]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoEnable fine grained locking within IPFilter, using mtx(9) and sx(9) allowing
darrenr [Fri, 24 Dec 2004 09:14:26 +0000 (09:14 +0000)]
Enable fine grained locking within IPFilter, using mtx(9) and sx(9) allowing
the the "needs giant" flag to be removed from the driver.

19 years ago``NULL is a specific instance of a null pointer constant; the generic is
keramida [Thu, 23 Dec 2004 23:45:25 +0000 (23:45 +0000)]
``NULL is a specific instance of a null pointer constant; the generic is
a "null pointer".''

Making good use of the excellent explanations sent to me by Ruslan
Ermilov, Garrett Wollman and Bruce Evans, correct the descriptions of
null pointers.  They are just "null pointers", not nil, not NULL or
".Dv NULL".

Suggested by: ru, wollman, bde
Reviewed by: ru, wollman
Pointy hat: keramida

19 years agoFix creation on "dlciX" hook in ng_sample.
rik [Thu, 23 Dec 2004 22:03:32 +0000 (22:03 +0000)]
Fix creation on "dlciX" hook in ng_sample.

Approved by: julian
MFC after: 3 days

19 years agoTurn NFS directio off until the stability issues are resolved.
ps [Thu, 23 Dec 2004 21:30:30 +0000 (21:30 +0000)]
Turn NFS directio off until the stability issues are resolved.

19 years agoUpdate disk->d_genid field when increasing sc->sc_genid.
pjd [Thu, 23 Dec 2004 21:15:15 +0000 (21:15 +0000)]
Update disk->d_genid field when increasing sc->sc_genid.

19 years ago- Give the timer, thermal, and error LVT entries an interrupt vector even
jhb [Thu, 23 Dec 2004 20:42:53 +0000 (20:42 +0000)]
- Give the timer, thermal, and error LVT entries an interrupt vector even
  though these aren't used yet.
- Add missing function prototypes for some static functions.
- Allow lvt_mode() to handle an LVT entry with a delivery mode of fixed.
- Consolidate code duplicated in lapic_init() and lapic_setup() to program
  the spurious vector register of a local APIC in a static lapic_enable()
  function.
- Dump the timer, thermal, error, and performance counter LVT entries
  during lapic_dump().
- Program LVT pins (currently only LINT0 and LINT1) after the local
  APIC has been software enabled via lapic_enable() since otherwise the
  LVT programming will not be able to unmask LVT sources.

19 years agoSome small style fixes.
jhb [Thu, 23 Dec 2004 20:35:51 +0000 (20:35 +0000)]
Some small style fixes.

19 years agoAdd some constants for the local APIC timer.
jhb [Thu, 23 Dec 2004 20:35:07 +0000 (20:35 +0000)]
Add some constants for the local APIC timer.

19 years agoAdd a simple 'intrcnt_add' function that other MD code can use to add a
jhb [Thu, 23 Dec 2004 20:34:18 +0000 (20:34 +0000)]
Add a simple 'intrcnt_add' function that other MD code can use to add a
single named counter to the interrupt counts without having to fake up an
entire interrupt source.

19 years agoModify pmap_enter_quick() so that it expects the page queues to be locked
alc [Thu, 23 Dec 2004 20:16:11 +0000 (20:16 +0000)]
Modify pmap_enter_quick() so that it expects the page queues to be locked
on entry and it assumes the responsibility for releasing the page queues
lock if it must sleep.

Remove a bogus comment from pmap_enter_quick().

Using the first change, modify vm_map_pmap_enter() so that the page queues
lock is acquired and released once, rather than each time that a page
is mapped.

19 years ago- Add a function to set the Task Priority Register (TPR) of the local APIC.
jhb [Thu, 23 Dec 2004 19:47:59 +0000 (19:47 +0000)]
- Add a function to set the Task Priority Register (TPR) of the local APIC.
  Currently this is only used to initiailize the TPR to 0 during initial
  setup.
- Reallocate vectors for the local APIC timer, error, and thermal LVT
  entries.  The timer entry is allocated from the top of the I/O interrupt
  range reducing the number of vectors available for hardware interrupts
  to 191.  Linux happens to use the same exact vector for its timer
  interrupt as well.  If the timer vector shared the same priority queue
  as the IPI handlers, then the frequency that the timer vector will
  eventually be firing at can interact badly with the IPIs resulting in
  the queue filling and the dreaded IPI stuck panics, hence it being located
  at the top of the previous priority queue instead.
- Fixup various minor nits in comments.

