]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
25 years agoCommit out caveat about hardlinks to directories since they are
Steve Price [Sun, 13 Dec 1998 23:35:01 +0000 (23:35 +0000)]
Commit out caveat about hardlinks to directories since they are
no longer possible.

PR: 8337

25 years agoFix typo - [0], [1], [2], not [0], [1], [3].
Eivind Eklund [Sun, 13 Dec 1998 23:32:44 +0000 (23:32 +0000)]
Fix typo - [0], [1], [2], not [0], [1], [3].

25 years agowcd_read_toc returns zero on success and EIO on failure, so change the
Steve Price [Sun, 13 Dec 1998 23:30:15 +0000 (23:30 +0000)]
wcd_read_toc returns zero on success and EIO on failure, so change the
test in open to return EIO if non-zero.

PR: 8316
Submitted by: Masatoshi TAMURA <tamrin@shinzan.kuee.kyoto-u.ac.jp>

25 years agosmall bug fix
Nick Hibma [Sun, 13 Dec 1998 23:26:41 +0000 (23:26 +0000)]
small bug fix

25 years agoCleanup usage string: -i takes an argument.
Steve Price [Sun, 13 Dec 1998 23:26:21 +0000 (23:26 +0000)]
Cleanup usage string: -i takes an argument.

PR: 8315
Submitted by: Amakawa Shuhei <amakawa@sf.t.u-tokyo.ac.jp>

25 years agoFix typo - sizeof(struct crtl *) -> sizeof(struct ctrl *).
Eivind Eklund [Sun, 13 Dec 1998 23:25:29 +0000 (23:25 +0000)]
Fix typo - sizeof(struct crtl *) -> sizeof(struct ctrl *).

XXX This still assume that bzero() over a field create null-pointers,
which seems a chancy proposition at best.

25 years agoRemoved unneeded definition
Nick Hibma [Sun, 13 Dec 1998 23:16:20 +0000 (23:16 +0000)]
Removed unneeded definition

25 years agoChange the wording to reflect that only the first character of the two
Steve Price [Sun, 13 Dec 1998 23:15:50 +0000 (23:15 +0000)]
Change the wording to reflect that only the first character of the two
arguments to changequote are significant.

PR: 8282

25 years agoAdded uhid again, sorting seems to work now
Nick Hibma [Sun, 13 Dec 1998 23:14:03 +0000 (23:14 +0000)]
Added uhid again, sorting seems to work now

25 years agoAdd support for another 3COM/USR PNP modem, USR3031.
Steve Price [Sun, 13 Dec 1998 23:12:54 +0000 (23:12 +0000)]
Add support for another 3COM/USR PNP modem, USR3031.

25 years agoAdded all the options to LINT with descriptions. Haven't tried to compile
Nick Hibma [Sun, 13 Dec 1998 23:06:16 +0000 (23:06 +0000)]
Added all the options to LINT with descriptions. Haven't tried to compile
the LINT kernel yet however...

25 years agodded the stubs for umodem and ucom (communications class driver). They are nothing...
Nick Hibma [Sun, 13 Dec 1998 23:04:35 +0000 (23:04 +0000)]
dded the stubs for umodem and ucom (communications class driver). They are nothing other than
the ugen driver with different variable names.

25 years agoRemove warning introduced by the last committer.
Eivind Eklund [Sun, 13 Dec 1998 23:00:48 +0000 (23:00 +0000)]
Remove warning introduced by the last committer.

25 years agoAnother extra pointer check.
Eivind Eklund [Sun, 13 Dec 1998 22:29:32 +0000 (22:29 +0000)]
Another extra pointer check.

25 years agoUpdated USB kernel sources to NetBSD sources of 1998-12-09.
Nick Hibma [Sun, 13 Dec 1998 22:27:42 +0000 (22:27 +0000)]
Updated USB kernel sources to NetBSD sources of 1998-12-09.
1 bug fix and several textual changes.
Preparing to feed back changes for port into NetBSD to create one source base.

25 years agoBlock for buffer overflow.
Eivind Eklund [Sun, 13 Dec 1998 21:02:28 +0000 (21:02 +0000)]
Block for buffer overflow.

25 years agoMade this actually work when there is an obj dir.
Bruce Evans [Sun, 13 Dec 1998 15:52:26 +0000 (15:52 +0000)]
Made this actually work when there is an obj dir.

