]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoThis commit was generated by cvs2svn to compensate for changes in r74473,
David E. O'Brien [Mon, 19 Mar 2001 19:46:16 +0000 (19:46 +0000)]
This commit was generated by cvs2svn to compensate for changes in r74473,
which included commits to RCS files with non-trunk default branches.

23 years agoVirgin import of the GCC 2.95.3 compilers
David E. O'Brien [Mon, 19 Mar 2001 19:46:16 +0000 (19:46 +0000)]
Virgin import of the GCC 2.95.3 compilers

23 years agoAdd missing comma when KERBEROS is defined.
Alfred Perlstein [Mon, 19 Mar 2001 19:36:11 +0000 (19:36 +0000)]
Add missing comma when KERBEROS is defined.

23 years agoTeach ftpd about the new GLOB_MAXPATH flag.
Jonathan Lemon [Mon, 19 Mar 2001 19:11:00 +0000 (19:11 +0000)]
Teach ftpd about the new GLOB_MAXPATH flag.

23 years agoIntroduce the GLOB_MAXPATH flag, which allows the user to limit the
Jonathan Lemon [Mon, 19 Mar 2001 19:10:06 +0000 (19:10 +0000)]
Introduce the GLOB_MAXPATH flag, which allows the user to limit the
number of paths which glob(3) will return.  Remove the hardcoded limit
from the last commit, which restores the previous unbounded behavior.

Document the new flag in the manual page.

23 years agoo Missed in addition of setfacl.h in setfacl(1)
Robert Watson [Mon, 19 Mar 2001 18:59:36 +0000 (18:59 +0000)]
o Missed in addition of setfacl.h in setfacl(1)

23 years agoo Missed in prior commit: getfacl(1) Makefile
Robert Watson [Mon, 19 Mar 2001 18:58:16 +0000 (18:58 +0000)]
o Missed in prior commit: getfacl(1) Makefile

23 years agoo Enable getfacl(1) and setfacl(1)
Robert Watson [Mon, 19 Mar 2001 18:09:56 +0000 (18:09 +0000)]
o Enable getfacl(1) and setfacl(1)

Obtained from: TrustedBSD Project

23 years agoo POSIX.2c Userland tool support for POSIX.1e ACLs -- getfacl retrieves ACLs
Robert Watson [Mon, 19 Mar 2001 18:09:25 +0000 (18:09 +0000)]
o POSIX.2c Userland tool support for POSIX.1e ACLs -- getfacl retrieves ACLs
  from files and directories, and setfacl sets ACLs on files and directories.

Submitted by: jedgar
Obtained from: TrustedBSD Project

23 years agoTry not to engage to ATA channels that are disabled by the BIOS.
Søren Schmidt [Mon, 19 Mar 2001 13:31:58 +0000 (13:31 +0000)]
Try not to engage to ATA channels that are disabled by the BIOS.

23 years agoBring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and
Alfred Perlstein [Mon, 19 Mar 2001 12:50:13 +0000 (12:50 +0000)]
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and
associated changes that had to happen to make this possible as well as
bugs fixed along the way.

  Bring in required TLI library routines to support this.

  Since we don't support TLI we've essentially copied what NetBSD
  has done, adding a thin layer to emulate direct the TLI calls
  into BSD socket calls.

  This is mostly from Sun's tirpc release that was made in 1994,
  however some fixes were backported from the 1999 release (supposedly
  only made available after this porting effort was underway).

  The submitter has agreed to continue on and bring us up to the
  1999 release.

  Several key features are introduced with this update:
    Client calls are thread safe. (1999 code has server side thread
    safe)
    Updated, a more modern interface.

  Many userland updates were done to bring the code up to par with
  the recent RPC API.

  There is an update to the pthreads library, a function
  pthread_main_np() was added to emulate a function of Sun's threads
  library.

  While we're at it, bring in NetBSD's lockd, it's been far too
  long of a wait.

  New rpcbind(8) replaces portmap(8) (supporting communication over
  an authenticated Unix-domain socket, and by default only allowing
  set and unset requests over that channel). It's much more secure
  than the old portmapper.

  Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded
  to support TI-RPC and to support IPV6.

  Umount(8) is also fixed to unmount pathnames longer than 80 chars,
  which are currently truncated by the Kernel statfs structure.

Submitted by: Martin Blapp <mb@imp.ch>
Manpage review: ru
Secure RPC implemented by: wpaul

23 years agoOn open create all the CD physical track devices according to the
Søren Schmidt [Mon, 19 Mar 2001 12:02:36 +0000 (12:02 +0000)]
On open create all the CD physical track devices according to the
TOC read from the CD, instead of cloning them when asked to.

23 years agoOnly allow root to attach/detach/etc ATA/ATAPI devices.
Søren Schmidt [Mon, 19 Mar 2001 11:55:13 +0000 (11:55 +0000)]
Only allow root to attach/detach/etc ATA/ATAPI devices.