19 years agoFix counting length of leadin for hook name (sizeof () => strlen ()).
rik [Thu, 23 Dec 2004 18:39:13 +0000 (18:39 +0000)]
Fix counting length of leadin for hook name (sizeof () => strlen ()).

Approved by:    julian
MFC after:      3 days

19 years agoWarn about the NOFOO -> NO_FOO conversion, and remind users to
ru [Thu, 23 Dec 2004 16:03:08 +0000 (16:03 +0000)]
Warn about the NOFOO -> NO_FOO conversion, and remind users to
read the COMMON ITEMS section of this file for proper upgrade
instructions.

19 years agoMove systm.h up, since it is required by mbuf.h.
glebius [Thu, 23 Dec 2004 13:09:37 +0000 (13:09 +0000)]
Move systm.h up, since it is required by mbuf.h.

Requested by: ru

19 years ago- sort includes
glebius [Thu, 23 Dec 2004 10:48:10 +0000 (10:48 +0000)]
- sort includes
- remove duplicate include sys/sysctl.h

19 years agoDon't call "objs" target in rescue.mk twice.
ru [Thu, 23 Dec 2004 10:16:46 +0000 (10:16 +0000)]
Don't call "objs" target in rescue.mk twice.

19 years agoMake "===> " prefixes look sane.
ru [Thu, 23 Dec 2004 10:13:17 +0000 (10:13 +0000)]
Make "===> " prefixes look sane.

19 years agoInclude bsd.compat.mk early from sys.mk, enough for makefiles
ru [Thu, 23 Dec 2004 08:51:50 +0000 (08:51 +0000)]
Include bsd.compat.mk early from sys.mk, enough for makefiles
using conditional statements to see the new spellings of NO_*
knobs (in case user still uses old spellings).

Reported by: kris

19 years agoDocument the 2130SLP
scottl [Thu, 23 Dec 2004 08:05:40 +0000 (08:05 +0000)]
Document the 2130SLP

19 years agoAdd OZ711M1
imp [Thu, 23 Dec 2004 05:32:02 +0000 (05:32 +0000)]
Add OZ711M1

Submitted by: Gordon Bergling

19 years agoAdd recognition of O2Micro 711M1.
imp [Thu, 23 Dec 2004 05:28:36 +0000 (05:28 +0000)]
Add recognition of O2Micro 711M1.

Submitted by: Gordon Bergling

19 years agoBetter rendering of the prototype for signal(3).
jkoshy [Thu, 23 Dec 2004 02:48:20 +0000 (02:48 +0000)]
Better rendering of the prototype for signal(3).

Suggested by: ru

19 years agoAssert the sem lock in sem_ref() and sem_rel(), as it is required to
rwatson [Thu, 23 Dec 2004 02:22:47 +0000 (02:22 +0000)]
Assert the sem lock in sem_ref() and sem_rel(), as it is required to
safely manipulate the reference count.

19 years agoAttempt to consistently use () around return values in calls to
rwatson [Thu, 23 Dec 2004 01:34:26 +0000 (01:34 +0000)]
Attempt to consistently use () around return values in calls to
return() in newer code (sysctl, ISN, timewait).

19 years agoRemove an XXXRW comment relating to whether or not the TCP timers are
rwatson [Thu, 23 Dec 2004 01:27:13 +0000 (01:27 +0000)]
Remove an XXXRW comment relating to whether or not the TCP timers are
MPSAFE: they are now believed to be.

Correct a typo in a second comment.

MFC after: 2 weeks

19 years agoRemove the now unused tcp_canceltimers() function. tcpcb timers are
rwatson [Thu, 23 Dec 2004 01:25:59 +0000 (01:25 +0000)]
Remove the now unused tcp_canceltimers() function.  tcpcb timers are
now stopped as part of tcp_discardcb().

