]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
25 years agoCopy from vendor branch. This mainly undoes most of the backing out of
Bruce Evans [Mon, 8 Feb 1999 06:29:10 +0000 (06:29 +0000)]
Copy from vendor branch.  This mainly undoes most of the backing out of
Russian zones/rules in rev.1.12.  ache objected mainly to the changes
in the Moscow zone names in rev.1.11 and those changes have been backed
out in the vendor branch.

Reviewed by: ache

25 years agoremove leftover garbage line.
Julian Elischer [Mon, 8 Feb 1999 05:53:39 +0000 (05:53 +0000)]
remove leftover garbage line.

25 years agoFix for PR 9309.
Julian Elischer [Mon, 8 Feb 1999 05:48:46 +0000 (05:48 +0000)]
Fix for PR 9309.
Divert was not feeding clean data to ifa_ifwithaddr() so it was
giving bad results.
Submitted by: kseel <kseel@utcorp.com>, Ruslan Ermilov <ru@ucb.crimea.ua>

25 years ago General cleanup. Remove #if 0's and remove useless register qualifiers.
Matthew Dillon [Mon, 8 Feb 1999 05:15:54 +0000 (05:15 +0000)]
General cleanup.  Remove #if 0's and remove useless register qualifiers.

25 years agoBumb up loader's version.
Daniel C. Sobral [Mon, 8 Feb 1999 04:14:31 +0000 (04:14 +0000)]
Bumb up loader's version.

Reviewed by: Michael Smith <msmith@freebsd.org>

25 years ago Adjust idle zero-page fill hysteresis based on tests. Use 2/3 and 4/5
Matthew Dillon [Mon, 8 Feb 1999 02:42:13 +0000 (02:42 +0000)]
Adjust idle zero-page fill hysteresis based on tests.  Use 2/3 and 4/5
    zero-fill levels.

    Adjust comment for ozfod in vmmeter.h - this counter represents
    non-optimal ( on the fly ) zero fills, not prefills.

25 years ago Include discrete ozfod as well as ozfod/zfod percentage.
Matthew Dillon [Mon, 8 Feb 1999 02:39:45 +0000 (02:39 +0000)]
Include discrete ozfod as well as ozfod/zfod percentage.

25 years ago If there are 4 or fewer disk devices, we have room to display additional
Matthew Dillon [Mon, 8 Feb 1999 02:11:52 +0000 (02:11 +0000)]
If there are 4 or fewer disk devices, we have room to display additional
    VM statistics.  zfod is moved and %slo-z ( percentage of zero-fills that
    were slow, i.e. not pre-zero'd ), and number of pages freed per second.

25 years agoFix white-space spamming.
David E. O'Brien [Mon, 8 Feb 1999 01:32:04 +0000 (01:32 +0000)]
Fix white-space spamming.

25 years ago Add hysteresis to alpha version of vm_page_zero_idle().
Matthew Dillon [Mon, 8 Feb 1999 00:47:32 +0000 (00:47 +0000)]
Add hysteresis to alpha version of vm_page_zero_idle().

25 years ago Rip out PQ_ZERO queue. PQ_ZERO functionality is now combined in with
Matthew Dillon [Mon, 8 Feb 1999 00:37:36 +0000 (00:37 +0000)]
Rip out PQ_ZERO queue.  PQ_ZERO functionality is now combined in with
    PQ_FREE.  There is little operational difference other then the kernel
    being a few kilobytes smaller and the code being more readable.

    * vm_page_select_free() has been *greatly* simplified.
    * The PQ_ZERO page queue and supporting structures have been removed
    * vm_page_zero_idle() revamped (see below)

    PG_ZERO setting and clearing has been migrated from vm_page_alloc()
    to vm_page_free[_zero]() and will eventually be guarenteed to remain
    tracked throughout a page's life ( if it isn't already ).

    When a page is freed, PG_ZERO pages are appended to the appropriate
    tailq in the PQ_FREE queue while non-PG_ZERO pages are prepended.
    When locating a new free page, PG_ZERO selection operates from within
    vm_page_list_find() ( get page from end of queue instead of beginning
    of queue ) and then only occurs in the nominal critical path case.  If
    the nominal case misses, both normal and zero-page allocation devolves
    into the same _vm_page_list_find() select code without any specific
    zero-page optimizations.

    Additionally, vm_page_zero_idle() has been revamped.  Hysteresis has been
    added and zero-page tracking adjusted to conform with the other changes.
    Currently hysteresis is set at 1/3 (lo) and 1/2 (hi) the number of free
    pages.  We may wish to increase both parameters as time permits.  The
    hysteresis is designed to avoid silly zeroing in borderline allocation/free
    situations.

