]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoo Replace reference to param.h with types.h
Robert Watson [Sat, 25 Aug 2001 15:19:37 +0000 (15:19 +0000)]
o Replace reference to param.h with types.h
o Verbosicize reference to .fa mode

Obtained from: TrustedBSD Project

23 years agoo Add a man page for the kernel vaccess() call, a utility function to
Robert Watson [Sat, 25 Aug 2001 15:04:23 +0000 (15:04 +0000)]
o Add a man page for the kernel vaccess() call, a utility function to
  generate access control decisions.

Obtained from: TrustedBSD Project

23 years agoo Note that the ASU accounting flag has now been deprecated.
Robert Watson [Sat, 25 Aug 2001 14:55:57 +0000 (14:55 +0000)]
o Note that the ASU accounting flag has now been deprecated.

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoo s/violate/override/ Capabilities are part of the system policy, not
Robert Watson [Sat, 25 Aug 2001 14:37:01 +0000 (14:37 +0000)]
o s/violate/override/  Capabilities are part of the system policy, not
  an exception to it.

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoDocument that IFF_MULTICAST should be ORd with IFF_POINTOPOINT or
Brian Somers [Sat, 25 Aug 2001 09:14:55 +0000 (09:14 +0000)]
Document that IFF_MULTICAST should be ORd with IFF_POINTOPOINT or
IFF_BROADCAST for the TUNSIFMODE ioctl.

23 years agoTUNSIFINFO now expects IFF_MULTICAST to be OR'd with either IFF_POINTOPOINT
Brian Somers [Sat, 25 Aug 2001 09:12:57 +0000 (09:12 +0000)]
TUNSIFINFO now expects IFF_MULTICAST to be OR'd with either IFF_POINTOPOINT
or IFF_BROADCAST.  If it's not, the IFF_MULTICAST is removed.

This is in line with how NetBSD & OpenBSD do it.

23 years agoArgh! Revert accidental commit.
Peter Wemm [Sat, 25 Aug 2001 07:42:27 +0000 (07:42 +0000)]
Argh! Revert accidental commit.

23 years agoFirst cut at getting the ISA routing working. If we can't route a PCI
Warner Losh [Sat, 25 Aug 2001 06:40:00 +0000 (06:40 +0000)]
First cut at getting the ISA routing working.  If we can't route a PCI
interrupt for the CSC interrupt, then we revert to ISA.  If we didn't
have an interrupt set up with hw.pcic.irq, then do polling.

Also, don't complain about ignoring function 1 for any devices except
pcic devices.  This would normally only show up if someone set
hw.pcic.ignore_function_1=1.

MFC: as soon as I can test it on some troublesome laptops.

23 years agoImplement idle zeroing of pages. I've been tinkering with this
Peter Wemm [Sat, 25 Aug 2001 05:00:44 +0000 (05:00 +0000)]
Implement idle zeroing of pages.  I've been tinkering with this
on and off since John Dyson left his work-in-progress.

It is off by default for now.  sysctl vm.zeroidle_enable=1 to turn it on.

There are some hacks here to deal with the present lack of preemption - we
yield after doing a small number of pages since we wont preempt otherwise.

This is basically Matt's algorithm [with hysteresis] with an idle process
to call it in a similar way it used to be called from the idle loop.

I cleaned up the includes a fair bit here too.

23 years agovm_page_zero_idle() is no longer MD.
Peter Wemm [Sat, 25 Aug 2001 04:54:25 +0000 (04:54 +0000)]
vm_page_zero_idle() is no longer MD.

23 years agoRemove a reference to MAP_INHERIT.
Peter Wemm [Sat, 25 Aug 2001 04:27:40 +0000 (04:27 +0000)]
Remove a reference to MAP_INHERIT.

23 years agoUpdate rendition of a typical distribution directory. Text updates to
Bruce A. Mah [Sat, 25 Aug 2001 04:22:24 +0000 (04:22 +0000)]
Update rendition of a typical distribution directory.  Text updates to
follow.

23 years agoAdd another comment.
Julian Elischer [Sat, 25 Aug 2001 02:44:38 +0000 (02:44 +0000)]
Add another comment.
check for 'teh's this time..

23 years agoOptionize UPAGES for the i386. As part of this I split some of the low
Peter Wemm [Sat, 25 Aug 2001 02:20:02 +0000 (02:20 +0000)]
Optionize UPAGES for the i386.  As part of this I split some of the low
level implementation stuff out of machine/globaldata.h to avoid exposing
UPAGES to lots more places.  The end result is that we can double
the kernel stack size with 'options UPAGES=4' etc.

