]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoThe syscall names are string constants, so make them consts.
David Malone [Tue, 29 Oct 2002 15:47:06 +0000 (15:47 +0000)]
The syscall names are string constants, so make them consts.

21 years agoAdd scr2txt to the required list of doc ports for "make release". I
Murray Stokely [Tue, 29 Oct 2002 15:13:28 +0000 (15:13 +0000)]
Add scr2txt to the required list of doc ports for "make release".  I
think this list of the bare minimal ports required should maybe be
moved to textproc/docproj/Makefile, so that when we add a new
dependency there, we can just add it to a required or optional list,
rather than add it there and forget about it here. ;)

Noticed by: jhay
Forgotten by: nik

21 years agoUse a single style for share/doc/ makefiles.
Ruslan Ermilov [Tue, 29 Oct 2002 15:01:59 +0000 (15:01 +0000)]
Use a single style for share/doc/ makefiles.

21 years agoCosmetics.
Ruslan Ermilov [Tue, 29 Oct 2002 15:00:10 +0000 (15:00 +0000)]
Cosmetics.

21 years agoSort SUBDIR list.
Ruslan Ermilov [Tue, 29 Oct 2002 14:58:00 +0000 (14:58 +0000)]
Sort SUBDIR list.
Don't mention SGML docs which now live elsewhere.

21 years agobsd.doc.mk changes:
Ruslan Ermilov [Tue, 29 Oct 2002 14:56:09 +0000 (14:56 +0000)]
bsd.doc.mk changes:

Don't gratuitously pipe thru a cat(1) if NODOCCOMPRESS.

Only create _stamp.extra when necessary.

Get rid of SOELIMPP and OBJS.

Use Groff version of soelim(1); we need its -I option
for the following to work.

Don't needlessly chdir to SRCDIR.  Only a few documents
need CD_HACK, and those that need it either use refer(1)
or .PSPIC macro which internally uses the .psbb call.

21 years agoThe agp device also exists on alpha and ia64.
Yoshihiro Takahashi [Tue, 29 Oct 2002 14:53:57 +0000 (14:53 +0000)]
The agp device also exists on alpha and ia64.

Submitted by: dfr

21 years agoDon't call warn(3).
Poul-Henning Kamp [Tue, 29 Oct 2002 14:45:43 +0000 (14:45 +0000)]
Don't call warn(3).

21 years agoDon't forget to '\n'-terminate new entries. This unbreaks chpass -a.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 13:58:42 +0000 (13:58 +0000)]
Don't forget to '\n'-terminate new entries.  This unbreaks chpass -a.

Submitted by: joerg

21 years ago- Some sections are for not only i386 but pc98.
Yoshihiro Takahashi [Tue, 29 Oct 2002 12:51:48 +0000 (12:51 +0000)]
- Some sections are for not only i386 but pc98.
- The agp driver is for i386 only.

21 years agoDon't try to dereference a NULL pw pointer. This would happen when
Joerg Wunsch [Tue, 29 Oct 2002 12:45:32 +0000 (12:45 +0000)]
Don't try to dereference a NULL pw pointer.  This would happen when
attempting to use chpass -a.

21 years agoFix `pkg_add -r' by backing out revs 1.34-1.38.
David E. O'Brien [Tue, 29 Oct 2002 12:17:43 +0000 (12:17 +0000)]
Fix `pkg_add -r' by backing out revs 1.34-1.38.
Revs 1.37-8 produce a bus error in some environments.
Revs 1.34-6 do not bus error, but write corrupted files.

21 years agoWhitespace cleanup
Dag-Erling Smørgrav [Tue, 29 Oct 2002 12:14:31 +0000 (12:14 +0000)]
Whitespace cleanup

21 years agoSince make release is toast anyway, add wood to the pyre:
Poul-Henning Kamp [Tue, 29 Oct 2002 12:13:36 +0000 (12:13 +0000)]
Since make release is toast anyway, add wood to the pyre:

This significantly rewamps libdisks discovery of existing disk
layout.

