]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoThis commit was generated by cvs2svn to compensate for changes in r121934,
Hartmut Brandt [Mon, 3 Nov 2003 09:05:49 +0000 (09:05 +0000)]
This commit was generated by cvs2svn to compensate for changes in r121934,
which included commits to RCS files with non-trunk default branches.

20 years agoVirgin import of NgATM shared kernel/user part 0.91
Hartmut Brandt [Mon, 3 Nov 2003 09:05:49 +0000 (09:05 +0000)]
Virgin import of NgATM shared kernel/user part 0.91

20 years agoHandle unaligned 4-byte loads. While in the neighborhood, remove the
Marcel Moolenaar [Mon, 3 Nov 2003 08:04:04 +0000 (08:04 +0000)]
Handle unaligned 4-byte loads. While in the neighborhood, remove the
cr.isr sanity check. We actually encounter insanities, which very
likely means that the insanity check itself is insane. Remove an empty
comment while I'm at it.

20 years agoRemove alpha-bitops.h now that it's unused. It's in the attic if
Marcel Moolenaar [Mon, 3 Nov 2003 07:10:53 +0000 (07:10 +0000)]
Remove alpha-bitops.h now that it's unused. It's in the attic if
we want it again.

20 years agoDon't divide the start and end of the chunk by the sector size when
Marcel Moolenaar [Mon, 3 Nov 2003 06:52:30 +0000 (06:52 +0000)]
Don't divide the start and end of the chunk by the sector size when
filling in the GPT entry. Both are already in sector numbers (LBA)
and exactly what we need for the entry. We now write a structurally
correct GPT partitioning.

20 years agoFix two bugs in the calculation of the last LBA of the GPT covered
Marcel Moolenaar [Mon, 3 Nov 2003 06:24:48 +0000 (06:24 +0000)]
Fix two bugs in the calculation of the last LBA of the GPT covered
part of the disk. The first appears to be a typo and instead of
dividing the media size with the sector size, we multiplied. The
second is an off-by-1 error that's the result of mixing up count
and index. The code in question is only applicable for virgin disks
and is used to create the "whole" chunk, which covers only the GPT
usable portion of the disk.

20 years agoRemove bogus RTFREE that was added in rev 1.47. The rmx code operates
Sam Leffler [Mon, 3 Nov 2003 06:11:44 +0000 (06:11 +0000)]
Remove bogus RTFREE that was added in rev 1.47.  The rmx code operates
directly on the radix tree and does not hold any routing table refernces.
This fixes the reference counting problems that manifested itself as a
panic during unmount of filesystems that were mounted by NFS over an
interface that had been removed.

Supported by: FreeBSD Foundation

20 years agoBuild lint(1) on ia64.
Marcel Moolenaar [Mon, 3 Nov 2003 05:14:47 +0000 (05:14 +0000)]
Build lint(1) on ia64.

20 years agoThe size of a long double is 16 bytes, not 12 bytes.
Marcel Moolenaar [Mon, 3 Nov 2003 05:09:57 +0000 (05:09 +0000)]
The size of a long double is 16 bytes, not 12 bytes.

20 years agoAdd a bogus definition of __va_list for use by lint. Make it visible
Marcel Moolenaar [Mon, 3 Nov 2003 05:04:09 +0000 (05:04 +0000)]
Add a bogus definition of __va_list for use by lint. Make it visible
only when lint is defined to protect builds with non-GNU compilers.

20 years agoUse VOP_UNLOCK/vrele instead of vput. td was erecived as a parameter
Alexander Kabaev [Mon, 3 Nov 2003 04:46:19 +0000 (04:46 +0000)]
Use VOP_UNLOCK/vrele instead of vput. td was erecived as a parameter
and one cannot be sure it is equal to curthread.

20 years agoBuild zzz(8) on ia64. It first checks ACPI before it checks for APM.
Marcel Moolenaar [Mon, 3 Nov 2003 04:25:55 +0000 (04:25 +0000)]
Build zzz(8) on ia64. It first checks ACPI before it checks for APM.
It will therefore always suspend using ACPI.

20 years ago - Remove kseq_find(), we no longer scan other cpu's run queues when we go
Jeff Roberson [Mon, 3 Nov 2003 03:27:22 +0000 (03:27 +0000)]
 - Remove kseq_find(), we no longer scan other cpu's run queues when we go
   idle.  They figure out that we're idle fast enough that the cache pollution
   introduces by scanning their run queue is more expensive than waiting
   a little longer.
 - Add kseq_setidle() to mark us as being idle.  Use this in place of
   kseq_find().
 - Remove kseq_load_highest(), kseq_find() was the only consumer of this
   interface.  kseq_balance() has it's own customized version that finds the
   lowest and highest loads simultaneously.

