]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoThis commit was generated by cvs2svn to compensate for changes in r118765,
njl [Mon, 11 Aug 2003 05:54:31 +0000 (05:54 +0000)]
This commit was generated by cvs2svn to compensate for changes in r118765,
which included commits to RCS files with non-trunk default branches.

21 years agoFix AcpiOsMapMemory to match the function definition. Don't use UINT32
njl [Mon, 11 Aug 2003 05:54:31 +0000 (05:54 +0000)]
Fix AcpiOsMapMemory to match the function definition.  Don't use UINT32
as a cast for a pointer.  Change has been submitted to the vendor.

Pointed out by: marcel, obrien

21 years agoMore pipe changes:
silby [Mon, 11 Aug 2003 05:51:51 +0000 (05:51 +0000)]
More pipe changes:

From alc:
Move pageable pipe memory to a seperate kernel submap to avoid awkward
vm map interlocking issues.  (Bad explanation provided by me.)

From me:
Rework pipespace accounting code to handle this new layout, and adjust
our default values to account for the fact that we now have a solid
limit on allocations.

Also, remove the "maxpipes" limit, as it no longer has a purpose.
(The limit on kva usage solves the problem of having two many pipes.)

21 years ago2nd part of making the name of the release running script overrideable.
obrien [Mon, 11 Aug 2003 04:21:58 +0000 (04:21 +0000)]
2nd part of making the name of the release running script overrideable.

21 years agoDon't provide mutexes for static busdma allocations.
scottl [Mon, 11 Aug 2003 03:39:41 +0000 (03:39 +0000)]
Don't provide mutexes for static busdma allocations.

21 years agoImplement cpu_set_upcall_kse(). Further tweaking may be needed after
marcel [Mon, 11 Aug 2003 01:38:23 +0000 (01:38 +0000)]
Implement cpu_set_upcall_kse(). Further tweaking may be needed after
testing.

21 years agoUse vm_page_hold() instead of vm_page_wire(). Otherwise, a multithreaded
alc [Mon, 11 Aug 2003 00:17:44 +0000 (00:17 +0000)]
Use vm_page_hold() instead of vm_page_wire().  Otherwise, a multithreaded
application could cause a wired page to be freed.  In general,
vm_page_hold() should be preferred for ephemeral kernel mappings of pages
borrowed from a user-level address space.  (vm_page_wire() should really be
reserved for indefinite duration pinning by the "owner" of the page.)

Discussed with: silby
Submitted by: tegge

21 years agoThe iBCS2 system call translator for statfs(2) did not check the
nectar [Sun, 10 Aug 2003 23:26:16 +0000 (23:26 +0000)]
The iBCS2 system call translator for statfs(2) did not check the
length parameter for validity.

Submitted by: David Rhodus <drhodus@catpa.com>

21 years agopanic() if we try to handle an out-of-range signal number in
nectar [Sun, 10 Aug 2003 23:05:37 +0000 (23:05 +0000)]
panic() if we try to handle an out-of-range signal number in
psignal()/tdsignal().  The test was historically in psignal().  It was
changed into a KASSERT, and then later moved to tdsignal() when the
latter was introduced.

Reviewed by: iedowse, jhb

21 years agoAdd or correct range checking of signal numbers in system calls and
nectar [Sun, 10 Aug 2003 23:04:55 +0000 (23:04 +0000)]
Add or correct range checking of signal numbers in system calls and
ioctls.

In the particular case of ptrace(), this commit more-or-less reverts
revision 1.53 of sys_process.c, which appears to have been erroneous.

Reviewed by: iedowse, jhb

21 years agoAdd some quick pathes to exit process when signal action is default and
davidxu [Sun, 10 Aug 2003 22:35:46 +0000 (22:35 +0000)]
Add some quick pathes to exit process when signal action is default and
signal can causes process to exit.

Reviewed by: deischen

21 years agoInitialize rtld lock just before turning on thread mode and
davidxu [Sun, 10 Aug 2003 22:30:20 +0000 (22:30 +0000)]
Initialize rtld lock just before turning on thread mode and
uninitialize rtld lock after thread mode shutdown.

21 years agoIf thread mode is not activated yet, just call __sys_fork() directly,
davidxu [Sun, 10 Aug 2003 22:20:41 +0000 (22:20 +0000)]
If thread mode is not activated yet, just call __sys_fork() directly,
otherwise masks all signals until fork() returns, in child process,
we reset library state before restoring signal masks until we reach
a safe to point.

