]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoBah, add in a missing space char I noticed when MFC'ing this.
John Baldwin [Wed, 22 Jan 2003 17:26:18 +0000 (17:26 +0000)]
Bah, add in a missing space char I noticed when MFC'ing this.

21 years agoNew errata item: adduser(8) passwds can't have special shell characters.
Bruce A. Mah [Wed, 22 Jan 2003 17:19:12 +0000 (17:19 +0000)]
New errata item:  adduser(8) passwds can't have special shell characters.

Submitted by: <robin@isometry.net>, mtm

21 years agoCorrectly identify the 3C920B chipset; although it may ship with some Nforce2
Mike Silbersack [Wed, 22 Jan 2003 17:14:33 +0000 (17:14 +0000)]
Correctly identify the 3C920B chipset; although it may ship with some Nforce2
boards, it's definitely not an Nvidia chip.

Info from: Nvidia's Linux Network driver & pciids.sourceforge.net

21 years agoThis commit was generated by cvs2svn to compensate for changes in r109686,
Dag-Erling Smørgrav [Wed, 22 Jan 2003 16:46:44 +0000 (16:46 +0000)]
This commit was generated by cvs2svn to compensate for changes in r109686,
which included commits to RCS files with non-trunk default branches.

21 years ago#include <sys/types.h> (perforce change 24073)
Dag-Erling Smørgrav [Wed, 22 Jan 2003 16:46:44 +0000 (16:46 +0000)]
#include <sys/types.h> (perforce change 24073)

21 years agoPrevent server-side glob(3) patterns from expanding
Yaroslav Tykhiy [Wed, 22 Jan 2003 16:25:22 +0000 (16:25 +0000)]
Prevent server-side glob(3) patterns from expanding
to a pathname that contains '\r' or '\n'.

Together with the earlier STAT bugfix, this must solve
the problem of such pathnames appearing in the FTP control
stream.

21 years agoList all of the assignment operators explicitly, for those who have been
Tony Finch [Wed, 22 Jan 2003 15:19:07 +0000 (15:19 +0000)]
List all of the assignment operators explicitly, for those who have been
traumatized by perl.

21 years agoForce early initialization of the resolver library, since the resolver
Dag-Erling Smørgrav [Wed, 22 Jan 2003 14:12:59 +0000 (14:12 +0000)]
Force early initialization of the resolver library, since the resolver
configuration files will no longer be available once sshd is chrooted.

PR: 39953, 40894
Submitted by: dinoex
MFC after: 3 days

21 years agoMake the arguments to d_close() match the arguments to d_open(), since
Joerg Wunsch [Wed, 22 Jan 2003 14:06:46 +0000 (14:06 +0000)]
Make the arguments to d_close() match the arguments to d_open(), since
otherwise the close operation is void, and the underlying device remains
marked as being busy.

Thanks to phk for finding this.

MFC after: 3 days

21 years agogetpeername() returns with no error but didn't fill struct sockaddr
Hajimu UMEMOTO [Wed, 22 Jan 2003 13:13:13 +0000 (13:13 +0000)]
getpeername() returns with no error but didn't fill struct sockaddr
correctly against PF_LOCAL.  It seems that the test always fails then
sockaddr was not filled.  So, I added else clause for workaround.
I doubt if it is right fix.  However, it is better than nothing.  I
found that NetBSD has same potential problem.  But, fortunately,
NetBSD has equivalent else clause.

MFC after: 1 week

21 years agoWhen computing a new seed for an existing user, opienewseed() would
Dag-Erling Smørgrav [Wed, 22 Jan 2003 10:55:36 +0000 (10:55 +0000)]
When computing a new seed for an existing user, opienewseed() would
incorrectly compute the length of the numeric portion of the previous
seed, causing the new seed to be one character shorter than the old
one.

This patch has been submitted to the vendor; I'm committing it right
away since the file is already off the vendor branch.

MFC after: 3 days

21 years agoAdd yet another Sony Vaio PS/2 mouse ID. This one is found on a VX88.
Jeffrey Hsu [Wed, 22 Jan 2003 03:31:08 +0000 (03:31 +0000)]
Add yet another Sony Vaio PS/2 mouse ID. This one is found on a VX88.

Submitted by: marcel

21 years agoRemove BAT invalidation. This is done later in the boot sequence,
Peter Grehan [Wed, 22 Jan 2003 01:47:05 +0000 (01:47 +0000)]
Remove BAT invalidation. This is done later in the boot sequence,
so isn't required here, and seems to cause problems when booting
from disk.

Approved by:  benno

21 years agoCatch up with the English version (w/ some translation fixes):
Hiroki Sato [Wed, 22 Jan 2003 01:45:14 +0000 (01:45 +0000)]
Catch up with the English version (w/ some translation fixes):

1.12  -> 1.20  errata/article.sgml
1.126 -> 1.129 hardware/common/dev.sgml
1.475 -> 1.480 relnotes/common/new.sgml

Submitted by: Hideyuki KURASHINA <rushani@jp.FreBSD.org>
References: [doc-jp-work 585]

21 years agoConvert remaining .s files to .S
Peter Grehan [Wed, 22 Jan 2003 01:34:37 +0000 (01:34 +0000)]
Convert remaining .s files to .S

Approved by:  benno

21 years ago- remove dead conditional CFLAG setting
Peter Grehan [Wed, 22 Jan 2003 01:34:10 +0000 (01:34 +0000)]
- remove dead conditional CFLAG setting
- switch locore .s -> .S

Approved by:  benno

21 years agoFix what I think is an off-by-one in certain worst-case scenarios
Peter Wemm [Wed, 22 Jan 2003 00:34:22 +0000 (00:34 +0000)]
Fix what I think is an off-by-one in certain worst-case scenarios
caused by rev 1.45.  (eg: the estimate being exactly half of the result.)

