]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agofree type too if we can't add the child.
John-Mark Gurney [Mon, 16 Jun 2003 19:18:06 +0000 (19:18 +0000)]
free type too if we can't add the child.

21 years agofix misspelling of ORIR_NOTFOUND
John-Mark Gurney [Mon, 16 Jun 2003 19:06:36 +0000 (19:06 +0000)]
fix misspelling of ORIR_NOTFOUND

21 years agoPartially back out rev 1.87 by nuking fifo_inactive() and moving the
Don Lewis [Mon, 16 Jun 2003 17:17:09 +0000 (17:17 +0000)]
Partially back out rev 1.87 by nuking fifo_inactive() and moving the
resource deallocation back to fifo_close().  This eliminates any
stale data that might be stuck in the socket buffers after all the
readers and writers have closed the fifo.

Tested by: Thorsten Schroeder <ths@katjusha.de>

21 years agoAlso remove the ones that snuck in at the bottom.
David E. O'Brien [Mon, 16 Jun 2003 16:12:05 +0000 (16:12 +0000)]
Also remove the ones that snuck in at the bottom.

21 years agoReference the new natmip(4) man page.
Hartmut Brandt [Mon, 16 Jun 2003 15:51:55 +0000 (15:51 +0000)]
Reference the new natmip(4) man page.

21 years agoRepo-copy of sys/pci/if_en_pci.c to the rest of the midway driver (sys/dev/en)
Hartmut Brandt [Mon, 16 Jun 2003 15:25:51 +0000 (15:25 +0000)]
Repo-copy of sys/pci/if_en_pci.c to the rest of the midway driver (sys/dev/en)
so that all of the driver sources are in one place. Adjust the configuration
files and the module build.

21 years agoMYADDR6 in ppp.link{up,down} should match even when IPCP is enabled.
Hajimu UMEMOTO [Mon, 16 Jun 2003 15:19:25 +0000 (15:19 +0000)]
MYADDR6 in ppp.link{up,down} should match even when IPCP is enabled.

MFC after: 1 week

21 years agoo BATTTIME and BATTPERCENT are 'int' not 'str'.
Maxim Konovalov [Mon, 16 Jun 2003 14:57:07 +0000 (14:57 +0000)]
o BATTTIME and BATTPERCENT are 'int' not 'str'.

PR: bin/48793
Submitted by: Stefan Farfeleder <stefan@fafoe.dyndns.org>
MFC after: 2 week

21 years agoFix the breakage introduced by rev. 1.43 of sys/dev/midway.c (don't commit
Hartmut Brandt [Mon, 16 Jun 2003 13:52:27 +0000 (13:52 +0000)]
Fix the breakage introduced by rev. 1.43 of sys/dev/midway.c (don't commit
on friday 13th and without making a universe). This adds struct and
constant definitions for ATM traffic parameters and re-enables the
build of the midway driver.

Tested by: make universe

21 years agoReset ncpus to 1 for bound thread group since there is only one
David Xu [Mon, 16 Jun 2003 13:14:52 +0000 (13:14 +0000)]
Reset ncpus to 1 for bound thread group since there is only one
thread in such group.
Change message text from kse_rel to kserel, it is better displayed
in top.

21 years agoIf ftpd is run with an -h option (hide host-specific info,)
Yaroslav Tykhiy [Mon, 16 Jun 2003 11:30:23 +0000 (11:30 +0000)]
If ftpd is run with an -h option (hide host-specific info,)
don't reveal the info in reply to the SYST command.

Get rid of using the "unix" macro at the same time.  It was a rather
poor way to check if the system was Unix since there were quite a
few Unix clones out there whose cc didn't define "unix" (e.g.,
NetBSD.)  It was also sensitive to the C standard used, which caused
unnecessary trouble:  With -std=c99, it should have been "__unix__",
and so on.

PR: bin/50690
Submitted by: Alex Semenyaka <alexs _at_ snark.ratmir.ru>
MFC after: 1 week

21 years agoo Pass a correct argument to printf(3).
Maxim Konovalov [Mon, 16 Jun 2003 09:44:53 +0000 (09:44 +0000)]
o Pass a correct argument to printf(3).

PR: bin/51750
Submitted by: Vasil Dimov <vd@datamax.bg>
MFC after: 2 weeks

21 years agoThis file was ignored by CVS in my last commit for some reason:
Poul-Henning Kamp [Mon, 16 Jun 2003 09:31:15 +0000 (09:31 +0000)]
This file was ignored by CVS in my last commit for some reason:

Remove pointless initialization of b_spc field, which now no longer
exists.

21 years agoo Correct usage example: ep(4) does not have 'link0' option.
Maxim Konovalov [Mon, 16 Jun 2003 09:11:29 +0000 (09:11 +0000)]
o Correct usage example: ep(4) does not have 'link0' option.
o Fix a typo.