25 years ago Backed out vm_map coalesce optimization - it resulted in 22% more page
Matthew Dillon [Mon, 8 Feb 1999 00:27:56 +0000 (00:27 +0000)]
Backed out vm_map coalesce optimization - it resulted in 22% more page
    faults for reasons unknown ( under investigation ).
    /usr/bin/time -l make in /usr/src/bin went from 67000 faults to 90000
    faults.

25 years agoChange the load address of the ELF dynamic linker from "2L*MAXDSIZ"
John Polstra [Sun, 7 Feb 1999 23:49:56 +0000 (23:49 +0000)]
Change the load address of the ELF dynamic linker from "2L*MAXDSIZ"
to an architecture-specific value defined in <machine/elf.h>.  This
solves problems on large-memory systems that have a high value for
MAXDSIZ.

The load address is controlled by a new macro ELF_RTLD_ADDR(vmspace).
On the i386 it is hard-wired to 0x08000000, which is the standard
SVR4 location for the dynamic linker.

On the Alpha, the dynamic linker is loaded MAXDSIZ bytes beyond
the start of the program's data segment.  This is the same place
a userland mmap(0, ...) call would put it, so it ends up just below
all the shared libraries.  The rationale behind the calculation is
that it allows room for the data segment to grow to its maximum
possible size.

These changes have been tested on the i386 for several months
without problems.  They have been tested on the Alpha as well,
though not for nearly as long.  I would like to merge the changes
into 3.1 within a week if no problems have surfaced as a result of
them.

25 years ago Remove MAP_ENTRY_IS_A_MAP 'share' maps. These maps were once used to
Matthew Dillon [Sun, 7 Feb 1999 21:48:23 +0000 (21:48 +0000)]
Remove MAP_ENTRY_IS_A_MAP 'share' maps.  These maps were once used to
    attempt to optimize forks but were essentially given-up on due to
    problems and replaced with an explicit dup of the vm_map_entry structure.
    Prior to the removal, they were entirely unused.

25 years ago Remove L1 cache coloring optimization ( leave L2 cache coloring opt ).
Matthew Dillon [Sun, 7 Feb 1999 20:45:15 +0000 (20:45 +0000)]
Remove L1 cache coloring optimization ( leave L2 cache coloring opt ).

    Rewrite vm_page_list_find() and vm_page_select_free() - make inline out
    of nominal case.

25 years agoReorder the pcvt(4) options to the appropriate section, so i can close
Joerg Wunsch [Sun, 7 Feb 1999 20:33:05 +0000 (20:33 +0000)]
Reorder the pcvt(4) options to the appropriate section, so i can close
docs/1855. :)

PR: docs/1855
Submitted by: Jeroen Ruigrok/Asmodai <asmodai@wxs.nl>

25 years agoFix a typo describing the filename of cvt-wtmp.
Joerg Wunsch [Sun, 7 Feb 1999 20:30:56 +0000 (20:30 +0000)]
Fix a typo describing the filename of cvt-wtmp.

Submitted by: Ruslan Ermilov <ru@ucb.crimea.ua>

25 years ago1) Fix our view of how ibcs2_semid_ds looks. There is no padding int there
Guido van Rooij [Sun, 7 Feb 1999 19:35:40 +0000 (19:35 +0000)]
1) Fix our view of how ibcs2_semid_ds looks. There is no padding int there
(SVR4 does have it so that;s probably the cause of this bug)
2) Add a wrapper function for translation between ibcs2_ipc_perm and
ipc_perm as I think we screwed up when defining the ipc_perm struct and
mixed up 'normal' and creator [ug]id's
3) Fix IBCS2_IPC_STAT semctl. The FreeBSD version needs a union semun
whereas the IBCS2 version gives a struct ibcs2_semid_ds.

Apparently this is all fixed in the SVR4 compatibility code.
PR: 7729

25 years agoCorrect server-side chap authentication comparison
Brian Somers [Sun, 7 Feb 1999 13:56:29 +0000 (13:56 +0000)]
Correct server-side chap authentication comparison
(broken with last commit).

25 years agoRemove forgotten diagnostics
Brian Somers [Sun, 7 Feb 1999 13:48:38 +0000 (13:48 +0000)]
Remove forgotten diagnostics

25 years agoIf user selects X, also auto-select compat22 (for now).
Jordan K. Hubbard [Sun, 7 Feb 1999 12:57:04 +0000 (12:57 +0000)]
If user selects X, also auto-select compat22 (for now).

25 years agoSMAILCF distro no longer exists.
Jordan K. Hubbard [Sun, 7 Feb 1999 12:43:14 +0000 (12:43 +0000)]
SMAILCF distro no longer exists.

