]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoDIS_8859-15 -> ISO_8859-15 rename
ache [Fri, 16 Mar 2001 07:31:55 +0000 (07:31 +0000)]
DIS_8859-15 -> ISO_8859-15 rename

23 years agoadd cnw driver to notes/lint
imp [Fri, 16 Mar 2001 07:29:58 +0000 (07:29 +0000)]
add cnw driver to notes/lint

23 years agoAdd cwn driver
imp [Fri, 16 Mar 2001 07:27:49 +0000 (07:27 +0000)]
Add cwn driver

23 years agoCommit port of cnw driver from Hiroyuki Aizu-san. This driver
imp [Fri, 16 Mar 2001 07:25:42 +0000 (07:25 +0000)]
Commit port of cnw driver from Hiroyuki Aizu-san.  This driver
supports Xircom netwave series of cards.  I have one of these cards
(but am trying to find one or two to test with), but it compiles and
aizu-san says it works.

I don't know if this supports 802.11 or not.  I've seen conflicting
information on this.

Submitted by: Hiroyuki Aizu <aizu@jaist.ac.jp>
Obtained from: NetBSD

23 years agoRemove the SMART ioctls, they are not in the official sources yet..
sos [Fri, 16 Mar 2001 07:23:59 +0000 (07:23 +0000)]
Remove the SMART ioctls, they are not in the official sources yet..

This should restore world...

23 years agoDIS_8859_15 -> ISO_8859-15 rename
ache [Fri, 16 Mar 2001 07:13:01 +0000 (07:13 +0000)]
DIS_8859_15 -> ISO_8859-15 rename

23 years agoResync to GENERIC. This backs out the spammage that I committed wrt
imp [Fri, 16 Mar 2001 06:33:44 +0000 (06:33 +0000)]
Resync to GENERIC.  This backs out the spammage that I committed wrt
killing ipv6 and some other things.

This makes GENERIC and NEWCARD the same, with OLDCARD stuff commented
out and the NEWCARD stuff included.  For the moment, pcic is commented
out (which has a old).  Plus invariants.  Plus ddb.

23 years agoVery small cosmetic tweak - avoid starting an extra xterm for the fvwm
jkh [Fri, 16 Mar 2001 03:32:13 +0000 (03:32 +0000)]
Very small cosmetic tweak - avoid starting an extra xterm for the fvwm
desktop case.

23 years ago/modules -> /boot/kernel
dd [Fri, 16 Mar 2001 02:26:48 +0000 (02:26 +0000)]
/modules -> /boot/kernel

PR: 25053
Approved by: nik

23 years agoMake it clear that the -v option doesn't actually set the date.
dd [Fri, 16 Mar 2001 02:20:24 +0000 (02:20 +0000)]
Make it clear that the -v option doesn't actually set the date.

PR: 24773
Approved by: nik

23 years agoExplain that TCP fragments with an offset of 1 are reported as being
dd [Fri, 16 Mar 2001 01:28:11 +0000 (01:28 +0000)]
Explain that TCP fragments with an offset of 1 are reported as being
dropped by rule -1 if logging is enabled.

PR: 25796
Submitted by: Crist J. Clark <cjclark@alum.mit.edu>
Approved by: nik

23 years agoCorrect descriptions of SOCK_RDM and SOCK_SEQPACKET.
dd [Fri, 16 Mar 2001 01:18:03 +0000 (01:18 +0000)]
Correct descriptions of SOCK_RDM and SOCK_SEQPACKET.

PR: 25797
Submitted by: Yuko Sasaki <yuko@veltec.co.jp>
Approved by: nik

23 years agoFix a comment
brian [Thu, 15 Mar 2001 23:25:55 +0000 (23:25 +0000)]
Fix a comment

PR: 25831
Submitted by: quinot@inf.enst.fr

23 years agoNote rules of enguagement.
obrien [Thu, 15 Mar 2001 23:20:13 +0000 (23:20 +0000)]
Note rules of enguagement.

23 years agoUse the newly introduced -W flag to netstat(1) to avoid truncated addresses.
des [Thu, 15 Mar 2001 20:46:35 +0000 (20:46 +0000)]
Use the newly introduced -W flag to netstat(1) to avoid truncated addresses.

23 years agoAdd a -W flag that tells netstat not to truncate addresses even if they are
des [Thu, 15 Mar 2001 20:46:04 +0000 (20:46 +0000)]
Add a -W flag that tells netstat not to truncate addresses even if they are
too long for the column they're printed in.

Move variable definitions out of netstat.h and into main.c.

Clean up some warnings.

23 years agoCorrect spelling of MNT_ASYNC.
dwmalone [Thu, 15 Mar 2001 19:52:04 +0000 (19:52 +0000)]
Correct spelling of MNT_ASYNC.

