]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agofix build w/ IEEE80211_DEBUG_REFCNT enabled
Sam Leffler [Thu, 1 May 2008 20:26:25 +0000 (20:26 +0000)]
fix build w/ IEEE80211_DEBUG_REFCNT enabled

16 years agoAdd a separate tree to track arena chunks that contain dirty pages.
Jason Evans [Thu, 1 May 2008 17:25:55 +0000 (17:25 +0000)]
Add a separate tree to track arena chunks that contain dirty pages.
This substantially improves worst case allocation performance, since
O(lg n) tree search can be used instead of O(n) tree iteration.

Use rb_wrap() instead of directly calling rb_*() macros.

16 years agoAdd rb_wrap(), which creates C function wrappers for most rb_*()
Jason Evans [Thu, 1 May 2008 17:24:37 +0000 (17:24 +0000)]
Add rb_wrap(), which creates C function wrappers for most rb_*()
macros.

Add rb_foreach_next() and rb_foreach_reverse_prev(), which make it
possible to re-synchronize tree iteration after the tree has been
modified.

Rename rb_tree_new() to rb_new().

16 years agoUse the correct bit when trying to force an interrupt through the HCC reg.
Bjoern A. Zeeb [Thu, 1 May 2008 13:10:03 +0000 (13:10 +0000)]
Use the correct bit when trying to force an interrupt through the HCC reg.
It's not a problem as this is a #ifdef notyet.

16 years agoo OpenBSD 4.3 added.
Maxim Konovalov [Thu, 1 May 2008 09:52:34 +0000 (09:52 +0000)]
o OpenBSD 4.3 added.

16 years agoDo not call ndis_setstate_80211() until we are ready to associate, the vap may
Andrew Thompson [Thu, 1 May 2008 05:11:33 +0000 (05:11 +0000)]
Do not call ndis_setstate_80211() until we are ready to associate, the vap may
not have been created yet and will panic. This requires ndis_scan() to always
set the SSID.

Reported by: Ben Kaduk

16 years agoUnify all the wifi *_ioctl routines
Andrew Thompson [Thu, 1 May 2008 04:55:00 +0000 (04:55 +0000)]
Unify all the wifi *_ioctl routines
 - Limit grabbing the lock to SIOCSIFFLAGS.
 - Move ieee80211_start_all() to SIOCSIFFLAGS.
 - Remove SIOCSIFMEDIA as it is not useful.
 - Limit ether_ioctl to only SIOCGIFADDR. SIOCSIFADDR and SIOCSIFMTU have no
   affect as there is no input/output path in the vap parent.  The vap code
   will handle the reinit of the mac address changes.
 - Split off ndis_ioctl_80211 as it was getting too different to wired devices.

This fixes a copyout while locked and a lock recursion.

Reviewed by: sam

16 years agore-enable WME by default; after a full day of testing on iwi I see no
Sam Leffler [Thu, 1 May 2008 03:49:59 +0000 (03:49 +0000)]
re-enable WME by default; after a full day of testing on iwi I see no
issues and the only way we'll identify them is for people to use it

16 years agoEnable autoflush of output to always show prompts. If piping the output
Sean Farley [Wed, 30 Apr 2008 23:27:02 +0000 (23:27 +0000)]
Enable autoflush of output to always show prompts.  If piping the output
prior to this patch, the prompt would not appear due to buffering.

Approved by: flz

16 years agoAttempt to make the print types more friendly to other architectures.
Julian Elischer [Wed, 30 Apr 2008 20:00:30 +0000 (20:00 +0000)]
Attempt to make the print types more friendly to other architectures.

Prodded by: Max Laier
Help from: BMS, jhb

16 years agoo add indirect array for field indices as the # stats is > 127 and doesn't
Sam Leffler [Wed, 30 Apr 2008 19:47:31 +0000 (19:47 +0000)]
o add indirect array for field indices as the # stats is > 127 and doesn't
  fit in a signed char
o change default output to something more useful for sta mode
o futz w/ various field names and widths; need to do full pass over this stuff

16 years agoupdate for new events
Sam Leffler [Wed, 30 Apr 2008 19:45:32 +0000 (19:45 +0000)]
update for new events

16 years agoAdd support for Microsoft Notebook Optical Mouse 3000 Model 1049.
Kai Wang [Wed, 30 Apr 2008 19:37:54 +0000 (19:37 +0000)]
Add support for Microsoft Notebook Optical Mouse 3000 Model 1049.