23 years agoImplement D_MD_ORDER (local extension) to get month/day order from locale
Andrey A. Chernov [Mon, 19 Mar 2001 11:55:12 +0000 (11:55 +0000)]
Implement D_MD_ORDER (local extension) to get month/day order from locale

23 years agoAdd D_MD_ORDER (local extension) to get month/day order from locale
Andrey A. Chernov [Mon, 19 Mar 2001 11:53:34 +0000 (11:53 +0000)]
Add D_MD_ORDER (local extension) to get month/day order from locale

23 years agoUpdate obsolete history info.
Ruslan Ermilov [Mon, 19 Mar 2001 11:05:32 +0000 (11:05 +0000)]
Update obsolete history info.

23 years agoInvalidate cached forwarding route (ipforward_rt) whenever a new route
Ruslan Ermilov [Mon, 19 Mar 2001 09:16:16 +0000 (09:16 +0000)]
Invalidate cached forwarding route (ipforward_rt) whenever a new route
is added to the routing table, otherwise we may end up using the wrong
route when forwarding.

PR: kern/10778
Reviewed by: silence on -net

23 years agoAdd an example of a novel technique to the IPv6 code. This technique seems
Poul-Henning Kamp [Mon, 19 Mar 2001 08:51:56 +0000 (08:51 +0000)]
Add an example of a novel technique to the IPv6 code.  This technique seems
to have been developed after the IPv6 code was written, it's called "checking
for error returns".

Now netstat at least doesnt't go beserk in jails.

23 years agoFix minor mdoc(7) and spelling nits.
Ruslan Ermilov [Mon, 19 Mar 2001 08:20:34 +0000 (08:20 +0000)]
Fix minor mdoc(7) and spelling nits.

23 years agomdoc(7) police:
Ruslan Ermilov [Mon, 19 Mar 2001 08:08:22 +0000 (08:08 +0000)]
mdoc(7) police:

- lowercase Nd argument
- mark function arguments with Fa
- mark defined values with Dv
- simply copying POSIX text for RETURN VALUES and ERRORS sections is not
  always a good idea.  POSIX uses the word "shall" indicating the behavior
  the correct implementation should follow.

23 years agoAdd sysctls for reading the tunables as suggested by des.
Søren Schmidt [Mon, 19 Mar 2001 08:04:54 +0000 (08:04 +0000)]
Add sysctls for reading the tunables as suggested by des.
Minor cleanups plus checks of the ->active state.
Cosmetics.

23 years agoCleanup the alloc/release code a bit.
Søren Schmidt [Mon, 19 Mar 2001 08:02:42 +0000 (08:02 +0000)]
Cleanup the alloc/release code a bit.

Fix length error on the bmio resource.

Fix the irq release code, zero out free'd irq.

23 years agoActivate atacontrol
Søren Schmidt [Mon, 19 Mar 2001 07:59:38 +0000 (07:59 +0000)]
Activate atacontrol

23 years agoSome minor cleanups to the code, no new functionality.
Søren Schmidt [Mon, 19 Mar 2001 07:58:47 +0000 (07:58 +0000)]
Some minor cleanups to the code, no new functionality.

23 years agoIncorporate most of Ruslans changes plus minor addtion by me
Søren Schmidt [Mon, 19 Mar 2001 07:57:25 +0000 (07:57 +0000)]
Incorporate most of Ruslans changes plus minor addtion by me

23 years agoCosmetic changes.
Søren Schmidt [Mon, 19 Mar 2001 07:48:19 +0000 (07:48 +0000)]
Cosmetic changes.

23 years agoLots of minor cleanup, plus a couple of interesting things.
Warner Losh [Mon, 19 Mar 2001 07:10:38 +0000 (07:10 +0000)]
Lots of minor cleanup, plus a couple of interesting things.
o Attempt to disable the slot when we detect that there are problems with
  it in our ISR.  This should make polling mode work better for more cards,
  but more work may be needed.  This "disabling" sets the card interrupt
  register to 0.  This worked for me for lots of tests in polling mode.
o Now that I've found datasheets, fix a boatload of magic numbers in the
  source to make it easier to understand.
o Use a table of names rather than a big case statement.
o Cull a few of the "unused" controller types that we map to other times
  that were a vestiage of PAO code that we never merged in the same way.
o Enforce legal IRQs.  You are no longer allowed to try to use IRQs that
  will fail on all known ISA/PCI <-> PCMCIA bridges.  The bridges do not
  have pins for these illegal interrupts, and all of them are listed as
  reserved and/or illegeal in the datasheets depending on which one you
  look at.
o Add comments about how IBM-AT based computers and NEC PC-98 based computers
  map these interrupts and which ones are valid.
o Always clear the bit that steers the management interrupt either to the
  value listed in the PCIC_STAT_INT register.  I've seen this bit get set
  on suspend/resume and after windows boot, and it does't hurt to clear it.
  NOTE: this might mean we can share this interrupt in the future.