PR: 25835
Submitted by: Tony Finch <dot@dotat.at>

23 years agoMake this compile with ${BDECFLAGS}, apply style(9), cleanup manpage.
ru [Thu, 15 Mar 2001 19:28:58 +0000 (19:28 +0000)]
Make this compile with ${BDECFLAGS}, apply style(9), cleanup manpage.

23 years agoLimit the number of paths that glob can return to MAX_GLOBENTRIES, which
jlemon [Thu, 15 Mar 2001 18:50:32 +0000 (18:50 +0000)]
Limit the number of paths that glob can return to MAX_GLOBENTRIES, which
is currently set to 10000.  This is intended to prevent glob from running
amok when a highly recursive path is provided (such as "../*/../*/../*/...")

Reviewed by: Diane Bruce <db@db.net>, jhb

23 years agoHmm, the last commit apparently only made it halfways
sos [Thu, 15 Mar 2001 16:43:55 +0000 (16:43 +0000)]
Hmm, the last commit apparently only made it halfways

23 years agoAdd atacontrol, a util to control variuos aspects of the ATA.4 driver,
sos [Thu, 15 Mar 2001 15:40:53 +0000 (15:40 +0000)]
Add atacontrol, a util to control variuos aspects of the ATA.4 driver,
please consult atacontrol.8 for usage..

23 years agoRemove the ATA_ENABLE_ATAPI_DMA, ATA_ENABLE_WC and ATA_ENABLE_TAGS
sos [Thu, 15 Mar 2001 15:38:57 +0000 (15:38 +0000)]
Remove the ATA_ENABLE_ATAPI_DMA, ATA_ENABLE_WC and ATA_ENABLE_TAGS
options, add the tuneables for this purpose.

23 years agoRemove the ATA_ENABLE_ATAPI_DMA, ATA_ENABLE_WC and ATA_ENABLE_TAGS
sos [Thu, 15 Mar 2001 15:37:26 +0000 (15:37 +0000)]
Remove the ATA_ENABLE_ATAPI_DMA, ATA_ENABLE_WC and ATA_ENABLE_TAGS
options, use the tuneables listed in ata.4 instead.

23 years agoProvide the interface to atacontrol and associated logic.
sos [Thu, 15 Mar 2001 15:36:25 +0000 (15:36 +0000)]
Provide the interface to atacontrol and associated logic.

see atacontrol(8) for more.

Also the ATA_ENABLE_ATAPI_DMA, ATA_ENABLE_WC and ATA_ENABLE_TAGS
options are gone, use the tuneables listed in ata.4 instead from
the loader (this makes it possible to switch off DMA before the
driver has to touch the devices on broken hardware).

23 years agoAdd the ata control device
sos [Thu, 15 Mar 2001 15:32:42 +0000 (15:32 +0000)]
Add the ata control device

23 years agoAdd the ata control device.
sos [Thu, 15 Mar 2001 15:26:38 +0000 (15:26 +0000)]
Add the ata control device.

23 years agonet/route.c:
ru [Thu, 15 Mar 2001 14:52:12 +0000 (14:52 +0000)]
net/route.c:

  A route generated from an RTF_CLONING route had the RTF_WASCLONED flag
  set but did not have a reference to the parent route, as documented in
  the rtentry(9) manpage.  This prevented such routes from being deleted
  when their parent route is deleted.

  Now, for example, if you delete an IP address from a network interface,
  all ARP entries that were cloned from this interface route are flushed.

  This also has an impact on netstat(1) output.  Previously, dynamically
  created ARP cache entries (RTF_STATIC flag is unset) were displayed as
  part of the routing table display (-r).  Now, they are only printed if
  the -a option is given.

netinet/in.c, netinet/in_rmx.c:

  When address is removed from an interface, also delete all routes that
  point to this interface and address.  Previously, for example, if you
  changed the address on an interface, outgoing IP datagrams might still
  use the old address.  The only solution was to delete and re-add some
  routes.  (The problem is easily observed with the route(8) command.)

  Note, that if the socket was already bound to the local address before
  this address is removed, new datagrams generated from this socket will
  still be sent from the old address.

PR: kern/20785, kern/21914
Reviewed by: wollman (the idea)

23 years agoAdd new ATA specific file, to be used with the upcoming atacontrol.
sos [Thu, 15 Mar 2001 13:56:08 +0000 (13:56 +0000)]
Add new ATA specific file, to be used with the upcoming atacontrol.

23 years agoMoved '#endif' to correct position. (previous commit broke pc98 support)
nyan [Thu, 15 Mar 2001 12:33:52 +0000 (12:33 +0000)]
Moved '#endif' to correct position. (previous commit broke pc98 support)

23 years agoWhen creating a package sort dependencies in such a way that if dependency
sobomax [Thu, 15 Mar 2001 10:47:00 +0000 (10:47 +0000)]
When creating a package sort dependencies in such a way that if dependency
A depends on dependency B then dependency A will be in all cases listed
before B, so ``pkg_add -r'' will fetch/install packages in the correct order.

Previously dependencies were sorted just by its names, which is why
``pkg_add -r'' never actually worked properly.