21 years agofix link_status() in non-Ethernet case
Max Khon [Wed, 22 Jan 2003 00:14:34 +0000 (00:14 +0000)]
fix link_status() in non-Ethernet case

MFC after: 1 week

21 years agoFinish update for cvs-1.11.5. Some build-time tunables were moved from
Peter Wemm [Tue, 21 Jan 2003 23:00:36 +0000 (23:00 +0000)]
Finish update for cvs-1.11.5.  Some build-time tunables were moved from
options.h to config.h.in and set via ./configure when built normally.
Export some of the build knobs to the Makefile here, overridable
from /etc/make.conf.  Also get the version strings right.
config.h was repocopied to config.h.proto, and we do a limited sed on it
at build time now.

21 years agoDont bother with unused rules for version.c
Peter Wemm [Tue, 21 Jan 2003 22:31:07 +0000 (22:31 +0000)]
Dont bother with unused rules for version.c

21 years agoTurn off Checkin-prog and Update-prog.
Peter Wemm [Tue, 21 Jan 2003 22:10:13 +0000 (22:10 +0000)]
Turn off Checkin-prog and Update-prog.

21 years agoUpdate for cvs-1.11.5 import (note to self: gotta fix the version tags)
Peter Wemm [Tue, 21 Jan 2003 22:04:20 +0000 (22:04 +0000)]
Update for cvs-1.11.5 import (note to self: gotta fix the version tags)

21 years agoThis commit was generated by cvs2svn to compensate for changes in r109661,
Peter Wemm [Tue, 21 Jan 2003 22:02:26 +0000 (22:02 +0000)]
This commit was generated by cvs2svn to compensate for changes in r109661,
which included commits to RCS files with non-trunk default branches.

21 years agoMerge cvs-1.11.2.1-20021201 -> 1.11.5 changes onto mainline
Peter Wemm [Tue, 21 Jan 2003 22:01:38 +0000 (22:01 +0000)]
Merge cvs-1.11.2.1-20021201 -> 1.11.5 changes onto mainline

21 years agoThis commit was generated by cvs2svn to compensate for changes in r109658,
Peter Wemm [Tue, 21 Jan 2003 21:58:47 +0000 (21:58 +0000)]
This commit was generated by cvs2svn to compensate for changes in r109658,
which included commits to RCS files with non-trunk default branches.

21 years agoThis commit was generated by cvs2svn to compensate for changes in r109655,
Peter Wemm [Tue, 21 Jan 2003 21:53:03 +0000 (21:53 +0000)]
This commit was generated by cvs2svn to compensate for changes in r109655,
which included commits to RCS files with non-trunk default branches.

21 years agoImport cvs-1.11.5 onto vendor branch
Peter Wemm [Tue, 21 Jan 2003 21:53:03 +0000 (21:53 +0000)]
Import cvs-1.11.5 onto vendor branch

21 years agoThere's absolutely no need for a struct-within-a-struct, so move the
Dag-Erling Smørgrav [Tue, 21 Jan 2003 20:33:27 +0000 (20:33 +0000)]
There's absolutely no need for a struct-within-a-struct, so move the
counters out of the inner struct and remove it.

21 years agoAdd missing SMP file locks around read-modify-write operations on
Jeffrey Hsu [Tue, 21 Jan 2003 20:20:48 +0000 (20:20 +0000)]
Add missing SMP file locks around read-modify-write operations on
the flag field.

Reviewed by: rwatson

21 years agoRemove OBE prototype for iszerodev() - it was replaced by the
Peter Wemm [Tue, 21 Jan 2003 19:26:17 +0000 (19:26 +0000)]
Remove OBE prototype for iszerodev() - it was replaced by the
D_MMAP_ANON device switch flag.

21 years agoFixes for a number of problems in the IOMMU code:
Thomas Moestl [Tue, 21 Jan 2003 18:22:26 +0000 (18:22 +0000)]
Fixes for a number of problems in the IOMMU code:
1.) Fix an off-by-one in the DVMA space handling, which would make it
    possible to allocate one page beyond the end of the DVMA area.
    This page was aliased to the first page. Apparently, this bug was
    responsible for the trashed nvram/eeprom some people were reporting,
    in conjunction with a number of unfortunate coincidences.
2.) Fix broken boundary and and lowaddr calculations.
3.) Fix a memory leak on an error path.
4.) Update a outdated comment to reflect the introduction of IOMMU_MAX_PRE,
    make the usage of IOMMU_MAX_PRE more consistent and KASSERT that the
    preallocation size is not 0.
5.) Fix a case where an error return was lost.
6.) When signalling an error to the caller by invoking the callback, do
    not use a segment pointer of NULL for compatability with existing
    drivers.

Also, increase the maximum segment number to 64; it is rather arbitrary,
with the exception of the of the stack space consumed by the segment
array.

Special thanks go to Harti Brandt <brandt@fokus.fraunhofer.de> for
spotting 4 and 5, and testing many iterations of patches.