23 years agoo Adding UFS_EXTATTR support in MFS required the inclusion of opt_ufs.h.
Robert Watson [Mon, 19 Mar 2001 07:04:17 +0000 (07:04 +0000)]
o Adding UFS_EXTATTR support in MFS required the inclusion of opt_ufs.h.
  I didn't realize we had an MFS module, so didn't update the
  dependencies there to reflect opt_ufs.h.

Obtained from: TrustedBSD Project

23 years agoo Enable UFS-based extended attribute support on MFS. Note that this change
Robert Watson [Mon, 19 Mar 2001 06:44:18 +0000 (06:44 +0000)]
o Enable UFS-based extended attribute support on MFS.  Note that this change
  is under-tested, and that MFS appears to be in the process of being
  deprecated in favor of FFS over md.  Note also that UFS_EXTATTR_AUTOSTART
  doesn't make much sense on MFS unless the MFSROOT is compiled in, so
  manual configuration is generally required.

Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 06:00:41 +0000 (06:00 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word.

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 05:56:02 +0000 (05:56 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word.

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 05:53:39 +0000 (05:53 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 05:48:58 +0000 (05:48 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word.  Part 2 of syscalls.master commit to catch rebuilt
  files.

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 05:44:15 +0000 (05:44 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word.

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 05:05:41 +0000 (05:05 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word.

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoo Rename "namespace" argument to "attrnamespace" as namespace is a C++
Robert Watson [Mon, 19 Mar 2001 05:04:30 +0000 (05:04 +0000)]
o Rename "namespace" argument to "attrnamespace" as namespace is a C++
  reserved word, causing breakage when a C++ program included libutil.h
  This change will be propagated elsewhere shortly.

Submitted by: jkh
Obtained from: TrustedBSD Project

23 years agoThis is an MFC candidate.
Justin T. Gibbs [Mon, 19 Mar 2001 04:40:35 +0000 (04:40 +0000)]
This is an MFC candidate.

aic7xxx.c:
Correct code that traverses the phase table.  A much too quick
push to staticize this structure resulted in non-functional
lookup code.  This corrects the printing of the phase where
a timeout occurred.

aic7xxx.reg:
Use FIFOQWDEMP as the name for bit 5 of DFSTATUS just like
the Adaptec data books.

aic7xxx.seq:
Refine the 2.1 PCI retry bug workaround for certain, non-ULTRA2,
controllers.  When the DMA of an SCB completes, it can take
some time for HDONE to come true after MREQPEN (PCI memory request
pending) falls.  If HDONE never comes true, we are in the hung
state and must manually drain the FIFO.  We used to test HDONE for
3 clock cycles to detect this condition.  This works on all of the
hardware I can personally test.  Some controllers were reported
to take 4 clock cycles, so the last version of this code waited
4 clock cycles.  This still didn't work for everyone.  To fix this,
I've adjusted the work around so that even if the hardware hasn't
hung, but we run the work-around code, the result is a long winded
way to complete the transfer, rather than a hang.

23 years agoo Change options FFS_EXTATTR and options FFS_EXTATTR_AUTOSTART to
Robert Watson [Mon, 19 Mar 2001 04:35:40 +0000 (04:35 +0000)]
o Change options FFS_EXTATTR and options FFS_EXTATTR_AUTOSTART to
  options UFS_EXTATTR and UFS_EXTATTR_AUTOSTART respectively.  This change
  reflects the fact that our EA support is implemented entirely at the
  UFS layer (modulo FFS start/stop/autostart hooks for mount and unmount
  events).  This also better reflects the fact that [shortly] MFS will also
  support EAs, as well as possibly IFS.

o Consumers of the EA support in FFS are reminded that as a result, they
  must change kernel config files to reflect the new option names.

Obtained from: TrustedBSD Project

23 years agoAdd the following POSIX 1003.1e functions and man pages:
Chris D. Faulhaber [Mon, 19 Mar 2001 03:19:51 +0000 (03:19 +0000)]
Add the following POSIX 1003.1e functions and man pages:
  o acl_calc_mask(): calculates the ACL mask entry associated with
    the given ACL.
  o acl_delete_entry(): remove a specified ACL entry from the given
    ACL.

Approved by: rwatson

23 years agoShow the bzero() bandwidth in kBps instead of Bps; use u_int32_t instead
Dag-Erling Smørgrav [Mon, 19 Mar 2001 00:28:04 +0000 (00:28 +0000)]
Show the bzero() bandwidth in kBps instead of Bps; use u_int32_t instead
of long and int64_t; and print the result as an unsigned long. This should
make the output from the bzero() test more readable, and avoid printing a
negative bandwidth. Note that this doesn't change the decision process,
since that is based on time elapsed, not on computed bandwidth.

23 years agopcm driver for S3 Sonicvibes chipset.
Orion Hodson [Mon, 19 Mar 2001 00:26:41 +0000 (00:26 +0000)]
pcm driver for S3 Sonicvibes chipset.

Reviewed by: Cameron Grant

23 years agoDocument fix for 'going nowhere without my init' install bug.
Wilko Bulte [Sun, 18 Mar 2001 22:28:10 +0000 (22:28 +0000)]
Document fix for 'going nowhere without my init' install bug.