Please send me reports if this does not work as expected on
i386 or sparc64 platforms.

I need to sort out alpha, pc98 and ia64 (in that order) before
testing on those platforms make a lot of sense.

Belived to work for: i386 sparc64
Unknown state: pc98 alpha ia64

21 years agoIn VarHead(), look into the correct (modified) string.
Juli Mallett [Tue, 29 Oct 2002 12:11:56 +0000 (12:11 +0000)]
In VarHead(), look into the correct (modified) string.

21 years agoUpdate for OpenSSH 3.5p1.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 10:18:00 +0000 (10:18 +0000)]
Update for OpenSSH 3.5p1.

21 years agoAdd a fortune for people who have read their classics.
Poul-Henning Kamp [Tue, 29 Oct 2002 10:16:06 +0000 (10:16 +0000)]
Add a fortune for people who have read their classics.

21 years agoResolve conflicts.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 10:16:02 +0000 (10:16 +0000)]
Resolve conflicts.

21 years agoProtect against tag expansion + fix some brainos.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 10:12:51 +0000 (10:12 +0000)]
Protect against tag expansion + fix some brainos.

21 years agoSome tricks I use when I upgrade.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 09:56:16 +0000 (09:56 +0000)]
Some tricks I use when I upgrade.

21 years agoCorrect shell code to expand globs in FREEBSD-Xlist
Dag-Erling Smørgrav [Tue, 29 Oct 2002 09:55:28 +0000 (09:55 +0000)]
Correct shell code to expand globs in FREEBSD-Xlist

21 years agoMore cruft.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 09:54:53 +0000 (09:54 +0000)]
More cruft.

21 years agoMention that the left-hand side of the comparison
Ruslan Ermilov [Tue, 29 Oct 2002 09:48:49 +0000 (09:48 +0000)]
Mention that the left-hand side of the comparison
conditional must always be a variable expansion.

Obtained from: PMake

Do not lie that debugging .for loops is a no-op.

21 years agoThis commit was generated by cvs2svn to compensate for changes in r106121,
Dag-Erling Smørgrav [Tue, 29 Oct 2002 09:43:00 +0000 (09:43 +0000)]
This commit was generated by cvs2svn to compensate for changes in r106121,
which included commits to RCS files with non-trunk default branches.

21 years agoVendor import of OpenSSH-portable 3.5p1.
Dag-Erling Smørgrav [Tue, 29 Oct 2002 09:43:00 +0000 (09:43 +0000)]
Vendor import of OpenSSH-portable 3.5p1.

21 years agoMostly support alternate manpage zipping.
David E. O'Brien [Tue, 29 Oct 2002 09:20:22 +0000 (09:20 +0000)]
Mostly support alternate manpage zipping.

21 years agoRemove unnecessary inclusion of <rune.h> to make it obvious that this file
Tim J. Robbins [Tue, 29 Oct 2002 09:03:57 +0000 (09:03 +0000)]
Remove unnecessary inclusion of <rune.h> to make it obvious that this file
does not use the deprecated rune system.

21 years agoLower a priority of "session drop" messages.
Maxim Konovalov [Tue, 29 Oct 2002 08:53:14 +0000 (08:53 +0000)]
Lower a priority of "session drop" messages.

Requested by: Eugene Grosbein <eugen@kuzbass.ru>
MFC after: 3 days

21 years agoTweak the warning language.
David E. O'Brien [Tue, 29 Oct 2002 08:41:12 +0000 (08:41 +0000)]
Tweak the warning language.

21 years agoFix various minor issues.
Poul-Henning Kamp [Tue, 29 Oct 2002 07:39:18 +0000 (07:39 +0000)]
Fix various minor issues.
Don't explode on 'write' because we access a pointer we just freed.

21 years agoSparc64 will not need the same hacks as alpha did. Hopefully alpha wont
Poul-Henning Kamp [Tue, 29 Oct 2002 07:38:05 +0000 (07:38 +0000)]
Sparc64 will not need the same hacks as alpha did.  Hopefully alpha wont
need them either.