25 years agoRemove strange-looking quotes. This part of the code is #if'ed out,
Daniel C. Sobral [Sun, 7 Feb 1999 12:12:29 +0000 (12:12 +0000)]
Remove strange-looking quotes. This part of the code is #if'ed out,
and the variable doesn't even exist (though it is referenced elsewhere).
Just make sure it produces error messages when Mike get back to it.

PR: kern/9934
Submitted by: Adrian Filipi-Martin

25 years agoMaxHeaderLines is now MaxHeadersLength (in bytes)
Peter Wemm [Sun, 7 Feb 1999 09:48:52 +0000 (09:48 +0000)]
MaxHeaderLines is now MaxHeadersLength (in bytes)

25 years agoRemove the standalone buildtools stuff, we don't use it here and it
Peter Wemm [Sun, 7 Feb 1999 09:45:12 +0000 (09:45 +0000)]
Remove the standalone buildtools stuff, we don't use it here and it
clutters things up.  (Suggested by wollman)

25 years agoMerge 8.9.2+patches -> 8.9.3 changes into mainline.
Peter Wemm [Sun, 7 Feb 1999 09:43:30 +0000 (09:43 +0000)]
Merge 8.9.2+patches -> 8.9.3 changes into mainline.

25 years agoThis commit was generated by cvs2svn to compensate for changes in r43730,
Peter Wemm [Sun, 7 Feb 1999 09:40:41 +0000 (09:40 +0000)]
This commit was generated by cvs2svn to compensate for changes in r43730,
which included commits to RCS files with non-trunk default branches.

25 years agoImport sendmail 8.9.3 onto vendor branch, replacing previous interim
Peter Wemm [Sun, 7 Feb 1999 09:40:41 +0000 (09:40 +0000)]
Import sendmail 8.9.3 onto vendor branch, replacing previous interim
8.9.2 + patches version.

Obtained from: ftp.sendmail.org

25 years ago When shadowing objects, adjust the page coloring of the shadowing object
Matthew Dillon [Sun, 7 Feb 1999 08:44:53 +0000 (08:44 +0000)]
When shadowing objects, adjust the page coloring of the shadowing object
    such that pages in the combined/shadowed object are consistantly
    colored.

Submitted by: "John S. Dyson" <dyson@iquest.net>

25 years ago- Add atkbd.4, atkbdc.4, splash.4 and vga.4.
Kazutaka YOKOTA [Sun, 7 Feb 1999 05:40:14 +0000 (05:40 +0000)]
- Add atkbd.4, atkbdc.4, splash.4 and vga.4.
- Update pcvt.4 and psm.4 to reflect recent keyboard driver changes.

25 years ago- Add FreeBSD 2.2.7, 2.2.8, 3.1, and 4.0 to the Fx and Os macros.
Kazutaka YOKOTA [Sun, 7 Feb 1999 05:36:30 +0000 (05:36 +0000)]
- Add FreeBSD 2.2.7, 2.2.8, 3.1, and 4.0 to the Fx and Os macros.
- Add 2.2.1 and 2.2.2 missing in a couple of places too.

(This is an interim measure and a better way of redefining these macros
is in discussion.)

25 years agoMake the ipx part of netstat work again.
John Hay [Sat, 6 Feb 1999 19:12:48 +0000 (19:12 +0000)]
Make the ipx part of netstat work again.

25 years agoHmm- I must be blind. There *was* already an appropriate typedef still
Matt Jacob [Sat, 6 Feb 1999 17:58:06 +0000 (17:58 +0000)]
Hmm- I must be blind. There *was* already an appropriate typedef still
left there. The typedef that was inserted negated the correct typedef.
This may have just been cvs automerge foo. Sorry.

25 years agoDon't dump core when p_stat is not in the expected range. This is
Bill Fenner [Sat, 6 Feb 1999 16:58:50 +0000 (16:58 +0000)]
Don't dump core when p_stat is not in the expected range.  This is
only likely to happen when you have a kernel<>userland mismatch,
but it's really annoying when top dumps core and leaves the terminal
in a mangled state; it's much nicer to print nicely formatted gibberish.

25 years agoI don't know what happened here- I submitted a *working* patch for
Matt Jacob [Sat, 6 Feb 1999 16:39:16 +0000 (16:39 +0000)]
I don't know what happened here- I submitted a *working* patch for
alpha to John Polstra yesterday and it was checked in. Then there
was a bunch of CVS activity, and it ended up with the *broken*
prototype being reinserted. All I see in the history is 'style nits'
as a comment, and I certainly agree with the latter of those two words.