Reviewed by: deischen

21 years agoTweak rtld lock to allow recursive on reader lock and detect recursive
davidxu [Sun, 10 Aug 2003 22:15:03 +0000 (22:15 +0000)]
Tweak rtld lock to allow recursive on reader lock and detect recursive
on writer lock. This is first cut at rwlock for rtld.

Submitted by: desichen

21 years agoIf thread mode is not activated yet, don't do extra work.
davidxu [Sun, 10 Aug 2003 22:07:28 +0000 (22:07 +0000)]
If thread mode is not activated yet, don't do extra work.

Reviewed by: deischen

21 years agoChange name of state_link() to state_polling() to make it
mbr [Sun, 10 Aug 2003 22:01:37 +0000 (22:01 +0000)]
Change name of state_link() to state_polling() to make it
more clear what it does.

Trim interface_active() to just do what it should do. Check
if we got link or not and if the NIC supports it. No special
treatment for mediachecks here anymore.

Simplify the code a lot, and remove doublicated parts.

Fix two minor spelling errors.

Add one missing #ifdef ENABLE_POLLING_MODE

Reviewed by: mdodd

21 years agoRename pmap_changebit() to pmap_clear_ptes() and remove the last
alc [Sun, 10 Aug 2003 21:53:55 +0000 (21:53 +0000)]
Rename pmap_changebit() to pmap_clear_ptes() and remove the last
parameter.  The new name better reflects what the function does and
how it is used.  The last parameter was always FALSE.

Note: In theory, gcc would perform constant propagation and dead code
elimination to achieve the same effect as removing the last parameter,
which is always FALSE.  In practice, recent versions do not.  So, there
is little point in letting unused code pessimize execution.

21 years agoo move cpu_reset() from vm_machdep.c to machdep.c.
marcel [Sun, 10 Aug 2003 21:33:07 +0000 (21:33 +0000)]
o move cpu_reset() from vm_machdep.c to machdep.c.
o reorder cpu_boot(), cpu_halt() and identifycpu().

No functional change.

21 years agoRevert revision 1.25 now that mergemaster does a 'make all' to build
gshapiro [Sun, 10 Aug 2003 20:48:59 +0000 (20:48 +0000)]
Revert revision 1.25 now that mergemaster does a 'make all' to build
the required .cf file(s).

Submitted by: ru
X-MFC after: mergemaster MFC

21 years agoThe BCM5704 ASIC has a smaller mbuf space than the 5700/1/2/3. Failure
wpaul [Sun, 10 Aug 2003 18:04:35 +0000 (18:04 +0000)]
The BCM5704 ASIC has a smaller mbuf space than the 5700/1/2/3. Failure
to configure this correctly yields many watchdog timeouts even on lightly
loaded machines. This is a common complaint from users with Dell 1750
servers with built-in dual 5704 NICs.

21 years agoLower initial drop value to 50, it is enough to hide linearity
ache [Sun, 10 Aug 2003 17:56:58 +0000 (17:56 +0000)]
Lower initial drop value to 50, it is enough to hide linearity

21 years agoFor type 0 rng lower initial drop to 50, it is enough to hide linearity
ache [Sun, 10 Aug 2003 17:49:55 +0000 (17:49 +0000)]
For type 0 rng lower initial drop to 50, it is enough to hide linearity
Reorganize historic #ifdef section

21 years agoSwitch from legacy to native mode for ICH4 and ICH5.
orion [Sun, 10 Aug 2003 14:57:47 +0000 (14:57 +0000)]
Switch from legacy to native mode for ICH4 and ICH5.

Submitted by: Shin-ichi YOSHIMOTO <yosimoto@waishi.jp>
Test by: Markko Merzin <markko@short.cut.ee>
PR: kern/53242
MFC after: 5 days

21 years agoDon't hardcode unit numer '0'.
ticso [Sun, 10 Aug 2003 14:28:24 +0000 (14:28 +0000)]
Don't hardcode unit numer '0'.
We can have multiple instances.

21 years agoAdd GB18030 and GBK to the list of encodings. Cross-reference the
tjr [Sun, 10 Aug 2003 11:51:14 +0000 (11:51 +0000)]
Add GB18030 and GBK to the list of encodings. Cross-reference the
new manual pages for the Big5, GB18030, GBK, and MSKanji encodings.