Pointy hats to: tmm

21 years agoFree resources when failing to set up the interrupt.
Thomas Moestl [Tue, 21 Jan 2003 17:28:36 +0000 (17:28 +0000)]
Free resources when failing to set up the interrupt.

21 years agobus_dmamap_sync() overhaul:
Thomas Moestl [Tue, 21 Jan 2003 17:22:52 +0000 (17:22 +0000)]
bus_dmamap_sync() overhaul:
- Remove NetBSD-style or-ed together BUS_DMASYNC operations, in some
  cases relaxing the (intended) syncing operation a bit.
- Add syncs before reading the descriptor rings.
- Try to combine syncs where possible to avoid overhead.
- Sync all maps before unloading them.

21 years agobus_dmamap_sync() overhaul:
Thomas Moestl [Tue, 21 Jan 2003 17:17:01 +0000 (17:17 +0000)]
bus_dmamap_sync() overhaul:
- Remove NetBSD-style or-ed together BUS_DMASYNC operations, in some
  cases relaxing the (intended) syncing operation a bit.
- Stop pretending that that we can sync part of a dmamap: replace the
  GEM_CDTXSYNC and GEM_CDRXSYNC macros with GEM_CDSYNC to sync the
  complete control map, and combine syncs wherever possible to avoid
  the overhead.
- Sync all maps before unloading them.
- Remove a few syncs which should be unnecessary.

21 years agoFix iommu_dvmamap_sync(): it was still operating as if the BUS_DMASYNC_*
Thomas Moestl [Tue, 21 Jan 2003 17:08:22 +0000 (17:08 +0000)]
Fix iommu_dvmamap_sync(): it was still operating as if the BUS_DMASYNC_*
constants where flag bits (as in NetBSD), although they are consecutively
numbered in FreeBSD. This would cause unnecessary flushing in the
BUS_DMASYNC_POSTWRITE case, but was otherwise mostly harmless.

21 years agoCorrect an off-by-one in the boundary check. Otherwise, resource
Thomas Moestl [Tue, 21 Jan 2003 17:02:21 +0000 (17:02 +0000)]
Correct an off-by-one in the boundary check. Otherwise, resource
allocations would fail if the desired allocation size was equal to
the boundary.

21 years agoDefine misc structs outside of struct firewire_comm.
Hidetoshi Shimokawa [Tue, 21 Jan 2003 16:37:01 +0000 (16:37 +0000)]
Define misc structs outside of struct firewire_comm.

21 years agoIgnore events of unused IR DMA.
Hidetoshi Shimokawa [Tue, 21 Jan 2003 16:24:35 +0000 (16:24 +0000)]
Ignore events of unused IR DMA.

21 years agoThis commit was generated by cvs2svn to compensate for changes in r109641,
Jacques Vidrine [Tue, 21 Jan 2003 14:19:06 +0000 (14:19 +0000)]
This commit was generated by cvs2svn to compensate for changes in r109641,
which included commits to RCS files with non-trunk default branches.

21 years agoAdd a missing include, needed to get a prototype for `des_read_pw_string'.
Jacques Vidrine [Tue, 21 Jan 2003 14:19:06 +0000 (14:19 +0000)]
Add a missing include, needed to get a prototype for `des_read_pw_string'.
This is particularly important for OpenSSL 0.9.7, as `des_read_pw_string'
is a macro there.  (This fix brought in on the vendor branch, because I
already committed it to Heimdal's CVS.)

21 years agoDefine OPENSSL_DES_LIBDES_COMPATIBILITY so that Heimdal will build with
Jacques Vidrine [Tue, 21 Jan 2003 14:08:24 +0000 (14:08 +0000)]
Define OPENSSL_DES_LIBDES_COMPATIBILITY so that Heimdal will build with
OpenSSL 0.9.7 when it is imported.  (This currently has no effect.)

21 years agoRemove NEXTDOSPART.
Yoshihiro Takahashi [Tue, 21 Jan 2003 14:02:11 +0000 (14:02 +0000)]
Remove NEXTDOSPART.

21 years agoUse NDOSPART instead of NEXTDOSPART.
Yoshihiro Takahashi [Tue, 21 Jan 2003 13:59:53 +0000 (13:59 +0000)]
Use NDOSPART instead of NEXTDOSPART.

21 years agoMFi386: revision 1.63.
Yoshihiro Takahashi [Tue, 21 Jan 2003 13:57:43 +0000 (13:57 +0000)]
MFi386: revision 1.63.

21 years agoNeeds wlan for wi.
Warner Losh [Tue, 21 Jan 2003 13:31:52 +0000 (13:31 +0000)]
Needs wlan for wi.

21 years agoDon't forget to properly quote input.
Mike Makonnen [Tue, 21 Jan 2003 12:11:35 +0000 (12:11 +0000)]
Don't forget to properly quote input.

Approved by: markm (mentor)(implicit)
Submitted by: Robin Breathe <robin@isometry.net>

21 years ago#ifdef NO_GEOM some bits here too.
Poul-Henning Kamp [Tue, 21 Jan 2003 11:39:19 +0000 (11:39 +0000)]
#ifdef NO_GEOM some bits here too.

21 years agoUpdate with the current list of scanners that the driver supports.
Josef Karthauser [Tue, 21 Jan 2003 11:39:13 +0000 (11:39 +0000)]
Update with the current list of scanners that the driver supports.