PR: misc/51955
Submitted by: Xander <x+freebsd-gnats@surfnet.nl>
MFC after: 1 week

21 years agoBound check for broken Configuration ROM.
Hidetoshi Shimokawa [Mon, 16 Jun 2003 08:43:22 +0000 (08:43 +0000)]
Bound check for broken Configuration ROM.

21 years agoForce media autodetection if the device has lost its parameter table.
Yaroslav Tykhiy [Mon, 16 Jun 2003 08:42:20 +0000 (08:42 +0000)]
Force media autodetection if the device has lost its parameter table.
Previously, any normal I/O on an fdc(4) device would fail with ENXIO
if the device had been opened in non-blocking mode and then closed
prior to the conventional access; that would last until the floppy
disk was ejected and re-inserted to raise the unit attention condition.

Add a clarifying comment.

21 years agoFix for big endian.
Hidetoshi Shimokawa [Mon, 16 Jun 2003 08:29:24 +0000 (08:29 +0000)]
Fix for big endian.

21 years agoIf in non-blocking mode, return EAGAIN instead of ENXIO
Yaroslav Tykhiy [Mon, 16 Jun 2003 07:50:17 +0000 (07:50 +0000)]
If in non-blocking mode, return EAGAIN instead of ENXIO
on an I/O attempt.  This is needed for consistency with
the concept of the half-opened state of fdc(4).

PR: kern/52338

21 years agoReplace evil abuse of geteblk() with malloc(9).
Poul-Henning Kamp [Mon, 16 Jun 2003 07:41:47 +0000 (07:41 +0000)]
Replace evil abuse of geteblk() with malloc(9).

21 years agoGet rid of the b_spc specialty field in struct buf by using an already
Poul-Henning Kamp [Mon, 16 Jun 2003 07:18:39 +0000 (07:18 +0000)]
Get rid of the b_spc specialty field in struct buf by using an already
available caller private field.

21 years agoRevert previos change for prior release/branch.
Hidetoshi Shimokawa [Mon, 16 Jun 2003 03:44:36 +0000 (03:44 +0000)]
Revert previos change for prior release/branch.

21 years agoThis never should have been here.
Mike Heffner [Mon, 16 Jun 2003 03:06:32 +0000 (03:06 +0000)]
This never should have been here.

21 years agoUpdate for latest lukemftp import.
Mike Heffner [Mon, 16 Jun 2003 02:37:42 +0000 (02:37 +0000)]
Update for latest lukemftp import.

21 years agoThis commit was generated by cvs2svn to compensate for changes in r116424,
Mike Heffner [Mon, 16 Jun 2003 02:36:38 +0000 (02:36 +0000)]
This commit was generated by cvs2svn to compensate for changes in r116424,
which included commits to RCS files with non-trunk default branches.

21 years agoImport the June 15, 2003 CVS version of NetBSD's ftp + patchset to fix
Mike Heffner [Mon, 16 Jun 2003 02:36:38 +0000 (02:36 +0000)]
Import the June 15, 2003 CVS version of NetBSD's ftp + patchset to fix
header file differences between FBSD and NBSD. Like lukemftpd, the
portable version of lukemftp lags too far behind CVS.

This import includes several big fixes plus a cleaned up manpage.

21 years agoLock the vm object when freeing pages.
Alan Cox [Sun, 15 Jun 2003 21:50:38 +0000 (21:50 +0000)]
Lock the vm object when freeing pages.

21 years agoIn the device attach routine, don't depend on uaa->iface being
Bill Paul [Sun, 15 Jun 2003 21:45:43 +0000 (21:45 +0000)]
In the device attach routine, don't depend on uaa->iface being
populated. Apparently, if you use an ehci controller, it's not.
Use usbd_device2interface_handle() to retrieve the interface handle.
NOTE: uaa->iface is populated in the probe routine, so I suspect the
fact that it's NULL in the attach routine is a bug in the ehci driver.

Also, don't depend on the PHY addresses returned by the AXE_CMD_READ_PHYID
command. The address is correct for my LinkSys NIC, but a user has
reported that with a D-Link NIC, the PHYID command returns address 4
while the attached Broadcom PHY is in fact strapped for address 0.
Instead, latch onto the first PHY address that returns valid data
during a readreg operation.

21 years agoFix yppasswdproc_update_master_1_svc() too.
Martin Blapp [Sun, 15 Jun 2003 21:24:45 +0000 (21:24 +0000)]
Fix yppasswdproc_update_master_1_svc() too.

Only call pw_mkdb if passfile == _PATH_MASTERPASSWD.
Otherwise, rename master.passwd to a temp filename, rename
the new passwd to master.passwd, and let yppwupdate update
passwd as it sees fit.