To be usefull, hovewer, this fix requires that all packages have been
rebuilt, so it will take some time until users would be able to feel
posititive improvements. For the same reasons it is desirable to propagate
these changes to the 4-stable package building cluster *before* 4.3 ports
freeze, so packages for 4.3-RELEASE would be properly prepared.

Prompted by: kris
Insanely appreciated by: obrien
Silently approved by: jkh, -ports

23 years agowith my mentor hat on...
billf [Thu, 15 Mar 2001 10:33:00 +0000 (10:33 +0000)]
with my mentor hat on...

remove the concept of a 'maintainer' of our make. there really isn't a
need for any one committer to hold an exclusive lock or serve as a filter
for this code.

23 years agoRevert previous change -- apparently it's not quite right. It broke
will [Thu, 15 Mar 2001 10:22:50 +0000 (10:22 +0000)]
Revert previous change -- apparently it's not quite right.  It broke
src/sys/modules/if_ef and possibly other things.  I tested the build with
a make based on rev. 1.26, and it worked fine.  Since I'm not particularly
inclined to figure out what's going on with this, it's probably prudent
just to back it out for now.

Found by: jkh
Suggested by: jhay

23 years agoFix double mention of ssh.
asmodai [Thu, 15 Mar 2001 09:24:40 +0000 (09:24 +0000)]
Fix double mention of ssh.

This file is already off the vendorbranch, nonetheless it needs to be
submitted back to the OpenSSH people.

PR: 25743
Submitted by: David Wolfskill <dhw@whistle.com>

23 years agoMention that MAXHOSTNAMELEN includes space for the NUL.
brian [Thu, 15 Mar 2001 08:35:53 +0000 (08:35 +0000)]
Mention that MAXHOSTNAMELEN includes space for the NUL.

23 years agoActually commit the new version of trimdomain *blush*
brian [Thu, 15 Mar 2001 08:30:05 +0000 (08:30 +0000)]
Actually commit the new version of trimdomain *blush*

Thanks for covering my blunder to: peter

23 years agoThe previous commit was based on knowledge obtained via Paul Saab.
peter [Thu, 15 Mar 2001 06:58:04 +0000 (06:58 +0000)]
The previous commit was based on knowledge obtained via Paul Saab.

23 years agoAdd a placeholder for the ServerWorks OSB4 device id. The SMBus
peter [Thu, 15 Mar 2001 06:56:51 +0000 (06:56 +0000)]
Add a placeholder for the ServerWorks OSB4 device id.  The SMBus
interface on this chip is compatable with the PIIX4.  The catch is that
this interferes with isab0 which wants to attach to the same PCI node.
It seems to work, but we only tested it on systems with no ISA cards.

23 years agoThe serverworks OSB4 pci->isa bridge has the same mapping register at
peter [Thu, 15 Mar 2001 06:51:45 +0000 (06:51 +0000)]
The serverworks OSB4 pci->isa bridge has the same mapping register at
offset 0x90 for the SMBus device as the PIIX4.

23 years agoKill the 4MB kernel limit dead. [I hope :-)].
peter [Thu, 15 Mar 2001 05:10:06 +0000 (05:10 +0000)]
Kill the 4MB kernel limit dead.  [I hope :-)].
For UP, we were using $tmp_stk as a stack from the data section.  If the
kernel text section grew beyond ~3MB, the data section would be pushed
beyond the temporary 4MB P==V mapping.  This would cause the trampoline
up to high memory to fault.  The hack workaround I did was to use all of
the page table pages that we already have while preparing the initial
P==V mapping, instead of just the first one.
For SMP, the AP bootstrap process suffered the same sort of problem and
got the same treatment.

MFC candidate - this breaks on 4.x just the same..

Thanks to: Richard Todd <rmtodd@ichotolot.servalan.com>

23 years agoJake essentially rewrote this. It is not by any stretch of the
peter [Thu, 15 Mar 2001 05:02:08 +0000 (05:02 +0000)]
Jake essentially rewrote this.  It is not by any stretch of the
imagination a derivative of what I did before.

23 years agoRegenerate after rwatson's commit to syscalls.master (rev 1.85)
peter [Thu, 15 Mar 2001 04:43:57 +0000 (04:43 +0000)]
Regenerate after rwatson's commit to syscalls.master (rev 1.85)

23 years agoThis include file has no business being here.
mdodd [Thu, 15 Mar 2001 03:38:20 +0000 (03:38 +0000)]
This include file has no business being here.

23 years agoDon't dump core when an attempt is made to login using protocol 2 with
green [Thu, 15 Mar 2001 03:15:18 +0000 (03:15 +0000)]
Don't dump core when an attempt is made to login using protocol 2 with
an invalid user name.

23 years agoo Update some of the kernel man pages associated with extended attributes
rwatson [Thu, 15 Mar 2001 03:13:58 +0000 (03:13 +0000)]
o Update some of the kernel man pages associated with extended attributes
  to reflect EA API change to explicit namespacing.

