]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoAdd more libc stuff.
Jake Burkholder [Sun, 18 Nov 2001 04:40:30 +0000 (04:40 +0000)]
Add more libc stuff.

Obtained from: netbsd

22 years agoCatch up to jmpbuf changes. Define offsets used from assmebly language
Jake Burkholder [Sun, 18 Nov 2001 04:35:09 +0000 (04:35 +0000)]
Catch up to jmpbuf changes.  Define offsets used from assmebly language
in another file so that it is easy to see what they are and keep them in
sync with the headers (grafted from the kernel generated assym.s).

22 years agoRemove fork and vfork wrappers. The kernel does the right thing.
Jake Burkholder [Sun, 18 Nov 2001 04:31:36 +0000 (04:31 +0000)]
Remove fork and vfork wrappers.  The kernel does the right thing.

22 years agoAdd .type and .size pseudo-ops to system call macros so that dynamic binaries
Jake Burkholder [Sun, 18 Nov 2001 04:29:20 +0000 (04:29 +0000)]
Add .type and .size pseudo-ops to system call macros so that dynamic binaries
will link.  They must be used for weak symbols as well as strong ones.

22 years agoMake jmpbuf the same size as ucontext_t so that it can be passed
Jake Burkholder [Sun, 18 Nov 2001 04:04:35 +0000 (04:04 +0000)]
Make jmpbuf the same size as ucontext_t so that it can be passed
to sigreturn.

Obtained from: alpha

22 years ago1. Split fp.h into fp.h and fsr.h so that the latter can be included
Jake Burkholder [Sun, 18 Nov 2001 04:00:22 +0000 (04:00 +0000)]
1. Split fp.h into fp.h and fsr.h so that the latter can be included
   in asm files.
2. Temporarily cause subnormal operands in floating point operations
   to be treated as zeros so that comlpetion of the operation does not
   need to be emulated.
3. Catch fp_exception_other and correctly skip over the unfinished
   instruction, but basically ignore them.  Emulating the instruction
   is not yet supported.
4. Zero td_retval[1] as well in syscall().

Submitted by: tmm (2, 3)

22 years agoAvoid missing ticks and hardclock stopping.
Jake Burkholder [Sun, 18 Nov 2001 03:47:30 +0000 (03:47 +0000)]
Avoid missing ticks and hardclock stopping.

Submitted by: tmm

22 years agoCatch up to new constants. (These commit messages should have a song.)
Jake Burkholder [Sun, 18 Nov 2001 03:45:13 +0000 (03:45 +0000)]
Catch up to new constants.  (These commit messages should have a song.)

22 years ago- M_ZERO already sets bif_dlist to zero; there is no need to
Andrew R. Reiter [Sun, 18 Nov 2001 03:41:20 +0000 (03:41 +0000)]
- M_ZERO already sets bif_dlist to zero; there is no need to
  do it again.

22 years ago1. Remove kdbframe. Bad idea.
Jake Burkholder [Sun, 18 Nov 2001 03:41:12 +0000 (03:41 +0000)]
1. Remove kdbframe.  Bad idea.
2. Add a TF_DONE macro, which fiddles a trapframe to make the retry on
   return from traps act like a done (advance past the trapping
   instruction instead of re-executing).
3. Flush the windows before entering the debugger, since it is no
   longer done in the breakpoint trap vector.
4. Print a warning if trace <pid> is attempted, it is not yet implemented.
5. Print traps better and decode system calls in traces.

Submitted by: rwatson (4)

22 years agoImplement SET. Set execption.s 1.12.
Jake Burkholder [Sun, 18 Nov 2001 03:31:01 +0000 (03:31 +0000)]
Implement SET.  Set execption.s 1.12.

Submitted by: tmm

22 years ago1. Convert the tstate saved in the pcb to a pstate and test for PSTATE_PEF
Jake Burkholder [Sun, 18 Nov 2001 03:28:28 +0000 (03:28 +0000)]
1. Convert the tstate saved in the pcb to a pstate and test for PSTATE_PEF
   to determine if a process is using floating point. in order to avoid
   sign extending a 13 bit immediate.
2. We don't need to context switch cwp anymore, it is better to just
   fiddle the save tstate on return from traps.  See exception.s 1.10
   and 1.12.
3. Completely remove pcb_cwp.
4. Implement vmapbuf, vunmapbuf and vm_fault_quick.  Completely remove
   TODOs from vm_machdep.c (yay!).

Submitted by: tmm (1, 3, 4)
Obtained from: existing archs (4)

22 years agoImplement hw.machine and hw.model sysctls.
Jake Burkholder [Sun, 18 Nov 2001 03:05:56 +0000 (03:05 +0000)]
Implement hw.machine and hw.model sysctls.

