]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoAlphas with libc.so between 2003/03/12 and 2003/03/29 cannot be
ru [Fri, 25 Apr 2003 20:11:17 +0000 (20:11 +0000)]
Alphas with libc.so between 2003/03/12 and 2003/03/29 cannot be
upgraded cleanly.  Document the reason and possible workarounds.

Requested by: gallatin

21 years agoRemove Giant from getpgid() and getsid() and tweak the logic to more
jhb [Fri, 25 Apr 2003 20:09:31 +0000 (20:09 +0000)]
Remove Giant from getpgid() and getsid() and tweak the logic to more
closely match that of 4.x.

21 years ago- Don't bother using the proc lock to test just P_SYSTEM as that is set in
jhb [Fri, 25 Apr 2003 20:06:30 +0000 (20:06 +0000)]
- Don't bother using the proc lock to test just P_SYSTEM as that is set in
  fork1() and never changes.
- The proc lock is enough to cover reading p_state, so push down sched_lock
  into the PRS_NORMAL case of the switch on p_state.

21 years ago- Push down Giant into the sysarch() calls that still need Giant.
jhb [Fri, 25 Apr 2003 20:04:02 +0000 (20:04 +0000)]
- Push down Giant into the sysarch() calls that still need Giant.
- Standardize on EINVAL rather than EOPNOTSUPP if the sysarch op value is
  invalid.

21 years agoPush down Giant around calls to proc_rwmem() in kern_ptrace. kern_ptrace()
jhb [Fri, 25 Apr 2003 20:02:16 +0000 (20:02 +0000)]
Push down Giant around calls to proc_rwmem() in kern_ptrace.  kern_ptrace()
should now be MP safe.

21 years agoPush Giant down into kern_sigaction() instead of locking it around calls
jhb [Fri, 25 Apr 2003 20:01:19 +0000 (20:01 +0000)]
Push Giant down into kern_sigaction() instead of locking it around calls
to kern_sigaction() in the various callers of the function.

21 years ago- Push down Giant around vnode operations in ktrace().
jhb [Fri, 25 Apr 2003 19:59:35 +0000 (19:59 +0000)]
- Push down Giant around vnode operations in ktrace().
- Mark the ktrace() and utrace() syscalls as being MP safe.
- Validate the facs argument to ktrace() prior to doing any vnode
  operations or acquiring any locks.
- Share lock the proctree lock over the entire section that calls
  ktrsetchildren() and ktrops().  We already did this for process groups.
  Doing it for the process case closes a small race where a process might
  go away after we look it up.  As a result of this, ktrstchildren() now
  just asserts that the proctree lock is locked rather than acquiring the
  lock itself.
- Add some missing comments to #else and #endif.

21 years agoLock Giant around calls to useracc() in osf1_sigreturn() so that the
jhb [Fri, 25 Apr 2003 19:52:30 +0000 (19:52 +0000)]
Lock Giant around calls to useracc() in osf1_sigreturn() so that the
function is MP safe.

21 years agoLock Giant around userland_sysctl() to get the hostname in osf1_sysinfo()
jhb [Fri, 25 Apr 2003 19:51:41 +0000 (19:51 +0000)]
Lock Giant around userland_sysctl() to get the hostname in osf1_sysinfo()
so that the function is MP safe.  The rest of the function doesn't need
Giant.

21 years agoUse a switch to convert the Linux sigprocmask flags to the equivalent
jhb [Fri, 25 Apr 2003 19:26:18 +0000 (19:26 +0000)]
Use a switch to convert the Linux sigprocmask flags to the equivalent
FreeBSD flags instead of just adding one to the Linux flags.  This should
be identical to the previous version except that I have at least one report
of this patch fixing problems people were having with Linux apps after my
last commit to this file.  It is safer to use the switch then to make
assumptions about the flag values anyways, esp. since we currently use
MD defines for the values of the flags and this is MI code.

Tested by: Michael Class <michael_class@gmx.net>

21 years agoMention PAE support.
jake [Fri, 25 Apr 2003 18:51:29 +0000 (18:51 +0000)]
Mention PAE support.

Reviewed by: bmah
Sponsored by: DARPA, Network Associates Laboratories

21 years agoWhen using `compat' mode, be sure to re-dispatch setpwent, endpwent,
nectar [Fri, 25 Apr 2003 18:25:19 +0000 (18:25 +0000)]
When using `compat' mode, be sure to re-dispatch setpwent, endpwent,
setgrent, and endgrent also.  (The previous NSS implementation used to
simply twiddle the internal data of the various modules directly.)

