]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoThis commit was generated by cvs2svn to compensate for changes in r80588,
obrien [Mon, 30 Jul 2001 03:09:46 +0000 (03:09 +0000)]
This commit was generated by cvs2svn to compensate for changes in r80588,
which included commits to RCS files with non-trunk default branches.

22 years agoVirgin import of Christos Zoulas's FILE 3.36.
obrien [Mon, 30 Jul 2001 03:09:46 +0000 (03:09 +0000)]
Virgin import of Christos Zoulas's FILE 3.36.

22 years agoRedo how we manage SCSI device settings- have a 3rd flags (nvram) that records
mjacob [Mon, 30 Jul 2001 01:00:21 +0000 (01:00 +0000)]
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records
either what's in NVRAM or what the safe defaults would be if we lack NVRAM.
Then we rename cur_XXXX to actv_XXXX (these are the currently active settings)
and the dev_XXX settings to goal_XXXX (these are the settings which we want
cur_XXXX to converge to).

This probably isn't entirely final as yet- but it's a lot closer to now
being what it should be, including allowing camcontrol to actually set
specific settings.

22 years agoRedo how we manage SCSI device settings- have a 3rd flags (nvram) that records
mjacob [Mon, 30 Jul 2001 00:59:32 +0000 (00:59 +0000)]
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records
either what's in NVRAM or what the safe defaults would be if we lack NVRAM.
Then we rename cur_XXXX to actv_XXXX (these are the currently active settings)
and the dev_XXX settings to goal_XXXX (these are the settings which we want
cur_XXXX to converge to).

Roll core minor.

22 years agoRedo how we manage SCSI device settings- have a 3rd flags (nvram) that records
mjacob [Mon, 30 Jul 2001 00:59:06 +0000 (00:59 +0000)]
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records
either what's in NVRAM or what the safe defaults would be if we lack NVRAM.
Then we rename cur_XXXX to actv_XXXX (these are the currently active settings)
and the dev_XXX settings to goal_XXXX (these are the settings which we want
cur_XXXX to converge to).

22 years agoKill the command (don't rerun it) if we had an AUTOSENSE failure.
mjacob [Mon, 30 Jul 2001 00:44:32 +0000 (00:44 +0000)]
Kill the command (don't rerun it) if we had an AUTOSENSE failure.
If we had an AUTOSENSE failure, we don't know what SENSE DATA
we had for a CHECK CONDITION. It's far better to assume failure
in this case.

22 years ago(Forced commit- last one missed the comments)
mjacob [Mon, 30 Jul 2001 00:30:58 +0000 (00:30 +0000)]
(Forced commit- last one missed the comments)

Handle both old and new TARGIOALLOCUNIT/TARGIOFREEUNIT cases- the new
one allows us to specify inquiry data we want to use.

Handle more of the CAM_DIS_DISCONNECT case.

22 years ago(Forced commit- last commit was inadvertant in that it missed the comments)
mjacob [Mon, 30 Jul 2001 00:27:54 +0000 (00:27 +0000)]
(Forced commit- last commit was inadvertant in that it missed the comments)

Move TARGCTLIOALLOCUNIT to OTARGCTLIOALLOCUNIT, TARGCTLIOFREEUNIT
to OTARGCTLIOFREEUNIT and redefine old associated structure to be
old_ioc_alloc_unit- deprecation but preservation of binaries.

Add new structure for same- but this one contains a pointer to
user defined INQUIRY data so you can define what the target
device looks like to the outside world.

22 years agobackout last commit- inadvertant
mjacob [Mon, 30 Jul 2001 00:22:57 +0000 (00:22 +0000)]
backout last commit- inadvertant

22 years agoscsi_targetio.h
mjacob [Mon, 30 Jul 2001 00:21:29 +0000 (00:21 +0000)]
scsi_targetio.h

22 years agoPropagate CAM_DIS_DISCONNECT on through:
mjacob [Mon, 30 Jul 2001 00:19:50 +0000 (00:19 +0000)]
Propagate CAM_DIS_DISCONNECT on through:

1. If we get frozen, unfreeze for disable disconnects.
2. Put CAM_DIS_DISCONNECT commands at the head of the work queue
(we have a target still connected and we can't run anything else
until this command completes).

If we had an error sending the last CTIO, unfreeze the queue anyway.

