]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoRewrite of the CAM error recovery code.
ken [Tue, 27 Mar 2001 05:45:52 +0000 (05:45 +0000)]
Rewrite of the CAM error recovery code.

Some of the major changes include:

- The SCSI error handling portion of cam_periph_error() has
  been broken out into a number of subfunctions to better
  modularize the code that handles the hierarchy of SCSI errors.
  As a result, the code is now much easier to read.

- String handling and error printing has been significantly
  revamped.  We now use sbufs to do string formatting instead
  of using printfs (for the kernel) and snprintf/strncat (for
  userland) as before.

  There is a new catchall error printing routine,
  cam_error_print() and its string-based counterpart,
  cam_error_string() that allow the kernel and userland
  applications to pass in a CCB and have errors printed out
  properly, whether or not they're SCSI errors.  Among other
  things, this helped eliminate a fair amount of duplicate code
  in camcontrol.

  We now print out more information than before, including
  the CAM status and SCSI status and the error recovery action
  taken to remedy the problem.

- sbufs are now available in userland, via libsbuf.  This
  change was necessary since most of the error printing code
  is shared between libcam and the kernel.

- A new transfer settings interface is included in this checkin.
  This code is #ifdef'ed out, and is primarily intended to aid
  discussion with HBA driver authors on the final form the
  interface should take.  There is example code in the ahc(4)
  driver that implements the HBA driver side of the new
  interface.  The new transfer settings code won't be enabled
  until we're ready to switch all HBA drivers over to the new
  interface.

src/Makefile.inc1,
lib/Makefile: Add libsbuf.  It must be built before libcam,
since libcam uses sbuf routines.

libcam/Makefile: libcam now depends on libsbuf.

libsbuf/Makefile: Add a makefile for libsbuf.  This pulls in the
sbuf sources from sys/kern.

bsd.libnames.mk: Add LIBSBUF.

camcontrol/Makefile: Add -lsbuf.  Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.

camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.

sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *.  This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.

Fix a typo.

cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.

Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.

cam.h: Add a new CAM flag, CAM_RETRY_SELTO.

Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().

cam_ccb.h: Add new transfer negotiation structures/types.

Change inq_len in the ccb_getdev structure to
be "reserved".  This field has never been
filled in, and will be removed when we next
bump the CAM version.

cam_debug.h: Fix typo.

cam_periph.c: Modularize cam_periph_error().  The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().

In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.

cam_xpt.c: Add new transfer negotiation code.  (ifdefed
out)

Add a new function, xpt_path_string().  This
is a string/sbuf analog to xpt_print_path().

scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code.  More of that code is shared
between userland the kernel.

scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.

Add a new error action, SS_REQSENSE.  (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.

Change the default actions around a bit.

scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO.  Selection
timeouts shouldn't be covered by a sense flag.

scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.

Get rid of the last vestiges of a read/write
interface.

libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.

aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.

sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.

Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.

Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.

Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.

kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.

Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken

23 years agoReplace dyn_fin_lifetime with dyn_ack_lifetime for half-closed state.
simokawa [Tue, 27 Mar 2001 05:28:30 +0000 (05:28 +0000)]
Replace dyn_fin_lifetime with dyn_ack_lifetime for half-closed state.
Half-closed state could last long for some connections and fin_lifetime
(default 20sec) is too short for that.

OK'ed by: luigi

23 years agolimit the amount of retries when sending data to prevent lockups.
alfred [Tue, 27 Mar 2001 05:03:49 +0000 (05:03 +0000)]
limit the amount of retries when sending data to prevent lockups.

23 years agoAt least install primes to the right place, for now. I suppose.
green [Tue, 27 Mar 2001 03:58:12 +0000 (03:58 +0000)]
At least install primes to the right place, for now.  I suppose.

Reminded by: everyone

23 years agoThe common wisdom is to use the largest number of cylinders per group.
obrien [Tue, 27 Mar 2001 01:34:58 +0000 (01:34 +0000)]
The common wisdom is to use the largest number of cylinders per group.
So bump the default from `16' to `22', which is the largest value allowed
with the current default block size.  This change increases the the
group size from 32MB/g to 44MB/g on a 4GB SCSI disk.

23 years agoDon't call the fictious `MAKEDEV' from the path "/sbin:/bin" when we
obrien [Tue, 27 Mar 2001 01:23:07 +0000 (01:23 +0000)]
Don't call the fictious `MAKEDEV' from the path "/sbin:/bin" when we
recurse.  Rather recurse on ourself (as we know our own name).

