]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
25 years agoAdd FIODTYPE ioctl for getting d_flags (type) info on a device.
Brian Feldman [Fri, 27 Aug 1999 16:35:37 +0000 (16:35 +0000)]
Add FIODTYPE ioctl for getting d_flags (type) info on a device.

Okayed by: phk

25 years agoFix the child's return path from fork so that fork will return 0
Andrew Gallatin [Fri, 27 Aug 1999 14:47:23 +0000 (14:47 +0000)]
Fix the child's return path from fork so that fork will return 0
in the child.  This corrects a problem where linux/alpha binaries see
the child's return value of fork as the parent's pid.  This happens because
linux/alpha binaries apparently check the return value directly, rather
than looking for a non-zero value in a4, as *BSD & OSF/1 do.

Reviewed by:dfr@nlsystems.com

25 years agoRemove a 4.3BSDish anachronism that claimed EPERM for an attempt to
Joerg Wunsch [Fri, 27 Aug 1999 14:25:00 +0000 (14:25 +0000)]
Remove a 4.3BSDish anachronism that claimed EPERM for an attempt to
mknod() a pathname containg a ``character with the high-order bit
set''.

25 years agoFix conflicts. This brings in rev 1.2 (FreeBSD native support -- do not
David E. O'Brien [Fri, 27 Aug 1999 11:02:42 +0000 (11:02 +0000)]
Fix conflicts.  This brings in rev 1.2 (FreeBSD native support -- do not
search the multitude of extended include file paths that do not exist)
into EGCS 1.1.2.

25 years agoSmall whitespace change to fix run-over lines introduced in the previous
Sheldon Hearn [Fri, 27 Aug 1999 10:59:40 +0000 (10:59 +0000)]
Small whitespace change to fix run-over lines introduced in the previous
commit.

Reported by: bde

25 years agoFix conflicts. This brings the "printf0" addition into EGCS 1.1.2.
David E. O'Brien [Fri, 27 Aug 1999 10:58:48 +0000 (10:58 +0000)]
Fix conflicts.  This brings the "printf0" addition into EGCS 1.1.2.

25 years agoAdd -m option to specify a creation mode whcih is not affected by
Sheldon Hearn [Fri, 27 Aug 1999 10:40:27 +0000 (10:40 +0000)]
Add -m option to specify a creation mode whcih is not affected by
the umask of the calling process.

PR: 13365
Reported by: James Howard <howardjp@wam.umd.edu>
Reviewed by: bde

25 years agoDon't accept %q length specifiers in the kernel (more precisely,
David E. O'Brien [Fri, 27 Aug 1999 10:05:08 +0000 (10:05 +0000)]
Don't accept %q length specifiers in the kernel (more precisely,
if compiling with -fformat-extensions). Gcc's format checker never
actually supported %q length specifiers.  It treats %q as an alias
for %ll, which is correct if quad_t is long long (e.g., on i386's)
and broken otherwise (e.g., on alphas).

quad_t's currently should be printed in the same way that they
already need to be printed to avoid compiler warnings on all
supported systems: cast them to a standard type that is at least
as large (long or long long) and use the length specifier for that
(%l or %ll).  This is problematic since long long isn't standard
yet.  C9x's intmax_t should be implemented soon.

Don't accept %L length specifiers in the kernel either.  The only
legitimate ones are for long doubles, but the kernel doesn't even
support plain doubles.  (gcc bogusly accepts %Ld as an alias for
%lld, and it sometimes prints "q" in error messages about "ll" and
"L" length specifiers, becauses it represents all these specifiers
as 'q'.)

Submitted by: bde

25 years agoFixed kernel format checking:
David E. O'Brien [Fri, 27 Aug 1999 10:02:56 +0000 (10:02 +0000)]
Fixed kernel format checking:
- plain %r and %z were disallowed.  The hard NULs in the warnings were
  hopefully caused by disallowing of plain formats being nonsense.
- new formats for shortening to a byte were allowed, but even the libc
  printf doesn't support them.
- old %hr and %hz formats were allowed, but the kernel printf doesn't
  support them.  The kernel doesn't support %hd either, but this is
  harder to fix.

Submitted by: bde

25 years agoFix conflicts.
David E. O'Brien [Fri, 27 Aug 1999 10:01:05 +0000 (10:01 +0000)]
Fix conflicts.
This brings revs 1.2 (-fformat-extensions) and 1.3 (printf0())
into EGCS 1.1.2.