Obtained from: TrustedBSD Project

23 years agoo Update getextattr and setextattr utilities to take into account the
rwatson [Thu, 15 Mar 2001 03:09:48 +0000 (03:09 +0000)]
o Update getextattr and setextattr utilities to take into account the
  revised EA interface with explicit namespacing.  Link against libutil
  to provide string/constant conversion for namespaces.  Document
  revised interface.

Obtained from: TrustedBSD Project

23 years agoo Update extattrctl to take into account the updated EA interface with
rwatson [Thu, 15 Mar 2001 03:04:35 +0000 (03:04 +0000)]
o Update extattrctl to take into account the updated EA interface with
  explicit namespaces.  Modify it to use libutil for string/constant
  namespace conversions.  Update the documentation to take into account
  the new interface.

Obtained from: TrustedBSD Project

23 years agoo To support new EA interface with explicit namespaces, introduce two
rwatson [Thu, 15 Mar 2001 03:00:39 +0000 (03:00 +0000)]
o To support new EA interface with explicit namespaces, introduce two
  utility functions which convert between string namespace names and
  numeric constants used by the interface.  Right now, two namespaces
  are supported, EXTATTR_NAMESPACE_SYSTEM ("system") and
  EXTATTR_NAMESPACE_USER ("user").  These functions are used by
  various userland EA utilities, rather than hard coding the routines
  all over the place.

Obtained from: TrustedBSD Project

23 years agoo Change the API and ABI of the Extended Attribute kernel interfaces to
rwatson [Thu, 15 Mar 2001 02:54:29 +0000 (02:54 +0000)]
o Change the API and ABI of the Extended Attribute kernel interfaces to
  introduce a new argument, "namespace", rather than relying on a first-
  character namespace indicator.  This is in line with more recent
  thinking on EA interfaces on various mailing lists, including the
  posix1e, Linux acl-devel, and trustedbsd-discuss forums.  Two namespaces
  are defined by default, EXTATTR_NAMESPACE_SYSTEM and
  EXTATTR_NAMESPACE_USER, where the primary distinction lies in the
  access control model: user EAs are accessible based on the normal
  MAC and DAC file/directory protections, and system attributes are
  limited to kernel-originated or appropriately privileged userland
  requests.

o These API changes occur at several levels: the namespace argument is
  introduced in the extattr_{get,set}_file() system call interfaces,
  at the vnode operation level in the vop_{get,set}extattr() interfaces,
  and in the UFS extended attribute implementation.  Changes are also
  introduced in the VFS extattrctl() interface (system call, VFS,
  and UFS implementation), where the arguments are modified to include
  a namespace field, as well as modified to advoid direct access to
  userspace variables from below the VFS layer (in the style of recent
  changes to mount by adrian@FreeBSD.org).  This required some cleanup
  and bug fixing regarding VFS locks and the VFS interface, as a vnode
  pointer may now be optionally submitted to the VFS_EXTATTRCTL()
  call.  Updated documentation for the VFS interface will be committed
  shortly.

o In the near future, the auto-starting feature will be updated to
  search two sub-directories to the ".attribute" directory in appropriate
  file systems: "user" and "system" to locate attributes intended for
  those namespaces, as the single filename is no longer sufficient
  to indicate what namespace the attribute is intended for.  Until this
  is committed, all attributes auto-started by UFS will be placed in
  the EXTATTR_NAMESPACE_SYSTEM namespace.

o The default POSIX.1e attribute names for ACLs and Capabilities have
  been updated to no longer include the '$' in their filename.  As such,
  if you're using these features, you'll need to rename the attribute
  backing files to the same names without '$' symbols in front.

o Note that these changes will require changes in userland, which will
  be committed shortly.  These include modifications to the extended
  attribute utilities, as well as to libutil for new namespace
  string conversion routines.  Once the matching userland changes are
  committed, a buildworld is recommended to update all the necessary
  include files and verify that the kernel and userland environments
  are in sync.  Note: If you do not use extended attributes (most people
  won't), upgrading is not imperative although since the system call
  API has changed, the new userland extended attribute code will no longer
  compile with old include files.

o Couple of minor cleanups while I'm there: make more code compilation
  conditional on FFS_EXTATTR, which should recover a bit of space on
  kernels running without EA's, as well as update copyright dates.

Obtained from: TrustedBSD Project

23 years agoFix make(1) bug: nested comments may be placed in .if, .else .if, and
will [Thu, 15 Mar 2001 02:51:11 +0000 (02:51 +0000)]
Fix make(1) bug:  nested comments may be placed in .if, .else .if, and
.endif statements but can't be placed in .elif.  Basically, the problem
was that ParseSkipLine() didn't handle comments the same way that
ParseReadLine() did, and thus you had errors with comments that are on a
conditional line (i.e. "^.") rather than a non-conditional line.

MFC candidate for 4.3-STABLE and 3.5-STABLE.