23 years agoUse SSIZE_MAX instead of INT_MAX, as kernel does ssize_t check
ache [Tue, 27 Mar 2001 01:16:44 +0000 (01:16 +0000)]
Use SSIZE_MAX instead of INT_MAX, as kernel does ssize_t check
Better explanation comment of FIXME section

23 years agoAdd ttyd0 which is needed on the Alpha when using the fix-it CDROM
obrien [Tue, 27 Mar 2001 01:06:58 +0000 (01:06 +0000)]
Add ttyd0 which is needed on the Alpha when using the fix-it CDROM
over a serial console.

23 years agoo Update copyright date
rwatson [Mon, 26 Mar 2001 19:55:35 +0000 (19:55 +0000)]
o Update copyright date
o Revise description in light of commits over last month including:
  - ACL editing library is now implemented
  - ACLs are now implemented

Obtained from: TrustedBSD Project

23 years agoAdded burncd to the SEE ALSO section.
sos [Mon, 26 Mar 2001 19:49:39 +0000 (19:49 +0000)]
Added burncd to the SEE ALSO section.

Idea by: "Akinori MUSHA" <knu@iDaemons.org>

23 years agoTreat mmap() error as fatal too, i.e. do exit(1) instead of return
ache [Mon, 26 Mar 2001 19:36:27 +0000 (19:36 +0000)]
Treat mmap() error as fatal too, i.e. do exit(1) instead of return

23 years agoAdjust FreeBSD 4.3 release date.
wosch [Mon, 26 Mar 2001 19:35:14 +0000 (19:35 +0000)]
Adjust FreeBSD 4.3 release date.

Approved by: jkh

23 years agorlines() checks:
ache [Mon, 26 Mar 2001 19:29:49 +0000 (19:29 +0000)]
rlines() checks:
1) really check for size overflow by checking negative value.
2) since mmap() not support files over INT_MAX size, add check for it
until either mmap() will be fixed or tail will be rewritted to handle
large files alternatively.
3) replace fseek(... file_size, SEEK_SET) with fseek(... 0L, SEEK_END)
to avoid off_t -> long cast
4) Use exit() if file is too big instead of warning and wrong logic
afterwards.

23 years agoNew release notes: netstat -W, sockstat -c and -l, FFS/EXT2FS security
bmah [Mon, 26 Mar 2001 18:04:53 +0000 (18:04 +0000)]
New release notes:  netstat -W, sockstat -c and -l, FFS/EXT2FS security
fixes (FreeBSD-SA-01:30).

Reorder netstat(1) notes to be adjacent.

23 years agoIntroduce support for POSIX.1e ACLs on UFS-based file systems. This
rwatson [Mon, 26 Mar 2001 17:53:19 +0000 (17:53 +0000)]
Introduce support for POSIX.1e ACLs on UFS-based file systems.  This
implementation is still experimental, and while fairly broadly tested,
is not yet intended for production use.  Support for POSIX.1e ACLs on
UFS will not be MFC'd to RELENG_4.

This implementation works by providing implementations of VOP_[GS]ETACL()
for FFS, as well as modifying the appropriate access control and file
creation routines.  In this implementation, ACLs are backed into extended
attributes; the base ACL (owner, group, other) permissions remain in the
inode for performance and compatibility reasons, so only the extended and
default ACLs are placed in extended attributes.  The logic for ACL
evaluation is provided by the fs-independent kern/kern_acl.c.

o Introduce UFS_ACL, a compile-time configuration option that enables
  support for ACLs on FFS (and potentially other UFS-based file systems).
o Introduce ufs_getacl(), ufs_setacl(), ufs_aclcheck(), which
  respectively get, set, and check the ACLs on the passed vnode.
o Introduce ufs_sync_acl_from_inode(), ufs_sync_inode_from_acl() to
  maintain access control information between inode permissions and
  extended attribute data.
o Modify ufs_access() to load a file access ACL and invoke
  vaccess_acl_posix1e() if ACLs are available on the file system
o Modify ufs_mkdir() and ufs_makeinode() to associate ACLs with newly
  created directories and files, inheriting from the parent directory's
  default ACL.
o Enable these new vnode operations and conditionally compiled code
  paths if UFS_ACL is defined.