Submitted by: tmm

22 years ago1. Remove bootinfo and just pass loader metadata to the kernel.
Jake Burkholder [Sun, 18 Nov 2001 03:02:53 +0000 (03:02 +0000)]
1. Remove bootinfo and just pass loader metadata to the kernel.
2. Remove mcontext.mc_sp, it is redundant.  Adjust spare space to make
   ucontext_t a nice size.
3. Raise pil in the debugger.

Submitted by: tmm (3)

22 years ago1. Implement ascopyto() and ascopyfrom() for copying to an alternate address
Jake Burkholder [Sun, 18 Nov 2001 02:47:26 +0000 (02:47 +0000)]
1. Implement ascopyto() and ascopyfrom() for copying to an alternate address
   space from kernel space and from an alternate address space to kernel
   space.
2. Remove the unused and unprototyped physcopy() and physzero() and replace
   with the more versatile ascopy() and aszero(), inspired by the above.
   These can be used to copy and zero physical pages of memory without mapping
   them into kernel space first.
3. Use magic numbers for the offsets in the jmpbuf structure like other
   platforms.
4. Use SET.

Submitted by:  tmm (1, 4)

22 years ago1. Fix a bug where the offsets of the alignment and mmu fault recorvery code
Jake Burkholder [Sun, 18 Nov 2001 02:33:37 +0000 (02:33 +0000)]
1. Fix a bug where the offsets of the alignment and mmu fault recorvery code
   in the window trap vectors were mixed up.  All this did is cause unnecesary
   traps and look wierd in traces.  Superfluous traps happen a lot in normal
   operation, so we are rather good at recovering from them.
2. Store the arguments for a ktr trace in the right place.
3. Use a generic trap vector for breakpoints.  It should not be special.
4. Save the frame pointer in the trap frame for kernel traps if DDB is compiled
   in, otherwsie we don't save the out registers for kernel traps and stack
   traces can't go through nested traps.
5. Apply the same fix to the return from kernel mode trap code as for user
   mode traps.  Ensure that the window we're returning to is the same one
   that we restore to by fiddling the cwp in the saved tstate.  This requires
   that we transfer the values loaded from the trap frame into alternate
   globals before restore-ing, but doing so is not very expensive and not
   worth worrying about.  Not changing the saved cwp can result in the register
   values magically changing on return from traps if we happen to have slept
   and the windows don't work out exactly the same.  Fix the trace just before
   the retry to account for different register usage.
6. Use a SET macro for loading address constants rather than a variation of
   set and setx.  set only works for 32 bit constants, while setx works for
   64 bit constants as well, but produces bloated code when unnecessary.
   Gas always generates the canonical 2 register, 6 instruction form, even
   when it could be optimized; set uses 1 register and 2 instructions.  At
   the moment we assume that the kernel binary is below 4GB so set is
   always sufficient, but the macro allows it to be configured.  Note that
   this has nothing to do with 32 vs. 64 bit address space, it only applies
   to addresses of symbols which are known at compile/link time.

Submitted by: tmm (6)

22 years agoRemove a sentence from the BUGS section that claims non page-aligned
Ian Dowse [Sun, 18 Nov 2001 00:47:45 +0000 (00:47 +0000)]
Remove a sentence from the BUGS section that claims non page-aligned
offsets don't work. It should really be documented that the returned
pointer can be in the middle of a fully-valid page when the offset
is not page-aligned, but I couldn't come up with suitable wording.

PR: kern/22754

22 years agoCorrect typo.
Dima Dorfman [Sun, 18 Nov 2001 00:33:12 +0000 (00:33 +0000)]
Correct typo.

PR: 32066
Submitted by: Peter Avalos <pavalos@theshell.com>

22 years agoAdd support for making the Coda /dev/cfsN devices.
Ian Dowse [Sun, 18 Nov 2001 00:20:08 +0000 (00:20 +0000)]
Add support for making the Coda /dev/cfsN devices.

PR: conf/22695
Submitted by: Kenneth Ingham <ingham@i-pi.com>

22 years agoFix a large number of -Wall, -Wformat and -W compiler warnings.
Ian Dowse [Sat, 17 Nov 2001 23:48:21 +0000 (23:48 +0000)]
Fix a large number of -Wall, -Wformat and -W compiler warnings.
These were mainly missing casts or wrong format strings in printf
statements, but there were also missing includes, unused variables,
functions and arguments.

The choice of `long' vs `int' still seems almost random in a lot
of places though.

22 years agoAdd missing Nvidia chips.
Sheldon Hearn [Sat, 17 Nov 2001 22:54:06 +0000 (22:54 +0000)]
Add missing Nvidia chips.