A symptom (group list set incorrectly in sshd) was
Reported by: Glenn Johnson <gjohnson@srrc.ars.usda.gov>

Sponsored by: DARPA, Network Associates Laboratories

21 years agoRemove a stray sentence.
jake [Fri, 25 Apr 2003 17:08:47 +0000 (17:08 +0000)]
Remove a stray sentence.

21 years ago - Lock the vm_object when iterating over its list of resident pages.
alc [Fri, 25 Apr 2003 16:30:02 +0000 (16:30 +0000)]
 - Lock the vm_object when iterating over its list of resident pages.

21 years agoConvert the midway driver to use busdma. Except for this conversion the
harti [Fri, 25 Apr 2003 16:14:03 +0000 (16:14 +0000)]
Convert the midway driver to use busdma. Except for this conversion the
following changes have been done:

- stylify. The original code was too hard to read.
- get rid of a number of compilation options (Adaptec-only, Eni-only, no-DMA).
- more debugging features.
- locking. This is not correct yet in the absence of interface layer locking,
  but is correct enough to not to cause lock order reversals.
- remove RAW mode. There are no users of this in the tree and I doubt that
  there are any.
- remove NetBSD compatibility code. There was no way to keep NetBSD non-busdma
  and FreeBSD busdma code together.
- if_en now buildable as a module.

This has been actively tested on sparc64 and i386 with ENI server and
client cards and an Adaptec card (thanks to kjc).

Reviewed by: mdodd, arr

21 years agoRegen.
jhb [Fri, 25 Apr 2003 15:59:44 +0000 (15:59 +0000)]
Regen.

21 years agoOops, the thr_* and jail_attach() syscall entries should be NOPROTO rather
jhb [Fri, 25 Apr 2003 15:59:18 +0000 (15:59 +0000)]
Oops, the thr_* and jail_attach() syscall entries should be NOPROTO rather
than STD.

21 years agoWhitespace fix up, no content changes. Translators can ignore this.
bmah [Fri, 25 Apr 2003 15:13:18 +0000 (15:13 +0000)]
Whitespace fix up, no content changes.  Translators can ignore this.

21 years agoUpdate HTT support text.
bmah [Fri, 25 Apr 2003 15:12:21 +0000 (15:12 +0000)]
Update HTT support text.

21 years agoRemove harmless invalid cast.
jake [Fri, 25 Apr 2003 15:07:58 +0000 (15:07 +0000)]
Remove harmless invalid cast.

Sponsored by: DARPA, Network Associates Laboratories

21 years agos/filesystem/file system/
bmah [Fri, 25 Apr 2003 14:42:53 +0000 (14:42 +0000)]
s/filesystem/file system/
s/bootloader/boot loader/

I hate search and replace sweeps like this as much as anyone else, but
in each case I used both alternatives within a single document, so I'm
trying to enforce some internal consistency.

21 years agoNew release notes: i386 boot loader 1.5TB limit, OpenSSH 3.6.1p1.
bmah [Fri, 25 Apr 2003 14:37:56 +0000 (14:37 +0000)]
New release notes:  i386 boot loader 1.5TB limit, OpenSSH 3.6.1p1.

Modified release notes:  Add note about 1.5TB limit to sysinstall note.

21 years agoThe documented login.conf variable for setting the login prompt is
des [Fri, 25 Apr 2003 11:57:20 +0000 (11:57 +0000)]
The documented login.conf variable for setting the login prompt is
"login_prompt".  This makes more sense than "prompt" which is what
login actually used, so change the code to match the documentation.

PR: docs/51396
MFC in: 3 days

21 years agoRepair the UFS2 superblock location consistency check so that it succeeds
tjr [Fri, 25 Apr 2003 10:07:50 +0000 (10:07 +0000)]
Repair the UFS2 superblock location consistency check so that it succeeds
on valid superblocks instead of issuing the error "not a BSD filesystem".
fs_sblockloc is a byte offset, not a fragment number. This change makes
quot work properly on UFS2 filesystems, which is important now that UFS2
is the default.

21 years agoNew release note: chown(8)/chgrp(1) -v -v.
johan [Fri, 25 Apr 2003 09:58:52 +0000 (09:58 +0000)]
New release note: chown(8)/chgrp(1) -v -v.