A few notes:

o This implementation is fairly widely tested, but still should be
  considered experimental.
o Currently, ACLs are not exported via NFS, instead, the summarizing
  file mode/etc from the inode is.  This results in conservative
  protection behavior, similar to the behavior of ACL-nonaware programs
  acting locally.
o It is possible that underlying binary data formats associated with
  this implementation may change.  Consumers of the implementation
  should expect to find their local configuration obsoleted in the
  next few months, resulting in possible loss of ACL data during an
  upgrade.
o The extended attributes interface and implementation is still
  undergoing modification to address portable interface concerns, as
  well as performance.
o Many applications do not yet correctly handle ACLs.  In general,
  due to the POSIX.1e ACL model, behavior of ACL-unaware applications
  will be conservative with respects to file protection; some caution
  is recommended.
o Instructions for configuring and maintaining ACLs on UFS will be
  committed in the near future; in the mean time it is possible to
  reference the README included in the last UFS ACL distribution
  placed in the TrustedBSD web site:

      http://www.TrustedBSD.org/downloads/

Substantial debugging, hardware, travel, or connectivity support for this
project was provided by: BSDi, Safeport Network Services, and NAI Labs.
Significant coding contributions were made by Chris Faulhaber.  Additional
support was provided by Brian Feldman, Thomas Moestl, and Ilmar Habibulin.

Reviewed by: jedgar, keichii, mckusick, trustedbsd-discuss, freebsd-fs
Obtained from: TrustedBSD Project

23 years agoFix error reporting of delayed send errors.
fenner [Mon, 26 Mar 2001 16:18:01 +0000 (16:18 +0000)]
Fix error reporting of delayed send errors.

23 years agosecure/ build fixes:
ru [Mon, 26 Mar 2001 14:53:33 +0000 (14:53 +0000)]
secure/ build fixes:

- TELNETOBJDIR is gone.  `buildworld' already installs libtelnet.a
  in ${WORLDTMP}/usr/lib, and we have LIBRARY_PATH pointing there.

- SSHDIR (formerly SSHSRC) is now shared between all SSH modules.
  New LIBSSH is introduced for libssh.a (an internal static lib).
  Previously, build without prior `obj' was broken; SSH modules
  always looked for libssh.a in ${.OBJDIR}.  Also, the dependancies
  on the libssh.a were missing.

- libtelnet/ did not install the crypto version of telnet.h into
  /usr/include/arpa.

- Removed BINOWN, BINMODE, BINDIR and SRCS with default values.

Reviewed by: markm

- MAN[1-9] -> MAN.

23 years agoBackout botched attempt to introduce MANSECT feature; it
ru [Mon, 26 Mar 2001 14:47:21 +0000 (14:47 +0000)]
Backout botched attempt to introduce MANSECT feature; it
doesn't work in "developer" mode (single module checkout).

23 years ago- Backout botched attempt to introduce MANSECT feature.
ru [Mon, 26 Mar 2001 14:42:20 +0000 (14:42 +0000)]
- Backout botched attempt to introduce MANSECT feature.
- MAN[1-9] -> MAN.

23 years ago- Backout botched attempt to introduce MANSECT feature.
ru [Mon, 26 Mar 2001 14:33:27 +0000 (14:33 +0000)]
- Backout botched attempt to introduce MANSECT feature.
- MAN[1-9] -> MAN.

23 years ago- Backout botched attempt to intoduce MANSECT feature.
ru [Mon, 26 Mar 2001 14:22:12 +0000 (14:22 +0000)]
- Backout botched attempt to intoduce MANSECT feature.
- MAN[1-9] -> MAN.

23 years agoDo not build (and install) both secure/ and standard versions
ru [Mon, 26 Mar 2001 12:49:05 +0000 (12:49 +0000)]
Do not build (and install) both secure/ and standard versions
of libtelnet, telnetd, and telnet.  This only worked because
secure/ was listed late in SUBDIR in Makefile.inc1.

Reviewed by: markm

23 years agoAdd missing NOSECURE check for `includes' target.
ru [Mon, 26 Mar 2001 12:46:17 +0000 (12:46 +0000)]
Add missing NOSECURE check for `includes' target.

Reviewed by: markm

23 years agoPrevious commit broke interlock locking for !LK_RETRY case.
bp [Mon, 26 Mar 2001 12:45:35 +0000 (12:45 +0000)]
Previous commit broke interlock locking for !LK_RETRY case.

23 years agoSend the remains (such as I have located) of "block major numbers" to
phk [Mon, 26 Mar 2001 12:41:29 +0000 (12:41 +0000)]
Send the remains (such as I have located) of "block major numbers" to
the bit-bucket.

23 years agoDecapitalise first letter of warning message.
sobomax [Mon, 26 Mar 2001 09:57:26 +0000 (09:57 +0000)]
Decapitalise first letter of warning message.

23 years agoIn which(1) mode resolve "/../", "/./" in arguments and PLIST contents. This
sobomax [Mon, 26 Mar 2001 09:31:41 +0000 (09:31 +0000)]
In which(1) mode resolve "/../", "/./" in arguments and PLIST contents. This
is done without realpath() to avoid symlinks resolving.