22 years agoWhen booted -v (eg bootverbose is non-zero), have pccard report what
imp [Mon, 30 Jul 2001 00:03:58 +0000 (00:03 +0000)]
When booted -v (eg bootverbose is non-zero), have pccard report what
resources it is attempting to assign to a child object.  This should
help people track down mysterious resource allocation problems more
easily.

# Unfortunately, it is harder to do the conflict check and report which
# resource failed if the driver itself doesn't.

22 years agoTwo recent commits in sys/ufs/ufs interacted badly with ext2fs
iedowse [Sun, 29 Jul 2001 22:26:01 +0000 (22:26 +0000)]
Two recent commits in sys/ufs/ufs interacted badly with ext2fs
because it shares ufs code. In ufs_fhtovp(), the test on i_effnlink
is invalid because ext2fs does not maintain this field. In ufs_close(),
i_effnlink is also tested, to determines whether or not to call
vn_start_write(). The ufs_fhtovp issue breaks NFS exporting of
ext2fs filesystems; I believe the other is harmless.

Fix both cases by checking um_i_effnlink_valid in the ufsmount
struct, and use i_nlink if necessary.

Noticed by: bde
Reviewed by: mckusick, bde

22 years agoCorrect the old length argument passed to sysctlbyname to be a pointer
tmm [Sun, 29 Jul 2001 22:01:55 +0000 (22:01 +0000)]
Correct the old length argument passed to sysctlbyname to be a pointer
to a size_t (not to an int).

MFC after: 2 days

22 years agoCorrect the old length argument passed to sysctl to be a pointer to a
tmm [Sun, 29 Jul 2001 21:56:14 +0000 (21:56 +0000)]
Correct the old length argument passed to sysctl to be a pointer to a
size_t (not to an int).

MFC after: 2 days

22 years agoDisallow ATAPI CD transfers that are not a multiple of the device block
tmm [Sun, 29 Jul 2001 21:01:13 +0000 (21:01 +0000)]
Disallow ATAPI CD transfers that are not a multiple of the device block
size (previously, the transfer size would be rounded up to a multiple of
the block size, which would overflow the buffer).
This fixes panics when doing things like trying to mount audio CD's.

PR: kern/21946
Review Timeout: sos

22 years ago(Re)Add an SSH module for PAM, heavily based on Andrew Korty's module
markm [Sun, 29 Jul 2001 18:31:09 +0000 (18:31 +0000)]
(Re)Add an SSH module for PAM, heavily based on Andrew Korty's module
from ports.

22 years agoAvoid any chance of being misunderstood as having libelled developers
sheldonh [Sun, 29 Jul 2001 15:08:14 +0000 (15:08 +0000)]
Avoid any chance of being misunderstood as having libelled developers
or developers' vendors without compromising the importance of warning
against bad practice.

Reported by: mjacob
MFC after: 1 week

22 years agoMove SHLIB_MAJOR to below LIB and add a comment about why NO_WARNS is set.
dd [Sun, 29 Jul 2001 13:22:41 +0000 (13:22 +0000)]
Move SHLIB_MAJOR to below LIB and add a comment about why NO_WARNS is set.

22 years agoInstall the man page and add mp.h to INCS.
dd [Sun, 29 Jul 2001 13:19:24 +0000 (13:19 +0000)]
Install the man page and add mp.h to INCS.

22 years agoAdd a manual page for the libmp interface. It isn't real great, but
dd [Sun, 29 Jul 2001 13:19:17 +0000 (13:19 +0000)]
Add a manual page for the libmp interface.  It isn't real great, but
then again neither is the interface it's documenting.

22 years agoConvert the olpt driver to using new-bus stuff.
nyan [Sun, 29 Jul 2001 11:11:45 +0000 (11:11 +0000)]
Convert the olpt driver to using new-bus stuff.

22 years agoMerged from sys/dev/syscons/syscons.c revisions 1.367 and 1.368.
nyan [Sun, 29 Jul 2001 10:52:07 +0000 (10:52 +0000)]
Merged from sys/dev/syscons/syscons.c revisions 1.367 and 1.368.

22 years agoDon't xref mt(1) just because it mentions ioctl.
dd [Sun, 29 Jul 2001 09:17:54 +0000 (09:17 +0000)]
Don't xref mt(1) just because it mentions ioctl.