Removed bogons, especially the include of bsd.subdir.mk.

Fixed style bugs.

25 years agoFixed disordering of cross references in previous commit.
Bruce Evans [Sun, 13 Dec 1998 15:32:26 +0000 (15:32 +0000)]
Fixed disordering of cross references in previous commit.

25 years agoUnbreak make world. Sorry about that. I'll have to figure out the steps
Nick Hibma [Sun, 13 Dec 1998 14:56:19 +0000 (14:56 +0000)]
Unbreak make world. Sorry about that. I'll have to figure out the steps
to add those programs properly.

25 years agoAdd a generic flag, CTLFLAG_SECURE, which can be used to mark a sysctl
Don Lewis [Sun, 13 Dec 1998 07:19:13 +0000 (07:19 +0000)]
Add a generic flag, CTLFLAG_SECURE, which can be used to mark a sysctl
variable unwriteable when securelevel > 0.
Reviewed by: jdp, eivind

25 years agoPR: bin/8624
Matthew Dillon [Sun, 13 Dec 1998 07:16:05 +0000 (07:16 +0000)]
PR: bin/8624

    Fixed intermediate calculation overflow when reporting users with > 2GB of
    disk space.

25 years agogetpgid() and getsid() were doing a comparision rather than an assignment,
Don Lewis [Sun, 13 Dec 1998 07:07:51 +0000 (07:07 +0000)]
getpgid() and getsid() were doing a comparision rather than an assignment,
which is fortunate, because otherwise another bug would allow them to be
used to stomp on the syscall return value of another process.

25 years agoPR: kern/8819
Matthew Dillon [Sun, 13 Dec 1998 06:50:17 +0000 (06:50 +0000)]
PR: kern/8819

    Trivial change of macro name from DEBUG to AWE_DEBUG to avoid conflict
    with kernel DEBUG option.

25 years agoPR: bin/9016
Matthew Dillon [Sun, 13 Dec 1998 06:40:18 +0000 (06:40 +0000)]
PR: bin/9016

    Fix bug with od/hd/hexdump. "*" lines are supposed to indicate one or
    duplicates of the previous line, but a small file with less then 16
    characters of zeros in it will be falsy identified as a repeat of
    the (non-existant) previous line.  i.e. the first line of output winds
    up being a "*".  Added a bit of code to handle the degenerate 'there is
    no previous line' case for the first line.

25 years agoPR: bin/5572
Matthew Dillon [Sun, 13 Dec 1998 06:07:38 +0000 (06:07 +0000)]
PR: bin/5572

    Prevent cron from going crazy if the time steps.  For example, if you
    have a system with hundreds of users and lots of different crontabs
    and your time steps back an hour, the old cron would then attempt to
    run an hours worth of cron jobs in a few seconds.

25 years ago Grr. removed. backed-out. sorry!
Matthew Dillon [Sun, 13 Dec 1998 04:39:01 +0000 (04:39 +0000)]
Grr. removed.  backed-out. sorry!

25 years ago Grrr. removed. backed out. sorry.
Matthew Dillon [Sun, 13 Dec 1998 04:38:28 +0000 (04:38 +0000)]
 Grrr.  removed. backed out. sorry.

25 years ago This needs to be commited now to fix usbd for make world
Matthew Dillon [Sun, 13 Dec 1998 04:32:13 +0000 (04:32 +0000)]
 This needs to be commited now to fix usbd for make world

25 years ago This needs to be commited now to fix usbd for make world
Matthew Dillon [Sun, 13 Dec 1998 04:31:15 +0000 (04:31 +0000)]
This needs to be commited now to fix usbd for make world

25 years agoPR: bin/8680
Matthew Dillon [Sun, 13 Dec 1998 03:44:19 +0000 (03:44 +0000)]
PR: bin/8680

    Minor adjustments to rwall.c.  Increased the size of a couple of char
    arrays and replaced a potentially dangeerous strcpy/strcat with snprintf().

25 years agoPR: bin/8680
Matthew Dillon [Sun, 13 Dec 1998 03:42:51 +0000 (03:42 +0000)]
PR: bin/8680

    Increase the size of a number of buffers and replace strcpy/strcat/sprintf
    with snprintf().  There was a minor and possibly not exploitable security
    hole related to one sprintf().