MFC after: 2 weeks

19 years agoRemove an annotation of a minor race relating to the update of
rwatson [Thu, 23 Dec 2004 01:21:54 +0000 (01:21 +0000)]
Remove an annotation of a minor race relating to the update of
multiple MIB entries using sysctl in short order, which might
result in unexpected values for tcp_maxidle being generated by
tcp_slowtimo.  In practice, this will not happen, or at least,
doesn't require an explicit comment.

MFC after: 2 weeks

19 years agoRemove temporary debugging printf that was used to detect the presence
rwatson [Thu, 23 Dec 2004 01:19:27 +0000 (01:19 +0000)]
Remove temporary debugging printf that was used to detect the presence
of a race that had previously caused a panic in order to determine if
the fix was for the right problem.  It was.

MFC after: 2 weeks

19 years agoIn sonewconn(), the s/if/while/ change to wait for room at the tail of
rwatson [Thu, 23 Dec 2004 01:16:21 +0000 (01:16 +0000)]
In sonewconn(), the s/if/while/ change to wait for room at the tail of
the accept queue is a feature, not a bug/issue, so remove the XXXRW
from the comment.

19 years agoRemove an XXXRW indicating atomic operations might be used as a
rwatson [Thu, 23 Dec 2004 01:07:12 +0000 (01:07 +0000)]
Remove an XXXRW indicating atomic operations might be used as a
substitute for a global mutex protecting the socket count and
generation number.

The observation that soreceive_rcvoob() can't return an mbuf
chain is a property, not a bug, so remove the XXXRW.

In sorflush, s/existing/previous/ for code when describing prior
behavior.

For SO_LINGER socket option retrieval, remove an XXXRW about why
we hold the mutex: this is correct and not dubious.

MFC after: 2 weeks

19 years agoIn soalloc(), simplify the mac_init_socket() handling to remove
rwatson [Thu, 23 Dec 2004 00:59:43 +0000 (00:59 +0000)]
In soalloc(), simplify the mac_init_socket() handling to remove
unnecessary use of a global variable and simplify the return case.
While here, use ()'s around return values.

In sodealloc(), remove a comment about why we bump the gencnt and
decrement the socket count separately.  It doesn't add
substantially to the reading, and clutters the function.

MFC after: 2 weeks

19 years agoFix a couple of typos.
keramida [Thu, 23 Dec 2004 00:27:03 +0000 (00:27 +0000)]
Fix a couple of typos.

PR: docs/75410
Submitted by: "Paul A.Hoadley" <paulh@logicsquad.net>
MFC after: 3 days

19 years ago- Add genid field to the metadata which will allow to improve reliability a bit.
pjd [Wed, 22 Dec 2004 23:09:32 +0000 (23:09 +0000)]
- Add genid field to the metadata which will allow to improve reliability a bit.
  After this change, when component is disconnected because of an I/O error,
  it will not be connected and synchronized automatically, it will be logged
  as broken and skipped. Autosynchronization can occur, when component is
  disconnected (on orphan event) and connected again - there were no I/O
  error, so there is no need to not connected the component, but when there were
  writes while it wasn't connected, it will be synchronized.
  This fix cases, when component is disconnected because of I/O error and can be
  connected again and again.
- Bump version number.
- Add version change history.
- Implement backward compatibility mechanism. After this change when metadata in
  old version is detected, it is automatically upgraded to the new (current)
  version.

19 years agoFurther fix the case mentioned in rev. 1.302. The
ru [Wed, 22 Dec 2004 22:00:01 +0000 (22:00 +0000)]
Further fix the case mentioned in rev. 1.302.  The
intent was (and still is) that if a user has say
CPUTYPE=i686 set in /etc/make.conf, we don't print
the assignment type warning unless TARGET_CPUTYPE
is overridden.

Unfortunately, the implementation was buggy, and
only recent changes to bsd.cpu.mk that swapped
canonical and alias values of some CPU types made
the bug apparent.

Here's what happens here.

- CPUTYPE=i686 is set in /etc/make.conf,
- bsd.cpu.mk reset it to "pentiumpro",
- Makefile.inc1 compares this canonical value
  with the result of the following test,

make -f /dev/null CPUTYPE=pentiumpro -V CPUTYPE

and expects the result to be "pentiumpro" too,
but "i686" is returned, here's why.  We have two
CPUTYPE variables, global, set to "i686" in
/etc/make.conf, and command-line (of a higher
precedence), set to "pentiumpro".