Submitted by: Philipp Mergenthaler <philipp.mergenthaler@stud.uni-karlsruhe.de>

22 years agoDon't capitalize variable names.
dd [Sun, 29 Jul 2001 09:17:16 +0000 (09:17 +0000)]
Don't capitalize variable names.

22 years agoioctl(2) can return EFAULT from copyin.
dd [Sun, 29 Jul 2001 09:16:07 +0000 (09:16 +0000)]
ioctl(2) can return EFAULT from copyin.

PR: 29285
Submitted by: Philipp Mergenthaler <philipp.mergenthaler@stud.uni-karlsruhe.de>

22 years agoEnable the new libmp in the build, and disable libgmp and its
dd [Sun, 29 Jul 2001 08:58:22 +0000 (08:58 +0000)]
Enable the new libmp in the build, and disable libgmp and its
henchmen.

22 years agoThis is the traditional BSD libmp interface implemented in terms of
dd [Sun, 29 Jul 2001 08:49:15 +0000 (08:49 +0000)]
This is the traditional BSD libmp interface implemented in terms of
the OpenSSL BIGNUM interface.  It is provided for compatibility only
and should not be used in new code.

22 years agoMake symlink $S/$M/include -> compile/FOO/machine at kernel-depend
kato [Sun, 29 Jul 2001 07:39:14 +0000 (07:39 +0000)]
Make symlink $S/$M/include -> compile/FOO/machine at kernel-depend
instead of geneassym.o.

22 years agoHook up the manpage.
obrien [Sun, 29 Jul 2001 07:21:39 +0000 (07:21 +0000)]
Hook up the manpage.

22 years agoRename the GLOB_MAXPATH flag of glob(3) to GLOB_LIMIT to be compatible
mikeh [Sun, 29 Jul 2001 00:52:37 +0000 (00:52 +0000)]
Rename the GLOB_MAXPATH flag of glob(3) to GLOB_LIMIT to be compatible
with NetBSD and OpenBSD. glob(3) will now return GLOB_NOSPACE with
errno set to 0 instead of GLOB_LIMIT when we match more than `gl_matchc'
patterns. GLOB_MAXPATH has been left as an alias of GLOB_LIMIT to
maintain backwards compatibility.

Reviewed by: sheldonh, assar
Obtained from: NetBSD/OpenBSD

22 years agoProperly handle wgetch(3) returning ERR. This prevents an abnormal exit
mp [Sat, 28 Jul 2001 22:40:10 +0000 (22:40 +0000)]
Properly handle wgetch(3) returning ERR. This prevents an abnormal exit
when a windows resize event (SIGWINCH) occurs.

Reported by: John Doe <rhamming2001@yahoo.com> and others on -stable.
Reviewed by: dd
MFC after: 1 week

22 years agoInclude value of command line argument that causes an error message or
se [Sat, 28 Jul 2001 21:34:56 +0000 (21:34 +0000)]
Include value of command line argument that causes an error message or
warning in said message, since fetch may be run from a makefile or script
which does not print the command line.

Approved by: des

22 years agoThe fix for schemeless and hostless URLs (rev. 1.27) broke the schemeless
se [Sat, 28 Jul 2001 21:28:14 +0000 (21:28 +0000)]
The fix for schemeless and hostless URLs (rev. 1.27) broke the schemeless
proxy specification, which seems to be valid according to the man page.

Change the logic to consider "hostname:port" a hostname and port instead
of a file URL.

Approved by: des

22 years agoAdd KOI8-U tables
ache [Sat, 28 Jul 2001 20:50:21 +0000 (20:50 +0000)]
Add KOI8-U tables

Submitted by: Olexander Kunytsa <kunia@istc.kiev.ua>

22 years agoDiff-reduce this with GENERIC.
markm [Sat, 28 Jul 2001 20:40:49 +0000 (20:40 +0000)]
Diff-reduce this with GENERIC.

OK'ed by: imp

22 years agoPermit direct swapping to NFS regular files using swapon(2). We
iedowse [Sat, 28 Jul 2001 20:18:38 +0000 (20:18 +0000)]
Permit direct swapping to NFS regular files using swapon(2). We
already allow this for NFS swap configured via BOOTP, so it is
known to work fine.

For many diskless configurations is is more flexible to have the
client set up swapping itself; it can recreate a sparse swap file
to save on server space for example, and it works with a non-NFS
root filesystem such as an in-kernel filesystem image.