PR: 25627
Bug found by: jhs
Fix submitted by: Seth Kingsley <sethk@osd.bsdi.com>  (thanks!!)

23 years agoremove bogus check -- for kernel threads we fork off of proc0, not curproc
gallatin [Thu, 15 Mar 2001 02:32:26 +0000 (02:32 +0000)]
remove bogus check -- for kernel threads we fork off of proc0, not curproc
This was causing panics when modules which create kthreads were loaded
after boot.

pointed out by: jake, jhb

23 years agoRemove the last remnants of circle queues.
dd [Thu, 15 Mar 2001 02:08:44 +0000 (02:08 +0000)]
Remove the last remnants of circle queues.

PR: 25184
Submitted by: Alex Kapranoff <alex@kapran.bitmcnit.bryansk.su>
Approved by: nik

23 years ago/devs -> /dev
dd [Thu, 15 Mar 2001 02:06:44 +0000 (02:06 +0000)]
/devs -> /dev

PR: 25252
Submitted by: Christian Weisgerber <naddy@mips.inka.de>
Approved by: nik, phk (via nik)

23 years agoMake tdelete(3), tfind(3), and twalk(3) links to tsearch(3).
dd [Thu, 15 Mar 2001 01:53:17 +0000 (01:53 +0000)]
Make tdelete(3), tfind(3), and twalk(3) links to tsearch(3).

Approved by: nik

23 years agoIt would help if trimdomain.c was actually committed. This is a stopgap
peter [Thu, 15 Mar 2001 00:15:22 +0000 (00:15 +0000)]
It would help if trimdomain.c was actually committed.  This is a stopgap
world-unbreaker until Brian Somers commits the one he intended to.

Pointy Hat to: brian

23 years agoDon't set the gateway address if the netmask is zero or we're on
ps [Wed, 14 Mar 2001 23:53:25 +0000 (23:53 +0000)]
Don't set the gateway address if the netmask is zero or we're on
the same network.  PXE does not do netmask calculations, so if the
gateway is set it will use it.

Submitted by: peter & FreeBSD cluster ACLs

23 years agoUse better descriptions (ones invovling words from the English language
jhb [Wed, 14 Mar 2001 21:33:36 +0000 (21:33 +0000)]
Use better descriptions (ones invovling words from the English language
anyways) for the accept filter modules.

23 years agoMAXHOSTNAMELEN includes space for a NUL.
brian [Wed, 14 Mar 2001 20:51:26 +0000 (20:51 +0000)]
MAXHOSTNAMELEN includes space for a NUL.
Don't roll our own version of trimdomain(), use the one in libutil.

Not objected to by: freebsd-audit

23 years agoMove trimdomain() into it's own source file and tidy things up a bit.
brian [Wed, 14 Mar 2001 20:51:15 +0000 (20:51 +0000)]
Move trimdomain() into it's own source file and tidy things up a bit.
Fix disorder in the Makefile.

Reviewed (mostly) by: bde

23 years agoMAXHOSTNAMELEN includes space for the NUL
brian [Wed, 14 Mar 2001 20:50:52 +0000 (20:50 +0000)]
MAXHOSTNAMELEN includes space for the NUL
Don't read past the end of the host passed to realhostname()

Not objected to by: freebsd-audit
Interface disliked by: imp

23 years agoAdd some performance features to the fxp driver. If the chip is not
jlemon [Wed, 14 Mar 2001 19:50:35 +0000 (19:50 +0000)]
Add some performance features to the fxp driver.  If the chip is not
a 82557 (e.g.: a newer chip) then:

   + enable MWI, if the PCI configuration indicates the system supports it
   + enable usage of extended TxCB, for better performance
   + enable hardware flow control.  FC frames will be passed up to the
     host only if promiscuous mode is enabled.

23 years agoWhen matching installed packages against glob keep track of which patterns
sobomax [Wed, 14 Mar 2001 19:46:43 +0000 (19:46 +0000)]
When matching installed packages against glob keep track of which patterns
actually triggered a match and which did not, and add patterns that didn't
into resulting list, so caller will have a chance to notify user that package
isn't installed. This should fix current, POLA-breaking behaviour when user
doesn't receive a notification if he specifies several packages, some of which
aren't installed.

23 years agoTypo. Move FreeBSD Id out of copyright section.
charnier [Wed, 14 Mar 2001 19:39:29 +0000 (19:39 +0000)]
Typo. Move FreeBSD Id out of copyright section.

23 years agoo In my merge, missed the one-line patch to ufs_vnops.c that removed
rwatson [Wed, 14 Mar 2001 18:27:04 +0000 (18:27 +0000)]
o In my merge, missed the one-line patch to ufs_vnops.c that removed
  the static prototype for ufs_readdir().  Note that ufs_readdir() was
  actually already non-static, the prototype was incorrect.

Submitted by: jedgar

23 years agoImprove comments about the sound drivers.
dcs [Wed, 14 Mar 2001 17:46:32 +0000 (17:46 +0000)]
Improve comments about the sound drivers.
Correct accf lines.