25 years agoReshuffle a number of menus to be more navigable. Delete extra
Jordan K. Hubbard [Sat, 6 Feb 1999 16:35:26 +0000 (16:35 +0000)]
Reshuffle a number of menus to be more navigable.  Delete extra
Root Password entry (PR#9291 - Jack O'Neill).

25 years agoChanged to use IICBUS_REPEATED_START
Roger Hardiman [Sat, 6 Feb 1999 10:56:09 +0000 (10:56 +0000)]
Changed to use IICBUS_REPEATED_START

Submitted by: Changed to use IICBUS_REPEATED_START

25 years agoTeach IPXrouted to handle the internal net properly.
John Hay [Sat, 6 Feb 1999 10:52:21 +0000 (10:52 +0000)]
Teach IPXrouted to handle the internal net properly.
PR: 9871
Submitted by: Boris Popov <bp@butya.kz>

25 years agoMake it possible to use lo0 as an internal IPX network. This is usefull
John Hay [Sat, 6 Feb 1999 10:48:11 +0000 (10:48 +0000)]
Make it possible to use lo0 as an internal IPX network. This is usefull
for mars_nwe server and nwfs.
PR: 9871
Submitted by: Boris Popov <bp@butya.kz>

25 years agoDon't use the pointer returned by MALLOC before checking if it is NULL.
John Hay [Sat, 6 Feb 1999 10:45:17 +0000 (10:45 +0000)]
Don't use the pointer returned by MALLOC before checking if it is NULL.
PR: 9871 (part of it)
Submitted by: Boris Popov <bp@butya.kz>

25 years agoSync with syscons for i386.
KATO Takenori [Sat, 6 Feb 1999 09:30:19 +0000 (09:30 +0000)]
Sync with syscons for i386.

Submitted by: Kazutaka YOKOTA <yokota@zodiac.mech.utsunomiya-u.ac.jp>

25 years agoAck! Correct a typo which crept into here and reshuffle the index back
Jordan K. Hubbard [Sat, 6 Feb 1999 08:45:20 +0000 (08:45 +0000)]
Ack!  Correct a typo which crept into here and reshuffle the index back
into alphabetical order.

25 years agoAdded underscores to some names in svr4_locore.s so that it compiles
Bruce Evans [Sat, 6 Feb 1999 08:41:06 +0000 (08:41 +0000)]
Added underscores to some names in svr4_locore.s so that it compiles
with -aout.  Added translation back to elf names in asnames.h as
usual.  The elf names were inconsistent in the aout case even
internally because a macro adds an underscore to just one of them.

Removed commented out code for a previous life of `svr4_esigcode'.
Didn't add an underscore to `svr4_esigcode' since it is correct for
aout although wrong for elf, like most internal names in assembler
files.  These names should be in a different namespace so that gprof
can ignore them.

Fixed some disorder in asnames.h.

25 years ago Flush delayed-write data out prior to issuing a rename rpc. This appears
Matthew Dillon [Sat, 6 Feb 1999 07:48:56 +0000 (07:48 +0000)]
Flush delayed-write data out prior to issuing a rename rpc.  This appears
    to fix the problem w/ NFSV3 whereby a make installworld would get into
    high-network-bandwidth situations continuously trying to retry nfs writes
    that fail with a 'stale file handle' error.

25 years agoHonor ${LDFLAGS} for building svr4_genassym. This is part of fixing
Bruce Evans [Sat, 6 Feb 1999 07:48:24 +0000 (07:48 +0000)]
Honor ${LDFLAGS} for building svr4_genassym.  This is part of fixing
world breakage (mainly for cross-world cases).  The world Makefile
attempts to build tools static so that nonexistent or wrong shared
libraries and interpreters don't get used.  This is broken anyway
since the world Makefile doesn't know about svr4_genassym.

Force building svr4_genassym static.  This is part of "fixing"
aout-to-elf-build breakage.  aout-to-elf-build abuses NOTOOLS to
avoid rebuilding all the aout tools.  This saves time and avoids
some complications.  However, it breaks all the internal tools --
they get linked to target libraries which might not work.  Cases
where the host can run the target's static libraries are "fixed"
by encrufting all Makefiles that build internal tools to build the
tools static.

Don't add .depend to CLEANFILES -- it just breaks the separation of
`make cleandepend' from `make clean'.

Removed some superflous explicit dependencies.

25 years ago Add hysteresis to the 'swap_pager_getswapspace; failed' console message.
Matthew Dillon [Sat, 6 Feb 1999 07:22:21 +0000 (07:22 +0000)]
Add hysteresis to the 'swap_pager_getswapspace; failed' console message.
    Also widen the hysteresis levels a little ( these really should be
    dynamically configured ).

