]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoAccept filters broke kernels compiled without options INET.
alfred [Mon, 20 Nov 2000 01:35:25 +0000 (01:35 +0000)]
Accept filters broke kernels compiled without options INET.
Make accept filters conditional on INET support to fix.

Pointed out by: bde
Tested and assisted by: Stephen J. Kiernan <sab@vegamuse.org>

23 years agoo Export cp_time ("CPU time statistics") using SYSCTL_OPAQUE.
rwatson [Mon, 20 Nov 2000 00:44:58 +0000 (00:44 +0000)]
o Export cp_time ("CPU time statistics") using SYSCTL_OPAQUE.
  This removes a reason that systat requires setgid kmem.  More to
  come.

23 years agoo Export nchstats ("VFS cache effectiveness statistics") using
rwatson [Mon, 20 Nov 2000 00:41:11 +0000 (00:41 +0000)]
o Export nchstats ("VFS cache effectiveness statistics") using
  SYSCTL_OPAQUE.  This removes a reason that systat requires
  setgid kmem.  More to come.

23 years agoo Export dmmax ("Maximum size of a swap block") using SYSCTL_INT.
rwatson [Mon, 20 Nov 2000 00:39:04 +0000 (00:39 +0000)]
o Export dmmax ("Maximum size of a swap block") using SYSCTL_INT.
  This removes a reason that systat requires setgid kmem.  More to
  come.

23 years agoDisable calling timeout() when doing bio. It was causing more prolems than
scottl [Sun, 19 Nov 2000 23:46:21 +0000 (23:46 +0000)]
Disable calling timeout() when doing bio.  It was causing more prolems than
solving.  This will be fixed for real soon.

23 years agoMake sbcompress use the new M_WRITABLE macro. Previously sbcompress
dwmalone [Sun, 19 Nov 2000 22:22:47 +0000 (22:22 +0000)]
Make sbcompress use the new M_WRITABLE macro. Previously sbcompress
could not compress into clusters. This could result in lots of
wasted clusters while recieving small packets from an interface
that uses clusters for all it's packets.

Patch is partially from BSDi (limiting the size of the copy) and
based on a patch for 4.1 by Ian Dowse <iedowse@maths.tcd.ie> and
myself.

Reviewed by: bmilekic
Obtained From: BSDi
Submitted by: iedowse

23 years agoStandardize spelling of "optimization" in comments
dougb [Sun, 19 Nov 2000 20:12:06 +0000 (20:12 +0000)]
Standardize spelling of "optimization" in comments

PR: conf/22622
Submitted by: knu

23 years agoWhen to run the periodic/daily event has had several rounds of discussion
dougb [Sun, 19 Nov 2000 18:16:46 +0000 (18:16 +0000)]
When to run the periodic/daily event has had several rounds of discussion
over the past couple years. The most recent came to the general consensus
that this was the best time, but no one actually made the change, so I'll
don my asbestos undies and dive in.

Please note that this time was chosen with input from people in various
countries with various methods and schedules for switching to and from DST.
There is no perfect time to schedule this job that works for everyone, but
this time both A) Works for more people, and B) Causes problems for fewer
people. And, ultimately, you can always change it if you need to.

23 years agoSome gratuitous whitespace changes.
brian [Sun, 19 Nov 2000 16:58:59 +0000 (16:58 +0000)]
Some gratuitous whitespace changes.

23 years agoDescribe how to tunnel properly.
brian [Sun, 19 Nov 2000 16:54:57 +0000 (16:54 +0000)]
Describe how to tunnel properly.

Requested by: Greg Quinlan <greg@macquarrie.com.au>

23 years agoAccept MSCHAPv2 by default.
brian [Sun, 19 Nov 2000 16:53:50 +0000 (16:53 +0000)]
Accept MSCHAPv2 by default.
Enable and accept MPPE by default.