21 years agoFix the case of the encoding name in the ENCODING line. Names are
tjr [Sun, 10 Aug 2003 11:41:38 +0000 (11:41 +0000)]
Fix the case of the encoding name in the ENCODING line. Names are
case-sensitive, and MSKANJI does not work.

21 years agoCross-reference gbk(5).
tjr [Sun, 10 Aug 2003 11:38:28 +0000 (11:38 +0000)]
Cross-reference gbk(5).

21 years agoCross-reference gbk(5) now that it exists. Fix a copy & paste error:
tjr [Sun, 10 Aug 2003 11:36:42 +0000 (11:36 +0000)]
Cross-reference gbk(5) now that it exists. Fix a copy & paste error:
one occurrence of GB 18030 should have been 11383.

21 years agoAdd a fairly minimal manual page for the GBK encoding.
tjr [Sun, 10 Aug 2003 11:34:35 +0000 (11:34 +0000)]
Add a fairly minimal manual page for the GBK encoding.

21 years agoAdd a cross reference to Unicode 3.0.
tjr [Sun, 10 Aug 2003 11:26:18 +0000 (11:26 +0000)]
Add a cross reference to Unicode 3.0.

21 years agoMake the name of the release running script overrideable.
obrien [Sun, 10 Aug 2003 09:54:22 +0000 (09:54 +0000)]
Make the name of the release running script overrideable.

21 years agoAdd cross references to the new character encoding manual pages,
tjr [Sun, 10 Aug 2003 09:25:52 +0000 (09:25 +0000)]
Add cross references to the new character encoding manual pages,
and to mbsinit(3) while I'm at it.

21 years agoAdd manual pages for the BIG5, GB18030 and MSKanji encodings. These may
tjr [Sun, 10 Aug 2003 09:23:51 +0000 (09:23 +0000)]
Add manual pages for the BIG5, GB18030 and MSKanji encodings. These may
need to be fleshed out a little, especially big5(5).

21 years agoNow that we can ignore up to 8KB of dirty registers, remove the RSE
marcel [Sun, 10 Aug 2003 08:04:21 +0000 (08:04 +0000)]
Now that we can ignore up to 8KB of dirty registers, remove the RSE
magic from exec_setregs(). In set_mcontext() we now also don't have
to worry that we entered the kernel with more that 512 bytes of
dirty registers on the kernel stack. Note that we cannot make any
assumptions anymore WRT to NaT collection points in exec_setregs(),
so we have to deal with them now.

21 years agoMF p4:
imp [Sun, 10 Aug 2003 03:47:29 +0000 (03:47 +0000)]
MF p4:

Minor style(9) nit.

21 years agoGrrr. There is a gratuitous difference in the RX descriptor status
wpaul [Sun, 10 Aug 2003 02:41:18 +0000 (02:41 +0000)]
Grrr. There is a gratuitous difference in the RX descriptor status
word between the 8139C+ and the 8169. The 8139C+ has a 'frame alignment
error bit' (bit 27) but the 8169 does not. Rather than simply mark this
bit as reserved, RealTek removed it completely and shifted the remaining
status bits one space to the left. This was causing rl_rxeofcplus()
to misparse the error and checksum bits.

To workaround this, rl_rxeofcplus() now shifts the rxstat word one
bit to the right before testing any of the status bits (but after
the frame length has been extracted).

21 years agoSome bridges, like o2 micro and maybe ricoh, have some issues between
imp [Sun, 10 Aug 2003 02:40:49 +0000 (02:40 +0000)]
Some bridges, like o2 micro and maybe ricoh, have some issues between
the time the card is inserted and the time that the card is
configured.  This can lead to interrupt storms.  The O2Micro suggested
workaround is to route the card function interrupt to IRQ1.  It
appears from my testing that this is an acceptable workaround for most
chipsets (there's still some issue with the ricoh chipset).

Also, only look at the NOT_A_CARD bit when the bridge tells us there's
a card present.  At least one test caused this to be true after the
card was removed, but the author couldn't recreate it with the
workaround in place.  The change is more conservative than the
previous code, but still has the work around that wasn't present in
the older code.