Reviewed by: imp
(RELENG-7) Tested by: Oliver Herold <oliver@akephalos.de>
PR: usb/121052
MFC after: 1 month

16 years agoremove old code to handle mcast address changes; this is all done through
Sam Leffler [Wed, 30 Apr 2008 17:00:32 +0000 (17:00 +0000)]
remove old code to handle mcast address changes; this is all done through
net80211 and pushed into the driver through non-ioctl callbacks

16 years agoEmit a warning when the network_interfaces variable is not set to AUTO.
Brooks Davis [Wed, 30 Apr 2008 16:29:15 +0000 (16:29 +0000)]
Emit a warning when the network_interfaces variable is not set to AUTO.

MFC after: 3 days

16 years agodisable default enabling of WME until we resolve driver regressions
Sam Leffler [Wed, 30 Apr 2008 16:05:57 +0000 (16:05 +0000)]
disable default enabling of WME until we resolve driver regressions

16 years ago Make ld use tradmips for output formats since we migrated to it.
Oleksandr Tymoshenko [Wed, 30 Apr 2008 12:44:58 +0000 (12:44 +0000)]
  Make ld use tradmips for output formats since we migrated to it.

  Approved by: cognet (mentor)

16 years agoWhen receiving delegated credentials, initialise our cred's linked list.
Doug Rabson [Wed, 30 Apr 2008 11:29:22 +0000 (11:29 +0000)]
When receiving delegated credentials, initialise our cred's linked list.
Add a bit more sanity checking for GSS-API mechanisms that claim to have
delegated creds but don't actually return a cred handle.

MFC after: 2 weeks

16 years agoUse global implementation of _gss_oid_equal.
Doug Rabson [Wed, 30 Apr 2008 11:27:15 +0000 (11:27 +0000)]
Use global implementation of _gss_oid_equal.

MFC after: 2 weeks

16 years agoAllow null oids in _gss_oid_equal().
Doug Rabson [Wed, 30 Apr 2008 11:25:34 +0000 (11:25 +0000)]
Allow null oids in _gss_oid_equal().

MFC after: 2 weeks

16 years agoEliminate an unused field from the pmap.
Alan Cox [Wed, 30 Apr 2008 06:05:36 +0000 (06:05 +0000)]
Eliminate an unused field from the pmap.

16 years agoDon't panic even if bus_dmamap_load(9) was failed. Just return
Pyun YongHyeon [Wed, 30 Apr 2008 02:49:24 +0000 (02:49 +0000)]
Don't panic even if bus_dmamap_load(9) was failed. Just return
ENOBUFS so callers can reuse previous mbuf.

Submitted by: Oleg  (agile.quad AT gmail DOT com)
MFC after: 1 week

16 years agomp_machdep.c is only conditional upon smp, not aim. If booke grows
Marcel Moolenaar [Wed, 30 Apr 2008 00:50:50 +0000 (00:50 +0000)]
mp_machdep.c is only conditional upon smp, not aim. If booke grows
support for smp, mp_machdep.c needs to be included as well.

16 years agofix build
Sam Leffler [Wed, 30 Apr 2008 00:11:08 +0000 (00:11 +0000)]
fix build

Submitted by: delphij

16 years ago o Add MIPS to the list of architectures with defined TLS_TCB_ALIGN
Oleksandr Tymoshenko [Tue, 29 Apr 2008 23:15:23 +0000 (23:15 +0000)]
  o Add MIPS to the list of architectures with defined TLS_TCB_ALIGN
  o Stick with TLS Variant II for MIPS for the moment.

  Approved by: imp

16 years ago Set QUANTUM_2POW_MIN and SIZEOF_PTR_2POW parameters for MIPS
Oleksandr Tymoshenko [Tue, 29 Apr 2008 22:56:05 +0000 (22:56 +0000)]
  Set QUANTUM_2POW_MIN and SIZEOF_PTR_2POW parameters for MIPS

  Approved by: imp

16 years agoDocument the kproc_kthread_add() call
Julian Elischer [Tue, 29 Apr 2008 22:43:15 +0000 (22:43 +0000)]
Document the kproc_kthread_add() call
and fix a small detail of its implementation.
MFC after: 1 week

16 years agoAdd new directories, scripts and wlanwds. Introduced with vap support.
Rui Paulo [Tue, 29 Apr 2008 22:05:26 +0000 (22:05 +0000)]
Add new directories, scripts and wlanwds. Introduced with vap support.