23 years agoFormat string paranoia
kris [Sun, 19 Nov 2000 13:30:36 +0000 (13:30 +0000)]
Format string paranoia

23 years agoFormat string error in warn()
kris [Sun, 19 Nov 2000 13:29:43 +0000 (13:29 +0000)]
Format string error in warn()

23 years agoUse mktemp -t to respect TMPDIR
kris [Sun, 19 Nov 2000 13:10:11 +0000 (13:10 +0000)]
Use mktemp -t to respect TMPDIR
Clean up temporary file at runtime

23 years agoCorrect typo
kris [Sun, 19 Nov 2000 13:09:36 +0000 (13:09 +0000)]
Correct typo

23 years agoFormat string paranoia
kris [Sun, 19 Nov 2000 12:46:16 +0000 (12:46 +0000)]
Format string paranoia

23 years agoConvert various calls to splhigh() to disable_intr() since splhigh() is
dfr [Sun, 19 Nov 2000 12:28:42 +0000 (12:28 +0000)]
Convert various calls to splhigh() to disable_intr() since splhigh() is
now a no-op.

23 years agoWe don't need <stddef.h> for offsetof() any more.
dfr [Sun, 19 Nov 2000 12:26:14 +0000 (12:26 +0000)]
We don't need <stddef.h> for offsetof() any more.

23 years agoUnlink the temporary file immediately so it is removed on exit.
kris [Sun, 19 Nov 2000 12:04:12 +0000 (12:04 +0000)]
Unlink the temporary file immediately so it is removed on exit.

Obtained from: OpenBSD

23 years agoMAXPATHNAMELEN -> MAXPATHLEN
kris [Sun, 19 Nov 2000 11:29:58 +0000 (11:29 +0000)]
MAXPATHNAMELEN -> MAXPATHLEN

Submitted by: ianm@kashmir.cit.nepean.uws.edu.au via OpenBSD
Obtained from: OpenBSD

23 years agoL_SET -> SEEK_SET
kris [Sun, 19 Nov 2000 10:56:14 +0000 (10:56 +0000)]
L_SET -> SEEK_SET

Obtained from: OpenBSD

23 years agoFormat string paranoia
kris [Sun, 19 Nov 2000 10:52:10 +0000 (10:52 +0000)]
Format string paranoia

Obtained from: OpenBSD

23 years agoDon't suggest people use getpid() to construct temporary filenames;
kris [Sun, 19 Nov 2000 10:30:42 +0000 (10:30 +0000)]
Don't suggest people use getpid() to construct temporary filenames;
point them to mkstemp() instead.

Obtained from: OpenBSD (based on)

23 years agoUpdate the documentation to describe the new mktemp() family behaviour.
kris [Sun, 19 Nov 2000 10:21:32 +0000 (10:21 +0000)]
Update the documentation to describe the new mktemp() family behaviour.
Also notes that mkstemp() first appeared in 4.4BSD (change obtained
from OpenBSD)

Reviewed by: sheldonh

23 years agoFix a buffer overflow from a long local hostname.
kris [Sun, 19 Nov 2000 10:08:26 +0000 (10:08 +0000)]
Fix a buffer overflow from a long local hostname.

Obtained from: OpenBSD

23 years agoDon't use sizeof() on a pointer when we really wanted to measure
kris [Sun, 19 Nov 2000 10:01:27 +0000 (10:01 +0000)]
Don't use sizeof() on a pointer when we really wanted to measure
the length of the array.

Noticed by: Christos Zoulas <christos@ZOULAS.COM>
Obtained from: OpenBSD

23 years ago- Protect the callout wheel with a separate spin mutex, callout_lock.
jake [Sun, 19 Nov 2000 06:02:32 +0000 (06:02 +0000)]
- Protect the callout wheel with a separate spin mutex, callout_lock.
- Use the mutex in hardclock to ensure no races between it and
  softclock.
- Make softclock be INTR_MPSAFE and provide a flag,
  CALLOUT_MPSAFE, which specifies that a callout handler does not
  need giant.  There is still no way to set this flag when
  regstering a callout.

Reviewed by: -smp@, jlemon

23 years agoCheck return code from login_tty. Allow getty to try and become
jwd [Sun, 19 Nov 2000 02:10:25 +0000 (02:10 +0000)]
Check return code from login_tty. Allow getty to try and become
a daemon and session leader (thus allowing getty to be run from
a shell command line or script).

