection generator, which generates and
closes the specified number of TCP connections sequentially and
synchronously.  Useful for trying to trigger races in the accept
code.

19 years agoFix fsbtodb() for UFS1. This fixes an overflow for file sizes >1 TB,
njl [Sat, 9 Oct 2004 20:16:06 +0000 (20:16 +0000)]
Fix fsbtodb() for UFS1.  This fixes an overflow for file sizes >1 TB,
allowing for sizes up to 4 TB.  This doesn't affect UFS2 since b is already
a 64 bit type, coincidental with daddr_t.

Submitted by: bde

19 years agoAdd a note to the man page warning users about possible lock order
csjp [Sat, 9 Oct 2004 20:07:33 +0000 (20:07 +0000)]
Add a note to the man page warning users about possible lock order
reversals+system lock ups if they are using ucred based rules
while running with debug.mpsafenet=1.

I am working on merging a shared locking mechanism into ipfw which
should take care of this problem, but it still requires a bit more
testing and review.

19 years agoIn the previous revision, I did not intend to change the default value
green [Sat, 9 Oct 2004 18:51:32 +0000 (18:51 +0000)]
In the previous revision, I did not intend to change the default value
of "nosleepwithlocks."

Submitted by: ru

19 years agoUse the FP_ILOG macros from <math.h> rather than hardcoded return values.
stefanf [Sat, 9 Oct 2004 17:14:28 +0000 (17:14 +0000)]
Use the FP_ILOG macros from <math.h> rather than hardcoded return values.
Also be prepared for FP_ILOGBNAN != INT_MAX.

Reviewed by: md5

19 years agoo Backout rev. 1.16, see 1.3 commit log for more info.
maxim [Sat, 9 Oct 2004 17:13:58 +0000 (17:13 +0000)]
o Backout rev. 1.16, see 1.3 commit log for more info.

Requested by: bde

o Remove unneeded sys/types.h and netinet/in.h from the synopsis and
the example.
o We do have struct in_addr in arpa/inet.h, so no need for netinet/in.h.
o Mention where AF_* constants defined are.

Educated by: bde

19 years agoAcquire the send socket buffer lock around tcp_output() activities
rwatson [Sat, 9 Oct 2004 16:48:51 +0000 (16:48 +0000)]
Acquire the send socket buffer lock around tcp_output() activities
reaching into the socket buffer.  This prevents a number of potential
races, including dereferencing of sb_mb while unlocked leading to
a NULL pointer deref (how I found it).  Potentially this might also
explain other "odd" TCP behavior on SMP boxes (although  haven't
seen it reported).

RELENG_5 candidate.

19 years agoAdd SOCKBUF_UNLOCK_ASSERT(), which asserts that the current thread does
rwatson [Sat, 9 Oct 2004 16:42:57 +0000 (16:42 +0000)]
Add SOCKBUF_UNLOCK_ASSERT(), which asserts that the current thread does
not hold the mutex for a socket buffer.

19 years agoFlush the register windows before we start changing the context.
kensmith [Sat, 9 Oct 2004 16:42:09 +0000 (16:42 +0000)]
Flush the register windows before we start changing the context.

Submitted by: Andrew Belashov <bel (at) orel.ru> (slightly modified)
Reviewed by: jake

19 years agoAdd support for the ICH6 in legacy mode.
sos [Sat, 9 Oct 2004 16:27:13 +0000 (16:27 +0000)]
Add support for the ICH6 in legacy mode.
The AHCI part is not supported yet, but is in the works.

5.3 RC1 candidate

19 years agoMake fsck WARNS=2 clean.
le [Sat, 9 Oct 2004 15:56:34 +0000 (15:56 +0000)]
Make fsck WARNS=2 clean.

19 years agoInclude <unistd.h> for {g,s}eteuid().
stefanf [Sat, 9 Oct 2004 15:36:13 +0000 (15:36 +0000)]
Include <unistd.h> for {g,s}eteuid().

19 years agoReturn 0, not NULL, from a function declared as returning int.
des [Sat, 9 Oct 2004 14:20:18 +0000 (14:20 +0000)]
Return 0, not NULL, from a function declared as returning int.

