]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoRewrite sendfile's header support so that headers are now sent in the first
Mike Silbersack [Sun, 1 Feb 2004 07:56:44 +0000 (07:56 +0000)]
Rewrite sendfile's header support so that headers are now sent in the first
packet along with data, instead of in their own packet.  When serving files
of size (packetsize - headersize) or smaller, this will result in one less
packet crossing the network.  Quick testing with thttpd and http_load has
shown a noticeable performance improvement in this case (350 vs 330 fetches
per second.)

Included in this commit are two support routines, iov_to_uio, and m_uiotombuf;
these routines are used by sendfile to construct the header mbuf chain that
will be linked to the rest of the data in the socket buffer.

20 years ago - Disable ithread binding in all cases for now. This doesn't make as much
Jeff Roberson [Sun, 1 Feb 2004 06:20:18 +0000 (06:20 +0000)]
 - Disable ithread binding in all cases for now.  This doesn't make as much
   sense with sched_4bsd as it does with sched_ule.
 - Use P_NOLOAD instead of the absence of td->td_ithd to determine whether or
   not a thread should be accounted for in sched_tdcnt.

20 years ago - Fix a problem where we did not drain the cache of buckets in the zone
Jeff Roberson [Sun, 1 Feb 2004 06:15:17 +0000 (06:15 +0000)]
 - Fix a problem where we did not drain the cache of buckets in the zone
   when uma_reclaim() was called.  This was introduced when the zone
   working-set algorithm was removed in favor of using the per cpu caches
   as the working set.

20 years agoCoalesce pipe allocations and frees. Previously, the pipe code
Robert Watson [Sun, 1 Feb 2004 05:56:51 +0000 (05:56 +0000)]
Coalesce pipe allocations and frees.  Previously, the pipe code
would allocate two 'struct pipe's from the pipe zone, and malloc a
mutex.

- Create a new "struct pipepair" object holding the two 'struct
  pipe' instances, struct mutex, and struct label reference.  Pipe
  structures now have a back-pointer to the pipe pair, and a
  'pipe_present' flag to indicate whether the half has been
  closed.

- Perform mutex init/destroy in zone init/destroy, avoiding
  reallocating the mutex for each pipe.  Perform most pipe structure
  setup in zone constructor.

- VM memory mappings for pageable buffers are still done outside of
  the UMA zone.

- Change MAC API to speak 'struct pipepair' instead of 'struct pipe',
  update many policies.  MAC labels are also handled outside of the
  UMA zone for now.  Label-only policy modules don't have to be
  recompiled, but if a module is recompiled, its pipe entry points
  will need to be updated.  If a module actually reached into the
  pipe structures (unlikely), that would also need to be modified.

These changes substantially simplify failure handling in the pipe
code as there are many fewer possible failure modes.

On half-close, pipes no longer free the 'struct pipe' for the closed
half until a full-close takes place.  However, VM mapped buffers
are still released on half-close.

Some code refactoring is now possible to clean up some of the back
references, etc; this patch attempts not to change the structure
of most of the pipe implementation, only allocation/free code
paths, so as to avoid introducing bugs (hopefully).

This cuts about 8%-9% off the cost of sequential pipe allocation
and free in system call tests on UP and SMP in my micro-benchmarks.
May or may not make a difference in macro-benchmarks, but doing
less work is good.

Reviewed by: juli, tjr
Testing help: dwhite, fenestro, scottl, et al

20 years ago - Revert rev 1.240 we no longer need a kthread for loadav().
Jeff Roberson [Sun, 1 Feb 2004 05:37:36 +0000 (05:37 +0000)]
 - Revert rev 1.240 we no longer need a kthread for loadav().

20 years agoUse `uname -n` as default for the HOSTNAME configuration variable.
Dag-Erling Smørgrav [Sun, 1 Feb 2004 04:51:25 +0000 (04:51 +0000)]
Use `uname -n` as default for the HOSTNAME configuration variable.

20 years ago - Use sched_load() rather than grabbing the sx lock and traversing the proc
Jeff Roberson [Sun, 1 Feb 2004 02:51:33 +0000 (02:51 +0000)]
 - Use sched_load() rather than grabbing the sx lock and traversing the proc
   table to discover the load.

20 years ago - Add a new member to struct kseq called ksq_sysload. This is intended to
Jeff Roberson [Sun, 1 Feb 2004 02:48:36 +0000 (02:48 +0000)]
 - Add a new member to struct kseq called ksq_sysload.  This is intended to
   track the load for the sched_load() function.  In the SMP case this member
   is not defined because it would be redundant with the ksg_load member
   which already tracks the non ithd load.
 - For sched_load() in the UP case simply return ksq_sysload.  In the SMP
   case traverse the list of kseq groups and sum up their ksg_load fields.