Continuously told that this would be faster by: terry

20 years agoCorrect rev 1.56 which (incorrectly) reversed the test used to
Sam Leffler [Mon, 3 Nov 2003 03:22:39 +0000 (03:22 +0000)]
Correct rev 1.56 which (incorrectly) reversed the test used to
decide if in_pcbpurgeif0 should be invoked.

Supported by: FreeBSD Foundation

20 years agoRewrite Write_Disk() so that it creates a GPT. Note that the code is
Marcel Moolenaar [Mon, 3 Nov 2003 03:18:34 +0000 (03:18 +0000)]
Rewrite Write_Disk() so that it creates a GPT. Note that the code is
basicly untested, but the guts is all there. I need to free up a disk
before I give it a spin.

20 years agoia64 tenderbox hates what I did wrt a printf, so cast the snot out of
Warner Losh [Mon, 3 Nov 2003 01:30:03 +0000 (01:30 +0000)]
ia64 tenderbox hates what I did wrt a printf, so cast the snot out of
it and hope for the best.

20 years ago - Increase the scope of the source object lock in vm_map_copy_entry().
Alan Cox [Mon, 3 Nov 2003 00:59:54 +0000 (00:59 +0000)]
 - Increase the scope of the source object lock in vm_map_copy_entry().

20 years agoBoth CC and CX versions of the 556 exist
Warner Losh [Mon, 3 Nov 2003 00:46:08 +0000 (00:46 +0000)]
Both CC and CX versions of the 556 exist

20 years agoSeparate out the old 3CXEM556 models (10Mbps only) from the faster
Warner Losh [Sun, 2 Nov 2003 23:30:22 +0000 (23:30 +0000)]
Separate out the old 3CXEM556 models (10Mbps only) from the faster
3CCFEM556 models (10/100Mbps).  I have only the former, and none of
the latter, so document that the latter might not work.

20 years agoConstify a few more strings that were hidden behind XE_DEBUG define.
Scott Mitchell [Sun, 2 Nov 2003 23:29:33 +0000 (23:29 +0000)]
Constify a few more strings that were hidden behind XE_DEBUG define.
Remove XE_VENDOR_ID_* defines in favour of PCMCIA_VENDOR_*.

Reviewed by: imp (mentor)

20 years agoChange the buffer length test in NEEDSP() so that it does not
Tim J. Robbins [Sun, 2 Nov 2003 23:20:24 +0000 (23:20 +0000)]
Change the buffer length test in NEEDSP() so that it does not
subtract one unsigned number from another potentially smaller
one, leading to wraparound (and heap corruption, eventually).

PR: 58813
MFC after: 2 weeks

20 years agoSet exit code to 1 in case at least one of the input files
Stefan Eßer [Sun, 2 Nov 2003 23:12:08 +0000 (23:12 +0000)]
Set exit code to 1 in case at least one of the input files
could not be opened.

20 years ago - Increase the scope of two vm object locks in vm_object_split().
Alan Cox [Sun, 2 Nov 2003 22:52:42 +0000 (22:52 +0000)]
 - Increase the scope of two vm object locks in vm_object_split().

20 years agoFix burning of CD's that got broken by the GEOM'ification.
Søren Schmidt [Sun, 2 Nov 2003 22:24:47 +0000 (22:24 +0000)]
Fix burning of CD's that got broken by the GEOM'ification.

GEOM was not designed to handle media that does not have
a size. Blank CD's are of that type, so cheat and set the
media size to -1. This allows burning to work, but makes
GEOM issue outofrange reads that makes the ATAPI subsystem
spew out a few warnings. GEOM should be tought about this.

GEOM was not designed to handle changing the sectorsize
between opens. Writing multitack CD's with both audio and
data tracks needs to change sector size on the fly. We
cheat here and stuff the current sectorsize into GEOM
private internals. GEOM should grow some clean way for this.

20 years agoUse a better way to take NO_LPR into account, and correctly check the
Mark Murray [Sun, 2 Nov 2003 22:13:36 +0000 (22:13 +0000)]
Use a better way to take NO_LPR into account, and correctly check the
NOCRYPT macro when looking at crypto.

Reminded by: ru

20 years agoOn ATA control commands return the registers in the request.
Søren Schmidt [Sun, 2 Nov 2003 22:04:53 +0000 (22:04 +0000)]
On ATA control commands return the registers in the request.

20 years agoRemoved (after the env(1)/printenv(1) split) extraneous arguments
Ruslan Ermilov [Sun, 2 Nov 2003 21:46:11 +0000 (21:46 +0000)]
Removed (after the env(1)/printenv(1) split) extraneous arguments
to the .Nm macro.

20 years agoCorrect the spelling of m_devget() in the sysctl variable description.
Ruslan Ermilov [Sun, 2 Nov 2003 21:43:29 +0000 (21:43 +0000)]
Correct the spelling of m_devget() in the sysctl variable description.