16 years agoAdd missing 't'.
Rui Paulo [Tue, 29 Apr 2008 22:02:27 +0000 (22:02 +0000)]
Add missing 't'.

Approved by: sam

16 years agodisable HT capabilities until we sort out firmware issues
Sam Leffler [Tue, 29 Apr 2008 21:43:16 +0000 (21:43 +0000)]
disable HT capabilities until we sort out firmware issues

16 years agoLock filedesc exclusively when modifying fd_[cr]dir.
Roman Divacky [Tue, 29 Apr 2008 21:40:11 +0000 (21:40 +0000)]
Lock filedesc exclusively when modifying fd_[cr]dir.
This is probably harmless but it's better to lock it
correctly.

Approved by: kib (mentor)

16 years agoIntel 4965 wireless driver (derived from openbsd driver of the same name)
Sam Leffler [Tue, 29 Apr 2008 21:36:17 +0000 (21:36 +0000)]
Intel 4965 wireless driver (derived from openbsd driver of the same name)

16 years agoAdd an option (compiled out by default)
Julian Elischer [Tue, 29 Apr 2008 21:23:21 +0000 (21:23 +0000)]
Add an option (compiled out by default)
to profile outoing packets for a number of mbuf chain
related parameters
e.g. number of mbufs, wasted space.
probably will do with further work later.

Reviewed by: various

16 years agoFix a comment typo.
Robert Watson [Tue, 29 Apr 2008 21:21:15 +0000 (21:21 +0000)]
Fix a comment typo.

MFC after: 3 days

16 years agoRework how kgdb manages kernel and vmcore files to be a bit more gdb-ish
John Baldwin [Tue, 29 Apr 2008 20:32:45 +0000 (20:32 +0000)]
Rework how kgdb manages kernel and vmcore files to be a bit more gdb-ish
so that kgdb can be used more like a normal gdb:
- Load the kernel via the standard 'exec' target and allow it to be changed
  via the 'file' command.
  - Instead of explicitly loading the kernel file as the mail symbol file
    during startup, just pass it to gdb_main() as the executable file.
  - Change the kld support (via shared libraries) to cache the address of
    the linker_files and linker_kernel_file variables in addition to the
    offsets of various members in 'struct linker_file'.
  - When a new symbol file is loaded, recompute the addresses and offsets
    used by the kld support code.
  - When a new symbol file is loaded, recalculate the ofs_fix variable to
    account for the different ways a trapframe can be passed to trap
    frame handlers in i386.  This is done by adding a MD
    kgdb_trgt_new_objfile() hook that is empty on all but i386.
  - Don't use the directory name of the kernel specified on the command
    line to find kernel modules in the kld support code.  Instead,
    extract the filename of the current executable via exec_bfd.  Now
    the 'kernel' variable is private to main.c again.
  - Make the 'add-kld' command explicitly fail if no executable is loaded.
- Make the support for vmcores a real core-dump target that opens the
  kernel and vmcore on open and closes the kvm connection when closed, etc.
  - The 'core' command can now be used to select a vmcore to use, either
    a crash dump file or /dev/mem for live debugging.
  - The 'detach' command can be used to detach from a vmcore w/o attaching
    to a new one.
  - kgdb no longer explicitly opens a core dump during startup and no longer
    has to use an atexit() hook to close the kvm connection on shutdown.
  - Symbols for kld's are automatically loaded anytime a core is opened.
    Also, the unread portion of dmesg is dumped just as it was done on kgdb
    startup previously.
- Don't require either a remote target or core dump if a kernel is specified.
  You can now just run 'kgdb kernel' similar to running gdb on an executable
  and later connect to a remote target or core dump.
- Use a more relaxed way to verify remote targets specified via -r.
  Instead of explicitly allowing a few non-file target specifications,
  just assume that if stat() on the arg and on "/dev/" + arg both fail
  that is some non-file target and pass it to gdb.
- Don't use a custom interpreter.  The existing kgdb_init() hook and the
  target_new_objfile() hook give us sufficient hooks during startup to
  setup kgdb-specific behavior now.
- Always add the 'proc', 'tid', and 'add-kld' commands on startup and not
  just if we have a core dump.  Currently the 'proc' and 'tid' commands do
  not work for remote targets (I will fix at least 'tid' in the next round
  of changes though).  However, the 'add-kld' command works fine for
  loading symbols for a kernel module on a remote target.
- Always setup the 'kld' shared library target operations instead of just
  if we have a core dump.  Although symbols for kernel modules are not
  automatically loaded when connecting to a remote target, you can do
  'info sharedlibrary' after connecting to the remote target and kgdb will
  find all the modules.  You can then use the 'sharedlibrary' command to
  load symbols from the module files.
- Change kthr_init() to free the existing list of kthr objects before
  generating a new one.  This allows it to be invoked multiple times
  w/o leaking memory.

