]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoDon't clobber the default for CFLAGS.
Bruce Evans [Fri, 31 Aug 2001 11:05:24 +0000 (11:05 +0000)]
Don't clobber the default for CFLAGS.

23 years agoDon't clobber the default for CFLAGS.
Bruce Evans [Fri, 31 Aug 2001 11:01:20 +0000 (11:01 +0000)]
Don't clobber the default for CFLAGS.

Reviewed by: dfr

23 years agoAdd manpage for inet_net_ntop(3) and inet_net_pton(3).
Ruslan Ermilov [Fri, 31 Aug 2001 10:54:44 +0000 (10:54 +0000)]
Add manpage for inet_net_ntop(3) and inet_net_pton(3).

Obtained from: NetBSD

23 years agoUse ``.Rv -std'' wherever possible.
Ruslan Ermilov [Fri, 31 Aug 2001 09:57:38 +0000 (09:57 +0000)]
Use ``.Rv -std'' wherever possible.

Submitted by: yar

23 years agoo Move definition of CAP_MAX_BUF_LEN to sys/capability.h
Robert Watson [Fri, 31 Aug 2001 07:07:28 +0000 (07:07 +0000)]
o Move definition of CAP_MAX_BUF_LEN to sys/capability.h

(I could have sworn I committed this before, but apparently I missed it
 during the merge, breaking world)

Submitted by: tmm
Obtained from: TrustedBSD Project
Pointed out by: Mike Barcroft <mike@FreeBSD.org>

23 years agoSimilar to changes on i386/alpha/etc pmap.c; converge on a similar
Peter Wemm [Fri, 31 Aug 2001 06:42:45 +0000 (06:42 +0000)]
Similar to changes on i386/alpha/etc pmap.c; converge on a similar
look/feel on pmap_new_proc() with some cosmetic style changes.

23 years agoConverge with i386/alpha/etc pmap.c for pmap_new_proc/pmap_dispose_proc().
Peter Wemm [Fri, 31 Aug 2001 06:30:27 +0000 (06:30 +0000)]
Converge with i386/alpha/etc pmap.c for pmap_new_proc/pmap_dispose_proc().

23 years agoSame as i386/i386/pmap.c: clean up some style. This is irrelevant since
Peter Wemm [Fri, 31 Aug 2001 06:25:28 +0000 (06:25 +0000)]
Same as i386/i386/pmap.c: clean up some style.  This is irrelevant since
it is inside #if 0'ed code, but it would be a shame if this stuff got
cut/pasted elsewhere.

23 years agoSame treatment as i386/isa/pmap.c - clean up some style to sync towards
Peter Wemm [Fri, 31 Aug 2001 06:19:08 +0000 (06:19 +0000)]
Same treatment as i386/isa/pmap.c - clean up some style to sync towards
KSE diff.

23 years agoCall OS-independent resume routine to execute _WAK .. etc.
Takanori Watanabe [Fri, 31 Aug 2001 05:36:48 +0000 (05:36 +0000)]
Call OS-independent resume routine to execute _WAK .. etc.
This should also recover GPE bit,comment says, though not implemented yet.

23 years agoRevert previous delta; <arpa/inet.h> isn't quite ready to stop including
Mike Barcroft [Fri, 31 Aug 2001 03:29:48 +0000 (03:29 +0000)]
Revert previous delta; <arpa/inet.h> isn't quite ready to stop including
<sys/types.h>.

23 years agoAttempt to make key behavior more clear.
Eric Melville [Fri, 31 Aug 2001 03:29:25 +0000 (03:29 +0000)]
Attempt to make key behavior more clear.

23 years agoo Deprecate byteorder(3) prototypes from <sys/types.h>, these are
Mike Barcroft [Fri, 31 Aug 2001 03:12:01 +0000 (03:12 +0000)]
o Deprecate byteorder(3) prototypes from <sys/types.h>, these are
  now prototyped indirectly in <arpa/inet.h>.
o Deprecate in_addr_t and in_port_t typedefs in <sys/types.h>, these
  are now typedef'd in <arpa/inet.h>.

Discussed with: bde
PR: 29946

23 years agoDo a style cleanup pass for the pmap_{new,dispose,etc}_proc() functions
Peter Wemm [Fri, 31 Aug 2001 02:28:01 +0000 (02:28 +0000)]
Do a style cleanup pass for the pmap_{new,dispose,etc}_proc() functions
to get them closer to the KSE tree.  I will do the other $machine/pmap.c
files shortly.

23 years agoRemove already commented out entry for pci/fxp.c
Jonathan Lemon [Fri, 31 Aug 2001 02:22:48 +0000 (02:22 +0000)]
Remove already commented out entry for pci/fxp.c

23 years agoThe fxp driver has lived in dev/fxp for some time, remove old files.
Jonathan Lemon [Fri, 31 Aug 2001 02:21:03 +0000 (02:21 +0000)]
The fxp driver has lived in dev/fxp for some time, remove old files.