Partially Reviewed by: bde

23 years agoAdd/fix the ability to split a file larger than 2 gigabytes.
jwd [Sun, 19 Nov 2000 01:44:20 +0000 (01:44 +0000)]
Add/fix the ability to split a file larger than 2 gigabytes.

Added $FreeBSD tag (in the way the already present sccsid is
done). I've been told the rcsid stuff may be of dubious value
so I'm curious to know if folks (still) use it.

Submitted by: Brad Chisholm <blc@bsdwins.com>

23 years agoImplement a low-memory deadlock solution.
dillon [Sat, 18 Nov 2000 23:06:26 +0000 (23:06 +0000)]
Implement a low-memory deadlock solution.

    Removed most of the hacks that were trying to deal with low-memory
    situations prior to now.

    The new code is based on the concept that I/O must be able to function in
    a low memory situation.  All major modules related to I/O (except
    networking) have been adjusted to allow allocation out of the system
    reserve memory pool.  These modules now detect a low memory situation but
    rather then block they instead continue to operate, then return resources
    to the memory pool instead of cache them or leave them wired.

    Code has been added to stall in a low-memory situation prior to a vnode
    being locked.

    Thus situations where a process blocks in a low-memory condition while
    holding a locked vnode have been reduced to near nothing.  Not only will
    I/O continue to operate, but many prior deadlock conditions simply no
    longer exist.

Implement a number of VFS/BIO fixes

(found by Ian): in biodone(), bogus-page replacement code, the loop
        was not properly incrementing loop variables prior to a continue
        statement.  We do not believe this code can be hit anyway but we
        aren't taking any chances.  We'll turn the whole section into a
        panic (as it already is in brelse()) after the release is rolled.

In biodone(), the foff calculation was incorrectly
        clamped to the iosize, causing the wrong foff to be calculated
        for pages in the case of an I/O error or biodone() called without
        initiating I/O.  The problem always caused a panic before.  Now it
        doesn't.  The problem is mainly an issue with NFS.

Fixed casts for ~PAGE_MASK.  This code worked properly before only
        because the calculations use signed arithmatic.  Better to properly
        extend PAGE_MASK first before inverting it for the 64 bit masking
        op.

In brelse(), the bogus_page fixup code was improperly throwing
        away the original contents of 'm' when it did the j-loop to
        fix the bogus pages.  The result was that it would potentially
        invalidate parts of the *WRONG* page(!), leading to corruption.

There may still be cases where a background bitmap write is
        being duplicated, causing potential corruption.  We have identified
        a potentially serious bug related to this but the fix is still TBD.
        So instead this patch contains a KASSERT to detect the problem
   and panic the machine rather then continue to corrupt the filesystem.
The problem does not occur very often..  it is very hard to
reproduce, and it may or may not be the cause of the corruption
people have reported.

Review by: (VFS/BIO: mckusick, Ian Dowse <iedowse@maths.tcd.ie>)
Testing by: (VM/Deadlock) Paul Saab <ps@yahoo-inc.com>

23 years agoAdd the splvm()'s suggested in PR 20609 to protect vm_pager_page_unswapped().
dillon [Sat, 18 Nov 2000 21:11:23 +0000 (21:11 +0000)]
Add the splvm()'s suggested in PR 20609 to protect vm_pager_page_unswapped().
The remainder of the PR is still open.

PR: kern/20609 (partial fix)

23 years ago This patchset fixes a large number of file descriptor race conditions.
dillon [Sat, 18 Nov 2000 21:01:04 +0000 (21:01 +0000)]
This patchset fixes a large number of file descriptor race conditions.
    Pre-rfork code assumed inherent locking of a process's file descriptor
    array.  However, with the advent of rfork() the file descriptor table
    could be shared between processes.  This patch closes over a dozen
    serious race conditions related to one thread manipulating the table
    (e.g. closing or dup()ing a descriptor) while another is blocked in
    an open(), close(), fcntl(), read(), write(), etc...

PR: kern/11629
Discussed with: Alexander Viro <viro@math.psu.edu>

23 years agoFix typo.
ben [Sat, 18 Nov 2000 20:19:26 +0000 (20:19 +0000)]
Fix typo.