21 years ago- Update some comments regarding hardware details of the 8169 and
wpaul [Sun, 10 Aug 2003 01:41:35 +0000 (01:41 +0000)]
- Update some comments regarding hardware details of the 8169 and
  note the existence of the 8169S and 8110S components. (The 8169
  is just a MAC, the 8169S and 8110S contain both a MAC and PHY.)

- Properly handle list and buffer addresses as 64-bit. The RX and
  TX DMA list addresses should be bus_addr_t's. Added RL_ADDR_HI()
  and RL_ADDR_LO() macros to obtain values for writing into chip
  registers.

- Set a slightly different TIMERINT value for 8169 NICs for improved
  performance.

- Change left out of previous commit log: added some additional
  hardware rev codes for other 10/100 chips and for the 8169S/8110S
  'rev C' gigE MACs.

21 years agoRe-enable pcfclock in the build.
obrien [Sun, 10 Aug 2003 01:35:36 +0000 (01:35 +0000)]
Re-enable pcfclock in the build.

21 years agoSilence unwanted 'gigabit link up' messages: do not treat the
wpaul [Sun, 10 Aug 2003 01:32:16 +0000 (01:32 +0000)]
Silence unwanted 'gigabit link up' messages: do not treat the
BGE_MACSTAT_MI_COMPLETE bit in the MAC status register as a link
change indicator. We turn this bit on now because some of the newer
chips need it, but it usually just means that reading/writing
an MII/GMII register has completed, not that a link change has
occured.

21 years agoInstead of unconditionally refusing to install if no swap partitions
das [Sun, 10 Aug 2003 01:04:05 +0000 (01:04 +0000)]
Instead of unconditionally refusing to install if no swap partitions
are specified, prompt the user with a yes/no box.

21 years agoUse get_mcontext in sendsig and set_mcontext in sigreturn instead of
jake [Sat, 9 Aug 2003 23:14:33 +0000 (23:14 +0000)]
Use get_mcontext in sendsig and set_mcontext in sigreturn instead of
frobbing things directly.

21 years agoDon't provide a mutex in the S/G list dma tag since it will never be defered.
scottl [Sat, 9 Aug 2003 23:07:21 +0000 (23:07 +0000)]
Don't provide a mutex in the S/G list dma tag since it will never be defered.
Fix some nearby style.

21 years agoAdd code to cope more with strange conditions after careful study of
imp [Sat, 9 Aug 2003 21:05:54 +0000 (21:05 +0000)]
Add code to cope more with strange conditions after careful study of
the standard.

1) When the bridge tells us that we have a card that isn't recognized, we
   use the force register to force the CV_TEST to run.  This test causes the
   bridge to re-evaluate the card.  Once this re-evaluation process happens,
   we get a new interrupt that may say it is ready to process.  We try this up
   to 20 times.  Tests have shown that this appears to correctly reset the
   'Unknown card type' problem that I saw on my Sony PCG-505TS.
2) Take a page from OLDCARD and always read the CSC register in the ISR.
   Some TI (and it seems maybe Ricoh) chipsets require this to behave
   properly.  This work around appears to work due to some power management
   protocols that were improperly implemented.  Maybe it can be removed when
   this driver supports the full PME# protocol described in the standards.
3) Minor additional debug printf when debugging is enabled.
4) Minor additional commentary for things that are obvious only after study.

# I'm committing this from my Sony PCG-505TS using shared PCI interrupts
# and NEWCARD, but there are some issues with the Ricoh bridge still, but
# at least now I can boot with the card inserted and have it work.

21 years agoCBB_SOCKET_STAT_ is too long. Change it to CBB_STATE_. Also, make
imp [Sat, 9 Aug 2003 20:49:22 +0000 (20:49 +0000)]
CBB_SOCKET_STAT_ is too long.  Change it to CBB_STATE_.  Also, make
some minor changes to more closely match the bit names used in the 7.0
version of the PCMCIA standard.  Also add bits from the FORCE
register.

21 years agoFix devices which do not support ifm_status. Always return
mbr [Sat, 9 Aug 2003 20:44:08 +0000 (20:44 +0000)]
Fix devices which do not support ifm_status. Always return
TRUE for them.

Reported by: mdodd
Tested by: Craig Rodrigues <rodrigc@crodrigues.org>

21 years agoWhen allocating memory, align it to the minimum alignment required for
imp [Sat, 9 Aug 2003 20:38:53 +0000 (20:38 +0000)]
When allocating memory, align it to the minimum alignment required for
cardbus bridges if the alignment requirements were less restrictive.