The following part of bsd.cpu.mk,

.  elif ${CPUTYPE} == "i686"
CPUTYPE = pentiumpro

which is responsible for converting aliases to
canonical values, sees the value of the CPUTYPE
command-line variable first, "pentiumpro", and
no conversion is done -- the net effect is that
CPUTYPE global stays with its old value "i686",
and "make -V CPUTYPE" (which prints variables
in the global context) returns "i686".

The fix was to pass the CPUTYPE in the test above
as an environment variable instead of as a command
line variable, i.e.,

CPUTYPE=pentiumpro make -f /dev/null -V CPUTYPE

This time, CPUTYPE global is still set to "i686"
initially (by /etc/make.conf), and an envieronment
variable CPUTYPE (of a lower precedence) is set
to "pentiumpro".  The .elif sees it's set to
"i686" and resets it to "pentiumpro", and so
"make -V" returns "pentiumpro".

NB: these various types of make(1) variables can
be very painful, especially when combined with
"make -V".

19 years agoAdd send buffer locking to uipc_send(). Without this locking a race can
alc [Wed, 22 Dec 2004 20:28:46 +0000 (20:28 +0000)]
Add send buffer locking to uipc_send().  Without this locking a race can
occur between a reader and a writer that results in a panic upon close,
e.g.,
"panic: sbflush_locked: cc 4 || mb 0xffffff0052afa400 || mbcnt 0"

Reviewed by: rwatson@
MFC after: 2 weeks

19 years ago'forget' command takes device names, not provider names.
pjd [Wed, 22 Dec 2004 17:50:44 +0000 (17:50 +0000)]
'forget' command takes device names, not provider names.

19 years agoInclude fcntl.h
phk [Wed, 22 Dec 2004 17:39:21 +0000 (17:39 +0000)]
Include fcntl.h
Include selinfo.h (don't rely on vnode.h to do so)
Check O_NONBLOCK instead of IO_NELAY
Don't include vnode.h

19 years agoDon't include filedesc.h
phk [Wed, 22 Dec 2004 17:38:43 +0000 (17:38 +0000)]
Don't include filedesc.h
Include fcntl.h
Include selinfo.h (don't rely on vnode.h to do so)
Check O_NONBLOCK instead of IO_NDELAY
Don't include vnode.h

19 years agoInclude fcntl.h
phk [Wed, 22 Dec 2004 17:37:57 +0000 (17:37 +0000)]
Include fcntl.h
Check O_NONBLOCK instead of IO_NDELAY
Include uio.h
Don't include vnode.h
Don't include filedesc.h

19 years agoInclude uio.h
phk [Wed, 22 Dec 2004 17:37:14 +0000 (17:37 +0000)]
Include uio.h
Check O_NONBLOCK instead if IO_NDELAY
Don't include vnode.h

19 years agoInclude fcntl.h
phk [Wed, 22 Dec 2004 17:36:38 +0000 (17:36 +0000)]
Include fcntl.h
Check O_NONBLOCK instead of IO_NDELAY.
Include selinfo.h instead of relying on vnode.h to do so.
Don't include vnode.h

19 years agoInclude fcntl.h
phk [Wed, 22 Dec 2004 17:35:52 +0000 (17:35 +0000)]
Include fcntl.h
check O_NONBLOCK instead of IO_NDELAY
Don't include vnode.h.

19 years agoInclude fcntl.h
phk [Wed, 22 Dec 2004 17:34:53 +0000 (17:34 +0000)]
Include fcntl.h
check O_NONBLOCK instead of IO_NDELAY
don't include vnode.h

19 years agoInclude fcntl.h
phk [Wed, 22 Dec 2004 17:34:25 +0000 (17:34 +0000)]
Include fcntl.h
Check O_NONBLOCK instead of IO_NDELAY
Don't include vnode.h

19 years agoCheck O_NONBLOCK instead of IO_NDELAY.
phk [Wed, 22 Dec 2004 17:32:53 +0000 (17:32 +0000)]
Check O_NONBLOCK instead of IO_NDELAY.
Don't include <sys/vnode.h>

19 years agoFix comment.
phk [Wed, 22 Dec 2004 17:32:27 +0000 (17:32 +0000)]
Fix comment.