PR: 22923
Submitted by: Jimmy Olgeni <olgeni@uli.it>

23 years agoFix typos.
ben [Sat, 18 Nov 2000 20:18:34 +0000 (20:18 +0000)]
Fix typos.

PR: 22922
Submitted by: Jimmy Olgeni <olgeni@uli.it>

23 years agoFix another typo.
ben [Sat, 18 Nov 2000 20:16:47 +0000 (20:16 +0000)]
Fix another typo.

PR: 22924
Submitted by: Jimmy Olgeni <olgeni@uli.it>

23 years agoWhile in the SYNOPSIS section, Nm macro always required an
ru [Sat, 18 Nov 2000 17:34:55 +0000 (17:34 +0000)]
While in the SYNOPSIS section, Nm macro always required an
argument to setup indentation for the rest of the synopsis
line and to automatically emit a line break for a non-first
synopsis line.  It does not require it anymore.

The following now works as expected:

.Sh NAME
.Nm utility
.Sh SYNOPSIS
.Nm
.Cm start
.Nm
.Cm stop

23 years agomdoc(7) police: fix errors uncovered by the new feature of the Nm macro.
ru [Sat, 18 Nov 2000 15:50:46 +0000 (15:50 +0000)]
mdoc(7) police: fix errors uncovered by the new feature of the Nm macro.

23 years agoFurther use of M_ZERO.
dwmalone [Sat, 18 Nov 2000 15:21:22 +0000 (15:21 +0000)]
Further use of M_ZERO.

Submitted by: josh@zipperup.org
Submitted by: Robert Drehmel <robd@gmx.net>
Approved by: msmith

23 years agoAdd the use of M_ZERO to netgraph.
dwmalone [Sat, 18 Nov 2000 15:17:43 +0000 (15:17 +0000)]
Add the use of M_ZERO to netgraph.

Submitted by: josh@zipperup.org
Submitted by: Robert Drehmel <robd@gmx.net>
Submitted by: archie
Approved by: archie

23 years agomdoc(7) police: now that Nm macro accepts punctuation characters
ru [Sat, 18 Nov 2000 14:44:30 +0000 (14:44 +0000)]
mdoc(7) police: now that Nm macro accepts punctuation characters
as argument, unbreak this page by escaping the `[' character.

Noticed by: sheldonh

23 years agoFix a braino ..
sos [Sat, 18 Nov 2000 12:14:35 +0000 (12:14 +0000)]
Fix a braino ..

23 years agodo not blindly assume 8khz is supported on open(). try for 8khz but respect
cg [Sat, 18 Nov 2000 03:43:04 +0000 (03:43 +0000)]
do not blindly assume 8khz is supported on open().  try for 8khz but respect
minspeed/maxspeed specified by the hw driver.

Submitted by: Andrew Gordon <arg@arg1.demon.co.uk>

23 years agoMake the order of values prompted for with the "-i" option match print_part()
obrien [Sat, 18 Nov 2000 02:55:43 +0000 (02:55 +0000)]
Make the order of values prompted for with the "-i" option match print_part()

23 years agoUse vop_defaultop() instead of ntfs_bypass().
bp [Sat, 18 Nov 2000 02:47:12 +0000 (02:47 +0000)]
Use vop_defaultop() instead of ntfs_bypass().

PR: kern/22756

23 years agoRelease sched_lock very briefly to give interrupts a chance to fire if we
jhb [Sat, 18 Nov 2000 00:21:00 +0000 (00:21 +0000)]
Release sched_lock very briefly to give interrupts a chance to fire if we
are in softclock() for a long time.  The old code already did an
splx()/slphigh() pair here, I just missed adding in the equivalent mutex
operations on sched_lock earlier.

23 years agoDon't attempt to cluster write buffers where the VMIO flag isn't set.
tegge [Fri, 17 Nov 2000 23:40:08 +0000 (23:40 +0000)]
Don't attempt to cluster write buffers where the VMIO flag isn't set.

23 years agoMake sure we don't cross stripe boundaries when reviving striped plexes.
des [Fri, 17 Nov 2000 23:40:01 +0000 (23:40 +0000)]
Make sure we don't cross stripe boundaries when reviving striped plexes.
This makes crash recovery work for stripe sizes that are not multiples of
DEFAULT_REVIVE_BLOCKSIZE (currently 64 kB).
While we're here, fix a few cosmetic nits.