21 years agoForced commit to note that WARNS was changed in the prev commit since
johan [Fri, 25 Apr 2003 09:02:59 +0000 (09:02 +0000)]
Forced commit to note that WARNS was changed in the prev commit since
this code is WARNS=5 clean. Try to keep it that way.

21 years agoMake fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet):
njl [Fri, 25 Apr 2003 09:01:54 +0000 (09:01 +0000)]
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet):
- Add fxp_start_body() and change fxp_start() to just acquire locks and
  then call fxp_start_body().  Places that would call fxp_start() with
  locks held (mutex recursion) now call fxp_start_body() directly.
  Remove MTX_RECURSE flag from sc_mtx. [gallatin]
- Change fxp_attach() to work without the softc lock, saving interrupt
  hooking until the head of fxp_attach().
- Call ether_ifattach() before overriding ifp parameters. This reverts
  part of 1.155.
- Remove multiple error paths in fxp_attach().
- Teardown interrupt in fxp_detach() before unlocking the softc.
- Make sure mutex is not held in fxp_release()
- Delete the miibus instance and/or self in fxp_release(), not in
  fxp_detach().  This can happen if attach fails partway through.
- Move ifmedia_removeall to fxp_release() since attach may fail after
  media have been allocated.
- Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr,
  fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog.
- Pass in ifp to fxp_intr_body since its callers sometimes already use
  it.
- Add compatibility define for INTR_MPSAFE for 4.x. [gallatin]
- You don't need to bzero softc.

Ideas from: gallatin, mux
Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP

21 years agoSimilar to chmod(1); make -v -v mean very verbose and show
johan [Fri, 25 Apr 2003 08:57:55 +0000 (08:57 +0000)]
Similar to chmod(1); make -v -v mean very verbose and show
the old and new uid/gid.

PR: 41341
Submitted by: Edward Brocklesby <nighthawk@unrealircd.com>
Reviewed by: bde@ (an older version)

21 years agoRestore bzip2(1) support in loader(8) after Peter's renames of
ru [Fri, 25 Apr 2003 06:44:56 +0000 (06:44 +0000)]
Restore bzip2(1) support in loader(8) after Peter's renames of
malloc() to Malloc(), and free() to Free(), in zalloc_malloc.c.

21 years ago - Relax the Giant required in vm_page_remove().
alc [Fri, 25 Apr 2003 06:35:05 +0000 (06:35 +0000)]
 - Relax the Giant required in vm_page_remove().
 - Remove the Giant required from vm_page_free_toq().  (Any locking
   errors will be caught by vm_page_remove().)

This remedies a panic that occurred when kmem_malloc(NOWAIT) performed
without Giant failed to allocate the necessary pages.

Reported by: phk

21 years agoRemoving comment in some text files for fixit floppy overflow.
kuriyama [Fri, 25 Apr 2003 06:08:11 +0000 (06:08 +0000)]
Removing comment in some text files for fixit floppy overflow.

Approved by: re (murray)

21 years agoImport latest changes from the vendor. This also is reported to fix the
scottl [Fri, 25 Apr 2003 05:37:04 +0000 (05:37 +0000)]
Import latest changes from the vendor.  This also is reported to fix the
driver at long last!
Many thanks to vaidas.damosevicius@if.lt for keeping this issue alive
and pursuing Intel for a fix, Intel/ICP for working on the driver, and
Sergey Osokin for bringing the original patches up to 5-CURRENT.

21 years agoChange default to be 'process all events before calling daemon.'
imp [Fri, 25 Apr 2003 02:13:42 +0000 (02:13 +0000)]
Change default to be 'process all events before calling daemon.'

Also, put a small work around into devd to prevent a hang on boot this
would cause because select used to return 2 rather than 0 for no
evetnts due to a bug I fixed a few days ago in subr_bus.c.  I'll
remove this workaround May 7th.  You have until then to upgrade your
kernel if you want to run a new devd with an older kernel.

21 years agoAdd warning about a potential hang on boot at devd after 20030507 if
imp [Fri, 25 Apr 2003 02:10:17 +0000 (02:10 +0000)]
Add warning about a potential hang on boot at devd after 20030507 if
you have an old kernel and a new userland.  I've just committed a fix
for devd to work around the kernel bug, but that goes away on May 7th.

