]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years ago - Assert that every page found in the active queue is an active page.
alc [Wed, 22 Oct 2003 03:08:24 +0000 (03:08 +0000)]
 - Assert that every page found in the active queue is an active page.

20 years agoReflect the fact that fxtv is now found in ports/multimedia/fxtv.
simon [Tue, 21 Oct 2003 23:36:46 +0000 (23:36 +0000)]
Reflect the fact that fxtv is now found in ports/multimedia/fxtv.

PR: docs/58355
Submitted by: Simon Barner <barner@in.tum.de>

20 years agoTurn off ISP_SMPLOCK- not to be turned on again.
mjacob [Tue, 21 Oct 2003 21:52:23 +0000 (21:52 +0000)]
Turn off ISP_SMPLOCK- not to be turned on again.

Until we can have perfect knowledge that all callers above us think it's okay
for us to sleep, releasing *our* locks of course, we don't dare try and sleep.

20 years agostop use of NI_WITHSCOPEID. it was deprecated.
ume [Tue, 21 Oct 2003 20:11:47 +0000 (20:11 +0000)]
stop use of NI_WITHSCOPEID.  it was deprecated.

Obtained from: KAME

20 years ago- change scope to zone.
ume [Tue, 21 Oct 2003 20:05:32 +0000 (20:05 +0000)]
- change scope to zone.
- change node-local to interface-local.
- better error handling of address-to-scope mapping.
- use in6_clearscope().

Obtained from: KAME

20 years agoAdd myself.
sergei [Tue, 21 Oct 2003 19:56:23 +0000 (19:56 +0000)]
Add myself.

Approved by: krion (mentor)

20 years ago - Assert that the containing vm object is locked in
alc [Tue, 21 Oct 2003 19:36:51 +0000 (19:36 +0000)]
 - Assert that the containing vm object is locked in
   vm_page_set_validclean().  (This function reads and modifies the
   vm page's valid field, which is synchronized by the lock on the
   containing vm object.)

20 years agoMerge the following from the English version:
hrs [Tue, 21 Oct 2003 19:29:10 +0000 (19:29 +0000)]
Merge the following from the English version:

1.183 -> 1.184 hardware/common/dev.sgml
1.640 -> 1.644 relnotes/common/new.sgml

20 years agoProperly unload the DMA SG list on errors.
sos [Tue, 21 Oct 2003 19:25:20 +0000 (19:25 +0000)]
Properly unload the DMA SG list on errors.

20 years agoFix the DMA problem that most severely hit on the DS3112a SATA chip
sos [Tue, 21 Oct 2003 19:20:37 +0000 (19:20 +0000)]
Fix the DMA problem that most severely hit on the DS3112a SATA chip
in connection with Marvell based SATA->PATA dongles.

The problem was caused by a combination of things working
together to make it hard to spot...

The ATA driver has always started the ATA command, then build
the SG list for DMA and then finally started the DMA engine.
While this is according to specs, it poses a potential
problem as some controllers apparently do not allow for unlimitted
time between starting the ATA command and starting the DMA engine.

At about the same time as ATAng was committed there were lots
of other changes applied, some of which was locking in parts
that causes the busdma load functions to take significantly
longer to load the SG list.

This pushed the time spent between starting the ATA command and
starting the DMA engine over the hill for some controllers
(especially the Silicon Image DS3112a) and caused what looked
like lost interrupts.

The solution is to get all the SG list work or rather all
busdma related stuff done before we even try to start anything.

This has the nice side effect of seperating busdma out the
way it should be, so the working of the ATA machinery is not
cluttered up with busdma droppings, making the code easier
to read and understand.

20 years agoModernize examples/usage. Kill 1200/300 baud modem entries and
imp [Tue, 21 Oct 2003 19:15:29 +0000 (19:15 +0000)]
Modernize examples/usage.  Kill 1200/300 baud modem entries and
instead use 14.4kbps and faster modems as examples.  Separate line
speed and baud rate and be careful when talking of one verses the
other.