21 years agothe 'd' partition hasn't been magic for years, so allocate it in natural
Poul-Henning Kamp [Tue, 29 Oct 2002 07:37:11 +0000 (07:37 +0000)]
the 'd' partition hasn't been magic for years, so allocate it in natural
order instead of last.

21 years agoAdd the write_spar64_disk.c to make life easier for testers.
Poul-Henning Kamp [Tue, 29 Oct 2002 07:35:36 +0000 (07:35 +0000)]
Add the write_spar64_disk.c to make life easier for testers.

This file depends on some major surgery in the rest of libdisk which is
not yet committed.

21 years agoDuh. #ifdef CTASSERT so we don't break user-land software.
Poul-Henning Kamp [Tue, 29 Oct 2002 06:43:57 +0000 (06:43 +0000)]
Duh.  #ifdef CTASSERT so we don't break user-land software.

21 years agoFix winChkName() to match when the last slot contains nothing but the
Semen Ustimenko [Tue, 29 Oct 2002 01:03:44 +0000 (01:03 +0000)]
Fix winChkName() to match when the last slot contains nothing but the
terminating zero (it was treated as length missmatch). The mtools create
such slots if the name len is the product of 13 (max number of unicode
chars fitting in directory slot).

MFC after: 1 week

21 years agoConvert most instances of "FreeBSD" to &os;.
Bruce A. Mah [Tue, 29 Oct 2002 00:59:36 +0000 (00:59 +0000)]
Convert most instances of "FreeBSD" to &os;.

Suggested by: ue

21 years agoo Change .Nd to be less redundant and more consistent with upcoming policy
Chris Costello [Tue, 29 Oct 2002 00:21:03 +0000 (00:21 +0000)]
o Change .Nd to be less redundant and more consistent with upcoming policy
  man pages.
o Move a misplaced comment.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Labs

21 years agoRemove unnecessary <sys/malloc.h> include.
Semen Ustimenko [Tue, 29 Oct 2002 00:20:47 +0000 (00:20 +0000)]
Remove unnecessary <sys/malloc.h> include.

Submitted by: marius@alchemy.franken.de
MFC after: 1 week

21 years agoSplit var.c into var.c and var_modify.c and move all the modification funcs
Juli Mallett [Mon, 28 Oct 2002 23:33:57 +0000 (23:33 +0000)]
Split var.c into var.c and var_modify.c and move all the modification funcs
to var_modify.c, for readability.  constify some low hanging fruit (string
manipulation functions) and the upper layers appropriately.  No longer use
the private strstr(3) implementation, while changing string code.

Tested by:      lots of successful make buildworld.

21 years agoPut a CTASSERT on the size of struct sun_disklabel.
Poul-Henning Kamp [Mon, 28 Oct 2002 23:31:33 +0000 (23:31 +0000)]
Put a CTASSERT on the size of struct sun_disklabel.

Sponsored by: DARPA & NAI Labs.

21 years agoScoop out examples illustrating the label text format and refer to
Chris Costello [Mon, 28 Oct 2002 23:06:04 +0000 (23:06 +0000)]
Scoop out examples illustrating the label text format and refer to
maclabel(7) instead.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Labs

21 years agoCorrectly spell "te" as "mls".
Chris Costello [Mon, 28 Oct 2002 23:03:25 +0000 (23:03 +0000)]
Correctly spell "te" as "mls".

21 years agoAdd a man page describing the Mandatory Access Control label format.
Chris Costello [Mon, 28 Oct 2002 22:54:54 +0000 (22:54 +0000)]
Add a man page describing the Mandatory Access Control label format.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Labs

21 years agoAdd the remaning part of the new libdisk interaction.
Poul-Henning Kamp [Mon, 28 Oct 2002 22:43:54 +0000 (22:43 +0000)]
Add the remaning part of the new libdisk interaction.

WARNING:  This is not a published interface, it is a stopgap measure for
WARNING:  libdisk so we can get 5.0-R out of the door.