20 years ago - Introduce and use vm_object_reference_locked(). Unlike
Alan Cox [Sun, 2 Nov 2003 21:30:10 +0000 (21:30 +0000)]
 - Introduce and use vm_object_reference_locked().  Unlike
   vm_object_reference(), this function must not be used to reanimate dead
   vm objects.  This restriction simplifies locking.

Reviewed by: tegge

20 years agoOnly the 3C1 is now known not to work. All my other cards work with
Warner Losh [Sun, 2 Nov 2003 20:20:28 +0000 (20:20 +0000)]
Only the 3C1 is now known not to work.  All my other cards work with
this driver.  I've not tested the '556B or the '556BI cards, however,
as I don't own them.

20 years agoMFp4:
Warner Losh [Sun, 2 Nov 2003 20:18:19 +0000 (20:18 +0000)]
MFp4:
o Fix MFC cards.  We were bogusly setting CCR_IOBASE[01] and CCR_IOLIMIT.
  now when we activate the resource, we adjust these for MFC cards, per the
  spec.
o Change type of pf_mfc_* to be bus_addr_t, which is more correct than
  long.

This makes my 3C362D/3C363D and 3CXEM556 cards work!  Woo Hoo!

20 years agoMFp4:
Warner Losh [Sun, 2 Nov 2003 20:13:39 +0000 (20:13 +0000)]
MFp4:
o Remove redundant $FreeBSD$
o Better comments about ep_get_macaddr.
o remove one tab in a switch statement (style only)
o Recognize ID 0x0035 as the device ID for the 3CXEM556 that I have.  This
  makes the 3CXEM556 work for me.  Not 100% sure this is the assigned ID,
  as I don't have the datasheets for this part, but it does work and get
  the correct ethrnet address.
o Comment about the whole fake IRQ 3 thing.  some need it, some don't, all
  work with it.

20 years agoMFp4:
Warner Losh [Sun, 2 Nov 2003 20:10:09 +0000 (20:10 +0000)]
MFp4:
o Minor tweak to error processing in ep_alloc.

20 years agoMFP4:
Warner Losh [Sun, 2 Nov 2003 20:08:58 +0000 (20:08 +0000)]
MFP4:
o define EP_W2_ALT_EEPROM.  we don't use it yet, but may soon.

20 years agorename variables.
Hajimu UMEMOTO [Sun, 2 Nov 2003 19:09:29 +0000 (19:09 +0000)]
rename variables.

Obtained from: KAME

20 years agoSince we do not honour the Confidential: field, and as an attempt to
Ceri Davies [Sun, 2 Nov 2003 17:35:33 +0000 (17:35 +0000)]
Since we do not honour the Confidential: field, and as an attempt to
decrease the number of misfiled PR submissions in general, do not
allow users to send a PR that contains errors.

Discussed with: core

20 years agocorrect %d/%u mismatch.
Hajimu UMEMOTO [Sun, 2 Nov 2003 12:28:04 +0000 (12:28 +0000)]
correct %d/%u mismatch.

Obtained from: KAME

20 years agoexit(3) with negative value does not make sense'
Hajimu UMEMOTO [Sun, 2 Nov 2003 11:43:07 +0000 (11:43 +0000)]
exit(3) with negative value does not make sense'

Obtained from: KAME

20 years agomake debugging with "setkey -x" a lot easier.
Hajimu UMEMOTO [Sun, 2 Nov 2003 11:26:42 +0000 (11:26 +0000)]
make debugging with "setkey -x" a lot easier.

Obtained from: KAME

20 years ago - Remove the ksq_loads[] array. We are only interested in three counts,
Jeff Roberson [Sun, 2 Nov 2003 10:56:48 +0000 (10:56 +0000)]
 - Remove the ksq_loads[] array.  We are only interested in three counts,
   the total load, the timeshare load, and the number of threads that can
   be migrated to another cpu.  Account for these seperately.
 - Introduce a KSE_CAN_MIGRATE() macro which determines whether or not a KSE
   can be migrated to another CPU.  Currently, this only checks to see if
   we're an interrupt handler.  Eventually this will also be used to support
   CPU binding.

20 years agoMake sure we get all user-written input. This simplifies the
Mark Murray [Sun, 2 Nov 2003 10:55:16 +0000 (10:55 +0000)]
Make sure we get all user-written input. This simplifies the
code considerably.

Submitted by: (forgotten)
[I'll happily acknowledge the submitter if he owns up!]

20 years agocleanup secasvar management; refcnt is key.c responsibility;
Hajimu UMEMOTO [Sun, 2 Nov 2003 10:49:47 +0000 (10:49 +0000)]
cleanup secasvar management; refcnt is key.c responsibility;
alloc/free is keydb.c responsibility.