25 years agoFix conflicts.
David E. O'Brien [Fri, 27 Aug 1999 09:36:15 +0000 (09:36 +0000)]
Fix conflicts.

25 years ago- Retain the previous vty and history buffers when setting up
Kazutaka YOKOTA [Fri, 27 Aug 1999 09:20:41 +0000 (09:20 +0000)]
- Retain the previous vty and history buffers when setting up
  the graphics mode.  This was the behavior prior to syscons.c
  rev 1.278, but broken in scvidctl.c rev 1.1.

25 years ago- Set the correct value to va_line_width while in the ioctl
Kazutaka YOKOTA [Fri, 27 Aug 1999 09:19:50 +0000 (09:19 +0000)]
- Set the correct value to va_line_width while in the ioctl
  FBIO_SETLINEWIDTH.

25 years ago- Ouch! Do not assume the VESA BIOS will preserve the upper 16 bits
Kazutaka YOKOTA [Fri, 27 Aug 1999 09:18:42 +0000 (09:18 +0000)]
- Ouch! Do not assume the VESA BIOS will preserve the upper 16 bits
  of EAX and EBX.

25 years agoAdd ``-i'' option, which is the same as ``-''. The -i option is the
Sheldon Hearn [Fri, 27 Aug 1999 08:59:32 +0000 (08:59 +0000)]
Add ``-i'' option, which is the same as ``-''. The -i option is the
recommended option in the manpage, but the - option remains for
backward compatibility and is documented as such.

PR: 13363
Reported by: James Howard <howardjp@wam.umd.edu>
Reviewed by: bde

25 years agoDon't install /etc/manpath.config.sample.
Ruslan Ermilov [Fri, 27 Aug 1999 08:43:58 +0000 (08:43 +0000)]
Don't install /etc/manpath.config.sample.

25 years agoA few style changes (fixes hopefully) and some more tidying up. Fix (?)
Peter Wemm [Fri, 27 Aug 1999 08:32:54 +0000 (08:32 +0000)]
A few style changes (fixes hopefully) and some more tidying up.  Fix (?)
the volatile cast warnings.

25 years agoRemove some vestiges of devfs direct calls.
Julian Elischer [Fri, 27 Aug 1999 08:31:20 +0000 (08:31 +0000)]
Remove some vestiges of devfs direct calls.

25 years agoI misunderstood the failure mode - revert BINMODE to 4554. I'm still
Jordan K. Hubbard [Fri, 27 Aug 1999 07:47:17 +0000 (07:47 +0000)]
I misunderstood the failure mode - revert BINMODE to 4554.  I'm still
going to remove ppp from the installation options in 5 days if ppp is
still broken, however, as it hasn't worked at installation time
for several weeks now and it's only causing people to fill my mailbox
with questions.

25 years ago'build' calls 'stage1'
Roger Hardiman [Fri, 27 Aug 1999 07:41:25 +0000 (07:41 +0000)]
'build' calls 'stage1'
The check for the exit status of the 'stage1' script was broken.
Change "X$?" != "X0" to "X?" != "X0"

25 years agoAdd PHK's make_dev() into more places where DEVFS used to be
Julian Elischer [Fri, 27 Aug 1999 07:26:26 +0000 (07:26 +0000)]
Add PHK's make_dev() into more places where DEVFS used to be
hooked in directly.

Alpha change checked by: Matthew Jacob <mjacob@feral.com>
i4b ISDN changes checked by: Udo Schweigert <ust@cert.siemens.de>
    and Hellmuth Michaelis <hm@hcs.de>
PC98 changes checked by: Takahashi Yoshihiro <nyan@FreeBSD.org>

25 years agoDon't return 0 for an unknown ioctl (!). This was breaking ppp(8).
Peter Wemm [Fri, 27 Aug 1999 06:53:34 +0000 (06:53 +0000)]
Don't return 0 for an unknown ioctl (!).  This was breaking ppp(8).
Slight tidy up while here.

25 years agoFix a typo.
Jordan K. Hubbard [Fri, 27 Aug 1999 06:25:37 +0000 (06:25 +0000)]
Fix a typo.