23 years agoMinor cleanup to the previous commit:
sos [Wed, 14 Mar 2001 14:00:09 +0000 (14:00 +0000)]
Minor cleanup to the previous commit:

Print what devices went away and which arrived.

Avoid timeout loop on missing status.

23 years agoAdd support for the IOMEGA Clik!
sos [Wed, 14 Mar 2001 12:12:46 +0000 (12:12 +0000)]
Add support for the IOMEGA Clik!

IOMEGA deserves a medal for making the most nonstandard ATAPI
devices, if they are ignorant or just not smart enough I don't
know, but somebody should help them out of their misery...

23 years agoRefine the detach/attach code.
sos [Wed, 14 Mar 2001 12:05:44 +0000 (12:05 +0000)]
Refine the detach/attach code.

Proberly fail outstanding bio requests on devices that are detached.

This makes it possible to change between disk/cdrom/dvd/whathaveyou
in a notebook, just by suspending it, changing the device in the
bay (or what you model calls it), unsuspend and the ATA driver
will figure out what disappeared and properly fail those, and attach
any new devices found.

23 years agoAdd a commented out entry for compat4x
murray [Wed, 14 Mar 2001 11:30:57 +0000 (11:30 +0000)]
Add a commented out entry for compat4x

PR: 25196
Submitted by: Chris Knight <chris@aims.com.au>

23 years agoDocument the HTTP proxy install method.
murray [Wed, 14 Mar 2001 11:08:28 +0000 (11:08 +0000)]
Document the HTTP proxy install method.

PR: 21451

23 years agoDocument the fact that you can install through an HTTP proxy
murray [Wed, 14 Mar 2001 11:01:05 +0000 (11:01 +0000)]
Document the fact that you can install through an HTTP proxy

PR: 21451
Approved by: jkh

23 years agoAttempt to fix the problem with -j builds, and du-uglify the asm code
kris [Wed, 14 Mar 2001 10:10:11 +0000 (10:10 +0000)]
Attempt to fix the problem with -j builds, and du-uglify the asm code
generation and assembly targets.

Help from: bde, obrien

23 years agoAvoid sending a pointless `CWD .' command which confuses some
murray [Wed, 14 Mar 2001 10:05:48 +0000 (10:05 +0000)]
Avoid sending a pointless `CWD .' command which confuses some
non-standard ftp servers.

PR: 21856

23 years agoArgh! Why can't I stop breaking the fvwm desktop option? *I* use
jkh [Wed, 14 Mar 2001 09:39:54 +0000 (09:39 +0000)]
Argh!  Why can't I stop breaking the fvwm desktop option?  *I* use
fvwm as my desktop, yet I've broken this damn thing 3 times in a row
now while all the desktops I don't actually use continue to work fine! :)

23 years ago- Do not touch the Giant.
tanimura [Wed, 14 Mar 2001 07:29:47 +0000 (07:29 +0000)]
- Do not touch the Giant.

Requested by: jhb

- Initialize a mutex prior to locking it.

Spotted by: dcs

23 years agoFix a lock reversal problem in the VM subsystem related to threaded
dillon [Wed, 14 Mar 2001 06:48:53 +0000 (06:48 +0000)]
Fix a lock reversal problem in the VM subsystem related to threaded
programs.   There is a case during a fork() which can cause a deadlock.

From Tor -
The workaround that consists of setting a flag in the vm map that
indicates that a fork is in progress and using that mark in the page
fault handling to force a revalidation failure.  That change will only
affect (pessimize) page fault handling during fork for threaded
(linuxthreads style) applications and applications using aio_*().

Submited by: tegge

23 years agoTemporarily remove the vm_map_simplify() call from vm_map_insert(). The
dillon [Wed, 14 Mar 2001 06:09:42 +0000 (06:09 +0000)]
Temporarily remove the vm_map_simplify() call from vm_map_insert().  The
call is correct, but it interferes with the massive hack called
vm_map_growstack().  The call will be returned after our stack handling
code is fixed.

Reported by: tegge

23 years agoo Implement "options FFS_EXTATTR_AUTOSTART", which depends on
rwatson [Wed, 14 Mar 2001 05:32:31 +0000 (05:32 +0000)]
o Implement "options FFS_EXTATTR_AUTOSTART", which depends on
  "options FFS_EXTATTR".  When extended attribute auto-starting
  is enabled, FFS will scan the .attribute directory off of the
  root of each file system, as it is mounted.  If .attribute
  exists, EA support will be started for the file system.  If
  there are files in the directory, FFS will attempt to start
  them as attribute backing files for attributes baring the same
  name.  All attributes are started before access to the file
  system is permitted, so this permits race-free enabling of
  attributes.  For attributes backing support for security
  features, such as ACLs, MAC, Capabilities, this is vital, as
  it prevents the file system attributes from getting out of
  sync as a result of file system operations between mount-time
  and the enabling of the extended attribute.  The userland
  extattrctl tool will still function exactly as previously.
  Files must be placed directly in .attribute, which must be
  directly off of the file system root: symbolic links are
  not permitted.  FFS_EXTATTR will continue to be able
  to function without FFS_EXTATTR_AUTOSTART for sites that do not
  want/require auto-starting.  If you're using the UFS_ACL code
  available from www.TrustedBSD.org, using FFS_EXTATTR_AUTOSTART
  is recommended.