Reviewed by: phk
Tested by: genesys

21 years agoThe page queue lock is already held in pmap_remove, change acquire/release
Jake Burkholder [Sun, 15 Jun 2003 21:06:49 +0000 (21:06 +0000)]
The page queue lock is already held in pmap_remove, change acquire/release
to assertion of ownership.  Serves me right for not booting a witness
kernel.

21 years agoRemove the evil BIOTOBUF macro, it is no longer used anywhere.
Poul-Henning Kamp [Sun, 15 Jun 2003 20:36:22 +0000 (20:36 +0000)]
Remove the evil BIOTOBUF macro, it is no longer used anywhere.

Retain b_bio as the first element of struct buf for now in case some code
somewhere still do the evil cast thing.

21 years agoIn specfs::vop_specstratey(), assert that the vnode and buffer agree about
Poul-Henning Kamp [Sun, 15 Jun 2003 20:31:04 +0000 (20:31 +0000)]
In specfs::vop_specstratey(), assert that the vnode and buffer agree about
the device.

21 years ago- Mirror vm_page_queue_mtx assertions added to the i386 pmap.
Jake Burkholder [Sun, 15 Jun 2003 19:54:50 +0000 (19:54 +0000)]
- Mirror vm_page_queue_mtx assertions added to the i386 pmap.
- Add vm page queue locking in certain places that are only needed on
  sparc64.

This should make pmap_qenter and pmap_qremove MP-safe.

Discussed with: alc

21 years agoI have not had any reports of trouble for a long time, so remove the
Poul-Henning Kamp [Sun, 15 Jun 2003 19:49:14 +0000 (19:49 +0000)]
I have not had any reports of trouble for a long time, so remove the
gentle versions of the vop_strategy()/vop_specstrategy() mismatch methods
and use vop_panic() instead.

21 years agoRemember to release the loader's heap.
Jake Burkholder [Sun, 15 Jun 2003 19:16:43 +0000 (19:16 +0000)]
Remember to release the loader's heap.

Reviewed by: tmm

21 years agoTake 2: Remove _both_ KASSERTS.
Poul-Henning Kamp [Sun, 15 Jun 2003 19:16:34 +0000 (19:16 +0000)]
Take 2:  Remove _both_ KASSERTS.

21 years agoDuh! I misread my handwritte notes: We do _not_ want to asser that
Poul-Henning Kamp [Sun, 15 Jun 2003 19:14:03 +0000 (19:14 +0000)]
Duh!  I misread my handwritte notes:  We do _not_ want to asser that
vp == bp->b_vp in specfs, that was the entire point of VOP_SPECSTRATEGY().

21 years agoAdd the same KASSERT to all VOP_STRATEGY and VOP_SPECSTRATEGY implementations
Poul-Henning Kamp [Sun, 15 Jun 2003 18:53:00 +0000 (18:53 +0000)]
Add the same KASSERT to all VOP_STRATEGY and VOP_SPECSTRATEGY implementations
to check that the buffer points to the correct vnode.

21 years agoAdd my timecounter paper from EuroBSDcon2002
Poul-Henning Kamp [Sun, 15 Jun 2003 18:49:46 +0000 (18:49 +0000)]
Add my timecounter paper from EuroBSDcon2002

21 years agoRemove in toto coda_strategy which incorrectly implemented vop_panic();
Poul-Henning Kamp [Sun, 15 Jun 2003 18:45:15 +0000 (18:45 +0000)]
Remove in toto coda_strategy which incorrectly implemented vop_panic();

21 years agoMerge entries from maho's USB device compatibility list.
Hiroki Sato [Sun, 15 Jun 2003 18:20:51 +0000 (18:20 +0000)]
Merge entries from maho's USB device compatibility list.

Obtained from: http://people.FreeBSD.org/~maho/USB/

21 years agoAlways set bio_resid properly in fdstrategy(),
Yaroslav Tykhiy [Sun, 15 Jun 2003 18:13:17 +0000 (18:13 +0000)]
Always set bio_resid properly in fdstrategy(),
as should every block device strategy routine.
There was at least one evil consequence of not doing so:
Some errors returned by fdstrategy() could be lost (EAGAIN,
in particular.)

PR: kern/52338 (in the audit-trail)
Discussed with: bde

21 years agoCheck whether the floppy type pointer has been set before trying
Yaroslav Tykhiy [Sun, 15 Jun 2003 16:18:58 +0000 (16:18 +0000)]
Check whether the floppy type pointer has been set before trying
to access floppy parameters through it.

Note: The DIOCGSECTORSIZE and DIOCGMEDIASIZE handlers withing
fdioctl() couldn't be just moved to below the existing check
for blocking mode because fd->ft can be non-NULL while still
in non-blocking mode (fd->ft can be set with the FD_STYPE ioctl.)

