]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoInfrastructure tweaks to allow having both an Elf32 and an Elf64 executable
Peter Wemm [Sat, 20 Jul 2002 02:56:12 +0000 (02:56 +0000)]
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable
handler in the kernel at the same time.  Also, allow for the
exec_new_vmspace() code to build a different sized vmspace depending on
the executable environment.  This is a big help for execing i386 binaries
on ia64.   The ELF exec code grows the ability to map partial pages when
there is a page size difference, eg: emulating 4K pages on 8K or 16K
hardware pages.

Flesh out the i386 emulation support for ia64.  At this point, the only
binary that I know of that fails is cvsup, because the cvsup runtime
tries to execute code in pages not marked executable.

Obtained from:  dfr (mostly, many tweaks from me).

22 years agoFix a warning:
Peter Wemm [Sat, 20 Jul 2002 01:09:35 +0000 (01:09 +0000)]
Fix a warning:
ffs_softdep.c:1630: warning: int format, different type arg (arg 2)

22 years agoAdd PCI IDs for the A, B, and C UHCI controllers in the ICH4.
John Baldwin [Fri, 19 Jul 2002 22:17:02 +0000 (22:17 +0000)]
Add PCI IDs for the A, B, and C UHCI controllers in the ICH4.

Sponsored by: The Weather Channel

22 years agoAdd PCI ID for the ICH4 ATA100 controller.
John Baldwin [Fri, 19 Jul 2002 22:14:54 +0000 (22:14 +0000)]
Add PCI ID for the ICH4 ATA100 controller.

Sponsored by: The Weather Channel

22 years agoSet P_NOLOAD on the pagezero kthread so that it doesn't artificially skew
Peter Wemm [Fri, 19 Jul 2002 21:06:01 +0000 (21:06 +0000)]
Set P_NOLOAD on the pagezero kthread so that it doesn't artificially skew
the loadav.  This is not real load.  If you have a nice process running in
the background, pagezero may sit in the run queue for ages and add one to
the loadav, and thereby affecting other scheduling decisions.

22 years ago o Use vm_page_alloc(... | VM_ALLOC_WIRED) in place of vm_page_wire().
Alan Cox [Fri, 19 Jul 2002 19:35:06 +0000 (19:35 +0000)]
 o Use vm_page_alloc(... | VM_ALLOC_WIRED) in place of vm_page_wire().

22 years agoOnly define SHLIB_LINK if SHLIB_NAME matches the *.so.* pattern.
Ruslan Ermilov [Fri, 19 Jul 2002 18:40:13 +0000 (18:40 +0000)]
Only define SHLIB_LINK if SHLIB_NAME matches the *.so.* pattern.
(Useful for RELENG_4's lib/libpam/modules.)

22 years agoAdd support for probing secondary buses on the ServerWorks Grand Champion
Andrew Gallatin [Fri, 19 Jul 2002 18:35:08 +0000 (18:35 +0000)]
Add support for probing secondary buses on the ServerWorks Grand Champion
chipset used for P4-Xeon machines

PR: kern/38894
Tested-by: "Marc G. Fournier" <scrappy@hub.org>
Submitted-by: Mark Tinguely (partially)
22 years agoAdd the tcps_sndrexmitbad statistic, keep track of late acks that caused
Matthew Dillon [Fri, 19 Jul 2002 18:29:38 +0000 (18:29 +0000)]
Add the tcps_sndrexmitbad statistic, keep track of late acks that caused
unnecessary retransmissions.

22 years agoWrap a line longer than 80 characters.
Maxime Henrion [Fri, 19 Jul 2002 17:44:44 +0000 (17:44 +0000)]
Wrap a line longer than 80 characters.

22 years agoMerge conflicts.
Mark Murray [Fri, 19 Jul 2002 16:38:21 +0000 (16:38 +0000)]
Merge conflicts.

22 years ago- Merge the mount options at MNT_UPDATE time with vfs_mergeopts().
Maxime Henrion [Fri, 19 Jul 2002 16:05:31 +0000 (16:05 +0000)]
- Merge the mount options at MNT_UPDATE time with vfs_mergeopts().
- Sanity check the mount options list (remove duplicates) with
  vfs_sanitizeopts().
- Fix some malloc(0)/free(NULL) bugs.