21 years agoTurn the annoying and long error message off. It was so
mbr [Sat, 9 Aug 2003 20:36:06 +0000 (20:36 +0000)]
Turn the annoying and long error message off. It was so
long that it was even hard to find the real error cause.

Requested by: rwatson

21 years agoAdd comments about some alignment that looks a little strange.
imp [Sat, 9 Aug 2003 20:33:53 +0000 (20:33 +0000)]
Add comments about some alignment that looks a little strange.

21 years agoMinor fixes to a couple of comments.
imp [Sat, 9 Aug 2003 20:31:25 +0000 (20:31 +0000)]
Minor fixes to a couple of comments.

21 years agoAdd a nsmb.conf manual page.
trhodes [Sat, 9 Aug 2003 19:11:52 +0000 (19:11 +0000)]
Add a nsmb.conf manual page.

PR: 45833
Submitted by: osa@FreeBSD.org (original version)
Reviewed by: ru, tjr, osa, -doc

21 years agoApparently I limited my testing of the ductape and bubblegum solution
marcel [Sat, 9 Aug 2003 19:11:20 +0000 (19:11 +0000)]
Apparently I limited my testing of the ductape and bubblegum solution
to acpidb. The same problem exists in iasl. Add JIT patching there
too.
Add a comment to both makefiles to increase the chance that both
kludges are removed when a real solution is committed.

21 years agoBackground: When proc_rwmem() wired and mapped a page, it also added
alc [Sat, 9 Aug 2003 18:01:19 +0000 (18:01 +0000)]
Background: When proc_rwmem() wired and mapped a page, it also added
a reference to the containing object.  The purpose of the reference
being to prevent the destruction of the object and an attempt to free
the wired page.  (Wired pages can't be freed.)  Unfortunately, this
approach does not work.  Some operations, like fork(2) that call
vm_object_split(), can move the wired page to a difference object,
thereby making the reference pointless and opening the possibility
of the wired page being freed.

A solution is to use vm_page_hold() in place of vm_page_wire().  Held
pages can be freed.  They are moved to a special hold queue until the
hold is released.

Submitted by: tegge

21 years agoo There are 6 trap disable bits in ar.fpsr, not five. Even though we
marcel [Sat, 9 Aug 2003 17:07:24 +0000 (17:07 +0000)]
o There are 6 trap disable bits in ar.fpsr, not five. Even though we
  didn't provide a constant for one of them (non-IEEE denormal trap),
  in an attempt to not support it probably, it's not we are left with
  the lower 5 bits.
o Properly mask the passed or returned fp_except_t. Not doing so
  causes instant core dumps by trying to write an invalid value to
  ar.fpsr. Now that we're masking, stop using exclusive-or to invert
  bits.

This fixes the illegal instruction fault encountered when building
mozilla.

21 years agoAdd libpthread to the alpha build.
deischen [Sat, 9 Aug 2003 15:29:52 +0000 (15:29 +0000)]
Add libpthread to the alpha build.

Requested by ru: Since the majority of archs can now support the
build of libpthread, rearrange the Makefile to treat libpthread
as an exception.

21 years agoCall bus_dmamap_create() to create the per-SRB buffer maps. Note that none
scottl [Sat, 9 Aug 2003 15:16:45 +0000 (15:16 +0000)]
Call bus_dmamap_create() to create the per-SRB buffer maps.  Note that none
of the SRB resources are freed when the driver unloads; maybe that will be
tackled another day.

21 years agoPerform first-aid: unbreak the build for amd64 and ia64 by patching
marcel [Sat, 9 Aug 2003 07:06:09 +0000 (07:06 +0000)]
Perform first-aid: unbreak the build for amd64 and ia64 by patching
osunixxf.c on the fly. This avoids having to pull it from the vendor
branch or otherwise pollute the repository with new short-lived files.

This should hold until the real fix arrives.

21 years agoAdd alpha support to libpthread. It compiles but hasn't been tested;
deischen [Sat, 9 Aug 2003 05:44:27 +0000 (05:44 +0000)]
Add alpha support to libpthread.  It compiles but hasn't been tested;
there is still some missing kernel support.

Reviewed by: marcel

21 years agoAlso ifdef the variable which becomes unused w/o INET6, fixing the build
silby [Sat, 9 Aug 2003 05:44:00 +0000 (05:44 +0000)]
Also ifdef the variable which becomes unused w/o INET6, fixing the build
after the previous commit.