21 years agoAdd an argument to get_mcontext() which specified whether the
deischen [Fri, 25 Apr 2003 01:50:30 +0000 (01:50 +0000)]
Add an argument to get_mcontext() which specified whether the
syscall return values should be cleared.  The system calls
getcontext() and swapcontext() want to return 0 on success
but these contexts can be switched to at a later time so
the return values need to be cleared in the saved register
sets.  Other callers of get_mcontext() would normally want
the context without clearing the return values.

Remove the i386-specific context saving from the KSE code.
get_mcontext() is not i386-specific any more.

Fix a bad pointer in the alpha get_mcontext() code.  The
context was being bcopy()'d from &td->tf_frame, but tf_frame
is itself a pointer, so the thread was being copied instead.
Spotted by jake.

Glanced at by:  jake
Reviewed by:    bde (months ago)

21 years agoConnect libpthread to the (i386) world. It is currently installed
deischen [Fri, 25 Apr 2003 01:34:43 +0000 (01:34 +0000)]
Connect libpthread to the (i386) world.  It is currently installed
as libkse and will once again be renamed libpthread after more
testing.

Approved by: re

21 years agoRemove the %gs restoring hack (already commented out).
deischen [Fri, 25 Apr 2003 01:31:56 +0000 (01:31 +0000)]
Remove the %gs restoring hack (already commented out).
Don't install man pages.

Temporarily (again) rename the library to libkse.  It will be put back
to libpthread after more wide-spread testing.

21 years agoUpdate the DRM to the latest from DRI CVS. Includes some bugfixes and removal
anholt [Fri, 25 Apr 2003 01:18:47 +0000 (01:18 +0000)]
Update the DRM to the latest from DRI CVS.  Includes some bugfixes and removal
of the infrastructure for the gamma driver which was removed a while back.
The DRM_LINUX option is removed because the handler is now provided by the
linux compat code itself.

21 years agoBack out previous delta to fix fsck on filesystems without an fstab entry,
jmallett [Fri, 25 Apr 2003 01:12:35 +0000 (01:12 +0000)]
Back out previous delta to fix fsck on filesystems without an fstab entry,
where we want to take the disklabel filesystem type of "4.2BSD" and use
fsck_4.2bsd on those filesystems.

Add a comment about why the code is there, now that we know:

         * XXX This is a kludge to make automatic filesystem type guessing
         * from the disklabel work for "4.2BSD" filesystems.  It does a
         * very limited subset of transliteration to a normalised form of
         * filesystem name, and we do not seem to enforce a filesystem
         * name character set.

21 years agoInclude altkstack pages in the RSS regardless of whether the process
tjr [Fri, 25 Apr 2003 00:20:40 +0000 (00:20 +0000)]
Include altkstack pages in the RSS regardless of whether the process
is swapped out. Pointed out by jhb.

21 years agoIn compat mode, we `redispatch' the lookup. It is probably a good
nectar [Thu, 24 Apr 2003 23:56:58 +0000 (23:56 +0000)]
In compat mode, we `redispatch' the lookup.  It is probably a good
idea to re-initialize `struct passwd', because e.g. pw_class might
get set by one module, but not by another.  Add another call to the
internal pwd_init function to accomplish this.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoAdd an ioctl handler for the DRM. This removes the need for the DRM_LINUX
anholt [Thu, 24 Apr 2003 23:36:35 +0000 (23:36 +0000)]
Add an ioctl handler for the DRM.  This removes the need for the DRM_LINUX
option, which has been a source of frustration for many users.

21 years agoRegen.
jhb [Thu, 24 Apr 2003 20:50:57 +0000 (20:50 +0000)]
Regen.

21 years agoDocument support for the new 'U' flag, which was implemented back in March
gad [Thu, 24 Apr 2003 20:50:01 +0000 (20:50 +0000)]
Document support for the new 'U' flag, which was implemented back in March
as revision 1.64 of newsyslog.c.  I plan to MFC these changes next week.

MFC after: 1 week

21 years agoFix the thr_create() entry by adding a trailing \. Also, sync up the
jhb [Thu, 24 Apr 2003 20:49:46 +0000 (20:49 +0000)]
Fix the thr_create() entry by adding a trailing \.  Also, sync up the
MP safe flag for thr_* with the main table.