This is mainly being done for the benefit of a MFC to RELENG_4 at some
point.  -current doesn't really need this so much since each interrupt
runs on its own kstack.

23 years agos/teh/the/
Peter Wemm [Sat, 25 Aug 2001 02:06:59 +0000 (02:06 +0000)]
s/teh/the/

23 years agoAdd an explanatory note that would have saved me an hour or two
Julian Elischer [Sat, 25 Aug 2001 01:16:47 +0000 (01:16 +0000)]
Add an explanatory note that would have saved me an hour or two
of confusion had it been there when I started reading the code..

23 years agoJohn Polstra
Matt Jacob [Fri, 24 Aug 2001 23:08:13 +0000 (23:08 +0000)]
John  Polstra

bogus additional write to WXREG_IMASK that enabled *all* interrupt causes
while I was trying to disable the chip. Oops.

23 years agoForce a commit on kern_mutex.c to explain reason for last commit but while
Bosko Milekic [Fri, 24 Aug 2001 23:00:59 +0000 (23:00 +0000)]
Force a commit on kern_mutex.c to explain reason for last commit but while
I'm at it also add a comment in mtx_validate() explaining the purpose
of the last change.

Basically, this fixes booting kernels compiled with MUTEX_DEBUG. What used
to happen is before we setidt from init386() [still using BTX idt], we
called mtx_init() on several mutex locks, notably Giant and some others.
This is a problem for MUTEX_DEBUG because it enables mtx_validate() which
calls kernacc(), some of which in turn requires Giant.
Fix by calling kernacc() from mtx_validate() only if (!cold).

23 years ago*** empty log message ***
Bosko Milekic [Fri, 24 Aug 2001 22:53:45 +0000 (22:53 +0000)]
*** empty log message ***

23 years agoUse Brian Sommers' named entry
Warner Losh [Fri, 24 Aug 2001 21:43:35 +0000 (21:43 +0000)]
Use Brian Sommers' named entry

23 years agoRename the DIAGNOSTICS section to RETURN VALUES and describe the
Brian Somers [Fri, 24 Aug 2001 21:39:27 +0000 (21:39 +0000)]
Rename the DIAGNOSTICS section to RETURN VALUES and describe the
return values a little more.  Specifically, mention that a return
of 0 from NgRecvData() and NgRecvMsg() means the socket has been
closed.

Suggested by: jkh

23 years agoAdd INHERIT_XXX defines for minherit() system call.
Matthew Dillon [Fri, 24 Aug 2001 19:48:07 +0000 (19:48 +0000)]
Add INHERIT_XXX defines for minherit() system call.
Remove MAP_INHERIT - it is no longer supported.

23 years agoOk, third time is the charm. VM_INHERIT_XXX -> INHERIT_XX (use the same
Matthew Dillon [Fri, 24 Aug 2001 19:45:59 +0000 (19:45 +0000)]
Ok, third time is the charm.  VM_INHERIT_XXX -> INHERIT_XX (use the same
semantics as PROT_XXX vs VM_PROT_XXX separating user-space defines from
kernel defines).

23 years agoOops, minherit() uses VM_INHERIT_XXX as the argument, not MAP_XXX. Properly
Matthew Dillon [Fri, 24 Aug 2001 19:43:45 +0000 (19:43 +0000)]
Oops, minherit() uses VM_INHERIT_XXX as the argument, not MAP_XXX.  Properly
document minherit().

23 years agoRemove support for the badly broken MAP_INHERIT (from -current only).
Matthew Dillon [Fri, 24 Aug 2001 19:29:56 +0000 (19:29 +0000)]
Remove support for the badly broken MAP_INHERIT (from -current only).

23 years agoUpdate the mmap.2 and minherit.2 manual pages. Add a short explanation and
Matthew Dillon [Fri, 24 Aug 2001 19:28:00 +0000 (19:28 +0000)]
Update the mmap.2 and minherit.2 manual pages.  Add a short explanation and
referal from mmap to minherit for MAP_INHERIT.  Fully document the
minherit.2 manual page (because frankly, my dear, however you think it
currently works is almost certainly wrong!).  I may soon re-implement
MAP_COPY because I believe we can support it properly now, but I will have
to call it something else and that is for a later time.