20 years agoWrap a long line in the previous commit
silby [Tue, 21 Oct 2003 18:48:49 +0000 (18:48 +0000)]
Wrap a long line in the previous commit

Suggested by: njl

20 years agoChange all SYSCTLS which are readonly and have a related TUNABLE
silby [Tue, 21 Oct 2003 18:28:36 +0000 (18:28 +0000)]
Change all SYSCTLS which are readonly and have a related TUNABLE
from CTLFLAG_RD to CTLFLAG_RDTUN so that sysctl(8) can provide
more useful error messages.

20 years agoHave sysctl print out a more useful error message when it detects that the
silby [Tue, 21 Oct 2003 16:49:30 +0000 (16:49 +0000)]
Have sysctl print out a more useful error message when it detects that the
user has attempted to write to a read only, tunable value.

20 years agoAdd another sysctl flag, CTLFLAG_TUN, which is a hint to the userland
silby [Tue, 21 Oct 2003 16:48:33 +0000 (16:48 +0000)]
Add another sysctl flag, CTLFLAG_TUN, which is a hint to the userland
sysctl that a given variable is tunable.

Also added is CTLFLAG_RDTUN, which is CTLFLAG_RD|CTLFLAG_TUN; TUN does
not always imply read-only, so RDTUN should be used where RD was used
before.

20 years agoLine-break sync to the MAC development branch.
rwatson [Tue, 21 Oct 2003 15:18:26 +0000 (15:18 +0000)]
Line-break sync to the MAC development branch.

20 years agoWe need to initialize bp->b_offset and bp->b_iooffset
simokawa [Tue, 21 Oct 2003 13:18:19 +0000 (13:18 +0000)]
We need to initialize bp->b_offset and bp->b_iooffset
becuase bp->b_blkno is ignored now.

20 years agoReject negative ngrp arguments in linux_setgroups() and linux_setgroups16();
tjr [Tue, 21 Oct 2003 11:00:33 +0000 (11:00 +0000)]
Reject negative ngrp arguments in linux_setgroups() and linux_setgroups16();
stops users being able to cause setgroups to clobber the kernel stack by
copying in data past the end of the linux_gidset array.

20 years agoUp the alignment requirement of the SC1100 to 16 bytes, it has all the bugs and then...
sos [Tue, 21 Oct 2003 08:53:29 +0000 (08:53 +0000)]
Up the alignment requirement of the SC1100 to 16 bytes, it has all the bugs and then some of its Cyrix inheritance.

Set the max_iosize for the Cyrix to 63K as the SC1100.

20 years agoRemove clause 3 in the copyrights (according to permission in src/COPYRIGHT)
phk [Tue, 21 Oct 2003 08:27:05 +0000 (08:27 +0000)]
Remove clause 3 in the copyrights (according to permission in src/COPYRIGHT)
Remove trailing whitespace.

20 years agoRemove "register" keywords.
phk [Tue, 21 Oct 2003 07:58:52 +0000 (07:58 +0000)]
Remove "register" keywords.
Use newspeak functions definitions.

20 years agoDon peril-sensitive sunglasses and mark pipe(2) as MPSAFE. I've beaten up
scottl [Tue, 21 Oct 2003 07:03:27 +0000 (07:03 +0000)]
Don peril-sensitive sunglasses and mark pipe(2) as MPSAFE.  I've beaten up
on it for the last 15 hours with no signs of problems.  It gives a small
(1%) gain on buildworld since pipe_read/pipe_write are already free of Giant.

20 years agoRetire bio_caller2 (alias for b_io.bio_caller2), this field is reserved
phk [Tue, 21 Oct 2003 06:58:58 +0000 (06:58 +0000)]
Retire bio_caller2 (alias for b_io.bio_caller2), this field is reserved
for dev_strategy() use.

Retire bio_driver[12] (aliases for b_io.bio_driver[12]) these fields are
reserved for device driver use and can as such never have any interest
in the buf end of things.