21 years agoVOP_FSYNC() expects to be called with the vnode locked, so lock fvp in
truckman [Thu, 24 Apr 2003 20:39:40 +0000 (20:39 +0000)]
VOP_FSYNC() expects to be called with the vnode locked, so lock fvp in
nfs_rename() before calling VOP_FSYNC() and unlock fvp immediately after.

Reviewed by: bde

21 years agoFix a bug with df on large (>1TB) nfsv3 file servers on 32 bit client
peter [Thu, 24 Apr 2003 20:36:32 +0000 (20:36 +0000)]
Fix a bug with df on large (>1TB) nfsv3 file servers on 32 bit client
machines where the 'long' number of blocks in struct statfs wont fit.
Instead of chosing an artificial 512 byte block size, simply scale it up
until we avoid an overflow.  NFSv3 reports the sizes in bytes, and the
blocksize is a figment of nfsclient's imagination.

21 years agoCatch up with nsdispatch.c: nsdispatch(3) is now `hidden' by
nectar [Thu, 24 Apr 2003 20:16:21 +0000 (20:16 +0000)]
Catch up with nsdispatch.c:  nsdispatch(3) is now `hidden' by
namespace.h.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoDon't complain about missing NSS methods when built statically. It is
nectar [Thu, 24 Apr 2003 19:57:31 +0000 (19:57 +0000)]
Don't complain about missing NSS methods when built statically.  It is
annoying and not very useful.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoFail to mount a device if the bytes per sector in the BPB is less than
jhb [Thu, 24 Apr 2003 18:19:19 +0000 (18:19 +0000)]
Fail to mount a device if the bytes per sector in the BPB is less than
DEV_BSIZE or if the number of FAT sectors is zero.

21 years agoUse $ipnat_program instead of the hardcoded path, which was wron
mtm [Thu, 24 Apr 2003 18:14:52 +0000 (18:14 +0000)]
Use $ipnat_program instead of the hardcoded path, which was wron
anyways.

Approved by: markm (mentor) (implicit)
Submitted by: leafy <leafy@leafy.idv.tw>

21 years agoCatch up with nsdispatch.c: nsdispatch(3) is now `hidden' by
nectar [Thu, 24 Apr 2003 18:05:48 +0000 (18:05 +0000)]
Catch up with nsdispatch.c:  nsdispatch(3) is now `hidden' by
namespace.h.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoCatch up with nsdispatch.c: nsdispatch(3) is now `hidden' by
nectar [Thu, 24 Apr 2003 17:41:20 +0000 (17:41 +0000)]
Catch up with nsdispatch.c:  nsdispatch(3) is now `hidden' by
namespace.h.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoMake P_PROTECTED not conflict with P_STOPPED_SIG. Replace
wes [Thu, 24 Apr 2003 16:20:14 +0000 (16:20 +0000)]
Make P_PROTECTED not conflict with P_STOPPED_SIG.  Replace
P_UNUSED100000 which is *truly* unused, until now.

Submitted by: Robert Drehmel <robert@zoot.drehmel.com>

21 years agoFix printf warnings for i386 with PAE.
simokawa [Thu, 24 Apr 2003 15:27:06 +0000 (15:27 +0000)]
Fix printf warnings for i386 with PAE.

Submitted by: harti

21 years agoinit_nsm() is executed after a call to daemon(*, 0), so error and
ghelmer [Thu, 24 Apr 2003 14:38:42 +0000 (14:38 +0000)]
init_nsm() is executed after a call to daemon(*, 0), so error and
warning messages should be logged rather than sent to /dev/null.

PR: bin/45461

21 years agoFix a typo in UINT_MAX definition.
kan [Thu, 24 Apr 2003 14:25:38 +0000 (14:25 +0000)]
Fix a typo in UINT_MAX definition.

21 years agoForced commit to note that the previous commit was made in response to
des [Thu, 24 Apr 2003 12:27:01 +0000 (12:27 +0000)]
Forced commit to note that the previous commit was made in response to

PR: bin/51245

21 years agoRemove a bogus null password check which assumed that a user with an empty
des [Thu, 24 Apr 2003 12:26:25 +0000 (12:26 +0000)]
Remove a bogus null password check which assumed that a user with an empty
password must necessarily have an empty pwd->pw_passwd.  Also add a check
that prevents users from setting a blank password unless the nullok option
was specified.  Root is still allowed to give anyone a blank password.