Submitted by: sobomax, Garrett Rooney <rooneg@electricjellyfish.net>

23 years agoFirst step towards plugging the "pccard is trying to map into a BIOS
imp [Mon, 26 Mar 2001 08:05:20 +0000 (08:05 +0000)]
First step towards plugging the "pccard is trying to map into a BIOS
region for CIS reading" problem:

Use bus_alloc_resource to get the memory that we'll be using.  Also
has the benefit of doing usage checking as well.  This gets rid of the
ugly kludge that we had before for mapping pmem to vmem.

Second, move PIOCSRESOURCE to its own routine and make it conform more
to style(9) in the process.

23 years agoRemoved _MANPAGES, MANDEPEND and MANSRC.
ru [Mon, 26 Mar 2001 08:04:11 +0000 (08:04 +0000)]
Removed _MANPAGES, MANDEPEND and MANSRC.

Added MAN which will eventually replace MAN[0-9] and MAN1aout.
For now, the old syntax is still supported.

Reviewed by: bde

23 years agoDo not depend on ``all-man'' if -DNOMAN.
ru [Mon, 26 Mar 2001 07:46:57 +0000 (07:46 +0000)]
Do not depend on ``all-man'' if -DNOMAN.

23 years agoDon't use MANDEPEND and MANSRC.
ru [Mon, 26 Mar 2001 07:28:26 +0000 (07:28 +0000)]
Don't use MANDEPEND and MANSRC.

23 years agoPrevent race condition by using msleep() instead of mtx_unlock()/tsleep().
bp [Mon, 26 Mar 2001 03:10:07 +0000 (03:10 +0000)]
Prevent race condition by using msleep() instead of mtx_unlock()/tsleep().

Reviewed by: alfred

23 years agoStop claiming to support the DEC EtherWorks II/III cards.
jkh [Mon, 26 Mar 2001 01:32:29 +0000 (01:32 +0000)]
Stop claiming to support the DEC EtherWorks II/III cards.

PR:             misc/18641

23 years agofix: text following `#else' violates ANSI standard
alfred [Mon, 26 Mar 2001 00:35:23 +0000 (00:35 +0000)]
fix: text following `#else' violates ANSI standard

Pointed out by: ${BDECFLAGS}

23 years agoDon't call daemon() and setup our signal handlers until after we check
alfred [Sun, 25 Mar 2001 23:32:55 +0000 (23:32 +0000)]
Don't call daemon() and setup our signal handlers until after we check
and do the unregister/reregister work.

Don't call syslog in the unregister/reregister code as we haven't called
openlog() yet.

Be a more conservative about accepting errno values from socket(2),
only EPROTONOSUPPORT means that the kernel isn't supporting it
something like INET6.  The other possible errnos would be returned
if there was a mistake in the socket(2) call so remove them from the
list of "acceptable" return values.

23 years agoDisable ipv6 when getnetconfigent("udp6"/"tcp6") fails.
alfred [Sun, 25 Mar 2001 23:28:03 +0000 (23:28 +0000)]
Disable ipv6 when getnetconfigent("udp6"/"tcp6") fails.

Submitted by: Martin Blapp <mb@imp.ch>

23 years agoTurn off building the ARC loader. I don't know of anyone currently working
obrien [Sun, 25 Mar 2001 23:07:44 +0000 (23:07 +0000)]
Turn off building the ARC loader.  I don't know of anyone currently working
on advancing this WIP.

23 years agofix whitespace bogons
cg [Sun, 25 Mar 2001 21:43:24 +0000 (21:43 +0000)]
fix whitespace bogons

23 years agoIssue a ``quit'' after other ppp commands given on the command line and
brian [Sun, 25 Mar 2001 20:06:08 +0000 (20:06 +0000)]
Issue a ``quit'' after other ppp commands given on the command line and
then wait for the connection to be closed by the peer.

This means that commands such as ``pppctl ... show links'' will
display the correct output again (rather than truncating it depending
on how much data arrived in the last packet).

23 years agoReplace pmap_unset() with rpcb_unset() which fixes the unregistering.
alfred [Sun, 25 Mar 2001 19:59:07 +0000 (19:59 +0000)]
Replace pmap_unset() with rpcb_unset() which fixes the unregistering.

Submitted by: Martin Blapp <mb@imp.ch>

23 years agoDeal with lack of IPv6 support gracefully.
alfred [Sun, 25 Mar 2001 19:57:58 +0000 (19:57 +0000)]
Deal with lack of IPv6 support gracefully.