Thanks to: jkh / obrien

23 years agoLibraries should _never_ call exit() themselves (or its alternate spelling
David E. O'Brien [Sun, 18 Mar 2001 21:30:11 +0000 (21:30 +0000)]
Libraries should _never_ call exit() themselves (or its alternate spelling
`err()').  libdisk does! and additionally libdisk gets confused on Alpha
disks with foreign disklabels, throws up its hands and exits.  This is
the cause of the "going no where without my init" install bug on the Alpha.

So now on the Alpha, rather than call err(), we print the error string and
continue processing.

Submitted by: jkh

23 years agoCorrect typo in usage.
Hajimu UMEMOTO [Sun, 18 Mar 2001 18:45:43 +0000 (18:45 +0000)]
Correct typo in usage.

Submitted by: David Malone <dwmalone@maths.tcd.ie>

23 years agoUse a module name of 'if_lnc' in both the PCI and ISA sections of
Ian Dowse [Sun, 18 Mar 2001 17:44:23 +0000 (17:44 +0000)]
Use a module name of 'if_lnc' in both the PCI and ISA sections of
the driver. Doing this breaks the ability to unload the unneeded
parts of the driver (e.g unload the PCI section when using an ISA
card), but currently ifconfig(8) expects an interface `XXX' to have
a driver name of `if_XXX'.

PR: kern/25582
Submitted by: Alexander N. Kabaev <kabaev@mail.ru>, imp (apparently
Warner suggested a similar fix some time ago).
Reviewed by: paul (who would prefer to see ifconfig changed instead)

23 years agoFix typo: s/DRIVERMODULE/DRIVER_MODULE/
Alexander Langer [Sun, 18 Mar 2001 16:54:39 +0000 (16:54 +0000)]
Fix typo: s/DRIVERMODULE/DRIVER_MODULE/

PR: 25901
Submitted by: Maxime Henrion <mux@qualys.com>

23 years agoAdd `ipv6_ifconfig_IFN_aliasN' directive to allow multiple aliases.
Hajimu UMEMOTO [Sun, 18 Mar 2001 16:07:21 +0000 (16:07 +0000)]
Add `ipv6_ifconfig_IFN_aliasN' directive to allow multiple aliases.

PR: conf/24239

23 years agos/protocol-cloned/cloned/ I have missed in 1.24 -> 1.25.
Ruslan Ermilov [Sun, 18 Mar 2001 13:50:58 +0000 (13:50 +0000)]
s/protocol-cloned/cloned/ I have missed in 1.24 -> 1.25.

23 years agoMake sure the cached forwarding route (ipforward_rt) is still up before
Ruslan Ermilov [Sun, 18 Mar 2001 13:04:07 +0000 (13:04 +0000)]
Make sure the cached forwarding route (ipforward_rt) is still up before
using it.  Not checking this may have caused the wrong IP address to be
used when processing certain IP options (see example below).  This also
caused the wrong route to be passed to ip_output() when forwarding, but
fortunately ip_output() is smart enough to detect this.

This example demonstrates the wrong behavior of the Record Route option
observed with this bug.  Host ``freebsd'' is acting as the gateway for
the ``sysv''.

1. On the gateway, we add the route to the destination.  The new route
   will use the primary address of the loopback interface, 127.0.0.1:

:  freebsd# route add 10.0.0.66 -iface lo0 -reject
:  add host 10.0.0.66: gateway lo0

2. From the client, we ping the destination.  We see the correct replies.
   Please note that this also causes the relevant route on the ``freebsd''
   gateway to be cached in ipforward_rt variable:

:  sysv# ping -snv 10.0.0.66
:  PING 10.0.0.66: 56 data bytes
:  ICMP Host Unreachable from gateway 192.168.0.115
:  ICMP Host Unreachable from gateway 192.168.0.115
:  ICMP Host Unreachable from gateway 192.168.0.115
:
:  ----10.0.0.66 PING Statistics----
:  3 packets transmitted, 0 packets received, 100% packet loss

3. On the gateway, we delete the route to the destination, thus making
   the destination reachable through the `default' route:

:  freebsd# route delete 10.0.0.66
:  delete host 10.0.0.66

4. From the client, we ping destination again, now with the RR option
   turned on.  The surprise here is the 127.0.0.1 in the first reply.
   This is caused by the bug in ip_rtaddr() not checking the cached
   route is still up befor use.  The debug code also shows that the
   wrong (down) route is further passed to ip_output().  The latter
   detects that the route is down, and replaces the bogus route with
   the valid one, so we see the correct replies (192.168.0.115) on
   further probes:

:  sysv# ping -snRv 10.0.0.66
:  PING 10.0.0.66: 56 data bytes
:  64 bytes from 10.0.0.66: icmp_seq=0. time=10. ms
:    IP options:  <record route> 127.0.0.1, 10.0.0.65, 10.0.0.66,
:                                192.168.0.65, 192.168.0.115, 192.168.0.120,
:                                0.0.0.0(Current), 0.0.0.0, 0.0.0.0
:  64 bytes from 10.0.0.66: icmp_seq=1. time=0. ms
:    IP options:  <record route> 192.168.0.115, 10.0.0.65, 10.0.0.66,
:                                192.168.0.65, 192.168.0.115, 192.168.0.120,
:                                0.0.0.0(Current), 0.0.0.0, 0.0.0.0
:  64 bytes from 10.0.0.66: icmp_seq=2. time=0. ms
:    IP options:  <record route> 192.168.0.115, 10.0.0.65, 10.0.0.66,
:                                192.168.0.65, 192.168.0.115, 192.168.0.120,
:                                0.0.0.0(Current), 0.0.0.0, 0.0.0.0
:
:  ----10.0.0.66 PING Statistics----
:  3 packets transmitted, 3 packets received, 0% packet loss
:  round-trip (ms)  min/avg/max = 0/3/10

23 years agoFix typo in the comment
Andrey A. Chernov [Sun, 18 Mar 2001 12:20:38 +0000 (12:20 +0000)]
Fix typo in the comment

23 years agoConvert to new locale format with md_order and without E[Ff]_fmt
Andrey A. Chernov [Sun, 18 Mar 2001 12:17:46 +0000 (12:17 +0000)]
Convert to new locale format with md_order and without E[Ff]_fmt
Add ampm_fmt to all locales

23 years agoRelax local FreeBSD restrictions on 3 chars abbrev. name length and %c format
Andrey A. Chernov [Sun, 18 Mar 2001 11:58:15 +0000 (11:58 +0000)]
Relax local FreeBSD restrictions on 3 chars abbrev. name length and %c format
since they not allows POSIXly legal locale data. Currently, if relaxed form
POSIXly legal locale data will be used right now, some programs will be broken,
but it means that either locale data or programs must be fixed, not the library.

Introduce non-standard md_order (month/day order) locale field to be used later
via nl_langinfo(). Currently %EF and %Ef emulated using this field, but they
planned for remove in future in favour of nl_langinfo() test field.

Implement %F per POSIX

23 years agoFix core noted in -stable with 'auth disable SRA'.
Nick Sayer [Sun, 18 Mar 2001 09:44:25 +0000 (09:44 +0000)]
Fix core noted in -stable with 'auth disable SRA'.

I just mistakenly commited this to RELENG_4. I have contacted Jordan to see
about how to fix this. Pass the pointy hat.

23 years agoMake 'A' and 'a', 'B' and 'b' the same, per POSIX
Andrey A. Chernov [Sun, 18 Mar 2001 08:41:06 +0000 (08:41 +0000)]
Make 'A' and 'a', 'B' and 'b' the same, per POSIX

23 years ago- Add iso88025_ifdetach().
Matthew N. Dodd [Sun, 18 Mar 2001 05:43:25 +0000 (05:43 +0000)]
- Add iso88025_ifdetach().
- Add support for 802.2 type IPX frames.
- Cleanup iso88025_output() and iso88025_output() a bit.

23 years ago- Define payload length constants for 4Mbps and 16Mbps.
Matthew N. Dodd [Sun, 18 Mar 2001 05:41:07 +0000 (05:41 +0000)]
- Define payload length constants for 4Mbps and 16Mbps.
- Use explicit sizes for header structure fields.
- Use __attribute__ ((__packed__)) for header structures.
- Define struct iso88025_rif; for future use.
- Prototype upcoming iso88025_ifdetach()
- Get rid of __P() constructs in prototypes.

23 years agoproperly document the loadConfig option
Jordan K. Hubbard [Sun, 18 Mar 2001 05:38:11 +0000 (05:38 +0000)]
properly document the loadConfig option

23 years agoo Caused FFS_EXTATTR_AUTOSTART to scan two sub-directories of ".attribute"
Robert Watson [Sun, 18 Mar 2001 04:04:23 +0000 (04:04 +0000)]
o Caused FFS_EXTATTR_AUTOSTART to scan two sub-directories of ".attribute"
  off of the file system root: "user" for user attributes, and "system"
  for system attributes.  When the scan occurs, attribute backing files
  discovered in those directories will be started in the respective
  namespaces.  This re-introduces support for auto-starting of user
  attributes, which was removed when the "$" prefix for system attributes
  was replaced with explicit namespacing.

  For users of the TrustedBSD UFS POSIX.1e ACL code, you'll need to:
    mv ${FSROOT}/'$posix1e.acl_access' ${FSROOT}/system/posix1e.acl_access
    mv ${FSROOT}/'$posix1e.acl_default' ${FSROOT}/system/posix1e.acl_default

  For users of the TrustedBSD POSIX.1e Capability code, you'll need to:
    mv ${FSROOT}/'$posix1e.cap' ${FSROOT}/system/posix1e.cap

  For users of the TrustedBSD MAC code, you'll need to:
    mv ${FSROOT}/'$freebsd.mac' ${FSROOT}/system/freebsd.mac

  Updated versions of relevant patches will be released in the near
  future.