Reviewed by: grog
Sponsored by: Enitel ASA (http://www.enitel.no/)

23 years ago"minute(s) period(s)" --> "minute(s)"
mph [Fri, 17 Nov 2000 23:08:03 +0000 (23:08 +0000)]
"minute(s) period(s)" --> "minute(s)"

23 years agoFix the `make -jX' (X>1) breakage.
obrien [Fri, 17 Nov 2000 21:25:15 +0000 (21:25 +0000)]
Fix the `make -jX' (X>1) breakage.

Based on patch submitted by: Makoto MATSUSHITA <matusita@jp.freebsd.org>
Reviewed by: marcel, bde

23 years ago- Split the run queue and sleep queue linkage, so that a process
jake [Fri, 17 Nov 2000 18:09:18 +0000 (18:09 +0000)]
- Split the run queue and sleep queue linkage, so that a process
may block on a mutex while on the sleep queue without corrupting
it.
- Move dropping of Giant to after the acquire of sched_lock.

Tested by: John Hay <jhay@icomtek.csir.co.za>
jhb

23 years agoReference security advisories FreeBSD-SA-00:{68,69,70}.
bmah [Fri, 17 Nov 2000 17:48:40 +0000 (17:48 +0000)]
Reference security advisories FreeBSD-SA-00:{68,69,70}.

23 years ago- Change extra sanity checks in cpu_switch() to be conditional on INVARIANTS
jhb [Fri, 17 Nov 2000 17:37:43 +0000 (17:37 +0000)]
- Change extra sanity checks in cpu_switch() to be conditional on INVARIANTS
  instead of DIAGNOSTIC.
- Remove the p_wchan check as it no longer applies since a process may be
  switched out during CURSIG() within msleep() or mawait().
- Remove an extra sanity check only needed during the early SMPng work.

23 years agomdoc(7) police: Nm is a must in the NAME section.
ru [Fri, 17 Nov 2000 17:11:26 +0000 (17:11 +0000)]
mdoc(7) police: Nm is a must in the NAME section.

23 years agoReturn to the stock version of this file as local FreeBSD mods
ru [Fri, 17 Nov 2000 16:30:34 +0000 (16:30 +0000)]
Return to the stock version of this file as local FreeBSD mods
to it have been made in mdoc(7) format, but now it is again in
man(7) format, and the vendor apparently incorporated our mods.

23 years agomdoc(7) police: eliminate warnings.
ru [Fri, 17 Nov 2000 15:40:57 +0000 (15:40 +0000)]
mdoc(7) police: eliminate warnings.

23 years agoFixed Nm macro so that it accepts punctuation characters as argument.
ru [Fri, 17 Nov 2000 15:30:41 +0000 (15:30 +0000)]
Fixed Nm macro so that it accepts punctuation characters as argument.
For example, neither of the following worked before:

     .Nm ,       name,
     .Nm .       name.
     .Nm :       name:
     .Dq Nm      ``name''
     .Op Nm .    [name].

23 years agomdoc(7) police: use certified section headers wherever possible.
ru [Fri, 17 Nov 2000 11:44:16 +0000 (11:44 +0000)]
mdoc(7) police: use certified section headers wherever possible.

23 years agoFix typos.
ben [Fri, 17 Nov 2000 09:16:29 +0000 (09:16 +0000)]
Fix typos.

PR: 22901, 22902
Submitted by: Jimmy Olgeni <olgeni@uli.it>

23 years agoThe default kernel filename is "kernel" again, not "kernel.ko".
msmith [Fri, 17 Nov 2000 04:43:56 +0000 (04:43 +0000)]
The default kernel filename is "kernel" again, not "kernel.ko".

Submitted by: mckusick

23 years agoAdd the 'gdt' and 'gdtd' devices for the ICP Vortex RAID controller family.
msmith [Fri, 17 Nov 2000 01:36:34 +0000 (01:36 +0000)]
Add the 'gdt' and 'gdtd' devices for the ICP Vortex RAID controller family.