PR: kern/52338
No MFC: Not applicable to STABLE

21 years agoVarious cr*() calls believed to be MPSAFE, since the uidinfo
Robert Watson [Sun, 15 Jun 2003 15:57:42 +0000 (15:57 +0000)]
Various cr*() calls believed to be MPSAFE, since the uidinfo
code is locked down.

21 years agoName a function argument "mbuf", not "buf", if it is
Yaroslav Tykhiy [Sun, 15 Jun 2003 14:47:12 +0000 (14:47 +0000)]
Name a function argument "mbuf", not "buf", if it is
a pointer to struct mbuf for clarity and consistency.

21 years agoAdd missing descriptions of macros M_ALIGN and MH_ALIGN.
Yaroslav Tykhiy [Sun, 15 Jun 2003 14:40:03 +0000 (14:40 +0000)]
Add missing descriptions of macros M_ALIGN and MH_ALIGN.
Remove a reference to the defunct macro M_COPY_PKTHDR;
document the new functions m_dup_pkthdr() and m_move_pkthdr(),
and the macro variant of the latter, M_MOVE_PKTHDR().

21 years agoAdd more markup to the mbuf(9) manpage. This includes:
Yaroslav Tykhiy [Sun, 15 Jun 2003 14:14:11 +0000 (14:14 +0000)]
Add more markup to the mbuf(9) manpage.  This includes:

- tagging plaintext "mbuf", "mbuf cluster", and "mbuf chain"
  with .Vt (variable type) since all of them are ways of managing
  data, i.e., they can be seen as data types;

- using .Vt/.Va instead of .Li (literal) where appropriate;

- tagging plaintext words that actually refer to function arguments
  with .Fa.

Suggested by: ru

21 years agoIf lint(1) is not "lint", don't try to lint the xlint compat-libs. A
Mark Murray [Sun, 15 Jun 2003 13:03:37 +0000 (13:03 +0000)]
If lint(1) is not "lint", don't try to lint the xlint compat-libs. A
third-party lint is probably being used.

21 years ago1. Add code to support bound thread. when blocked, a bound thread never
David Xu [Sun, 15 Jun 2003 12:51:26 +0000 (12:51 +0000)]
1. Add code to support bound thread. when blocked, a bound thread never
   schedules an upcall. Signal delivering to a bound thread is same as
   non-threaded process. This is intended to be used by libpthread to
   implement PTHREAD_SCOPE_SYSTEM thread.
2. Simplify kse_release() a bit, remove sleep loop.

21 years agoUse .Va, not .Fa, to refer to structure members.
Yaroslav Tykhiy [Sun, 15 Jun 2003 12:42:41 +0000 (12:42 +0000)]
Use .Va, not .Fa, to refer to structure members.

mdoc(7) contains an ambiguous statement on the issue,
but our mdoc(7) police's opinion is solid.

Suggested by: ru

21 years agoIf the device goes away during ulpt_reset(), make sure not to call
Ian Dowse [Sun, 15 Jun 2003 11:55:50 +0000 (11:55 +0000)]
If the device goes away during ulpt_reset(), make sure not to call
ulpt_status() afterwards. This fixes a crash that can occur if a
USB printer is power-cycled when printing is just starting. The
problem is similar to that fixed in revision 1.33, but it is much
less likely to occur.

MFC after: 1 week

21 years agoDon't overwrite the static panicstr buffer for secondary and further
Ian Dowse [Sun, 15 Jun 2003 11:43:00 +0000 (11:43 +0000)]
Don't overwrite the static panicstr buffer for secondary and further
panics. Before revision 1.38, we used to just point panicstr at the
format string if panicstr was NULL, but since we now use a static
buffer for the formatted panic message, we have to be careful to
only write to it during the first panic.

Pointed out by: bde

21 years agoUse __builtin_alloca() on compilers that have it. Keep the prototype for
Dag-Erling Smørgrav [Sun, 15 Jun 2003 11:01:52 +0000 (11:01 +0000)]
Use __builtin_alloca() on compilers that have it.  Keep the prototype for
the benefit of lint and non-{GNU,Intel} compilers.

21 years agoRemove the definition of alloca(), it does not belong here.
Dag-Erling Smørgrav [Sun, 15 Jun 2003 10:59:44 +0000 (10:59 +0000)]
Remove the definition of alloca(), it does not belong here.

21 years agoKill prototype for __msgwrite too. It is not needed anymore.
Martin Blapp [Sun, 15 Jun 2003 10:55:39 +0000 (10:55 +0000)]
Kill prototype for __msgwrite too. It is not needed anymore.