Submitted by: Martin Blapp <mb@imp.ch>

23 years agofix -o port=xxx
alfred [Sun, 25 Mar 2001 19:18:43 +0000 (19:18 +0000)]
fix -o port=xxx

23 years agorelease resources if one of the speculative probes in opti_detect() fails,
cg [Sun, 25 Mar 2001 19:09:06 +0000 (19:09 +0000)]
release resources if one of the speculative probes in opti_detect() fails,
otherwise resource_list_alloc panics when opti_detect tries its next probe.

23 years agothe softc is not a mutex, don't try to lock it.
cg [Sun, 25 Mar 2001 18:56:48 +0000 (18:56 +0000)]
the softc is not a mutex, don't try to lock it.

Submitted by: George Reid <greid@ukug.uk.freebsd.org>

23 years agoUpdate for file 3.34.
obrien [Sun, 25 Mar 2001 18:38:47 +0000 (18:38 +0000)]
Update for file 3.34.

23 years agoThis commit was generated by cvs2svn to compensate for changes in r74784,
obrien [Sun, 25 Mar 2001 18:37:04 +0000 (18:37 +0000)]
This commit was generated by cvs2svn to compensate for changes in r74784,
which included commits to RCS files with non-trunk default branches.

23 years agoVirgin import of Christos Zoulas's FILE 3.34.
obrien [Sun, 25 Mar 2001 18:37:04 +0000 (18:37 +0000)]
Virgin import of Christos Zoulas's FILE 3.34.

23 years agoBah. 'ln -sf' -> 'ln -fs'. I need to read my email more closely in the
scottl [Sun, 25 Mar 2001 15:51:43 +0000 (15:51 +0000)]
Bah.  'ln -sf' -> 'ln -fs'.  I need to read my email more closely in the
mornings.

Pointy hat again Obtained from: roam@orbitel.org

23 years ago'ln -s' -> 'ln -sf' for the afa link.
scottl [Sun, 25 Mar 2001 14:11:55 +0000 (14:11 +0000)]
'ln -s' -> 'ln -sf' for the afa link.

Pointy hat Obtained from: roam@orbitel.bg

23 years agoMake header files conform to style(9).
brian [Sun, 25 Mar 2001 12:05:10 +0000 (12:05 +0000)]
Make header files conform to style(9).

Reviewed by (*): bde

(*) alias_local.h only got a cursory glance.

23 years agoFix .Xr sd -> da.
dirk [Sun, 25 Mar 2001 11:57:25 +0000 (11:57 +0000)]
Fix .Xr sd -> da.

23 years agoIdentify obsolete ports
brian [Sun, 25 Mar 2001 11:35:22 +0000 (11:35 +0000)]
Identify obsolete ports

23 years ago(MFC candidate since this is already a merge from /sys/net only.)
joerg [Sun, 25 Mar 2001 09:59:23 +0000 (09:59 +0000)]
(MFC candidate since this is already a merge from /sys/net only.)

Merge rev's 1.65 and 1.66 from sys/net/if_spppsubr.c (implement the
`restart' option, and fix a blatant bug with PAP authentication).

The i4b implementation of this file should be merged back, but for now,
we need this here as well.

Reviewed by: gj

23 years agoThis is another MFC candidate.
joerg [Sun, 25 Mar 2001 09:53:07 +0000 (09:53 +0000)]
This is another MFC candidate.

Fix a serious bug in sppp where anyone could obtain a successful PAP
authentication by supplying a null password.  I've only stumpled across
the PR while browsing for all sppp-related PRs.

Should we also file a security advisory for this?

PR: 21592
Submitted by: <dli@3bc.de> Dirk Liebke

23 years agoRemove some unused stuff
peter [Sun, 25 Mar 2001 07:21:04 +0000 (07:21 +0000)]
Remove some unused stuff

23 years agoI need to add to the previous commit:
markm [Sun, 25 Mar 2001 07:02:11 +0000 (07:02 +0000)]
I need to add to the previous commit:

Tested by: dougb

23 years agoAllow bog-standard ioctls through. There are really handled in
markm [Sun, 25 Mar 2001 06:55:19 +0000 (06:55 +0000)]
Allow bog-standard ioctls through. There are really handled in
higher layers, but there needs to be a "no-error" return here.

23 years agoMerge various changes from OpenBSD and NetBSD.
mikeh [Sun, 25 Mar 2001 04:57:05 +0000 (04:57 +0000)]
Merge various changes from OpenBSD and NetBSD.

o remove panic() in favor of err(3) and use err(3) functions
  consistently throughout
o use stat(2)'s S_IS* macros rather than S_IF*
o [r]index -> str[r]chr
o convert some static buffers to dynamic ones
o use real tempfiles rather than reopening the same templates
o rename some functions that clash with libc
o convert wait_status from union to int and use wait(2) status macros
o fix multiple potential buffer overflows
o fix a few comments
o add $FreeBSD$

Reviewed by: nra, nectar (earlier version)

23 years agoRemove an extraneous declaration.
brian [Sun, 25 Mar 2001 03:34:29 +0000 (03:34 +0000)]
Remove an extraneous declaration.