23 years agoGo back to populating data_len in struct ngpppoe_init_data.
brian [Thu, 16 Nov 2000 23:15:42 +0000 (23:15 +0000)]
Go back to populating data_len in struct ngpppoe_init_data.

23 years agoGo back to using data_len in struct ngpppoe_init_data after discussions
brian [Thu, 16 Nov 2000 23:14:53 +0000 (23:14 +0000)]
Go back to using data_len in struct ngpppoe_init_data after discussions
with Julian and Archie.

Implement a new ``sizedstring'' parse type for dealing with field pairs
consisting of a uint16_t followed by a data field of that size, and use
this to deal with the data_len and data fields.

Written by: Archie with some input by me
Agreed in principle by: julian

23 years agoFix a bug where a statically initialized condition variable
deischen [Thu, 16 Nov 2000 22:50:33 +0000 (22:50 +0000)]
Fix a bug where a statically initialized condition variable
was not getting properly initialized in pthread_cond_signal()
and pthread_cond_broadcast().  Reportedly, this can cause
an application to die.

MFC candidate

Submitted by: ade

23 years agoBe a bit more precise about what ``nat deny_incoming yes'' does.
brian [Thu, 16 Nov 2000 21:50:50 +0000 (21:50 +0000)]
Be a bit more precise about what ``nat deny_incoming yes'' does.

23 years agoThe recent changes to msleep() and mawait() resulted in timeout() and
jhb [Thu, 16 Nov 2000 21:20:52 +0000 (21:20 +0000)]
The recent changes to msleep() and mawait() resulted in timeout() and
untimeout() not being called with Giant in those functions.  For now,
use the sched_lock to protect the callout wheel in softclock() and in
the various timeout and callout functions.

Noticed by: tegge

23 years agoFix a couple of typos
marko [Thu, 16 Nov 2000 20:18:45 +0000 (20:18 +0000)]
Fix a couple of typos

Approved by: Warner

23 years agoWhen checking the device code in the probe routine, leave the chip in
wpaul [Thu, 16 Nov 2000 19:56:09 +0000 (19:56 +0000)]
When checking the device code in the probe routine, leave the chip in
16-bit mode. Technically, pcn_probe() is destructive because once the
chip goes into 32-bit mode, the only way to get it out again is a
hardware reset. And once the device is in 32-bit mode, the lnc driver
won't be able to talk to it. So if pcn_probe() is called before the
lnc probe routine, and pcn_probe() rejects the chip as one it doesn't
support, the lnc driver will be SOL.

I don't like this. I think it's a design flaw that you can't switch
the chip out of 32-bit mode once it's selected. The only 'right'
solution is for the pcn driver to support all of the PCI devices
in 32-bit mode, however I don't have samples of all the PCnet series
cards for testing.

23 years agoDelete 4 lines of misleading/incorrect comments.
deischen [Thu, 16 Nov 2000 19:15:56 +0000 (19:15 +0000)]
Delete 4 lines of misleading/incorrect comments.

23 years agoMake the question mark (`?') a valid punctuation character.
ru [Thu, 16 Nov 2000 17:04:14 +0000 (17:04 +0000)]
Make the question mark (`?') a valid punctuation character.
One can now, for example, write:

.Sh EXAMPLES
Have you read the
.Xr mdoc 7 ?
.Pp
Do you like
.Fx ?

The mdoc.samples(7) page has been automatically updated to
record this change.

23 years agoAdd kernel option NETGRAPH_ONE2MANY.
archie [Thu, 16 Nov 2000 16:59:26 +0000 (16:59 +0000)]
Add kernel option NETGRAPH_ONE2MANY.

23 years agoFix problem with extra space after an Xr, because aZ was being called twice.
ru [Thu, 16 Nov 2000 16:05:13 +0000 (16:05 +0000)]
Fix problem with extra space after an Xr, because aZ was being called twice.
Fix the no-space-after-a-period-in-macro problem.

Obtained from: NetBSD

23 years agovx is now optional rather than taking a count. Reflect that in the
imp [Thu, 16 Nov 2000 15:16:41 +0000 (15:16 +0000)]
vx is now optional rather than taking a count.  Reflect that in the
files.  Also a minor white space nit.