22 years agoBuild standard directory for kerberos 5 (Heimdal) database.
markm [Sat, 28 Jul 2001 20:00:54 +0000 (20:00 +0000)]
Build standard directory for kerberos 5 (Heimdal) database.

22 years agoUpgraded launchpad for kerberos. Noe kerberos IV OR kerberos 5
markm [Sat, 28 Jul 2001 19:57:57 +0000 (19:57 +0000)]
Upgraded launchpad for kerberos. Noe kerberos IV OR kerberos 5
may be started at boot for kerberos servers.

22 years agoFix the environment handling:
markm [Sat, 28 Jul 2001 19:53:10 +0000 (19:53 +0000)]
Fix the environment handling:

However, there's still a bug in login.c
because you copy the environment *before* the call to pam_open_session,
which won't set the necessary variables set by /usr/ports/security/pam_ssh.

Submitted by: Volker Stolz <stolz@hyperion.informatik.rwth-aachen.de>

22 years agoRemove a bootverbose diagnostic that makes bootverbose just too verbose.
gibbs [Sat, 28 Jul 2001 18:45:52 +0000 (18:45 +0000)]
Remove a bootverbose diagnostic that makes bootverbose just too verbose.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r80508,
obrien [Sat, 28 Jul 2001 18:44:42 +0000 (18:44 +0000)]
This commit was generated by cvs2svn to compensate for changes in r80508,
which included commits to RCS files with non-trunk default branches.

22 years agoImport of Binutils from the FSF 2.11 branch [checked out 19-July-2001].
obrien [Sat, 28 Jul 2001 18:44:42 +0000 (18:44 +0000)]
Import of Binutils from the FSF 2.11 branch [checked out 19-July-2001].

22 years agoNote MFC of IPFilter 3.4.20.
dd [Sat, 28 Jul 2001 14:37:32 +0000 (14:37 +0000)]
Note MFC of IPFilter 3.4.20.

22 years agoMerged from sys/i86/isa/npx.c revision 1.107.
kato [Sat, 28 Jul 2001 13:08:31 +0000 (13:08 +0000)]
Merged from sys/i86/isa/npx.c revision 1.107.

22 years agoFinally add the manpage for the IEEE 802.1Q VLAN
yar [Sat, 28 Jul 2001 12:27:06 +0000 (12:27 +0000)]
Finally add the manpage for the IEEE 802.1Q VLAN
network interface.

22 years agoNote MFC of tcpdump 3.6.3.
dd [Sat, 28 Jul 2001 12:24:18 +0000 (12:24 +0000)]
Note MFC of tcpdump 3.6.3.

I didn't do this when I merged the delta to RELENG_4 because I thought
&merged; didn't apply to contributed software since there is one entry
per application which gets updated with the new version number, as
opposed to all the other programs, which get one entry per update.
However, the previous commit removed &merged; from the IPFilter entry,
so perhaps I just didn't look long enough when I did the tcpdump
merge.

22 years agoIPFilter updated to 3.4.20.
dd [Sat, 28 Jul 2001 12:21:19 +0000 (12:21 +0000)]
IPFilter updated to 3.4.20.

22 years agofix conflicts created by import
darrenr [Sat, 28 Jul 2001 12:08:15 +0000 (12:08 +0000)]
fix conflicts created by import

22 years agoThis commit was generated by cvs2svn to compensate for changes in r80486,
darrenr [Sat, 28 Jul 2001 11:59:33 +0000 (11:59 +0000)]
This commit was generated by cvs2svn to compensate for changes in r80486,
which included commits to RCS files with non-trunk default branches.

22 years agoImport IPFilter version 3.4.20
darrenr [Sat, 28 Jul 2001 11:59:33 +0000 (11:59 +0000)]
Import IPFilter version 3.4.20

22 years agoThis commit was generated by cvs2svn to compensate for changes in r80484,
darrenr [Sat, 28 Jul 2001 11:59:32 +0000 (11:59 +0000)]
This commit was generated by cvs2svn to compensate for changes in r80484,
which included commits to RCS files with non-trunk default branches.