20 years agoRemove KASSERTS on B_PHYS for vmapbuf() and vunmapbuf(), B_PHYS is going
phk [Tue, 21 Oct 2003 06:53:10 +0000 (06:53 +0000)]
Remove KASSERTS on B_PHYS for vmapbuf() and vunmapbuf(), B_PHYS is going
away.

20 years agobaud rate and bit rate were horribly confused in this document. Baud
imp [Tue, 21 Oct 2003 03:22:49 +0000 (03:22 +0000)]
baud rate and bit rate were horribly confused in this document.  Baud
rate is how fast modems exchange symbols.  Bit rate is how many bits
per second the serial port nominally communicates at.  Try to use bit
rate consistently where that's what is ment.  The default data rate is
now 9600 baud.  300 baud being default pre-dates 4.4-LITE1.  Document
that tip doesn't respect the system default rate for a given channel.

Sorry guys, but no ucbvax, kremvax or kgbvax added to the example.

Noted by: bde

20 years agoRemove md_bspstore from the MD fields of struct thread. Now that
marcel [Tue, 21 Oct 2003 01:13:49 +0000 (01:13 +0000)]
Remove md_bspstore from the MD fields of struct thread. Now that
the backing store is at a fixed address, there's no need for a
per-thread variable.

20 years agoAddition.
phk [Mon, 20 Oct 2003 21:22:41 +0000 (21:22 +0000)]
Addition.

20 years agorevert default for idle polling to zero until we can resolve the
sam [Mon, 20 Oct 2003 21:14:24 +0000 (21:14 +0000)]
revert default for idle polling to zero until we can resolve the
livelock problem

20 years agoWhen a numeric field overflows its width, try formatting the number in
phk [Mon, 20 Oct 2003 20:13:50 +0000 (20:13 +0000)]
When a numeric field overflows its width, try formatting the number in
'kilo' or 'mega' with appropriate suffix instead of filling the field
with stars.

20 years ago - If a thread is not bound to a kse return 0 from sched_pctcpu().
jeff [Mon, 20 Oct 2003 19:55:21 +0000 (19:55 +0000)]
 - If a thread is not bound to a kse return 0 from sched_pctcpu().

Reported by:  pawel.worach@nordea.com

20 years agoSpelling, grammar fixes.
gsutter [Mon, 20 Oct 2003 19:17:15 +0000 (19:17 +0000)]
Spelling, grammar fixes.

Reviewed by: imp

20 years ago - Remove some long unused code.
alc [Mon, 20 Oct 2003 18:57:01 +0000 (18:57 +0000)]
 - Remove some long unused code.

20 years agoInitialize the buf's b_object in pbgetvp(). Clear it in pbrelvp(). (This
alc [Mon, 20 Oct 2003 18:24:38 +0000 (18:24 +0000)]
Initialize the buf's b_object in pbgetvp().  Clear it in pbrelvp().  (This
facilitates synchronization of the vm page's valid field using the
vm object's lock.)

Suggested by: tegge

20 years agofix build: linux_to_bsd_msf_lba is no longer used because of previous commit
sam [Mon, 20 Oct 2003 17:56:10 +0000 (17:56 +0000)]
fix build: linux_to_bsd_msf_lba is no longer used because of previous commit

20 years agoenclose IPv6 part with ifdef INET6.
ume [Mon, 20 Oct 2003 16:19:01 +0000 (16:19 +0000)]
enclose IPv6 part with ifdef INET6.

Obtained from: KAME

20 years agoMark dup as MPSAFE. Giant was pushed into dup ages ago, but it looks
dwmalone [Mon, 20 Oct 2003 16:16:03 +0000 (16:16 +0000)]
Mark dup as MPSAFE. Giant was pushed into dup ages ago, but it looks
like it was missed in syscalls.master.

Spotted by: alc

20 years agocorrect linkmtu handling.
ume [Mon, 20 Oct 2003 15:27:48 +0000 (15:27 +0000)]
correct linkmtu handling.