25 years ago Add reference to login.conf(5)
Matthew Dillon [Sun, 13 Dec 1998 02:48:43 +0000 (02:48 +0000)]
Add reference to login.conf(5)

25 years agoPR: docs/9045
Matthew Dillon [Sun, 13 Dec 1998 02:47:00 +0000 (02:47 +0000)]
PR: docs/9045

    Update chpass(1) manual page.  Change the definition of the class
    field and also reorganize the field list to match that of the
    /etc/master.passwd file, to avoid confusion.

25 years agoPR: docs/9050
Matthew Dillon [Sun, 13 Dec 1998 02:34:56 +0000 (02:34 +0000)]
PR: docs/9050

    Add reference to required include file #include <stdlib.h> for
    getloadavg(3) function call.

25 years ago Add required #include references to manual page
Matthew Dillon [Sun, 13 Dec 1998 02:32:59 +0000 (02:32 +0000)]
Add required #include references to manual page

25 years agoPR: bin/3478
Matthew Dillon [Sun, 13 Dec 1998 01:53:50 +0000 (01:53 +0000)]
PR: bin/3478

    Have pwd_mkdb lock the source file while rebuilding the database.  When
    called by programs such as vipw, the source file is a temporary file and
    this does not conflict with the lock on /etc/master.passwd already held
    by vipw.  When run manually, however, master.passwd is typically specified
    as the argument and the locking prevents other programs from messing with
    master.passwd during the database rebuild.

    Also pwd_mkdb uses a blocking exclusive lock as it may be called from
    a script.  The -N option was added to cause pwd_mkdb to get the lock
    non-blocking and exit with an error if the attempt fails, again useful
    for scripts.

25 years ago oops. Fix indentation of the 'for' loop I just added.
Matthew Dillon [Sun, 13 Dec 1998 01:39:32 +0000 (01:39 +0000)]
oops.  Fix indentation of the 'for' loop I just added.

25 years ago Handle the race condition where vipw may lock a password file which has
Matthew Dillon [Sun, 13 Dec 1998 01:36:45 +0000 (01:36 +0000)]
Handle the race condition where vipw may lock a password file which has
    just been replaced.  After our lock succeeds we check if st_nlink is 0
    and if it is we close the descriptor and retry our open/lock sequence.

25 years agoPull in fixes similar to those recently committed to the bt driver:
Justin T. Gibbs [Sun, 13 Dec 1998 00:05:04 +0000 (00:05 +0000)]
Pull in fixes similar to those recently committed to the bt driver:

 - For transactions of 0 length, us a non-residual checking CCB type.

 - Preserve command status if our interrupt handler completes a command
   while we are polling for completion in aha_cmd.

25 years agoReturn ENODEV instead of EINVAL when a particular exchange or move
Justin T. Gibbs [Sat, 12 Dec 1998 23:52:46 +0000 (23:52 +0000)]
Return ENODEV instead of EINVAL when a particular exchange or move
operation exceeds the capabilities of the changer device.

25 years agoReviewed by: freebsd-current
Matthew Dillon [Sat, 12 Dec 1998 23:26:53 +0000 (23:26 +0000)]
Reviewed by: freebsd-current

    Update manual pages for rc(8) and rc.conf(5) based on recent changes
    to rc.local and rc.conf[.local].

25 years ago Clarify rc's handling of rc.local
Matthew Dillon [Sat, 12 Dec 1998 23:08:34 +0000 (23:08 +0000)]
   Clarify rc's handling of rc.local

25 years agoReviewed by: cvs-current
Matthew Dillon [Sat, 12 Dec 1998 23:05:22 +0000 (23:05 +0000)]
Reviewed by: cvs-current

    Delete rc.local from CVS tree, its remaining functionality has been
    moved to /etc/rc.  /etc/rc still supports an rc.local but it is now
    a 100% user-controlled file.

25 years agoReviewed by: cvs-current
Matthew Dillon [Sat, 12 Dec 1998 23:04:21 +0000 (23:04 +0000)]
Reviewed by: cvs-current

    Commit changes to rc and rc.local, removing the remaining minimal
    functionality of rc.local into rc and commenting it out of rc.local
    prior to the deletion of rc.local from the CVS tree.