22 years agoImport IPFilter version 3.4.20
darrenr [Sat, 28 Jul 2001 11:59:32 +0000 (11:59 +0000)]
Import IPFilter version 3.4.20

22 years agoChange permissions back to 4554/554
brian [Sat, 28 Jul 2001 11:58:41 +0000 (11:58 +0000)]
Change permissions back to 4554/554

Suggested by: kris

22 years agofix conflicts created by import
darrenr [Sat, 28 Jul 2001 11:58:26 +0000 (11:58 +0000)]
fix conflicts created by import

22 years agoIf the peer sends a REQ without the IPADDR option, only reject it
brian [Sat, 28 Jul 2001 11:32:08 +0000 (11:32 +0000)]
If the peer sends a REQ without the IPADDR option, only reject it
once.  If they repeat the request (again without the IPADDR option)
ACK it.

I've had reports that some ppp implementations will not assign
themselves an IP number.  This should negotiate with such things.

MFC after: 3 days

22 years agoThe security officer requested this be backed out for discussion.
obrien [Sat, 28 Jul 2001 04:44:09 +0000 (04:44 +0000)]
The security officer requested this be backed out for discussion.

22 years ago#ifdef some 5.0 code with freebsd_version to reduce diffs with stable.
imp [Sat, 28 Jul 2001 04:25:11 +0000 (04:25 +0000)]
#ifdef some 5.0 code with freebsd_version to reduce diffs with stable.

22 years agoIt is spelled INTR_FAST in current and INTR_TYPE_FAST in stable, so try to
imp [Sat, 28 Jul 2001 04:08:25 +0000 (04:08 +0000)]
It is spelled INTR_FAST in current and INTR_TYPE_FAST in stable, so try to
make allowances.

22 years agoStable requires machine/clock.h to quiet warnings. It isn't
imp [Sat, 28 Jul 2001 04:04:05 +0000 (04:04 +0000)]
Stable requires machine/clock.h to quiet warnings.  It isn't
strictly necessary on current, but having it in here makes the diffs with
stable smaller and doesn't hurt anything except for phk's redundant include
finder.

22 years agoTo reduce diffs with -stable, ifdef selinfo.h or select.h include
imp [Sat, 28 Jul 2001 03:47:10 +0000 (03:47 +0000)]
To reduce diffs with -stable, ifdef selinfo.h or select.h include

22 years agoMinor whitespace nit
imp [Sat, 28 Jul 2001 03:46:35 +0000 (03:46 +0000)]
Minor whitespace nit

22 years agoUse spaces instead of hard tabs in the diagram.
imp [Sat, 28 Jul 2001 03:45:55 +0000 (03:45 +0000)]
Use spaces instead of hard tabs in the diagram.

22 years agoRemove s_strl*(). I am not sure what was thought they accomplished.
obrien [Sat, 28 Jul 2001 01:59:58 +0000 (01:59 +0000)]
Remove s_strl*().  I am not sure what was thought they accomplished.

When reading the code I had to stop, say "ok, what does *these*
modifications of strl*() do?  Pull out grep.  Oh, not in add/, maybe above
in ../lib/?  Yep.  So what do they do?  Comments above them are misleading,
guess I'll have to read the code.  Oh, they just test strl* against the
size and return the result of the test.  Now I can continue to read the
code I was.

The uses of s_strl*() then test that result and errx()'s.
Lets think about the "optimized" code I am removing:

In general the compiler pushes the three args to strl* onto the stack and calls
s_strl*.  s_strl* has to indirectly access 3 args from the stack.  Then push
them on the stack a 2nd time for the real strl* call.  s_strl* then pops the
return from strl* off the stack; or moves it from the register it was returned
in, to the register where tests can happen.  s_strl* then pops the three
arguments to strl*.  Perform the test, push the result of the test, or move it
from the result register to the return value register.  The caller to s_strl*
now has to either pop the return value of s_strl* or move it from the return
value register to the test register.  The caller then pops the three args to
s_strl* off the stack (the same args that s_strl* itself had to pop off after
the real call to strl*).  The s_strl* caller then performs a simular test to
what has already been done, and conditionally jumps.  By doing things this way, we've given the compiler optimizer less to work with.

Also, please don't forget the that call to s_strl* has possibly jumped to code
not in the cache due to being far away from the calling code, thus causing a
pipeline stall.

So where is the "optimization" from s_strl*?
It isn't code clarity.
It isn't code execution speed.  It isn't code size either.