23 years agoRemove MAP_NOEXTEND. It came from 4.4-lite and not only was never
Matthew Dillon [Fri, 24 Aug 2001 18:22:46 +0000 (18:22 +0000)]
Remove MAP_NOEXTEND.  It came from 4.4-lite and not only was never
implemented, but mmap()'s default behavior is *already* to not extend
files.  Only write() or ftruncate() can extend a file.

23 years agoStyle nits:
John Baldwin [Fri, 24 Aug 2001 17:46:58 +0000 (17:46 +0000)]
Style nits:
- Don't use punctuation or newlines in panic messages.
- Remove excess blank lines.

Requested and partially submitted by: bde

23 years agoExplain what we're doing in pcic_pci_shutdown a little better. Try to
Warner Losh [Fri, 24 Aug 2001 17:29:42 +0000 (17:29 +0000)]
Explain what we're doing in pcic_pci_shutdown a little better.  Try to
explain the subtle side effects that are going to happen and why we go
ahead and ack the interrupt source.  This stuff is tricky to get
right.

Also, emperical tests have shown that doing a shutdown in attach to be
ineffectual, so remove it from there.  Analysis of the code paths
shows that nearly identical writes to these registers happen in later
parts of the code.  The hanging problem on thinkpads when we change
the interrupt routing type is something else.

23 years agoAxe a commented, unused #define related to the old giant lock.
John Baldwin [Fri, 24 Aug 2001 16:52:31 +0000 (16:52 +0000)]
Axe a commented, unused #define related to the old giant lock.

23 years agoINTREN() and INTRDIS() have been MP safe for a while. Don't say that they
John Baldwin [Fri, 24 Aug 2001 16:51:47 +0000 (16:51 +0000)]
INTREN() and INTRDIS() have been MP safe for a while.  Don't say that they
are "soon to be" MP safe in the comments.

23 years agoRemove references to the old giant kernel lock in various comments.
John Baldwin [Fri, 24 Aug 2001 16:46:15 +0000 (16:46 +0000)]
Remove references to the old giant kernel lock in various comments.

23 years agoPrevent passing a null pointer as a filename to vn_open(),
Peter Pentchev [Fri, 24 Aug 2001 15:49:30 +0000 (15:49 +0000)]
Prevent passing a null pointer as a filename to vn_open(),
if for some reason expand_name() failed to build a core file name.

PR: 29931
Submitted by: Foldi Tamas <crow@kapu.hu>
Reviewed by: dd, -arch
MFC after: 1 month

23 years agoHandle return values of 0 from NgRecvMsg() properly.
Brian Somers [Fri, 24 Aug 2001 14:53:41 +0000 (14:53 +0000)]
Handle return values of 0 from NgRecvMsg() properly.

23 years agoUnderstand that a return value of 0 from NgRecvMsg() means that the
Brian Somers [Fri, 24 Aug 2001 14:52:38 +0000 (14:52 +0000)]
Understand that a return value of 0 from NgRecvMsg() means that the
socket was closed.

This prevents erroneous ``Unexpected netgraph version'' from turning
up in the log.

23 years agoExplain what the return value from NgRecvMsg() and NgRecvData() means,
Brian Somers [Fri, 24 Aug 2001 14:52:05 +0000 (14:52 +0000)]
Explain what the return value from NgRecvMsg() and NgRecvData() means,
specifically that 0 means the socket has been closed.

23 years agoFd is macro too, so use \&Fd
Andrey A. Chernov [Fri, 24 Aug 2001 11:12:58 +0000 (11:12 +0000)]
Fd is macro too, so use \&Fd

23 years agoFildes -> Fd too (started from big letter)
Andrey A. Chernov [Fri, 24 Aug 2001 11:05:11 +0000 (11:05 +0000)]
Fildes -> Fd too (started from big letter)

23 years agoChange
Andrey A. Chernov [Fri, 24 Aug 2001 10:59:32 +0000 (10:59 +0000)]
Change
start means ...
to
.Fa l_start
means ...

23 years agoFix up English from previous 3 revisions.
Sheldon Hearn [Fri, 24 Aug 2001 10:36:29 +0000 (10:36 +0000)]
Fix up English from previous 3 revisions.

There is no such argument 'fildes' in the SYNOPSIS.  It's called 'fd'.