23 years agoPrototypes for cap_equal_np() and cap_subset_np().
Robert Watson [Fri, 31 Aug 2001 02:17:43 +0000 (02:17 +0000)]
Prototypes for cap_equal_np() and cap_subset_np().

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoNote that ASU is now deprecated.
Robert Watson [Fri, 31 Aug 2001 02:15:13 +0000 (02:15 +0000)]
Note that ASU is now deprecated.

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoRevert previous "fix"; bus mice still have to be manually probed even in
Mike Smith [Fri, 31 Aug 2001 02:14:34 +0000 (02:14 +0000)]
Revert previous "fix"; bus mice still have to be manually probed even in
the presence of PnP data.

Submitted by: yokota

23 years agoo Use .Fx to refer to FreeBSD
Robert Watson [Fri, 31 Aug 2001 02:12:54 +0000 (02:12 +0000)]
o Use .Fx to refer to FreeBSD

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoo Remove definition of CAP_MAX_BUF_LEN since it is defined in
Robert Watson [Fri, 31 Aug 2001 02:11:59 +0000 (02:11 +0000)]
o Remove definition of CAP_MAX_BUF_LEN since it is defined in
  sys/capability.h now.

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoTidy up cpu_fork() a little. This is mainly for changes I've been
Peter Wemm [Fri, 31 Aug 2001 02:11:24 +0000 (02:11 +0000)]
Tidy up cpu_fork() a little.  This is mainly for changes I've been
working on in the KSE area.

23 years ago Introduce implementations of POSIX.1e non-portable form capability
Robert Watson [Fri, 31 Aug 2001 02:07:48 +0000 (02:07 +0000)]
 Introduce implementations of POSIX.1e non-portable form capability
  support functions:
     cap_subset_np()    - Is cap1 a subset of cap2
     cap_equal_np()     - Is cap1 equal to cap2

o Introduce implementations of POSIX.1e capability support functions:
     cap_copy_ext()     - Externalize capability
     cap_copy_int()     - Internalize capability
     cap_size()         - Determine size required for cap_copy_ext()

Submitted by: tmm
Obtained from: TrustedBSD Project

23 years agoProperly move cursor when home and end keys are used.
Eric Melville [Fri, 31 Aug 2001 01:56:06 +0000 (01:56 +0000)]
Properly move cursor when home and end keys are used.

23 years agoCleanup
Matthew Dillon [Fri, 31 Aug 2001 01:26:30 +0000 (01:26 +0000)]
Cleanup

23 years agoGet rid of most of the GIANT_XXX assertion defines. Nobody is going to use
Matthew Dillon [Fri, 31 Aug 2001 00:51:35 +0000 (00:51 +0000)]
Get rid of most of the GIANT_XXX assertion defines.  Nobody is going to use
them, including me.

23 years agoGiant pushdown syscalls in kern/uipc_syscalls.c. Affected calls:
Matthew Dillon [Fri, 31 Aug 2001 00:37:34 +0000 (00:37 +0000)]
Giant pushdown syscalls in kern/uipc_syscalls.c.  Affected calls:

recvmsg(), sendmsg(), recvfrom(), accept(), getpeername(), getsockname(),
socket(), connect(), accept(), send(), recv(), bind(), setsockopt(), listen(),
sendto(), shutdown(), socketpair(), sendfile()

23 years agoAdd a UGAR() macro to simplify the diff's for the Giant pushdown.
John Baldwin [Fri, 31 Aug 2001 00:36:29 +0000 (00:36 +0000)]
Add a UGAR() macro to simplify the diff's for the Giant pushdown.

23 years agoMake my Grandfather famous by showing the world my middle name,
Adrian Chadd [Fri, 31 Aug 2001 00:25:16 +0000 (00:25 +0000)]
Make my Grandfather famous by showing the world my middle name,
as seems to be the trend.

(Thanks Josef. :-)

23 years agoGiant Pushdown: sysv shm, sem, and msg calls.
Matthew Dillon [Fri, 31 Aug 2001 00:02:18 +0000 (00:02 +0000)]
Giant Pushdown:  sysv shm, sem, and msg calls.

23 years agoAdd names for the TI PCI-1210, TI PCI-4410 and TI PCI-4450 parts. I had
Warner Losh [Thu, 30 Aug 2001 22:48:47 +0000 (22:48 +0000)]
Add names for the TI PCI-1210, TI PCI-4410 and TI PCI-4450 parts.  I had
this for a while, and don't know how it didn't make it into the tree.