Reported by: Terry Jones <terry@cliffs.ucsd.edu>

25 years agoI don't know who decided that an install mode of 4544 was correct, but
Jordan K. Hubbard [Fri, 27 Aug 1999 06:01:27 +0000 (06:01 +0000)]
I don't know who decided that an install mode of 4544 was correct, but
for a binary it's far from it.  This is why sysinstall is unable to run
ppp at startup time and causing weirdies in -current.

25 years agoRebuild the list of interface numbers to names if we're trying
Brian Somers [Thu, 26 Aug 1999 20:01:20 +0000 (20:01 +0000)]
Rebuild the list of interface numbers to names if we're trying
to evaluate an interface number that didn't previously exist.

25 years agoAdd readonly OID ``net.inet.tcp.tcbhashsize'' so it is possible to
Jonathan Lemon [Thu, 26 Aug 1999 19:52:17 +0000 (19:52 +0000)]
Add readonly OID ``net.inet.tcp.tcbhashsize'' so it is possible to
discover the size of the TCB hashtable on a running system.

25 years agoHandle the new Makefile infrastructure in the doc/ repository.
Nik Clayton [Thu, 26 Aug 1999 19:50:14 +0000 (19:50 +0000)]
Handle the new Makefile infrastructure in the doc/ repository.

Set ALLLANG to install all the languages, otherwise set DOC_LANG
to just the languages (and encodings) you want to install.

Default to building the html-split (lots of small HTML files, with
links to go between them), html (one big honking HTML file), and plain
text versions of the documentation.  Does not compress any of the docs
prior to doing the installation.

Testing and feedback:   Jack O'Neill <jack@germanium.xtalwind.net>,
Cockups and typos:      nik

25 years agoPull the FDP maintained documents (that's the FAQ, the Handbook, the
Nik Clayton [Thu, 26 Aug 1999 19:48:39 +0000 (19:48 +0000)]
Pull the FDP maintained documents (that's the FAQ, the Handbook, the
tutorials, and all the language translations thereof) out of here.  They
still install here, but the installation routine in doc/ sorts that
out.

25 years agoSync with -stable version
Roger Hardiman [Thu, 26 Aug 1999 18:36:36 +0000 (18:36 +0000)]
Sync with -stable version

25 years agoComment out references to ssh to allow builds to complete.
Roger Hardiman [Thu, 26 Aug 1999 17:58:58 +0000 (17:58 +0000)]
Comment out references to ssh to allow builds to complete.

25 years agoDon't follow symlinks on coredumps.
Dima Ruban [Thu, 26 Aug 1999 17:32:48 +0000 (17:32 +0000)]
Don't follow symlinks on coredumps.

Reviewed by: dillon && security-officer

25 years agoDon't suggest sysctl(8) as a means for discovering what filesystem types
Sheldon Hearn [Thu, 26 Aug 1999 16:47:03 +0000 (16:47 +0000)]
Don't suggest sysctl(8) as a means for discovering what filesystem types
are available; suggest lsvfs(1) instead.

Reported by: Neil Blakey-Milner <nbm@rucus.ru.ac.za>

25 years ago1. Add support for months and years in relative time spec (now + 1 year)
Nick Sayer [Thu, 26 Aug 1999 16:18:38 +0000 (16:18 +0000)]
1. Add support for months and years in relative time spec (now + 1 year)

2. Rip out dateadd() and re-do the plus operator handler to use mktime()
instead (per wollman).

Reviewed by: wollman

25 years agoVALIASED doesn't exist any more.
Poul-Henning Kamp [Thu, 26 Aug 1999 15:49:54 +0000 (15:49 +0000)]
VALIASED doesn't exist any more.

25 years agoUpdate so it applies cleanly
Roger Hardiman [Thu, 26 Aug 1999 15:17:24 +0000 (15:17 +0000)]
Update so it applies cleanly

25 years agoUpdate patch so it applies cleanly
Roger Hardiman [Thu, 26 Aug 1999 15:13:00 +0000 (15:13 +0000)]
Update patch so it applies cleanly

25 years agoSimplify the handling of VCHR and VBLK vnodes using the new dev_t:
Poul-Henning Kamp [Thu, 26 Aug 1999 14:53:31 +0000 (14:53 +0000)]
Simplify the handling of VCHR and VBLK vnodes using the new dev_t:

        Make the alias list a SLIST.