20 years ago - Keep a variable 'sched_tdcnt' that is used for the local implementation
Jeff Roberson [Sun, 1 Feb 2004 02:46:47 +0000 (02:46 +0000)]
 - Keep a variable 'sched_tdcnt' that is used for the local implementation
   of sched_load().  This variable tracks the number of running and runnable
   non ithd threads.  This removes the need to traverse the proc table and
   discover how many threads are runnable.

20 years ago - Add a sched API entry point that returns the system load. This load should
Jeff Roberson [Sun, 1 Feb 2004 02:44:35 +0000 (02:44 +0000)]
 - Add a sched API entry point that returns the system load.  This load should
   not include any ithreads.
 - Document the difference between sched_load() and sched_runnable() as they
   are very similar.

20 years agoMove KASSERT regarding td_critnest to after the value of td is set to
Robert Watson [Sun, 1 Feb 2004 02:31:36 +0000 (02:31 +0000)]
Move KASSERT regarding td_critnest to after the value of td is set to
curthread, to avoid warning and incorrect behavior.

Hoped not to mind: jeff

20 years ago - Assert that td_critnest > 0 in critical_exit() to catch cases of
Jeff Roberson [Sun, 1 Feb 2004 01:24:54 +0000 (01:24 +0000)]
 - Assert that td_critnest > 0 in critical_exit() to catch cases of
   unbalanced uses of the critical_* api.

20 years agoClarify more about LC_*/LANG priorities.
Andrey A. Chernov [Sat, 31 Jan 2004 23:29:20 +0000 (23:29 +0000)]
Clarify more about LC_*/LANG priorities.

20 years agoMerge vfscanf.c,v 1.35.
David Schultz [Sat, 31 Jan 2004 23:18:53 +0000 (23:18 +0000)]
Merge vfscanf.c,v 1.35.

20 years agoFix the handling of negative hexadecimal numbers in integer formats.
David Schultz [Sat, 31 Jan 2004 23:16:09 +0000 (23:16 +0000)]
Fix the handling of negative hexadecimal numbers in integer formats.
Note that this bug is unrelated to recent work in this area; it seems
to have been present since revision 1.1.

Obtained from: NetBSD

20 years agoFix an error in a KASSERT string: it's pipe_free_kmem(), not
Robert Watson [Sat, 31 Jan 2004 23:03:22 +0000 (23:03 +0000)]
Fix an error in a KASSERT string: it's pipe_free_kmem(), not
pipespace(), that contains this KASSERT.

20 years agoUse memcpy plus a manual NUL termination when copying the interface name
Brooks Davis [Sat, 31 Jan 2004 22:59:03 +0000 (22:59 +0000)]
Use memcpy plus a manual NUL termination when copying the interface name
from the sdl because strlcpy requires that the source string be
NUL-terminated unlike strncpy.

Submitted by: Peter Jeremy <PeterJeremy at optushome dot com dot au>

20 years agoMove the checks for whether or not to build pppctl to
Daniel Eischen [Sat, 31 Jan 2004 22:18:09 +0000 (22:18 +0000)]
Move the checks for whether or not to build pppctl to
the architecture-dependent sections of the Makefile.

Submitted by: ru

20 years agodeischen had a good idea on how to make my previous commit a little better,
Tom Rhodes [Sat, 31 Jan 2004 22:02:03 +0000 (22:02 +0000)]
deischen had a good idea on how to make my previous commit a little better,
update accordingly.

Suggested by: deischen

20 years agoCompiled longrun.c when defined options CPU_ENABLE_LONGRUN,
Takeshi Shibagaki [Sat, 31 Jan 2004 20:14:44 +0000 (20:14 +0000)]
Compiled longrun.c when defined options CPU_ENABLE_LONGRUN,
and fixed wrong comparation in cpu vendor. Longrun function
was re-enabled.

20 years agoLC_ALL not always take priority over other LC_*
Andrey A. Chernov [Sat, 31 Jan 2004 19:15:32 +0000 (19:15 +0000)]
LC_ALL not always take priority over other LC_*

Obtained from:  NetBSD
PR:             62047

20 years agoAdd the Canon N1240U to the mix. It appears to be similar to the
Warner Losh [Sat, 31 Jan 2004 18:12:18 +0000 (18:12 +0000)]
Add the Canon N1240U to the mix.  It appears to be similar to the
1220U and supported by sane

# someone should update the man page to include all the devices that
# uscanner supports.

20 years agoUpdate this manual page for the libkse -> libpthread changes.
Tom Rhodes [Sat, 31 Jan 2004 17:56:45 +0000 (17:56 +0000)]
Update this manual page for the libkse -> libpthread changes.