MFC after: 1 week

16 years agoMore recommendations
Warner Losh [Tue, 29 Apr 2008 19:55:18 +0000 (19:55 +0000)]
More recommendations

16 years agoAdd support for the BCM5906[M] adapters. These adapters only support
John Baldwin [Tue, 29 Apr 2008 19:47:13 +0000 (19:47 +0000)]
Add support for the BCM5906[M] adapters.  These adapters only support
10/100 operation and place the mailbox registers at a different offset.
They also do not have an EEPROM, so the MAC address must be read from
NVRAM instead.

MFC after: 1 month
PR: kern/118975
Submitted by: benjsc, Thomas Nyström  thn at saeab dot se
Submitted by: sephe (original patch for DragonflyBSD)

16 years ago- Add a global variable 'fbsdcoreops_suppress_target' that can be set to
John Baldwin [Tue, 29 Apr 2008 17:47:25 +0000 (17:47 +0000)]
- Add a global variable 'fbsdcoreops_suppress_target' that can be set to
  force the FreeBSD multithreaded core target to not register any target
  for handling core dumps.  This is analogous to the
  'coreops_suppress_target' variable that GDB provides for suppressing the
  default core dump target.  KGDB will use this new variable so it can
  provide its own core dump target that uses libkvm to work with vmcore
  files.
- Adjust the long name and documentation of the FreeBSD multithreaded core
  dump target so it better matches what GDB's core dump target uses.

MFC after: 1 week
Reviewed by: davidxu, marcel

16 years ago_setjmp.o was missing a dependency on "machine".
Ruslan Ermilov [Tue, 29 Apr 2008 17:42:42 +0000 (17:42 +0000)]
_setjmp.o was missing a dependency on "machine".

16 years agoDon't forget to clean the "machine" symlink on amd64, otherwise bad
Ruslan Ermilov [Tue, 29 Apr 2008 17:37:01 +0000 (17:37 +0000)]
Don't forget to clean the "machine" symlink on amd64, otherwise bad
things may happen.

Reported by: phk
MFC after: 3 days

16 years agoThis bloke has his priorities straight:
Poul-Henning Kamp [Tue, 29 Apr 2008 13:02:21 +0000 (13:02 +0000)]
This bloke has his priorities straight:

+"Can you be more stupid than aggravating the judge AND your lawyer?
+No? Oh yes you can: You can aggravate the whole kernel community."
+               -- Alexander Lyamin (about Hans Reisers murder trial)

Found at: http://www.news.com/8301-10784_3-9925607-7.html?tag=nefd.top

16 years ago Define INLINE_LIMIT and additional CFLAGS for mips.
Oleksandr Tymoshenko [Tue, 29 Apr 2008 11:28:10 +0000 (11:28 +0000)]
  Define INLINE_LIMIT and additional CFLAGS for mips.

  Approved by: cognet (mentor)

16 years ago Define KINFO_PROC_SIZE for mips.
Oleksandr Tymoshenko [Tue, 29 Apr 2008 11:17:45 +0000 (11:17 +0000)]
  Define KINFO_PROC_SIZE for mips.

  Approved by: cognet (mentor)

16 years agoMention -U and -ai arguments to mergemaster in a comment for the
Robert Watson [Tue, 29 Apr 2008 09:08:33 +0000 (09:08 +0000)]
Mention -U and -ai arguments to mergemaster in a comment for the
general mergemaster line.

MFC after: 3 days
Suggested by: Ben Laurie

16 years agoFix some section references.
Pawel Jakub Dawidek [Tue, 29 Apr 2008 08:16:05 +0000 (08:16 +0000)]
Fix some section references.

16 years agoThe referenced section name is 'Formats', not 'FORMATS'.
Pawel Jakub Dawidek [Tue, 29 Apr 2008 07:35:31 +0000 (07:35 +0000)]
The referenced section name is 'Formats', not 'FORMATS'.

16 years agoDon't depend on the modification time of the "zfs" subdir.
Ruslan Ermilov [Tue, 29 Apr 2008 06:54:12 +0000 (06:54 +0000)]
Don't depend on the modification time of the "zfs" subdir.