25 years agoBASE=4.0
Jordan K. Hubbard [Sat, 6 Feb 1999 06:51:30 +0000 (06:51 +0000)]
BASE=4.0

25 years ago Be nice when no swap is configured in system
Matthew Dillon [Sat, 6 Feb 1999 06:43:56 +0000 (06:43 +0000)]
Be nice when no swap is configured in system

25 years ago Make 'top' handle case w/ new swapper where no swap is configured
Matthew Dillon [Sat, 6 Feb 1999 06:33:55 +0000 (06:33 +0000)]
Make 'top' handle case w/ new swapper where no swap is configured

25 years ago Handle case where no swap is configured
Matthew Dillon [Sat, 6 Feb 1999 06:31:57 +0000 (06:31 +0000)]
Handle case where no swap is configured

25 years agoDecouple pap & chap output routines from the corresponding
Brian Somers [Sat, 6 Feb 1999 02:54:47 +0000 (02:54 +0000)]
Decouple pap & chap output routines from the corresponding
input routines and take advantage of the new init/continue
interface in libradius.  This allows a timely response on
other links in an MP setup while RADIUS requests are in
progress as well as the ability to handle other data from
the peer in parallel.  It should also make the future addition
of PAM support trivial.

While I'm in there, validate pap & chap header IDs if
``idcheck'' is enabled (the default) for other FSM packet
types.

NOTE: This involved integrating the generation of chap
      challenges and the validation of chap responses
      (and commenting what's going on in those routines).
      I currently have no way of testing ppps ability
      to respond to M$Chap CHALLENGEs correctly, so if
      someone could do the honours, it'd be much
      appreciated (it *looks* ok!).

Sponsored by: Internet Business Solutions Ltd., Switzerland

25 years agoBack out rev 1.89. The #include is necessary for the #if that tests
Peter Wemm [Sat, 6 Feb 1999 02:28:52 +0000 (02:28 +0000)]
Back out rev 1.89.  The #include is necessary for the #if that tests
the value that config(8) provides when intpm is configured.  It seems
that somebody forgot to rerun config at the time and commented this out
instead.

25 years agoUse snd_nxt, not rcv_nxt, when calculating the ISS during TIME_WAIT.
Bill Fenner [Sat, 6 Feb 1999 00:47:45 +0000 (00:47 +0000)]
Use snd_nxt, not rcv_nxt, when calculating the ISS during TIME_WAIT.
This was missed in the 4.4-Lite2 merge.

Noticed by: Mohan Parthasarathy <Mohan.Parthasarathy@eng.Sun.COM> and
jayanth@loc201.tandem.com (vijayaraghavan_jayanth)
on the tcp-impl mailing list.

25 years agoBruce gently nudged me into doing this right
Matt Jacob [Fri, 5 Feb 1999 22:54:37 +0000 (22:54 +0000)]
Bruce gently nudged me into doing this right

25 years agoAdd one more "escape" for marking internal variables.
Jordan K. Hubbard [Fri, 5 Feb 1999 22:25:13 +0000 (22:25 +0000)]
Add one more "escape" for marking internal variables.

25 years agoCorrect an "&" operator which should have been "&&".
John Polstra [Fri, 5 Feb 1999 22:24:26 +0000 (22:24 +0000)]
Correct an "&" operator which should have been "&&".

Submitted by: mjacob

25 years agoTotally change the way variables are accounted for in sysinstall.
Jordan K. Hubbard [Fri, 5 Feb 1999 22:15:52 +0000 (22:15 +0000)]
Totally change the way variables are accounted for in sysinstall.
Now we know which variables are internal and which need to be
backed to /etc/rc.conf.site.  rc.conf is not touched now.

Also kget kernel change information back properly and set up a loader.rc
file to use it.

25 years agoOops, really don't generate compiler warnings about missing braces.
Bruce Evans [Fri, 5 Feb 1999 16:58:22 +0000 (16:58 +0000)]
Oops, really don't generate compiler warnings about missing braces.

25 years agoDon't generate compiler warnings about missing braces.
Bruce Evans [Fri, 5 Feb 1999 16:49:18 +0000 (16:49 +0000)]
Don't generate compiler warnings about missing braces.

25 years agoUse O_RDONLY and O_RDWR for open() instead of number.
Jun Kuriyama [Fri, 5 Feb 1999 16:00:17 +0000 (16:00 +0000)]
Use O_RDONLY and O_RDWR for open() instead of number.