19 years agoForcibly create symlinked headers, otherwise the build process may fail
joerg [Sat, 9 Oct 2004 13:51:28 +0000 (13:51 +0000)]
Forcibly create symlinked headers, otherwise the build process may fail
if the target link already existed (e. g. -DNO_KERNELCLEAN).

19 years agoRemove inlined m_tag_free(). Rename _m_tag_free() to m_tag_free()
glebius [Sat, 9 Oct 2004 13:25:19 +0000 (13:25 +0000)]
Remove inlined m_tag_free(). Rename _m_tag_free() to m_tag_free()
and make it visible (same way as in OpenBSD). Describe usage in manpage.

This change is useful for creating custom free methods, which
call default free method at their end.

While here, make malloc declaration for mbuf tags more informative.

Approved by: julian (mentor), sam
MFC after: 1 month

19 years agostyle(9)
obrien [Sat, 9 Oct 2004 08:31:21 +0000 (08:31 +0000)]
style(9)

19 years agoDon't "implicitly order all sleep locks before spin locks" in witness
green [Sat, 9 Oct 2004 08:16:37 +0000 (08:16 +0000)]
Don't "implicitly order all sleep locks before spin locks" in witness
when the spin lock in question isn't -- it's the critical_enter() that
KDB set.  No more panic in DDB for console -> syscons -> tty -> knote
operations.

19 years agoDon't use matchlvl attach arg. It seems to be not initialized
takawata [Sat, 9 Oct 2004 07:48:31 +0000 (07:48 +0000)]
Don't use matchlvl attach arg. It seems to be not initialized
in FreeBSD probe mechanism.

19 years agoPort NetBSD auxio driver. The driver was modified to use led(4) and can
yongari [Sat, 9 Oct 2004 07:31:03 +0000 (07:31 +0000)]
Port NetBSD auxio driver. The driver was modified to use led(4) and can
be used to announce various system activity.
The auxio device provides auxiliary I/O functions and is found on various
SBus/EBus UltraSPARC models. At present, only front panel LED is
controlled by this driver.

Approved by:    jake (mentor)
Reviewed by:    joerg
Tested by:      joerg

19 years agoDon't count RNBC (internal buffer full) towards the RX error count since it's
scottl [Sat, 9 Oct 2004 07:27:03 +0000 (07:27 +0000)]
Don't count RNBC (internal buffer full) towards the RX error count since it's
not really an error.

Submitted by: Gerrit Nagelhout

19 years agoDelete a .Fn that had no argument and
keramida [Sat, 9 Oct 2004 04:03:08 +0000 (04:03 +0000)]
Delete a .Fn that had no argument and
properly terminate a .Bl with a matching .El

MFC after: 3 days

19 years ago3 important fixes for growfs:
scottl [Sat, 9 Oct 2004 02:53:47 +0000 (02:53 +0000)]
3 important fixes for growfs:

1) ginode() is passed a cylinder group number and inode number.  The inode
number is relative to the cg.  Use this relative number rather than the
absolute inode number when searching the cg inode bitmap to see if the inode
is allocated.  Using the absolute number quickly runs the check off the end
of the array and causes invalid inodes to be referenced.

2) ginode() checks the absolute indoe number to make sure that it is greater
than ROOTINO.  However, the caller loops through all of the possible inode
numbers and directly passes in values that are < ROOTINO.  Instead of halting
the program with an error, just return NULL.

3) When allocating new cylinder groups, growfs was initializing all of the
inodes in the group regardless of this only being required for UFS1.  Not
doing this for UFS2 provides a significant performance increase.

These fixes allow growing a filesystem beyond a trivial amount and have
been tested to grow an 8GB filesystem to 1.9TB.  Much more testing would
be appreciated.

Obtained from: Sandvine, Inc.

19 years agoif system scope thread didn't set timeout, don't call clock_gettime syscall
davidxu [Fri, 8 Oct 2004 22:57:30 +0000 (22:57 +0000)]
if system scope thread didn't set timeout, don't call clock_gettime syscall
before and after sleeping.

Reviewed by: deischen

19 years agoOnly do the geometry translations on ad* devices, other devices seems to
sos [Fri, 8 Oct 2004 21:27:27 +0000 (21:27 +0000)]
Only do the geometry translations on ad* devices, other devices seems to
have their own way of life.
Those other devices translations should be moved here as well.