Noticed by: alc

21 years agoAdd signalcontext() which will be needed by libpthread.
deischen [Sat, 9 Aug 2003 05:37:54 +0000 (05:37 +0000)]
Add signalcontext() which will be needed by libpthread.

Reviewed by: marcel

21 years agoAdd more detail about the filesystem-specific options available to the user.
bms [Sat, 9 Aug 2003 04:21:18 +0000 (04:21 +0000)]
Add more detail about the filesystem-specific options available to the user.
Administer zen master treatment to the side of this new committer's head.

PR: docs/53688
Submitted by: bms
Approved by: jake (mentor)

21 years agoPR: docs/53688
bms [Sat, 9 Aug 2003 04:07:42 +0000 (04:07 +0000)]
PR: docs/53688
Submitted by: bms
Approved by: jake (mentor)

21 years agoAdd the POSIX 1003.1-2001 posix_madvise() interface.
bms [Sat, 9 Aug 2003 03:23:24 +0000 (03:23 +0000)]
Add the POSIX 1003.1-2001 posix_madvise() interface.

PR: standards/54634
Reviewed by: das
Approved by: jake (mentor)

21 years ago#ifdef INET6 the if_loop packet defrag; since only the ipv6 stack (might)
silby [Sat, 9 Aug 2003 02:55:11 +0000 (02:55 +0000)]
#ifdef INET6 the if_loop packet defrag; since only the ipv6 stack (might)
require this to be done, there's no reason to waste time doing it if
ipv6 isn't compiled in.

MFC after: 1 week

21 years agoFix sign-extension bug for 32 and 64-bit values. For 64-bit values
marcel [Sat, 9 Aug 2003 01:55:37 +0000 (01:55 +0000)]
Fix sign-extension bug for 32 and 64-bit values. For 64-bit values
this involves the sign-extension of the high and low "word". Both
of which are 32-bit. The bug is especially harmful on ia64, where
0x9fffffffe0000000 is a common address (base of register stack).
This was invariably displayed as 0xffffffffe0000000.

The sign-extension is fixed by using {b|l}e{16|32|64}dec() where
applicable. Since elfdump(1) is not a bootstrap tool, dependency
on these functions is not a problem.

21 years agoAdd the PCI Id of the Intersil prism3 (mini-PCI) WLAN chip. The Vaio
marcel [Fri, 8 Aug 2003 23:33:42 +0000 (23:33 +0000)]
Add the PCI Id of the Intersil prism3 (mini-PCI) WLAN chip. The Vaio
PCG-505BX (for example) has one of those:

wi0: <Intersil Prism3> mem 0xf8000000-0xf8000fff at device 2.0 on pci2
wi0: 802.11 address: 00:02:8a:94:d8:73
wi0: using RF:PRISM3(Mini-PCI)
wi0: Intersil Firmware: Primary (1.1.1), Station (1.5.6)
wi0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps

21 years agoInitialize bp->bio_driver1 in idad_strategy().
mdodd [Fri, 8 Aug 2003 23:01:12 +0000 (23:01 +0000)]
Initialize bp->bio_driver1 in idad_strategy().
Removed by accident in 1.38.

This fixes the trap 12 everyone was seeing in ida_construct_qcb().

PR:  kern/53245

21 years ago - Remove GIANT_REQUIRED from pipespace().
alc [Fri, 8 Aug 2003 22:38:15 +0000 (22:38 +0000)]
 - Remove GIANT_REQUIRED from pipespace().
 - Remove a duplicate initialization from pipe_create().

21 years agoo Add code to GC freed KSEs and KSE groups
davidxu [Fri, 8 Aug 2003 22:20:59 +0000 (22:20 +0000)]
o Add code to GC freed KSEs and KSE groups
o Fix a bug in kse_free_unlocked(), kcb_dtor shouldn't be called because
  the KSE is cached and will be resued in _kse_alloc().

Reviewed by: deischen

21 years agoSince it builds and seems to work OK, add libpthread to the amd64 build.
deischen [Fri, 8 Aug 2003 21:16:06 +0000 (21:16 +0000)]
Since it builds and seems to work OK, add libpthread to the amd64 build.