Obtained from: KAME

20 years agoFix typo: s/dev_db/dev_mkdb/
bmah [Mon, 20 Oct 2003 15:17:44 +0000 (15:17 +0000)]
Fix typo:  s/dev_db/dev_mkdb/

Pointed out by: Michel Lavondes <fox@vader.aacc.cc.md.us>

20 years agoRemember to check the DE_WHITEOUT flag in the case where a cloned
phk [Mon, 20 Oct 2003 15:08:10 +0000 (15:08 +0000)]
Remember to check the DE_WHITEOUT flag in the case where a cloned
device is hidden by a devfs(8) rule.

Spotted by:  Adam Nowacki <ptnowak@bsk.vectranet.pl>

20 years agoIf just gcc could make up its mind...
sos [Mon, 20 Oct 2003 14:28:37 +0000 (14:28 +0000)]
If just gcc could make up its mind...

20 years agoBring the description of st_[cma]time modification conditions a bit
dds [Mon, 20 Oct 2003 13:52:22 +0000 (13:52 +0000)]
Bring the description of st_[cma]time modification conditions a bit
closer to reality.  More work remains to be done.  st_mtime should
be the most complete based on IEEE Std 1003.1, 2003 Edition, a
review of ufs_vnops.c, and some experimentation.

20 years agoOnly return valid DMA error bits.
sos [Mon, 20 Oct 2003 13:45:11 +0000 (13:45 +0000)]
Only return valid DMA error bits.

20 years agoUp delay from 10 to 100ms after reset, this helps some slow devices
sos [Mon, 20 Oct 2003 13:44:33 +0000 (13:44 +0000)]
Up delay from 10 to 100ms after reset, this helps some slow devices
get their act together before we start probing.

20 years agoOnly announce ECC errors when its only that.
sos [Mon, 20 Oct 2003 13:32:42 +0000 (13:32 +0000)]
Only announce ECC errors when its only that.

20 years agoFix some security bugs in the SVR4 emulator:
tjr [Mon, 20 Oct 2003 10:38:48 +0000 (10:38 +0000)]
Fix some security bugs in the SVR4 emulator:
- Return NULL instead of returning memory outside of the stackgap
  in stackgap_alloc() (FreeBSD-SA-00:42.linux)
- Check for stackgap_alloc() returning NULL in svr4_emul_find(),
  and clean_pipe().
- Avoid integer overflow on large nfds argument in svr4_sys_poll()
- Reject negative nbytes argument in svr4_sys_getdents()
- Don't copy out past the end of the struct componentname
  pathname buffer in svr4_sys_resolvepath()
- Reject out-of-range signal numbers in svr4_sys_sigaction(),
  svr4_sys_signal(), and svr4_sys_kill().
- Don't malloc() user-specified lengths in show_ioc() and
  show_strbuf(), place arbitrary limits instead.
- Range-check lengths in si_listen(), ti_getinfo(), ti_bind(),
  svr4_do_putmsg(), svr4_do_getmsg(), svr4_stream_ti_ioctl().

Some fixes obtain from OpenBSD.

20 years agoUpdate to using a 56k modem connection, plus add some verbage about
imp [Mon, 20 Oct 2003 10:26:54 +0000 (10:26 +0000)]
Update to using a 56k modem connection, plus add some verbage about
what the entry does for people that are not used to reading
/etc/remote files every day.  Keep arpavax, because it is a cool name,
but remove the phone number listed.  arpavax hasn't been answering
that number for a while :-)

20 years agoWhite space only changes. Put each sentence on a line by itself, per
imp [Mon, 20 Oct 2003 10:04:55 +0000 (10:04 +0000)]
White space only changes.  Put each sentence on a line by itself, per
our normal practice.  In anticipation of bigger changes to this file.

20 years agoWe dont support CDROMREADAUDIO anymore.
sos [Mon, 20 Oct 2003 09:51:00 +0000 (09:51 +0000)]
We dont support CDROMREADAUDIO anymore.

20 years agoRemove no longer existant CDIOCREADAUDIO ioctl.
sos [Mon, 20 Oct 2003 09:29:40 +0000 (09:29 +0000)]
Remove no longer existant CDIOCREADAUDIO ioctl.