19 years agoWhen send()ing to syslogd return ENOBUFS keep trying until success.
glebius [Fri, 8 Oct 2004 21:15:21 +0000 (21:15 +0000)]
When send()ing to syslogd return ENOBUFS keep trying until success.

This fixes a case, when DoSed syslogd completely loses messages.

PR: bin/72366
Discussed with: dwmalone, millert@OpenBSD.org
Approved by: julian (mentor)
Obtained from: OpenBSD (rev. 1.17, 1.21 by millert)
MFC after: 3 months

19 years agoUse a real (and compact) list for a filename list; this helps groff
keramida [Fri, 8 Oct 2004 20:44:49 +0000 (20:44 +0000)]
Use a real (and compact) list for a filename list; this helps groff
split the lines in more reasonable places too, both in tty/ps output.

MFC after: 3 days

19 years agoEliminate linked list used to track inodes with an initial link
truckman [Fri, 8 Oct 2004 20:44:47 +0000 (20:44 +0000)]
Eliminate linked list used to track inodes with an initial link
count of zero and instead encode this information in the inode state.
Pass 4 performed a linear search of this list for each inode in
the file system, which performs poorly if the list is long.

Reviewed by:    sam & keramida (an earlier version of the patch), mckusick
MFC after: 1 month

19 years agoMake less wrong for desciptions of signal handling
julian [Fri, 8 Oct 2004 20:40:30 +0000 (20:40 +0000)]
Make less wrong for desciptions of signal handling

MFC in: 1 week

19 years agoTo avoid pushing the paragraph text too far from the left border, making
keramida [Fri, 8 Oct 2004 20:31:33 +0000 (20:31 +0000)]
To avoid pushing the paragraph text too far from the left border, making
line-splitting extremely difficult for groff, indent the .Bl items by
the standard `indent' length instead of an indent large enough to hold
the maximal tag name.

Reviewed by: ru
MFC after: 3 days

19 years agoClose a .Bl list with an .El to fix a groff warning.
keramida [Fri, 8 Oct 2004 20:22:55 +0000 (20:22 +0000)]
Close a .Bl list with an .El to fix a groff warning.

19 years agoFix critical stability problems that can cause UMA mbuf cluster
green [Fri, 8 Oct 2004 20:19:29 +0000 (20:19 +0000)]
Fix critical stability problems that can cause UMA mbuf cluster
state management corruption, mbuf leaks, general mbuf corruption,
and at least on i386 a first level splash damage radius that
encompasses up to about half a megabyte of the memory after
an mbuf cluster's allocation slab.  In short, this has caused
instability nightmares anywhere the right kind of network traffic
is present.

When the polymorphic refcount slabs were added to UMA, the new types
were not used pervasively.  In particular, the slab management
structure was turned into one for refcounts, and one for non-refcounts
(supposed to be mostly like the old slab management structure),
but the latter was almost always used through out.  In general, every
access to zones with UMA_ZONE_REFCNT turned on corrupted the
"next free" slab offset offset and the refcount with each other and
with other allocations (on i386, 2 mbuf clusters per 4096 byte slab).

Fix things so that the right type is used to access refcounted zones
where it was not before.  There are additional errors in gross
overestimation of padding, it seems, that would cause a large kegs
(nee zones) to be allocated when small ones would do.  Unless I have
analyzed this incorrectly, it is not directly harmful.

19 years agoAdd a version of netsend that uses the interval timer rather than
rwatson [Fri, 8 Oct 2004 19:23:11 +0000 (19:23 +0000)]
Add a version of netsend that uses the interval timer rather than
explicit clock reads to set an overall duration to the send, and
blasts rather than trying to clock output.  The goal of netblast,
unlike netsend, is to send as many UDP packets as possible; the
cost is that there's no ability to control the rate, and there's
less accuracy in the timing as the interval timer granularity is
relatively low.