21 years agoRegen
Josef Karthauser [Tue, 21 Jan 2003 11:38:17 +0000 (11:38 +0000)]
Regen

21 years agoAdd support for Epson 1660 scanner.
Josef Karthauser [Tue, 21 Jan 2003 11:37:54 +0000 (11:37 +0000)]
Add support for Epson 1660 scanner.

Submitted by: Bruno Schwander <bruno@tinkerbox.org>
MFC after: 3 days

21 years agouse 'void *' instead of 'caddr_t' for useracc, kernacc, vslock and vsunlock.
Alfred Perlstein [Tue, 21 Jan 2003 11:34:57 +0000 (11:34 +0000)]
use 'void *' instead of 'caddr_t' for useracc, kernacc, vslock and vsunlock.

21 years ago#ifdef NO_GEOM all of this file.
Poul-Henning Kamp [Tue, 21 Jan 2003 10:40:46 +0000 (10:40 +0000)]
#ifdef NO_GEOM all of this file.

21 years agoBandaid to make the kernel compile until the scsi-crew can find out what
Poul-Henning Kamp [Tue, 21 Jan 2003 10:31:43 +0000 (10:31 +0000)]
Bandaid to make the kernel compile until the scsi-crew can find out what
is happening.

21 years agoMake this compile with DEBUG defined now that WARNS=0 has been removed
Tim J. Robbins [Tue, 21 Jan 2003 10:06:04 +0000 (10:06 +0000)]
Make this compile with DEBUG defined now that WARNS=0 has been removed
from the Makefile:
- Print pointers with %p instead of %x.
- Include missing headers to get prototypes.

Noticed by: benno

21 years agoAdd yet another Sony Vaio PS/2 mouse ID. This one is found in a R505GL.
Marcel Moolenaar [Tue, 21 Jan 2003 10:00:32 +0000 (10:00 +0000)]
Add yet another Sony Vaio PS/2 mouse ID. This one is found in a R505GL.
The ACPI data for the mouse includes the compatibility ID, but we
apparently don't make use of it.

21 years agoAdd a missing '{'
Poul-Henning Kamp [Tue, 21 Jan 2003 09:57:01 +0000 (09:57 +0000)]
Add a missing '{'

21 years agoCatch up to WAIT/NOWAIT cleanup.
Alfred Perlstein [Tue, 21 Jan 2003 08:57:35 +0000 (08:57 +0000)]
Catch up to WAIT/NOWAIT cleanup.

21 years agoRemove M_TRYWAIT/M_WAITOK/M_WAIT. Callers should use 0.
Alfred Perlstein [Tue, 21 Jan 2003 08:56:16 +0000 (08:56 +0000)]
Remove M_TRYWAIT/M_WAITOK/M_WAIT.  Callers should use 0.
Merge M_NOWAIT/M_DONTWAIT into a single flag M_NOWAIT.

21 years agopccardd and pccardc are for i386 and pc98 only. Don't build them on
Warner Losh [Tue, 21 Jan 2003 08:42:27 +0000 (08:42 +0000)]
pccardd and pccardc are for i386 and pc98 only.  Don't build them on
other architectures.

Pointed out by: peter

21 years agoRemove the fake disklabel, no longer needed.
Søren Schmidt [Tue, 21 Jan 2003 08:33:48 +0000 (08:33 +0000)]
Remove the fake disklabel, no longer needed.
Fix potential malloc panic.

Prodded by: phk

21 years ago- Restore header printing to wi_printaplist()
Matthew N. Dodd [Tue, 21 Jan 2003 07:28:57 +0000 (07:28 +0000)]
- Restore header printing to wi_printaplist()
- Move APRATE switch() code to function for clarity.
- Conditionally call wi_printaplist() alone if more than one 'L'
  is preset.
- Add the 'Q' flag to suppress printing of extraneous information
  in wi_printaplist().
- Re-order second getopt() in main().

Reviewed by:  imp

21 years agopreserve the order of tags copied by m_tag_copy_chain
Sam Leffler [Tue, 21 Jan 2003 06:14:38 +0000 (06:14 +0000)]
preserve the order of tags copied by m_tag_copy_chain

Obtained from: OpenBSD

21 years agoThe PowerPC-specfic scripttempl file was depreciated in favor of the generic
David E. O'Brien [Tue, 21 Jan 2003 06:12:56 +0000 (06:12 +0000)]
The PowerPC-specfic scripttempl file was depreciated in favor of the generic
ELF one.

21 years agoRun kldxref on sparc64, it works now.
Jake Burkholder [Tue, 21 Jan 2003 05:52:48 +0000 (05:52 +0000)]
Run kldxref on sparc64, it works now.

21 years agoFix module dependency (pre)loading on sparc64 by relocating the variables
Jake Burkholder [Tue, 21 Jan 2003 05:46:46 +0000 (05:46 +0000)]
Fix module dependency (pre)loading on sparc64 by relocating the variables
read from the raw kld files.

Submitted by: Hartmut Brandt <brandt@fokus.gmd.de>
PR: 46870
Tested on: alpha (obrien), i386, sparc64

21 years ago - Add a VM_WAIT in the appropriate cases where vm_page_alloc() fails and flags
Jeff Roberson [Tue, 21 Jan 2003 05:44:52 +0000 (05:44 +0000)]
 - Add a VM_WAIT in the appropriate cases where vm_page_alloc() fails and flags
   indicate that uma_small_alloc should not.  This code should be refactored so
   that there is not so much cross arch duplication.