23 years agoMove /etc/defaults/make.conf to /usr/share/examples/etc/make.conf as
Alexander Langer [Thu, 30 Aug 2001 22:44:51 +0000 (22:44 +0000)]
Move /etc/defaults/make.conf to /usr/share/examples/etc/make.conf as
discussed on the arch@ mailinglist (after repo-copy).

sys.mk will .error if it finds /etc/defaults/make.conf but include
it anyways (this is the same behaviour as with the make.conf.local
removal).

/usr/share/examples/etc/make.conf has BDEFLAGS commented out now,
since it's only an example file.

Adjust all textes that talk about make.conf or defaults/make.conf to
match the new situation.

23 years agoAdd safety belts. A control endpoint doesn't have an endpoint descriptor.
Nick Hibma [Thu, 30 Aug 2001 21:45:28 +0000 (21:45 +0000)]
Add safety belts. A control endpoint doesn't have an endpoint descriptor.

23 years agoAdd Xref to make.conf(5).
Alexander Langer [Thu, 30 Aug 2001 21:44:46 +0000 (21:44 +0000)]
Add Xref to make.conf(5).

23 years agoNuke all ports related options. This file has to be re-edited once
Alexander Langer [Thu, 30 Aug 2001 21:40:55 +0000 (21:40 +0000)]
Nuke all ports related options.  This file has to be re-edited once
the ports-team has decided about the ports.conf issue.

23 years agoadd a missing -o.
Luigi Rizzo [Thu, 30 Aug 2001 21:39:14 +0000 (21:39 +0000)]
add a missing -o.
There might be some other bug related to OBJDIR handling, once
fixed these changes can be MFC immediately.

23 years agoNote the new cards supported by the asr and aac drivers.
Scott Long [Thu, 30 Aug 2001 21:37:09 +0000 (21:37 +0000)]
Note the new cards supported by the asr and aac drivers.

MFC after: 1 day

23 years agoCorrected a typo.
Valentino Vaschetto [Thu, 30 Aug 2001 21:12:35 +0000 (21:12 +0000)]
Corrected a typo.

Submitted by: Seth Kingsley <seth.kingsley@windriver.com>
Approved by: murray

23 years agoNew release note: SA-01:58.
Bruce A. Mah [Thu, 30 Aug 2001 21:06:43 +0000 (21:06 +0000)]
New release note:  SA-01:58.

23 years agoAdd some missing words, markup tweaks.
Bruce A. Mah [Thu, 30 Aug 2001 20:56:42 +0000 (20:56 +0000)]
Add some missing words, markup tweaks.

Commit from: SIGCOMM 2001

23 years agoTake CAM_REQUEUE_REQ out of the class of things we were trying to honor
Matt Jacob [Thu, 30 Aug 2001 20:54:00 +0000 (20:54 +0000)]
Take CAM_REQUEUE_REQ out of the class of things we were trying to honor
retry count on.

MFC after: 4 weeks

23 years agoTry to discard some ungetc data in saved internal buffer checks too,
Andrey A. Chernov [Thu, 30 Aug 2001 20:49:47 +0000 (20:49 +0000)]
Try to discard some ungetc data in saved internal buffer checks too,
if offset tends to be negative.

23 years agogoto dumb; if can't obtain curoff for whence != SEEK_CUR cases, as supposed
Andrey A. Chernov [Thu, 30 Aug 2001 20:19:45 +0000 (20:19 +0000)]
goto dumb; if can't obtain curoff for whence != SEEK_CUR cases, as supposed

23 years agoAdd more EOVERFLOW checks.
Andrey A. Chernov [Thu, 30 Aug 2001 19:54:04 +0000 (19:54 +0000)]
Add more EOVERFLOW checks.

When file offset tends to be negative due to internal and ungetc buffers
additions counted, try to discard some ungetc data first, then return EBADF.
Later one can happens if lseek(fileno(fd),...) called f.e. POSIX says that
ungetc beyond beginning of the file results are undefined, so we can just
discard some of ungetc data in that case.

Don't rely on gcc cast when checking for overflow, use OFF_MAX.

Cosmetique.

23 years agoFix kernel crash when VLANs are combined with ng_ether(4), by attaching
Archie Cobbs [Thu, 30 Aug 2001 19:09:10 +0000 (19:09 +0000)]
Fix kernel crash when VLANs are combined with ng_ether(4), by attaching
interfaces of type IFT_L2VLAN as well as IFT_ETHER during module load.

Submitted by: yar

23 years agoRemove the MPSAFE keyword from the parser for syscalls.master.
Matthew Dillon [Thu, 30 Aug 2001 18:50:57 +0000 (18:50 +0000)]
Remove the MPSAFE keyword from the parser for syscalls.master.
Instead introduce the [M] prefix to existing keywords.  e.g.
MSTD is the MP SAFE version of STD.  This is prepatory for a
massive Giant lock pushdown.  The old MPSAFE keyword made
syscalls.master too messy.