19 years agoUpdate a quirk for the ASUS P5A to disable the timer. It appears to work fine
njl [Fri, 8 Oct 2004 17:56:47 +0000 (17:56 +0000)]
Update a quirk for the ASUS P5A to disable the timer.  It appears to work fine
with acpi but the timer runs twice as fast.  Note that the main problem
(system doesn't work properly with acpi disabled) should be fixed separately.

Changes:
* Add a quirk to disable the timer
* Merge the P5A and P5A-B quirks since they appear to be based on the
  same ASL.

PR: i386/72450
Tested by: Kevin Oberman <oberman es.net>
MFC after: 3 days

19 years agoFix sis, bfe and ndis in the same way dc was fixed:
mlaier [Fri, 8 Oct 2004 16:14:42 +0000 (16:14 +0000)]
Fix sis, bfe and ndis in the same way dc was fixed:
  Do not tell the hardware to send when there were no packets enqueued.

Found and reviewed by: green
MFC after: 1 days

19 years agoNew release notes:
hrs [Fri, 8 Oct 2004 15:51:02 +0000 (15:51 +0000)]
New release notes:
SA-04:15, and BIND update (8.3.1->9.3.0).

Update release notes:
update a list of network interfaces which support polling(4), and
fix a typo (s/lukemftp/lukemftpd/).

19 years agoRemove a reference to a non-existent syscall: _thr_exit(). The
mtm [Fri, 8 Oct 2004 14:48:02 +0000 (14:48 +0000)]
Remove a reference to a non-existent syscall: _thr_exit(). The
actual name is thr_exit(). How this ever worked is beyond me.

19 years agoRemove an unused variable.
mtm [Fri, 8 Oct 2004 14:23:49 +0000 (14:23 +0000)]
Remove an unused variable.

Submitted by: Pawel Worach <pawel.worach@telia.com>

19 years agoFix an mdoc warning:
keramida [Fri, 8 Oct 2004 13:50:18 +0000 (13:50 +0000)]
Fix an mdoc warning:
Using a macro as first argument cancels effect of .Li (#2984)

MFC after: 3 days

19 years agoUse int format string, not a long format string.
rwatson [Fri, 8 Oct 2004 12:28:28 +0000 (12:28 +0000)]
Use int format string, not a long format string.

19 years agoChange pfil starvation prevention from fail-open to fail-close.
mlaier [Fri, 8 Oct 2004 12:07:20 +0000 (12:07 +0000)]
Change pfil starvation prevention from fail-open to fail-close.
We return ENOBUF to indicate the problem, which is an errno that should be
handled well everywhere.

Requested & Submitted by: green
Silently okay'ed by: The rest of the firewall gang
MFC after: 3 days

19 years agoOnly print progress statistics once per second.
phk [Fri, 8 Oct 2004 10:50:40 +0000 (10:50 +0000)]
Only print progress statistics once per second.

19 years ago- sort struct rtentry fields in man page in some order as they are in struct
glebius [Fri, 8 Oct 2004 09:57:12 +0000 (09:57 +0000)]
- sort struct rtentry fields in man page in some order as they are in struct
- remove RTF_PRCLONING
- add rt_mtx field
- rename rt_metrics -> rt_metrics_lite
- mention that only 3 metrics are really used in rt_metrics_lite

Reviewed by: ru, andre

19 years agoAdd more MLINKS for rtalloc.9
glebius [Fri, 8 Oct 2004 09:52:56 +0000 (09:52 +0000)]
Add more MLINKS for rtalloc.9

Approved by: ru

19 years agoMore grammar cleanups from MDocGuard^WRuslan.
glebius [Fri, 8 Oct 2004 09:52:00 +0000 (09:52 +0000)]
More grammar cleanups from MDocGuard^WRuslan.

19 years agoMake pte_load_store() an atomic operation in all cases, not just i386 PAE.
alc [Fri, 8 Oct 2004 08:23:43 +0000 (08:23 +0000)]
Make pte_load_store() an atomic operation in all cases, not just i386 PAE.

Restructure pmap_enter() to prevent the loss of a page modified (PG_M) bit
in a race between processors.  (This restructuring assumes the newly atomic
pte_load_store() for correct operation.)

Reviewed by: tegge@
PR: i386/61852

19 years agoUse generic tty processing code instead of local copy.
phk [Fri, 8 Oct 2004 06:45:10 +0000 (06:45 +0000)]
Use generic tty processing code instead of local copy.