Discussed with: deischen

20 years agoUse strtol to parse the <%d> tags in printline() and printsys().
Ian Dowse [Sat, 31 Jan 2004 17:44:27 +0000 (17:44 +0000)]
Use strtol to parse the <%d> tags in printline() and printsys().
Sort local variables declarations in these functions.

Submitted by: bde

20 years agoApply s/LIBC_R/LIBPTHREAD/ to the DPADD variable. The LDADD variable
Marcel Moolenaar [Sat, 31 Jan 2004 17:28:09 +0000 (17:28 +0000)]
Apply s/LIBC_R/LIBPTHREAD/ to the DPADD variable. The LDADD variable
already had s/libc_r/libpthread/ applied to it.

20 years agoAdd LIBPTHREAD.
Marcel Moolenaar [Sat, 31 Jan 2004 17:25:18 +0000 (17:25 +0000)]
Add LIBPTHREAD.

20 years agoAdd some logic so that pppctl isn't built if the correct threading library
Daniel Eischen [Sat, 31 Jan 2004 17:12:05 +0000 (17:12 +0000)]
Add some logic so that pppctl isn't built if the correct threading library
is not present.

Noticed by: ru

20 years agoCorrect the link for the profiling library.
Daniel Eischen [Sat, 31 Jan 2004 17:00:04 +0000 (17:00 +0000)]
Correct the link for the profiling library.

20 years agoIntroduce the SO_BINTIME option which takes a high-resolution timestamp
Poul-Henning Kamp [Sat, 31 Jan 2004 10:40:25 +0000 (10:40 +0000)]
Introduce the SO_BINTIME option which takes a high-resolution timestamp
at packet arrival.

For benchmarking purposes SO_BINTIME is preferable to SO_TIMEVAL
since it has higher resolution and lower overhead.  Simultaneous
use of the two options is possible and they will return consistent
timestamps.

This introduces an extra test and a function call for SO_TIMEVAL, but I have
not been able to measure that.

20 years agoBump the NFCv3/TCP defaults for rsize and wsize from 8K to 32K to match
David E. O'Brien [Sat, 31 Jan 2004 10:40:15 +0000 (10:40 +0000)]
Bump the NFCv3/TCP defaults for rsize and wsize from 8K to 32K to match
Solaris and HP-UX.  This increases read performance for large files across NFS.

PR: 62024 & 26324
Submitted by: Bjoern Groenvall <bg@sics.se>

20 years agoCatch up with the 20040130 src/UPDATING entry, and spell libpthread
Ruslan Ermilov [Sat, 31 Jan 2004 09:41:05 +0000 (09:41 +0000)]
Catch up with the 20040130 src/UPDATING entry, and spell libpthread
as -lpthread.  (Alpha and SPARC64 still spell it -lkse, but we hope
this to change by 5.3-RELEASE.)

Reviewed by: deischen

20 years agoOverhaul of kerberos5/ makefiles. Most significant changes are:
Ruslan Ermilov [Sat, 31 Jan 2004 08:15:57 +0000 (08:15 +0000)]
Overhaul of kerberos5/ makefiles.  Most significant changes are:

- Dropped support for standalone builds, this was only partially
  supported anyway, and required so much magic in makefiles that
  made life dangerous (e.g., by using the custom yacc rules).

- Got rid of .OBJDIR in makefiles -- makes building of individual
  files possible again.

- Made the .x.c transformations -j safe.

- Reprogrammed LDADD to fix static build of some utilities that
  was broken.

- Fixed LDFLAGS and DPADD in the WITH_OPENLDAP case -- positively
  affects the contents of .depend files.

- Removed redundant .h's from SRCS, only kept those that are
  generated.

- libkrb5/ INCS were bogusly installed again with libgssapi/.

- Made build-tools real tools with their own makefiles in
  separate directories.  This allows us to properly track
  their dependencies, etc.

- Faster build, 21% less of makefile code!

Approved by: nectar
Reviewed by: markm
Silence on: arch

20 years agoInstall a 'disklable' alias.
David E. O'Brien [Sat, 31 Jan 2004 07:39:45 +0000 (07:39 +0000)]
Install a 'disklable' alias.