Begin comments MP-Safe procedures with the comment:
/*
 * MPSAFE
 */
This comments means that the procedure may be called without
Giant held (The procedure itself may still need to obtain
Giant temporarily to do its thing).

sv_prepsyscall() is now MP SAFE and assumed to be MP SAFE
sv_transtrap() is now MP SAFE and assumed to be MP SAFE

ktrsyscall() and ktrsysret() are now MP SAFE (Giant Pushdown)
trapsignal() is now MP SAFE (Giant Pushdown)

Places which used to do the if (mtx_owned(&Giant)) mtx_unlock(&Giant)
test in syscall[2]() in */*/trap.c now do not.  Instead they
explicitly unlock Giant if they previously obtained it, and then
assert that it is no longer held to catch broken system calls.

Rebuild syscall tables.

23 years agoAxe all the plists and supporting scripts to manage them and use them to
John Baldwin [Thu, 30 Aug 2001 18:11:42 +0000 (18:11 +0000)]
Axe all the plists and supporting scripts to manage them and use them to
roll dists.

23 years agoOverhaul this script a bit:
John Baldwin [Thu, 30 Aug 2001 18:10:26 +0000 (18:10 +0000)]
Overhaul this script a bit:
- build_x.sh now does all the steps to build an X dist rather than being
  a child script of package_x_dists.sh
- Update the usage information to list the ports you need to install
  before running this script as well as needing to set CVSROOT.
- Make sure CVSROOT is set.  If not, exit with an error.
- We now take two parameters: a work directory and an output directory.
  The work directory is used as scratch space.  All of the bindist will
  end up in the output directory.
- Only apply XF86.patch to the checked out XFree86 port if it exists.
- Use XFree86's build-bindist tool to package up the dists using their
  distfiles rather than using our own packing lists that have to be
  manually updated each time the port changes.

Example usage:
env CVSROOT=/home/ncvs ./build_x.sh /usr/xtmp /usr/x11dists

23 years agoThe current XFree86 port needs some help in the BUILD_XDIST case. Commit
John Baldwin [Thu, 30 Aug 2001 17:53:00 +0000 (17:53 +0000)]
The current XFree86 port needs some help in the BUILD_XDIST case.  Commit
this patch here until I can get the port fixed.

23 years agoRevert the previous delta and apply a better fix which corrects
Mike Barcroft [Thu, 30 Aug 2001 17:13:15 +0000 (17:13 +0000)]
Revert the previous delta and apply a better fix which corrects
a check on the final snprintf and reduces duplicated code.

Submitted by: brian

23 years agoRemove old acpica module
Mike Smith [Thu, 30 Aug 2001 17:11:11 +0000 (17:11 +0000)]
Remove old acpica module

23 years agoACPI no longer has an ISA attachment.
Mike Smith [Thu, 30 Aug 2001 17:00:33 +0000 (17:00 +0000)]
ACPI no longer has an ISA attachment.

23 years agoClear SA_FLAG_ERR_PENDING for MTREW, MTERASE and MTRETENS ioctl cases.
Matt Jacob [Thu, 30 Aug 2001 16:25:24 +0000 (16:25 +0000)]
Clear SA_FLAG_ERR_PENDING for MTREW, MTERASE and MTRETENS ioctl cases.
Clear residual counts after a successful samount (the user doesn't
care that we got an N-kbyte residual on our test read).

Change a lot of error handling code.

1. If we end up in saerror, check more carefully about the kind of
error. If it is a CAM_SCSI_STATUS_ERROR and it is a read/write
command, we'll be handling this in saerror.  If it isn't a read/write
command, check to see whether this is just an EOM/EOP check condition-
if it is, just set residual and return normally. A residual and
then a NO SENSE check condiftion with the ASC of 0 and ASCQ of
between 1 and 4 are normal 'signifying' events, not errors per se,
and we shouldn't give the command to cam_periph_error to do something
relatively unpredictable with.

2. If we get a Bus Reset, had a BDR sent, or get the cam status of
CAM_REQUEUE_REQ, check the retry count on the command. The default
error handler, cam_periph_error, doesn't honor retry count in these
cases. This may change in the future, but for now, make sure we
set EIO and return without calling cam_periph_error if the retry
count for the command with an error is zero.

3. Clean up the pending error case goop and handle cases more
sensibly.

The rules are:

 If command was a Write:

  If we got a SSD_KEY_VOLUME_OVERFLOW, the resid is
  propagated and we set ENOSPC as the error.

  Else if we got an EOM condition- just mark EOM pending.

And set a residual of zero. For the longest time I was just
        propagating residual from the sense data- but my tape
        comparison tests were always failing because all drives I
        tested with actually *do* write the data anyway- the EOM
        (early warning) condition occurred *prior* to all of the
        data going out to media- that is, it was still buffered by
        the drive. This case is described in SCSI-2, 10.2.14,
        paragraph #d for the meaning of 'information field'. A
        better fix for this would be to issue a WFM command of zero
        to cause the drive to flush any buffered data, but this
        would require a fairly extensive rewrite.