New device names are {tty,cua}G$(adapter)$(port)[.lock,.init]

19 years agoMention the dc(4) driver's support, and ipfw(8)'s classifier support.
green [Fri, 8 Oct 2004 03:33:31 +0000 (03:33 +0000)]
Mention the dc(4) driver's support, and ipfw(8)'s classifier support.

19 years agoReference altq(4) instead of pf.conf(5).
green [Fri, 8 Oct 2004 03:31:09 +0000 (03:31 +0000)]
Reference altq(4) instead of pf.conf(5).

Tip of the hat to: mlaier

19 years agoCommit forgotten documentation for "diverted" rules.
green [Fri, 8 Oct 2004 03:24:10 +0000 (03:24 +0000)]
Commit forgotten documentation for "diverted" rules.

19 years agoSince net/net_osdep.c contained only one function that could be
brooks [Fri, 8 Oct 2004 00:24:30 +0000 (00:24 +0000)]
Since net/net_osdep.c contained only one function that could be
trivially implemented as a macro, do that and remove it.  NetBSD did
this quite a while ago.

19 years ago1. Incorporate most of Ruslan's improvements to where and how the
dougb [Fri, 8 Oct 2004 00:14:28 +0000 (00:14 +0000)]
1. Incorporate most of Ruslan's improvements to where and how the
/etc/namedb symlink is created.

2. Incorporate Brian's suggestion to make the link relative. This
is necessary to handle situations (such as mergemaster) where the
user is building a tree in a seperate environment. This will also
fix the problem with the way DESTDIR is set in 'make release'.

3. Add a new knob, NO_BIND_MTREE, as suggested by the folks who
already have stuff in /var/named that they don't want me to mess with.

4. Update make.conf(5) with the new stuff, and correct a few paths
that have changed since I last updated it.

19 years agoStyle. Use ETHER_IS_MULTICAST() appropriately instead of masking off the bit.
bms [Thu, 7 Oct 2004 20:56:29 +0000 (20:56 +0000)]
Style. Use ETHER_IS_MULTICAST() appropriately instead of masking off the bit.

Reviewed by: jmallett

19 years agoRemove obsolete references to PFIL_HOOKS and RANDOM_IP_ID
mlaier [Thu, 7 Oct 2004 20:49:52 +0000 (20:49 +0000)]
Remove obsolete references to PFIL_HOOKS and RANDOM_IP_ID

MFC after: 1 days

19 years agoBack out v1.58... We still don't know what is causing the specific
kensmith [Thu, 7 Oct 2004 20:36:56 +0000 (20:36 +0000)]
Back out v1.58...  We still don't know what is causing the specific
problem I had but it's happening in code that is messing around with
register windows - I'm willing to live with that piece being sensitive
to this and it looks like the other problems we had reported lately
are not fixed by using -O instead of -O2.

Sorry for the churn.  Looks like I need a second pointy hat.  Someone
tells me they stack well.  :-))))

19 years agoBe sure to always return 0 for negative access requests.
pjd [Thu, 7 Oct 2004 20:13:23 +0000 (20:13 +0000)]
Be sure to always return 0 for negative access requests.

Reported by: Maciej Kucharz <qk@comp.waw.pl>

19 years agoMove the PC98 specific geometry "gunk" to geom_pc98.c where it belongs.
sos [Thu, 7 Oct 2004 17:37:09 +0000 (17:37 +0000)]
Move the PC98 specific geometry "gunk" to geom_pc98.c where it belongs.
This also adds support for bigger disks on the controller I have access to,
and maybe others if I understood the adhoc methods used on those.

Those with more PC98 bigdrive controllers it is hereby invited to add/fix
support for those in geom_pc98.c and not using #ifdef PC98 all over the place.

19 years agoOoops. Forgot the man page for pbio.4 in my port from RELENG_4.
imp [Thu, 7 Oct 2004 16:25:21 +0000 (16:25 +0000)]
Ooops.  Forgot the man page for pbio.4 in my port from RELENG_4.

19 years agoPort pbio to HEAD.
imp [Thu, 7 Oct 2004 16:21:03 +0000 (16:21 +0000)]
Port pbio to HEAD.