Obtained from: http://www.yourvote.com/pci/

22 years agoGive a more useful diagnostic when an extraneous hard link to a
Ian Dowse [Sat, 17 Nov 2001 22:46:36 +0000 (22:46 +0000)]
Give a more useful diagnostic when an extraneous hard link to a
directory is encountered. This includes the full path of the
directory that will be removed if the user answers "y" to the
"REMOVE?" question.

PR: bin/226851
Submitted by: KOIE Hide <hide@koie.org>
MFC after: 1 week

22 years agoAdd two new accounts/groups for sendmail:
Gregory Neil Shapiro [Sat, 17 Nov 2001 21:24:45 +0000 (21:24 +0000)]
Add two new accounts/groups for sendmail:

smmsp - sendmail 8.12 operates as a set-group-ID binary (instead of
set-user-ID).  This new user/group will be used for command line
submissions.  UID/GID 25 is suggested in the sendmail documentation and has
been adopted by other operating systems such as OpenBSD and Solaris 9.

mailnull - The default value for DefaultUser is now set to the uid and gid
of the first existing user mailnull, sendmail, or daemon that has a
non-zero uid.  If none of these exist, sendmail reverts back to the old
behavior of using uid 1 and gid 1.  Currently FreeBSD uses daemon for
DefaultUser but I would prefer not to use an account used by other
programs, hence the addition of mailnull.  UID/GID 26 has been chosen for
this user.

This was discussed on -arch on October 18-19, 2001.

MFC after: 1 week

22 years agoRemove config.sgml from dependencies. Forgot this in my last commit.
Bruce A. Mah [Sat, 17 Nov 2001 20:53:23 +0000 (20:53 +0000)]
Remove config.sgml from dependencies.  Forgot this in my last commit.

22 years agoNo UserConfig under -CURRENT (at least so I've been led to believe)
Bruce A. Mah [Sat, 17 Nov 2001 20:28:44 +0000 (20:28 +0000)]
No UserConfig under -CURRENT (at least so I've been led to believe)
so take out the text that discusses it.

22 years agoAdd a comment that the chroot area cannot reside on a "nodev"-mounted
Bruce A. Mah [Sat, 17 Nov 2001 19:54:22 +0000 (19:54 +0000)]
Add a comment that the chroot area cannot reside on a "nodev"-mounted
filesystem.

PR: 12712
Submitted by: Sergei Laskavy <laskavy@Berkeley.Gambit.Msk.SU>

22 years agoMake test(1) a builtin command of our sh(1) for efficiency. The
Akinori MUSHA [Sat, 17 Nov 2001 19:10:11 +0000 (19:10 +0000)]
Make test(1) a builtin command of our sh(1) for efficiency.  The
binary size increase is 3,784 bytes (about 0.6%).

I don't drop the printf builtin while I'm here because some /etc/rc.*
scripts seem to use it before mounting /usr where printf(1) resides.

Reviewed by: arch (sheldonh)
Inspired by: NetBSD, ksh
Clued by: ume (on how the printf builtin is used)

22 years agoFix missing holdsock()->fgetsock()
Matthew Dillon [Sat, 17 Nov 2001 18:43:13 +0000 (18:43 +0000)]
Fix missing holdsock()->fgetsock()

Submitted by: Hisashi Hiramoto <hiramoto@phys.chs.nihon-u.ac.jp>

22 years agoBe very generous with timeouts for synchronize cache. We may wait a very long
Matt Jacob [Sat, 17 Nov 2001 18:26:00 +0000 (18:26 +0000)]
Be very generous with timeouts for synchronize cache. We may wait a very long
time in the cases where it really sends the drive out to lunch, but it also
allows us to catch very wierd edge cases of strange drives that might take
a very long time (emulated disk drives over a network, e.g.).

22 years agoThe Infortrend IFT-3102 multihost U2 SCSI to U2 SCSI controller
Ian Dowse [Sat, 17 Nov 2001 17:43:06 +0000 (17:43 +0000)]
The Infortrend IFT-3102 multihost U2 SCSI to U2 SCSI controller
doesn't support the synchronise cache command.

PR: kern/21752
Submitted by: Nick R. Colakovic <nickc@corp.firstindustrial.com>
MFC after: 1 week

22 years agoPrint the correct return code for successful Linux syscalls.
Dag-Erling Smørgrav [Sat, 17 Nov 2001 17:18:36 +0000 (17:18 +0000)]
Print the correct return code for successful Linux syscalls.