16 years ago- Fix makefile so it doesn't break the build in some corner cases. [1]
Ruslan Ermilov [Tue, 29 Apr 2008 06:48:00 +0000 (06:48 +0000)]
- Fix makefile so it doesn't break the build in some corner cases. [1]
- Remove an extra copy of zfs.c.

Reported by: yar [1]

16 years agoFix compiling problem.
David Xu [Tue, 29 Apr 2008 05:48:05 +0000 (05:48 +0000)]
Fix compiling problem.

16 years agoUse UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE to save
David Xu [Tue, 29 Apr 2008 03:58:18 +0000 (03:58 +0000)]
Use UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE to save
time in kernel(avoid VM lookup).

16 years agoIntroduce command UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE
David Xu [Tue, 29 Apr 2008 03:48:48 +0000 (03:48 +0000)]
Introduce command UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE
to allow userland to specify that an address is not shared by multiple
processes.

16 years agoCheck for integer overflow before calling sbrk(2), since it uses a
Jason Evans [Tue, 29 Apr 2008 01:32:42 +0000 (01:32 +0000)]
Check for integer overflow before calling sbrk(2), since it uses a
signed increment argument, but the size is an unsigned integer.

16 years agomention DX8400/50
Warner Losh [Tue, 29 Apr 2008 00:53:12 +0000 (00:53 +0000)]
mention DX8400/50

16 years agoAdd Epson DX8400 scanner.
Warner Losh [Tue, 29 Apr 2008 00:51:20 +0000 (00:51 +0000)]
Add Epson DX8400 scanner.

PR: 123148
Submitted by: Bill Squire

16 years agoANSIfy function prototypes. While I am there, constify some parameters and
Xin LI [Mon, 28 Apr 2008 21:08:42 +0000 (21:08 +0000)]
ANSIfy function prototypes.  While I am there, constify some parameters and
make use of C99 sparse initialization for static variables, this makes talk(1)
to compile cleanly with WARNS=6.

16 years agoCheck packet directions more properly instead of just checking received
Jung-uk Kim [Mon, 28 Apr 2008 19:42:11 +0000 (19:42 +0000)]
Check packet directions more properly instead of just checking received
interface is null.

PR: kern/123138
Submitted by: Dmitry (hanabana at mail dot ru)
MFC after: 1 week

16 years agoUse kgdb_parse() instead of libkvm(3) to read the first instruction from
John Baldwin [Mon, 28 Apr 2008 18:27:19 +0000 (18:27 +0000)]
Use kgdb_parse() instead of libkvm(3) to read the first instruction from
"calltrap" to see which method is used to pass trap frames.  This seg
faulted on remote gdb connections (where libkvm isn't used).

MFC after: 3 days

16 years agoEliminate pointless casts from kmem_suballoc().
Alan Cox [Mon, 28 Apr 2008 17:25:27 +0000 (17:25 +0000)]
Eliminate pointless casts from kmem_suballoc().

16 years agoCorrect minor typos in SCTP man pages.
Robert Watson [Mon, 28 Apr 2008 16:57:56 +0000 (16:57 +0000)]
Correct minor typos in SCTP man pages.

MFC after: 3 days

16 years agoRemove the 'add_kld_command' arg from load_kld(). It is always true since
John Baldwin [Mon, 28 Apr 2008 15:26:11 +0000 (15:26 +0000)]
Remove the 'add_kld_command' arg from load_kld().  It is always true since
the auto-loading of kld's switched to hooking into gdb's shared library
support.

MFC after: 1 week

16 years agoHandle endianness for mips
Oleksandr Tymoshenko [Mon, 28 Apr 2008 14:54:17 +0000 (14:54 +0000)]
Handle endianness for mips

Approved by: cognet (mentor)

16 years agoDocument 'show mount' command.
Pawel Jakub Dawidek [Mon, 28 Apr 2008 14:11:23 +0000 (14:11 +0000)]
Document 'show mount' command.

16 years agoSigh, when reapplying the patch to HEAD, I somehow forgot to commit this file.
Stefan Farfeleder [Mon, 28 Apr 2008 07:26:34 +0000 (07:26 +0000)]
Sigh, when reapplying the patch to HEAD, I somehow forgot to commit this file.