19 years agoDon't include vnode.h.
phk [Wed, 22 Dec 2004 17:31:44 +0000 (17:31 +0000)]
Don't include vnode.h.
Check O_NONBLOCK instead of IO_NDELAY

19 years agoCorrect speling erors.
wollman [Wed, 22 Dec 2004 17:31:28 +0000 (17:31 +0000)]
Correct speling erors.

19 years agoDon't include vnode.h
phk [Wed, 22 Dec 2004 17:31:10 +0000 (17:31 +0000)]
Don't include vnode.h

19 years agoInclude fcntl.h not vnode.h.
phk [Wed, 22 Dec 2004 17:30:38 +0000 (17:30 +0000)]
Include fcntl.h not vnode.h.
Include uio.h instead of relying on vnode.h to do so.
Check O_NONBLOCK not IO_NDELAY.

19 years agoCheck O_NONBLOCK not IO_NDELAY.
phk [Wed, 22 Dec 2004 17:29:37 +0000 (17:29 +0000)]
Check O_NONBLOCK not IO_NDELAY.

Don't include vnode.h

19 years ago#include fcntl.h not vnode.h. Check O_NONBLOCK not IO_NDELAY.
phk [Wed, 22 Dec 2004 17:29:02 +0000 (17:29 +0000)]
#include fcntl.h not vnode.h.  Check O_NONBLOCK not IO_NDELAY.

19 years ago#include of <sys/vnode.h> not necesary.
phk [Wed, 22 Dec 2004 17:28:34 +0000 (17:28 +0000)]
#include of <sys/vnode.h> not necesary.

19 years agoBe consistent about flag values passed to device drivers read/write
phk [Wed, 22 Dec 2004 17:05:44 +0000 (17:05 +0000)]
Be consistent about flag values passed to device drivers read/write
methods:

Read can see O_NONBLOCK and O_DIRECT.

Write can see O_NONBLOCK, O_DIRECT and O_FSYNC.

In addition O_DIRECT is shadowed as IO_DIRECT for now for backwards
compatibility.

19 years agoShuffle numeric values of the IO_* flags to match the O_* flags from
phk [Wed, 22 Dec 2004 16:25:50 +0000 (16:25 +0000)]
Shuffle numeric values of the IO_* flags to match the O_* flags from
fcntl.h.

This is in preparation for making the flags passed to device drivers be
consistently from fcntl.h for all entrypoints.

Today open, close and ioctl uses fcntl.h flags, while read and write
uses vnode.h flags.

19 years agoPunctuation marks should be separate arguments in groff macros.
keramida [Wed, 22 Dec 2004 16:20:12 +0000 (16:20 +0000)]
Punctuation marks should be separate arguments in groff macros.

Noticed by: ru

19 years agoUse .Dv NULL when referring to NULL C pointers, instead of "nil".
keramida [Wed, 22 Dec 2004 16:15:52 +0000 (16:15 +0000)]
Use .Dv NULL when referring to NULL C pointers, instead of "nil".

19 years agoLook into machine-specific manpage subdirectories too.
ru [Wed, 22 Dec 2004 16:04:58 +0000 (16:04 +0000)]
Look into machine-specific manpage subdirectories too.

PR: bin/72243
MFC after: 3 weeks

19 years agoCross reference init(8) too, instead of vaguely referring to the
keramida [Wed, 22 Dec 2004 15:44:21 +0000 (15:44 +0000)]
Cross reference init(8) too, instead of vaguely referring to the
"initialization process".

19 years agoFixed the only warning and mark as WARNS=6 clean.
ru [Wed, 22 Dec 2004 15:25:51 +0000 (15:25 +0000)]
Fixed the only warning and mark as WARNS=6 clean.