 Else if the command was a READ:

  If we got a SSD_KEY_BLANK_CHECK-
If we have a One Filemark EOT model- mark EOM as pending,
otherwise set EIO as the erorr.
  Else if we found a Filemark-
If we're in Fixed Block mode- mark EOF pending.

 If we had an ILI (Incorrect Length Indicator)-
  If the residual is less than zero, whine about tape record
  being too big for user's buffer, otherwise if we were in
  Fixed Block mode, mark EIO as pending.

All 'pending' conditions mean that the command in question completes
without error indication. It had succeeded, but a signifying event
occurred during its execution which will apply to the *next* command
that would be exexcuted. Except for the one EOM case above, we always
propagate residual.

Now, way back in sastart- if we notice any of the PENDING bits set,
we don't run the command we've just pulled off the wait queue. Instead,
we then figure out it's disposition based upon a previous command's
association with a signifying event.

 If SA_FLAG_EOM_PENDING is set, we don't set an error. We just complete
 the command with residual set to the request count (not data moved,
 but no error). We continue on.

 If SA_FLAG_EOF_PENDING- if we have this, it's only because we're in
 Fixed Block mode- in which case we traverse all waiting buffers (which
 we can get in fixed block mode because physio has split things up) and
 mark them all as no error, but no data moved and complete them.

 If SA_FLAG_EIO_PENDING, just mark the buffer with an EIO error
 and complete it.

Then we clear all of the pending state bits- we're done.

MFC after: 4 weeks

23 years agoRecognise VIA Apollo KT133A bridge.
Doug Rabson [Thu, 30 Aug 2001 14:17:48 +0000 (14:17 +0000)]
Recognise VIA Apollo KT133A bridge.

PR: 30061
Submitted by: John Merryweather Cooper <jmcoopr@webmail.bmi.net>
MFC after: 1 week

23 years agoRestore the `-perm +mode' feature.
Ruslan Ermilov [Thu, 30 Aug 2001 13:17:58 +0000 (13:17 +0000)]
Restore the `-perm +mode' feature.

Broken in the "close a PR" race, in revision 1.30.
Note that the patch in the PR did not have this bug!

23 years agoMake my Dad famous by showing the world my middle name, as seems
Josef Karthauser [Thu, 30 Aug 2001 13:17:26 +0000 (13:17 +0000)]
Make my Dad famous by showing the world my middle name, as seems
to be the trend.

23 years agoDoh! Fix a comma that disappeared along the way.
Scott Long [Thu, 30 Aug 2001 13:09:09 +0000 (13:09 +0000)]
Doh!  Fix a comma that disappeared along the way.

Pointy-hat by: peter

23 years agoDS10 has it's USB interface disabled (by Compaq/DEC). I wonder where I got
Wilko Bulte [Thu, 30 Aug 2001 12:46:53 +0000 (12:46 +0000)]
DS10 has it's USB interface disabled (by Compaq/DEC). I wonder where I got
the info that USB works in FreeBSD??

23 years agoFixed some typos, verboseness and misformatting in comments.
Bruce Evans [Thu, 30 Aug 2001 12:30:58 +0000 (12:30 +0000)]
Fixed some typos, verboseness and misformatting in comments.

23 years agoLike su(1), make PAM use mandatory. Remove parts of the authentication
Mark Murray [Thu, 30 Aug 2001 11:27:36 +0000 (11:27 +0000)]
Like su(1), make PAM use mandatory. Remove parts of the authentication
logic that are handled by PAM. Fix documentation to reflect this.

23 years agoFixed breakage of NOSHARED worlds in previous commit. -lreadline was
Bruce Evans [Thu, 30 Aug 2001 10:37:28 +0000 (10:37 +0000)]
Fixed breakage of NOSHARED worlds in previous commit.  -lreadline was
added but not its postrequisite -ltermcap.

Fixed breakage of DPADD in previous commit.  ${LIBREADLINE} was misspelled
-lreadline.  This should have been fatal since there is no file named
-lreadline, but it worked because of an undcumented bugfeature in make(1)
(or its configuration files): missing source files named -l* are silently
assumed to be up to date libraries.  `make checkdpadd' also fails to detect
this error.

23 years agoAdd support for sending ATAPI commands via ioctl.
Søren Schmidt [Thu, 30 Aug 2001 09:47:17 +0000 (09:47 +0000)]
Add support for sending ATAPI commands via ioctl.

23 years agoFix buffer overflow in queue file handling.
Kris Kennaway [Thu, 30 Aug 2001 09:26:54 +0000 (09:26 +0000)]
Fix buffer overflow in queue file handling.