25 years agoBack out revs 1.181 and 1.182 which upset a few people. I hope those
John Birrell [Sat, 12 Dec 1998 22:00:49 +0000 (22:00 +0000)]
Back out revs 1.181 and 1.182 which upset a few people. I hope those
(3?) people will make an effort to help those who would have benefitted from
this change. And just telling them that they should read and understand
the significance of each message posted to -current is not really good
enough IMHO.

25 years agoPR: kern/8990
Matthew Dillon [Sat, 12 Dec 1998 21:45:49 +0000 (21:45 +0000)]
PR: kern/8990

    If timer calculation results in degenerate value (0), force it to 1
    to avoid divide-by-zero panic later on in calls to IGMP_RANDOM_DELAY().
    I considered simply adding 1 to the timer calculation, but was unsure
    if the calculation was part of the IGMP standard or not so did not want
    to mess with it for all cases.

25 years ago sendmsg() didn't like the MSG_EOR flag and returned an error. Removing the
Matthew Dillon [Sat, 12 Dec 1998 21:30:17 +0000 (21:30 +0000)]
sendmsg() didn't like the MSG_EOR flag and returned an error.  Removing the
    flag makes portal tcp operation work.

25 years agoPR: kern/8965
Matthew Dillon [Sat, 12 Dec 1998 21:07:09 +0000 (21:07 +0000)]
PR: kern/8965
Obtained from: Stephen Clawson <sclawson@cs.utah.edu>

    Wakeup anyone waiting on a mount point prior to returning from umount,
    whether an error occurs or not.  Fixes a stat/NFS-umount race and other
    potential future problems.  Fix taken from bug/pr which also indicated
    that the same fix has already been applied to OpenBSD and NetBSD.

25 years agoPR: bin/9031
Matthew Dillon [Sat, 12 Dec 1998 20:56:53 +0000 (20:56 +0000)]
PR: bin/9031

    Changed unbounded strcpy() to snprintf() to fix buffer overrun exploit

25 years agoAdd a simple library for accessing i/o ports and memory on the alpha.
Doug Rabson [Sat, 12 Dec 1998 18:05:06 +0000 (18:05 +0000)]
Add a simple library for accessing i/o ports and memory on the alpha.
This is only intended for use by the X server.

25 years agoClose PR bin/8753 pwd_mkdb problem when having comments in passwd file
Foxfair Hu [Sat, 12 Dec 1998 16:08:41 +0000 (16:08 +0000)]
Close PR bin/8753 pwd_mkdb problem when having comments in passwd file

Submitted by Chia-liang Kao clkao@CirX.ORG .

25 years agoFixed warning in usr.sbin/usbd
Nick Hibma [Sat, 12 Dec 1998 11:57:59 +0000 (11:57 +0000)]
Fixed warning in usr.sbin/usbd

25 years agoAdded usbd, usbdevs directories
Nick Hibma [Sat, 12 Dec 1998 11:47:31 +0000 (11:47 +0000)]
Added usbd, usbdevs directories
Submitted by: MIHIRA Sanpei Yoshiro

25 years agoWhen no driver was found for a device, the message 'not probed' appeared
Nick Hibma [Sat, 12 Dec 1998 11:30:04 +0000 (11:30 +0000)]
When no driver was found for a device, the message 'not probed' appeared
This is odd, especially in the case of USB where the driver is found
in several tries: vendor specific, class specific, interface specific.
The mouse driver is found at the interface specific level...
Reviewed by: Doug Rabson (dfr@freebsd.org)

25 years agoAdd support for the YE-Data external PCMCIA floppy driver. This
Warner Losh [Sat, 12 Dec 1998 08:16:01 +0000 (08:16 +0000)]
Add support for the YE-Data external PCMCIA floppy driver.  This
floppy is used on the toshiba Libretto line of subnotebook computers.
It differs from a normal floppy in that you must use PIO rather than
DMA to transfer the data.

To enable this, you must add options "FDC_YE" to your kernel.  I don't
have a machine that has a floppy and a pcmcia slot to test to make
sure that this doesn't impact normal floppy units, so I've left this as
an option.

I have ported this to -current and made an attempt to ensure that the
indentation conforms to style(9), aka the bruce filter.