Submitted by: Vitezslav Novy <vita@fio.cz>
PR: 32036
MFC after: 1 week

22 years agoAdd quirk for Fujitsu M2513A MO drives. These drives hang at various
Ian Dowse [Sat, 17 Nov 2001 14:46:22 +0000 (14:46 +0000)]
Add quirk for Fujitsu M2513A MO drives. These drives hang at various
operations due to the synchronize cache command.

PR: kern/21674
Submitted by: W.Scholten <whs@xs4all.nl>
MFC after: 1 week

22 years agoFix pthread_join so that it works if the target thread exits while
Daniel Eischen [Sat, 17 Nov 2001 14:28:39 +0000 (14:28 +0000)]
Fix pthread_join so that it works if the target thread exits while
the joining thread is in a signal handler.

Reported by: Loren James Rittle <rittle@labs.mot.com>
MFC after: 1 week

22 years agoMFi386: revisions from 1.26 to 1.30.
Yoshihiro Takahashi [Sat, 17 Nov 2001 13:58:04 +0000 (13:58 +0000)]
MFi386: revisions from 1.26 to 1.30.

22 years agoMFi386: revision 1.12
Yoshihiro Takahashi [Sat, 17 Nov 2001 13:57:32 +0000 (13:57 +0000)]
MFi386: revision 1.12

22 years agoMFi386: revision 1.326.
Yoshihiro Takahashi [Sat, 17 Nov 2001 13:48:31 +0000 (13:48 +0000)]
MFi386: revision 1.326.

22 years agoMFi386: revision 1.160
Yoshihiro Takahashi [Sat, 17 Nov 2001 13:44:33 +0000 (13:44 +0000)]
MFi386: revision 1.160

22 years agoMFi386: revision 1.483
Yoshihiro Takahashi [Sat, 17 Nov 2001 13:16:51 +0000 (13:16 +0000)]
MFi386: revision 1.483

22 years agoChange mode for var/db/ipf to 0700
Guido van Rooij [Sat, 17 Nov 2001 12:21:29 +0000 (12:21 +0000)]
Change mode for var/db/ipf to 0700

22 years agoIf 'VX' is given, realloc(foo,0) will bail, it shouldn't.
Poul-Henning Kamp [Sat, 17 Nov 2001 10:57:04 +0000 (10:57 +0000)]
If 'VX' is given, realloc(foo,0) will bail, it shouldn't.

PR: 29376
Submitted by: Farooq Mela <fmela0@sm.socccd.cc.ca.us>

22 years agoGive struct socket structures a ref counting interface similar to
Matthew Dillon [Sat, 17 Nov 2001 03:07:11 +0000 (03:07 +0000)]
Give struct socket structures a ref counting interface similar to
vnodes.  This will hopefully serve as a base from which we can
expand the MP code.  We currently do not attempt to obtain any
mutex or SX locks, but the door is open to add them when we nail
down exactly how that part of it is going to work.

22 years agoFix the non-KSTACK_GUARD case.. It has been broken since the KSE
Peter Wemm [Sat, 17 Nov 2001 01:56:04 +0000 (01:56 +0000)]
Fix the non-KSTACK_GUARD case.. It has been broken since the KSE
commit.  ptek was not been initialized.

22 years agoStart bringing i386/pmap.c into line with cleanups that were done to
Peter Wemm [Sat, 17 Nov 2001 01:38:32 +0000 (01:38 +0000)]
Start bringing i386/pmap.c into line with cleanups that were done to
alpha pmap.  In particular -
- pd_entry_t and pt_entry_t are now u_int32_t instead of a pointer.
  This is to enable cleaner PAE and x86-64 support down the track sor
  that we can change the pd_entry_t/pt_entry_t types to 64 bit entities.
- Terminate "unsigned *ptep, pte" with extreme prejudice and use the
  correct pt_entry_t/pd_entry_t types.
- Various other cosmetic changes to match cleanups elsewhere.
- This eliminates a boatload of casts.
- use VM_MAXUSER_ADDRESS in place of UPT_MIN_ADDRESS in a couple of places
  where we're testing user address space limits.  Assuming the page tables
  start directly after the end of user space is not a safe assumption.
There is still more to go.

22 years agoForward declare struct ifnet - this fixes a warning in tdfx_pci.c
Peter Wemm [Sat, 17 Nov 2001 01:22:07 +0000 (01:22 +0000)]
Forward declare struct ifnet - this fixes a warning in tdfx_pci.c

22 years agoFix printf warnings (int/long)
Peter Wemm [Sat, 17 Nov 2001 01:20:02 +0000 (01:20 +0000)]
Fix printf warnings (int/long)
#if 0 around unused ifname_bsd_to_linux() function

22 years agoFix warning in debug printf. This is a long on alpha, and int on i386,
Peter Wemm [Sat, 17 Nov 2001 01:16:10 +0000 (01:16 +0000)]
Fix warning in debug printf.  This is a long on alpha, and int on i386,
but printed with %ld always.