21 years agoAdded the new knob MINIROOT that, when set, causes the creation
ru [Fri, 8 Aug 2003 21:00:57 +0000 (21:00 +0000)]
Added the new knob MINIROOT that, when set, causes the creation
of the bootable UFS image, miniroot.ufs, fully identical to the
contents of the "bootonly" CD-ROM.  The image is made available
on FTP as miniroot/miniroot.ufs.gz, for resemblance with NetBSD.

This replaces the boot.flp image for sparc64, making the latter
as coming with NO_FLOPPIES.

Reviewed by: jhb
Tested by: hrs

21 years agoCopyin the thread mailbox flags from the correct location
deischen [Fri, 8 Aug 2003 20:23:10 +0000 (20:23 +0000)]
Copyin the thread mailbox flags from the correct location
in the mailbox.

21 years agoDrop MIP6 code. We don't support MIP6 yet.
ume [Fri, 8 Aug 2003 20:14:28 +0000 (20:14 +0000)]
Drop MIP6 code.  We don't support MIP6 yet.

MFC after: 1 week

21 years agoMake this WARNS=6 clean by prototyping 'usage'.
johan [Fri, 8 Aug 2003 19:51:01 +0000 (19:51 +0000)]
Make this WARNS=6 clean by prototyping 'usage'.

Tested by: make universe

21 years agoRemove stale DCE 1.1 pointers to the IBM site. The URLs aren't valid
marcel [Fri, 8 Aug 2003 19:18:43 +0000 (19:18 +0000)]
Remove stale DCE 1.1 pointers to the IBM site. The URLs aren't valid
anymore. This also fixes long line bugs caused by the lengthy URLs :-)

21 years agoFix markup for uuid_equal()
marcel [Fri, 8 Aug 2003 19:12:28 +0000 (19:12 +0000)]
Fix markup for uuid_equal()

21 years agoFix two (2) bugs in one (1) statement:
marcel [Fri, 8 Aug 2003 19:03:37 +0000 (19:03 +0000)]
Fix two (2) bugs in one (1) statement:
o  fix the len argument of memcmp(3) to be the size of the node field
   of the uuid structure, not the size of the uuid structure itself.
   We're comparing the node fields...
o  uuid_compare(3) is specified to return -1, 0 or 1, depending on
   the outcome of the comparison. memcmp(3) returns the difference
   between the first differing bytes. Hence, we cannot ever return
   the return value of memcmp(3) as-is.

PR: standards/55370
Submitted by: Konstantin Oznobihin <bork@rsu.ru>

21 years agoAdd support for automatically computing the required file system
ru [Fri, 8 Aug 2003 18:23:00 +0000 (18:23 +0000)]
Add support for automatically computing the required file system
size and average inode size.

Reviewed by: jhb

21 years agoadd rtsol_flags.
ume [Fri, 8 Aug 2003 17:43:58 +0000 (17:43 +0000)]
add rtsol_flags.

MFC after: 1 week

21 years agoNote in STANDARDS that -gno are not (yet?) POSIX conformant.
schweikh [Fri, 8 Aug 2003 17:04:17 +0000 (17:04 +0000)]
Note in STANDARDS that -gno are not (yet?) POSIX conformant.

Spotted by: harti
MFC after: 3 days

21 years agoKNF, correct typos and cleanup spaces.
ume [Fri, 8 Aug 2003 16:56:01 +0000 (16:56 +0000)]
KNF, correct typos and cleanup spaces.

Obtained from: KAME
MFC after: 1 week

21 years agodrop the code for the environment where getifaddrs(3) is not
ume [Fri, 8 Aug 2003 16:52:31 +0000 (16:52 +0000)]
drop the code for the environment where getifaddrs(3) is not
supported.

Obtained from: KAME
MFC after: 1 week

21 years agostop use of register.
ume [Fri, 8 Aug 2003 16:47:06 +0000 (16:47 +0000)]
stop use of register.

Obtained from: KAME
MFC after: 1 week

21 years agoProcess the "O bit."
ume [Fri, 8 Aug 2003 16:42:37 +0000 (16:42 +0000)]
Process the "O bit."
If the value of OtherConfigFlag changes from FALSE to TRUE, the
host should invoke the stateful autoconfiguration protocol,
requesting information.
[RFC 2462 Section 5.5.3]

Obtained from: KAME
MFC after: 1 week

21 years ago__FUNCTION__ --> __func__
ume [Fri, 8 Aug 2003 16:38:23 +0000 (16:38 +0000)]
__FUNCTION__ --> __func__

Obtained from: KAME
MFC after: 1 week

21 years agoNew release notes: rl(4) supports RealTek 8139C+ and 8169, ccd(4)
bmah [Fri, 8 Aug 2003 15:59:34 +0000 (15:59 +0000)]
New release notes:  rl(4) supports RealTek 8139C+ and 8169, ccd(4)
support for raw disks, acpidb(8), iasl(8).