Reviewed by: rwatson (some time ago)

22 years agoThis commit was generated by cvs2svn to compensate for changes in r100360,
Mark Murray [Fri, 19 Jul 2002 15:19:23 +0000 (15:19 +0000)]
This commit was generated by cvs2svn to compensate for changes in r100360,
which included commits to RCS files with non-trunk default branches.

22 years agoSnapshot of NetBSD's (x)lint. Little functionality, just some code
Mark Murray [Fri, 19 Jul 2002 15:19:23 +0000 (15:19 +0000)]
Snapshot of NetBSD's (x)lint. Little functionality, just some code
tidy-ups.

22 years agos/inline/__inline/
Mark Murray [Fri, 19 Jul 2002 14:12:27 +0000 (14:12 +0000)]
s/inline/__inline/

22 years agoMerge from import. The import was done ages ago, and this diff has been
Mark Murray [Fri, 19 Jul 2002 14:10:35 +0000 (14:10 +0000)]
Merge from import. The import was done ages ago, and this diff has been
carried since then.

22 years agos/inline/__inline/
Mark Murray [Fri, 19 Jul 2002 13:49:37 +0000 (13:49 +0000)]
s/inline/__inline/

22 years ago"inline" fixing. Replace "inline" with "__inline" to make more BSD
Mark Murray [Fri, 19 Jul 2002 13:38:43 +0000 (13:38 +0000)]
"inline" fixing. Replace "inline" with "__inline" to make more BSD
standard (and easier to define away with support in cdefs.h).
Also convert two function-like macros to static inline functions
for lint and the debugger.

22 years agoUpdate to current reality: change COMPILEDIR and work on "R"ead-only-data
Poul-Henning Kamp [Fri, 19 Jul 2002 10:29:07 +0000 (10:29 +0000)]
Update to current reality: change COMPILEDIR and work on "R"ead-only-data
symbols as well.

22 years agoRemove broken and incomplete support for old releases of System V,
Tim J. Robbins [Fri, 19 Jul 2002 08:09:04 +0000 (08:09 +0000)]
Remove broken and incomplete support for old releases of System V,
don't support system that implement getcwd(3) with a pipe to /bin/pwd.

22 years agoAdd the ebus sio attachment.
David E. O'Brien [Fri, 19 Jul 2002 08:03:14 +0000 (08:03 +0000)]
Add the ebus sio attachment.

22 years agoDon't install any old cruft present in the tree, including
Ruslan Ermilov [Fri, 19 Jul 2002 07:51:58 +0000 (07:51 +0000)]
Don't install any old cruft present in the tree, including
editor backups, .orig or .rej files, etc.  Make transition
from SHARED=symlinks to SHARED=copies and vice versa work.

Submitted by: des, ru

22 years agoAdd support to UFS2 to provide storage for extended attributes.
Kirk McKusick [Fri, 19 Jul 2002 07:29:39 +0000 (07:29 +0000)]
Add support to UFS2 to provide storage for extended attributes.
As this code is not actually used by any of the existing
interfaces, it seems unlikely to break anything (famous
last words).

The internal kernel interface to manipulate these attributes
is invoked using two new IO_ flags: IO_NORMAL and IO_EXT.
These flags may be specified in the ioflags word of VOP_READ,
VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that
you want to do I/O to the normal data part of the file and
IO_EXT means that you want to do I/O to the extended attributes
part of the file. IO_NORMAL and IO_EXT are mutually exclusive
for VOP_READ and VOP_WRITE, but may be specified individually
or together in the case of VOP_TRUNCATE. For example, when
removing a file, VOP_TRUNCATE is called with both IO_NORMAL
and IO_EXT set. For backward compatibility, if neither IO_NORMAL
nor IO_EXT is set, then IO_NORMAL is assumed.

Note that the BA_ and IO_ flags have been `merged' so that they
may both be used in the same flags word. This merger is possible
by assigning the IO_ flags to the low sixteen bits and the BA_
flags the high sixteen bits. This works because the high sixteen
bits of the IO_ word is reserved for read-ahead and help with
write clustering so will never be used for flags. This merge
lets us get away from code of the form:

        if (ioflags & IO_SYNC)
                flags |= BA_SYNC;

For the future, I have considered adding a new field to the
vattr structure, va_extsize. This addition could then be
exported through the stat structure to allow applications to
find out the size of the extended attribute storage and also
would provide a more standard interface for truncating them
(via VOP_SETATTR rather than VOP_TRUNCATE).