Obtained from: TrustedBSD Project

23 years agoFix a couple of things in the internal mbuf allocation interface:
Bosko Milekic [Sat, 17 Mar 2001 23:23:24 +0000 (23:23 +0000)]
Fix a couple of things in the internal mbuf allocation interface:

- Make sure that m_mballoc() really doesn't allow over nmbufs mbufs to
  be allocated from mb_map. In the case where nmbufs-reserved space is not
  an exact multiple of PAGE_SIZE (which it should be, but anyway...), we
  hold nmbufs as an absolute maximum which need not ever be reached.

- Clean up m_clalloc(); make it more consistent in the sense that the first
  argument `ncl' really means "the number of clusters ensured to be allocated"
  and not "the number of pages worth of clusters to be allocated," as was
  previously the case. This also makes it consistent with m_mballoc() as well
  as the comment that preceeds it.

Reviewed by: jlemon

23 years agoIgnore comments in /etc/passwd
Brian Somers [Sat, 17 Mar 2001 21:22:29 +0000 (21:22 +0000)]
Ignore comments in /etc/passwd

PR: 25845
Submitted by: Udo Schweigert <ust@cert.siemens.de>

23 years agoRemove unused locale/nls directories.
Andrey A. Chernov [Sat, 17 Mar 2001 21:14:52 +0000 (21:14 +0000)]
Remove unused locale/nls directories.
Corresponding fixes for "make release" already commited elsewhere.

23 years agoDon't link et_EE with 8859-1
Andrey A. Chernov [Sat, 17 Mar 2001 21:02:17 +0000 (21:02 +0000)]
Don't link et_EE with 8859-1

23 years agoDIS -> ISO
Andrey A. Chernov [Sat, 17 Mar 2001 20:15:21 +0000 (20:15 +0000)]
DIS -> ISO

23 years agoAttempt to make "make release" work again. Without a nls/fi_FI.DIS_8859-15
Poul-Henning Kamp [Sat, 17 Mar 2001 19:02:12 +0000 (19:02 +0000)]
Attempt to make "make release" work again.  Without a nls/fi_FI.DIS_8859-15
tcsh barfs.

23 years agodestroy child devices on detach to prevent ever-increasing numbers of
Cameron Grant [Sat, 17 Mar 2001 16:04:25 +0000 (16:04 +0000)]
destroy child devices on detach to prevent ever-increasing numbers of
pcm/midi devices trying to attach if the module is repeatedly loaded and
unloaded.

23 years agoCorrect the arguments to the extattrct()l call so that the enable and
Thomas Moestl [Sat, 17 Mar 2001 15:59:59 +0000 (15:59 +0000)]
Correct the arguments to the extattrct()l call so that the enable and
disable commands work as documented.

Approved by: rwatson

23 years agoNuke non-standard EAI_RESNULL.
Hajimu UMEMOTO [Sat, 17 Mar 2001 14:25:23 +0000 (14:25 +0000)]
Nuke non-standard EAI_RESNULL.

23 years agoMerged from sys/i386/conf/GENERIC revision 1.305 (moved the fxp driver
KATO Takenori [Sat, 17 Mar 2001 14:04:24 +0000 (14:04 +0000)]
Merged from sys/i386/conf/GENERIC revision 1.305 (moved the fxp driver
into the miibus section.)

23 years agos/over 3500 ports/over 4500 ports/
Peter Wemm [Sat, 17 Mar 2001 12:37:28 +0000 (12:37 +0000)]
s/over 3500 ports/over 4500 ports/

MFC candidate. :-)  The RELENG_4 version says 3000 still.

23 years agoUnrevert something which shouldn't have been reverted in the previous revert.
Poul-Henning Kamp [Sat, 17 Mar 2001 11:43:36 +0000 (11:43 +0000)]
Unrevert something which shouldn't have been reverted in the previous revert.

(Yes, I'm going to bed now to get some much needed sleep...)

23 years agoFix some further style nits
Brian Somers [Sat, 17 Mar 2001 10:09:52 +0000 (10:09 +0000)]
Fix some further style nits

Pointed out by: bde

23 years agoAdd a new entrypoint to the hashes in libmd:
Poul-Henning Kamp [Sat, 17 Mar 2001 10:00:50 +0000 (10:00 +0000)]
Add a new entrypoint to the hashes in libmd:
  char *
  FooFileChunk(const char *filename, char *buf, off_t offset, off_t length)
Which only hashes part of a file.
Implement FooFile() in terms of this function.

Submitted by: roam

23 years agoUse a generic implementation of the Fowler/Noll/Vo hash (FNV hash).
Peter Wemm [Sat, 17 Mar 2001 09:31:06 +0000 (09:31 +0000)]
Use a generic implementation of the Fowler/Noll/Vo hash (FNV hash).
Make the name cache hash as well as the nfsnode hash use it.