21 years agoFix the master yppasswd routines, so they really work
Martin Blapp [Sun, 15 Jun 2003 10:37:22 +0000 (10:37 +0000)]
Fix the master yppasswd routines, so they really work
for root on ypmaster. yppasswd_local() did use YPPASSWDPROG
instead of MASTER_YPPASSWDPROG, and the domain was not set,
resulting in a coredump during xdr-encode.

Reviewed by: des

21 years agoFix the master yppasswd routines, so they really work
Martin Blapp [Sun, 15 Jun 2003 10:36:53 +0000 (10:36 +0000)]
Fix the master yppasswd routines, so they really work
for root on ypmaster. Fix ypclnt_havepasswdd() to work
for AF_LOCAL connections.

Reviewed by: des

21 years agoOnly call pw_mkdb if passfile == _PATH_MASTERPASSWD.
Martin Blapp [Sun, 15 Jun 2003 10:34:11 +0000 (10:34 +0000)]
Only call pw_mkdb if passfile == _PATH_MASTERPASSWD.
Otherwise, rename master.passwd to a temp filename, rename
the new passwd to master.passwd, and let yppwupdate update
passwd as it sees fit.

PR: 52601, 7968
Reviewed by: des
Submitted by: Dan Nelson <dnelson@allantgroup.com>

21 years agoReplace the old SCM_CREDS cred procedures. They can now be
Martin Blapp [Sun, 15 Jun 2003 10:32:01 +0000 (10:32 +0000)]
Replace the old SCM_CREDS cred procedures. They can now be
replaced just fine with getpeereid() and the whole code
gets a lot simpler. We don't break the ABI, since all server
programms use __rpc_get_local_uid(), and we just change library
internals.

Reviewed by: des

21 years agoerr() on allocation failure. WARNS=9 compliant
Philippe Charnier [Sun, 15 Jun 2003 09:28:17 +0000 (09:28 +0000)]
err() on allocation failure. WARNS=9 compliant
use #if 0, #ifndef lint, #endif /* not lint */, #endif ordering
when a message is provided, use errx() instead of err().

21 years agoDocument changes to default output.
Greg Lehey [Sun, 15 Jun 2003 08:46:36 +0000 (08:46 +0000)]
Document changes to default output.

21 years agoPrint out the current recording devices in the default printout.
Greg Lehey [Sun, 15 Jun 2003 08:46:04 +0000 (08:46 +0000)]
Print out the current recording devices in the default printout.

Discussed with: cg

Break some particularly long lines.

Still not enough for:  bde.

21 years agoRemove an unnecessary forward declaration.
Alan Cox [Sun, 15 Jun 2003 07:28:33 +0000 (07:28 +0000)]
Remove an unnecessary forward declaration.

21 years agoNow that the kernel access control for quotactl(2) appears to work
Robert Watson [Sun, 15 Jun 2003 06:54:36 +0000 (06:54 +0000)]
Now that the kernel access control for quotactl(2) appears to work
properly, clean up quota(1).  quota(1) has the ability to query
quotas either directly from the kernel, or if that fails, by reading
the quota.user or quota.group files specified for the file system
in /etc/fstab.  The setuid bit existed solely (apparently) to let
non-operator users query their quotas and consumption when quotas
weren't enabled for the file system.

o Remove the setuid bit from quota(1).

o Remove the logic used by quota(1) when running setuid to prevent
  users from querying the quotas of other users or groups.  Note
  that this papered over previously broken kernel access control;
  if you queried directly using the system call, you could access
  some of the data "restricted" by quota(1).

In the new world order, the ability to inspect the (live) quotas of
other uids and gids via the kernel is controlled by the privilege
requirement sysctl.  The ability to query via the file is controlled
by the file permissions on the quota database backing files
(root:operator, group readable by default).

21 years agoNow that the kernel access control for quotactl(2) appears to work
Robert Watson [Sun, 15 Jun 2003 06:46:24 +0000 (06:46 +0000)]
Now that the kernel access control for quotactl(2) appears to work
properly, clean up quota(1).  quota(1) has the ability to query
quotas either directly from the kernel, or if that fails, by reading
the quota.user or quota.group files specified for the file system
in /etc/fstab.  The setuid bit existed solely (apparently) to let
non-operator users query their quotas and consumption when quotas
weren't enabled for the file system.

o Remove the setuid bit from quota(1).

o Remove the logic used by quota(1) when running setuid to prevent
  users from querying the quotas of other users or groups.  Note
  that this papered over previously broken kernel access control.