Sponsored by: DARPA & NAI Labs

21 years agoAdd support for the new libdisk interaction.
Poul-Henning Kamp [Mon, 28 Oct 2002 22:42:20 +0000 (22:42 +0000)]
Add support for the new libdisk interaction.

Sponsored by: DARPA & NAI Labs.

21 years agoFix the wizardmodes 'write' command to not explode: Don't reference
Poul-Henning Kamp [Mon, 28 Oct 2002 22:40:49 +0000 (22:40 +0000)]
Fix the wizardmodes 'write' command to not explode:  Don't reference
the name in the disk tree we just deleted when we reopen.

21 years agoEnd of translation of section 1.5
Sebastien Gioria [Mon, 28 Oct 2002 21:46:50 +0000 (21:46 +0000)]
End of translation of section 1.5

21 years agoAdd note: cvs checkout -P needed for fresh tree.
Warner Losh [Mon, 28 Oct 2002 21:33:10 +0000 (21:33 +0000)]
Add note: cvs checkout -P needed for fresh tree.

21 years agoTrim extraneous #else and #endif MAC comments per style(9).
Robert Watson [Mon, 28 Oct 2002 21:17:53 +0000 (21:17 +0000)]
Trim extraneous #else and #endif MAC comments per style(9).

21 years agoCorrect a typo in a previously commented include entry that was made
Robert Watson [Mon, 28 Oct 2002 19:50:06 +0000 (19:50 +0000)]
Correct a typo in a previously commented include entry that was made
visible in the recent commit.

21 years agoRemove all reference to 'struct oldmac', since it's no longer required
Robert Watson [Mon, 28 Oct 2002 19:44:05 +0000 (19:44 +0000)]
Remove all reference to 'struct oldmac', since it's no longer required
with the new VFS/EA semantics in the MAC framework.  Move the per-policy
structures out to per-policy include files, removing all policy-specific
defines and structures out of the base framework includes and
implementation, making mac_biba and mac_mls entirely self-contained.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoClarify style(9) WRT comments following #endif, #else.
Robert Watson [Mon, 28 Oct 2002 19:33:22 +0000 (19:33 +0000)]
Clarify style(9) WRT comments following #endif, #else.

The closing comment is required only for long conditionally defined
code sections, with the exception of lint cases.  Attempt to document
also the logic for using '!' before the SOMETIMESSOMETHGINGHERE.
The goal of these comments is to make complex cases more
comprehensible, not to require them in all cases.  The rules here are
derived from behavior used in 90+% of the kernel source code.

Reviewed by and discussed with: jhb, bde, mike

21 years agoAdd a return type for mac_biba_high_single(), apparently lost in an
Robert Watson [Mon, 28 Oct 2002 19:18:29 +0000 (19:18 +0000)]
Add a return type for mac_biba_high_single(), apparently lost in an
earlier merge.  It's a 'static int'.

21 years agoRename mac_biba_subject_equal_ok() to mac_biba_subject_privileged()
Robert Watson [Mon, 28 Oct 2002 19:17:39 +0000 (19:17 +0000)]
Rename mac_biba_subject_equal_ok() to mac_biba_subject_privileged()
to evolve the notion of Biba privilege a bit.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoZero the trusted_interface buffer before starting parsing.
Robert Watson [Mon, 28 Oct 2002 19:16:44 +0000 (19:16 +0000)]
Zero the trusted_interface buffer before starting parsing.
Print a warning if a requested interface name is longer than
IFNAMSIZ.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoAn inappropriate ASSERT slipped in during the recent merge of the
Robert Watson [Mon, 28 Oct 2002 18:53:53 +0000 (18:53 +0000)]
An inappropriate ASSERT slipped in during the recent merge of the
reboot checking; remove.

21 years agoReduce namespace pollution to userland.
Scott Long [Mon, 28 Oct 2002 17:08:07 +0000 (17:08 +0000)]
Reduce namespace pollution to userland.