Submitted by: millert@openbsd.org, gad
Reported by: X-Force <xforce@iss.net>

23 years agorestore(8) doesn't need to be setgid `tty', and never did.
Ruslan Ermilov [Thu, 30 Aug 2001 09:18:55 +0000 (09:18 +0000)]
restore(8) doesn't need to be setgid `tty', and never did.

At the times, restore(8) and rrestore(8) were the different
utilities.  rrestore(8) was installed setuid `root', while
restore(8) with usual ownership and privileges.  Later on,
on August 28, 1991 (what a coincidence!), rrestore(8) code
was merged with restore(8).  The setgid `tty' bit then was
accidentally put.

23 years agoAdd ACPI attachments.
Mike Smith [Thu, 30 Aug 2001 09:17:03 +0000 (09:17 +0000)]
Add ACPI attachments.

23 years agoCorrect usage of ISA_PNP_PROBE
Mike Smith [Thu, 30 Aug 2001 09:16:53 +0000 (09:16 +0000)]
Correct usage of ISA_PNP_PROBE

23 years agoSafety-check against empty PnP ID lists.
Mike Smith [Thu, 30 Aug 2001 09:14:28 +0000 (09:14 +0000)]
Safety-check against empty PnP ID lists.

23 years agoUnbreak group operations by reverting previous delta which removed the
Dima Dorfman [Thu, 30 Aug 2001 06:32:17 +0000 (06:32 +0000)]
Unbreak group operations by reverting previous delta which removed the
assignment of `l' in `gr_update' to the return value of snprintf.  It
claimed to have fixed the case where snprintf returned -1--in fact, it
broke the entire routine.  Not setting `l' here causes fileupdate() to
invariably fail with EINVAL because it does its own check to assert
that the parameter isn't -1.

23 years agoNew release notes: ntp suite updated to 4.1.0, daemon(8).
Dima Dorfman [Thu, 30 Aug 2001 05:34:37 +0000 (05:34 +0000)]
New release notes: ntp suite updated to 4.1.0, daemon(8).

Updated release notes: ACPI 20010816.

23 years agoAdd a new utility that runs an executable detached from the
Mike Barcroft [Thu, 30 Aug 2001 04:48:02 +0000 (04:48 +0000)]
Add a new utility that runs an executable detached from the
controlling terminal.

Discussed on: -hackers
Obtained from: BSD/OS

23 years agoRemove whitespace at end of line I happened to find during my last edit.
Alexander Langer [Thu, 30 Aug 2001 02:32:00 +0000 (02:32 +0000)]
Remove whitespace at end of line I happened to find during my last edit.

23 years agoAdd support for proper URI encoding, using strvisx(3)'s VIS_HTTPSTYLE.
Alexander Langer [Thu, 30 Aug 2001 02:30:33 +0000 (02:30 +0000)]
Add support for proper URI encoding, using strvisx(3)'s VIS_HTTPSTYLE.

Requests through a proxy are still broken for URIs that contain
blanks, since this required a bigger rewrite of the whole function.

23 years agoMinor improvements to arc4random():
Mike Silbersack [Thu, 30 Aug 2001 01:15:25 +0000 (01:15 +0000)]
Minor improvements to arc4random():

- Decrease reseeding interval from every 64 to every 16384 runs
  to reduce entropy usage.
- Add time based reseeding.  (Every 5 minutes.)
- Throw away the first 256 words of output as suggested in
  "Weaknesses in the Key Scheduling Algorithm of RC4."

Reviewed by: Mark Murray
MFC After: jkh says ok

23 years agoMention that the ACPI module load can be disabled by unsetting $acpi_load
Mike Smith [Thu, 30 Aug 2001 01:05:28 +0000 (01:05 +0000)]
Mention that the ACPI module load can be disabled by unsetting $acpi_load

23 years agoNote that compiling ACPI into the kernel is deprecated for normal use.
Mike Smith [Thu, 30 Aug 2001 00:58:57 +0000 (00:58 +0000)]
Note that compiling ACPI into the kernel is deprecated for normal use.

23 years agoAdd support for HTTP/1.1 name-based virtual hosts. Also, use
Mike Barcroft [Thu, 30 Aug 2001 00:57:35 +0000 (00:57 +0000)]
Add support for HTTP/1.1 name-based virtual hosts.  Also, use
asprintf(3) when creating the request string, as the length of
a path is defined as unlimited by the standard and limiting the
total request to 4K is awfully arbitrary.

PR: 30054
Submitted by: Joseph Mallett <jmallett@xMach.org>
MFC after: 8 days

23 years agoBuild the ACPI module by default.
Mike Smith [Thu, 30 Aug 2001 00:55:22 +0000 (00:55 +0000)]
Build the ACPI module by default.