I am also contemplating adding a pathconf parameter (for
concreteness, lets call it _PC_MAX_EXTSIZE) which would
let an application determine the maximum size of the extended
atribute storage.

Sponsored by: DARPA & NAI Labs.

22 years ago o Duplicate an odd side-effect of vm_page_wire() in vm_page_allocate()
Alan Cox [Fri, 19 Jul 2002 03:33:04 +0000 (03:33 +0000)]
 o Duplicate an odd side-effect of vm_page_wire() in vm_page_allocate()
   when VM_ALLOC_WIRED is specified: set the PG_MAPPED bit in flags.
 o In both vm_page_wire() and vm_page_allocate() add a comment saying
   that setting PG_MAPPED does not belong there.

22 years agoClear up confusion in ugly code. ^T gave wrong results for RSS.
Julian Elischer [Thu, 18 Jul 2002 21:19:56 +0000 (21:19 +0000)]
Clear up confusion in ugly code.  ^T gave wrong results for RSS.
I misinterpretted this code when changing it to handle threads.
(there are still issues here)
Submitted by: Ian Dowse <iedowse@maths.tcd.ie>

22 years agoTry to give a more descriptive error message for the pilot error of
Joerg Wunsch [Thu, 18 Jul 2002 20:52:17 +0000 (20:52 +0000)]
Try to give a more descriptive error message for the pilot error of
attempting to export the non-root of a filesystem with -alldirs.  This
pilot error seems to be very common, and the "could not remount" error
message doesn't give much hints about the real reason.  See the old PR
below for an example.

While i was at it, make it possible to entirely omit the often
annoying error message in that case by specifying the "quiet" exports
flag.  This allows to specify something like

/cdrom -alldirs,ro,quiet <where to export to>

which will silently fail if nothing is mounted under /cdrom, but do
the rigth thing as soon as you mount something.

While doing this, i've put the embedded example in the exports(5) man
page into a subsection of its own as it ought to be.

Thanks for Paul Southworth for reminding me about this problem.

PR: bin/4448
MFC after: 1 month

22 years agoIntroduce two new sysctl's:
Matthew Dillon [Thu, 18 Jul 2002 19:06:12 +0000 (19:06 +0000)]
Introduce two new sysctl's:

net.inet.tcp.rexmit_min (default 3 ticks equiv)

    This sysctl is the retransmit timer RTO minimum,
    specified in milliseconds.  This value is
    designed for algorithmic stability only.

net.inet.tcp.rexmit_slop (default 200ms)

    This sysctl is the retransmit timer RTO slop
    which is added to every retransmit timeout and
    is designed to handle protocol stack overheads
    and delayed ack issues.

Note that the *original* code applied a 1-second
RTO minimum but never applied real slop to the RTO
calculation, so any RTO calculation over one second
would have no slop and thus not account for
protocol stack overheads (TCP timestamps are not
a measure of protocol turnaround!).  Essentially,
the original code made the RTO calculation almost
completely irrelevant.

Please note that the 200ms slop is debateable.
This commit is not meant to be a line in the sand,
and if the community winds up deciding that increasing
it is the correct solution then it's easy to do.
Note that larger values will destroy performance
on lossy networks while smaller values may result in
a greater number of unnecessary retransmits.

22 years agoRemove unused calls to inet_addr()
Brian Somers [Thu, 18 Jul 2002 18:50:05 +0000 (18:50 +0000)]
Remove unused calls to inet_addr()

22 years agoCorrection to the previous revision: define SHLIB_LINK if SHLIB_NAME
Ruslan Ermilov [Thu, 18 Jul 2002 17:44:20 +0000 (17:44 +0000)]
Correction to the previous revision: define SHLIB_LINK if SHLIB_NAME
is defined (whether or not LIB is defined).

22 years ago o Remove the acquisition and release of Giant from the idle priority thread
Alan Cox [Thu, 18 Jul 2002 17:40:07 +0000 (17:40 +0000)]
 o Remove the acquisition and release of Giant from the idle priority thread
   that pre-zeroes free pages.
 o Remove GIANT_REQUIRED from some low-level page queue functions.  (Instead
   assertions on the page queue lock are being added to the higher-level
   functions, like vm_page_wire(), etc.)