22 years agoMissing KSE s/curproc/curthread/
Peter Wemm [Sat, 17 Nov 2001 01:09:53 +0000 (01:09 +0000)]
Missing KSE s/curproc/curthread/

22 years agoReturn EOPNOTSUPP for unknown module events.
Ian Dowse [Sat, 17 Nov 2001 00:46:08 +0000 (00:46 +0000)]
Return EOPNOTSUPP for unknown module events.

PR: kern/18473
Submitted by: "Jeroen C. van Gelderen" <gelderen@systemics.com>

22 years agoNew release notes: timezone database update, Linux emulation update
Bruce A. Mah [Sat, 17 Nov 2001 00:44:41 +0000 (00:44 +0000)]
New release notes:  timezone database update, Linux emulation update
for linux_base-7 port.

22 years agoFix some warnings on 64 bit platforms.
Peter Wemm [Sat, 17 Nov 2001 00:42:02 +0000 (00:42 +0000)]
Fix some warnings on 64 bit platforms.

22 years agoWhen laying out objects in a ZONE_INTERRUPT zone, allow them to cross
Jonathan Lemon [Sat, 17 Nov 2001 00:40:48 +0000 (00:40 +0000)]
When laying out objects in a ZONE_INTERRUPT zone, allow them to cross
a page boundary, since we've already allocated all our contiguous kva
space up front.  This eliminates some memory wastage, and allows us to
actually reach the # of objects were specified in the zinit() call.

Reviewed by: peter, dillon

22 years agoutime/stime.tv_sec are elapsed times, not relative to 1970. We can
Peter Wemm [Sat, 17 Nov 2001 00:26:57 +0000 (00:26 +0000)]
utime/stime.tv_sec are elapsed times, not relative to 1970.  We can
safely print them as longs.  Even if ^T overflows after a process
has accumulated 68 years of user or system time, it is no big deal.

22 years agoZap a number of #ifdef sunos blocks, and all of the `register'
Ian Dowse [Sat, 17 Nov 2001 00:06:55 +0000 (00:06 +0000)]
Zap a number of #ifdef sunos blocks, and all of the `register'
keywords.

22 years agoYou cannot cast a time_t to quad_t and printf it with %lld. quad_t is
Peter Wemm [Fri, 16 Nov 2001 23:53:48 +0000 (23:53 +0000)]
You cannot cast a time_t to quad_t and printf it with %lld. quad_t is
64 bits, not long long.

22 years agoGive a sensible error message when the filesystem to be dumped is
Ian Dowse [Fri, 16 Nov 2001 22:13:44 +0000 (22:13 +0000)]
Give a sensible error message when the filesystem to be dumped is
not listed in /etc/fstab. Previously, the user would be greeted
with "DUMP: bad sblock magic number" when dump tried to parse
the directory contents as an FFS filesystem.

PR: bin/12789
Submitted by: Bob Willcox <bob@pmr.com>

22 years agoFix a number of misspellings of "dependency" and "dependencies" in
Ian Dowse [Fri, 16 Nov 2001 21:08:40 +0000 (21:08 +0000)]
Fix a number of misspellings of "dependency" and "dependencies" in
comments and function names.

PR: kern/8589
Submitted by: Rajesh Vaidheeswarran <rv@fore.com>

22 years agoHandle the IBCS2 FIONREAD ioctl. I have only tested that this
Ian Dowse [Fri, 16 Nov 2001 20:32:32 +0000 (20:32 +0000)]
Handle the IBCS2 FIONREAD ioctl. I have only tested that this
compiles, but the patch looks reasonable.

PR: i386/5784
Submitted by: Remy NONNENMACHER <remy@synx.com>

22 years agoAxe SMP support, the ses device, and the atapifd device from the kern.flp
John Baldwin [Fri, 16 Nov 2001 18:33:19 +0000 (18:33 +0000)]
Axe SMP support, the ses device, and the atapifd device from the kern.flp
boot kernel so it fits again.  This actually gives us quite a bit of
breathing room, so some more ethernet drivers might be turned on now in a
later commit.

22 years agoCorrectly call THREAD_UNLOCK() if the recursive call trap is sprung.
Poul-Henning Kamp [Fri, 16 Nov 2001 18:15:33 +0000 (18:15 +0000)]
Correctly call THREAD_UNLOCK() if the recursive call trap is sprung.