23 years agoCopy from kern_lockf.c: remove extra check
Andrey A. Chernov [Fri, 24 Aug 2001 10:22:16 +0000 (10:22 +0000)]
Copy from kern_lockf.c: remove extra check

23 years agoRemove extra check unneded now
Andrey A. Chernov [Fri, 24 Aug 2001 10:20:26 +0000 (10:20 +0000)]
Remove extra check unneded now

23 years agoOn today's kernels masking with ~KERNBASE is turning out to be less
Peter Wemm [Fri, 24 Aug 2001 09:56:44 +0000 (09:56 +0000)]
On today's kernels masking with ~KERNBASE is turning out to be less
than useful.  It still hits at least 8 digits. Adjust for reality.

This is still not satisfactory for the alpha if you add "-O paddr".

23 years agoHandle kvm_getprocs() returning nothing. Dont pass -1 to a size
Peter Wemm [Fri, 24 Aug 2001 09:53:26 +0000 (09:53 +0000)]
Handle kvm_getprocs() returning nothing.  Dont pass -1 to a size
for malloc().

23 years agoMake ps -M corefile work again. This has been broken for quite some time.
Peter Wemm [Fri, 24 Aug 2001 09:43:44 +0000 (09:43 +0000)]
Make ps -M corefile work again.  This has been broken for quite some time.
kvm_proclist() was aborting when it saw the ithreads with no pgrp.

23 years agoBanish hard-coded KERNBASE references from savecore. Dynamically
Peter Wemm [Fri, 24 Aug 2001 09:26:17 +0000 (09:26 +0000)]
Banish hard-coded KERNBASE references from savecore.  Dynamically
adjust to whatever kernbase is in the kernel that we are dumping.

23 years agoDynamically adapt to kernbase changes on crashdumps, falling back to
Peter Wemm [Fri, 24 Aug 2001 09:12:04 +0000 (09:12 +0000)]
Dynamically adapt to kernbase changes on crashdumps, falling back to
KERNBASE if the "kernbase" symbol is not present on older kernels.

23 years agoDynamically adjust to the value of KERNBASE in a crashdump, with
Peter Wemm [Fri, 24 Aug 2001 08:53:30 +0000 (08:53 +0000)]
Dynamically adjust to the value of KERNBASE in a crashdump, with
a fallback for old kernels without the "kernbase" symbol.

23 years agoExport the actual KERNBASE to the symbol table. We can use nlist() to get
Peter Wemm [Fri, 24 Aug 2001 08:29:54 +0000 (08:29 +0000)]
Export the actual KERNBASE to the symbol table.  We can use nlist() to get
this without having to second guess it in userland.

23 years agoMove cpu_fxsr definition to C code (so debug info is generated) and where
Peter Wemm [Fri, 24 Aug 2001 08:27:24 +0000 (08:27 +0000)]
Move cpu_fxsr definition to C code (so debug info is generated) and where
it is easily #ifdef'ed so that we dont miss unintentional references to it.

23 years agoMove code to shutdown the pcic_pci_shutdown. Call it on system
Warner Losh [Fri, 24 Aug 2001 07:43:35 +0000 (07:43 +0000)]
Move code to shutdown the pcic_pci_shutdown.  Call it on system
shutdown and also before we get going with the device initialization.
This may fix the hangs some people are seeing on warmboot.  It appears
that some machines will reset the cardbus bridge on boot, while others
don't.  So we turn off the card, and ack the interrupts (which likely
is a nop in the shutdown case since we're still fielding interrupts).
This should turn off the interrupts.

Since I don't have hardware that hangs on reboot, I'm committing this
without testing that aspect of the patch (it causes no harm on my
Dell).

23 years agoTwo fixes. First, put into place a more generic chipset specific
Warner Losh [Fri, 24 Aug 2001 07:30:18 +0000 (07:30 +0000)]
Two fixes.  First, put into place a more generic chipset specific
initialization structure.  Warn the user for those chipsets that
aren't yet customized that they might not work.  Second, try to power
off the slot on attach and ack the interrupts.  I don't know, but this
might solve the hangs that people will see on Thinkpads if they set
hw.pcic.init_routing=1.

23 years agoGet rid of redundant device_set_desc. Both the PCI and PCCARD versions
Nick Sayer [Fri, 24 Aug 2001 05:47:50 +0000 (05:47 +0000)]
Get rid of redundant device_set_desc. Both the PCI and PCCARD versions
do this in the probe routine now.

23 years agoRemove a duplicate "that".
Kenneth D. Merry [Fri, 24 Aug 2001 02:24:58 +0000 (02:24 +0000)]
Remove a duplicate "that".