In collaboration with: peter

22 years agoMFCs noted: OPIE 2.4, OpenBSD m4 20020426, GNU tar 1.3.25.
Bruce A. Mah [Thu, 18 Jul 2002 16:06:46 +0000 (16:06 +0000)]
MFCs noted:  OPIE 2.4, OpenBSD m4 20020426, GNU tar 1.3.25.

22 years agoBeautify. This has the side effect of improving portability and
Mark Murray [Thu, 18 Jul 2002 15:56:46 +0000 (15:56 +0000)]
Beautify. This has the side effect of improving portability and
making lint work cleaner.

Inspired to do this by: jhb

22 years agoVoid functions cannot return values.
Mark Murray [Thu, 18 Jul 2002 15:53:11 +0000 (15:53 +0000)]
Void functions cannot return values.

22 years agoRemove boot2 temporarily.
Yoshihiro Takahashi [Thu, 18 Jul 2002 14:50:41 +0000 (14:50 +0000)]
Remove boot2 temporarily.

22 years agoAdd initialization code for the AMD Elan sc520 which maps the MMCR
Poul-Henning Kamp [Thu, 18 Jul 2002 12:56:54 +0000 (12:56 +0000)]
Add initialization code for the AMD Elan sc520 which maps the MMCR
into KVM and sets the i8254 frequency to the correct value.

22 years agoTo force install(1) to always compare files before installing, one
Ruslan Ermilov [Thu, 18 Jul 2002 12:54:55 +0000 (12:54 +0000)]
To force install(1) to always compare files before installing, one
now needs to set COPY=-C as -C is no longer compatible with the -d
option.  It is also likely to be renamed to INSTALL_COPY soon.
Update documentation to reflect this change.

PR: bin/40724

22 years agoRemove the statically allocated array that holds OpenFirmware memory mappings
Benno Rice [Thu, 18 Jul 2002 12:43:08 +0000 (12:43 +0000)]
Remove the statically allocated array that holds OpenFirmware memory mappings
during pmap_bootstrap.  Instead, temporarily help ourselves to some memory
from phys_avail since we won't need it post-boostrap.

22 years agoMajor rework of how we copy data into kernel space.
Benno Rice [Thu, 18 Jul 2002 12:39:02 +0000 (12:39 +0000)]
Major rework of how we copy data into kernel space.

We now talk to the memory and mmu instances directly rather than using the
OpenFirmware "claim" method.

22 years agoInstall scripts via FILES (purposedly not via SCRIPTS that would
Ruslan Ermilov [Thu, 18 Jul 2002 12:33:01 +0000 (12:33 +0000)]
Install scripts via FILES (purposedly not via SCRIPTS that would
strip the suffixes).

22 years agoDon't assume file descriptors fit in a short, use an int instead.
Tim J. Robbins [Thu, 18 Jul 2002 12:24:35 +0000 (12:24 +0000)]
Don't assume file descriptors fit in a short, use an int instead.

22 years agoAdded groff(1) and natd(8) entries.
Ruslan Ermilov [Thu, 18 Jul 2002 12:12:15 +0000 (12:12 +0000)]
Added groff(1) and natd(8) entries.

22 years agos/${INSTALL} -c/${INSTALL} ${COPY}/
Ruslan Ermilov [Thu, 18 Jul 2002 12:07:49 +0000 (12:07 +0000)]
s/${INSTALL} -c/${INSTALL} ${COPY}/

22 years agos/install -c/${INSTALL} ${COPY}/
Ruslan Ermilov [Thu, 18 Jul 2002 12:02:20 +0000 (12:02 +0000)]
s/install -c/${INSTALL} ${COPY}/

22 years agoAdd an entry for the AMD Elan SC520 hostbridge. I do not belive we can
Poul-Henning Kamp [Thu, 18 Jul 2002 10:57:20 +0000 (10:57 +0000)]
Add an entry for the AMD Elan SC520 hostbridge.  I do not belive we can
identify this gadget on the CPUID result alone, so I intend to activate
the necessary magic (i8254 frequency for instance) for it based on the
precense of the on-chip host to PCI bridge.