23 years agoTemporary build structure for GDB 5.0 so people can test the new version
obrien [Sun, 25 Mar 2001 02:20:37 +0000 (02:20 +0000)]
Temporary build structure for GDB 5.0 so people can test the new version
before pulling the switch to making it the default version.

23 years agoIf we're adding an interface address that is already one of the interface's
brian [Sun, 25 Mar 2001 01:34:27 +0000 (01:34 +0000)]
If we're adding an interface address that is already one of the interface's
aliases with the same netmask and destination, don't remove it and then
re-add exactly the same thing.

This means that static (non-sticky) routes that use the interface address
(or destination address) as a destination will not suddenly evaporate when
IPCP comes up (not unless the negotiated IPs have changed anyway).

23 years agoMove the atomic() mbstat.m_drops incrementing to the MGET(HDR) and
bmilekic [Sat, 24 Mar 2001 23:47:52 +0000 (23:47 +0000)]
Move the atomic() mbstat.m_drops incrementing to the MGET(HDR) and
MCLGET macros in order to avoid incrementing the drop count twice.
Otherwise, in some cases, we may increment m_drops once in m_mballoc()
for example, and increment it again in m_mballoc_wait() if the
wait fails.

23 years agomega-commit.
cg [Sat, 24 Mar 2001 23:10:29 +0000 (23:10 +0000)]
mega-commit.

this introduces a new buffering mechanism which results in dramatic
simplification of the channel manager.

as several structures have changed, we take the opportunity to move their
definitions into the source files where they are used, make them private and
de-typedef them.

the sound drivers are updated to use snd_setup_intr instead of
bus_setup_intr, and to comply with the de-typedefed structures.

the ac97, mixer and channel layers have been updated with finegrained
locking, as have some drivers- not all though.  the rest will follow soon.

23 years agoAdd apsfilter to "important package" set.
jkh [Sat, 24 Mar 2001 23:05:44 +0000 (23:05 +0000)]
Add apsfilter to "important package" set.

Submitted by: andreas

23 years agoCatch these modules up with obrien's recent cleanup commit.
peter [Sat, 24 Mar 2001 22:33:43 +0000 (22:33 +0000)]
Catch these modules up with obrien's recent cleanup commit.

23 years agoDrop 'KMODDEPS = snd_pcm' - it isn't used.
peter [Sat, 24 Mar 2001 22:28:49 +0000 (22:28 +0000)]
Drop 'KMODDEPS = snd_pcm' - it isn't used.

23 years agoBack out 'KMODDEPS = miibus' addition in last commit. This isn't used.
peter [Sat, 24 Mar 2001 22:21:59 +0000 (22:21 +0000)]
Back out 'KMODDEPS = miibus' addition in last commit.  This isn't used.

23 years agoAdd new categories science and ukrainian.
asami [Sat, 24 Mar 2001 21:52:25 +0000 (21:52 +0000)]
Add new categories science and ukrainian.

23 years agoAdd #if'ed support for the future option CAM_NEW_TRAN_CODE.
groudier [Sat, 24 Mar 2001 21:11:31 +0000 (21:11 +0000)]
Add #if'ed support for the future option CAM_NEW_TRAN_CODE.

23 years agowhen writing to one of the setup registers, a 32-bit write is required.
cg [Sat, 24 Mar 2001 20:36:59 +0000 (20:36 +0000)]
when writing to one of the setup registers, a 32-bit write is required.

this apparently fixes problems initialising certain es1371/es1373/ct5880
revisions.

Confirmed working by: Richard J Kuhns <rjk@grauel.com>
PR: i386/25944

23 years agoRevert the warning elimination in rev 1.292. It breaks the Alpha build with:
obrien [Sat, 24 Mar 2001 17:28:06 +0000 (17:28 +0000)]
Revert the warning elimination in rev 1.292.  It breaks the Alpha build with:

/usr/src/usr.sbin/sysinstall/menus.c:1323: initializer element is not computable at load time
/usr/src/usr.sbin/sysinstall/menus.c:1323: (near initialization for `MenuNetworking.items[9].aux')

23 years agoAdd bandaid to get ISA only systems to link.
sos [Sat, 24 Mar 2001 16:19:07 +0000 (16:19 +0000)]
Add bandaid to get ISA only systems to link.