Reviewed by: jake
Spotted by: tmm
Tested on: alpha, sparc64
Pointy hat to: jeff and everyone who cut and pasted the bad code. :-)

21 years agohook fast_ipsec.4 up
Sam Leffler [Tue, 21 Jan 2003 05:35:56 +0000 (05:35 +0000)]
hook fast_ipsec.4 up

21 years agofast ipsec protocols
Sam Leffler [Tue, 21 Jan 2003 05:35:08 +0000 (05:35 +0000)]
fast ipsec protocols

21 years agoNew errata entry for ipfw(4).
Bruce A. Mah [Tue, 21 Jan 2003 05:30:41 +0000 (05:30 +0000)]
New errata entry for ipfw(4).

Submitted by: maxim
Reviewed by: rwatson, bmah

21 years agoThe FTP daemon was vulnerable to a DoS where an attacker could bind()
Crist J. Clark [Tue, 21 Jan 2003 05:13:02 +0000 (05:13 +0000)]
The FTP daemon was vulnerable to a DoS where an attacker could bind()
up port 20 for an extended period of time and thus lock out all other
users from establishing PORT data connections. Don't hold on to the
bind() while we loop around waiting to see if we can make our
connection.

Being a DoS, it has security implications, giving it a short MFC
time.

MFC after: 1 day

21 years agoAugment the model number list.
David E. O'Brien [Tue, 21 Jan 2003 04:43:16 +0000 (04:43 +0000)]
Augment the model number list.

21 years agoRewrite the SMP filedesc locking in knote_attach() in order to
Jeffrey Hsu [Tue, 21 Jan 2003 04:05:49 +0000 (04:05 +0000)]
Rewrite the SMP filedesc locking in knote_attach() in order to
  1.  eliminate unnecessary loop which frees and re-allocates
the just allocated array
  2.  eliminate the newsize recomputation
  3.  eliminate unnecessary unlock and relock around free
  4.  correctly match the free with the malloc into M_KQUEUE instead of M_TEMP
  5.  eliminate conditional assignment of oldlist, which is equivalent to a
simple assignment
  6.  eliminate the oldlist temporary variable completely

Reviewed by:    jhb

21 years agoGC an unused reference to vop_refreshlabel_desc; reference to
Robert Watson [Tue, 21 Jan 2003 04:05:37 +0000 (04:05 +0000)]
GC an unused reference to vop_refreshlabel_desc; reference to
opt_mac.h was removed previously so it was never compiled in.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoFix kldxref on sparc64 by allowing non-trivial relocations to be performed
Jake Burkholder [Tue, 21 Jan 2003 03:51:53 +0000 (03:51 +0000)]
Fix kldxref on sparc64 by allowing non-trivial relocations to be performed
on variables read out of raw kld files.  Unlike other platforms the value
will be in an Elf_Rela, not in the data section of the elf file.

Submitted by: Hartmut Brandt <brandt@fokus.gmd.de>
PR: 46730
Tested on: alpha (obrien), i386, sparc64

21 years agoPerform VOP_GETATTR() before mac_check_vnode_exec() so that
Robert Watson [Tue, 21 Jan 2003 03:26:28 +0000 (03:26 +0000)]
Perform VOP_GETATTR() before mac_check_vnode_exec() so that
the cached attributes are available to MAC modules.

Submitted by:   mike halderman <mrh@nosc.mil>
Obtained from: TrustedBSD Project

21 years agoResolve relative relocations in klds before trying to parse the module's
Jake Burkholder [Tue, 21 Jan 2003 02:42:44 +0000 (02:42 +0000)]
Resolve relative relocations in klds before trying to parse the module's
metadata.  This fixes module dependency resolution by the kernel linker on
sparc64, where the relocations for the metadata are different than on other
architectures; the relative offset is in the addend of an Elf_Rela record
instead of the original value of the location being patched.
Also fix printf formats in debug code.

Submitted by: Hartmut Brandt <brandt@fokus.gmd.de>
PR: 46732
Tested on: alpha (obrien), i386, sparc64

21 years agoFix a missed goal.period -> goal.offset change. In
Justin T. Gibbs [Mon, 20 Jan 2003 23:51:18 +0000 (23:51 +0000)]
Fix a missed goal.period -> goal.offset change.  In
this case, the bug resulted in comparing a period
against an offset.

21 years agoinclude cdefs.h so as to unbreak the libc build
Andrew Gallatin [Mon, 20 Jan 2003 22:05:39 +0000 (22:05 +0000)]
include cdefs.h so as to unbreak the libc build

21 years agoReference maclabel(7).
Chris Costello [Mon, 20 Jan 2003 21:25:00 +0000 (21:25 +0000)]
Reference maclabel(7).

Sponsored by: DARPA, Network Associates Laboratories

21 years agoConvert badsect to use libufs.
Juli Mallett [Mon, 20 Jan 2003 21:24:34 +0000 (21:24 +0000)]
Convert badsect to use libufs.

21 years agoAdd libufs to dependencies.
Juli Mallett [Mon, 20 Jan 2003 21:18:44 +0000 (21:18 +0000)]
Add libufs to dependencies.

21 years agoProperly mark up column lists. This does not affect output; I just had
Chris Costello [Mon, 20 Jan 2003 21:15:03 +0000 (21:15 +0000)]
Properly mark up column lists.  This does not affect output; I just had
the arguments to .Bl incorrect.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoMake tunefs use libufs, it seems to do well enough for printing / setting
Juli Mallett [Mon, 20 Jan 2003 21:15:02 +0000 (21:15 +0000)]
Make tunefs use libufs, it seems to do well enough for printing / setting
things.