Obtained from: KAME

20 years agoConvert the Big5, EUC, MSKanji and UTF-8 encoding methods to implement
Tim J. Robbins [Sun, 2 Nov 2003 10:09:33 +0000 (10:09 +0000)]
Convert the Big5, EUC, MSKanji and UTF-8 encoding methods to implement
mbrtowc() and wcrtomb() directly. GB18030, GBK and UTF2 are left
unconverted; GB18030 will be done eventually, but GBK and UTF2 may just
be removed, as they are subsets of GB18030 and UTF-8 respectively.

20 years agoRemove headers copied from i386 and either useless or wrong on ia64.
Marcel Moolenaar [Sun, 2 Nov 2003 09:19:07 +0000 (09:19 +0000)]
Remove headers copied from i386 and either useless or wrong on ia64.
An example of useless is bios.h. An example of wrong is msdos.h (due
to the use of long for 32-bit fields).

display.h cannot be removed because it's used by syscons. That header
however has no platform dependency and shouldn't really be here.

Removal if these headers may cause build failures in the ports tree.
It's the ports that need fixing in that case.

Tested with: buildworld, LINT

20 years ago- do not quit from key_sendup() even if writes to non-target
Hajimu UMEMOTO [Sun, 2 Nov 2003 09:13:33 +0000 (09:13 +0000)]
- do not quit from key_sendup() even if writes to non-target
  socket fails.
- remove an unneeded function.
- fix pfkey stat.
- fix comment.

Obtained from: KAME

20 years agoo Add PART_EFI so that we can use it instead of PART_FAT on ia64
Marcel Moolenaar [Sun, 2 Nov 2003 08:58:57 +0000 (08:58 +0000)]
o  Add PART_EFI so that we can use it instead of PART_FAT on ia64
   to better deal with the fact that we need an EFI partition and
   that we need to have a mountpoint for it.
o  When creating a new partition, add EFI to the list of types
   the user can select from. This makes it easy to create an EFI.
o  Do not include wizard.c on ia64.
o  The user cannot create a partition on ia64 that's a multiple of
   the cylinder size. We don't have a notion of cyclinders.

20 years agoTurn Write_Disk() into a stub for now. It needs to be rewritten to
Marcel Moolenaar [Sun, 2 Nov 2003 08:41:18 +0000 (08:41 +0000)]
Turn Write_Disk() into a stub for now. It needs to be rewritten to
write out a GPT and not a MBR.

20 years agoo Move Int_Open_Disk() from disk.c to open_disk.c for use by all
Marcel Moolenaar [Sun, 2 Nov 2003 08:39:08 +0000 (08:39 +0000)]
o  Move Int_Open_Disk() from disk.c to open_disk.c for use by all
   platforms except ia64 and use Int_Open_Disk() in open_ia64_disk.c
   on ia64. We need to know more than GEOM can provide us so we're
   forced to read from the disk. Move uuid_type() to open_ia64_disk.c
   and remove all references on non-ia64.
o  Pass the GEOM conftxt to Int_Open_Disk() so that only Open_Disk()
   needs to know about GEOM and libdisk can more easily be used with
   media not handled by GEOM.
o  Create an ia64 specific definiton of struct disk on ia64, because
   we don't need/have most of the fields other platforms need and
   other fields not applicable on platforms other than ia64.
o  Do not compile change.c on ia64. It's too PC specific.
o  In Fixup_Names() in create_chunk.c, try all partition numbers
   that are valid for the GPT disk. We have the total number of
   partitions that can be allocated in the disk structure on ia64.
   Also, use the GPT partition naming if we're creating one under
   a chunk of type "whole". It's a GPT partition in that case.
o  In Create_Chunk(), compile-out the PC specific code on ia64 that
   checks BIOS geometry restrictions.
o  In Debug_Disk() in disk.c, dump the ia64 specific fields.
o  Save the partition index in the chunk on ia64 so that we can
   preserve it when we write the data back to disk. This avoids that
   partitions get moved around or swapped after installing FreeBSD,
   which may render a disk unusable.

20 years agoDo not fill in d_ncylinders, d_ntracks and d_nsectors in the disklabel
Marcel Moolenaar [Sun, 2 Nov 2003 08:07:25 +0000 (08:07 +0000)]
Do not fill in d_ncylinders, d_ntracks and d_nsectors in the disklabel
on ia64. Not only do we not have a disklabel by default, we also do
not have a notion of cylinders, tracks and sectors.

20 years agoCompile-out support for allfreebsd, dedicate, sanitize and bios on
Marcel Moolenaar [Sun, 2 Nov 2003 08:01:18 +0000 (08:01 +0000)]
Compile-out support for allfreebsd, dedicate, sanitize and bios on
ia64. These commands are PC specific and not supported.