Reviewed by: nate, markm
Submitted by: David Horwitt (dhorwitt@ucsd.edu)

25 years ago(0) Fast INDEX generation. Only print out the directory name and
Satoshi Asami [Sat, 12 Dec 1998 07:39:30 +0000 (07:39 +0000)]
(0) Fast INDEX generation.  Only print out the directory name and
    don't recurse in "make describe".  The new INDEX target in
    ports/Makefile invokes a perl script to recurse and convert them
    into package names.

    While I'm here, change the name of targets and move them around a
    little bit for the sake of consistency.

    It is also probably worth noting here that the meaning of the
    "build dependency" list in INDEX has been changed slightly
    changed.  The old list was "build depends and its build depends"
    -- not particularly useful if you had things like autoconf, which
    run-depend on gm4 (you install all the things listed here and
    you'll get an autoconf that won't run).

    It is now "build depends and its run depends" -- you install
    everything listed here, and you'll be able to build the port.
Submitted by: steve

(0') Fast README.html generation.  It uses ports/INDEX to find
    dependencies instead of embarking on to a recursive loop.
Submitted by: steve

(1) Remove NO_WRKDIR and NO_EXTRACT.  Their functionality are easily
    replacable with NO_WRKSUBDIR=t and EXTRACT_ONLY= (nothing on right
    side), and they get in the way of read-only port trees.

(2) Surround first few variable definitions with ".if !defined()".
    This will make cross-compilation easier and also speed up make
    processes.

(3) Call sysctl with absolute path.  Prefer the one in /sbin over the
    one in /usr/sbin.

(4) Add four new variables

    PKGINSTALL?= ${PKGDIR}/INSTALL
    PKGDEINSTALL?= ${PKGDIR}/DEINSTALL
    PKGREQ?= ${PKGDIR}/REQ
    PKGMESSAGE?= ${PKGDIR}/MESSAGE

    and use them in PKG_ARGS.  Frobbing with PKG_ARGS directly is
    strongly discouraged.

(5) Change PKG_SUFX to ".tar" (instead of ".tgz") if PKG_NOCOMPRESS is
    defined.  This is intended only for our own use.

(6) Add more sites to MASTER_SITE_GNU.
Submitted by: billf

(7) Override MANUAL_PACKAGE_BUILD if PARALLEL_PACKAGE_BUILD is
    defined.  This is intended only for our own use.

(8) Add new target "ignorelist" which will print out the package name
    if the port is not going to be built on this machine.  This is
    intended only for our own use.

(9) Make mtree a little quieter.

25 years agoAdd working client and server sample applications.
Poul-Henning Kamp [Fri, 11 Dec 1998 21:48:20 +0000 (21:48 +0000)]
Add working client and server sample applications.

Submitted by: Mike Spengler <mks@circe.networkcs.com>

25 years agoSet and propagate the correct cause code values.
Poul-Henning Kamp [Fri, 11 Dec 1998 21:47:47 +0000 (21:47 +0000)]
Set and propagate the correct cause code values.

Submitted by: Mike Spengler <mks@circe.networkcs.com>

25 years agoCopy application name and connection attributes for accept() sockets from
Poul-Henning Kamp [Fri, 11 Dec 1998 21:47:21 +0000 (21:47 +0000)]
Copy application name and connection attributes for accept() sockets from
   listen() socket.

Allow setsockopt(APPLNAME) after socket is connected.

25 years ago(almost) null commit, recording message for previous commit:
Poul-Henning Kamp [Fri, 11 Dec 1998 21:42:57 +0000 (21:42 +0000)]
(almost) null commit, recording message for previous commit:

s/_NET_IF_HDLC_H_/_NET_IF_SPPP_H_/

Unfold almost correct and hideous beyond reason, boolean expression,
making it more correct at the same time.

25 years ago*** empty log message ***
Poul-Henning Kamp [Fri, 11 Dec 1998 21:40:13 +0000 (21:40 +0000)]
*** empty log message ***

25 years agoReviewed by: freebsd-current
Matthew Dillon [Fri, 11 Dec 1998 17:06:16 +0000 (17:06 +0000)]
Reviewed by: freebsd-current

    Fix signal/library corruption by blocking all signals except during
    select().  The reported corruption was with reentrancy in the malloc lib.

25 years agoMerged from Lite2 (one bcopy -> memmove, one significant change:
Bruce Evans [Fri, 11 Dec 1998 11:21:47 +0000 (11:21 +0000)]
Merged from Lite2 (one bcopy -> memmove, one significant change:
don't unlink _PATH_NOLOGIN for the -k case even if shutdown terminates
abnormally.  NetBSD already has this change).