22 years agoPacify users who get all bent out of shape when they see the "xl%d: command
wpaul [Fri, 27 Jul 2001 20:55:45 +0000 (20:55 +0000)]
Pacify users who get all bent out of shape when they see the "xl%d: command
never completed" message. The RX reset takes longer complete than it
used to, a lot longer in fact than xl_wait() is prepared to wait.
When we do the RX reset in xl_reset(), this cases xl_wait() to time out
and whine. We wait a little extra time now after the RX reset, which
should silence the warning.

Thanks to obrien for finally getting me a box with a NIC that
causes this problem for me to tinker with.

22 years agoRemember to zero out certain things that we malloc() and/or contigmalloc().
wpaul [Fri, 27 Jul 2001 19:38:56 +0000 (19:38 +0000)]
Remember to zero out certain things that we malloc() and/or contigmalloc().

22 years agoDisable the dirhash sanity check that panics if an unused directory
iedowse [Fri, 27 Jul 2001 18:45:41 +0000 (18:45 +0000)]
Disable the dirhash sanity check that panics if an unused directory
entry (d_ino == 0) is found in a position that is not the start of
a DIRBLKSIZ block.

While such entries cannot occur normally (ufs always extends the
previous entry to cover the free space instead), they do not cause
problems and fsck does not fix them, so panicking is bad.

22 years agoUncomment a return(ENXIO) that I commented out for debugging purposes.
wpaul [Fri, 27 Jul 2001 18:28:37 +0000 (18:28 +0000)]
Uncomment a return(ENXIO) that I commented out for debugging purposes.

22 years agoMake ancontrol WARNS=2 clean.
brooks [Fri, 27 Jul 2001 18:26:42 +0000 (18:26 +0000)]
Make ancontrol WARNS=2 clean.

22 years agoMove route(8) entries out of Contributed Software section; they were
bmah [Fri, 27 Jul 2001 17:20:33 +0000 (17:20 +0000)]
Move route(8) entries out of Contributed Software section; they were
hanging out with a routed(8) entry.

Submitted by: dd

22 years agoTake -Wconversion out of BDECFLAGS. It is not particularly useful for
peter [Fri, 27 Jul 2001 16:23:53 +0000 (16:23 +0000)]
Take -Wconversion out of BDECFLAGS.  It is not particularly useful for
us anyway because it doesn't work right on the x86 and alpha.  On
K&R code, small ints would be promoted to int.  ANSI-C doesn't require
this and the small ints can be passed taking 8 or 16 bits of stack
space.  However, the x86 abi that we use *does* promote to 32 bit,
and the alpha ABI passes them in 64 bit registers so we dont have
that aspect of the problem here.  Losing float precision by having it
cast down to int because the funtion prototype specifies int is the
least of our problems.  -Wmissing-prototypes helps here anyway.

22 years agoIntroduce two new tunables from the boot loader.
imp [Fri, 27 Jul 2001 16:07:02 +0000 (16:07 +0000)]
Introduce two new tunables from the boot loader.

hw.pcic.irq Globally set the IRQ for all pcic devices' management
interrupt (aka card status change or CSC interrupt)
This is what used to be known as
machdep.pccard.pcic_irq (which has been retained for
now for compatibility).
hw.pcic.ignore_fuction_1 Ignores function 1 for all PCIC bridges by not
attaching to them.  Lucent released a huge batch
of cards that were imporperly manufactuered (lacking
the 0 ohm resister to disable slot 1).  This is
a big hammer to keep those cards from causing problems
(I've had 4 people contact me saying my patches
worked great once they added a kludge to always ignore
function 1, or until they soldered these resistors
in place!).

No clue where to document these.  They act as both boot loader environment
variables, as well as read-only sysctls after boot.

At the same time, sort sys/systm.h in its proper order after sys/sysctl.h.

22 years agoCleanup duplication in Aironet headers
brooks [Fri, 27 Jul 2001 16:05:21 +0000 (16:05 +0000)]
Cleanup duplication in Aironet headers

PR: kern/29210
Submitted by: Doug Ambrisko <ambrisko@ambrisko.com>
MFC after: 3 days

22 years agoRevert previous accidental commit. FWIW, it was part of enabling
peter [Fri, 27 Jul 2001 15:57:17 +0000 (15:57 +0000)]
Revert previous accidental commit.  FWIW, it was part of enabling
VM caching of disks through mmap() and stopping syncing of open files
that had their last reference in the fs removed (ie: their unsync'ed
pages get discarded on close already, so I made it stop syncing too).