Technical Reviewed by: ru

20 years agoRemove unnecessary vm object reference and deallocate calls from ffs_read()
Alan Cox [Sat, 31 Jan 2004 05:42:58 +0000 (05:42 +0000)]
Remove unnecessary vm object reference and deallocate calls from ffs_read()
and ffs_write().  These calls trace their origins to the dead vfs_ioopt
code, first appearing in revision 1.39 of ufs_readwrite.c.

Observed by: bde
Discussed with: tegge

20 years agoNow that libpthread is the default threading library, remove the
Marcel Moolenaar [Sat, 31 Jan 2004 05:05:45 +0000 (05:05 +0000)]
Now that libpthread is the default threading library, remove the
compatibility link from libc_r to libpthread (previously a link
from libc_r to libkse).

20 years ago1. Garbage-collected LIBDES (now in a different library), LIBPERL (banished
Bruce Evans [Sat, 31 Jan 2004 03:15:50 +0000 (03:15 +0000)]
1. Garbage-collected LIBDES (now in a different library), LIBPERL (banished
to ports) and LIBRESOLV (now in a different library.

2. Added comments about nonexistent libraries LIBPC and LIBPLOT.

Submitted by: ru (1)

20 years agoFixed insertion sort errors for LIBBLUETOOTH and LIBSDP.
Bruce Evans [Sat, 31 Jan 2004 03:09:10 +0000 (03:09 +0000)]
Fixed insertion sort errors for LIBBLUETOOTH and LIBSDP.

20 years agoRemoved XXX comments about some libraries only being in the secure
Bruce Evans [Sat, 31 Jan 2004 02:51:41 +0000 (02:51 +0000)]
Removed XXX comments about some libraries only being in the secure
distribution.  This is not the place to document this, especially
now that the secure distribution is the normal one.

Reviewed by: ru

20 years agoFixed some style bugs (long lines).
Bruce Evans [Sat, 31 Jan 2004 02:17:11 +0000 (02:17 +0000)]
Fixed some style bugs (long lines).

20 years agoIn Write_Disk(), fix the non-error case where we returned to the
Marcel Moolenaar [Fri, 30 Jan 2004 20:52:54 +0000 (20:52 +0000)]
In Write_Disk(), fix the non-error case where we returned to the
caller without closing the disk device and freeing allocated
memory. Not closing the disk device prevents GEOM from retasting
after spoiling.

Pointy hat: marcel

20 years agoDrop -mcpu=pentiumpro from the default CPU flags for i386 as it doesn't
John Baldwin [Fri, 30 Jan 2004 19:51:26 +0000 (19:51 +0000)]
Drop -mcpu=pentiumpro from the default CPU flags for i386 as it doesn't
perform better than the default setting for most i386 CPUs.

Requested by: bde

20 years agoBe more robust in the probe. We dont want to get into a loop with
Søren Schmidt [Fri, 30 Jan 2004 19:16:08 +0000 (19:16 +0000)]
Be more robust in the probe. We dont want to get into a loop with
reinitting when we try to identify devices. If they dont interrupt
on identify we retry once. If this fails we simply ignore that device.

20 years agoMove recent devfs path rules note from userland section to kernel
Bruce A. Mah [Fri, 30 Jan 2004 18:29:14 +0000 (18:29 +0000)]
Move recent devfs path rules note from userland section to kernel
section; update manpage entity reference appropriately.

Requested by: cperciva

20 years agoNew errata: SA-04:01.
Bruce A. Mah [Fri, 30 Jan 2004 17:53:48 +0000 (17:53 +0000)]
New errata:  SA-04:01.

20 years agoArgh. Fix SA-04:01 note so that it actually makes sense.
Bruce A. Mah [Fri, 30 Jan 2004 17:50:35 +0000 (17:50 +0000)]
Argh.  Fix SA-04:01 note so that it actually makes sense.

20 years agoNew release notes: SA-04:01, devfs(8) rules work on directories,
Bruce A. Mah [Fri, 30 Jan 2004 17:46:05 +0000 (17:46 +0000)]
New release notes:  SA-04:01, devfs(8) rules work on directories,
libdisk fix for pc98, libpthread default, boot floppy overhaul.

Modified release notes:  Rewrite Project Evil note.

20 years agoMechanical whitespace cleanup.
Dag-Erling Smørgrav [Fri, 30 Jan 2004 16:26:29 +0000 (16:26 +0000)]
Mechanical whitespace cleanup.