Spotted-by: bde

21 years agoFix a bug in the cherry-picker kkey generator routine.
Poul-Henning Kamp [Mon, 28 Oct 2002 11:06:58 +0000 (11:06 +0000)]
Fix a bug in the cherry-picker kkey generator routine.

WARNING:  You need to backup and restore the _unencrypted_ contents
WARNING:  of your GBDE disks when you take this update!

Sponsored by: DARPA & NAI Labs.

21 years agolibfetch is now WARNS5 clean in the non-SSL case.
Dag-Erling Smørgrav [Mon, 28 Oct 2002 10:37:31 +0000 (10:37 +0000)]
libfetch is now WARNS5 clean in the non-SSL case.

21 years agoCross-reference putc(3).
Tim J. Robbins [Mon, 28 Oct 2002 10:35:18 +0000 (10:35 +0000)]
Cross-reference putc(3).

21 years agoAllow the admin to specify a different NAS identifier than the hostname.
Dag-Erling Smørgrav [Mon, 28 Oct 2002 10:28:46 +0000 (10:28 +0000)]
Allow the admin to specify a different NAS identifier than the hostname.

Submitted by: Boris Kovalenko <boris@ntmk.ru>

21 years agoFix an off-by-one error (> where >= should have been used) which caused
Dag-Erling Smørgrav [Mon, 28 Oct 2002 10:19:03 +0000 (10:19 +0000)]
Fix an off-by-one error (> where >= should have been used) which caused
_fetch_writev() to incorrectly report EPIPE in certain cases.

Also fix a number of const warnings by using __DECONST(), plus a signed /
unsigned comparison by casting the rhs to ssize_t.

Submitted by: fenner, Craig Rodrigues <rodrigc@attbi.com>

21 years agoSimplify by using inference rules properly.
Tim J. Robbins [Mon, 28 Oct 2002 09:05:43 +0000 (09:05 +0000)]
Simplify by using inference rules properly.

21 years agoEmptify.
Ruslan Ermilov [Mon, 28 Oct 2002 08:44:59 +0000 (08:44 +0000)]
Emptify.

21 years agoMove SCSI drivers to third floppy disk (drivers.flp) to give "make
Murray Stokely [Mon, 28 Oct 2002 08:34:48 +0000 (08:34 +0000)]
Move SCSI drivers to third floppy disk (drivers.flp) to give "make
release" a chance of finishing on the Alpha platform.

The actual split between drivers on disks 2 and 3 should be optimized
so that most users don't need the third disk, but for now, I'm just
trying to get it working.

21 years agoHandle boundary cases more correctly; mblen(s, 0) and mbtowc(NULL, s, 0)
Tim J. Robbins [Mon, 28 Oct 2002 08:24:46 +0000 (08:24 +0000)]
Handle boundary cases more correctly; mblen(s, 0) and mbtowc(NULL, s, 0)
return -1 regardless of what s points to, mbtowc(&w, s, 1) sets w to a
null wide character when s points to a null byte. This seems to be closer
to what most other implementations do, but the C99 standard contradicts
itself for these cases.

21 years agoAdd more compatibility junk.
Poul-Henning Kamp [Mon, 28 Oct 2002 07:50:47 +0000 (07:50 +0000)]
Add more compatibility junk.

21 years agoClose a race window in kse_create(): signal delivered after SIGPENDING call
David Xu [Mon, 28 Oct 2002 07:37:06 +0000 (07:37 +0000)]
Close a race window in kse_create(): signal delivered after SIGPENDING call
but before we call kse_link().

21 years agoRemove stale information from these two manpage, and point the readers
Luigi Rizzo [Mon, 28 Oct 2002 07:24:58 +0000 (07:24 +0000)]
Remove stale information from these two manpage, and point the readers
to the one up-to-date page which is ipfw(8).

MFC after: 3 days

21 years agoMisc fixes from Chris Pepper, plus additional explainations on
Luigi Rizzo [Mon, 28 Oct 2002 07:03:56 +0000 (07:03 +0000)]
Misc fixes from Chris Pepper, plus additional explainations on
dummynet operation.