As a special tweak, create an unsigned version of register_t.  This allows
us to use a special tweak for the 64 bit versions that significantly
speeds up the i386 version (ie: int64 XOR int64 is slower than int64
XOR int32).

The code layout is a little strange for the string function, but I was
able to get between 5 to 10% improvement over the original version I
started with. The layout affects gcc code generation choices and this way
was fastest on x86 and alpha.

Note that 'CPUTYPE=p3' etc makes a fair difference to this.  It is
around 45% faster with -march=pentiumpro on a p6 cpu.

23 years agoBackout my previous commit, it appearantly doesn't work for asynchronous
Poul-Henning Kamp [Sat, 17 Mar 2001 08:59:31 +0000 (08:59 +0000)]
Backout my previous commit, it appearantly doesn't work for asynchronous
messages.

23 years agoUnbreak "make installworld"
Mark Murray [Sat, 17 Mar 2001 07:38:57 +0000 (07:38 +0000)]
Unbreak "make installworld"

23 years agoDramatically improve the **lame** nfs_hash(). This is based on the
Peter Wemm [Sat, 17 Mar 2001 05:43:01 +0000 (05:43 +0000)]
Dramatically improve the **lame** nfs_hash().  This is based on the
Fowler / Noll / Vo Hash (http://www.isthe.com/chongo/tech/comp/fnv/).

This improves hash coverage a *massive* amount.  We were seeing one
set of machines that were using 0.84% of their 131072 entry nfsnode
hash buckets with maximum chain lengths of up to ~500 entries.  The
machine was spending nearly 100% of its time in 'system'.
A test with this has pushed the coverage from a few perCent up to 91%
utilization with a max chain length of 11.

Submitted by:  David Filo

23 years agoFix the building of the aac driver as a module. Add a comment about enabling
Scott Long [Sat, 17 Mar 2001 04:42:33 +0000 (04:42 +0000)]
Fix the building of the aac driver as a module.  Add a comment about enabling
debug information for the driver.

23 years agoAlways call resource_int_value function for getting portsize and msize.
Yoshihiro Takahashi [Sat, 17 Mar 2001 04:23:12 +0000 (04:23 +0000)]
Always call resource_int_value function for getting portsize and msize.
It was not set resource size (portsize/msize) if resource address was set.

This is MFC candidate.

23 years agoReally fix NWAY negotiation for the 82553 PHY. Locking down the
Jonathan Lemon [Sat, 17 Mar 2001 02:50:20 +0000 (02:50 +0000)]
Really fix NWAY negotiation for the 82553 PHY.  Locking down the
media interface selection should not imply disabling NWAY negotiaton
as well.

Problem pointed out by: peter

23 years agoStop cron from shouting in the syslog. Use setproctitle() instead, since
Peter Wemm [Sat, 17 Mar 2001 00:21:54 +0000 (00:21 +0000)]
Stop cron from shouting in the syslog.  Use setproctitle() instead, since
smashing argv[0] doesn't have the intended effect of changing the ps(1)
output these days.

PR: 25850

23 years agoThis is an MFC candidate.
Justin T. Gibbs [Sat, 17 Mar 2001 00:12:15 +0000 (00:12 +0000)]
This is an MFC candidate.

Add the AAC_DEBUG option to enable debugging in the aac driver.

Correct a race condition in the interrupt handler where the
controller may queue a fib to a response queue after the driver
has serviced the queue but before the interrupt is cleared.
This could leave a completed fib stranded in the response queue
unless another I/O completed and generated another interrupt.

Reviewed by: msmith

23 years agoAdd the AAC_DEBUG option to enable debugging in the aac driver.
Justin T. Gibbs [Sat, 17 Mar 2001 00:09:47 +0000 (00:09 +0000)]
Add the AAC_DEBUG option to enable debugging in the aac driver.

Reviewed by: msmith

23 years agoClean up the grammar a bit from the last commit.
Murray Stokely [Fri, 16 Mar 2001 22:55:55 +0000 (22:55 +0000)]
Clean up the grammar a bit from the last commit.

23 years agoWhen doing a recv(.. MSG_WAITALL) for a message which is larger than
Jonathan Lemon [Fri, 16 Mar 2001 22:37:06 +0000 (22:37 +0000)]
When doing a recv(.. MSG_WAITALL) for a message which is larger than
the socket buffer size, the receive is done in sections.  After completing
a read, call pru_rcvd on the underlying protocol before blocking again.

This allows the the protocol to take appropriate action, such as
sending a TCP window update to the peer, if the window happened to
close because the socket buffer was filled.  If the protocol is not
notified, a TCP transfer may stall until the remote end sends a window
probe.

23 years agoFix a few things in the aic(4) driver:
Kenneth D. Merry [Fri, 16 Mar 2001 22:20:19 +0000 (22:20 +0000)]
Fix a few things in the aic(4) driver:

 - enable 10MHz (fast SCSI) operation on boards that support it.  (only
   aic6360 boards with fast SCSI enabled can do it)

 - bounds check sync periods and offsets passed in from the transport layer

 - tell the user which resource allocation failed (for the ISA probe) if we
   weren't able to allocate an IRQ, DRQ or I/O port.