20 years agoI'll continue to babysit the Kerberos-related stuff.
Jacques Vidrine [Fri, 30 Jan 2004 15:40:33 +0000 (15:40 +0000)]
I'll continue to babysit the Kerberos-related stuff.

20 years agoUse the official ng_timeout function to trigger sending. This means,
Hartmut Brandt [Fri, 30 Jan 2004 15:34:57 +0000 (15:34 +0000)]
Use the official ng_timeout function to trigger sending. This means,
that we can get rid of of all the spl*() calls, because ng_timeout
handles the locking issues.

20 years agoAdd NEC uPD72873.
Hidetoshi Shimokawa [Fri, 30 Jan 2004 14:30:19 +0000 (14:30 +0000)]
Add NEC uPD72873.

Submitted by: Christian Laursen <xi@borderworlds.dk>

20 years agoUse device_identify and bus_add_child methods to add a firewire
Hidetoshi Shimokawa [Fri, 30 Jan 2004 14:28:11 +0000 (14:28 +0000)]
Use device_identify and bus_add_child methods to add a firewire
bus on fwohci. This should fix attach failure caused by a race
between firewire and fwochi initialization for the kernel module.

20 years agoUpdate the MINIMALDOCPORTS variable.
Yoshihiro Takahashi [Fri, 30 Jan 2004 14:10:24 +0000 (14:10 +0000)]
Update the MINIMALDOCPORTS variable.
  - Add ports/Tools to get some scripts.
  - Add missing dependency (print/ghostscript-gnu-nox11). [1]
  - Remove unused ports (devel/bison and textproc/mkcatalog). [1]

Submitted by: jhay [1]

20 years agoBump __FreeBSD_version to note the change of default thread library
Daniel Eischen [Fri, 30 Jan 2004 13:43:51 +0000 (13:43 +0000)]
Bump __FreeBSD_version to note the change of default thread library
from libc_r to libpthread.

20 years agoMFi386: revision 1.397 (cosmetic changes)
Yoshihiro Takahashi [Fri, 30 Jan 2004 13:15:42 +0000 (13:15 +0000)]
MFi386: revision 1.397 (cosmetic changes)

20 years agoMFi386: revision 1.1122 (typos and cosmetic changes)
Yoshihiro Takahashi [Fri, 30 Jan 2004 13:06:57 +0000 (13:06 +0000)]
MFi386: revision 1.1122 (typos and cosmetic changes)

20 years agoAdd a note about the libkse -> libpthread switch.
Daniel Eischen [Fri, 30 Jan 2004 12:24:21 +0000 (12:24 +0000)]
Add a note about the libkse -> libpthread switch.

Reviewed by: imp

20 years agos/-lc_r/-lpthread/
Daniel Eischen [Fri, 30 Jan 2004 12:14:50 +0000 (12:14 +0000)]
s/-lc_r/-lpthread/

20 years agoMake gcc -pthread link to -lpthread instead of -lc_r.
Daniel Eischen [Fri, 30 Jan 2004 12:14:19 +0000 (12:14 +0000)]
Make gcc -pthread link to -lpthread instead of -lc_r.

20 years agoChange libkse back to libpthread and make it the default
Daniel Eischen [Fri, 30 Jan 2004 12:13:17 +0000 (12:13 +0000)]
Change libkse back to libpthread and make it the default
thread library for i386, amd64, and ia64.  For alpha
and sparc64 the library is not changed and remains libkse,
and links are installed so that libpthread -> libc_r.

The gcc -pthread option will be changed in a separate
commit so that it links to -lpthread instead of -lc_r.

Approved by: re@

20 years agoadd hprop to the build and add LDAP mods to hprop.
Michael Reifenberger [Fri, 30 Jan 2004 11:06:48 +0000 (11:06 +0000)]
add hprop to the build and add LDAP mods to hprop.
add kadmind.8

Submitted by: Alex Deiter <tiamat@komi.mts.ru>

20 years agoAdd a test for what was broken in rev. 1.28 and fixed in rev. 1.29
Ruslan Ermilov [Fri, 30 Jan 2004 09:46:28 +0000 (09:46 +0000)]
Add a test for what was broken in rev. 1.28 and fixed in rev. 1.29
of make/str.c.

20 years agoRemove NetBSD'isms (add FreeBSD'isms?), which makes gre(4) working again.
Maxim Sobolev [Fri, 30 Jan 2004 09:03:01 +0000 (09:03 +0000)]
Remove NetBSD'isms (add FreeBSD'isms?), which makes gre(4) working again.