Submitted by: "Sergey A. Osokin" <osa@freebsd.org.ru>

23 years agoIn order for the MA301 to pass interrupts from the PCCard, you must
Nick Sayer [Fri, 24 Aug 2001 02:14:26 +0000 (02:14 +0000)]
In order for the MA301 to pass interrupts from the PCCard, you must
go into the PLX 9052's interrupt control register and turn on the magic
interrupt enable bit.

Partial thanks are due to OpenBSD for pointing out that the chip is a
PLX 9052 and pointing me to the datasheet PDF.

23 years agoo Clarify comments in vaccess_acl_posix1e() ACL evaluation routine so
Robert Watson [Fri, 24 Aug 2001 01:41:42 +0000 (01:41 +0000)]
o Clarify comments in vaccess_acl_posix1e() ACL evaluation routine so
  as to improve readability and accuracy.

Obtained from: TrustedBSD Project

23 years agoReflect that the lge(4) driver has been tested on the Alpha.
Bruce A. Mah [Fri, 24 Aug 2001 00:30:50 +0000 (00:30 +0000)]
Reflect that the lge(4) driver has been tested on the Alpha.

Reviewed by: wpaul

23 years agobegin executed --> being executed
Dima Dorfman [Fri, 24 Aug 2001 00:18:37 +0000 (00:18 +0000)]
begin executed --> being executed

Obtained from: NetBSD

23 years agoif_wi can be either pci (without pccard) or pccard
Nick Sayer [Thu, 23 Aug 2001 23:58:49 +0000 (23:58 +0000)]
if_wi can be either pci (without pccard) or pccard

23 years ago- callout_stop() now returns an integer value instead of void so
John Baldwin [Thu, 23 Aug 2001 23:52:59 +0000 (23:52 +0000)]
- callout_stop() now returns an integer value instead of void so
  describe it.
- Add a return values section.

23 years agoSpelling fix: depreciated -> deprecated.
John Baldwin [Thu, 23 Aug 2001 23:41:53 +0000 (23:41 +0000)]
Spelling fix: depreciated -> deprecated.

23 years agoRemove refernces to M_ASLEEP/asleep/await.
John Baldwin [Thu, 23 Aug 2001 23:40:53 +0000 (23:40 +0000)]
Remove refernces to M_ASLEEP/asleep/await.

23 years agoRemove references to asleep/await.
John Baldwin [Thu, 23 Aug 2001 23:37:31 +0000 (23:37 +0000)]
Remove references to asleep/await.

23 years agoUse witness_upgrade/downgrade for sx_try_upgrade/downgrade.
John Baldwin [Thu, 23 Aug 2001 22:51:22 +0000 (22:51 +0000)]
Use witness_upgrade/downgrade for sx_try_upgrade/downgrade.

23 years agoWhitespace nit.
John Baldwin [Thu, 23 Aug 2001 22:47:51 +0000 (22:47 +0000)]
Whitespace nit.

23 years agoAdd witness_upgrade() and witness_downgrade() for handling upgrades and
John Baldwin [Thu, 23 Aug 2001 22:47:05 +0000 (22:47 +0000)]
Add witness_upgrade() and witness_downgrade() for handling upgrades and
downgrades of shared/exclusive locks.

23 years agoConvert some KASSERT()'s into if (foo) panic() because they are testing
John Baldwin [Thu, 23 Aug 2001 22:44:47 +0000 (22:44 +0000)]
Convert some KASSERT()'s into if (foo) panic() because they are testing
how locks are managed by the rest of the kernel, not verifying the internal
integrity of witness itself.

23 years agoKilled reference to MAP_INHERIT which is not supported in FreeBSD.
David Greenman [Thu, 23 Aug 2001 22:39:52 +0000 (22:39 +0000)]
Killed reference to MAP_INHERIT which is not supported in FreeBSD.

23 years agoCorrect the comment about bpfattach() to match reality.
Dima Dorfman [Thu, 23 Aug 2001 22:38:08 +0000 (22:38 +0000)]
Correct the comment about bpfattach() to match reality.