Encouraged by: Nate

25 years agos/card.conf/pccard.conf/
Jun Kuriyama [Fri, 5 Feb 1999 15:59:57 +0000 (15:59 +0000)]
s/card.conf/pccard.conf/

25 years agoStyle nit: sort elf_brand_inuse() prototype in with the others.
Mark Newton [Fri, 5 Feb 1999 13:52:15 +0000 (13:52 +0000)]
Style nit: sort elf_brand_inuse() prototype in with the others.
Prototype also needed to be added for !ELFCLASS32 case.

Submitted by: bde

25 years agoOops, the last commit contained a wrong patch. This is the correct one.
Kazutaka YOKOTA [Fri, 5 Feb 1999 12:58:40 +0000 (12:58 +0000)]
Oops, the last commit contained a wrong patch.  This is the correct one.

25 years agoFollowup to yokota's last commit:
Dag-Erling Smørgrav [Fri, 5 Feb 1999 12:40:16 +0000 (12:40 +0000)]
Followup to yokota's last commit:

 - va_mode_flags -> va_info.vi_flags
 - scanline width may be different from screen width

25 years ago- Don't assume the line length in the video memory is always the same as
Kazutaka YOKOTA [Fri, 5 Feb 1999 11:52:13 +0000 (11:52 +0000)]
- Don't assume the line length in the video memory is always the same as
  the screen width.
- Store the current video mode information in the `video_adapter' struct.
- The size of the `v_offscreensize' field in the VESA mode information
  block is u_int16, not u_int8.

25 years agoSync with sys/i386/isa/sio.c revision up to 1.228.
KATO Takenori [Fri, 5 Feb 1999 11:37:40 +0000 (11:37 +0000)]
Sync with sys/i386/isa/sio.c revision up to 1.228.

Submitted by: Takahashi Yoshihiro <nyan@wyvern.cc.kogakuin.ac.jp>

25 years agoAllow an alternate to rad_send_request() for programs that
Brian Somers [Fri, 5 Feb 1999 11:23:44 +0000 (11:23 +0000)]
Allow an alternate to rad_send_request() for programs that
don't wish to wait for the RADIUS server to respond.
Reviewed by: jdp

25 years agoBump MFS up to 2.88MB. Now there's room for quite a bit more stuff
Jordan K. Hubbard [Fri, 5 Feb 1999 10:09:11 +0000 (10:09 +0000)]
Bump MFS up to 2.88MB. Now there's room for quite a bit more stuff
(like a dhcp client?).

25 years agoNote that nextboot requires the legacy boot code.
Robert Nordier [Fri, 5 Feb 1999 09:58:39 +0000 (09:58 +0000)]
Note that nextboot requires the legacy boot code.

25 years agoUpdate snapshot build machine names.
Jordan K. Hubbard [Fri, 5 Feb 1999 09:54:59 +0000 (09:54 +0000)]
Update snapshot build machine names.

25 years agoIndicate that LKMs are obsolete, just to prevent confusion.
Jordan K. Hubbard [Fri, 5 Feb 1999 09:35:46 +0000 (09:35 +0000)]
Indicate that LKMs are obsolete, just to prevent confusion.

25 years agoIndicate that we're using XFree86 3.3.3.1
Jordan K. Hubbard [Fri, 5 Feb 1999 09:28:16 +0000 (09:28 +0000)]
Indicate that we're using XFree86 3.3.3.1

25 years agoquiet the alpha compiler
Matt Jacob [Fri, 5 Feb 1999 08:49:34 +0000 (08:49 +0000)]
quiet the alpha compiler

25 years agoadd a note about exclusive open
Matt Jacob [Fri, 5 Feb 1999 08:06:05 +0000 (08:06 +0000)]
add a note about exclusive open

25 years agodocument new features
Matt Jacob [Fri, 5 Feb 1999 07:59:07 +0000 (07:59 +0000)]
document new features

25 years agominor tweaks
Matt Jacob [Fri, 5 Feb 1999 07:57:26 +0000 (07:57 +0000)]
minor tweaks

25 years ago The elf loader sets the permissions on bss to VM_PROT_READ|VM_PROT_WRITE
Matthew Dillon [Fri, 5 Feb 1999 07:49:29 +0000 (07:49 +0000)]
The elf loader sets the permissions on bss to VM_PROT_READ|VM_PROT_WRITE
    rather then VM_PROT_ALL.  obreak, on the otherhand, uses VM_PROT_ALL.
    This prevents vm_map_insert() from being able to coalesce the heap and
    creates an extra map entry.  Since current architectures ignore
    VM_PROT_EXECUTE anyway, and since not having VM_PROT_EXECUTE on data/bss
    may provide protection in the future, obreak now uses read+write rather
    then all (r+w+x).

    This is an optimization, not a bug fix.