MFC after: 3 days

21 years agoRemove a line that reflected behavior that does not actually exist.
Chris Costello [Mon, 28 Oct 2002 03:46:32 +0000 (03:46 +0000)]
Remove a line that reflected behavior that does not actually exist.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs

21 years agoHandle hints for the atspeaker device.
Matthew N. Dodd [Mon, 28 Oct 2002 02:00:09 +0000 (02:00 +0000)]
Handle hints for the atspeaker device.
Document same.

21 years agoRemove mf.a from sapic_read() and sapic_write(). We only care
Marcel Moolenaar [Mon, 28 Oct 2002 01:59:27 +0000 (01:59 +0000)]
Remove mf.a from sapic_read() and sapic_write(). We only care
about ordering and not acceptance. The removal of mf.a leaves
behind the mf that accompanied it.

21 years agoUp WARNS to 3 if not building with crypto.
Bill Fenner [Mon, 28 Oct 2002 01:41:28 +0000 (01:41 +0000)]
Up WARNS to 3 if not building with crypto.

Approved by: des

21 years agoRemove mf.a (the acceptance form of the memory fence instruction)
Marcel Moolenaar [Mon, 28 Oct 2002 01:00:57 +0000 (01:00 +0000)]
Remove mf.a (the acceptance form of the memory fence instruction)
from all low-level bus space support functions. There's no need
to actually force the read/write to be accepted by the platform
before we can do anything else. We still have the mf instruction
there, which forces ordering. This too is not required given the
semantices of the bus space I/O functions, but it's not at all
clear to me if there are any poorly written device drivers that
depend on the strict ordering by the processor. The motto here is
to take small steps...

21 years agoMake vmstat -i work:
Marcel Moolenaar [Mon, 28 Oct 2002 00:50:39 +0000 (00:50 +0000)]
Make vmstat -i work:
o  Properly set the pointer to the counter for each interrupt and
   update the intrnames table.
o  Remove Alpha cruft from intrcnt.h.
o  Create INTRNAME_LEN as the single entity that defines the width
   of the names in the intrnames table (incl. terminatinf '\0').

21 years agoCreate a small library function, check_utility_compat(3), to determine
Garrett Wollman [Mon, 28 Oct 2002 00:15:43 +0000 (00:15 +0000)]
Create a small library function, check_utility_compat(3), to determine
whether a named utility should behave in FreeBSD 4.x-compatible mode
or in a standard mode (default standard).  The configuration is done
malloc(3)-style, with either an environment variable or a symlink.

Update expr(1) to use this new interface.

21 years agoFix a case in kern_rename() where a vn_finished_write() call was
Ian Dowse [Sun, 27 Oct 2002 23:23:51 +0000 (23:23 +0000)]
Fix a case in kern_rename() where a vn_finished_write() call was
missed. This bug has been present since the vn_start_write() and
vn_finished_write() calls were first added in revision 1.159. When
the case is triggered, any attempts to create snapshots on the
filesystem will deadlock and also prevent further write activity
on that filesystem.

21 years agoIn ipi_send(), perform a mf instruction prior to initiating the IPI.
Marcel Moolenaar [Sun, 27 Oct 2002 23:00:46 +0000 (23:00 +0000)]
In ipi_send(), perform a mf instruction prior to initiating the IPI.
This guarantees that loads and stores emitted before the fence are
made visible before the IPI becomes pended.
Remove the mf.a instruction after initiating the IPI. There's no
guarantee that the IPI becomes pended prior to subsequent reads or
writes. Even if there was a guarantee, it would mostly be without
any benefit.

21 years agoSerial + Parallel + Ethernet + NFS section translation
Sebastien Gioria [Sun, 27 Oct 2002 20:26:56 +0000 (20:26 +0000)]
Serial + Parallel + Ethernet + NFS section translation

21 years agoTake care of the case for the default speed (no -s option) also.
Nate Lawson [Sun, 27 Oct 2002 19:44:57 +0000 (19:44 +0000)]
Take care of the case for the default speed (no -s option) also.