OK'd by: dds

19 years agoRename the QUEUEING section to QUEUEING/ALTQ to make it easier to find the
mlaier [Thu, 7 Oct 2004 15:39:02 +0000 (15:39 +0000)]
Rename the QUEUEING section to QUEUEING/ALTQ to make it easier to find the
appropiate section when redirected from ALTQ(4).

MFC after: 2 days

19 years agoForced commit:
mtm [Thu, 7 Oct 2004 14:51:23 +0000 (14:51 +0000)]
Forced commit:

Remove references to the FreeBSD keyword.
Make a note in UPDATING about the recent change regarding
the keyword.

MFC after: 3 days

19 years agoUse PTHREAD_SCOPE_SYSTEM to decide what should be done.
davidxu [Thu, 7 Oct 2004 14:23:15 +0000 (14:23 +0000)]
Use PTHREAD_SCOPE_SYSTEM to decide what should be done.

19 years agoWhen running with debug.mpsafenet=0, initialize IP multicast routing
rwatson [Thu, 7 Oct 2004 14:13:35 +0000 (14:13 +0000)]
When running with debug.mpsafenet=0, initialize IP multicast routing
callouts as non-CALLOUT_MPSAFE.  Otherwise, they may trigger an
assertion regarding Giant if they enter other parts of the stack from
the callout.

MFC after: 3 days
Reported by: Dikshie < dikshie at ppk dot itb dot ac dot id >

19 years agoF\98
mtm [Thu, 7 Oct 2004 14:06:23 +0000 (14:06 +0000)]
F\98

19 years agoRemove the requirement for the FreeBSD keyword as it no longer
mtm [Thu, 7 Oct 2004 13:55:26 +0000 (13:55 +0000)]
Remove the requirement for the FreeBSD keyword as it no longer
makes any sense.

Discussed with: dougb, brooks
MFC after: 3 days

19 years agoFollow kernel change, restore signal mask correctly by using a command
davidxu [Thu, 7 Oct 2004 13:52:18 +0000 (13:52 +0000)]
Follow kernel change, restore signal mask correctly by using a command
of kse_thr_interrupt.

19 years agoAdd an execve command for kse_thr_interrupt to allow libpthread to
davidxu [Thu, 7 Oct 2004 13:50:10 +0000 (13:50 +0000)]
Add an execve command for kse_thr_interrupt to allow libpthread to
restore signal mask correctly, this is required by POSIX.

Reviewed by: deischen

19 years agoAdd a minimal altq.4 manpage to tell about the kernel options and where to
mlaier [Thu, 7 Oct 2004 12:10:25 +0000 (12:10 +0000)]
Add a minimal altq.4 manpage to tell about the kernel options and where to
find more information. Also move the "SUPPORTED DEVICES" section from altq.9
to altq.4, where is belongs.

Requested by: green
MFC after: 2 days

19 years agoAdd SHARP to the pool of drives that doesn not need byteswapping of
sos [Thu, 7 Oct 2004 11:43:25 +0000 (11:43 +0000)]
Add SHARP to the pool of drives that doesn not need byteswapping of
the model etc fields from identify.

19 years agoMdoc and grammar fixes.
glebius [Thu, 7 Oct 2004 10:59:09 +0000 (10:59 +0000)]
Mdoc and grammar fixes.

Submitted by: ru

19 years ago- note that RTF_PRCLONING is obsoleted
glebius [Thu, 7 Oct 2004 10:14:23 +0000 (10:14 +0000)]
- note that RTF_PRCLONING is obsoleted
- explain locking differencies between rtalloc(), rtalloc_ign(), rtalloc1()
  and rtfree(), RTFREE().

Reviewed by: andre (1.5 months ago)
OK'd by: ru

19 years ago- Be more userfriendly and allow to specify gbde device name in those forms:
pjd [Thu, 7 Oct 2004 10:02:46 +0000 (10:02 +0000)]
- Be more userfriendly and allow to specify gbde device name in those forms:
device
device.bde
/dev/device
/dev/device.bde
- Fix stop routine:
+ There don't have to be file system mounted on gbde device,
  so ignore errors from umount(8).
+ Only detach existing gbde devices.