21 years agoAdd nullok to the pam_unix line.
des [Thu, 24 Apr 2003 12:22:42 +0000 (12:22 +0000)]
Add nullok to the pam_unix line.

21 years agoIt seems that 1 was not a magic value as I thought, but a coincidence.
des [Thu, 24 Apr 2003 12:12:06 +0000 (12:12 +0000)]
It seems that 1 was not a magic value as I thought, but a coincidence.
Instead of applying the adjustment to processes with a start time of 1,
apply it to all processes with a start time of less than 3600.

None of this would be necessary if the start times were recorded in ticks
instead of seconds and microseconds.

21 years agoDo a better job of calculating the RSS for swapped-out processes:
tjr [Thu, 24 Apr 2003 11:03:04 +0000 (11:03 +0000)]
Do a better job of calculating the RSS for swapped-out processes:
don't include the kernel stacks of swapped-out threads in the page count,
but do include the alternate kernel stack. jhb provided some helpful
comments on this.

PR: 49102

21 years agodevice s/acpica/acpi/
hmp [Thu, 24 Apr 2003 09:54:18 +0000 (09:54 +0000)]
device s/acpica/acpi/

Noticed by: Stefan Farfeleder <stefan@fafoe.dyndns.org>
Approved by: des (mentor)

21 years agoAristotle dies in 02 Oct 322 BC not in March.
maxim [Thu, 24 Apr 2003 09:54:02 +0000 (09:54 +0000)]
Aristotle dies in 02 Oct 322 BC not in March.

PR: misc/46440
Submitted by: Sergey A. Osokin <osa@freebsd.org.ru>
MFC after: 4 days

21 years agoFix printf warning for RELENG_4.
simokawa [Thu, 24 Apr 2003 09:27:17 +0000 (09:27 +0000)]
Fix printf warning for RELENG_4.

21 years agoFix a typo: "Managment" -> "Management".
maxim [Thu, 24 Apr 2003 09:16:01 +0000 (09:16 +0000)]
Fix a typo: "Managment" -> "Management".

PR: bin/47343
Submitted by: Jeremy C. Reed <reed@reedmedia.net>
MFC after: 4 days

21 years agoo Add support for running /etc/rc.early
mtm [Thu, 24 Apr 2003 08:27:29 +0000 (08:27 +0000)]
o Add support for running /etc/rc.early
o Re-run through /etc/sysctl.conf a second time just before we set
  the securelevel.

Approved by: markm (mentor) (implicit)
Reviewed by: dougb

21 years agoMake ipfilter, ipnat, ipmon, and ipfs behave more like the old rc.
mtm [Thu, 24 Apr 2003 08:20:47 +0000 (08:20 +0000)]
Make ipfilter, ipnat, ipmon, and ipfs behave more like the old rc.
o group them together so they run one right after another
o use the NetBSD supplied ipfs script instead of tacking
  it on to the end of ipnat
o Load the ipl module in ipnat and ipfilter, if it's not already
  loaded
o In ipmon and ipnat show a warning if neither ipfilter nor
  ipnat is enabled or the ipl module is not loaded, and exit

Approved by: markm (mentor) (implicit)
Tested by: leafy <leafy@leafy.idv.tw>

21 years agoFree mount credentials (mnt_cred) when freeing the mount struct
tjr [Thu, 24 Apr 2003 08:16:06 +0000 (08:16 +0000)]
Free mount credentials (mnt_cred) when freeing the mount struct
in failure cases to avoid leaking struct ucreds, and ultimately
leaking struct uidinfo references.

21 years agoAdd some IEEE 1394 OHCI chips.
simokawa [Thu, 24 Apr 2003 07:29:52 +0000 (07:29 +0000)]
Add some IEEE 1394 OHCI chips.

Partially submitted by: Tetsuya Ryuchi <ryuchi@ryuchi.org>
PR: misc/51336

21 years ago - Move swap_pager_isswapped()'s prototype to a more logical place.
alc [Thu, 24 Apr 2003 05:29:27 +0000 (05:29 +0000)]
 - Move swap_pager_isswapped()'s prototype to a more logical place.

21 years ago - Acquire the vm_object's lock when performing vm_object_page_clean().
alc [Thu, 24 Apr 2003 04:31:25 +0000 (04:31 +0000)]
 - Acquire the vm_object's lock when performing vm_object_page_clean().
 - Add a parameter to vm_pageout_flush() that tells vm_pageout_flush()
   whether its caller has locked the vm_object.  (This is a temporary
   measure to bootstrap vm_object locking.)