        Drop the "fast recycling" optimization of vnodes (including
        the returning of a prexisting but stale vnode from checkalias).
        It doesn't buy us anything now that we don't hardlimit
        vnodes anymore.

        Rename checkalias2() and checkalias() to addalias() and
        addaliasu() - which takes dev_t and udev_t arg respectively.

        Make the revoke syscalls use vcount() instead of VALIASED.

        Remove VALIASED flag, we don't need it now and it is faster
        to traverse the much shorter lists than to maintain the
        flag.

        vfs_mountedon() can check the dev_t directly, all the vnodes
        point to the same one.

Print the devicename in specfs/vprint().

Remove a couple of stale LFS vnode flags.

Remove unimplemented/unused LK_DRAINED;

25 years agoInitialize the dev->si_bsize fields.
Poul-Henning Kamp [Thu, 26 Aug 1999 14:46:10 +0000 (14:46 +0000)]
Initialize the dev->si_bsize fields.

Submitted by: tegge
Reviewed by: phk

25 years ago. mention the existence of various flags in the synopsis, so it's more
Joerg Wunsch [Thu, 26 Aug 1999 13:41:43 +0000 (13:41 +0000)]
. mention the existence of various flags in the synopsis, so it's more
  obvious to the casual reader
. add flag 0x80 description

25 years agoChanged to get the page size using a sysctl rather than using DEFAULT_PAGE_SIZE
Roger Hardiman [Thu, 26 Aug 1999 11:11:48 +0000 (11:11 +0000)]
Changed to get the page size using a sysctl rather than using DEFAULT_PAGE_SIZE

Suggested by: Alan Cox

25 years agoThis commit was generated by cvs2svn to compensate for changes in r50397,
David E. O'Brien [Thu, 26 Aug 1999 09:30:50 +0000 (09:30 +0000)]
This commit was generated by cvs2svn to compensate for changes in r50397,
which included commits to RCS files with non-trunk default branches.

25 years agoVirgin import of gcc from EGCS 1.1.2
David E. O'Brien [Thu, 26 Aug 1999 09:30:50 +0000 (09:30 +0000)]
Virgin import of gcc from EGCS 1.1.2

25 years agoFix DMA macros to work on alpha.
Doug Rabson [Thu, 26 Aug 1999 08:41:41 +0000 (08:41 +0000)]
Fix DMA macros to work on alpha.

25 years agoDon't schedule the next transaction if the pipe is being aborted.
Doug Rabson [Thu, 26 Aug 1999 08:40:55 +0000 (08:40 +0000)]
Don't schedule the next transaction if the pipe is being aborted.

25 years agoMake the behaviour of `read -e', ie. treating backslashes as special,
Thomas Gellekum [Thu, 26 Aug 1999 08:16:27 +0000 (08:16 +0000)]
Make the behaviour of `read -e', ie. treating backslashes as special,
the default. Add -r option for the read builtin to reverse this.

PR: 13274
Reviewed by: cpiazza, hoek, sheldonh

25 years agoAligh mbufs to sizeof(long), not 4
Brian Somers [Thu, 26 Aug 1999 06:37:53 +0000 (06:37 +0000)]
Aligh mbufs to sizeof(long), not 4

25 years agoProperly re-enable the transmitter in the TX error handler.
Bill Paul [Thu, 26 Aug 1999 05:31:53 +0000 (05:31 +0000)]
Properly re-enable the transmitter in the TX error handler.

25 years agoHandle buses with multiple PHYs correctly.
Bill Paul [Thu, 26 Aug 1999 05:30:33 +0000 (05:30 +0000)]
Handle buses with multiple PHYs correctly.

25 years agoChange asf function to use the new name of the list in kern_linker.c.
Greg Lehey [Thu, 26 Aug 1999 04:01:37 +0000 (04:01 +0000)]
Change asf function to use the new name of the list in kern_linker.c.
See revision 1.35 of sys/kern/kern_linker.c.

25 years agoread_drive: Remove references to vp->v_lastr, which is going away.
Greg Lehey [Thu, 26 Aug 1999 03:27:21 +0000 (03:27 +0000)]
read_drive: Remove references to vp->v_lastr, which is going away.