21 years agoFix a bug that caused i386 to produce broken binaries for big-endian
ru [Fri, 8 Aug 2003 15:44:17 +0000 (15:44 +0000)]
Fix a bug that caused i386 to produce broken binaries for big-endian
64-bit platforms.  (size_t is 32bit on i386, but Elf_Shdr.sh_size is
64bit on SPARC64.)

21 years agoImplement signalcontext.
jake [Fri, 8 Aug 2003 15:40:28 +0000 (15:40 +0000)]
Implement signalcontext.

21 years agoccd(4) can now operate on raw disks and other GEOM providers.
das [Fri, 8 Aug 2003 15:17:15 +0000 (15:17 +0000)]
ccd(4) can now operate on raw disks and other GEOM providers.
Remove the paragraph that says it can't.

Discussed with: phk

21 years agoFix comment
ache [Fri, 8 Aug 2003 14:02:20 +0000 (14:02 +0000)]
Fix comment

21 years agoAdd ko_KR.CP949 locale
ache [Fri, 8 Aug 2003 13:48:15 +0000 (13:48 +0000)]
Add ko_KR.CP949 locale

PR:             55341
Submitted by:   Hye-Shik Chang <perky@freebsd.org>

21 years agoAdd ko_KR.CP949 locale
ache [Fri, 8 Aug 2003 13:43:49 +0000 (13:43 +0000)]
Add ko_KR.CP949 locale

PR:             55341
submitted by:   Hye-Shik Chang <perky@freebsd.org>

21 years agoMFi386: revision 1.201.
nyan [Fri, 8 Aug 2003 12:45:42 +0000 (12:45 +0000)]
MFi386: revision 1.201.

21 years agoAdd /rescue/ and a reference to its manual page rescue(8).
simon [Fri, 8 Aug 2003 11:15:57 +0000 (11:15 +0000)]
Add /rescue/ and a reference to its manual page rescue(8).

Approved by: ceri (mentor)

21 years agoAllow gcc driver to process -r option iself, do not use -Wl,-r to
kan [Fri, 8 Aug 2003 03:41:13 +0000 (03:41 +0000)]
Allow gcc driver to process -r option iself, do not use -Wl,-r to
bypass it. Doing otherwise did not allow compiler to detect and disable
conflicting options generated from specs.

Reported by: jake

21 years agoHook acpidb(8) and iasl(8) up to the build and unhook amldb(8). It will
njl [Fri, 8 Aug 2003 03:20:59 +0000 (03:20 +0000)]
Hook acpidb(8) and iasl(8) up to the build and unhook amldb(8).  It will
be going away after a short while but still available in ports.

21 years agoDefault to 5 seconds before sleeping to give some machines time to stabilize.
njl [Fri, 8 Aug 2003 03:19:22 +0000 (03:19 +0000)]
Default to 5 seconds before sleeping to give some machines time to stabilize.
This doesn't break anything on my laptop and some claim it helps them.

21 years agoMFi386 1.422 & 1.423: lock page queues in pmap_insert_entry().
alc [Fri, 8 Aug 2003 01:52:03 +0000 (01:52 +0000)]
MFi386 1.422 & 1.423: lock page queues in pmap_insert_entry().

21 years agoMFi386 1.422 & 1.423: lock page queues in pmap_insert_entry().
marcel [Fri, 8 Aug 2003 00:30:26 +0000 (00:30 +0000)]
MFi386 1.422 & 1.423: lock page queues in pmap_insert_entry().

21 years ago0) preallocate per-interface context structures without the ifnet lock held
billf [Thu, 7 Aug 2003 21:27:17 +0000 (21:27 +0000)]
0) preallocate per-interface context structures without the ifnet lock held
1) avoid immediately calling bzero() after malloc() by passing M_ZERO
2) do not initialize individual members of the global context to zero
3) remove an unused assignment of ifctx in bootpc_init()

Reviewed by: tegge