25 years agoFixed missing 'p' and `-' flags and other defects in the usage message.
Bruce Evans [Fri, 11 Dec 1998 11:04:19 +0000 (11:04 +0000)]
Fixed missing 'p' and `-' flags and other defects in the usage message.

Fixed some style bugs.

25 years agoAdd directories for KLD examples.
Doug Rabson [Fri, 11 Dec 1998 10:45:29 +0000 (10:45 +0000)]
Add directories for KLD examples.

25 years agoSample modules for KLD.
Doug Rabson [Fri, 11 Dec 1998 10:44:31 +0000 (10:44 +0000)]
Sample modules for KLD.

PR: misc/8621
Submitted by: Rajesh Vaidheeswarran <rv@fore.com>

25 years agoMerged from Lite2 (just an English fix or pun removal).
Bruce Evans [Fri, 11 Dec 1998 10:35:38 +0000 (10:35 +0000)]
Merged from Lite2 (just an English fix or pun removal).

Updated date.  I think dates in man pages should be changed at least
when a new feature is described.

25 years agoFixed disordered options in synopsis.
Bruce Evans [Fri, 11 Dec 1998 10:25:55 +0000 (10:25 +0000)]
Fixed disordered options in synopsis.

25 years agoFixed missing `p' flag in synopsis.
Bruce Evans [Fri, 11 Dec 1998 10:20:54 +0000 (10:20 +0000)]
Fixed missing `p' flag in synopsis.

Broken in: previous commit

25 years agoSync with sys/i386/isa/lpt.c revision 1.72.
KATO Takenori [Fri, 11 Dec 1998 08:48:21 +0000 (08:48 +0000)]
Sync with sys/i386/isa/lpt.c revision 1.72.

25 years agoShould be sysctl -n. Yesterday wasn't one of my better days. Doh.
John Birrell [Fri, 11 Dec 1998 08:25:12 +0000 (08:25 +0000)]
Should be sysctl -n. Yesterday wasn't one of my better days. Doh.

Reported by: Ben Smithurst & Makoto Matsushita

25 years agoSync with sys/i386/i386/machdep.c revision 1.317.
KATO Takenori [Fri, 11 Dec 1998 08:04:33 +0000 (08:04 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.317.

25 years agoSome fixes to handle fixed mode and variable mode more sensibly- and also
Matt Jacob [Fri, 11 Dec 1998 07:19:36 +0000 (07:19 +0000)]
Some fixes to handle fixed mode and variable mode more sensibly- and also
incorporate some notion of which revision the device is. If it's < SCSI2, for
example, READ BLOCK LIMITS is not a MANDATORY command.

At any rate, the initial state is to try and read block limits to get a notion
of the smallest and largest record size as well as the granularity. However,
this doesn't mean that the device should actually *in* fixed block mode should
the max && min  be equal... *That* choice is (for now) determined by whether
the device comes up with a blocksize of nonzero. If so, then it's a fixed block
preferred device, otherwise not (this will change again soon).

When actually doing I/O, and you're in fixed length mode, the block count is
*not* the byte count divided by the minimum block size- it's the byte count
divided by the current blocksize (or use shift/mask shortcuts if that worked
out...).

Then when you *change* the blocksize via an ioctl, make sure this actually
propagates to the stored notion of blocksize (and update the shift/mask
shortcuts).

Misc Other:
When doing a mode select, only use the SCSI_SAME_DENSITY (0x7f) code if
the device is >= SCSI2- otherwise just use the saved density code.

Recover from the ripple of ILLEGAL REQUEST not being 'retried' in that
RESERVE/RELEASE is not a mandatory command for < SCSI2 (so ignore it if it
fails).

25 years agoIf we failed to probe/attach somehow, we never have a sc->sc_hcca
Julian Elischer [Fri, 11 Dec 1998 06:02:06 +0000 (06:02 +0000)]
If we failed to probe/attach somehow, we never have a sc->sc_hcca
but the present PCI probe code still thinks we are there as the pci attach
can't return an error code.

This means we are in the shared interrupt list, but have not been set up.
If we are sharing ints with another device, ohci_intr will be called and will
coredump on a NULL reference. So just return if it is called when not set up.