Reviewed-by: dillon

25 years agocomplete_rqe: explicitly reinitialize b.b_iodone. This shouldn't have
Greg Lehey [Thu, 26 Aug 1999 03:26:39 +0000 (03:26 +0000)]
complete_rqe: explicitly reinitialize b.b_iodone.  This shouldn't have
      changed, but we've seen some weird cases of b_iodone
      being NULL, and they don't make any sense.

Suffered-by: Bernd Walter <ticso@cicely.de>
25 years agoconfig_plex: don't allocate a rangelock struct.
Greg Lehey [Thu, 26 Aug 1999 03:25:38 +0000 (03:25 +0000)]
config_plex: don't allocate a rangelock struct.

lockrange: correctly expand rangelock struct, including expanding a
   null struct.  Previously lockrange would attempt to lock a
   NULL pointer under these circumstances.

Reported-by: Ian Freislich <iang@uunet.co.za>

25 years agoUpdate and tidy descriptions of info -V.
Greg Lehey [Thu, 26 Aug 1999 03:24:02 +0000 (03:24 +0000)]
Update and tidy descriptions of info -V.
Describe use of -w option with start.
Tidy up descriptions of scale factor suffixes.

25 years agoAdd a verbose mode to show what files are being copied.
Michael Haro [Thu, 26 Aug 1999 02:44:56 +0000 (02:44 +0000)]
Add a verbose mode to show what files are being copied.
Idea taken from obrien.

Reviewed by: obrien

25 years agounifdef -DINTERNAL_LS - it's too useful to be off by default. If anyone
Peter Wemm [Thu, 26 Aug 1999 00:45:36 +0000 (00:45 +0000)]
unifdef -DINTERNAL_LS - it's too useful to be off by default.  If anyone
really dislikes this, we could add a switch to disable it at runtime and
check in popen.c.

25 years agoUse .p2align to ensure consistant a.out/elf alignment. I'd have used
Peter Wemm [Wed, 25 Aug 1999 23:50:03 +0000 (23:50 +0000)]
Use .p2align to ensure consistant a.out/elf alignment.  I'd have used
SUPERALIGN_TEXT, but this is inline assembler and after cpp has run.
Inspired by bde's comments on linux_locore.s.

25 years agoIt would help if things that MUST be 32 bit aligned were really 32 bit
Peter Wemm [Wed, 25 Aug 1999 23:42:02 +0000 (23:42 +0000)]
It would help if things that MUST be 32 bit aligned were really 32 bit
aligned.  If I recall correctly, this is to ensure apic_imen can be
accessed in a single bus cycle.  Also, use TEXT_ALIGN rather than a
.align 2 (which means 2 byte align on ELF and 4 byte align on a.out)

25 years agoReplace ".align 2" with "ALIGN_TEXT" as per recent discussion re: the linux
Mark Newton [Wed, 25 Aug 1999 22:55:56 +0000 (22:55 +0000)]
Replace ".align 2" with "ALIGN_TEXT" as per recent discussion re: the linux
emulator.

Suggested by: bde

25 years agoMake a place to store the devfs hook for the block device, as the same
Julian Elischer [Wed, 25 Aug 1999 22:50:12 +0000 (22:50 +0000)]
Make a place to store the devfs hook for the block device, as the same
specinfo is used to identify both raw and block version sof a device.

25 years agoLooks like I missed a little white space error, fixed.
Rodney W. Grimes [Wed, 25 Aug 1999 21:49:37 +0000 (21:49 +0000)]
Looks like I missed a little white space error, fixed.

25 years agoFix some mdoc(7) style inconsistences
Brian Somers [Wed, 25 Aug 1999 21:39:40 +0000 (21:39 +0000)]
Fix some mdoc(7) style inconsistences
Submitted by: Alexey M. Zelkin" <phantom@cris.net>

25 years agoAdd -e option to usage().
Doug White [Wed, 25 Aug 1999 21:31:08 +0000 (21:31 +0000)]
Add -e option to usage().

25 years agoMention our support of the xp1000 family
Andrew Gallatin [Wed, 25 Aug 1999 21:14:03 +0000 (21:14 +0000)]
Mention our support of the xp1000 family

25 years agoDon't getopt -M option and don't process it unless ENABLE_AMQ_MOUNT is on.
David E. O'Brien [Wed, 25 Aug 1999 20:06:03 +0000 (20:06 +0000)]
Don't getopt -M option and don't process it unless ENABLE_AMQ_MOUNT is on.