21 years agoRe-implement kernel access control for quotactl() as found in the
Robert Watson [Sun, 15 Jun 2003 06:36:19 +0000 (06:36 +0000)]
Re-implement kernel access control for quotactl() as found in the
UFS quota implementation.  Push some quite broken access control
logic out of ufs_quotactl() into the individual command
implementations in ufs_quota.c; fix that logic.  Pass in the thread
argument to any quotactl command that will need to perform access
control.

o quotaon() requires privilege (PRISON_ROOT).

o quotaoff() requires privilege (PRISON_ROOT).

o getquota() requires that:

    If the type is USRQUOTA, either the effective uid match the
    requested quota ID, that the unprivileged_get_quota flag be
    set, or that the thread be privileged (PRISON_ROOT).

    If the type is GRPQUOTA, require that either the thread be
    a member of the group represented by the requested quota ID,
    that the unprivileged_get_quota flag be set, or that the
    thread be privileged (PRISON_ROOT).

o setquota() requires privilege (PRISON_ROOT).

o setuse() requires privilege (PRISON_ROOT).

o qsync() requires no special privilege (consistent with what
  was present before, but probably not very useful).

Add a new sysctl, security.bsd.unprivileged_get_quota, which when
set to a non-zero value, will permit unprivileged users to query user
quotas with non-matching uids and gids.  Set this to 0 by default
to be mostly consistent with the previous behavior (the same for
USRQUOTA, but not for GRPQUOTA).

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoTighten up the string->integer conversion in sysctl(8):
Robert Watson [Sun, 15 Jun 2003 06:26:08 +0000 (06:26 +0000)]
Tighten up the string->integer conversion in sysctl(8):

(1) Reject zero-length strings for CTLTYPE_INT, _UINT, _LONG,
    _ULONG.  Do not silently convert to 0.

(2) When converting CTLTYPE_INT, _UINT, _LONG, and _ULONG, check the
    end pointer generated by strtol() and strtoul() rather than
    discarding it.  Reject the string if either none of the string
    was useful for conversion to an integer, or if there was
    trailing garbage.

I.e., we will not allow you to set a numeric sysctl to a value unless
we can completely convert the string argument to a numeric value.
I tripped over this when I put the following in /etc/sysctl.conf:

   kern.maxfiles="4000"

Ouch.

21 years agoDisable unneeded devices and options.
Yoshihiro Takahashi [Sun, 15 Jun 2003 04:31:52 +0000 (04:31 +0000)]
Disable unneeded devices and options.

21 years agoRemove CPU_ATHLON_SSE_HACK option.
Yoshihiro Takahashi [Sun, 15 Jun 2003 04:29:37 +0000 (04:29 +0000)]
Remove CPU_ATHLON_SSE_HACK option.

21 years agoInclude "../Makefile.inc"
Yoshihiro Takahashi [Sun, 15 Jun 2003 04:27:22 +0000 (04:27 +0000)]
Include "../Makefile.inc"

21 years agoMost power management devices don't exist on pc98.
Yoshihiro Takahashi [Sun, 15 Jun 2003 04:24:28 +0000 (04:24 +0000)]
Most power management devices don't exist on pc98.

21 years agoThe en module has been broken for the last 40 hours. Disconnect it
Warner Losh [Sun, 15 Jun 2003 04:15:29 +0000 (04:15 +0000)]
The en module has been broken for the last 40 hours.  Disconnect it
from the tree until it is fixed.  Since it is an atm driver, it isn't
commonly used so this will not negatively impact too many people.
harti can reconnect it when he resurfaces and corrects the en module
problems.  This should allow snapshots to start succeeding again.

Reported by: lots of people

21 years ago - Increase the ksegrp's cpu time history buffer to 250ms.
Jeff Roberson [Sun, 15 Jun 2003 04:14:25 +0000 (04:14 +0000)]
 - Increase the ksegrp's cpu time history buffer to 250ms.
 - Decrease the history buffer divisor to 2 so that we remember more of the
   old behavior.

21 years ago- Enable support for building Configuration ROM.
Hidetoshi Shimokawa [Sun, 15 Jun 2003 04:09:26 +0000 (04:09 +0000)]
- Enable support for building Configuration ROM.
- Improve probe message for S800 chips.

21 years agoP_THREADED was renamed to P_SA, follow up.
David Xu [Sun, 15 Jun 2003 03:42:04 +0000 (03:42 +0000)]
P_THREADED was renamed to P_SA, follow up.

21 years agoFix crom_add_simple_text().
Hidetoshi Shimokawa [Sun, 15 Jun 2003 03:22:30 +0000 (03:22 +0000)]
Fix crom_add_simple_text().
Bytewap Configuration ROM image.

21 years agoAdd CSRVAL_VENDOR_PRIVATE.
Hidetoshi Shimokawa [Sun, 15 Jun 2003 03:19:55 +0000 (03:19 +0000)]
Add CSRVAL_VENDOR_PRIVATE.