21 years agomove sysctl's under hw.hifn; change max batching default to 1 (no batching)
Sam Leffler [Mon, 20 Jan 2003 21:11:56 +0000 (21:11 +0000)]
move sysctl's under hw.hifn; change max batching default to 1 (no batching)
based on what I learned from the Broadcom h/w

21 years agomove sysctl's under hw.ubsec
Sam Leffler [Mon, 20 Jan 2003 21:07:30 +0000 (21:07 +0000)]
move sysctl's under hw.ubsec

21 years agoReference the MAC chapter in the user Handbook.
Chris Costello [Mon, 20 Jan 2003 21:07:11 +0000 (21:07 +0000)]
Reference the MAC chapter in the user Handbook.

Sponsored by: DARPA, Network Associates Laboratories

21 years agocorrect default setting of hw.wi.txerate; it must be -1, not zero,
Sam Leffler [Mon, 20 Jan 2003 21:06:58 +0000 (21:06 +0000)]
correct default setting of hw.wi.txerate; it must be -1, not zero,
to silence all msgs (like the old driver)

21 years agocorrect sysctl names and move them to hw.wi: hw.wi.txerate controls the rate
Sam Leffler [Mon, 20 Jan 2003 20:55:37 +0000 (20:55 +0000)]
correct sysctl names and move them to hw.wi: hw.wi.txerate controls the rate
at which tx errors are printed (default to 0); hw.wi.debug control the debug
msgs and is only present when WI_DEBUG is defined at compile time (the default
for the moment)

Requested by: imp

21 years agoaic7xxx.reg:
Justin T. Gibbs [Mon, 20 Jan 2003 20:44:55 +0000 (20:44 +0000)]
aic7xxx.reg:
Add a constant for the controller's stack size and the
maximum scsi offset.

aic7xxx.seq:
Style nit.  The source is implied to be the destination
unless overridden in an "and" instruction.

Update target mode code for changes in identify seen
sequencer flags.

aic7xxx_pci.c:
Ensure that the PCIERRGENDIS bit is set in the
PCIERRGEN config space register.  Perhaps this
is a reason for the spurios parity errors reported
on U160 controllers.

Honor the AHC_NO_BIOS_INIT flag.

Allow PCI interrupt reporting to be disabled,
by clearing the PERRRESEN bit in the command
register.  This option is now enabled via a new
softc flag: AHC_DISABLE_PCI_PERR.

Disable SERR and pause the controller prior to performing
our mmapped I/O test.  This should handle the case of
controllers that do not "auto-access pause".  For legacy
controllers, use SCB ram instead of scratch ram since
the latter may contain settings left over from the BIOS
that we will use if an seeprom is not found.

Make use of new ahc_inl/outl() inlines.

aic7xxx.h:
Reformat a few comments to follow driver style.

Add a controller flags that indicate that a controller
has not been initialized by the BIOS and whether to
disable PCI parity errors..

Remove stack probing softc members.

Add a few more syncrate constants that are useful in speed
fallback calculations.

Add the SHOW_MASKED_ERRORS debug flag.

aic7xxx.h:
aic7xxx.c:
Implement the SCB_SILENT flag.  This is useful for
hushing up the driver during DV or other operations
that we expect to cause transmission errors.  The
messages will still print if the SHOW_MASKED_ERRORS
debug option is enabled.

aic7xxx_inline.h:
Implement ahc_[in|out][w|l|q].  This removes the need
for manual 'or and shift" type operations throughout
the driver.

aic7xxx.c:
Move SELTO dignostic so that the SCB is still valid
when we use it for printing path information.

If we are narrow, limit syncrate to Ultra2.

Don't clobber ppr_options when forcing a renegotiation.
The current ppr_options may be referenced while queuing
new commands.  Don't set our width to unknown when forcing
negotiation on narrow controllers.  This will confuse the
negotiation code into negotiating with a wide message on
narrow controllers.

Add an "asserting atn" diagnostic with controller/target
information.

Remove the probe_stack code.  The stack is always
4 deep on legacy controllers, so probing is pointless.
This also avoids an issue where probing the stack would
upset the aic7770.

In ahc_reset(), record whether or not we found the
controller in a reset state.  If the controller was
already reset, assume that no BIOS has initialized
the controller and ignore left over scratch ram
settings.

Fix an ifdef bug that caused sequencer debugging to
be enabled always.

Clear the ultraenb flag in our tstate during startup.
The ultraenbled'ness of a device is recorded in the user
transfer settings.  tstate->ultraenb bitmask indicates
which devices we have negotiated an ultra speed with.
Just after initialization, we are async.  Setting the
ultraenb flag while async seems to be harmless, but it
was confusing to see the ULTRAENB flag set in the SCB.

Enhance residual diagnostic to indicate if the residual
if for sense information or normal data transfers.

Indicate the features, bugs, and flags set in the softc
that are used to control firmware patch download when
booting verbose.

In ahc_dump_card_state() fix a logic reversal.  The
SCSIPHASE register only exists on U160 controllers.
The SCSISIGI register exists on all controllers.  Not
the other way around.  Also print out the ERROR register.

Allow ahc_dump_card_state() to be called when the sequencer
is not paused.  Add dump card state markers as in the U320
driver.