23 years agoUpdate for slightly changed ACPI OSD module. Make debugging code conditional
Mike Smith [Thu, 30 Aug 2001 00:53:13 +0000 (00:53 +0000)]
Update for slightly changed ACPI OSD module.  Make debugging code conditional
on ACPI_DEBUG, so we can set it in /etc/make.conf.

23 years agoAdd support for attaching PnP-aware ISA drivers to ACPI.
Mike Smith [Thu, 30 Aug 2001 00:50:58 +0000 (00:50 +0000)]
Add support for attaching PnP-aware ISA drivers to ACPI.

Always parse ACPI device resource settings (current resources only)
and attach the resources to the device before probe/attaching.

23 years agoRetarget the resource parser slightly. We only fetch current resources
Mike Smith [Thu, 30 Aug 2001 00:49:34 +0000 (00:49 +0000)]
Retarget the resource parser slightly.  We only fetch current resources
for the device now (we should really just be parsing a passed-in resource
buffer).

Wrap long lines so this is (more) readable.

Support Address16 and Address32 resources, in the CONSUMER case.

Support DRQs so that we can handle ISA devices.

Support ExtendedIrqs (we ignore most of their attributes)

Add a placeholder device for system memory and system resources.  This
takes the place of the nexus placeholder, which only attaches to ISA.

23 years agoNote that now that some ISA devices will attach to ACPI, we need to
Mike Smith [Thu, 30 Aug 2001 00:45:42 +0000 (00:45 +0000)]
Note that now that some ISA devices will attach to ACPI, we need to
keep the ivar indexes that ISA uses free.

23 years agoDon't parse our resources ourself, the ACPI core code must do it.
Mike Smith [Thu, 30 Aug 2001 00:45:12 +0000 (00:45 +0000)]
Don't parse our resources ourself, the ACPI core code must do it.

23 years agoNuke the (fairly bogus) attachment of *all* ACPI devices to ISA.
Mike Smith [Thu, 30 Aug 2001 00:44:29 +0000 (00:44 +0000)]
Nuke the (fairly bogus) attachment of *all* ACPI devices to ISA.

23 years agoAdd missing acpi_disabled() call so that this driver can be disabled.
Mike Smith [Thu, 30 Aug 2001 00:44:01 +0000 (00:44 +0000)]
Add missing acpi_disabled() call so that this driver can be disabled.

23 years agoTeach the loader how to find the system ACPI information, and autoload
Mike Smith [Thu, 30 Aug 2001 00:42:12 +0000 (00:42 +0000)]
Teach the loader how to find the system ACPI information, and autoload
the ACPI module if the system apperars to be ACPI compliant.

This is an initial cut; the load should really be done by Forth support
code, and we should check both the BIOS build date and a blacklist.

23 years agoo Remove some GCCisms in src/powerpc/include/endian.h.
Mike Barcroft [Thu, 30 Aug 2001 00:04:19 +0000 (00:04 +0000)]
o Remove some GCCisms in src/powerpc/include/endian.h.
o Unify <machine/endian.h>'s across all architectures.
o Make bswapXX() functions use a different spelling of u_int16_t and
  friends to reduce namespace pollution.  The bswapXX() functions
  don't actually exist, but we'll probably import these at some
  point.  Atleast one driver (if_de) depends on bswapXX() for big
  endian cases.
o Deprecate byteorder(3) prototypes from <sys/types.h>, these are
  now prototyped indirectly in <arpa/inet.h>.
o Deprecate in_addr_t and in_port_t typedefs in <sys/types.h>, these
  are now typedef'd in <arpa/inet.h>.
o Change byteorder(3) prototypes to use standards compliant uint32_t
  (spelled __uint32_t to reduce namespace pollution).
o Document new preferred headers and standards compliance.

Discussed with: bde
PR: 29946
Reviewed by: bmilekic

23 years agowhen newreno is turned on, if dupacks = 1 or dupacks = 2 and
Jayanth Vijayaraghavan [Wed, 29 Aug 2001 23:54:13 +0000 (23:54 +0000)]
when newreno is turned on, if dupacks = 1 or dupacks = 2 and
new data is acknowledged, reset the dupacks to 0.
The problem was spotted when a connection had its send buffer full
because the congestion window was only 1 MSS and was not being incremented
because dupacks was not reset to 0.

Obtained from: Yahoo!

23 years agoo Fix some checks on snprintf(3) to prevent miscalculations.
Mike Barcroft [Wed, 29 Aug 2001 23:51:14 +0000 (23:51 +0000)]
o Fix some checks on snprintf(3) to prevent miscalculations.
o This fixes a memory leak that can occur on some URL's.