20 years agoWhen a driver successfully created a device on demand, we can directly
phk [Mon, 20 Oct 2003 07:04:09 +0000 (07:04 +0000)]
When a driver successfully created a device on demand, we can directly
pick up the DEVFS inode number from the dev_t and find our directory
entry from that, we don't need to scan the directory to find it.

This also solves an issue with on-demand devices in subdirectories.

Submitted by: cognet

20 years ago - Synchronize access to a vm page's valid field using the containing
alc [Mon, 20 Oct 2003 05:57:55 +0000 (05:57 +0000)]
 - Synchronize access to a vm page's valid field using the containing
   vm object's lock.

20 years agoPut the RSE backing store at a fixed address. This change is triggered
marcel [Mon, 20 Oct 2003 05:34:10 +0000 (05:34 +0000)]
Put the RSE backing store at a fixed address. This change is triggered
by libguile that needs to know the base of the RSE backing store. We
currently do not export the fixed address to userland by means of a
sysctl so user code needs to hardcode it for now. This will be revisited
later.

The RSE backing store is now at the bottom of region 4. The memory stack
is at the top of region 4. This means that the whole region is usable
for the stacks, giving a 61-bit stack space.

Port: lang/guile (depended of x11/gnome2)

20 years ago - Remove comments referring to functions that no longer exist.
alc [Mon, 20 Oct 2003 05:16:27 +0000 (05:16 +0000)]
 - Remove comments referring to functions that no longer exist.

20 years agoThis test relies on the concurrency level being 1; make it so.
deischen [Mon, 20 Oct 2003 04:23:49 +0000 (04:23 +0000)]
This test relies on the concurrency level being 1; make it so.

20 years agoVarious style and type fixes in my last commit.
cognet [Mon, 20 Oct 2003 04:10:20 +0000 (04:10 +0000)]
Various style and type fixes in my last commit.

Suggested by: mux

20 years ago - Hold the vm object's lock around calls to vm_page_set_validclean().
alc [Mon, 20 Oct 2003 04:05:24 +0000 (04:05 +0000)]
 - Hold the vm object's lock around calls to vm_page_set_validclean().

20 years agodocument the fact that kqueue will immediately return and not timeout when
jmg [Mon, 20 Oct 2003 02:29:16 +0000 (02:29 +0000)]
document the fact that kqueue will immediately return and not timeout when
nevents is 0.

PR: kern/45291

20 years agoFix m_head handling in sis_encap so that the correct mbuf is always handed
silby [Sun, 19 Oct 2003 23:28:02 +0000 (23:28 +0000)]
Fix m_head handling in sis_encap so that the correct mbuf is always handed
to BPF_MTAP.

20 years agoFix a problem where m_defrag would allocate a new mbuf to replace the
silby [Sun, 19 Oct 2003 23:05:19 +0000 (23:05 +0000)]
Fix a problem where m_defrag would allocate a new mbuf to replace the
chain passed into dc_encap, which dc_start was unaware of.  This caused
the old (now invalid) mbuf to be passed to BPF_MTAP.

Spotted by: Kenjiro Cho <kjc@csl.sony.co.jp>

20 years agoAdd a new macro M_ASSERTVALID which ensures that the mbuf in question
silby [Sun, 19 Oct 2003 22:33:41 +0000 (22:33 +0000)]
Add a new macro M_ASSERTVALID which ensures that the mbuf in question
is non-free.  (More checks can/should be added in the future.)

Use M_ASSERTVALID in BPF_MTAP so that we catch when freed mbufs are
passed in, even if no bpf listeners are active.

Inspired by a bug in if_dc caught by Kenjiro Cho.

20 years agoAdd a testcase which validates that the same buffer can be passed to
phk [Sun, 19 Oct 2003 22:12:23 +0000 (22:12 +0000)]
Add a testcase which validates that the same buffer can be passed to
rijndael_blockDecrypt() as both input and output.