21 years ago1. Migrate TDF_UPCALLING from td_flags to td_pflags.
David Xu [Sun, 15 Jun 2003 03:18:58 +0000 (03:18 +0000)]
1. Migrate TDF_UPCALLING from td_flags to td_pflags.
2. Add a flag TDF_SA, it will be used to distinguish SA
   based thread from bound thread.

21 years agofix a couple typos in the comments
John-Mark Gurney [Sun, 15 Jun 2003 03:08:37 +0000 (03:08 +0000)]
fix a couple typos in the comments

From NetBSD's bootpd (in src/usr.sbin/bootp)

21 years agofix bootpd to use fd_set. For some reason on Sparc, using int with
John-Mark Gurney [Sun, 15 Jun 2003 03:05:14 +0000 (03:05 +0000)]
fix bootpd to use fd_set.  For some reason on Sparc, using int with
select is broken.

21 years agomake rarpd use in_addr_t instead of u_long since this is neccessary
John-Mark Gurney [Sun, 15 Jun 2003 03:00:22 +0000 (03:00 +0000)]
make rarpd use in_addr_t instead of u_long since this is neccessary
on 64bit platforms

ok'd by: silence

21 years ago - Cap the growth of sleep and run time in sched_exit_kse().
Jeff Roberson [Sun, 15 Jun 2003 02:52:29 +0000 (02:52 +0000)]
 - Cap the growth of sleep and run time in sched_exit_kse().

21 years agoFix some style problems, some of which are old, some new, and some
David Schultz [Sun, 15 Jun 2003 02:21:13 +0000 (02:21 +0000)]
Fix some style problems, some of which are old, some new, and some
inherited from UFS.

Requested by: bde, njl

21 years ago - Fix the maximum slice value. I accidentally checked in a value of '2'
Jeff Roberson [Sun, 15 Jun 2003 02:18:29 +0000 (02:18 +0000)]
 - Fix the maximum slice value.  I accidentally checked in a value of '2'
   which meant no process would run for longer than 20ms.
 - Slightly redo the interactivity scorer.  It follows the same algorithm but
   in a slightly more correct way.  Previously values above half were
   incorrect.
 - Lower the interactivity threshold to 20.  It seems that in testing non-
   interactive tasks are hardly ever near there and expensive interactive
   tasks can sometimes surpass it.  This area needs more testing.
 - Remove an unnecessary KTR.
 - Fix a case where an idle thread that had an elevated priority due to
   priority prop. would be placed back on the idle queue.
 - Delay setting NEEDRESCHED until userret() for threads that haad their
   priority elevated while in kernel.  This gives us the same context switch
   optimization as SCHED_4BSD.
 - Limit the child's slice to 1 in sched_fork_kse() so we detect its behavior
   more quickly.
 - Inhert some of the run/slp time from the child in sched_exit_ksegrp().
 - Redo some of the priority comparisons so they are more clear.
 - Throttle the frequency of sched_pctcpu_update() so that rounding errors
   do not make it invalid.

21 years agoMerge the following from the English version:
Hiroki Sato [Sun, 15 Jun 2003 02:10:37 +0000 (02:10 +0000)]
Merge the following from the English version:

1.158 -> 1.159 hardware/common/dev.sgml

21 years agocheck_drive: If the partition isn't a Vinum drive, release it again
Greg Lehey [Sun, 15 Jun 2003 01:42:01 +0000 (01:42 +0000)]
check_drive: If the partition isn't a Vinum drive, release it again
and return NULL.

vinum_scandisk: Don't handle NULL device pointers.
        Only look at compatibility partition for i386.  This
        is a kludge which should go away once I have adequate
        documentation for the New World Order.

Together, these fixes remove occasional error messages about
non-existent drives.  They may also fix a number of problems that have
been reported without a PR.

PRs: None

21 years agoDon't set TDF_UPCALLING in cpu_set_upcall, it is not only used by
David Xu [Sun, 15 Jun 2003 00:54:07 +0000 (00:54 +0000)]
Don't set TDF_UPCALLING in cpu_set_upcall, it is not only used by
SA process.

21 years agoRename P_THREADED to P_SA. P_SA means a process is using scheduler
David Xu [Sun, 15 Jun 2003 00:31:24 +0000 (00:31 +0000)]
Rename P_THREADED to P_SA. P_SA means a process is using scheduler
activations.

21 years agoWhite space fix.
Hidetoshi Shimokawa [Sun, 15 Jun 2003 00:17:38 +0000 (00:17 +0000)]
White space fix.

21 years agoUse #ifdef __alpha__, not __alpha.
Alan Cox [Sun, 15 Jun 2003 00:12:42 +0000 (00:12 +0000)]
Use #ifdef __alpha__, not __alpha.