19 years agoAdd support for Pentium M, Pentium 3M and Pentium 4M.
des [Thu, 7 Oct 2004 09:56:48 +0000 (09:56 +0000)]
Add support for Pentium M, Pentium 3M and Pentium 4M.

PR: i386/72340
Submitted by: Rong-En Fan <rafan@infor.org>
MFC after: 2 weeks

19 years agoUse generic ttycode instead of local copy.
phk [Thu, 7 Oct 2004 06:19:11 +0000 (06:19 +0000)]
Use generic ttycode instead of local copy.

19 years agoOnly try to attach if parent device actually exists.
pjd [Thu, 7 Oct 2004 06:00:06 +0000 (06:00 +0000)]
Only try to attach if parent device actually exists.
I used ugly "/dev/${parent}" instead of "${parentdev}", because "/dev/"
prefix for devices listed in gbde_devices variable is optional.

Reported by: Sean McNeil <sean@mcneil.com>

19 years agoNote libthr ABI breakage.
mtm [Thu, 7 Oct 2004 05:45:39 +0000 (05:45 +0000)]
Note libthr ABI breakage.

19 years agoRegen to unbreak world.
davidxu [Thu, 7 Oct 2004 01:09:46 +0000 (01:09 +0000)]
Regen to unbreak world.

Pointy hat to: mtm

19 years agoBack out rev 1.240; it is unnecessary. In particular,
das [Wed, 6 Oct 2004 23:53:49 +0000 (23:53 +0000)]
Back out rev 1.240; it is unnecessary.  In particular,
p1 == curthread, so _PHOLD(p1) will not have to block
to swap in p1.

Noticed by: jhb

19 years agoDon't set the BIO_ONQUEUE debugging flag until we actually put the bio
phk [Wed, 6 Oct 2004 20:59:59 +0000 (20:59 +0000)]
Don't set the BIO_ONQUEUE debugging flag until we actually put the bio
onto a queue.  This made the ENOMEM handling an instant panic.

19 years agoUse generic tty code instead of local copies.
phk [Wed, 6 Oct 2004 20:01:49 +0000 (20:01 +0000)]
Use generic tty code instead of local copies.

19 years agoBack out v1.49. Recent findings suggest sparc64 may not be ready for
kensmith [Wed, 6 Oct 2004 19:55:14 +0000 (19:55 +0000)]
Back out v1.49.  Recent findings suggest sparc64 may not be ready for
-O2 on kernel compiles after all.  While working on adding a KASSERT
to sparc64/sparc64/rwindow.c I found that it was "position sensitive",
putting it above a call to flushw() instead of below caused corruption
of processes on the system.  jake and jhb have both confirmed there is
no obvious explanation for that.  The exact same kernel code does not
have the process corruption problem if compiled with -O instead of -O2.
There have been signs of similar issues floated on the sparc64@ mailing
list, lets see if this helps make them go away.

Note this isn't an optimal fix as far as the file format goes, if this
disgusts too many people I'll fix it the right way.  Since compiling
with something other than -O is a known problem this format would prevent
a change to the default causing grief.  And this may also help motivate
finding out what the compiler is doing wrong so we can shift back to
using -O2. :-)

My turn for the pointy hat...  One of the florescent ones...

MFC after: 2 days

19 years agoFix the PC98 lockups on boot.
sos [Wed, 6 Oct 2004 19:46:08 +0000 (19:46 +0000)]
Fix the PC98 lockups on boot.
The interchannel locking for PC98 needed to be updated to match the
rest of the locking in ATA.

19 years agoGeoms without softc are geoms which are initialized, so wait for them.
pjd [Wed, 6 Oct 2004 18:47:15 +0000 (18:47 +0000)]
Geoms without softc are geoms which are initialized, so wait for them.

19 years agoCorrect grammos. Added BUGS: may mishandle nonstandard manual pages,
schweikh [Wed, 6 Oct 2004 18:31:45 +0000 (18:31 +0000)]
Correct grammos. Added BUGS: may mishandle nonstandard manual pages,
especially third party junk^Wpages, like emulators/mtools(1) and friends.