Submitted by: Alan Cox <alc@cs.rice.edu>

25 years agominor tweaks
Matt Jacob [Fri, 5 Feb 1999 07:48:26 +0000 (07:48 +0000)]
minor tweaks

25 years agoExtend unit numbers to a full 10 bits (split into sections
Matt Jacob [Fri, 5 Feb 1999 07:32:52 +0000 (07:32 +0000)]
Extend unit numbers to a full 10 bits (split into sections
of the minor). Establish and use a control mode open. Control
mode opens may open the device without locking, but are prohibited
from all but some ioctls. MTIOCGET always works. MTIOCERRSTAT
works, but the clearing of latched error status is contingent
upon whether another application has the device open, in which
case an interruptible perip acquire is done. MTSETBSIZ, MTSETDNSTY
and MTCOMP also require a periph aquire.

Relative fileno and blkno are tracked. Note that just about any
error will make these undefined, and if you space to EOD or use
hardware block positioning, these are also lost until the next
UNLOAD or REWIND.

Driver state is also tracked and recorded in the unit softc
to be passed back in mt_dsreg for a MTIOCGET call.
Thanks to Dan Strick for suggesting this.

Reintroduce 2 filemarks at EOD for all but QIC devices. I
really think it's wrong, but there is a lot of 3rd party
software that depends upon this (not the least of which is
tcopy). Introduce a SA_QUIRK_1FM to ensure that some devices
can be marked as only being able to do 1 FM at EOD.

At samount time force a load to BOT if we aren't mounted. If the
LOAD command fails, use the REWIND command (e.g., for the IBM 3590
which for some gawdawful reason doesn't support the LOAD (to BOT)
command).

Also at samount time, if you don't know fixed or variable, try to
*set* to one of the known fixed (or variable, for special case)
density codes. We only have to do this once per boot, so it's not
that painful. This is another way to try and figure out the wierd
QIC devices without having to quirk everything in the universe.

A substantial amount of cleanup as to what operations can and what
operations cannot be retried. Don't retry space operations if they
fail- it'll just lead to lossage.

Not yet done is invalidating mounts correctly after errors. ENOTIME.

25 years agoadd QIC 525,320,1320,3080 codes
Matt Jacob [Fri, 5 Feb 1999 07:19:23 +0000 (07:19 +0000)]
add QIC 525,320,1320,3080 codes

25 years agoCorrect a format mismatch on 64-bit architectures. This should
John Polstra [Fri, 5 Feb 1999 06:18:54 +0000 (06:18 +0000)]
Correct a format mismatch on 64-bit architectures.  This should
fix the erroneous values in the procfs "map" file on the Alpha.

25 years agoRemove some dead debug code
Greg Lehey [Fri, 5 Feb 1999 05:09:22 +0000 (05:09 +0000)]
Remove some dead debug code
build_write_request: get the check for failure right

25 years agoAdditional note on last rev: The rationale for this is to allow you
Mark Newton [Fri, 5 Feb 1999 03:47:47 +0000 (03:47 +0000)]
Additional note on last rev:  The rationale for this is to allow you
to run Solaris executables (or executables from any other ELF system)
directly off the CD-ROM without having to waste megabytes of disk
by copying them to another filesystem just to brand them.

25 years agoCreated sysctl kern.fallback_elf_brand. Defaults to "none", which will
Mark Newton [Fri, 5 Feb 1999 03:43:18 +0000 (03:43 +0000)]
Created sysctl kern.fallback_elf_brand.  Defaults to "none", which will
give the same behaviour produced before today.  If sysadmin sets it
to a valid ELF brand, ELF image activator will attempt to run unbranded
ELF exectutables as if they were branded with that value.

Suggested by: Dima Ruban <dima@best.net>

25 years agoAdd driver states that will be reported (for SCSI) in mt_dsreg.
Matt Jacob [Fri, 5 Feb 1999 02:47:30 +0000 (02:47 +0000)]
Add driver states that will be reported (for SCSI) in mt_dsreg.

25 years agoPrint relative (mt_fileno, mt_blkno) position, if known.
Matt Jacob [Fri, 5 Feb 1999 02:46:21 +0000 (02:46 +0000)]
Print relative (mt_fileno, mt_blkno) position, if known.
Print driver state if not NIL.

25 years agofinally document new commands
Matt Jacob [Fri, 5 Feb 1999 02:45:08 +0000 (02:45 +0000)]
finally document new commands

25 years agoRemove the FTP_PASSIVE_MODE "fix"; libftpio handles this.
Bill Fenner [Fri, 5 Feb 1999 01:01:17 +0000 (01:01 +0000)]
Remove the FTP_PASSIVE_MODE "fix"; libftpio handles this.