21 years agoWhen filling out a kinfo_proc structure, if we come across a process
des [Thu, 24 Apr 2003 03:37:59 +0000 (03:37 +0000)]
When filling out a kinfo_proc structure, if we come across a process
whose p_stats->p_start has the magic value 1, replace it with boottime.
Some users were apparently confused by the fact that ps(1) reported a
start time in early 1970 for system processes.

21 years agoDon't print anything for fault at cpu_switch_load_gs, just like other
davidxu [Thu, 24 Apr 2003 01:48:59 +0000 (01:48 +0000)]
Don't print anything for fault at cpu_switch_load_gs, just like other
code to recover fault in doreti because of invalid segment registers,
silently push error to userland.

21 years agoDocument support for the new -s and -R command-line options, and
gad [Thu, 24 Apr 2003 01:34:40 +0000 (01:34 +0000)]
Document support for the new -s and -R command-line options, and
the 'N' flag.  These were coded in March as revisions 1.55 and 1.56
of newsyslog.c.  I intend to MFC all the matching changes next week.
This also reorganizes the description of the 'flags' field to give
list of the valid flags, instead of a long paragraph explaining
each of the possible values.

Obtained from: NetBSD (in spirit at least, for -s and N)
MFC after: 1 week

21 years agoDon't convert the kernel ident to uppercase when writing the Makefile.
des [Thu, 24 Apr 2003 00:52:58 +0000 (00:52 +0000)]
Don't convert the kernel ident to uppercase when writing the Makefile.

21 years agoTweak to previous commit: increment ifp->if_iqdrops if the m_copy() fails.
archie [Wed, 23 Apr 2003 23:45:57 +0000 (23:45 +0000)]
Tweak to previous commit: increment ifp->if_iqdrops if the m_copy() fails.

Suggested by: Neelkanth Natu <neelnatu@yahoo.com>

21 years agoSync with latest and greatest Apple sources (which, among other things, fix
jkh [Wed, 23 Apr 2003 23:42:23 +0000 (23:42 +0000)]
Sync with latest and greatest Apple sources (which, among other things, fix
the style flag).

Submitted by: Conrad Minshall <conrad@apple.com>
Reviewed by: jkh
Obtained from: Apple Computer, Inc.

21 years agoRemove obsolete compatibility ifdef.
imp [Wed, 23 Apr 2003 23:41:48 +0000 (23:41 +0000)]
Remove obsolete compatibility ifdef.

21 years agoCopy CIS3 and CIS4 strings into the kernel too.
imp [Wed, 23 Apr 2003 23:40:28 +0000 (23:40 +0000)]
Copy CIS3 and CIS4 strings into the kernel too.

PR: 51333
Submitted by: Scott Mitchell

21 years ago1. Allow drivers to query CIS strings from OLDCARD.
imp [Wed, 23 Apr 2003 23:39:21 +0000 (23:39 +0000)]
1. Allow drivers to query CIS strings from OLDCARD.
2. Include backwards compatibility good for the moment (eventually will
   be turned off in current, but allow for a short transition period).

PR: 51333
Submited by: Scott Mitchell (1)
MFC after: 2 weeks

21 years agoAdd accessor for cis4 string. Neither OLDCARD nor NEWCARD support
imp [Wed, 23 Apr 2003 23:32:53 +0000 (23:32 +0000)]
Add accessor for cis4 string.  Neither OLDCARD nor NEWCARD support
this yet, but that will change soon.

Pr: 51333
Submited by: Scott Mitchell

21 years agoAdd a macro to get the current thread mailbox pointer.
deischen [Wed, 23 Apr 2003 21:49:34 +0000 (21:49 +0000)]
Add a macro to get the current thread mailbox pointer.

21 years agoRemove the i386-specific hack (well, we only run on i386 anyways)
deischen [Wed, 23 Apr 2003 21:48:29 +0000 (21:48 +0000)]
Remove the i386-specific hack (well, we only run on i386 anyways)
to always set %gs when resuming a thread.

Install this library as libpthread instead of libkse.

21 years agoProtect thread errno from being changed while operating
deischen [Wed, 23 Apr 2003 21:46:50 +0000 (21:46 +0000)]
Protect thread errno from being changed while operating
on behalf of the KSE.