20 years agoTurn Track_Aligned(), Prev_Track_Aligned(), Next_Track_Aligned(),
Marcel Moolenaar [Sun, 2 Nov 2003 07:58:19 +0000 (07:58 +0000)]
Turn Track_Aligned(), Prev_Track_Aligned(), Next_Track_Aligned(),
Cyl_Aligned(), Prev_Cyl_Aligned() and Next_Cyl_Aligned() into
tautologies on ia64. GPT removes all notion of tracks, heads and
sectors per track, so there are no alignment considerations.

20 years agoAdd an additional check to the tcp_twrecycleable function; I had
Mike Silbersack [Sun, 2 Nov 2003 07:47:03 +0000 (07:47 +0000)]
Add an additional check to the tcp_twrecycleable function; I had
previously only considered the send sequence space.  Unfortunately,
some OSes (windows) still use a random positive increments scheme for
their syn-ack ISNs, so I must consider receive sequence space as well.

The value of 250000 bytes / second for Microsoft's ISN rate of increase
was determined by testing with an XP machine.

20 years agoo In Print_Chunk(), don't print the address of the chunk on ia64. It
Marcel Moolenaar [Sun, 2 Nov 2003 07:44:59 +0000 (07:44 +0000)]
o  In Print_Chunk(), don't print the address of the chunk on ia64. It
   doesn't have any meaning and only results in lines longer than 80
   characters.
o  In Delete_Chunk2(), also look for chunks of type "part" under
   chunks of type "whole" on ia64. They're not only under chunks of
   type "freebsd" there.

20 years agoTrim more devices from the hardware list: ida(4), ciss(4), iir(4),
Bruce A. Mah [Sun, 2 Nov 2003 07:41:03 +0000 (07:41 +0000)]
Trim more devices from the hardware list:  ida(4), ciss(4), iir(4),
ie(4).

20 years agoDNS should not necessarily be named(8), tweak the comment a bit.
Ruslan Ermilov [Sun, 2 Nov 2003 07:31:44 +0000 (07:31 +0000)]
DNS should not necessarily be named(8), tweak the comment a bit.

20 years agoMFV: The -d option.
Ruslan Ermilov [Sun, 2 Nov 2003 07:27:40 +0000 (07:27 +0000)]
MFV: The -d option.

20 years agoSort encodings.
Ruslan Ermilov [Sun, 2 Nov 2003 07:25:24 +0000 (07:25 +0000)]
Sort encodings.

20 years agoStyle.
Ruslan Ermilov [Sun, 2 Nov 2003 06:47:39 +0000 (06:47 +0000)]
Style.

20 years agoThe '574 family works (I had a bad dongle which is why I thought it
Warner Losh [Sun, 2 Nov 2003 05:33:12 +0000 (05:33 +0000)]
The '574 family works (I had a bad dongle which is why I thought it
might not be).
The modem combo cards do not (confirmed today)

20 years agoTake care not to call vput if thread used in corresponding vget
Alexander Kabaev [Sun, 2 Nov 2003 04:52:53 +0000 (04:52 +0000)]
Take care not to call vput if thread used in corresponding vget
wasn't curthread, i.e. when we receive a thread pointer to use
as a function argument. Use VOP_UNLOCK/vrele in these cases.

The only case there td != curthread known at the moment is
boot() calling sync with thread0 pointer.

This fixes the panic on shutdown people have reported.

20 years ago - In sched_prio() only force us onto the current queue if our priority is
Jeff Roberson [Sun, 2 Nov 2003 04:25:59 +0000 (04:25 +0000)]
 - In sched_prio() only force us onto the current queue if our priority is
   being elevated (numerically smaller).

20 years ago - Rename SCHED_PRI_NTHRESH to SCHED_SLICE_NTHRESH since it is only used in
Jeff Roberson [Sun, 2 Nov 2003 04:10:15 +0000 (04:10 +0000)]
 - Rename SCHED_PRI_NTHRESH to SCHED_SLICE_NTHRESH since it is only used in
   slice assignment.  Add a comment describing what it does.
 - Remove a stale XXX comment, the nice should not impact the interactivity,
   nice adjustments only effect non-interactive tasks in ULE.
 - Don't allow nice -20 tasks to totally starve nice 0 tasks.  Give them at
   least SCHED_SLICE_MIN ticks.  We still allow nice 0 tasks to starve nice
   +20 tasks as intended.

20 years ago - Remove PRIO_TOTAL.
Jeff Roberson [Sun, 2 Nov 2003 03:50:16 +0000 (03:50 +0000)]
 - Remove PRIO_TOTAL.

Submitted by: bde

20 years ago - Remove uses of PRIO_TOTAL and replace them with SCHED_PRI_NRESV
Jeff Roberson [Sun, 2 Nov 2003 03:49:32 +0000 (03:49 +0000)]
 - Remove uses of PRIO_TOTAL and replace them with SCHED_PRI_NRESV
 - SCHED_PRI_NRESV does not have the off by one error in PRIO_TOTAL so we
   do not have to account for it in the few places that we use it.