PR: 29967
Submitted by: Joseph Mallett <jmallett@xMach.org>

23 years agoCorrect a typo in a comment: FIN_WAIT2 -> FIN_WAIT_2
Dima Dorfman [Thu, 23 Aug 2001 22:34:29 +0000 (22:34 +0000)]
Correct a typo in a comment: FIN_WAIT2 -> FIN_WAIT_2

PR: 29970
Submitted by: Joseph Mallett <jmallett@xMach.org>

23 years agoImplement `stty ek` as documented.
Dima Dorfman [Thu, 23 Aug 2001 22:31:13 +0000 (22:31 +0000)]
Implement `stty ek` as documented.

PR: 24063
Submitted by: Seth Kingsley <sethk@osd.bsdi.com>

23 years ago$units is the canonical name for the number of units to create; use it.
Dima Dorfman [Thu, 23 Aug 2001 22:19:53 +0000 (22:19 +0000)]
$units is the canonical name for the number of units to create; use it.

PR: 29971
Submitted by: Joseph Mallett <jmallett@xMach.org>

23 years agoSpelling police: queueing -> queuing
Dima Dorfman [Thu, 23 Aug 2001 21:58:38 +0000 (21:58 +0000)]
Spelling police: queueing -> queuing

PR: 29985
Submitted by: Joseph Mallett <jmallett@xMach.org>

23 years agoSpelling police: relabelled -> relabeled
Dima Dorfman [Thu, 23 Aug 2001 21:52:38 +0000 (21:52 +0000)]
Spelling police: relabelled -> relabeled

PR: 29985
Submitted by: {aac,amr,mlx}.4

23 years agoCapitalize Linux where appropriate.
Dima Dorfman [Thu, 23 Aug 2001 21:51:16 +0000 (21:51 +0000)]
Capitalize Linux where appropriate.

Submitted by: Joseph Mallett <jmallett@xMach.org>

23 years agoThis commit was generated by cvs2svn to compensate for changes in r82230,
Jeroen Ruigrok van der Werven [Thu, 23 Aug 2001 21:46:59 +0000 (21:46 +0000)]
This commit was generated by cvs2svn to compensate for changes in r82230,
which included commits to RCS files with non-trunk default branches.

23 years agoString format patch
Jeroen Ruigrok van der Werven [Thu, 23 Aug 2001 21:46:59 +0000 (21:46 +0000)]
String format patch

23 years agoSpelling police: withough -> without
Dima Dorfman [Thu, 23 Aug 2001 21:36:18 +0000 (21:36 +0000)]
Spelling police: withough -> without

Submitted by: Joseph Mallett <jmallett@xMach.org>

23 years agoThe name of the kernel really isn't KERNEL now.
Dima Dorfman [Thu, 23 Aug 2001 21:17:28 +0000 (21:17 +0000)]
The name of the kernel really isn't KERNEL now.

Submitted by: setantae@submonkey.net

23 years agoNew release note: SA-01:56.
Bruce A. Mah [Thu, 23 Aug 2001 21:14:42 +0000 (21:14 +0000)]
New release note:  SA-01:56.

Fix entry on SA-01:55, which was mislabeled as SA-01:54.  Also fix a
typo (I still can't spell "compatibility" correctly, it seems).

23 years ago``recognized -> recognize'' where appropriate.
Dima Dorfman [Thu, 23 Aug 2001 21:13:31 +0000 (21:13 +0000)]
``recognized -> recognize'' where appropriate.

Obtained from: OpenBSD

23 years agoAdd a new kernel option RESTARTABLE_PANICS. If this option is present,
John Baldwin [Thu, 23 Aug 2001 20:32:21 +0000 (20:32 +0000)]
Add a new kernel option RESTARTABLE_PANICS.  If this option is present,
then one can restart from a panic by resetting the panicstr variable to
NULL.  This commit conditionalizes the previously committed functionality
on this variable.  It also removes the __dead2 attribute from the panic()
function so that when one continues from a panic() the behavior will
be predictable.

23 years agoMisc. changes:
Jimmy Olgeni [Thu, 23 Aug 2001 19:54:11 +0000 (19:54 +0000)]
Misc. changes:

* Rename cvsup-bin to cvsup (cvsup-bin does not exist anymore).
* Add the net/rsync port to the CDROM packages.
* Add the misc/compat* ports.

Approved by: jkh

23 years agoNow we implement l_len<0 per POSIX, describe it.
Andrey A. Chernov [Thu, 23 Aug 2001 19:00:34 +0000 (19:00 +0000)]
Now we implement l_len<0 per POSIX, describe it.