22 years agoFix cut/paste blunder. Serves me right for doing a last minute tweak
peter [Fri, 27 Jul 2001 15:52:49 +0000 (15:52 +0000)]
Fix cut/paste blunder.  Serves me right for doing a last minute tweak
to what I had for some time.

Submitted by: bde

22 years agoBe consistent about how we quote commands that could be entered by the
dd [Fri, 27 Jul 2001 12:25:55 +0000 (12:25 +0000)]
Be consistent about how we quote commands that could be entered by the
user.

22 years agoThe ACPI timer register corruption problem is resolved in the PIIX4
msmith [Fri, 27 Jul 2001 09:01:13 +0000 (09:01 +0000)]
The ACPI timer register corruption problem is resolved in the PIIX4
starting with the PIIX4M.  Restrict enabling the workaround to those
chips known to be buggy.

22 years agoMinor nits merged from my stable tree:
imp [Fri, 27 Jul 2001 07:47:35 +0000 (07:47 +0000)]
Minor nits merged from my stable tree:
o kill blank line that I introduced in cardinfo.h
o Delete unused variable wasinactive.
o return 0 from pccard_resume.
o Set the state and lastsate initially to be empty.
o move comment above code for interrupt dispatching.
o Powerstate interface is now available as of 430002, not 500000 (note that
  this change will be not 100% correct since the power state stuff didn't
  enter current until well after 500000, but it is good enough for the two
  branche we have going now).

22 years agoAdditional clarification.
imp [Fri, 27 Jul 2001 07:32:59 +0000 (07:32 +0000)]
Additional clarification.

22 years agoAttempt to fix and document interactions between suspend/resume and pccardc
imp [Fri, 27 Jul 2001 07:21:42 +0000 (07:21 +0000)]
Attempt to fix and document interactions between suspend/resume and pccardc
power x 0.

pccardc power x 0 used to disable the slot.  But a suspend/resume
would reactivate the pccard.  It no longer does that.  Now the
disabling of the slot is sticy until it is reset with power x 1 or the
card is ejected.  This seems closer to correct behavior to me.

o Process all card state changes the same using pccard_do_stat_change().
o Cleanup disabling the card so that we can preserve the state after
  the change.  Basically, don't set it to empty as often as we do.
o On suspend, the new state is "empty" and the laststate is "suspend"
o Document state machine with a diagram of states and edges.  The
  edges are labeld to tell the reader what event causes the external
  state changes.
o "machdep.pccard.pcic_resume_reset" may be obsolete now.  We always
  call the bridge driver's resume method on resume now.  Otherwise cards
  won't automatically show up.  If it needs to stay, I'll add it back.

22 years agoGive hopefully better diagnostics about the card types we reject.
imp [Fri, 27 Jul 2001 07:14:12 +0000 (07:14 +0000)]
Give hopefully better diagnostics about the card types we reject.

22 years agoImprove wording for pccard memory assignement.
imp [Fri, 27 Jul 2001 06:32:20 +0000 (06:32 +0000)]
Improve wording for pccard memory assignement.
Document intended correct behavior for pccardc power and how it interacts
with suspend/resume.

22 years agoRemove -traditional from CFLAGS and add unistd.h to header.h so this
brooks [Fri, 27 Jul 2001 04:40:04 +0000 (04:40 +0000)]
Remove -traditional from CFLAGS and add unistd.h to header.h so this
actually compiles.  Hopefully, this code didn't rely on some weird
side effect of -traditional.

22 years agoAdd a short man page describing the vfs_unmountall() function.
chris [Fri, 27 Jul 2001 03:54:59 +0000 (03:54 +0000)]
Add a short man page describing the vfs_unmountall() function.

MFC after: 1 week

22 years agoAdd cross-references for the new kldsym(2) man page.
chris [Fri, 27 Jul 2001 03:03:36 +0000 (03:03 +0000)]
Add cross-references for the new kldsym(2) man page.

22 years agoAdd a new kldsym(2) man page.
chris [Fri, 27 Jul 2001 02:56:16 +0000 (02:56 +0000)]
Add a new kldsym(2) man page.

22 years agoMake PMAP_SHPGPERPROC tunable. One shouldn't need to recompile a kernel
peter [Fri, 27 Jul 2001 01:08:59 +0000 (01:08 +0000)]
Make PMAP_SHPGPERPROC tunable.  One shouldn't need to recompile a kernel
for this, since it is easy to run into with large systems with lots of
shared mmap space.