Obtained from: am-utils-6.0.1s11

25 years agoUse ALIGN_TEXT macro for alignment to avoid ambiguity.
Marcel Moolenaar [Wed, 25 Aug 1999 19:04:39 +0000 (19:04 +0000)]
Use ALIGN_TEXT macro for alignment to avoid ambiguity.

Pointed out by: bde

25 years agowhat is a devuce?
Matt Jacob [Wed, 25 Aug 1999 18:05:52 +0000 (18:05 +0000)]
what is a devuce?

25 years agoStyle clean-up:
Sheldon Hearn [Wed, 25 Aug 1999 16:01:45 +0000 (16:01 +0000)]
Style clean-up:

* All variables are now embraced: ${foo}

* All comparisons against some value now take the form:
  [ "${foo}" ? "value" ]
  where ? is a comparison operator

* All empty string tests now take the form:
  [ -z "${foo}" ]

* All non-empty string tests now take the form:
  [ -n "${foo}" ]

Submitted by: jkh

25 years agoFix linux_newlstat in that it doesn't return the attributes of its containing
Marcel Moolenaar [Wed, 25 Aug 1999 15:23:54 +0000 (15:23 +0000)]
Fix linux_newlstat in that it doesn't return the attributes of its containing
directory. Also, update arguments of NDINIT for both newstat and newlstat.

While I'm at it, fix style bugs in all {s|ls|fs}tat syscalls.

Reported by: bde

25 years agoDocument apmd stuff.
Mitsuru IWASAKI [Wed, 25 Aug 1999 15:13:24 +0000 (15:13 +0000)]
Document apmd stuff.

25 years agoRemove obsolete BUGS section which describes something which stopped
Jordan K. Hubbard [Wed, 25 Aug 1999 14:38:58 +0000 (14:38 +0000)]
Remove obsolete BUGS section which describes something which stopped
being relevant when we ripped out the BSD/VAX code.

Submitted by: Guy Harris <gharris@flashcom.net>

25 years agoDeclaration of strtouq() is not missing from stdlib.h, so don't redeclare
Dmitrij Tejblum [Wed, 25 Aug 1999 14:22:20 +0000 (14:22 +0000)]
Declaration of strtouq() is not missing from stdlib.h, so don't redeclare
it.

25 years agoFix {g|s}etgroups semantics. We use cr_groups[0] to hold egid. This means that
Marcel Moolenaar [Wed, 25 Aug 1999 14:11:01 +0000 (14:11 +0000)]
Fix {g|s}etgroups semantics. We use cr_groups[0] to hold egid. This means that
egid will be twice in the set and that setting cr_groups[0] will change egid.
This is simply solved by ignoring cr_groups[0]. That is; linux_getgroups does
not return cr_groups[0] and linux_setgroups does not touch it.

Noticed by: bde
Brought to my attention by: sheldonh

25 years agoFix lingo: 'is not implemented or obsoleted' is not the same as 'is obsoleted
Marcel Moolenaar [Wed, 25 Aug 1999 13:17:05 +0000 (13:17 +0000)]
Fix lingo: 'is not implemented or obsoleted' is not the same as 'is obsoleted
or not implemented'.

25 years agoIntroduce vn_isdisk(struct vnode *vp) function, and use it to test for diskness.
Poul-Henning Kamp [Wed, 25 Aug 1999 12:24:39 +0000 (12:24 +0000)]
Introduce vn_isdisk(struct vnode *vp) function, and use it to test for diskness.

25 years agoAdd a couple of missing but unimportant break; statements.
Poul-Henning Kamp [Wed, 25 Aug 1999 11:44:11 +0000 (11:44 +0000)]
Add a couple of missing but unimportant break; statements.

25 years agoChange all UNIMPL syscalls to STD and add them to linux_dummy. Now we always
Marcel Moolenaar [Wed, 25 Aug 1999 11:19:03 +0000 (11:19 +0000)]
Change all UNIMPL syscalls to STD and add them to linux_dummy. Now we always
know if and when an unimplemented or obsoleted syscall is being used. Make the
message more end-user friendly.

And as long as we're here, rename some unimplemeted syscalls (linux_phys ->
linux_umount2, linux_vm86 -> linux_vm86old, linux_new_vm86 -> linux_vm86).