This property is important because inside rijndael we can get away
with allocating just a 16 byte "work" buffer on the stack (which
is very cheap), whereas the calling code would need to allocate the
full sized buffer, and in all likelyhood would have to do so with
an expensive malloc(9).

20 years agoChange the default mode for lost+found from 01777 to 0700. The
iedowse [Sun, 19 Oct 2003 21:49:44 +0000 (21:49 +0000)]
Change the default mode for lost+found from 01777 to 0700. The
original intention of the less restrictive permissions was to allow
users to move or delete recovered files that they own. However, it
is better to not create world-writable directories by default; the
administrator can always pre-create lost+found if different permissions
are desired.

Reviewed by: mckusick

20 years ago- revert to old rijndael code. new rijndael code broke gbde.
ume [Sun, 19 Oct 2003 21:28:34 +0000 (21:28 +0000)]
- revert to old rijndael code.  new rijndael code broke gbde.
- since aes-xcbc-mac and aes-ctr require functions in new
  rijndael code, aes-xcbc-mac and aes-ctr are disabled for now.

20 years agofalloc allocates a file structure and adds it to the file descriptor
dwmalone [Sun, 19 Oct 2003 20:41:07 +0000 (20:41 +0000)]
falloc allocates a file structure and adds it to the file descriptor
table, acquiring the necessary locks as it works. It usually returns
two references to the new descriptor: one in the descriptor table
and one via a pointer argument.

As falloc releases the FILEDESC lock before returning, there is a
potential for a process to close the reference in the file descriptor
table before falloc's caller gets to use the file. I don't think this
can happen in practice at the moment, because Giant indirectly protects
closes.

To stop the file being completly closed in this situation, this change
makes falloc set the refcount to two when both references are returned.
This makes life easier for several of falloc's callers, because the
first thing they previously did was grab an extra reference on the
file.

Reviewed by: iedowse
Idea run past: jhb

20 years ago - Add vm object locking to vfs_clean_pages() and vfs_bio_set_validclean().
alc [Sun, 19 Oct 2003 20:39:06 +0000 (20:39 +0000)]
 - Add vm object locking to vfs_clean_pages() and vfs_bio_set_validclean().
   This is to synchronize access to the vm page's valid field by
   vm_page_set_validclean().

20 years agoRemove KASSERT check for negative bio_offsets, add "normal" EIO
phk [Sun, 19 Oct 2003 19:06:54 +0000 (19:06 +0000)]
Remove KASSERT check for negative bio_offsets, add "normal" EIO
error return for same.

20 years agoThe re(4) driver works on CURRENT/alpha.
bmah [Sun, 19 Oct 2003 18:35:40 +0000 (18:35 +0000)]
The re(4) driver works on CURRENT/alpha.

Reported by: naddy@mips.inka.de (Christian Weisgerber) on alpha@

20 years agoModified release notes: Fixed minor grammatical bugs in atkbd(4) bug
bmah [Sun, 19 Oct 2003 18:32:48 +0000 (18:32 +0000)]
Modified release notes:  Fixed minor grammatical bugs in atkbd(4) bug
and safe(4).

20 years agoFinish the removal of the bst/bsh confusion.
imp [Sun, 19 Oct 2003 17:38:04 +0000 (17:38 +0000)]
Finish the removal of the bst/bsh confusion.

20 years agoRemove debug printf().
mux [Sun, 19 Oct 2003 14:33:00 +0000 (14:33 +0000)]
Remove debug printf().

20 years agoImplement partially /proc/<pid>/maps.
cognet [Sun, 19 Oct 2003 14:13:51 +0000 (14:13 +0000)]
Implement partially /proc/<pid>/maps.
It looks enough to make SImics run.

Reviewed by: des

20 years agoVarious NFS-related bug fixes (mostly related to using variable_get()
des [Sun, 19 Oct 2003 13:37:12 +0000 (13:37 +0000)]
Various NFS-related bug fixes (mostly related to using variable_get()
instead of variable_cmp() to test boolean rc.conf variables).