Reported by: Jaakko Heinonen

16 years agovm_map_fixed(), unlike vm_map_find(), does not update "addr", so it can be
Alan Cox [Mon, 28 Apr 2008 05:30:23 +0000 (05:30 +0000)]
vm_map_fixed(), unlike vm_map_find(), does not update "addr", so it can be
passed by value.

16 years agoThe first argment of mtdbatu or mtibatu is part of the encoding.
Marcel Moolenaar [Mon, 28 Apr 2008 03:04:41 +0000 (03:04 +0000)]
The first argment of mtdbatu or mtibatu is part of the encoding.
It needs to be constant, so eliminate the loop and "hand-unroll".

16 years agoMFp4: SMP support
Marcel Moolenaar [Sun, 27 Apr 2008 22:33:43 +0000 (22:33 +0000)]
MFp4: SMP support

16 years agorestore the hal's channel list when doing getradiocaps so it's in sync with
Sam Leffler [Sun, 27 Apr 2008 22:03:56 +0000 (22:03 +0000)]
restore the hal's channel list when doing getradiocaps so it's in sync with
the 802.11 layer's list

16 years agoEliminate track_modified_needed(), better known as pmap_track_modified()
Marcel Moolenaar [Sun, 27 Apr 2008 21:04:54 +0000 (21:04 +0000)]
Eliminate track_modified_needed(), better known as pmap_track_modified()
on other platforms.  We no longer need it because we do not create managed
mappings within the clean submap.

Pointed out by: alc

16 years ago- Fix bugs where the value of arithmetic expansion$((...)) was trucated
Stefan Farfeleder [Sun, 27 Apr 2008 20:46:45 +0000 (20:46 +0000)]
- Fix bugs where the value of arithmetic expansion$((...)) was trucated
  to type int.
- Change the type used for arithmetic expansion to intmax_t (ie. 64 bit on all
  currently supported FreeBSD architectures).  SUSv3 requires at least type
  long but allows for larger types.  Other shells (eg. bash, zsh, NetBSD's sh)
  do that too.

PR: 122659
Submitted by: Jaakko Heinonen (minor modifications by me)

16 years agoMFp4: SMP support
Marcel Moolenaar [Sun, 27 Apr 2008 19:51:34 +0000 (19:51 +0000)]
MFp4: SMP support

16 years agoMake sure tmpstk is aligned and make it 8KB in size -- not 8KB+16.
Marcel Moolenaar [Sun, 27 Apr 2008 19:03:14 +0000 (19:03 +0000)]
Make sure tmpstk is aligned and make it 8KB in size -- not 8KB+16.

16 years agoAdd support files for compiling with soft-float. This has been
Marcel Moolenaar [Sun, 27 Apr 2008 18:34:34 +0000 (18:34 +0000)]
Add support files for compiling with soft-float. This has been
copied from ARM and modified to warrant the duplication. Oh,
and to make it work for PowerPC :-)

16 years agoRemove mfsvr():
Marcel Moolenaar [Sun, 27 Apr 2008 17:13:22 +0000 (17:13 +0000)]
Remove mfsvr():
o  The function is defined unconditionally but depends on SPR_SVR,
   which is defined conditionally.
o  spr.h defines mfspr() and mtspr(), which is no worse to use.

16 years agoFix include guard spelling.
Robert Watson [Sun, 27 Apr 2008 15:51:49 +0000 (15:51 +0000)]
Fix include guard spelling.

MFC after: 3 days
Submitted by: diego

16 years agoWhen writing trailers in sendfile(2), don't call kern_writev()
Robert Watson [Sun, 27 Apr 2008 15:50:00 +0000 (15:50 +0000)]
When writing trailers in sendfile(2), don't call kern_writev()
while holding the socket buffer lock.  These leads to an
immediate panic due to recursing the socket buffer lock.  This
bug was introduced in uipc_syscalls.c:1.240, but masked by
another bug until that was fixed in uipc_syscalls.c:1.269.

Note that the current fix isn't perfect, but better than
panicking: normally we guarantee that simultaneous invocations
of a system call to write on a stream socket won't be
interlaced, which is ensured by use of the socket buffer sleep
lock.  This is guaranteed for the sendfile headers, but not
trailers.  In practice, this is likely not a problem, but
should be fixed.

MFC after: 3 days
Pointy hat to: andre (1.240), cperciva (1.269)

16 years agoAdd a breif example for the wlan change. Include an explicit pointer
Warner Losh [Sun, 27 Apr 2008 04:07:36 +0000 (04:07 +0000)]
Add a breif example for the wlan change.  Include an explicit pointer
to rc.conf(5) to remind people where to look for all the details.
People without network connectivity forget basics like this...  This
is in keeping with historic UPDATING entries which try to provide
basic information in the entry, and a pointer to more extensive
information documenting the new thing.