This fixes the symptom and not the cause.
The right answer is to let the PCI system know that the attach failed,
or to fail earlier (in the PCI probe).
The attach() is a void fn() so it can't return failure..

25 years agoConvert dadump to use reasonable data types so that some casting is unecessary.
Justin T. Gibbs [Fri, 11 Dec 1998 03:54:43 +0000 (03:54 +0000)]
Convert dadump to use reasonable data types so that some casting is unecessary.

25 years agoDo not attempt to retry commands that fail with ILLEGAL REQUEST status.
Justin T. Gibbs [Fri, 11 Dec 1998 03:53:05 +0000 (03:53 +0000)]
Do not attempt to retry commands that fail with ILLEGAL REQUEST status.

25 years agoRemove unsued variable.
Justin T. Gibbs [Fri, 11 Dec 1998 03:50:35 +0000 (03:50 +0000)]
Remove unsued variable.

If we are not transfering any data, use a non S/G ccb type that doesn't
return residual information.  It seems that some firmware revisions dislike
S/G ccbs with 0 length S/G lists.

Correct bt_cmd() so that we always honor command status that was latched
by our interrupt routine while polling for completion..

25 years agoFixed same as Julian did in uhci_pci.c and fixed the fetching of the revision ohci_pc...
Nick Hibma [Fri, 11 Dec 1998 00:09:54 +0000 (00:09 +0000)]
Fixed same as Julian did in uhci_pci.c and fixed the fetching of the revision ohci_pci.c; clean up

25 years agoChange to the current directory before doing the install. I !love make.
John Birrell [Fri, 11 Dec 1998 00:09:39 +0000 (00:09 +0000)]
Change to the current directory before doing the install. I !love make.

25 years agoAdd a '-p' flag to shutdown which corresponds to the '-p' flag to halt,
Mike Smith [Thu, 10 Dec 1998 23:54:02 +0000 (23:54 +0000)]
Add a '-p' flag to shutdown which corresponds to the '-p' flag to halt,
requesting a system power-off after shutdown.

25 years agoReversed the order of the USB devices, makes sure the ugen device is probed last
Nick Hibma [Thu, 10 Dec 1998 23:36:46 +0000 (23:36 +0000)]
Reversed the order of the USB devices, makes sure the ugen device is probed last

25 years agoPerform APM power-off on power-off request, not halt request.
Mike Smith [Thu, 10 Dec 1998 23:36:14 +0000 (23:36 +0000)]
Perform APM power-off on power-off request, not halt request.

25 years agoFix typo.
Jun Kuriyama [Thu, 10 Dec 1998 22:07:06 +0000 (22:07 +0000)]
Fix typo.

25 years agoThe shift and the masking were in the wrong order for extracting
Julian Elischer [Thu, 10 Dec 1998 22:07:05 +0000 (22:07 +0000)]
The shift and the masking were in the wrong order for extracting
the INTERFACE type byte from the longword register.

25 years agoCALL -> PCALL for sigaltstack for libc_r.
John Birrell [Thu, 10 Dec 1998 20:36:24 +0000 (20:36 +0000)]
CALL -> PCALL for sigaltstack for libc_r.

25 years agoDon't hide mknod, it doesn't need a wrapper and never has had one.
John Birrell [Thu, 10 Dec 1998 20:27:52 +0000 (20:27 +0000)]
Don't hide mknod, it doesn't need a wrapper and never has had one.

25 years agoRemove some compiler warnings.
Julian Elischer [Thu, 10 Dec 1998 20:11:47 +0000 (20:11 +0000)]
Remove some compiler warnings.

25 years agoRename one of the two devfs_link's to devfs_makelink.
Eivind Eklund [Thu, 10 Dec 1998 19:57:01 +0000 (19:57 +0000)]
Rename one of the two devfs_link's to devfs_makelink.

25 years agoSilence gcc -Wall -ansi -pedantic.
Bill Paul [Thu, 10 Dec 1998 19:02:07 +0000 (19:02 +0000)]
Silence gcc -Wall -ansi -pedantic.

Pointed out by: Eivind

25 years agoBoy, this was tricky to find:
Brian Somers [Thu, 10 Dec 1998 18:36:53 +0000 (18:36 +0000)]
Boy, this was tricky to find:
  Remove any dial timer that might be hanging around at
  datalink_Destroy() time.  This timer may be left running
  after the link is closed (making sure it's not automatically
  opened again too soon).