Change prototype for linux_newuname from `struct linux_newuname_t *' into
`struct linux_new_utsname *'. This change is reflected in linux.h and
linux_misc.c.

25 years agoChange all UNIMPL syscalls to STD and add them to linux_dummy. Now we always
Marcel Moolenaar [Wed, 25 Aug 1999 11:17:36 +0000 (11:17 +0000)]
Change all UNIMPL syscalls to STD and add them to linux_dummy. Now we always
know if and when an unimplemented or obsoleted syscall is being used. Make the
message more end-user friendly.

And as long as we're here, rename some unimplemeted syscalls (linux_phys ->
linux_umount2, linux_vm86 -> linux_vm86old, linux_new_vm86 -> linux_vm86).

Change prototype for linux_newuname from `struct linux_newuname_t *' into
`struct linux_new_utsname *'. This change is reflected in linux.h and
linux_misc.c.

Lastly, make line-continuation and indentation more uniform.

25 years agoFix stack misalignment for Linux binaries caused by `linux_sigcode' not being
Marcel Moolenaar [Wed, 25 Aug 1999 09:05:29 +0000 (09:05 +0000)]
Fix stack misalignment for Linux binaries caused by `linux_sigcode' not being
a multiple of 4 bytes in size. This solves the recent SIGBUS errors for
glibc2.1 configurations.

Explained by: bde

25 years agoSort cross-references. Move mount(8) down the list where it belongs, in
Chris Costello [Wed, 25 Aug 1999 08:45:31 +0000 (08:45 +0000)]
Sort cross-references.  Move mount(8) down the list where it belongs, in
particular.

25 years agoSome very minor changes to the ISA dma code.
Doug Rabson [Wed, 25 Aug 1999 08:25:58 +0000 (08:25 +0000)]
Some very minor changes to the ISA dma code.

25 years agoRename 'bios_jmp' to 'bios16_jmp' to make it clear what it's related to.
Mike Smith [Wed, 25 Aug 1999 06:56:37 +0000 (06:56 +0000)]
Rename 'bios_jmp' to 'bios16_jmp' to make it clear what it's related to.

25 years agoUse the far jump for the base of the page arithmatic rather than the
Peter Wemm [Wed, 25 Aug 1999 06:44:32 +0000 (06:44 +0000)]
Use the far jump for the base of the page arithmatic rather than the
calling function, otherwise Bad Things Happen(tm) when bios16_call is
not in the same page as bios_jmp.

Reviewed by:  msmith

25 years agoMake DEVFS use PHK's specinfo struct as the source of dev_t and devsw.
Julian Elischer [Wed, 25 Aug 1999 04:55:20 +0000 (04:55 +0000)]
Make DEVFS use PHK's specinfo struct as the source of dev_t and devsw.

In lookup() however it's the other way around as we need to supply the
dev_t for the vnode, so devfs still has a copy of it stashed away.

Sourcing it from the vnode in the vnops however is useful as it makes
a lot of the code almost the same as that in specfs.

25 years agoTry to get /usr/include/readline back into the snapshots.
David E. O'Brien [Wed, 25 Aug 1999 03:40:55 +0000 (03:40 +0000)]
Try to get /usr/include/readline back into the snapshots.

Submitted by: bde

25 years agoDon't open the swap file. The open descriptor for it hasn't been used
Bruce Evans [Wed, 25 Aug 1999 03:01:54 +0000 (03:01 +0000)]
Don't open the swap file.  The open descriptor for it hasn't been used
for over 5 years since we switched to using procfs for kvm_uread().
This cleanup was motivated by recent breakage of the default swap file
(/dev/drum) when swapon() has not been called.

25 years agoMake DEVFS ops reflect some of the changes that have recently occured in the
Julian Elischer [Wed, 25 Aug 1999 02:04:40 +0000 (02:04 +0000)]
Make DEVFS ops reflect some of the changes that have recently occured in the
specfs version of the same ops. Merging with phk's dev_t changes is not yet
complete.

i.e. devfs still uses it's own devsw pointer rather than following
devsw(a_vp->v_rdev), and it's own copy of the dev_t.

This fixes some broken actions re: syscons and DEVFS due to bitrot in
devfs vops.