23 years agoFix typo ); -> ,
asmodai [Sat, 24 Mar 2001 15:25:04 +0000 (15:25 +0000)]
Fix typo ); -> ,

23 years agoUnbreak build on alpha.
ume [Sat, 24 Mar 2001 15:17:27 +0000 (15:17 +0000)]
Unbreak build on alpha.
  - Move in_port_t to sys/types.h.
  - Nuke in_addr_t from each endian.h.

Reported by: jhb

23 years agoSort includes and remove bogus typedef of modeventhand_t, since this
alex [Sat, 24 Mar 2001 12:18:45 +0000 (12:18 +0000)]
Sort includes and remove bogus typedef of modeventhand_t, since this
isn't used in the example any more.

Pointed out by: bde

23 years agoList of distfile contents to delete using `rm' and `find -delete'.
obrien [Sat, 24 Mar 2001 09:37:48 +0000 (09:37 +0000)]
List of distfile contents to delete using `rm' and `find -delete'.
Simular to the FREEBSD-Xlist used with `tar -X ... -xf'.
This file is typically used when one starts with the GDB anoncvs repo
rather than a release tarball.

23 years agoSigh. I should know better than to commit a one line change from a PR
obrien [Sat, 24 Mar 2001 09:01:31 +0000 (09:01 +0000)]
Sigh.  I should know better than to commit a one line change from a PR
w/o a full compile test, even when the PR seemed so authoritative on the
subject...

Install sstream, not the non-existent sstring.

23 years agoRestore the careful preservation of .depend (rev 1.78 of Makefile.i386)
obrien [Sat, 24 Mar 2001 08:44:28 +0000 (08:44 +0000)]
Restore the careful preservation of .depend (rev 1.78 of Makefile.i386)
that I removed in my last commit dealing with `make depend' bogons.
This commit has some races, but hopefully they are too short to matter.
Unfortuneatly, neither .newdep nor .olddep is removed by `make clean'.

Submitted by: bde

23 years agoSilence (harmless) warnings.
markm [Sat, 24 Mar 2001 08:38:42 +0000 (08:38 +0000)]
Silence (harmless) warnings.

23 years agoRound 1 of Brucification inspired changes.
peter [Sat, 24 Mar 2001 08:37:56 +0000 (08:37 +0000)]
Round 1 of Brucification inspired changes.

23 years agoCleaner way of adding -fschg (ie, correctly implement).
obrien [Sat, 24 Mar 2001 08:31:45 +0000 (08:31 +0000)]
Cleaner way of adding -fschg (ie, correctly implement).
"INSTALLFLAGS" belongs to individual Makefiles.
"_INSTALLFLAGS" is for global additions.

Submitted by: bde

23 years agoFix a problem where we were switching npxproc from underneath processes
obrien [Sat, 24 Mar 2001 08:27:57 +0000 (08:27 +0000)]
Fix a problem where we were switching npxproc from underneath processes
running in process context in order to run interrupt handlers.  This
caused a big smashing of the stack on AMD K6, K5 and Intel Pentium (ie, P5)
processors because we are using npxproc as a flag to indicate whether
the state has been pushed onto the stack.

Submitted by: bde

23 years agoInstall sstring.
obrien [Sat, 24 Mar 2001 08:19:31 +0000 (08:19 +0000)]
Install sstring.

PR: 25927
Submitted by: Jeremy Norris <ishmael27@home.com>

23 years agoGarbage collect these ancient bits.
obrien [Sat, 24 Mar 2001 08:15:22 +0000 (08:15 +0000)]
Garbage collect these ancient bits.

23 years agoAdd cvs tag
dougb [Sat, 24 Mar 2001 07:20:36 +0000 (07:20 +0000)]
Add cvs  tag

23 years agoUse (..., "%s", foo) instead of (..., foo) to avoid a warning about a
jhb [Sat, 24 Mar 2001 06:26:47 +0000 (06:26 +0000)]
Use (..., "%s", foo) instead of (..., foo) to avoid a warning about a
non-constant format string when calling kthread_create() to create an
ithread.

23 years ago- Define and use MAXCPU like the alpha and i386 instead of NCPUS.
jhb [Sat, 24 Mar 2001 06:22:57 +0000 (06:22 +0000)]
- Define and use MAXCPU like the alpha and i386 instead of NCPUS.
- Sort the sys/mutex.h include in mp_machdep.c into a closer to correct
  location.

23 years agoStick a prototype for handleclock() in machine/clock.h and include it
jhb [Sat, 24 Mar 2001 06:20:48 +0000 (06:20 +0000)]
Stick a prototype for handleclock() in machine/clock.h and include it
interrupt.c to quiet a warning.

23 years agoPreserve ssh directory across upgrades.
jkh [Sat, 24 Mar 2001 05:40:43 +0000 (05:40 +0000)]
Preserve ssh directory across upgrades.