19 years ago- Fixed handling of manpage subdirectories:
ru [Wed, 22 Dec 2004 15:24:48 +0000 (15:24 +0000)]
- Fixed handling of manpage subdirectories:

  catman /usr/share/man/man8
  cd /usr/share/man; catman man8

- Don't print false warnings about invalid cat pages which are
  machine-specific cat page subdirectories (visible with -v).

- Fixed one memory leak.

19 years agoRemove single line containing the word "KLD" ommitted in ps.1:1.80.
rwatson [Wed, 22 Dec 2004 09:04:47 +0000 (09:04 +0000)]
Remove single line containing the word "KLD" ommitted in ps.1:1.80.

Pointed out by: ru (some time ago)

19 years agoDue to unknown reasons, Disk_Names() returns SCSI CDROM as a valid
yongari [Wed, 22 Dec 2004 08:26:48 +0000 (08:26 +0000)]
Due to unknown reasons, Disk_Names() returns SCSI CDROM as a valid
disk. This is main reason why sysinstall presents SCSI CDROM to
available disks in Fdisk/Label menu. In addition, adding a blank
SCSI CDROM to the menu generates floating point exception in sparc64.
Disk_Names() just extracts sysctl "kern.disks". Why GEOM treats SCSI
CDROM as a disk is beyond me and that should be investigated.
For temporary workaround, ignore SCSI CDROM device.

PR: sparc64/72962
Tested by: R. Tyler Ballance < tyler AT tamu DOT edu>
MFC after: 1 week

19 years agoPlug memory leak.
yongari [Wed, 22 Dec 2004 08:17:18 +0000 (08:17 +0000)]
Plug memory leak.

MFC after: 1 week

19 years agoInclude wchar.h to improve our chances of finding
kientzle [Wed, 22 Dec 2004 06:40:28 +0000 (06:40 +0000)]
Include wchar.h to improve our chances of finding
WCHAR_MAX.  This might fix a portability problem on HP_UX.

Thanks to: Susan Kim

19 years agoPortability nit: Some platforms require stdio.h before bzlib.h.
kientzle [Wed, 22 Dec 2004 06:30:14 +0000 (06:30 +0000)]
Portability nit: Some platforms require stdio.h before bzlib.h.

Thanks to: Kurt J. Lidl

19 years agoAdd __archive_strappend_int to format a decimal
kientzle [Wed, 22 Dec 2004 06:12:40 +0000 (06:12 +0000)]
Add __archive_strappend_int to format a decimal
number into a variable-length archive_string.

19 years agoAdd a fallback version of fnmatch() to the bsdtar
kientzle [Wed, 22 Dec 2004 06:08:04 +0000 (06:08 +0000)]
Add a fallback version of fnmatch() to the bsdtar
source code.  Include configure logic to pick up the
system one when it exists and use the fallback version
when it doesn't exist.  Set the default for FreeBSD
to use the system version.

With this, bsdtar should now be quite portable.

19 years agoGC an unused #include
kientzle [Wed, 22 Dec 2004 05:54:10 +0000 (05:54 +0000)]
GC an unused #include

19 years agoFix -W long option handling.
kientzle [Wed, 22 Dec 2004 03:38:28 +0000 (03:38 +0000)]
Fix -W long option handling.
In particular:
  -W excl=text
fails because "excl" is a prefix of both "exclude" and "exclude-from".  But,
  -W exclude=text
is okay because it matches "exclude" exactly.

Thanks to: Jose F Nieves
MFC after: 7 days

19 years agoDon't truncate major/minor numbers written to the legacy
kientzle [Wed, 22 Dec 2004 02:35:37 +0000 (02:35 +0000)]
Don't truncate major/minor numbers written to the legacy
ustar fields.  Later, we're going to permit numeric extensions
for these fields, so we can support large values here.  In particular,
this allows GNU tar to correctly extract such entries even
though it doesn't support the pax extended attributes.

Note: r1.18 and r1.17.2.1 of this file allowed similar treatment
of the uid/gid fields.

Thanks to: Ben Mesander

19 years agoTune the bidding for tar archives. This
kientzle [Wed, 22 Dec 2004 00:49:16 +0000 (00:49 +0000)]
Tune the bidding for tar archives.  This
improves the recognition of hardlink entries
with/without bodies (which is implemented through
a look-ahead that uses the bid function).