Submitted by: hmp

20 years agoModified release note:
hrs [Sun, 19 Oct 2003 12:58:57 +0000 (12:58 +0000)]
Modified release note:
Fix a typo in IP_ONESBCAST item.

MFC noted:
ipfw(4) limit rule processing fix.

20 years agoMFi386: revision 1.577.
nyan [Sun, 19 Oct 2003 11:35:44 +0000 (11:35 +0000)]
MFi386: revision 1.577.

20 years agos/5.0-CURRENT/FreeBSD 5.x/
phk [Sun, 19 Oct 2003 10:24:59 +0000 (10:24 +0000)]
s/5.0-CURRENT/FreeBSD 5.x/

20 years agoFix LINT build by correcting a missed change.
bms [Sun, 19 Oct 2003 09:31:07 +0000 (09:31 +0000)]
Fix LINT build by correcting a missed change.

20 years agoDisable irqs before entering the power-off state. This is not known
njl [Sun, 19 Oct 2003 05:56:59 +0000 (05:56 +0000)]
Disable irqs before entering the power-off state.  This is not known
to fix any problems but is similar to how Linux implements this
function.

20 years agoTidy up loose ends in the idle process. Call the MI cpu_idle() function
peter [Sun, 19 Oct 2003 02:43:57 +0000 (02:43 +0000)]
Tidy up loose ends in the idle process.  Call the MI cpu_idle() function
for all platforms now.

XXX alpha/sparc64/powerpc should fill in the function.

Submitted by:  bde

20 years agoAdd a stub cpu_idle() function for sparc64, alpha, powerpc. This is a
peter [Sun, 19 Oct 2003 02:36:07 +0000 (02:36 +0000)]
Add a stub cpu_idle() function for sparc64, alpha, powerpc.  This is a
MI declared function so it should be everywhere.

20 years agoFix broken su -m behaviour :
cognet [Sun, 19 Oct 2003 02:09:36 +0000 (02:09 +0000)]
Fix broken su -m behaviour :
chshell must return 0 if the shell is not a standard shell, or else it is
possible to use an account without a valid shell.

Reviewed by: des

20 years agoUse npxdrop in cpu_thread_exit to save some cycles.
davidxu [Sun, 19 Oct 2003 00:57:10 +0000 (00:57 +0000)]
Use npxdrop in cpu_thread_exit to save some cycles.
Clear FPU pcb flags for new upcall thread, these flags needn't
be inherited, the new thread should start from clean FPU status.

20 years agoBefore trying to restart/stop the MTA or MSP queue runner, make sure
gshapiro [Sun, 19 Oct 2003 00:31:50 +0000 (00:31 +0000)]
Before trying to restart/stop the MTA or MSP queue runner, make sure
we are configured to use an MTA or MSP queue runner.  Otherwise the
script used to complain about the missing PID file.

PR: conf/57075
Submitted by: Jonathan Chen <jonc@chen.org.nz>
MFC after: 2 weeks

20 years agoAdding WorkAroundBrokenAAAA seems a necessity even in the submit.mc given
gshapiro [Sun, 19 Oct 2003 00:03:13 +0000 (00:03 +0000)]
Adding WorkAroundBrokenAAAA seems a necessity even in the submit.mc given
the number of broken DNS servers out there in the world.  Since we are
diverging from the sendmail.org submit.mc, it's time to make our own
freebsd.submit.mc.

PR: conf/57733
Reviewed by: nork
MFC after: 2 weeks

20 years agoDon't confuse tags and handles.
imp [Sun, 19 Oct 2003 00:03:10 +0000 (00:03 +0000)]
Don't confuse tags and handles.

20 years ago - Synchronize access to a vm page's valid field using the containing
alc [Sun, 19 Oct 2003 00:01:56 +0000 (00:01 +0000)]
 - Synchronize access to a vm page's valid field using the containing
   vm object's lock.
 - Reduce the scope of the vm page queues lock in two places.

20 years agoNote that PAE has been MFC'd into RELENG_4 and will appear in FreeBSD
simon [Sat, 18 Oct 2003 22:54:45 +0000 (22:54 +0000)]
Note that PAE has been MFC'd into RELENG_4 and will appear in FreeBSD
4.9.