20 years agoTake the plunge and make this driver be INTR_FAST. This re-arranges the
Scott Long [Fri, 30 Jan 2004 07:04:39 +0000 (07:04 +0000)]
Take the plunge and make this driver be INTR_FAST.  This re-arranges the
interrupt handler so that no locks are needed, and schedules the
command completion routine with a taskqueue_fast.  This also corrects the
locking in the command thread and removes the need for operation flags.

Simple load tests show that this is now considerably faster than FreeBSD 4.x
in the SMP case when multiple i/o tasks are running.

20 years agoFix a typo[1] and merge from the English version:
Hiroki Sato [Fri, 30 Jan 2004 05:42:22 +0000 (05:42 +0000)]
Fix a typo[1] and merge from the English version:

1.59  -> 1.61  errata/article.sgml

Pointed out by[1]: Daisuke Takahashi <takaha@mue.biglobe.ne.jp>
References[1]: [doc-jp-work 828]

20 years agoIn checkLabels(), deal with the fact that on ia64 we do not have a
Marcel Moolenaar [Fri, 30 Jan 2004 05:21:32 +0000 (05:21 +0000)]
In checkLabels(), deal with the fact that on ia64 we do not have a
disklabel.

20 years agoGRR. MFi386: white space spam
Peter Wemm [Fri, 30 Jan 2004 03:02:45 +0000 (03:02 +0000)]
GRR. MFi386: white space spam

20 years agoMerge some more changes from i386.
Peter Wemm [Fri, 30 Jan 2004 00:24:45 +0000 (00:24 +0000)]
Merge some more changes from i386.

20 years agoAdd a test for what was fixed in revs. 1.28 and 1.29 of make/str.c.
Ruslan Ermilov [Thu, 29 Jan 2004 22:51:36 +0000 (22:51 +0000)]
Add a test for what was fixed in revs. 1.28 and 1.29 of make/str.c.

20 years agoConsistently capitalize acronyms like POSIX and NIC. Style.
Jens Schweikhardt [Thu, 29 Jan 2004 21:24:53 +0000 (21:24 +0000)]
Consistently capitalize acronyms like POSIX and NIC. Style.

20 years agoFix typos and remove whitespace at EOL.
Jens Schweikhardt [Thu, 29 Jan 2004 21:07:54 +0000 (21:07 +0000)]
Fix typos and remove whitespace at EOL.

20 years ago- Tweak order of arch fs options to stick all floppy options together and
John Baldwin [Thu, 29 Jan 2004 19:01:58 +0000 (19:01 +0000)]
- Tweak order of arch fs options to stick all floppy options together and
  all MFS root options together and in the same order on all archs.
- Drop support for using the no longer existing 2.88 floppy image for CD
  booting on i386 via the EMUL_BOOT variable.
- Use a stock loader on the boot floppy rather than one without forth.
- Force the gzip of the mfsroot in release.9 to succeed.  This is useful
  when doing lots of rereleases when testing floppy changes.
- If an acpi kernel module exists, stick it on the boot floppy and for
  now always load it after the kernel.

20 years agoSync with bsdlabel/Makefile.
David E. O'Brien [Thu, 29 Jan 2004 18:15:19 +0000 (18:15 +0000)]
Sync with bsdlabel/Makefile.

20 years agoAdd a very basic manpage.
David E. O'Brien [Thu, 29 Jan 2004 18:12:27 +0000 (18:12 +0000)]
Add a very basic manpage.

20 years agoAdd kk_KZ.PT154
Andrey A. Chernov [Thu, 29 Jan 2004 16:28:22 +0000 (16:28 +0000)]
Add kk_KZ.PT154

Submitted by:   Birsh T <tim@zhezu.kz>

20 years agoLatest version of this file from InterNIC. This version updates the IP
Doug Barton [Thu, 29 Jan 2004 16:24:30 +0000 (16:24 +0000)]
Latest version of this file from InterNIC. This version updates the IP
address of b.root-servers.net, and various comments.

20 years agoTwo nits from rev. 1.835:
Ruslan Ermilov [Thu, 29 Jan 2004 16:21:00 +0000 (16:21 +0000)]
Two nits from rev. 1.835:

- Fixed spelling of NO_FLOPPIES to unbreak snapshot building on
  architectures that don't support them (as installation source).

- Don't rely on CVS setting execution bits on the shell script.
  (This is also consistent with the rest of this makefile.)