MFC after: 7 days

19 years agoPortability: the included fts.c should now
kientzle [Wed, 22 Dec 2004 00:21:28 +0000 (00:21 +0000)]
Portability: the included fts.c should now
compile on many non-FreeBSD systems, including
Linux, Solaris, Darwin, etc.

Thanks to: many people who pointed out the same portability problem

19 years agoAnd this, kids, is what happens if you act on a patch without looking
roam [Tue, 21 Dec 2004 20:02:50 +0000 (20:02 +0000)]
And this, kids, is what happens if you act on a patch without looking
at the mailing list to see if there isn't any more discussion just
around the corner.  Here's the update to the other part of the manpage
that deals with the hw.ata.atapi_dma loader tunable and initializing
ATAPI devices in PIO or DMA mode.

PR: 75021
Submitted by: Craig Rodrigues <rodrigc@crodrigues.org>
Spotted by: Joel Dahl <joel@automatvapen.se>

19 years agoNow, when force device destruction is done on shutdown, hide warning,
pjd [Tue, 21 Dec 2004 19:50:18 +0000 (19:50 +0000)]
Now, when force device destruction is done on shutdown, hide warning,
that device cannot be destroyed immediately, under debug=1.

Suggested by: simon

19 years agoReflect the change of the hw.ata.atapi_dma sysctl default in rev. 1.217 of
roam [Tue, 21 Dec 2004 19:41:21 +0000 (19:41 +0000)]
Reflect the change of the hw.ata.atapi_dma sysctl default in rev. 1.217 of
src/sys/dev/ata/ata-all.c from PIO to DMA.

Submitted by: Craig Rodrigues <rodrigc@crodrigues.org>
MFC after: 1 week

19 years agoImprove reliability and clean up code a bit.
pjd [Tue, 21 Dec 2004 19:30:59 +0000 (19:30 +0000)]
Improve reliability and clean up code a bit.
For more details check src/sys/geom/mirror/g_mirror.c rev.1.47,1.48,1.49,1.50.

19 years agoUse vtopde() instead of pmap_pde() in pmap_kextract(); vtopde() is smaller
alc [Tue, 21 Dec 2004 19:25:56 +0000 (19:25 +0000)]
Use vtopde() instead of pmap_pde() in pmap_kextract(); vtopde() is smaller
and faster in cases, such as pmap_kextract(), where the pde is known to
exist.

19 years agoWait a bit after creating device.
pjd [Tue, 21 Dec 2004 19:03:10 +0000 (19:03 +0000)]
Wait a bit after creating device.

19 years agoWhen RAID3 device is configured in "verify" mode, synchronization of the
pjd [Tue, 21 Dec 2004 18:48:59 +0000 (18:48 +0000)]
When RAID3 device is configured in "verify" mode, synchronization of the
parity component is now force on creation.
Change regression test which check "verify" mode to work properly after
this change.

19 years agoThis should not be permitted, but some GEOM classes held the topology lock
pjd [Tue, 21 Dec 2004 18:42:51 +0000 (18:42 +0000)]
This should not be permitted, but some GEOM classes held the topology lock
while doing g_(read|write)_data() (e.g. BSD). This can cause a deadlock
in MIRROR class. Not sure if this is safe to drop the topology lock in BSD
class, so change the code in MIRROR class to avoid this deadlock.

19 years agoImplement g_topology_try_lock().
pjd [Tue, 21 Dec 2004 18:32:46 +0000 (18:32 +0000)]
Implement g_topology_try_lock().

No objection from: phk

19 years agoUse the source of the termcap database when available.
ru [Tue, 21 Dec 2004 15:16:36 +0000 (15:16 +0000)]
Use the source of the termcap database when available.

19 years agoThe "fixate" command no longer terminates command processing,
ru [Tue, 21 Dec 2004 14:57:37 +0000 (14:57 +0000)]
The "fixate" command no longer terminates command processing,
so that "eject" following it can happen.

19 years agoDocument -F in usage() and SYNOPSIS.
ru [Tue, 21 Dec 2004 14:53:44 +0000 (14:53 +0000)]
Document -F in usage() and SYNOPSIS.