o This support is implemented by adding an invocation of
  ufs_extattr_autostart() to ffs_mountfs().  In addition,
  several new supporting calls are introduced in
  ufs_extattr.c:

    ufs_extattr_autostart(): start EAs on the specified mount
    ufs_extattr_lookup(): given a directory and filename,
                          return the vnode for the file.
    ufs_extattr_enable_with_open(): invoke ufs_extattr_enable()
                          after doing the equililent of vn_open()
                          on the passed file.
    ufs_extattr_iterate_directory(): iterate over a directory,
                          invoking ufs_extattr_lookup() and
                          ufs_extattr_enable_with_open() on each
                          entry.

o This feature is not widely tested, and therefore may contain
  bugs, caution is advised.  Several changes are in the pipeline
  for this feature, including breaking out of EA namespaces into
  subdirectories of .attribute (this is waiting on the updated
  EA API), as well as a per-filesystem flag indicating whether
  or not EAs should be auto-started.  This is required because
  administrators may not want .attribute auto-started on all
  file systems, especially if non-administrators have write access
  to the root of a file system.

Obtained from: TrustedBSD Project

23 years agoupdate this to be marginally closer to reality
mjacob [Wed, 14 Mar 2001 04:31:36 +0000 (04:31 +0000)]
update this to be marginally closer to reality

23 years agoClean up usage- ct_reserved is really ct_syshandle now.
mjacob [Wed, 14 Mar 2001 04:14:58 +0000 (04:14 +0000)]
Clean up usage- ct_reserved is really ct_syshandle now.

23 years agoFirst cut of target mode swizzling for non-little endian machines.
mjacob [Wed, 14 Mar 2001 04:14:22 +0000 (04:14 +0000)]
First cut of target mode swizzling for non-little endian machines.
It's probably wrong but it's a start.

23 years agoMote that how the pad bytes can be divided in half and used by either
mjacob [Wed, 14 Mar 2001 04:13:30 +0000 (04:13 +0000)]
Mote that how the pad bytes can be divided in half and used by either
the target mode code or outer layers.

Increase cd_tagval to be 32 bits since it will have to now carry 16
bits of parallel SCSI ATIO handle as well as a normal tag (if any).

23 years agoIn order to save ourselves grief with the SUNPRO compiler under
mjacob [Wed, 14 Mar 2001 04:11:56 +0000 (04:11 +0000)]
In order to save ourselves grief with the SUNPRO compiler under
Solaris (which, for reasons unknown to me, chokes on u_int16_t
as a typedef of unsigned short if used in a transitional (mixed K&R
and ANSI) way), we'll go the extra mile and fully ANSIfy things.

23 years agoOK, *now* we only sort the file once (red face).
jkh [Wed, 14 Mar 2001 03:26:40 +0000 (03:26 +0000)]
OK, *now* we only sort the file once (red face).

23 years agoDocument the lock and unlock commands. This fixes a segmentation fault.
dd [Wed, 14 Mar 2001 03:24:30 +0000 (03:24 +0000)]
Document the lock and unlock commands.  This fixes a segmentation fault.

PR: 25187
Approved by: nik

23 years agoDocument the lock and unlock commands.
dd [Wed, 14 Mar 2001 03:22:28 +0000 (03:22 +0000)]
Document the lock and unlock commands.

PR: 25187
Approved by: nik

23 years agoPoint people looking to upgrade from source at "The Cutting Edge" in
dd [Wed, 14 Mar 2001 03:12:06 +0000 (03:12 +0000)]
Point people looking to upgrade from source at "The Cutting Edge" in
the Handbook, rather than some nonexistent tutorial.

PR: 25556
Approved by: nik

23 years agoIn xpt_set_transfer_settings(), force async if either the period or the
ken [Tue, 13 Mar 2001 22:35:28 +0000 (22:35 +0000)]
In xpt_set_transfer_settings(), force async if either the period or the
offset is set to 0.

Re-arrange the DT limiting code so that we don't end up setting the period
to 0xa if the user really wants async.  The previous behavior seemed to
confuse the aic(4) driver.

PR: kern/22733
Reviewed by: gibbs

23 years agoRead information from local rc.conf file(s) to get proper settings for
gshapiro [Tue, 13 Mar 2001 20:58:39 +0000 (20:58 +0000)]
Read information from local rc.conf file(s) to get proper settings for
'make start'.