Submitted by: Edwin Groothuis <edwin@mavetju.org>
PR: conf/25633

23 years agoThis is kind of a hack, but it should work. Currently, world is broken
peter [Sat, 24 Mar 2001 04:40:49 +0000 (04:40 +0000)]
This is kind of a hack, but it should work.  Currently, world is broken
because libc/rpc/key_call.c references uname(), and ps/print.c also
defines uname(), and ps is linked statically.  This leads to a symbol
clash.  The userland uname(3) kinda sucked anyway as the hostname
etc was too short.  And since the libc rpc interface now uses
the utsname.nodename which gets truncated, I was tempted into doing
something about it.  Create a new userland uname function, called
__xuname() which takes an extra argument that allows you to change
the size of the fields.  uname() becomes a static inline function
in sys/utsname.h that passes the extra argument in.  struct utsname
has its field members expanded by default now in userland.
We still provide a 'uname' externally linkable function for things
that either think that they ``know'' the utsname format and assume
32 character strings and bypass the include file, or objects that
are linked against old libcs.  ie: just about every plausible
case that I can think of is covered.  Should we ever change the
default lengths again, a libc major bump should not be required
as the size is now passed to the function.

XXX the uname(2) in the kernel is for FreeBSD 1.1 binary compatability!
All the uname(3) functions that are exported to userland are actually
implemented in libc with sysctl.  uname(1) uses sysctl directly and
does not call uname(3).

PR: bin/4688

23 years agoJust use the proc lock to protect read accesses to p_pptr rather than the
jhb [Sat, 24 Mar 2001 04:00:01 +0000 (04:00 +0000)]
Just use the proc lock to protect read accesses to p_pptr rather than the
more expensive proctree lock.

23 years agoProtect p_wmesg and p_wchan with sched_lock while checking for deadlocks
jhb [Sat, 24 Mar 2001 03:57:44 +0000 (03:57 +0000)]
Protect p_wmesg and p_wchan with sched_lock while checking for deadlocks
with other byte range file locks.

23 years agoafterstep doesn't need an explicit xterm started for it either.
jkh [Sat, 24 Mar 2001 03:17:35 +0000 (03:17 +0000)]
afterstep doesn't need an explicit xterm started for it either.

23 years agoThis commit was generated by cvs2svn to compensate for changes in r74722,
obrien [Sat, 24 Mar 2001 01:58:31 +0000 (01:58 +0000)]
This commit was generated by cvs2svn to compensate for changes in r74722,
which included commits to RCS files with non-trunk default branches.

23 years agoImport the setjump/longjump exception handling fixes from GCC 2.95.3.test3
obrien [Sat, 24 Mar 2001 01:58:31 +0000 (01:58 +0000)]
Import the setjump/longjump exception handling fixes from GCC 2.95.3.test3
that were removed from GCC 2.95.3.test4 and the subsequent release due
to problems on HP-UX.  However, they work just fine on all the BSD's.

W/o these patches the following program segmentation faults if compiled
with -O2 (but not -Os or -O or -O0):

#include <stdio.h>

class A {
public:
  A() { printf("c'tor A\n"); }
  ~A(){ printf("d'tor A\n"); }
};

class foo : public A {
public:
  foo()  { printf("C'tor foo\n"); throw 8; }
  ~foo() { printf("D'tor foo\n"); }
};

int main(){
  try { foo fii; }
  catch (int){ printf("catch ...\n"); }
  return 0;
}

23 years agos/dependant/dependent
jkh [Sat, 24 Mar 2001 01:26:51 +0000 (01:26 +0000)]
s/dependant/dependent

Submitted by: rwatson

23 years agorevert rev 1.3: restore MAN1 directive to unbreak world
alfred [Sat, 24 Mar 2001 01:17:51 +0000 (01:17 +0000)]
revert rev 1.3: restore MAN1 directive to unbreak world

23 years agoSlightly de-bogify the uname() declarations.
peter [Sat, 24 Mar 2001 00:34:12 +0000 (00:34 +0000)]
Slightly de-bogify the uname() declarations.
1: there is no global 'struct utsname utsname' in the kernel.
2: the __P() stuff handles the __STDC__ stuff, so the extra ifdef is
   redundant.

23 years agoInstall /etc/primes.
green [Sat, 24 Mar 2001 00:33:05 +0000 (00:33 +0000)]
Install /etc/primes.

23 years agoNow that libfetch uses the high port range by default, add a -U option to
des [Sat, 24 Mar 2001 00:32:11 +0000 (00:32 +0000)]
Now that libfetch uses the high port range by default, add a -U option to
make it use the low (default) port range instead.