Obtained from: yahoo

22 years agoThe kernel *does* now define setjmp/longjmp. Dont duplicate it here.
peter [Fri, 27 Jul 2001 00:44:36 +0000 (00:44 +0000)]
The kernel *does* now define setjmp/longjmp.  Dont duplicate it here.
In file included from ../../../dev/vinum/vinumhdr.h:77,
                 from ../../../dev/vinum/vinum.c:44:
../../../dev/vinum/vinumext.h:165: warning: redundant redeclaration of `setjmp' in same scope
../../../sys/systm.h:96: warning: previous declaration of `setjmp'
../../../dev/vinum/vinummemory.c:44: warning: redundant redeclaration of `longjmp' in same scope
../../../sys/systm.h:97: warning: previous declaration of `longjmp'

22 years agoFix a warning.
peter [Fri, 27 Jul 2001 00:04:39 +0000 (00:04 +0000)]
Fix a warning.

22 years agoPatch up some style(9) stuff in tcp_new_isn()
peter [Fri, 27 Jul 2001 00:03:49 +0000 (00:03 +0000)]
Patch up some style(9) stuff in tcp_new_isn()

22 years agos/OpemBSD/OpenBSD/
peter [Fri, 27 Jul 2001 00:01:48 +0000 (00:01 +0000)]
s/OpemBSD/OpenBSD/

22 years agoMASK_FPU_SW didn't do what it was expected to do.
peter [Thu, 26 Jul 2001 23:47:04 +0000 (23:47 +0000)]
MASK_FPU_SW didn't do what it was expected to do.

22 years agoPrevent chpass(1) from writing -1 in the UID or GID fields
mike [Thu, 26 Jul 2001 23:27:10 +0000 (23:27 +0000)]
Prevent chpass(1) from writing -1 in the UID or GID fields
in the password file by properly casting integers.

PR: 27757
Reviewed by: des
Approved by: des
MFC after: 4 days

22 years agoCheck the state of the slot when we resume. Set it to empty if we no
imp [Thu, 26 Jul 2001 23:23:36 +0000 (23:23 +0000)]
Check the state of the slot when we resume.  Set it to empty if we no
longer have a pccard in the slot.  This fixes the problem where pccard
would say that a card had been inserted on resume.  This also appears
to make the insert/remove events more reliable after a resume as well,
but that may be a different bug I need to hunt down.

22 years agoUse the tunable maxusers rather than the compile-time one. Evaluate and
peter [Thu, 26 Jul 2001 23:08:31 +0000 (23:08 +0000)]
Use the tunable maxusers rather than the compile-time one.  Evaluate and
initialize in the right order to make derivative settings work right.
eg: at compile time, nmbufs was double nmbclusters.  For POLA this should
work the same at runtime.

22 years agoCall the early tunable setup functions as soon as kern_envp is available.
peter [Thu, 26 Jul 2001 23:06:44 +0000 (23:06 +0000)]
Call the early tunable setup functions as soon as kern_envp is available.
Some things depend on hz being set not long after this.

22 years agoMove MAXUSERS out of opt_param.h to make sure that other code doesn't
peter [Thu, 26 Jul 2001 23:05:35 +0000 (23:05 +0000)]
Move MAXUSERS out of opt_param.h to make sure that other code doesn't
use it rather than the tunable version.

22 years agoparam.c no longer gets special treatment. Use the source tree version.
peter [Thu, 26 Jul 2001 23:04:46 +0000 (23:04 +0000)]
param.c no longer gets special treatment.  Use the source tree version.

22 years agoMove param.c out of the conf directory and make it fully dynamic.
peter [Thu, 26 Jul 2001 23:04:03 +0000 (23:04 +0000)]
Move param.c out of the conf directory and make it fully dynamic.
Tunables are now derived at boot time from maxusers.  ie: change maxusers
via a tunable and all the derivative settings change.  You can change
the other tunables individually as well.  Even hz etc is tunable.

22 years agoStrongly recommend that ispfw(4)'s f/w be used with this driver if available.
mjacob [Thu, 26 Jul 2001 22:43:40 +0000 (22:43 +0000)]
Strongly recommend that ispfw(4)'s f/w be used with this driver if available.

MFC after: 1 week