21 years agoFix typo in comment: inlcude -> include.
Jimmy Olgeni [Mon, 20 Jan 2003 20:23:56 +0000 (20:23 +0000)]
Fix typo in comment: inlcude -> include.

21 years agoaic79xx.c:
Justin T. Gibbs [Mon, 20 Jan 2003 20:17:35 +0000 (20:17 +0000)]
aic79xx.c:
Implement the SCB_SILENT flag.  This is useful for
hushing up the driver during DV or other operations
that we expect to cause transmission errors.  The
messages will still print if the SHOW_MASKED_ERRORS
debug option is enabled.

Save and restore the NEGOADDR address when setting
new transfer settings.  The sequencer performs lookups
in the negotiation table too and it expects NEGOADDR
to remain consistent across pause/unpause sessions.

Consistently use "offset" instead of "period" to determine
if we are running sync or not.

Add a SHOW_MESSAGES diagnostic for when we assert ATN
during message processing.

Print out IU, QAS, and RTI features when showing transfer options.

Limit the syncrate after all option conformance
changes have taken place in ahd_devlimited_syncrate.
Changes in options may change the final syncrate we
accept.

Keep a copy of the hs_mailbox in our softc so that
we can perform read/modify/write operations on the
hs_mailbox without having to pause the sequencer to
read the last written value.  Use the ENINT_COALESS
flag in the hs_mailbox to toggle interrupt coalessing.

Add entrypoints for enabling interrupt coalessing and
setting both a timeout (how long to wait for commands
to be coalessed) and a maximum commands to coaless value.
Add a statistics timer that decides when to enable or
disable interrupt coalessing based on load.

Add a routine, ahd_reset_cmds_pending() which is used
to update the CMDS_PENDING sequencer variable whenever
error recovery compeltes SCBs without notifying the
sequencer.  Since ahd_reset_cmds_pending is called
during ahd_unpause() only if we've aborted SCBs, its
call to ahd_flush_qoutfifo should not cause recursion
through ahd_run_qoutfifo().  A panic has been added to
ensure that this recursion does not occur.

In ahd_search_qinfifo, update the CMDS_PENDING sequencer
variable directly.  ahd_search_qinififo can be called
in situations where using ahd_reset_cmds_pending() might
cause recursion.  Since we can safely determine the
exact number to reduce CMDS_PENDING by in this scenario
without running the qoutfifo, the manual update is sufficient.

Clean up diagnostics.
Add ahd_flush_qoutfifo() which will run the qoutfifo
as well as complete any commands sitting on the
sequencer's COMPLETE_SCB lists or the good status FIFO.
Use this routine in several places that did similar
things in an add-hoc, but incomplete, fashion.  A call
to this routine was also added to ahd_abort_scbs() to
close a race.

In ahd_pause_and_flushwork() only return once selections
are safely disabled.  Flush all completed commands via
ahd_flush_qoutfifo().

Remove "Now packetized" diagnostic now that this
information is incorperated into the actual negotiation
messages that are displayed.

When forcing renegotiation, don't clober the current
ppr_options.  Much of the driver uses this information
to determine if we are currently packetized or not.

Remove some stray spaces at column 1 in ahd_set_tags.

When complaining about getting a host message loop
request with no pending messages, print out the
SCB_CONTROL register down on the card.

Modify the ahd_sent_msg() routine to handle a search
for an outgoing identify message.  Use this to detect
a msg reject on an identify message which typically
indicates that the target thought we were packetized.
Force a renegotiation in this case.

In ahd_search_qinfifo(), wait more effectively for SCB
DMA activities to cease.  We also disable SCB fetch
operations since we are about to change the qinfifo
and any fetch in progress will likely be invalidated.

In ahd_qinfifo_count(), fix the qinfifo empty case.

In ahd_dump_card_state(), print out CCSCBCTL in the
correct mode.

If we are a narrow controller, don't set the current
width to unknown when forcing a future negotiation.
This just confuses the code into attempting a wide
negotiation on a narrow bus.

Add support for task management function completions.

Modify ahd_handle_devreset so that it can handle
lun resets in addition to target resets.  Use
ahd_handle_devreset for lun and target reset task
management functions.

Handle the abort task TMF race case better.  We now
wait until any current selections are over and then
set the TMF back to zero.  This should cause the sequencer
to ignore the abort TMF completion should it occur.

Correct a bug in the illegal phase handler that
caused us to drop down to narrow when handling the
unexpected command phase case after 3rd party
reset of a packetized device.

Indicate the features, bugs, and flags set in the softc
that are used to control firmware patch download when
        booting verbose.

aic79xx.h:
Add coalessing and HS_MAILBOX fields.

Add per-softc variables for the stats "daemon".

Add a debug option for interrupt coalessing activities.

Add two new softc flags:
o AHD_UPDATE_PEND_CMDS
  Run ahd_reset_cmds_pending() on the next unpause.

o AHD_RUNNING_QOUTFIFO
  Used to catch recursion through ahd_run_qoutfifo().

aic79xx.reg:
Correct register addresses related to the software timer
and the DFDBCTL register.

Add constants paramaterizing the software timer.

Add scratch ram locations for storing interrupt coalessing
tunables.

Break INTMASK in SEQITNCTL out into INTMASK1 and INTMASK2.
In at least the REV A, these are writable bits.  We make
use of that for a swtimer workaround in the sequencer.

Since HS_MAILBOX autoclears, provide a sequencer variable
to store its contents.