19 years ago- Fix the compile to chase the p_rux changes.
jhb [Wed, 6 Oct 2004 17:10:56 +0000 (17:10 +0000)]
- Fix the compile to chase the p_rux changes.
- Add a comment noting that the ru_[us]times values being read aren't
  actually valid and need to be computed from the raw values.

Submitted by: many (1)

19 years agoAdd missing 'static'
imp [Wed, 6 Oct 2004 15:18:12 +0000 (15:18 +0000)]
Add missing 'static'

19 years agoLock file for gbde devices is optional.
pjd [Wed, 6 Oct 2004 14:42:35 +0000 (14:42 +0000)]
Lock file for gbde devices is optional.

19 years agoClose a race between a thread exiting and the freeing of it's stack.
mtm [Wed, 6 Oct 2004 14:23:00 +0000 (14:23 +0000)]
Close a race between a thread exiting and the freeing of it's stack.
After some discussion the best option seems to be to signal the thread's
death from within the kernel. This requires that thr_exit() take an
argument.

Discussed with: davidxu, deischen, marcel
MFC after: 3 days

19 years agoLook out for geoms without softc.
pjd [Wed, 6 Oct 2004 14:15:47 +0000 (14:15 +0000)]
Look out for geoms without softc.

Reported by: tegge

19 years agoAllocate red zone and stack space together and then split red zone from
davidxu [Wed, 6 Oct 2004 08:11:07 +0000 (08:11 +0000)]
Allocate red zone and stack space together and then split red zone from
allocated space, orignal code left red zone unallocated, but those space
can be allocated by user code, and result was providing no protection.

19 years agoFor older systems with ACPI which don't have a pci <-> pci bridge,
imp [Wed, 6 Oct 2004 07:26:52 +0000 (07:26 +0000)]
For older systems with ACPI which don't have a pci <-> pci bridge,
allocate unallocated memory resources from the top 32MB of the address
space rather than the top 2GB.  While the latter works on some
chipsets, it fails badly on others.  32MB is more conservative and
matches what cheap harware from this era is hardwired to pass.

19 years agoFor legacy PCI bridges, limit memory allocation to the top 32MB of
imp [Wed, 6 Oct 2004 07:22:58 +0000 (07:22 +0000)]
For legacy PCI bridges, limit memory allocation to the top 32MB of
RAM.  Many older, legacy bridges only allow allocation from this
range.  This only appies to devices who don't have their memory
assigned by the BIOS (since we allocate the ranges so assigned
exactly), so should have minimal impact.

Hoewver, for CardBus bridges (cbb), they rarely get the resources
allocated by the BIOS, and this patch helps them greatly.  Typically
the 'bad Vcc' messages are caused by this problem.

19 years agoMention "-d directory" in usage().
brooks [Wed, 6 Oct 2004 04:47:37 +0000 (04:47 +0000)]
Mention "-d directory" in usage().

Pointy hat to: brooks

19 years agoDon't recurse the BPF descriptor lock during the BIOCSDLT operation
green [Wed, 6 Oct 2004 04:25:37 +0000 (04:25 +0000)]
Don't recurse the BPF descriptor lock during the BIOCSDLT operation
(and panic).  To try to finish making BPF safe, at the very least,
the BPF descriptor lock really needs to change into a reader/writer
lock that controls access to "settings," and a mutex that controls
access to the selinfo/knote/callout.  Also, use of callout_drain()
instead of callout_stop() (which is really a much more widespread
issue).

19 years agofixed too delayed routing cache expiry. (tvtohz() converts a time interval to ticks...
suz [Wed, 6 Oct 2004 03:32:26 +0000 (03:32 +0000)]
fixed too delayed routing cache expiry.  (tvtohz() converts a time interval to ticks, whereas hzto() converts an absolute time to ticks)

Obtained from: KAME

19 years agoAdd the Madison II, which is the second generation Madison. The Madison II
marcel [Wed, 6 Oct 2004 02:43:28 +0000 (02:43 +0000)]
Add the Madison II, which is the second generation Madison. The Madison II
is model 2 in the Itanium 2 family and has up to 9MB of L3 cache and clocks
higher than 1.5Ghz. There's no LV variant AFAICT.