21 years agoImplement the new 1003.1-2001 pathconf() keys, including the Advisory
Garrett Wollman [Sun, 27 Oct 2002 18:09:49 +0000 (18:09 +0000)]
Implement the new 1003.1-2001 pathconf() keys, including the Advisory
Information option.  Other filesystem implementations should do something
similar.

With advice from: mckusick, phk

21 years agoChange the way support for asynchronous I/O is indicated to applications
Garrett Wollman [Sun, 27 Oct 2002 18:07:41 +0000 (18:07 +0000)]
Change the way support for asynchronous I/O is indicated to applications
to conform to 1003.1-2001.  Make it possible for applications to actually
tell whether or not asynchronous I/O is supported.

Since FreeBSD's aio implementation works on all descriptor types, don't
call down into file or vnode ops when [f]pathconf() is asked about
_PC_ASYNC_IO; this avoids the need for every file and vnode op to know about
it.

21 years agoAs promised, downgrade the #error into a #warning.
Garrett Wollman [Sun, 27 Oct 2002 18:03:53 +0000 (18:03 +0000)]
As promised, downgrade the #error into a #warning.

21 years agoUpdate limits and configuration parameters for 1003.1/TC1/D6.
Garrett Wollman [Sun, 27 Oct 2002 18:03:02 +0000 (18:03 +0000)]
Update limits and configuration parameters for 1003.1/TC1/D6.
Implement new sysconf keys.  Change the implenentation of
_SC_ASYNCHRONOUS_IO in preparation for the next set of changes.

Move some limits which had been in <sys/syslimits.h> to <limits.h> where
they belong.  They had only ever been in syslimits.h to provide for the
kernel implementation of the CTL_USER MIB branch, which went away with
newsysctl years ago.  (There is a #error in <sys/syslimits.h> which I
will downgrade in the next commit.)

21 years agoAdd used #include <limits.h>.
Garrett Wollman [Sun, 27 Oct 2002 17:46:53 +0000 (17:46 +0000)]
Add used #include <limits.h>.

21 years agoDo not include <sys/syslimits.h> directly; it is not intended for general
Garrett Wollman [Sun, 27 Oct 2002 17:44:33 +0000 (17:44 +0000)]
Do not include <sys/syslimits.h> directly; it is not intended for general
consumption.

21 years agoMove speed units conversion to right before it is used. This catches the
Nate Lawson [Sun, 27 Oct 2002 17:43:22 +0000 (17:43 +0000)]
Move speed units conversion to right before it is used.  This catches the
case when the -s option is not used.

21 years agoDon't complain about not knowing the remote file size when working in
Dag-Erling Smørgrav [Sun, 27 Oct 2002 17:33:08 +0000 (17:33 +0000)]
Don't complain about not knowing the remote file size when working in
quiet mode.

PR: bin/37079
Submitted by: Nicolas Rachinsky <list@rachinsky.de>

21 years agoDon peril sensitive sun glasses and change the default system call vector
Jake Burkholder [Sun, 27 Oct 2002 17:22:43 +0000 (17:22 +0000)]
Don peril sensitive sun glasses and change the default system call vector
for sparc64 from trap #9 to trap #65.  This is one of the ABI "blessed"
system call vectors and is different from any other system that we might
want to emulate, making the emulation easier by reducing the number of
code paths that need to be shared.  Compatibility with old applications
is provided with COMPAT_FREEBSD4.
Add defines for a few special traps that we may need to implement for
compatibility with 32bit applications, and add comments on which vectors
are used for what in other systems, and which are available.
Pass magic flags to trap() for deprecated or unimplemented system call
vectors so they will deliver SIGSYS instead of SIGILL.

This piggy backs nicely with the recent sigaction(2) system call number
change, and provided the rules are followed for upgrading past it, this
change should not be noticed.