Pointed out by: knu

22 years agoExcept for a comment in zone.tab (about the location of the ISO 3166 table),
Garrett Wollman [Fri, 16 Nov 2001 17:23:23 +0000 (17:23 +0000)]
Except for a comment in zone.tab (about the location of the ISO 3166 table),
we are now using the vendor files verbatim.

22 years agoBack out the previous fix to the leading zero problem, I hadn't
Poul-Henning Kamp [Fri, 16 Nov 2001 17:07:47 +0000 (17:07 +0000)]
Back out the previous fix to the leading zero problem, I hadn't
noticed it in there already.  That should teach me to check exit
code from cvsup.

22 years agoReject leading zeros in dev_stdclone().
Poul-Henning Kamp [Fri, 16 Nov 2001 17:05:07 +0000 (17:05 +0000)]
Reject leading zeros in dev_stdclone().

PR: 32019
Submitted by: fenner

22 years agoo Allow rarpd to accept an additional '-t directory' argument, specifying
Robert Watson [Fri, 16 Nov 2001 16:49:21 +0000 (16:49 +0000)]
o Allow rarpd to accept an additional '-t directory' argument, specifying
  an alternative to /tftpboot.  This is useful it you're using tftpd
  with an alternative root (using -s), and would like rarpd to respond
  selectively to RARP requests using the same criteria as tftp.

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

22 years agoo Update licenses, comments.
Robert Watson [Fri, 16 Nov 2001 15:59:46 +0000 (15:59 +0000)]
o Update licenses, comments.

Obtained from: TrustedBSD Project

22 years agoCreate link from directory.3 to readdir_r.3.
Dima Dorfman [Fri, 16 Nov 2001 15:49:38 +0000 (15:49 +0000)]
Create link from directory.3 to readdir_r.3.

PR: 32028
Submitted by: andrew@ugh.net.au

22 years agoRegenerate with 1.19 of pccarddevs
Warner Losh [Fri, 16 Nov 2001 06:59:27 +0000 (06:59 +0000)]
Regenerate with 1.19 of pccarddevs

22 years agoMerge in most (hopefully all) of the devices that OpenBSD has that
Warner Losh [Fri, 16 Nov 2001 06:58:47 +0000 (06:58 +0000)]
Merge in most (hopefully all) of the devices that OpenBSD has that
NetBSD/FreeBSD doesn't have listed.

22 years agoo Reflect moving of extattr_namespace calls to libc from libutil; no
Robert Watson [Fri, 16 Nov 2001 05:10:35 +0000 (05:10 +0000)]
o Reflect moving of extattr_namespace calls to libc from libutil; no
  longer need to link libutil into command-line extended attribute
  utilities.

Obtained from: TrustedBSD Project

22 years agoo Reflect repo-copy of extattr.[c3] from libutil to libc, moving
Robert Watson [Fri, 16 Nov 2001 05:09:45 +0000 (05:09 +0000)]
o Reflect repo-copy of extattr.[c3] from libutil to libc, moving
  extattr namespace routines to the libc/posix1e directory.  While
  the extattr calls are not strictly POSIX.1e, POSIX.1e wasn't
  strictly ever approved, so I think that's OK.

Obtained from: TrustedBSD Project

22 years agoo Document 'nocheckmail' login capability.
Robert Watson [Fri, 16 Nov 2001 04:39:43 +0000 (04:39 +0000)]
o Document 'nocheckmail' login capability.

22 years agoo Add support for a 'nocheckmail' capability, which (if true) prevents
Robert Watson [Fri, 16 Nov 2001 04:39:16 +0000 (04:39 +0000)]
o Add support for a 'nocheckmail' capability, which (if true) prevents
  the 'You have mail.' check.  This is useful for sites that rely on
  remote mail access, rather than a local mail spool.  Due to the
  behavior of login_getcapbool(), the negated form is required so as
  to have appropriate results.
o This behavior may have to be independently added to sshd due to
  redundant implementation.

22 years agoDon't cleandir also if MODULES_WITH_WORLD.
Makoto Matsushita [Fri, 16 Nov 2001 03:02:34 +0000 (03:02 +0000)]
Don't cleandir also if MODULES_WITH_WORLD.

MFC after: 3 days

22 years agoUse the md device not olny a version of the system is 5 but 5 and over.
Yoshihiro Takahashi [Fri, 16 Nov 2001 03:01:50 +0000 (03:01 +0000)]
Use the md device not olny a version of the system is 5 but 5 and over.