Requested by: bde

20 years ago - Change sched_interact_update() to only accept slp+runtime values between
Jeff Roberson [Sun, 2 Nov 2003 03:36:33 +0000 (03:36 +0000)]
 - Change sched_interact_update() to only accept slp+runtime values between
   0 and SCHED_SLP_RUN_MAX * 2.  This allows us to simplify the algorithm
   quite a bit.  Before, it dealt with arbitrary values which required us
   to do nasty integer division tricks that didn't quite work out correctly.
 - Chnage sched_wakeup() to detect conditions where the slp+runtime could
   exceed SCHED_SLP_RUN_MAX * 2.  This can happen if we go to sleep for
   longer than 6 seconds.  In this case, we'll just clear the runtime and
   set the sleep time to the max.
 - Define a new function, sched_interact_fork() which updates the slp+runtime
   of a newly forked thread.  We want to limit the amount of history retained
   from the parent so that we learn the child's behavior quickly.  We don't,
   however want to decay it to nothing.  Previously, we would simply divide
   each parameter by 100 whenever we forked.  After a few forks the values
   would reach 0 and tasks would not be considered interactive.
 - Add another KTR entry, cleanup some existing entries.
 - Remove a useless sched_interact_update() from sched_priority().  This is
   already done by the callers that require it.

20 years agoDo if_xname conversion. I missed this one because it's pc98 only.
Brooks Davis [Sat, 1 Nov 2003 23:24:38 +0000 (23:24 +0000)]
Do if_xname conversion.  I missed this one because it's pc98 only.

20 years ago - Increase the scope of two vm object locks in vm_object_collapse().
Alan Cox [Sat, 1 Nov 2003 23:06:41 +0000 (23:06 +0000)]
 - Increase the scope of two vm object locks in vm_object_collapse().
 - Remove the acquisition and release of Giant from vm_object_coalesce().

20 years agoo Do not define WITH_SYSCONS and WITH_MICE on ia64. We cannot have
Marcel Moolenaar [Sat, 1 Nov 2003 20:17:11 +0000 (20:17 +0000)]
o  Do not define WITH_SYSCONS and WITH_MICE on ia64. We cannot have
   a generic kernel with syscons.
o  Do not define WITH_SLICES on ia64. We only label because we're
   going to create GPT disks.

20 years agoo Compile-out "wizard" mode on ia64.
Marcel Moolenaar [Sat, 1 Nov 2003 20:14:06 +0000 (20:14 +0000)]
o  Compile-out "wizard" mode on ia64.
o  Also allow swap and filesystem partitions outside a freebsd slice.
   This is typically the case for GPT.
o  Allow chunks of type "whole" to be displayed at the top. This is
   to allow a GPT disk to be labeled. We need a slice out of which we
   can make partitions, but a GPT disk doesn't have slices. For GPT
   disks a chunk of type "whole" can then be used as a placeholder.

20 years agoo Compile-out "wizard" mode on ia64.
Marcel Moolenaar [Sat, 1 Nov 2003 20:04:12 +0000 (20:04 +0000)]
o  Compile-out "wizard" mode on ia64.
o  Do not set bootblocks on ia64. It's not even a functionality in
   libdisk on ia64.

20 years agoRevert the first part of my previous change.
Tom Rhodes [Sat, 1 Nov 2003 16:57:19 +0000 (16:57 +0000)]
Revert the first part of my previous change.

Requested by: wollman

20 years agoThe copywrite is not a 'static char', remove the #ifdefs and move the copywrite up
Tom Rhodes [Sat, 1 Nov 2003 15:58:06 +0000 (15:58 +0000)]
The copywrite is not a 'static char', remove the #ifdefs and move the copywrite up
into the commented out 'copywrite' section.

Include sys/linker.h for kldload(3).

20 years agoRemove now unused variable.
Alexander Kabaev [Sat, 1 Nov 2003 15:04:50 +0000 (15:04 +0000)]
Remove now unused variable.

20 years agoMake the cardbus driver a derived class of the pci driver. In theory, this
Doug Rabson [Sat, 1 Nov 2003 12:45:03 +0000 (12:45 +0000)]
Make the cardbus driver a derived class of the pci driver. In theory, this
should allow many of the pci methods to be re-staticised.

20 years agoGEOM'ify atapi-cd. Original patch by phk, subtle changes by me.
Søren Schmidt [Sat, 1 Nov 2003 09:44:33 +0000 (09:44 +0000)]
GEOM'ify atapi-cd. Original patch by phk, subtle changes by me.