25 years agoPrint usage via fprintf(stderr, ..) instead of errx() to avoid progname prefix.
Archie Cobbs [Fri, 5 Feb 1999 00:42:14 +0000 (00:42 +0000)]
Print usage via fprintf(stderr, ..) instead of errx() to avoid progname prefix.
Submitted by: Philippe Charnier <charnier@xp11.frmug.org>

25 years agoMore nit-picking in the debugging instructions.
Greg Lehey [Fri, 5 Feb 1999 00:34:21 +0000 (00:34 +0000)]
More nit-picking in the debugging instructions.
Get the rc.conf description right.

25 years ago Fix race in pipe read code whereby a blocked lock can allow another
Matthew Dillon [Thu, 4 Feb 1999 23:50:49 +0000 (23:50 +0000)]
Fix race in pipe read code whereby a blocked lock can allow another
    process to sneak in and write to or close the pipe.  The read code
    enters a 'piperd' state after doing the lock operation without
    checking to see if the state changed, which can cause the process
    to wait forever.

    The code has also been documented more.

25 years agoreplace previous stupid comment with one more appropriate
Adam David [Thu, 4 Feb 1999 22:34:23 +0000 (22:34 +0000)]
replace previous stupid comment with one more appropriate
where it will be easily found

25 years agoCan't use elf_brand_inuse() here because iBCS2 doesn't use ELF. D'oh!
Mark Newton [Thu, 4 Feb 1999 21:21:38 +0000 (21:21 +0000)]
Can't use elf_brand_inuse() here because iBCS2 doesn't use ELF.  D'oh!
Inlined the same logic elf_brand_inuse() utilizes instead.

25 years agoAdded call to elf_brand_inuse() to prevent you from cutting your own
Mark Newton [Thu, 4 Feb 1999 21:20:13 +0000 (21:20 +0000)]
Added call to elf_brand_inuse() to prevent you from cutting your own
legs out from under you.

25 years agoadd back ctl for sa and add more units
Matt Jacob [Thu, 4 Feb 1999 20:35:07 +0000 (20:35 +0000)]
add back ctl for sa and add more units

25 years ago vp->v_object must be valid after normal flow of vfs_object_create()
Matthew Dillon [Thu, 4 Feb 1999 18:25:39 +0000 (18:25 +0000)]
vp->v_object must be valid after normal flow of vfs_object_create()
    completes, change if() to KASSERT().  This is not a bug, we are
    simplify clarifying and optimizing the code.

    In if/else in vfs_object_create(), the failure of both conditionals
    will lead to a NULL object.  Exit gracefully if this case occurs.
    ( this case does not normally occur, but needed to be handled ).

Obtained from: Eivind Eklund <eivind@FreeBSD.org>

25 years agoremind that apm is required in order for timekeeping to work
Adam David [Thu, 4 Feb 1999 18:08:55 +0000 (18:08 +0000)]
remind that apm is required in order for timekeeping to work

25 years ago Fix bug in a KASSERT I introduced in vm_page_qcollapse() rev 1.139.
Matthew Dillon [Thu, 4 Feb 1999 17:47:52 +0000 (17:47 +0000)]
Fix bug in a KASSERT I introduced in vm_page_qcollapse() rev 1.139.

    Since paging is in progress, page scan in vm_page_qcollapse() must be
    protected at atleast splbio() to prevent pages from being ripped out from
    under the scan.

25 years agoMake FICL_TRACE default.
Daniel C. Sobral [Thu, 4 Feb 1999 17:13:49 +0000 (17:13 +0000)]
Make FICL_TRACE default.

Add freebsd.fr, for FreeBSD specific forth source. Add $ and %, to
replace the lost @ and - functionality of include. $ has the opposite
behavior of @ though, since the default behavior was inverted.

25 years agoMake use of prototypes to silence warnings.
Daniel C. Sobral [Thu, 4 Feb 1999 17:06:46 +0000 (17:06 +0000)]
Make use of prototypes to silence warnings.

Change include() so it will be able to load files with forth code,
instead of just builtins. Remove #@- from the include section of the
help file, since they don't work in the new version of include, unless
BOOT_FORTH is not defined.

Change bf_run() so it will return the result returned by ficlExec(). Also,
make bf_run() push "interpret" to be executed by ficlExec(), since ficlExec()
doesn't do it by itself. (Things worked previously because nothing
recursed through ficlExec() by the way of bf_run()).

Change/extend comments on builtin behavior.

Search for "interpret" at the end of bf_init(), so /boot/boot.4th can
provide it's own version.

Remove dead code.