Add a kse_reinit function to reinitialize a reused KSE.

Submitted by: davidxu

21 years agoAdd a new sys/limits.h file which in turn depends on machine/_limits.h
kan [Wed, 23 Apr 2003 21:41:59 +0000 (21:41 +0000)]
Add a new sys/limits.h file which in turn depends on machine/_limits.h
to get actual constant values. This is in preparation for machine/limits.h
retirement.

Discussed on: standards@
Submitted by: Craig Rodrigues <rodrigc@attbi.com>  (*)
Modified by: kan

21 years agoIntroduce a g_waitfor_event() function which posts an event and waits for
phk [Wed, 23 Apr 2003 21:28:27 +0000 (21:28 +0000)]
Introduce a g_waitfor_event() function which posts an event and waits for
it to be run (or cancelled) and use this instead of home-rolled versions.

21 years agoDocument support for the <default> entry, which was implemented in
gad [Wed, 23 Apr 2003 21:26:12 +0000 (21:26 +0000)]
Document support for the <default> entry, which was implemented in
revision 1.50 of newsyslog.c , back in February.  I intend to MFC
those changes soon.

MFC after: 4 days

21 years agoMore of the event stuff can now be private to geom_event.c
phk [Wed, 23 Apr 2003 20:54:42 +0000 (20:54 +0000)]
More of the event stuff can now be private to geom_event.c

21 years agoRename g_call_me() to g_post_event(), and give it a flag
phk [Wed, 23 Apr 2003 20:46:12 +0000 (20:46 +0000)]
Rename g_call_me() to g_post_event(), and give it a flag
argument to determine if we can M_WAITOK in malloc.

21 years agostyle(9): use NULL when assigning/comparing a char *,
johan [Wed, 23 Apr 2003 20:33:57 +0000 (20:33 +0000)]
style(9): use NULL when assigning/comparing a char *,
  consitently compare fileid with 0 using == and !=.

Reviewed by: silence on @audit

21 years agoMake this WARNS=5 clean by adding () in 'if (a || (b && c))'.
johan [Wed, 23 Apr 2003 20:26:21 +0000 (20:26 +0000)]
Make this WARNS=5 clean by adding () in 'if (a || (b && c))'.

Look at by: sebastian.ssmoller@web.de on audit@

21 years agoRemove the now unused hardcoded g_post_event() event support.
phk [Wed, 23 Apr 2003 20:25:33 +0000 (20:25 +0000)]
Remove the now unused hardcoded g_post_event() event support.

21 years agoMake this WARNS=5 clean by using %j and (uintmax_t).
johan [Wed, 23 Apr 2003 20:24:13 +0000 (20:24 +0000)]
Make this WARNS=5 clean by using %j and (uintmax_t).

Look at by: sebastian.ssmoller@web.de on audit@

21 years agoThese are WARNS=5 clean, mark them as such to keep them that way.
johan [Wed, 23 Apr 2003 20:22:33 +0000 (20:22 +0000)]
These are WARNS=5 clean, mark them as such to keep them that way.

Look at by: sebastian.ssmoller@web.de on audit@

21 years agoUse bus_space*() instead of in*() and out*().
jhay [Wed, 23 Apr 2003 20:22:32 +0000 (20:22 +0000)]
Use bus_space*() instead of in*() and out*().

21 years agoTurn EV_NEW_PROVIDER into a g_call_me() event.
phk [Wed, 23 Apr 2003 20:16:13 +0000 (20:16 +0000)]
Turn EV_NEW_PROVIDER into a g_call_me() event.

21 years agoConvert EV_SPOILED event to use g_call_me().
phk [Wed, 23 Apr 2003 20:06:38 +0000 (20:06 +0000)]
Convert EV_SPOILED event to use g_call_me().

21 years agoRemove Giant from osigblock(), osigsetmask(), and kern_sigaltstack().
jhb [Wed, 23 Apr 2003 19:49:18 +0000 (19:49 +0000)]
Remove Giant from osigblock(), osigsetmask(), and kern_sigaltstack().

21 years agoTurn the hardwired NEW_CLASS event into a g_call_me() event.
phk [Wed, 23 Apr 2003 19:34:38 +0000 (19:34 +0000)]
Turn the hardwired NEW_CLASS event into a g_call_me() event.