20 years agoShorten the code by removing one "do-nothing" function, replacing it
Mark Murray [Sat, 1 Nov 2003 09:31:54 +0000 (09:31 +0000)]
Shorten the code by removing one "do-nothing" function, replacing it
with nullop(), which is in kern_conf.c.

20 years agoFix cable detection on AMD chips.
Søren Schmidt [Sat, 1 Nov 2003 09:30:15 +0000 (09:30 +0000)]
Fix cable detection on AMD chips.

20 years ago - Modify swap_pager_copy() and its callers such that the source and
Alan Cox [Sat, 1 Nov 2003 08:57:26 +0000 (08:57 +0000)]
 - Modify swap_pager_copy() and its callers such that the source and
   destination objects are locked on entry and exit.  Add comments to
   the callers noting that the locks can be released by swap_pager_copy().
 - Remove several instances of GIANT_REQUIRED.

20 years agoFixed breakage of my world in rev.1.30. #include <sys/time.h> instead of
Bruce Evans [Sat, 1 Nov 2003 08:43:54 +0000 (08:43 +0000)]
Fixed breakage of my world in rev.1.30.  #include <sys/time.h> instead of
depending on namespace pollution in <sys/stat.h> for the declarations of
struct timeval and utimes().

Fixed some style bugs in rev.1.30 and some nearby style bugs (mainly
unsorting and missing or extra blank lines).

Removed a wrong comment that was obtained from NetBSD in rev.1.14.  It said
that chflags() reset the times that were set "above" by utimes(), but
utimes wasn't "above" in FreeBSD until rev.1.30, and chflags() does't
actually reset the times.

20 years agoRemove TODO comment about creating a macro version of towctrans().
Tim J. Robbins [Sat, 1 Nov 2003 08:20:58 +0000 (08:20 +0000)]
Remove TODO comment about creating a macro version of towctrans().
Remove unnecessary inclusion of <ctype.h>.

20 years agoFix a typo that caused the optimized single-byte locale path not to be taken.
Tim J. Robbins [Sat, 1 Nov 2003 08:18:18 +0000 (08:18 +0000)]
Fix a typo that caused the optimized single-byte locale path not to be taken.

20 years ago- Add a new function tcp_twrecycleable, which tells us if the ISN which
Mike Silbersack [Sat, 1 Nov 2003 07:30:08 +0000 (07:30 +0000)]
- Add a new function tcp_twrecycleable, which tells us if the ISN which
we will generate for a given ip/port tuple has advanced far enough
for the time_wait socket in question to be safely recycled.

- Have in_pcblookup_local use tcp_twrecycleable to determine if
time_Wait sockets which are hogging local ports can be safely
freed.

This change preserves proper TIME_WAIT behavior under normal
circumstances while allowing for safe and fast recycling whenever
ephemeral port space is scarce.

20 years agoFix a few style glitches in the previous commit and make the
Mike Silbersack [Sat, 1 Nov 2003 07:06:04 +0000 (07:06 +0000)]
Fix a few style glitches in the previous commit and make the
tunable error message more brief.

Suggested by: bde

20 years agoMore trimmage of the hardware notes: sf(4), tx(4), wb(4), nge(4),
Bruce A. Mah [Sat, 1 Nov 2003 06:08:52 +0000 (06:08 +0000)]
More trimmage of the hardware notes:  sf(4), tx(4), wb(4), nge(4),
ste(4), sk(4), le(4), lge(4), txp(4), bge(4).

These are all Ethernet drivers whose manual pages already list
specific supported devices.

20 years agoTemporarily undo parts of the stuct mount locking commit by jeff.
Alexander Kabaev [Sat, 1 Nov 2003 05:51:54 +0000 (05:51 +0000)]
Temporarily undo parts of the stuct mount locking commit by jeff.
It is unsafe to hold a mutex across vput/vrele calls.

This will be redone when a better locking strategy is agreed upon.

Discussed with: jeff

20 years agoAdd a manual page for the utrace() system call.
Tim J. Robbins [Sat, 1 Nov 2003 05:36:14 +0000 (05:36 +0000)]
Add a manual page for the utrace() system call.

Obtained from: NetBSD

20 years agoAllow mbrtowc() and wcrtomb() to be implemented directly, instead of
Tim J. Robbins [Sat, 1 Nov 2003 05:13:13 +0000 (05:13 +0000)]
Allow mbrtowc() and wcrtomb() to be implemented directly, instead of
as wrappers around the deprecated 4.4BSD rune functions. This paves the
way for state-dependent encodings, which the rune API does not support.
- Add __emulated_sgetrune() and __emulated_sputrune(), which are
  implementations of sgetrune() and sputrune() in terms of
  mbrtowc() and wcrtomb().
- Rename the old rune-wrapper mbrtowc() and wcrtomb() functions to
  __emulated_mbrtowc() and __emulated_wcrtomb().
- Add __mbrtowc and __wcrtomb function pointers, which point to the
  current locale's conversion functions, or the __emulated versions.
- Implement mbrtowc() and wcrtomb() as calls to these function pointers.
- Make the "NONE" encoding implement mbrtowc() and wcrtomb() directly.