22 years ago(VM_MAX_KERNEL_ADDRESS - KERNBASE) / PAGE_SIZE may not fit in an integer.
Peter Wemm [Thu, 18 Jul 2002 10:28:00 +0000 (10:28 +0000)]
(VM_MAX_KERNEL_ADDRESS - KERNBASE) / PAGE_SIZE  may not fit in an integer.
Use lmin(long, long), not min(u_int, u_int).  This is a problem here on
ia64 which has *way* more than 2^32 pages of KVA.  281474976710655 pages
to be precice.

22 years agoAvoid using ints or shorts to store process id's, use pid_t instead.
Tim J. Robbins [Thu, 18 Jul 2002 10:22:42 +0000 (10:22 +0000)]
Avoid using ints or shorts to store process id's, use pid_t instead.
The pgrp member of struct job was declared as a short and could not store
every possible process group ID value, the rest of them were benign because
pid_t happens to be an int.

22 years agoMake the message that is printed when the foreground process is terminated
Tim J. Robbins [Thu, 18 Jul 2002 09:37:51 +0000 (09:37 +0000)]
Make the message that is printed when the foreground process is terminated
by a signal the same as pdksh/bash/sh before rev. 1.34.

22 years agoMove the definition of FreeBSD 4.6 back to where it belongs.
Ruslan Ermilov [Thu, 18 Jul 2002 09:32:15 +0000 (09:32 +0000)]
Move the definition of FreeBSD 4.6 back to where it belongs.

22 years agoAdd definition for FreeBSD 4.6.1.
Murray Stokely [Thu, 18 Jul 2002 09:22:33 +0000 (09:22 +0000)]
Add definition for FreeBSD 4.6.1.

22 years agoRevert last change.
Murray Stokely [Thu, 18 Jul 2002 09:15:54 +0000 (09:15 +0000)]
Revert last change.

Requested by: ru

22 years agoIntegrate the hw.pcic.pd6722_vsense tunable from the nomads list.
Warner Losh [Thu, 18 Jul 2002 08:13:45 +0000 (08:13 +0000)]
Integrate the hw.pcic.pd6722_vsense tunable from the nomads list.
This allows one to select the method of 3.3V card detection from the
three possible choices (none (0), the "6710 way" (1) and the "6729
way" (2)).  The default is the 6710 way, since it works in the most
cases.  The datasheets for the 6722 suggest that the '29 way is more
correct, but experience has shown this method to cause some laptops to
hang solid.  See source code for details until I update the man page.

Submitted by: shibata-san (chiharu shibata <chi@bd.mbn.or.jp>)

22 years agoSome strange hacks for the clpd6729:
Warner Losh [Thu, 18 Jul 2002 08:05:00 +0000 (08:05 +0000)]
Some strange hacks for the clpd6729:
o It needs to have pcic_isa_intr intrrupt handler
o for pci interrupts, in the func interrupt handler it needs to check the isa
  registers rather than the pci ones for card present.
o better commentary for some of the strangeness of the 6729 on pci
o fix some crunchy comments to better reflect reality.

With this I almost have the WL200 working, but an interrupt storm
after attach is causing problems for reasons unknown.  This code
doesn't seem to break the normal clpd6729 case, and I'd like others
with 6729 cards to try to test it (there were some that were used for
external pccard slots in pci only systems).

22 years agoTeach mdoc about FreeBSD 4.6.1
Murray Stokely [Thu, 18 Jul 2002 08:04:12 +0000 (08:04 +0000)]
Teach mdoc about FreeBSD 4.6.1

22 years agoThe Compaq WL200 is a CL-PD6729 based pci card with a prism 2 pcmcia
Warner Losh [Thu, 18 Jul 2002 06:01:35 +0000 (06:01 +0000)]
The Compaq WL200 is a CL-PD6729 based pci card with a prism 2 pcmcia
card behind it (without the pcmcia form factor).  This entry gets to
the point of attaching, but there's something wrong with the '29
support, so it doesn't quite work yet.

22 years agoMove IPFW2 definition before including ip_fw.h
Luigi Rizzo [Thu, 18 Jul 2002 05:18:41 +0000 (05:18 +0000)]
Move IPFW2 definition before including ip_fw.h

Make indentation of new parts consistent with the style used for this file.

22 years agoRemove spurious "echo '.'".
Dima Dorfman [Thu, 18 Jul 2002 05:00:23 +0000 (05:00 +0000)]
Remove spurious "echo '.'".