Reviewed by: jhb

20 years agoAdd kk_KZ.PT154
Andrey A. Chernov [Thu, 29 Jan 2004 16:17:25 +0000 (16:17 +0000)]
Add kk_KZ.PT154

20 years agoThese scripts run the protocol test suites for the SSCOP, SSCF-UNI and
Hartmut Brandt [Thu, 29 Jan 2004 16:01:57 +0000 (16:01 +0000)]
These scripts run the protocol test suites for the SSCOP, SSCF-UNI and
UNI protocols. The actual test suites are not in these directories because
of their size. One needs to install the atmsupport port (the script
will remind you, if it cannot find the port).

The SSCOP test suite includes booth the ETSI and the ITU-T test suite,
the SSCF-UNI test suite is home grown and the UNI test suite is the
P2MP ETSI test suite. Others may follow.

20 years agoAdd a simple test script for the ng_atmpif node.
Hartmut Brandt [Thu, 29 Jan 2004 15:58:06 +0000 (15:58 +0000)]
Add a simple test script for the ng_atmpif node.

Submitted by: Vincent Jardin <vjardin@free.fr>

20 years agoThe name of small boot floppy should be boot-small.flp to avoid conflicting
Yoshihiro Takahashi [Thu, 29 Jan 2004 15:14:59 +0000 (15:14 +0000)]
The name of small boot floppy should be boot-small.flp to avoid conflicting
normal size floppy.

Reviewed by: jhb

20 years agoStyle(9) option sorting
Guido van Rooij [Thu, 29 Jan 2004 15:14:03 +0000 (15:14 +0000)]
Style(9) option sorting

Submitted by: Ruslan Ermilov <ru@FreeBSD.org>

20 years agoCleanups/cosmetics.
Søren Schmidt [Thu, 29 Jan 2004 15:03:01 +0000 (15:03 +0000)]
Cleanups/cosmetics.

20 years agoFix pathname so 'make tags' in a kernel build directory looks in
Ken Smith [Thu, 29 Jan 2004 14:58:22 +0000 (14:58 +0000)]
Fix pathname so 'make tags' in a kernel build directory looks in
the right place for the $MACHINE/include directory.

Approved by: rwatson (mentor)

20 years agoFix manpage and usage() to reflect that -a can be used in combination
Guido van Rooij [Thu, 29 Jan 2004 14:42:26 +0000 (14:42 +0000)]
Fix manpage and usage() to reflect that -a can be used in combination
with -o

Submitted by: Ruslan Ermilov <ru@FreeBSD.org>
Pointed out by: Ceri Davies <ceri@submonkey.net

20 years agoRemoved duplicate $FreeBSD$ tag.
Ruslan Ermilov [Thu, 29 Jan 2004 14:36:16 +0000 (14:36 +0000)]
Removed duplicate $FreeBSD$ tag.

20 years agoAdd dependency to snd_via8233
Michael Reifenberger [Thu, 29 Jan 2004 14:11:03 +0000 (14:11 +0000)]
Add dependency to snd_via8233

20 years agoUnbreak -o fstab and -o current in combination with -a
Guido van Rooij [Thu, 29 Jan 2004 13:16:51 +0000 (13:16 +0000)]
Unbreak -o fstab and -o current in combination with -a

MFC after: 2 weeks

20 years agoFixed breakage of scheduling in rev.1.29 of subr_4bsd.c. The
Bruce Evans [Thu, 29 Jan 2004 12:35:11 +0000 (12:35 +0000)]
Fixed breakage of scheduling in rev.1.29 of subr_4bsd.c.  The
"scheduler" here has very little to do with scheduling.  It is actually
the swapper, and it really must be the last SYSINIT'ed item like its
comment says, since proc0 metamorphoses into swapper by calling
scheduler() last in mi_start(), and scheduler() never returns..  Rev.1.29
of subr_4bsd.c broke this by adding another SI_ORDER_FIRST item
(kproc_start() for schedcpu_thread() onto the SI_SUB_RUN_SCHEDULER_LIST.
The sorting of SYSINITs with identical orders (at all levels) is
apparently nondeterministic, so this resulted in schedule() sometimes
being called second last and schedcpu_thread() not being called at all.

This quick fix just changes the code to almost match the comment
(SI_ORDER_FIRST -> SI_ORDER_ANY).  "LAST" is misspelled "ANY", and
there is no way to ensure that there is only 1 very lst SYSINIT.
A more complete fix would remove the SYSINIT obfuscation.

20 years agoI update the rwlock code in libthr to be more standards compliant and
Mike Makonnen [Thu, 29 Jan 2004 12:03:17 +0000 (12:03 +0000)]
I update the rwlock code in libthr to be more standards compliant and
what do I get for my troubles? libc breaks offcourse!