21 years agoIf someone tries to mount a union filesystem with another unionfs as
David Schultz [Sat, 14 Jun 2003 23:56:27 +0000 (23:56 +0000)]
If someone tries to mount a union filesystem with another unionfs as
the upper layer, fail gracefully instead of panicing.

MFC after: 3 days

21 years agoIntroduce malloc types M_UNDCACHE and M_UNPATH for important
David Schultz [Sat, 14 Jun 2003 23:48:20 +0000 (23:48 +0000)]
Introduce malloc types M_UNDCACHE and M_UNPATH for important
unionfs-related data structures to aid in debugging memory leaks.
Use NULL and NULLVP instead of 0 as appropriate.

MFC after: 3 days

21 years agoFactor out the process of freeing ``directory caches'', which unionfs
David Schultz [Sat, 14 Jun 2003 23:27:29 +0000 (23:27 +0000)]
Factor out the process of freeing ``directory caches'', which unionfs
directory vnodes use to refer to their constituent vnodes, into
union_dircache_free().  Also s/union_dircache/union_dircache_get/ and
tweak the structure of union_dircache_r().

MFC after: 3 days

21 years agoMigrate the thread stack management functions from the machine-dependent
Alan Cox [Sat, 14 Jun 2003 23:23:55 +0000 (23:23 +0000)]
Migrate the thread stack management functions from the machine-dependent
to the machine-independent parts of the VM.  At the same time, this
introduces vm object locking for the non-i386 platforms.

Two details:

1. KSTACK_GUARD has been removed in favor of KSTACK_GUARD_PAGES.  The
different machine-dependent implementations used various combinations
of KSTACK_GUARD and KSTACK_GUARD_PAGES.  To disable guard page, set
KSTACK_GUARD_PAGES to 0.

2. Remove the (unnecessary) clearing of PG_ZERO in vm_thread_new.  In
5.x, (but not 4.x,) PG_ZERO can only be set if VM_ALLOC_ZERO is passed
to vm_page_alloc() or vm_page_grab().

21 years agoCatch up man page with reality in rc.d/named.
Mike Makonnen [Sat, 14 Jun 2003 22:43:40 +0000 (22:43 +0000)]
Catch up man page with reality in rc.d/named.

Pointed out by: Simon L. Nielsen <simon@nitro.dk>

21 years agoMerge common XPT_CALC_GEOMETRY functions into a single convenience function.
Nate Lawson [Sat, 14 Jun 2003 22:40:08 +0000 (22:40 +0000)]
Merge common XPT_CALC_GEOMETRY functions into a single convenience function.
Devices below may experience a change in geometry.

* umass(4) drives exactly 1 GB in size now no longer use extended geometry.

21 years agoUpdate a comment about symlinking named's pid file to correctly
Mike Makonnen [Sat, 14 Jun 2003 22:26:30 +0000 (22:26 +0000)]
Update a comment about symlinking named's pid file to correctly
reflect the code.

Suggested by: maxim

21 years agoMerge common XPT_CALC_GEOMETRY functions into a single convenience function.
Nate Lawson [Sat, 14 Jun 2003 22:17:41 +0000 (22:17 +0000)]
Merge common XPT_CALC_GEOMETRY functions into a single convenience function.
Devices below may experience a change in geometry.

* Due to a bug, aic(4) never used extended geometry.  Changes all drives
  >1G to now use extended translation.
* sbp(4) drives exactly 1 GB in size now no longer use extended geometry.
* umass(4) drives exactly 1 GB in size now no longer use extended geometry.

For all other controllers in this commit, this should be a no-op.

Looked over by: scottl

21 years agoMerge common XPT_CALC_GEOMETRY functions into a single convenience function.
Nate Lawson [Sat, 14 Jun 2003 22:17:38 +0000 (22:17 +0000)]
Merge common XPT_CALC_GEOMETRY functions into a single convenience function.
Devices below may experience a change in geometry.

* Due to a bug, aic(4) never used extended geometry.  Changes all drives
  >1G to now use extended translation.
* sbp(4) drives exactly 1 GB in size now no longer use extended geometry.
* umass(4) drives exactly 1 GB in size now no longer use extended geometry.

For all other controllers in this commit, this should be a no-op.

PR:
Submitted by:
Looked over by: scottl
Approved by:
Obtained from:
MFC after:

21 years agoFix the KMOD for the lpbb device.
John-Mark Gurney [Sat, 14 Jun 2003 20:43:33 +0000 (20:43 +0000)]
Fix the KMOD for the lpbb device.

21 years agoamdpm was omitted from the build of i2c controllers. Fixed.
Nicolas Souchu [Sat, 14 Jun 2003 20:32:46 +0000 (20:32 +0000)]
amdpm was omitted from the build of i2c controllers. Fixed.