22 years agoApply same fix as Rev 1.19 of /etc/rc.diskless2: create sendmail
Brian S. Dean [Thu, 18 Jul 2002 05:00:22 +0000 (05:00 +0000)]
Apply same fix as Rev 1.19 of /etc/rc.diskless2: create sendmail
required directories if sendmail_enable is not set to "none".

Suggested by:   gordon

22 years agoAnonymize the "portmap" program to get better compatibility with
Doug Barton [Thu, 18 Jul 2002 05:00:21 +0000 (05:00 +0000)]
Anonymize the "portmap" program to get better compatibility with
rpcbind in -current.

Submitted by:   Alexander Kabaev <ak03@gte.com>

22 years agoMake nisdomainname=NO DTRT
Doug Barton [Thu, 18 Jul 2002 05:00:20 +0000 (05:00 +0000)]
Make nisdomainname=NO DTRT

Submitted by:   des, via Mike Makonnen <makonnen@pacbell.net>

22 years agoCleanup some pollution from the NetBSD sync, and add gif setup.
Doug Barton [Thu, 18 Jul 2002 05:00:19 +0000 (05:00 +0000)]
Cleanup some pollution from the NetBSD sync, and add gif setup.

Submitted by:   Mike Makonnen <makonnen@pacbell.net>

22 years agoFix a typo that caused dhclient not to work.
Gordon Tetlow [Thu, 18 Jul 2002 05:00:18 +0000 (05:00 +0000)]
Fix a typo that caused dhclient not to work.

Submitted by:   Dennis Kristensen <snicki@snicki.dk>
Reviewed by:    Mike Makonnen <makonnen@pacbell.net>

22 years agoMerge in all the changes that Mike Makonnen has been maintaining for a
Gordon Tetlow [Thu, 18 Jul 2002 05:00:17 +0000 (05:00 +0000)]
Merge in all the changes that Mike Makonnen has been maintaining for a
while. This is only the script pieces, the glue for the build comes next.

Submitted by:   Mike Makonnen <makonnen@pacbell.net>
Reviewed by:    silence on -current and -hackers
Prodded by:     rwatson

22 years agobe able to configure to run an IPv6 routing daemon even on
Hajimu UMEMOTO [Thu, 18 Jul 2002 05:00:16 +0000 (05:00 +0000)]
be able to configure to run an IPv6 routing daemon even on
an end node.

Requested by: Masachika ISHIZUKA <ishizuka@ish.org>
MFC after: 1 week

22 years agofixed to make mbuf chain.
Hajimu UMEMOTO [Thu, 18 Jul 2002 04:18:56 +0000 (04:18 +0000)]
fixed to make mbuf chain.

Obtained from: KAME
MFC after: 1 week

22 years ago o Introduce an argument, VM_ALLOC_WIRED, that requests vm_page_alloc()
Alan Cox [Thu, 18 Jul 2002 04:08:10 +0000 (04:08 +0000)]
 o Introduce an argument, VM_ALLOC_WIRED, that requests vm_page_alloc()
   to return a wired page.
 o Use VM_ALLOC_WIRED within Alpha's pmap_growkernel().  Also, because
   Alpha's pmap_growkernel() calls vm_page_alloc() from within a critical
   section, specify VM_ALLOC_INTERRUPT instead of VM_ALLOC_SYSTEM.  (Only
   VM_ALLOC_INTERRUPT is implemented entirely with a spin mutex.)
 o Assert that the page queues mutex is held in vm_page_wire()
   on Alpha, just like the other platforms.

22 years agoUse pmap_kenter() rather than vtopte() and bashing the page tables
Peter Wemm [Thu, 18 Jul 2002 00:42:53 +0000 (00:42 +0000)]
Use pmap_kenter() rather than vtopte() and bashing the page tables
directly.

22 years agoFix a transcription typo. s/ACPI_PTR/ACPI_POINTER/
Peter Wemm [Wed, 17 Jul 2002 23:49:55 +0000 (23:49 +0000)]
Fix a transcription typo.  s/ACPI_PTR/ACPI_POINTER/

22 years agoia64 does not have the same degree of stealth include file nesting,
Peter Wemm [Wed, 17 Jul 2002 23:43:55 +0000 (23:43 +0000)]
ia64 does not have the same degree of stealth include file nesting,
so it needs an explicit #include <machine/frame.h> to get 'struct
trapframe'.  The fact that it needs this at this level is rather bogus
but it will not compile without it.