Pointy hat to: brian

23 years agoMega update to the aac driver.
Scott Long [Wed, 29 Aug 2001 23:34:05 +0000 (23:34 +0000)]
Mega update to the aac driver.
1.  Correctly handle commands initiated by the adapter.  These commands
    are defered to a kthread responsible for their processing, then are
    properly returned to the controller.
2.  Add/remove disk devices when notified by the card that a container was
    created/deleted/offline.
3.  Implement crashdump functionality.
4.  Support all ioctls needed for the management CLI to work.  The linux
    version of this app can be found at the Dell or HP website.  A native
    version will be forthcoming.

MFC-after: 4.4-RELEASE

23 years agoFix a bug in lseek which caused the loader to fail on some gzipped
John Polstra [Wed, 29 Aug 2001 23:33:22 +0000 (23:33 +0000)]
Fix a bug in lseek which caused the loader to fail on some gzipped
kernels.  The error message was "elf_loadexec: cannot seek".

Libstand maintains a read-ahead buffer for each open file, so that
it can read in chunks of 512 bytes for greater efficiency.  When
the loader tries to lseek forward in a file by a small amount, it
sometimes happens that the target file offset is already in the
read-ahead buffer.  But the lseek code simply discarded the contents
of that buffer and performed a seek directly on the underlying
file.  This resulted in an attempt to seek backwards in the file,
since some of the data has already been read into the read-ahead
buffer.  Gzipped data streams cannot seek backwards, so an error
was returned.

This commit adds code which checks to see if the desired file offset
is already in the read-ahead buffer.  If it is, the code simply
adjusts the buffer pointer and length, thereby avoiding a reverse
seek on the gzipped data stream.

I incorporated a suggestion from Matt Dillon which saved a little
bit of code in this fix.

Reviewed by: dillon, gallatin, jhb

23 years agoxref to ispfw(4) in see also
Wilko Bulte [Wed, 29 Aug 2001 21:13:26 +0000 (21:13 +0000)]
xref to ispfw(4) in see also

Approved by: mjacob
MFC after: 2 days

23 years agoFirst, The Ricoh 5C47x chips don't have the disable the 3e0 bits.
Warner Losh [Wed, 29 Aug 2001 20:33:08 +0000 (20:33 +0000)]
First, The Ricoh 5C47x chips don't have the disable the 3e0 bits.
Second, the TI 1130 need to have the PCI_INTR set, not cleared.

This gets Soren's machine working with NEWCARD again.

# The whole initialization is a mess and needs to be organized ala OLDCARD.

23 years agochown syntax is user:group now.
Alexander Langer [Wed, 29 Aug 2001 19:59:30 +0000 (19:59 +0000)]
chown syntax is user:group now.

23 years agoDocument defaults/periodic.conf.
Alexander Langer [Wed, 29 Aug 2001 19:45:27 +0000 (19:45 +0000)]
Document defaults/periodic.conf.

23 years agoRIP all ports options, as discussed on arch@.
Alexander Langer [Wed, 29 Aug 2001 19:38:17 +0000 (19:38 +0000)]
RIP all ports options, as discussed on arch@.

23 years agoFix linux_getcwd() so that if the cwd isn't cached (__getcwd() fails),
Andrew Gallatin [Wed, 29 Aug 2001 19:05:27 +0000 (19:05 +0000)]
Fix linux_getcwd() so that if the cwd isn't cached (__getcwd() fails),
the cwd is looked up inside the kernel. The native getcwd() in libc
handles this in userland if __getcwd() fails.

Obtained from: NetBSD via OpenBSD
Tested by: Chris Casey <chriss@phys.ksu.edu>, Markus Holmberg <markush@acc.umu.se>
Reviewed by: Darrell Anderson <anderson@cs.duke.edu>
PR: kern/24315

23 years agosmbfs_advlock: simplify overflow checks (copy from kern_lockf.c)
Andrey A. Chernov [Wed, 29 Aug 2001 18:59:04 +0000 (18:59 +0000)]
smbfs_advlock: simplify overflow checks (copy from kern_lockf.c)
minor formatting issues to minimize differences

23 years agoadvlock: simplify overflow checks
Andrey A. Chernov [Wed, 29 Aug 2001 18:53:53 +0000 (18:53 +0000)]
advlock: simplify overflow checks

23 years agolseek: simplify overflow checks
Andrey A. Chernov [Wed, 29 Aug 2001 18:35:53 +0000 (18:35 +0000)]
lseek: simplify overflow checks

23 years agoo Introduce support for _CAPABILITY_NEEDMACROS, reducing namespace
Robert Watson [Wed, 29 Aug 2001 18:01:41 +0000 (18:01 +0000)]
o Introduce support for _CAPABILITY_NEEDMACROS, reducing namespace
  pollution for non-POSIX.1e macros.
o Introduce CAP_UNITE(), CAP_NONZERO().
o Disable aspects of CAP_SETPCAP, which is Linux-specific (in particular,
  remove it from CAP_ALL_ON).
o Improve commenting.

Submitted by: tmm
Obtained from: TrustedBSD Project