19 years agoMake "burncd ... fixate eject" really work as promised.
ru [Tue, 21 Dec 2004 14:49:10 +0000 (14:49 +0000)]
Make "burncd ... fixate eject" really work as promised.

19 years agoNOPAM -> NO_PAM
ru [Tue, 21 Dec 2004 12:49:24 +0000 (12:49 +0000)]
NOPAM -> NO_PAM

19 years agoNOIPSEC -> TRACEROUTE_NO_IPSEC
ru [Tue, 21 Dec 2004 12:44:20 +0000 (12:44 +0000)]
NOIPSEC -> TRACEROUTE_NO_IPSEC

19 years agoNOFORTH -> NO_FORTH
ru [Tue, 21 Dec 2004 12:32:15 +0000 (12:32 +0000)]
NOFORTH -> NO_FORTH

19 years agoNOCLEAN -> NO_CLEAN
ru [Tue, 21 Dec 2004 12:21:26 +0000 (12:21 +0000)]
NOCLEAN -> NO_CLEAN
NOCLEANDIR -> NO_CLEANDIR

19 years agoFix the build of netrate/ with WARNS=5 and add a Makefile.inc to
keramida [Tue, 21 Dec 2004 12:14:34 +0000 (12:14 +0000)]
Fix the build of netrate/ with WARNS=5 and add a Makefile.inc to
make it the default for all the netrate test tools.

Approved by: rwatson

19 years agoNOSHARE -> NO_SHARE
ru [Tue, 21 Dec 2004 12:13:23 +0000 (12:13 +0000)]
NOSHARE -> NO_SHARE

19 years agoNOHTML -> NO_HTML
ru [Tue, 21 Dec 2004 12:05:11 +0000 (12:05 +0000)]
NOHTML -> NO_HTML

19 years agoOverhaul ppp(8) build options so they are safe to use in
ru [Tue, 21 Dec 2004 12:01:15 +0000 (12:01 +0000)]
Overhaul ppp(8) build options so they are safe to use in
/etc/make.conf:

NOALIAS -> retired (support provided by PPP_NO_NAT)
NOATM -> PPP_NO_ATM (also subject to NO_ATM global)
NODES -> PPP_NO_DES (support was broken, now recovered)
NOI4B -> PPP_NO_I4B (also subject to NO_I4B global)
NOKLDLOAD -> PPP_NO_KLDLOAD
NONAT -> PPP_NO_NAT
NONETGRAPH -> PPP_NO_NETGRAPH
NOPAM -> PPP_NO_PAM (will be subject to NO_PAM global)
NORADIUS -> PPP_NO_RADIUS
NOSUID -> retired (support provided by PPP_NO_SUID)
PPP_NOSUID -> PPP_NO_SUID

19 years ago- The IQ80321 has a jumper which changes the PCI device id from 4 to 8 for the
cognet [Tue, 21 Dec 2004 11:36:09 +0000 (11:36 +0000)]
- The IQ80321 has a jumper which changes the PCI device id from 4 to 8 for the
GigE controller, so handle this.
- Use the outbound window 0 if the PCI mem requested is in its range, instead
of inconditionally use the outbound window 1.
This should be enough to get FreeBSD/arm to work on the IQ80321 board as well.

Reported and tested by: Jia-Shiun Li <jiashiun at gmail dot com>

19 years agoThe C define is NONAT.
ru [Tue, 21 Dec 2004 11:12:05 +0000 (11:12 +0000)]
The C define is NONAT.

19 years agoNOINET6 -> NO_INET6
ru [Tue, 21 Dec 2004 10:49:29 +0000 (10:49 +0000)]
NOINET6 -> NO_INET6

19 years agoNOGAMES -> NO_GAMES
ru [Tue, 21 Dec 2004 10:36:54 +0000 (10:36 +0000)]
NOGAMES -> NO_GAMES

19 years agoNOCRYPT -> NO_CRYPT
ru [Tue, 21 Dec 2004 10:16:04 +0000 (10:16 +0000)]
NOCRYPT -> NO_CRYPT

19 years agoNOSHARED -> NO_SHARED
ru [Tue, 21 Dec 2004 09:59:45 +0000 (09:59 +0000)]
NOSHARED -> NO_SHARED