Submitted by: bde

23 years agomdoc(7)ify.
ru [Thu, 16 Nov 2000 13:21:17 +0000 (13:21 +0000)]
mdoc(7)ify.

23 years agoDescribe -deny_incoming better, highlight some keywords,
ru [Thu, 16 Nov 2000 12:20:54 +0000 (12:20 +0000)]
Describe -deny_incoming better, highlight some keywords,
add myself to the AUTHORS section.

23 years agofix typo; compliment -> complement.
ben [Thu, 16 Nov 2000 12:19:19 +0000 (12:19 +0000)]
fix typo; compliment -> complement.

PR: 22840
Submitted by: Jimmy Olgeni <olgeni@uli.it>

23 years agoremove trailing fullstop from SEE ALSO.
ben [Thu, 16 Nov 2000 11:42:11 +0000 (11:42 +0000)]
remove trailing fullstop from SEE ALSO.

Reviewed by: grog

23 years agoReplace a `dagger' sign with a `double dagger' one.
ru [Thu, 16 Nov 2000 11:22:42 +0000 (11:22 +0000)]
Replace a `dagger' sign with a `double dagger' one.
The former looks ugly on grotty(1) devices.

23 years agoPut the probe verboseness behind bootverbose
sos [Thu, 16 Nov 2000 10:52:00 +0000 (10:52 +0000)]
Put the probe verboseness behind bootverbose

23 years agoNew netgraph node type ng_one2many(4).
archie [Thu, 16 Nov 2000 05:58:33 +0000 (05:58 +0000)]
New netgraph node type ng_one2many(4).

23 years agoAdd ichsmb.4 to the list.
archie [Thu, 16 Nov 2000 03:43:56 +0000 (03:43 +0000)]
Add ichsmb.4 to the list.

23 years agoDon't release and acquire Giant in mi_switch(). Instead, release and
jhb [Thu, 16 Nov 2000 02:16:44 +0000 (02:16 +0000)]
Don't release and acquire Giant in mi_switch().  Instead, release and
acquire Giant as needed in functions that call mi_switch().  The releases
need to be done outside of the sched_lock to avoid potential deadlocks
from trying to acquire Giant while interrupts are disabled.

Submitted by: witness

23 years agoremove redundant declaration of bsd_to_linux_sigset()
gallatin [Thu, 16 Nov 2000 02:08:40 +0000 (02:08 +0000)]
remove redundant declaration of bsd_to_linux_sigset()

reviewed by: marcel

23 years agofix glaring bugs in rt signals -- copyout the right signal mask in
gallatin [Thu, 16 Nov 2000 02:07:05 +0000 (02:07 +0000)]
fix glaring bugs in rt signals -- copyout the right signal mask in
linux_rt_sendsig() and restore the same signal mask linux does
in rt_sigreturn().  This gets us saving/restoring all 64-bits of the
linux sigset_t in rt signals.

Reviewed by: marcel

23 years agoArgh, add in a missing release of the sched_lock.
jhb [Thu, 16 Nov 2000 01:16:54 +0000 (01:16 +0000)]
Argh, add in a missing release of the sched_lock.

23 years agoCURSIG() calls functions that acquire sleep mutexes, so it is not a good
jhb [Thu, 16 Nov 2000 01:07:19 +0000 (01:07 +0000)]
CURSIG() calls functions that acquire sleep mutexes, so it is not a good
idea to be holding the sched_lock while we are calling it.  As such,
release sched_lock before calling CURSIG() in msleep() and mawait() and
reacquire it after CURSIG() returns.

Submitted by: witness

23 years agoUse the linux_connect() on alpha rather than passing directly through
gallatin [Thu, 16 Nov 2000 01:05:53 +0000 (01:05 +0000)]
Use the linux_connect() on alpha rather than passing directly through
to our native connect().  This is required to deal with the differences
in the way linux handles connects on non-blocking sockets.