23 years agoFix type-o
Matthew Dillon [Fri, 16 Mar 2001 22:18:26 +0000 (22:18 +0000)]
Fix type-o

Submitted by: okazaki

23 years agoFix inquiry length detection for the ses(4) driver. It was using the
Kenneth D. Merry [Fri, 16 Mar 2001 22:16:54 +0000 (22:16 +0000)]
Fix inquiry length detection for the ses(4) driver.  It was using the
inq_len member of the ccb_getdev structure, but we've never filled that
value in..

So we now get the length from the inquiry data returned by the drive.
(Since we will fetch as much inquiry data as the drive claims to support.)

Reviewed by: mjacob
Reported by: Andrzej Tobola <san@iem.pw.edu.pl>

23 years agoFix a number of warnings that come up when compiling with BDECFLAGS.
Garance A Drosehn [Fri, 16 Mar 2001 21:54:12 +0000 (21:54 +0000)]
Fix a number of warnings that come up when compiling with BDECFLAGS.

23 years agoprevent gdb from panic'ing an alpha with "mutex Giant not owned"
Andrew Gallatin [Fri, 16 Mar 2001 21:44:21 +0000 (21:44 +0000)]
prevent gdb from panic'ing an alpha with "mutex Giant not owned"

reviewed by: jhb

23 years agoFix a typo which would cause containers between 1GB and 2GB to have the wrong
Mike Smith [Fri, 16 Mar 2001 21:43:32 +0000 (21:43 +0000)]
Fix a typo which would cause containers between 1GB and 2GB to have the wrong
geometry reported.

Submitted by: mark tinguely <tinguely@web.cs.ndsu.NoDak.edu>

23 years agofix a typo preventing the second dma channel being released
Cameron Grant [Fri, 16 Mar 2001 21:04:14 +0000 (21:04 +0000)]
fix a typo preventing the second dma channel being released
use isa_dma_release when releasing dma channels

23 years agodon't leak memory allocated for feeders at module unload
Cameron Grant [Fri, 16 Mar 2001 20:58:44 +0000 (20:58 +0000)]
don't leak memory allocated for feeders at module unload
kill the fake channel when unregistering

23 years ago<sys/queue.h> makeover.
Poul-Henning Kamp [Fri, 16 Mar 2001 20:00:53 +0000 (20:00 +0000)]
<sys/queue.h> makeover.

23 years agoFix a style(9) nit.
Poul-Henning Kamp [Fri, 16 Mar 2001 19:36:23 +0000 (19:36 +0000)]
Fix a style(9) nit.

23 years agoAdd -ledit to ngctl.
Poul-Henning Kamp [Fri, 16 Mar 2001 19:35:07 +0000 (19:35 +0000)]
Add -ledit to ngctl.

Somebody submitted this long time ago, and it has been sitting in my
tree for months because I thought archie would pick it up.

Submitted by: (sorry, lost track)

23 years agoBump MAX_GLOBENTRIES up to 16384, so it is a power of two. Add
Jonathan Lemon [Fri, 16 Mar 2001 19:05:20 +0000 (19:05 +0000)]
Bump MAX_GLOBENTRIES up to 16384, so it is a power of two.  Add
some comments explaining that this is an arbitrary limit.

Requested by:  jkh

23 years agonuke IPSEC_SRCSEL which does not do the right thing.
Hajimu UMEMOTO [Fri, 16 Mar 2001 17:52:48 +0000 (17:52 +0000)]
nuke IPSEC_SRCSEL which does not do the right thing.
adjust state->ro if the tunnel endpoint is offlink.
KAME PR 233.

PR: kern/21079

23 years agoo The revenge of the mdoc(7) police:
Robert Watson [Fri, 16 Mar 2001 17:42:38 +0000 (17:42 +0000)]
o The revenge of the mdoc(7) police:

   - These pages abused Ar macro (they should have used Fa).
   - NULL and other numeric constants should be marked with Dv.
   - VOP_* in the ERRORS section for the EOPNOTSUPP entry should be marked
     with Fn.

Submitted by: ru

23 years agoo The mdoc(7) police strike with avengance:
Robert Watson [Fri, 16 Mar 2001 15:02:30 +0000 (15:02 +0000)]
o The mdoc(7) police strike with avengance:

   - These pages abused Ar macro (they should have used Fa).
   - NULL constant should be marked with Dv.
   - VOP_* in the ERRORS section for the EOPNOTSUPP entry should be marked
     with Fn.

Submitted by: ru

23 years agoEnable some undocumented bits in the DP83840 PHY, which is needed
Jonathan Lemon [Fri, 16 Mar 2001 14:17:02 +0000 (14:17 +0000)]
Enable some undocumented bits in the DP83840 PHY, which is needed
when using it with the Intel fxp driver.

Tested by: peter, Ollivier Robert <roberto@eurocontrol.fr>,
   Peter Schultz <pete@jocose.org>