20 years agoAdd the cpu_idle_hook() function pointer so that other idlers can be
njl [Sat, 18 Oct 2003 22:25:07 +0000 (22:25 +0000)]
Add the cpu_idle_hook() function pointer so that other idlers can be
hooked at runtime.  Make C1 sleep (e.g., HLT) be the default.  This
prepares the way for further ACPI sleep states.

20 years ago - Synchronize access to the page's valid field in
alc [Sat, 18 Oct 2003 21:30:29 +0000 (21:30 +0000)]
 - Synchronize access to the page's valid field in
   vnode_pager_generic_getpages() using the containing object's lock.

20 years ago - Increase the object lock's scope in vm_contig_launder() so that access
alc [Sat, 18 Oct 2003 21:09:21 +0000 (21:09 +0000)]
 - Increase the object lock's scope in vm_contig_launder() so that access
   to the object's type field and the call to vm_pageout_flush() are
   synchronized.
 - The above change allows for the eliminaton of the last parameter
   to vm_pageout_flush().
 - Synchronize access to the page's valid field in vm_pageout_flush()
   using the containing object's lock.

20 years agoFix a typo. The module has the EISA front-end commented out, therefore the
bms [Sat, 18 Oct 2003 20:44:23 +0000 (20:44 +0000)]
Fix a typo. The module has the EISA front-end commented out, therefore the
error may not have been picked up right away.

Reviewed by: mdodd
Submitted by: Stuart Walsh

20 years agoInitialize b_iooffset before calling VOP_[SPEC]STRATEGY
phk [Sat, 18 Oct 2003 19:49:46 +0000 (19:49 +0000)]
Initialize b_iooffset before calling VOP_[SPEC]STRATEGY

20 years agoInitialize b_iooffset before calling strategy
phk [Sat, 18 Oct 2003 19:48:21 +0000 (19:48 +0000)]
Initialize b_iooffset before calling strategy

20 years agoOnly automatically create an 'a' partition when there is nothing
phk [Sat, 18 Oct 2003 19:32:35 +0000 (19:32 +0000)]
Only automatically create an 'a' partition when there is nothing
but a 'c' partition.

20 years agoCorrections to revision 1.305
alc [Sat, 18 Oct 2003 18:48:17 +0000 (18:48 +0000)]
Corrections to revision 1.305
 - Specifying VM_MAP_WIRE_HOLESOK should not assume that the start
   address is the beginning of the map.  Instead, move to the first
   entry after the start address.
 - The implementation of VM_MAP_WIRE_HOLESOK was incomplete.  This
   caused the failure of mlockall(2) in some circumstances.

20 years agoNew release note:
hrs [Sat, 18 Oct 2003 18:18:17 +0000 (18:18 +0000)]
New release note:
Fix ipfw(4) limit rule processing.

20 years agoRetire b_pblkno which was an alias for a bio field which is for
phk [Sat, 18 Oct 2003 18:17:55 +0000 (18:17 +0000)]
Retire b_pblkno which was an alias for a bio field which is for
device drivers only.

20 years agoDon't report b_pblkno, it is going away.
phk [Sat, 18 Oct 2003 17:59:02 +0000 (17:59 +0000)]
Don't report b_pblkno, it is going away.

20 years agoDo not initialize bp->b_pblkno, it is going away.
phk [Sat, 18 Oct 2003 17:57:48 +0000 (17:57 +0000)]
Do not initialize bp->b_pblkno, it is going away.

20 years agoRetire bio_blkno entirely.
phk [Sat, 18 Oct 2003 17:53:34 +0000 (17:53 +0000)]
Retire bio_blkno entirely.

bio_offset is the field drivers should use.
bio_pblkno remains as a convenient place to store the number of
the device drivers.

20 years agoEliminate use bio_blkno.
phk [Sat, 18 Oct 2003 17:51:26 +0000 (17:51 +0000)]
Eliminate use bio_blkno.