This gets the private beta of the Compaq Linux/alpha JDK working
on FreeBSD/alpha

Approved by: marcel

23 years agomake the fcntl() flags match what the linux/alpha port uses, not
gallatin [Thu, 16 Nov 2000 00:58:07 +0000 (00:58 +0000)]
make the fcntl() flags match what the linux/alpha port uses, not
what linux/i386 uses

23 years agoinstall/build accf_* manpages
alfred [Thu, 16 Nov 2000 00:57:52 +0000 (00:57 +0000)]
install/build accf_* manpages

23 years agoXref accf_http and accf_data
alfred [Wed, 15 Nov 2000 22:54:06 +0000 (22:54 +0000)]
Xref accf_http and accf_data

23 years agoDocument the accf_data accept filter: "dataready"
alfred [Wed, 15 Nov 2000 22:53:02 +0000 (22:53 +0000)]
Document the accf_data accept filter: "dataready"

23 years agoDocument the accf_http accept filter "httpready".
alfred [Wed, 15 Nov 2000 22:46:21 +0000 (22:46 +0000)]
Document the accf_http accept filter "httpready".

23 years ago- Rename await() to mawait(). mawait() is to await() as msleep() is to
jhb [Wed, 15 Nov 2000 22:39:35 +0000 (22:39 +0000)]
- Rename await() to mawait().  mawait() is to await() as msleep() is to
  tsleep().  Namely, mawait() takes an extra argument which is a mutex
  to drop when going to sleep.  Just as with msleep(), if the priority
  argument includes the PDROP flag, then the mutex will be dropped and will
  not be reacquired when the process wakes up.
- Add in a backwards compatible macro await() that passes in NULL as the
  mutex argument to mawait().

23 years ago- Replace a KASSERT() that knew too much about mutex internals with a
jhb [Wed, 15 Nov 2000 22:30:48 +0000 (22:30 +0000)]
- Replace a KASSERT() that knew too much about mutex internals with a
  mtx_assert() that ensures the mutex we release during msleep() is both
  not recursed and owned by the current process.

23 years ago- Convert references from tsleep() -> msleep()
jhb [Wed, 15 Nov 2000 22:27:38 +0000 (22:27 +0000)]
- Convert references from tsleep() -> msleep()
- Fix a buglet in a comment above await()

23 years ago- Add a new macro DROP_GIANT_NOSWITCH() that is similar to DROP_GIANT()
jhb [Wed, 15 Nov 2000 22:12:33 +0000 (22:12 +0000)]
- Add a new macro DROP_GIANT_NOSWITCH() that is similar to DROP_GIANT()
  except that it uses the MTX_NOSWITCH flag while it releases Giant via
  mtx_exit().
- Add a mtx_recursed() primitive.  This primitive should only be used on
  a mutex owned by the current process.  It will return non-zero if the
  mutex is recursively owned, or zero otherwise.
- Add two new flags MA_RECURSED and MA_NOTRECURSED that can be used in
  conjuction with MA_OWNED to control the assertion checked by mtx_assert().
- Fix some of the KTR tracepoint strings to use %p when displaying the lock
  field of a mutex, which is a uintptr_t.

23 years agoInclude the right headers to get the DDB #define and the db_active variable.
jhb [Wed, 15 Nov 2000 22:08:16 +0000 (22:08 +0000)]
Include the right headers to get the DDB #define and the db_active variable.

23 years ago- Replace some instances of sched_ithd with sched_swi in KTR tracepoints.
jhb [Wed, 15 Nov 2000 22:05:23 +0000 (22:05 +0000)]
- Replace some instances of sched_ithd with sched_swi in KTR tracepoints.
- Assert that Giant is not owned during the main loop of sithd_loop().

23 years agoCorrect the Bintuils src path from the debugging version I accidently
obrien [Wed, 15 Nov 2000 22:05:00 +0000 (22:05 +0000)]
Correct the Bintuils src path from the debugging version I accidently
committed.

23 years agoAssert that Giant is not owned during the main loop of ithd_loop().
jhb [Wed, 15 Nov 2000 22:03:26 +0000 (22:03 +0000)]
Assert that Giant is not owned during the main loop of ithd_loop().

23 years agoDeclare the 'witness_spin_check' properly as a per-CPU variable in the
jhb [Wed, 15 Nov 2000 22:02:05 +0000 (22:02 +0000)]
Declare the 'witness_spin_check' properly as a per-CPU variable in the
non-SMP case.