21 years agoSlight amendment to rev 1.34: instead of considering any short read an
Dag-Erling Smørgrav [Sun, 27 Oct 2002 17:20:49 +0000 (17:20 +0000)]
Slight amendment to rev 1.34: instead of considering any short read an
error, only report an error if no data was read at all (unless len was
0 to start with).  Otherwise, the final read of practically any transfer
will end in a fatal error.

21 years agoDon't roll our own clean target, the default one
Maxime Henrion [Sun, 27 Oct 2002 17:06:03 +0000 (17:06 +0000)]
Don't roll our own clean target, the default one
works better.

21 years agoMerged various changes from src/sbin/fdisk/fdisk.c revisions up to 1.66.
Yoshihiro Takahashi [Sun, 27 Oct 2002 16:43:59 +0000 (16:43 +0000)]
Merged various changes from src/sbin/fdisk/fdisk.c revisions up to 1.66.

21 years agoIntroduce _fetch_writev(), which is the conn_t version of writev(2). In
Dag-Erling Smørgrav [Sun, 27 Oct 2002 16:11:21 +0000 (16:11 +0000)]
Introduce _fetch_writev(), which is the conn_t version of writev(2).  In
the SSL case, it is no different from the old _fetch_write(), but in the
non-SSL case it uses writev(2) to send the entire vector as a single
packet (provided it can fit in one packet).  Implement _fetch_write()
and _fetch_putln() in terms of _fetch_writev().

This should improve performance in the non-SSL case (by reducing protocol
overhead) and solve the problem where too-smart-for-their-own-good
firewalls reject FTP packets that do not end in CRLF.

PR: bin/44123
Submitted by: fenner

21 years agoCentrally manage enforcement of {reboot,swapon,sysctl} using the
Robert Watson [Sun, 27 Oct 2002 15:50:49 +0000 (15:50 +0000)]
Centrally manage enforcement of {reboot,swapon,sysctl} using the
mac_enforce_system toggle, rather than several separate toggles.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoEliminate two cases of undefined behaviour: total in _fetch_write() was
Dag-Erling Smørgrav [Sun, 27 Oct 2002 15:43:40 +0000 (15:43 +0000)]
Eliminate two cases of undefined behaviour: total in _fetch_write() was
not initialized before use, and _http_growbuf() did not return a value
on success.

Reported by: Peter Edwards <pmedwards@eircom.net>
MFC after: 2 weeks

21 years agoWhitespace and indentation cleanup.
Dag-Erling Smørgrav [Sun, 27 Oct 2002 15:32:51 +0000 (15:32 +0000)]
Whitespace and indentation cleanup.

21 years agoForgot to disable alarms after fetchXGet() in previous commit.
Dag-Erling Smørgrav [Sun, 27 Oct 2002 15:32:06 +0000 (15:32 +0000)]
Forgot to disable alarms after fetchXGet() in previous commit.

21 years agoAdd an ETA timer that kicks in after 30 seconds.
Dag-Erling Smørgrav [Sun, 27 Oct 2002 15:15:13 +0000 (15:15 +0000)]
Add an ETA timer that kicks in after 30 seconds.

Re-add alarm(2) calls around the calls to fetchStat(3) and fetchXGet(3),
since these calls can still time out on DNS lookups or TCP connect(2).

Remove the alarm(2) calls in the main loop, since all methods properly
handle transfer timeouts (as opposed to connection timeouts).

Set the sigalrm flag if a timeout occurs in the main loop.

Move the signal: label up a little so we still set the atime and mtime
when the transfer times out or is interrupted, so that restarted transfers
will work as expected (as long as the file still exists).

MFC after: 2 weeks

21 years agoBack out the previous commit, and fix the bug rather than try to hide its
Dag-Erling Smørgrav [Sun, 27 Oct 2002 15:08:21 +0000 (15:08 +0000)]
Back out the previous commit, and fix the bug rather than try to hide its
symptoms: make timeouts and short transfers fatal, and set errno to an
appropriate value (ETIMEDOUT for a timeout, EPIPE for a short transfer).

MFC after: 2 weeks