All of this emulation mess will be removed, together with rune support,
in FreeBSD 6.

20 years ago - Additional vm object locking in vm_object_split()
Alan Cox [Sat, 1 Nov 2003 04:54:23 +0000 (04:54 +0000)]
 - Additional vm object locking in vm_object_split()
 - New vm object locking assertions in vm_page_insert() and
   vm_object_set_writeable_dirty()

20 years agoMore trimming of specific device instances from the hardware notes to
Bruce A. Mah [Sat, 1 Nov 2003 04:41:32 +0000 (04:41 +0000)]
More trimming of specific device instances from the hardware notes to
reduce duplication of information from the manual pages: xl(4), cs(4),
ath(4), sbni(4), lmc(4).

20 years agoDo not bother walking mount point vnode list just to calculate
Alexander Kabaev [Sat, 1 Nov 2003 04:36:50 +0000 (04:36 +0000)]
Do not bother walking mount point vnode list just to calculate
the number of vnodes. Use precomputed mp->mnt_nvnodelistsize
value instead.

20 years agoo check hal ABI version to catch driver-HAL mismatches
Sam Leffler [Sat, 1 Nov 2003 03:37:33 +0000 (03:37 +0000)]
o check hal ABI version to catch driver-HAL mismatches
o print MAC, PHY, and radio h/w revisions at attach

20 years agoThis commit was generated by cvs2svn to compensate for changes in r121837,
Sam Leffler [Sat, 1 Nov 2003 03:19:51 +0000 (03:19 +0000)]
This commit was generated by cvs2svn to compensate for changes in r121837,
which included commits to RCS files with non-trunk default branches.

20 years agoVersion 0.9.5.17:
Sam Leffler [Sat, 1 Nov 2003 03:19:51 +0000 (03:19 +0000)]
Version 0.9.5.17:

o change os glue API to be compatible with Linux so hal.o's can
  be used on any system
o add ABI version to catch driver-HAL mismatches
o move hal version information from ah_osdep.c to binary component
o remove ath_hal_wait os glue component
o assign constant values to all enums to avoid potential compiler
  incompatibilities
o add support for 3Com badged cards (PCI vendor ID)
o add support for IBM mini-pci cards (PCI device ID)
o expose MAC, PHY, and radio hardware revisions
o support for big-endian platforms
o new method to set slot time in us
o bug fix for 5211: beacon timers not setup correctly
o bug fix for 5212: don't crash when handed a 5112 radio

20 years agoCopy cur's snap_time to last when refreshing statistics. Fixes problem
Tim J. Robbins [Sat, 1 Nov 2003 02:06:02 +0000 (02:06 +0000)]
Copy cur's snap_time to last when refreshing statistics. Fixes problem
where MB/s and tps statistics would always be zero, presumably because
they were being averaged out over the time between now and when the
system booted instead of a few seconds.

PR: 58683

20 years agoFix the logic to match the new name of the tunable.
Nate Lawson [Sat, 1 Nov 2003 01:05:53 +0000 (01:05 +0000)]
Fix the logic to match the new name of the tunable.

Pointed out by: iwasaki

20 years agoTrim the hifn(4) device instances from the hardware list.
Bruce A. Mah [Sat, 1 Nov 2003 00:57:33 +0000 (00:57 +0000)]
Trim the hifn(4) device instances from the hardware list.

Add an entry for safe(4) crypto accelerators.

20 years agoChange the reset video option to be positive (hw.acpi.reset_video).
Nate Lawson [Sat, 1 Nov 2003 00:18:29 +0000 (00:18 +0000)]
Change the reset video option to be positive (hw.acpi.reset_video).
Requested by: jhb

Initialize the real mode stack.  This is needed at least for the return
address from the lcall.
Requested by: takawata

Fix style bugs in acpi_wakecode.S
Requested by: bde

Remove the kernel option now that we have the tunable.

20 years agoRemove a bogus PCI ID entry.
Scott Long [Sat, 1 Nov 2003 00:13:43 +0000 (00:13 +0000)]
Remove a bogus PCI ID entry.

20 years agoWe want the length of the string, not the size of its pointer.
Brooks Davis [Sat, 1 Nov 2003 00:03:20 +0000 (00:03 +0000)]
We want the length of the string, not the size of its pointer.

20 years agoCast the NULL to a pointer type for 64 bit architectures
Gregory Neil Shapiro [Fri, 31 Oct 2003 22:12:09 +0000 (22:12 +0000)]
Cast the NULL to a pointer type for 64 bit architectures

Submitted by: harti
MFC after: 4 days