16 years agoRemove two unused declarations. These variables are now fields within
Alan Cox [Sun, 27 Apr 2008 00:10:45 +0000 (00:10 +0000)]
Remove two unused declarations.  These variables are now fields within
vm.h's struct kva_md_info.

16 years agoRemove an XXX comment. sys/amd64/amd64/pmap.c revision 1.516 explains why
Alan Cox [Sun, 27 Apr 2008 00:06:01 +0000 (00:06 +0000)]
Remove an XXX comment.  sys/amd64/amd64/pmap.c revision 1.516 explains why
"pmap_initialized" no longer exists.

16 years agoMFamd64 revision 1.544/i386 revision 1.549
Alan Cox [Sat, 26 Apr 2008 22:39:58 +0000 (22:39 +0000)]
MFamd64 revision 1.544/i386 revision 1.549
  Retire pmap_track_modified().  We no longer need it because we do not
  create managed mappings within the clean submap.  To prevent regressions,
  add assertions blocking the creation of managed mappings within the clean
  submap.

Approved by: imp

16 years agoAdd a missing quote symbol.
Yaroslav Tykhiy [Sat, 26 Apr 2008 22:15:03 +0000 (22:15 +0000)]
Add a missing quote symbol.

16 years agoJust met Kurt Vonnegut's quote remarkably relevant
Yaroslav Tykhiy [Sat, 26 Apr 2008 22:12:31 +0000 (22:12 +0000)]
Just met Kurt Vonnegut's quote remarkably relevant
to open software development (as well as to the other
kinds of human activity.)

16 years agoInclude the QUICC bus front-end in the module for PowerPC.
Marcel Moolenaar [Sat, 26 Apr 2008 18:41:42 +0000 (18:41 +0000)]
Include the QUICC bus front-end in the module for PowerPC.

16 years agoWith rev 1.24 of sys/powerpc/powermac/macio.c, we now get a
Marcel Moolenaar [Sat, 26 Apr 2008 18:40:59 +0000 (18:40 +0000)]
With rev 1.24 of sys/powerpc/powermac/macio.c, we now get a
total of 6 interrupt resources for scc(4) on macio(4). This
is 3 per channel, of which the 1st of each channel is the
interrupt associated with the SCC. The other 2 are for DMA
operation.
Change scc_bfe_attach() to accept an argument that's the
number of interrupts per channel (ipc) and change each bus
front-end (bfe) to pass that argument through a wrapper
for the device_attach method.

For now, we only allocate the 1st interrupt of each channel
to perserve behaviour.

16 years agoTake into account the size of the interrupt cell. It's determined
Marcel Moolenaar [Sat, 26 Apr 2008 18:35:44 +0000 (18:35 +0000)]
Take into account the size of the interrupt cell. It's determined
by the parent for interrupt resources. This corrects parsing of
the interrupts property.

With parsing of the property fixed, add all interrupts to the
resource list. Bump the max. number of interrupts from 5 to 6
as scc(4) attached to macio(4) has 6 interrupts (3 per channel).

Submitted by: Nathan Whitehorn <nathanw@uchicago.edu>

16 years agoRecognize Cicada CS8244 phy chip (among others, can be found on MPC8572DS
Rafal Jaworowski [Sat, 26 Apr 2008 18:07:24 +0000 (18:07 +0000)]
Recognize Cicada CS8244 phy chip (among others, can be found on MPC8572DS
development systems).

Obtained from: Freescale, Semihalf

16 years agoUse RSTCR for resetting the MPC8572 (the old way does not apply).
Rafal Jaworowski [Sat, 26 Apr 2008 18:03:00 +0000 (18:03 +0000)]
Use RSTCR for resetting the MPC8572 (the old way does not apply).

Obtained from: Freescale, Semihalf

16 years agoIntroduce a dedicated file for MPC85xx-specific routines. Move cpu_reset()
Rafal Jaworowski [Sat, 26 Apr 2008 17:57:29 +0000 (17:57 +0000)]
Introduce a dedicated file for MPC85xx-specific routines. Move cpu_reset()
there, as it's not relevant to Book-E specification, but is an implementation
detail, directly dependent on the given SoC version.