Add SEQINT codes for handling task management completions.

aic79xx.seq:
Correct ignore wide residue processing check for
a wide negotiation being in effect.  We must be
in the SCSI register window in order to access the
negotiation table.

Use the software timer and a commands completed count to
implement interrupt coalessing.  The command complete is
deferred until either the maximum command threshold or a
the expiration of a command deferral timer.  If we have
more SCBs to complete to the host (sitting in COMPLETE_SCB
lists), always try to coaless them up to our coalessing limit.
If coalessing is enabled, but we have fewer commands oustanting
than the host's min coalessing limit, complete the command
immediately.

Add code to track the number of commands outstanding.
Commands are outstanding from the time they are placed
into the execution queue until the DMA to post completion
is setup.

Add a workaround for intvec_2 interrupts on the H2A4.
In H2A4, the mode pointer is not saved for intvec2, but
is restored on iret.  This can lead to the restoration
of a bogus mode ptr.  Manually clear the intmask bits and
do a normal return to compensate.  We use intvec_2 to
track interrupt coalessing timeouts.

Since we cannot disable the swtimer's countdown, simply
mask its interrupt once we no longer care about it firing.

In idle_loop_cchan, update LOCAL_HS_MAILBOX everytime
we are notified of an HS_MAILBOX update via the
HS_MAILBOX_ACT bit in QOFF_CTLSTA.  We have to use a
local copy of persistant portions of the HS_MAILBOX as
the mailbox auto-clears on any read.

Move the test for the cfg4istat interrupt up an instruction
to hopefully close a race between the next outgoing selection
and our disabling of selections.

Add a missing ret to the last instruction in load_overrun_buf.

Add notifications to the host of task management
completions as well as the completions for commands
that completed successfully before their corresponding
TMF could be sent.

Hold a critical section during select-out processing
until we have a fully identified connection.  This
removes a race condition with the legacy abort handler.

Correct a few spelling errors in some comments.

aic79xx_inline.h:
Call ahd_reset_cmds_pending() in ahd_unpause if required.

Update cmdcmplt interrupt statistics in our interrupt
handler.

Allow callers to ahd_send_scb() to set the task management
function.

aic79xx_pci.c:
Disable SERR and pause the controller prior to performing
our mmapped I/O test.  The U320 controllers do not support
"auto-access-pause".

aic79xx_osm.c:
Set the task management function now that
ahd_send_scb() doesn't do it for us.  We
also perform a lun reset in response to BDR
requests to packetized devices.

21 years agoFix swapping to a file, it was broken when SPECSTRATEGY was introduced.
Matthew Dillon [Mon, 20 Jan 2003 20:00:32 +0000 (20:00 +0000)]
Fix swapping to a file, it was broken when SPECSTRATEGY was introduced.

21 years agofix -L (scan ap's): use WI_RID_SCAN_APS instead of WI_RDI_SCAN_REQ to start
Sam Leffler [Mon, 20 Jan 2003 19:46:46 +0000 (19:46 +0000)]
fix -L (scan ap's): use WI_RID_SCAN_APS instead of WI_RDI_SCAN_REQ to start
a scan so the driver can use the best/right op for the card

21 years agogc dead code
Sam Leffler [Mon, 20 Jan 2003 19:45:09 +0000 (19:45 +0000)]
gc dead code

21 years ago#if 0 one more line to make this compile without subr_disklabel.c.
Poul-Henning Kamp [Mon, 20 Jan 2003 19:43:46 +0000 (19:43 +0000)]
#if 0 one more line to make this compile without subr_disklabel.c.

Hopefully forgiven by: scottl

21 years agoAdd headings so that we can tell where one errata entry ends and another
Chris Costello [Mon, 20 Jan 2003 19:34:05 +0000 (19:34 +0000)]
Add headings so that we can tell where one errata entry ends and another
begins.

Reviewed by: bmah

21 years agoWrite out the debug messages (from the previous commit) only if
Mike Makonnen [Mon, 20 Jan 2003 19:30:15 +0000 (19:30 +0000)]
Write out the debug messages (from the previous commit) only if
the evaluated command exists.

Approved by: markm (mentor)(implicit)

21 years agoMFbed: translation updates
Udo Erdelhoff [Mon, 20 Jan 2003 19:14:22 +0000 (19:14 +0000)]
MFbed: translation updates

errata/article.sgml: 1.18  -> 1.19
hardware/common/dev.sgml: 1.128 -> 1.129
relnotes/common/new.sgml: 1.480 -> 1.481

21 years agoo add BIOCGDLTLIST and BIOCSDLT ioctls to get the data link type list
Sam Leffler [Mon, 20 Jan 2003 19:08:46 +0000 (19:08 +0000)]
o add BIOCGDLTLIST and BIOCSDLT ioctls to get the data link type list
  and set the link type for use by libpcap and tcpdump
o move mtx unlock in bpfdetach up; it doesn't need to be held so long
o change printf in bpf_detach to distinguish it from the same one in bpfsetdlt

Note there are locking issues here related to ioctl processing; they
have not been addressed here.

Submitted by: Guy Harris <guy@alum.mit.edu>
Obtained from: NetBSD (w/ locking modifications)

21 years agoAdd a missing OR.
Mike Makonnen [Mon, 20 Jan 2003 18:57:16 +0000 (18:57 +0000)]
Add a missing OR.

Approved by: markm (mentor)(implicit)