25 years agoWhen using ``set device !someprogram'', when ``someprogram''
Brian Somers [Thu, 10 Dec 1998 18:36:30 +0000 (18:36 +0000)]
When using ``set device !someprogram'', when ``someprogram''
exits, it causes a select() exception.
Handle these select() exceptions on link descriptors in pretty
much the same way as loss of carrier rather than dropping out
in confusion.

25 years agoSupport ESS1868 (and probably ESS688 & ESS1668).
Brian Somers [Thu, 10 Dec 1998 18:36:10 +0000 (18:36 +0000)]
Support ESS1868 (and probably ESS688 & ESS1668).
Submitted by: Max Khon <fjoe@husky.iclub.nsu.ru>

25 years agoSilence gcc -Wall -ansi -pedantic and fix minor bug in xl_txeof().
Bill Paul [Thu, 10 Dec 1998 16:18:43 +0000 (16:18 +0000)]
Silence gcc -Wall -ansi -pedantic and fix minor bug in xl_txeof().

Pointed out by: Eivind

25 years agolinux_pipe does not preserve the edx register. Linux and
Jordan K. Hubbard [Thu, 10 Dec 1998 13:47:18 +0000 (13:47 +0000)]
linux_pipe does not preserve the edx register.  Linux and
programs using glibc expect edx to be preserved accross syscalls.
As a result, linux programs running in emulation mode can
have whatever value may be represented by edx clobbered.

PR: 9038
Submitted-By: Richard Seaman, Jr. <dick@tar.com>

25 years agoAdd a test for hw.machine == i386 before trying to run ldconfig for
John Birrell [Thu, 10 Dec 1998 08:06:59 +0000 (08:06 +0000)]
Add a test for hw.machine == i386 before trying to run ldconfig for
legacy aout support.

25 years agoAdd logic to check if any of the BIN1 files do not already exist in
John Birrell [Thu, 10 Dec 1998 05:34:11 +0000 (05:34 +0000)]
Add logic to check if any of the BIN1 files do not already exist in
${DESTDIR}/etc and an install target to install the missing ones. This
allows new files like pam.conf to be installed by the first installworld
after the file is added, but avoid clobbering files that might be
customized. This should save some support questions.

25 years agoRevamp the way that exceptional message handling is performed so that it
Justin T. Gibbs [Thu, 10 Dec 1998 04:14:50 +0000 (04:14 +0000)]
Revamp the way that exceptional message handling is performed so that it
is more robust and common code can be used for both the target and iniator
roles.  The mechanism for tracking negotiation state has also been simplified.

Add support for sync/wide negotiation in target mode and fix many of
the target mode bugs running at higher speeds uncovered.  Make a first
stab at getting all of the bus skew delays correct.  Sync+Wide dataout
transfers still cause problems, but this may be an initiator problem.

Ensure that we exit BITBUCKET mode if the controller is restarted.

Add support for target mode only firmware downloads.  This has been
tested on the aic7880, but should mean that we can perform target mode
on any aic7xxx controller.  Mixed mode (initiator and target roles in
the same firmware load) is currently only supported on the aic7890, but
with optimization, may fit on chips with less instruction space.

25 years agoConvert debugging printfs to the CAM_DEBUG macro.
Justin T. Gibbs [Thu, 10 Dec 1998 04:07:42 +0000 (04:07 +0000)]
Convert debugging printfs to the CAM_DEBUG macro.

Allow sync transfers if the controller supports it.  Wide will follow
as soon as I get the kinks worked out of wide target transfers in the
aic7xxx driver (currently the only target mode driver in the tree).

25 years agoExpand the hba_misc fied in the Path Inquiry ccb to allow a controller driver
Justin T. Gibbs [Thu, 10 Dec 1998 04:05:49 +0000 (04:05 +0000)]
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver
to specify that it does not provide initiator services (PIM_NOINITIATOR)
and that the initial bus reset for device probing should be avoided
(PIM_NOBUSRESET).

Modify the XPT layer to honor these flags.

25 years agoRemove debugging printfs.
Justin T. Gibbs [Thu, 10 Dec 1998 04:00:03 +0000 (04:00 +0000)]
Remove debugging printfs.