25 years agoAdd missing quote.
Doug White [Wed, 25 Aug 1999 01:12:55 +0000 (01:12 +0000)]
Add missing quote.

25 years agoFix comment to match reality..
Julian Elischer [Wed, 25 Aug 1999 00:26:34 +0000 (00:26 +0000)]
Fix comment to match reality..
vop_strategy gets a vnode argument these days.

25 years agoFix patch so it applies without asking for a filename.
Doug White [Tue, 24 Aug 1999 23:28:26 +0000 (23:28 +0000)]
Fix patch so it applies without asking for a filename.

25 years agoDevfs now adds itself to the hook that PHK supplies.
Julian Elischer [Tue, 24 Aug 1999 20:30:15 +0000 (20:30 +0000)]
Devfs now adds itself to the hook that PHK supplies.
block devices are still not handled correctly, though a kludge
may make them "ok".

25 years agoInitialize the si_bsize fields for the MFS bogodevices.
Poul-Henning Kamp [Tue, 24 Aug 1999 18:35:33 +0000 (18:35 +0000)]
Initialize the si_bsize fields for the MFS bogodevices.

(This broke MFS rootfs and thereby installation)

25 years agoRemove unneeded prototypes.
Poul-Henning Kamp [Tue, 24 Aug 1999 18:31:06 +0000 (18:31 +0000)]
Remove unneeded prototypes.

25 years agoWork around a bad design in some PnP BIOS code whereby the BIOS can reach
Mike Smith [Tue, 24 Aug 1999 16:42:08 +0000 (16:42 +0000)]
Work around a bad design in some PnP BIOS code whereby the BIOS can reach
off the top of our constructed stack segment while it's trying to copy a
maximally-sized PnP argument frame around.

25 years agoMove ipxrouted.dmp from /tmp to /var/log to make it more difficult to hijack.
John Hay [Tue, 24 Aug 1999 13:15:39 +0000 (13:15 +0000)]
Move ipxrouted.dmp from /tmp to /var/log to make it more difficult to hijack.

PR: 13286

25 years agoFix bug introduced in rev 1.28, which causes kernel build to break for
Sheldon Hearn [Tue, 24 Aug 1999 08:39:41 +0000 (08:39 +0000)]
Fix bug introduced in rev 1.28, which causes kernel build to break for
the case where DEBUG is defined but not DIAGNOSTIC. ffs_checkblk is
declared conditionally on DIAGNOSTIC, not DEBUG.

PR: 13314
Reviewed by: bde

25 years agoCosmetic: Correct the Id string.
Alan Cox [Tue, 24 Aug 1999 06:54:20 +0000 (06:54 +0000)]
Cosmetic: Correct the Id string.

Submitted by: Peter Jeremy <jeremyp@gsmx07.alcatel.com.au>

25 years agoMake the comments nice, short-but-sweet XXX format.
Brian Feldman [Tue, 24 Aug 1999 06:10:14 +0000 (06:10 +0000)]
Make the comments nice, short-but-sweet XXX format.

Submitted by: sheldonh

25 years agoWhen the SYSINIT() was removed, it was replaced with a make_dev on-demand
Brian Feldman [Tue, 24 Aug 1999 05:58:35 +0000 (05:58 +0000)]
When the SYSINIT() was removed, it was replaced with a make_dev on-demand
creation of /dev/drum via calling swapon. However, the make_dev has a
bogus (insofar that it hasn't been added yet) cdevsw, so later we end
up crashing with a null pointer dereference on the swap vp's specinfo.
The specinfo points to a dev_t with a major of 254 (uninitialized), and
we get a crash on its d_strategy being called.

The simple solution to this is to call cdevsw_add before the make_dev
is ever used. This fixes the panic which occurred upon swapping.

25 years agoCorrect the mode information for the VGA mode X.
Kazutaka YOKOTA [Tue, 24 Aug 1999 04:26:54 +0000 (04:26 +0000)]
Correct the mode information for the VGA mode X.
- This mode uses four planes rather than one.
- "# of bytes in a scan line" = "# of pixels in the line"/2

25 years agoThis cleans up all the white space errors so that the next
Rodney W. Grimes [Tue, 24 Aug 1999 03:51:34 +0000 (03:51 +0000)]
This cleans up all the white space errors so that the next
commit is easier to understand.