16 years agoImprove handling of Local Access Windows on MPC85xx systems:
Rafal Jaworowski [Sat, 26 Apr 2008 17:47:28 +0000 (17:47 +0000)]
Improve handling of Local Access Windows on MPC85xx systems:

- detect number of LAWs in run time and initalize accordingly
- introduce decode windows target IDs used in MPC8572
- other minor updates

Obtained from: Freescale, Semihalf

16 years agoMove System Revision defines to a bit better place, add MPC8572 systems IDs.
Rafal Jaworowski [Sat, 26 Apr 2008 17:39:55 +0000 (17:39 +0000)]
Move System Revision defines to a bit better place, add MPC8572 systems IDs.

16 years agoEnable NFSLOCKD for MPC85XX kernel to comply with recent NFS rework.
Rafal Jaworowski [Sat, 26 Apr 2008 17:37:13 +0000 (17:37 +0000)]
Enable NFSLOCKD for MPC85XX kernel to comply with recent NFS rework.

16 years ago- Use the revamped code from the gem(4) PCI front-end, which
Marius Strobl [Sat, 26 Apr 2008 14:17:21 +0000 (14:17 +0000)]
- Use the revamped code from the gem(4) PCI front-end, which
  doesn't require parts of the Expansion ROM to be copied around,
  for obtaining the MAC address on !OFW platforms.
- Don't unnecessarily cache bus space tag and handle nor RIDs
  in the softcs of the front-ends.
- Don't use function calls in initializers.
- Let the SBus front-end depend on sbus(4).

16 years agoRemove some remnant alpha hacks.
Marius Strobl [Sat, 26 Apr 2008 14:13:48 +0000 (14:13 +0000)]
Remove some remnant alpha hacks.

Approved by: PCI-maintainers (imp, jhb)

16 years agoIncrease the default MUTEX_ADAPTIVE_SPINS to 2000, after further
Kris Kennaway [Sat, 26 Apr 2008 13:19:07 +0000 (13:19 +0000)]
Increase the default MUTEX_ADAPTIVE_SPINS to 2000, after further
testing it turns out 200 was too short to give good adaptive
performance.

Reviewed by:   jeff
MFC after:     1 week

16 years ago* Correct a mis-merge that leaked the PROC_LOCK [1]
Kris Kennaway [Sat, 26 Apr 2008 13:16:55 +0000 (13:16 +0000)]
* Correct a mis-merge that leaked the PROC_LOCK [1]
* Return ENOENT on error instead of 0 [2]

Submitted by: rdivacky [1], kib [2]

16 years agoImplement 'show mount' command in DDB. Without argument, it prints short
Pawel Jakub Dawidek [Sat, 26 Apr 2008 13:04:48 +0000 (13:04 +0000)]
Implement 'show mount' command in DDB. Without argument, it prints short
info about all currently mounted file systems. When an address is given
as an argument, prints detailed info about the given mount point.

MFC after: 2 weeks

16 years agoFreeBSD/mips gdb build suspport. From the mips2-jnpr branch.
Warner Losh [Sat, 26 Apr 2008 12:22:46 +0000 (12:22 +0000)]
FreeBSD/mips gdb build suspport.  From the mips2-jnpr branch.

16 years agoAdd mips support to libm, from mips2-jnpr perforce branch.
Warner Losh [Sat, 26 Apr 2008 12:20:29 +0000 (12:20 +0000)]
Add mips support to libm, from mips2-jnpr perforce branch.

16 years agoBring in mips threads support from perforce mips2-jnpr branch.
Warner Losh [Sat, 26 Apr 2008 12:17:57 +0000 (12:17 +0000)]
Bring in mips threads support from perforce mips2-jnpr branch.

16 years agoAdd mips support libc from the mips2-jnpr branch of perforce.
Warner Losh [Sat, 26 Apr 2008 12:08:02 +0000 (12:08 +0000)]
Add mips support libc from the mips2-jnpr branch of perforce.

16 years agoRemove the MD isa_irq_pending() and the underlying PCI-specific
Marius Strobl [Sat, 26 Apr 2008 11:01:38 +0000 (11:01 +0000)]
Remove the MD isa_irq_pending() and the underlying PCI-specific
infrastructure. Its only consumer ever was sio(4) and thus was
unused on sparc64 since removing the last traces of sio(4) in
sparc64 configuration files in favor for uart(4) over three
years ago. If similar functionality is required again it should
be brought back as an MD intr_pending() which works for all
busses by using for example interrupt controller hooks.