Requested by: matusita
MFC after: 1 day

22 years agoCosmetic tweak to eliminate some diffs
Peter Wemm [Fri, 16 Nov 2001 02:44:55 +0000 (02:44 +0000)]
Cosmetic tweak to eliminate some diffs

22 years agoMerge another missing part of i386/pmap.c rev 1.220. Dont blindly clear
Peter Wemm [Fri, 16 Nov 2001 02:41:31 +0000 (02:41 +0000)]
Merge another missing part of i386/pmap.c rev 1.220.  Dont blindly clear
the PG_BUSY flag without using the code that wakes up something else that
may be sleeping on it.

22 years agoAlthough the 'bool' type is referenced in the list of capabilities, it
Robert Watson [Fri, 16 Nov 2001 02:40:09 +0000 (02:40 +0000)]
Although the 'bool' type is referenced in the list of capabilities, it
is not defined in the capability type list.  Provide a definition for
'bool', if a slightly less than elegant one.  Note that this definition
does not include the complete scope of available behavior defined
in cgetcap(3), and could probably be improved.

22 years agoOops, I accidently merged a whitespace error from the original commit.
Peter Wemm [Fri, 16 Nov 2001 02:31:20 +0000 (02:31 +0000)]
Oops, I accidently merged a whitespace error from the original commit.
(whitespace at end of line in rev 1.264 pmap.c).  Fix them all.

22 years agoMerge rev 1.264 from i386/pmap.c (tegge via alfred):
Peter Wemm [Fri, 16 Nov 2001 02:28:33 +0000 (02:28 +0000)]
Merge rev 1.264 from i386/pmap.c (tegge via alfred):
Protect against an infinite loop when prefaulting pages.  This can
happen when the vm system maps past the end of an object or tries
to map a zero length object, the pmap layer misses the fact that
offsets wrap into negative numbers and we get stuck.

22 years agoMerge rev 1.202 from i386/pmap.c (back in 1998 by John Dyson):
Peter Wemm [Fri, 16 Nov 2001 02:25:29 +0000 (02:25 +0000)]
Merge rev 1.202 from i386/pmap.c (back in 1998 by John Dyson):
Make flushing dirty pages work correctly on filesystems that
unexpectedly do not complete writes even with sync I/O requests.
This should help the behavior of mmaped files when using
softupdates (and perhaps in other circumstances also.)

22 years agoMerge rev 1.293 of i386/pmap.c - skip PG_UNMANAGED in pmap_collect()
Peter Wemm [Fri, 16 Nov 2001 02:20:40 +0000 (02:20 +0000)]
Merge rev 1.293 of i386/pmap.c - skip PG_UNMANAGED in pmap_collect()

22 years agoConverge/fix some debug code (#if 0'ed on alpha, but whatever)
Peter Wemm [Fri, 16 Nov 2001 02:17:18 +0000 (02:17 +0000)]
Converge/fix some debug code (#if 0'ed on alpha, but whatever)
- use NPTEPG/NPDEPG instead of magic 1024 (important for PAE)
- use pt_entry_t instead of unsigned (important for PAE)
- use vm_offset_t instead of unsigned for va's (important for x86-64)

22 years agoConverge with i386/pmap.c - dont refer to curproc, use curthread.
Peter Wemm [Fri, 16 Nov 2001 02:06:06 +0000 (02:06 +0000)]
Converge with i386/pmap.c - dont refer to curproc, use curthread.

22 years agoSwitch warnings and strict back on again in a way that's compatible
Josef Karthauser [Fri, 16 Nov 2001 02:02:42 +0000 (02:02 +0000)]
Switch warnings and strict back on again in a way that's compatible
with -stable as well as -current.

Reviewed by: imp

22 years agoMerge part of i386/pmap.c rev 1.220 that got missed in alpha/pmap.c
Peter Wemm [Fri, 16 Nov 2001 02:02:21 +0000 (02:02 +0000)]
Merge part of i386/pmap.c rev 1.220 that got missed in alpha/pmap.c
rev 1.10

22 years agoAs part of a general cleanup and reconvergence of related pmap code,
Peter Wemm [Fri, 16 Nov 2001 01:56:34 +0000 (01:56 +0000)]
As part of a general cleanup and reconvergence of related pmap code,
start tidying up some loose ends.  The DEBUG_VA stuff has long since
passed its use-by date.  It wasn't used on ia64 but got cut/pasted there.

22 years agoNew release notes: IPFilter supports IPv6, ipfs(8).
Bruce A. Mah [Fri, 16 Nov 2001 01:48:31 +0000 (01:48 +0000)]
New release notes:  IPFilter supports IPv6, ipfs(8).

Reminded by: devet@devet.org (Arjan de Vet)

22 years agoFix a leftover client comment, long line fix.
Peter Wemm [Thu, 15 Nov 2001 23:49:02 +0000 (23:49 +0000)]
Fix a leftover client comment, long line fix.