23 years agoExpansions and clarification of some existing release notes items.
Bruce A. Mah [Thu, 23 Aug 2001 18:39:04 +0000 (18:39 +0000)]
Expansions and clarification of some existing release notes items.

Submitted by: Alex Kapranoff <kapr@acm.org>

23 years agoFix a bug in the ste_setmulti() routine. The NIC has 4 16-bit multicast
Bill Paul [Thu, 23 Aug 2001 18:22:55 +0000 (18:22 +0000)]
Fix a bug in the ste_setmulti() routine. The NIC has 4 16-bit multicast
hash registers, not 2 32-bit ones. This would prevent the multicasr filter
from being programmed correctly in some cases.

23 years agoStupid error from my side in prev. commit: || -> &&
Andrey A. Chernov [Thu, 23 Aug 2001 18:02:29 +0000 (18:02 +0000)]
Stupid error from my side in prev. commit: || -> &&

23 years agoClear the sx_xholder pointer when downgrading an exclusive lock.
John Baldwin [Thu, 23 Aug 2001 17:57:37 +0000 (17:57 +0000)]
Clear the sx_xholder pointer when downgrading an exclusive lock.

23 years agovn_stat(): if va_size (u_quad_t) > OFF_MAX, return EOVERFLOW, don't copy it
Andrey A. Chernov [Thu, 23 Aug 2001 17:56:48 +0000 (17:56 +0000)]
vn_stat(): if va_size (u_quad_t) > OFF_MAX, return EOVERFLOW, don't copy it
blindly to st_size

23 years agoCopy yet one check for SEEK_END overflow
Andrey A. Chernov [Thu, 23 Aug 2001 17:12:42 +0000 (17:12 +0000)]
Copy yet one check for SEEK_END overflow

23 years agoAdd yet one check for SEEK_END overflow
Andrey A. Chernov [Thu, 23 Aug 2001 17:09:23 +0000 (17:09 +0000)]
Add yet one check for SEEK_END overflow

23 years agoBail if we go beyond the directory size, not just if we hit it.
Andrew Gallatin [Thu, 23 Aug 2001 17:08:26 +0000 (17:08 +0000)]
Bail if we go beyond the directory size, not just if we hit it.
Certain ISO fs's (like the one for 4.4-RC1 disc1 on alpha)
trigger this, and we end up opening a null file name.  This causes us to get
a false match for "kernel.ko" when it does not exist.

23 years agolseek: fix check for vattr.va_size overflow. Check suggested by bde simple not
Andrey A. Chernov [Thu, 23 Aug 2001 17:01:25 +0000 (17:01 +0000)]
lseek: fix check for vattr.va_size overflow. Check suggested by bde simple not
works with unsigned types.

23 years agoImplement l_len<0 per POSIX check.
Andrey A. Chernov [Thu, 23 Aug 2001 16:13:59 +0000 (16:13 +0000)]
Implement l_len<0 per POSIX check.
Check for valid l_whence too.

23 years agoCopy my newly introduced l_len<0 'oops' fix from kern_lockf.c
Andrey A. Chernov [Thu, 23 Aug 2001 16:06:14 +0000 (16:06 +0000)]
Copy my newly introduced l_len<0 'oops' fix from kern_lockf.c

23 years agoOops, fix my broken handling of new l_len<0 case
Andrey A. Chernov [Thu, 23 Aug 2001 16:00:27 +0000 (16:00 +0000)]
Oops, fix my broken handling of new l_len<0 case

23 years agoCopy POSIX l_len<0 handling from kern_lockf.c
Andrey A. Chernov [Thu, 23 Aug 2001 15:44:24 +0000 (15:44 +0000)]
Copy POSIX l_len<0 handling from kern_lockf.c

23 years agoOriginally BSD return EINVAL for l_len < 0, but now POSIX wants it too,
Andrey A. Chernov [Thu, 23 Aug 2001 15:40:30 +0000 (15:40 +0000)]
Originally BSD return EINVAL for l_len < 0, but now POSIX wants it too,
so implement POSIX l_len < 0 handling.

23 years agoFix typo in repo damage message.
Warner Losh [Thu, 23 Aug 2001 15:39:27 +0000 (15:39 +0000)]
Fix typo in repo damage message.

Submitted by: Vadim Ostranitsyn <vadim@tsu.ru>

23 years agoCosmetique: correct English in comments
Andrey A. Chernov [Thu, 23 Aug 2001 14:49:02 +0000 (14:49 +0000)]
Cosmetique: correct English in comments