PR: conf/25639
Submitted by: Esa Karkkainen <ejk@pp.htv.fi>

23 years agofix a panic triggerable by anyone with read/write access to the audio
cg [Tue, 13 Mar 2001 18:43:22 +0000 (18:43 +0000)]
fix a panic triggerable by anyone with read/write access to the audio
devices.  opening /dev/{dsp,dspW,audio}0 and then opening a different device
from that list and closing it resulted in a panic when any operation is
performed on the first fd.

we prevent this happening by denying the second open unless it uses the same
minor device as the first.

PR: kern/25519

23 years agoRFC768 (UDP) requires that "if the computed checksum is zero, it
ru [Tue, 13 Mar 2001 17:07:06 +0000 (17:07 +0000)]
RFC768 (UDP) requires that "if the computed checksum is zero, it
is transmitted as all ones".  This got broken after introduction
of delayed checksums as follows.  Some guys (including Jonathan)
think that it is allowed to transmit all ones in place of a zero
checksum for TCP the same way as for UDP.  (The discussion still
takes place on -net.)  Thus, the 0 -> 0xffff checksum fixup was
first moved from udp_output() (see udp_usrreq.c, 1.64 -> 1.65)
to in_cksum_skip() (see sys/i386/i386/in_cksum.c, 1.17 -> 1.18,
INVERT expression).  Besides that I disagree that it is valid for
TCP, there was no real problem until in_cksum.c,v 1.20, where the
in_cksum() was made just a special version of in_cksum_skip().
The side effect was that now every incoming IP datagram failed to
pass the checksum test (in_cksum() returned 0xffff when it should
actually return zero).  It was fixed next day in revision 1.21,
by removing the INVERT expression.  The latter also broke the
0 -> 0xffff fixup for UDP checksums.

Before this change:
: tcpdump: listening on lo0
: 127.0.0.1.33005 > 127.0.0.1.33006:  udp 0 (ttl 64, id 1)
:                          4500 001c 0001 0000 4011 7cce 7f00 0001
:                          7f00 0001 80ed 80ee 0008 0000

After this change:
: tcpdump: listening on lo0
: 127.0.0.1.33005 > 127.0.0.1.33006:  udp 0 (ttl 64, id 1)
:                          4500 001c 0001 0000 4011 7cce 7f00 0001
:                          7f00 0001 80ed 80ee 0008 ffff

23 years ago- Sync up with stable by adding in the Netgraph and Sound module sections,
jhb [Tue, 13 Mar 2001 16:16:35 +0000 (16:16 +0000)]
- Sync up with stable by adding in the Netgraph and Sound module sections,
  the agp module, and the accept filter modules.
- Remove an extraneous blank line.

23 years agoFix typo: compatability -> compatibility
olgeni [Tue, 13 Mar 2001 14:15:16 +0000 (14:15 +0000)]
Fix typo: compatability -> compatibility

23 years agoCount and show incoming UDP datagrams with no checksum.
ru [Tue, 13 Mar 2001 13:26:06 +0000 (13:26 +0000)]
Count and show incoming UDP datagrams with no checksum.

23 years agoMultiply empty lines pasted as single line, fix it
ache [Tue, 13 Mar 2001 10:23:01 +0000 (10:23 +0000)]
Multiply empty lines pasted as single line, fix it

23 years agoFix a race condition in the transmit path caused by the controller
phk [Tue, 13 Mar 2001 09:29:37 +0000 (09:29 +0000)]
Fix a race condition in the transmit path caused by the controller
caching a bit more than we were told initially.

More statistics gathering as well.

23 years agoDont call device close and ioctl functions if device has disappeared.
sos [Tue, 13 Mar 2001 08:45:05 +0000 (08:45 +0000)]
Dont call device close and ioctl functions if device has disappeared.

Reviewed by: phk

23 years agoDocument (with warnings) the otherwise undocumented `F' command.
jkh [Tue, 13 Mar 2001 06:52:07 +0000 (06:52 +0000)]
Document (with warnings) the otherwise undocumented `F' command.

23 years agoBe a better rc.conf citizen and create an initial file which:
jkh [Tue, 13 Mar 2001 06:42:12 +0000 (06:42 +0000)]
Be a better rc.conf citizen and create an initial file which:

1. Has a time-stamp to show when it was created

2. Sorts and uniq's the output to only contain single instances of a
   given setting. This doesn't mean you still can't have settings which
   override one another, that's still possible since it's too much
   trouble to do the redundancy checking here.

Requested by:  lots of people

23 years agoCommit some tweaks I have had laying around my tree for over a year now.
peter [Tue, 13 Mar 2001 05:56:48 +0000 (05:56 +0000)]
Commit some tweaks I have had laying around my tree for over a year now.
Always set the APM "device" description.  Some minor style tweaks.

23 years agoBy convention, the moduledata is static unless there is a reason for it
peter [Tue, 13 Mar 2001 05:55:43 +0000 (05:55 +0000)]
By convention, the moduledata is static unless there is a reason for it
to not be.