22 years agoDo not allow leading zeros on device names in dev_stdclone().
Bill Fenner [Thu, 15 Nov 2001 23:27:46 +0000 (23:27 +0000)]
Do not allow leading zeros on device names in dev_stdclone().

PR: kern/32019
Reviewed by: phk

22 years agoAllow bit 21 of EFLAGS register (PSL_ID) be changed in the use-mode without
Maxim Sobolev [Thu, 15 Nov 2001 23:01:29 +0000 (23:01 +0000)]
Allow bit 21 of EFLAGS register (PSL_ID) be changed in the use-mode without
ill effects. This should fix problems threaded programs are having with
auto-detecting CPU type.

Reported by: Joe Clarke <marcus@marcuscom.com>
Tested by: Joe Clarke <marcus@marcuscom.com>
Reviewed by: jhb
MFC after: 1 week

22 years agoo Update copyright dates, comments...
Robert Watson [Thu, 15 Nov 2001 22:56:40 +0000 (22:56 +0000)]
o Update copyright dates, comments...

Obtained from: TrustedBSD Project

22 years agoo extattrctl initattr, when pre-allocating store for extended attributes,
Robert Watson [Thu, 15 Nov 2001 22:50:06 +0000 (22:50 +0000)]
o extattrctl initattr, when pre-allocating store for extended attributes,
  computed a a chunksize that didn't include the extended attribute
  header.  This was a non-fatal error, in that it was just writing out
  zeros anyway, but did have the effect of not pre-allocating the
  right amount of disk space.  This fix calculates chunksize to include
  the attribute header.

Submitted by: Dale Rahn
Sponsored by: DARPA, UPenn POSSE Project
Obtained from: OpenBSD

22 years agoUse MTX_QUIET for the lock operations during clock interrupts so their logs
John Baldwin [Thu, 15 Nov 2001 19:54:48 +0000 (19:54 +0000)]
Use MTX_QUIET for the lock operations during clock interrupts so their logs
don't drown out more useful log messages.

22 years agoAdd a couple of returns to making recovering from a failed witness_assert()
John Baldwin [Thu, 15 Nov 2001 19:46:36 +0000 (19:46 +0000)]
Add a couple of returns to making recovering from a failed witness_assert()
more sane in the RESTARTABLE_PANICS case.

22 years ago'error' is now unused, since its only purpose was to catch the
Robert Watson [Thu, 15 Nov 2001 19:42:08 +0000 (19:42 +0000)]
'error' is now unused, since its only purpose was to catch the
results of suser_td().

22 years agoDocument the atkbd flags of 0x03 and be explicit to mention this might
Jeroen Ruigrok van der Werven [Thu, 15 Nov 2001 19:42:07 +0000 (19:42 +0000)]
Document the atkbd flags of 0x03 and be explicit to mention this might
fit some dockingstation keyboard probing.

PR: 23681
Submitted by: yokota [PR issued by:
Claude Lefrancois <lmcclef@lmc.ericsson.se>]

22 years agoTemporarily disable unloading, as it appears to be broken (read: causes
Robert Watson [Thu, 15 Nov 2001 19:41:14 +0000 (19:41 +0000)]
Temporarily disable unloading, as it appears to be broken (read: causes
panics)

22 years agoNew release notes: Jitter in load average computation, newsyslog(8)
Bruce A. Mah [Thu, 15 Nov 2001 19:32:57 +0000 (19:32 +0000)]
New release notes:  Jitter in load average computation, newsyslog(8)
support for bzip2(1).

MFCs noted:  gx(4), edquota(8) -f.

22 years agoo Rely on /dev/snp* permissions, rather than using suser() in snpopen()
Robert Watson [Thu, 15 Nov 2001 19:28:49 +0000 (19:28 +0000)]
o Rely on /dev/snp* permissions, rather than using suser() in snpopen()
  to prevent inappropriate use of the snoop device.  This means that,
  with appropriate chmod'ing, arbitrary users may make use of the snoop
  device.

22 years agoRemove definition of witness and comment stating that this file implements
John Baldwin [Thu, 15 Nov 2001 19:08:55 +0000 (19:08 +0000)]
Remove definition of witness and comment stating that this file implements
witness.  Witness moved off to subr_witness.c a while ago.

22 years agoAdd card_if.h to SRCS.
Makoto Matsushita [Thu, 15 Nov 2001 18:59:04 +0000 (18:59 +0000)]
Add card_if.h to SRCS.

src/sys/dev/an/if_an_pccard.c rev 1.10 change requires to include
"card_if.h" but Makefile doesn't know about this file.  Without this,
kernel build will fail at this driver.