Reimplement a hack (in libthr) that allows libc to use
rwlocks without initializing them first. The hack was reimplemented
so that only a private libc version of the rwlock locking functions
initializes an uninitialized rwlock. The application version will
correctly fail.

20 years agoWhen suspending a thread if the timeout was very short or
Mike Makonnen [Thu, 29 Jan 2004 09:44:36 +0000 (09:44 +0000)]
When suspending a thread if the timeout was very short or
the system call got interrupted and the absolute timeout is
converted to a relative timeout, it may happen that we get a
negative number. In such a case, simply set the timeout to
zero so that if the event that the thread wants to wait for has
happened it can still return successfully, but if it hasn't
happened then the thread doesn't suspend indefinitely. This should
fix certain applications (including mozilla) that seem to hang
indefinitely sometimes.

Noticed and debugged by: Morten Johansen <root@morten-johansen.net>

20 years agoAdd reference to environ(7)
Andrey A. Chernov [Thu, 29 Jan 2004 09:27:24 +0000 (09:27 +0000)]
Add reference to environ(7)

20 years agoFix a bug where we never managed to include the sense data we wanted to send.
Matt Jacob [Thu, 29 Jan 2004 06:36:30 +0000 (06:36 +0000)]
Fix a bug where we never managed to include the sense data we wanted to send.

20 years agoGo back to using AUTHMODE_AUTO if WEP is on. In some cases, the Centrino
Bill Paul [Thu, 29 Jan 2004 03:16:58 +0000 (03:16 +0000)]
Go back to using AUTHMODE_AUTO if WEP is on. In some cases, the Centrino
won't associate in BSS mode if you use AUTHMODE_SHARED. I probably don't
understand enough to know when SHARED should be used vs. OPEN or WPA.
For now, go back to what works.

20 years agoFix longstanding buglet- for centrally handled CTIO2s we were checking the wrong
Matt Jacob [Thu, 29 Jan 2004 02:40:43 +0000 (02:40 +0000)]
Fix longstanding buglet- for centrally handled CTIO2s we were checking the wrong
bit for this being the  last CTIO2.  It didn't matter since it really was the
last CTIO2 and the resources recycled, but still....

Add in CTIO3 define for future DAC work.

20 years agoAssert process lock in ptracestop(), since we're going to rely
Robert Watson [Thu, 29 Jan 2004 00:58:21 +0000 (00:58 +0000)]
Assert process lock in ptracestop(), since we're going to rely
on it, and later unlock it.

20 years agoWhen UMA_MD_SMALL_ALLOC is defined, pmap_kextract will be called
Peter Grehan [Thu, 29 Jan 2004 00:45:41 +0000 (00:45 +0000)]
When UMA_MD_SMALL_ALLOC is defined, pmap_kextract will be called
for direct-mapped addresses. Assume that any address less than KVA
is one of these and return it. Also assert that an address is KVA
does have a valid mapping - callers of pmap_kextract don't check
the return value, since they assume that they have a valid virtual
address.

20 years agoImplement UMA_MD_SMALL_ALLOC, since the BAT registers allow direct
Peter Grehan [Thu, 29 Jan 2004 00:32:22 +0000 (00:32 +0000)]
Implement UMA_MD_SMALL_ALLOC, since the BAT registers allow direct
addressing of memory. Makes a substantial improvement for apps that
stress the limited amount of KVM on PPC (e.g. untarring the ports tree).

uma_machdep.c stolen from amd64/ia64.

20 years agoRe-add debug register support.
Peter Wemm [Thu, 29 Jan 2004 00:07:29 +0000 (00:07 +0000)]
Re-add debug register support.
Some other minor tweaks snuck in here, including supporting more
discontiguous memory segments and some cosmetic tweaks.

20 years agoRe-add user_dbreg_trap() for debug register support
Peter Wemm [Thu, 29 Jan 2004 00:05:03 +0000 (00:05 +0000)]
Re-add user_dbreg_trap() for debug register support

20 years agoTake another shot at the invariants calls to __panic. They hadn't been
Peter Wemm [Thu, 29 Jan 2004 00:02:54 +0000 (00:02 +0000)]
Take another shot at the invariants calls to __panic.  They hadn't been
updated for the regparm ABI on amd64.
Context switch debug regs.
Update for fpu simplification
Don't needlessly reload %cr3, in case the cpu has the tlb flush filter
turned off.  Re-add LAZY_SWITCH stubs.