22 years agoThis is a userland tool, not a kernel component or libstand client.
Peter Wemm [Wed, 17 Jul 2002 23:41:58 +0000 (23:41 +0000)]
This is a userland tool, not a kernel component or libstand client.

22 years agoPacify gcc on ia64
Peter Wemm [Wed, 17 Jul 2002 23:32:13 +0000 (23:32 +0000)]
Pacify gcc on ia64

22 years agoI don't know how the minimum retransmit timeout managed to get set to
Matthew Dillon [Wed, 17 Jul 2002 23:32:03 +0000 (23:32 +0000)]
I don't know how the minimum retransmit timeout managed to get set to
one second but it badly breaks throughput on networks with minor packet
loss.

Complaints by: at least two people tracked down to this.
MFC after: 3 days

22 years agoFix some typos in 1.68 from over a week ago.
Peter Wemm [Wed, 17 Jul 2002 23:22:50 +0000 (23:22 +0000)]
Fix some typos in 1.68 from over a week ago.

22 years agoCap the initial PV and PTE table preallocations. Otherwise we explode
Peter Wemm [Wed, 17 Jul 2002 23:21:59 +0000 (23:21 +0000)]
Cap the initial PV and PTE table preallocations.  Otherwise we explode
on the Itanium2 system I have when we use up *all* of the initial 256MB
direct mapped region before we are ready to dynamically expand it.

The machine that I have has 4 cpus and a very big hole in the middle.
This makes the bogus '(last_address - first_address) / PAGE_SIZE'
calculations especially dangerous and caused many millions of initial
PV/PTE's to be preallocated.

22 years agoBe sure to use a logical address for the SAL table. For some reason the
Peter Wemm [Wed, 17 Jul 2002 23:17:49 +0000 (23:17 +0000)]
Be sure to use a logical address for the SAL table.  For some reason the
phsysical address is still mapped at this stage of boot on the Itanium1
SDV boxes we have.  But Itanium2 does *not* let us get away with this.

22 years agoUpdate for new ACPICA import. Gah.
Peter Wemm [Wed, 17 Jul 2002 23:14:41 +0000 (23:14 +0000)]
Update for new ACPICA import.  Gah.

22 years agoAdd END markers to asm functions so that debuggers can find their size.
Jake Burkholder [Wed, 17 Jul 2002 22:20:41 +0000 (22:20 +0000)]
Add END markers to asm functions so that debuggers can find their size.

22 years agoAvoid trying to set PG_G on the first 4MB when we set up the 4MB page.
Peter Wemm [Wed, 17 Jul 2002 21:47:05 +0000 (21:47 +0000)]
Avoid trying to set PG_G on the first 4MB when we set up the 4MB page.
This solves the SMP panic for at least one system.  I'd still like to know
why my xeon works though.

Tested by: bmilekic

22 years agoAvoid passing NULL to freehostent(3).
Yaroslav Tykhiy [Wed, 17 Jul 2002 19:29:25 +0000 (19:29 +0000)]
Avoid passing NULL to freehostent(3).

MFC after: 1 week

22 years agoFix a reversed test.
Julian Elischer [Wed, 17 Jul 2002 19:20:48 +0000 (19:20 +0000)]
Fix a reversed test.
Fix some style nits.
Fix a KASSERT message.
Add/fix some comments.

Submitted by: bde@freebsd.org

22 years agoMake sure the process state for the idle proc is set correctly
Julian Elischer [Wed, 17 Jul 2002 19:18:45 +0000 (19:18 +0000)]
Make sure the process state for the idle proc is set correctly
from the beginning.

22 years agoFix setting parameters for getipnodebyaddr(3):
Yaroslav Tykhiy [Wed, 17 Jul 2002 19:07:07 +0000 (19:07 +0000)]
Fix setting parameters for getipnodebyaddr(3):
o "struct addrinfo" contains a pointer to "struct sockaddr,"
  not "struct sockaddr" itself
o the function takes a pointer to "struct in*_addr", not to
  "struct sockaddr," so the address length must be corresponding

MFC after: 1 week

22 years agoAdd myself.
Ville Skyttä [Wed, 17 Jul 2002 19:00:37 +0000 (19:00 +0000)]
Add myself.

Approved by: knu (mentor)

22 years agoIn pam_init_ctx(), register a cleanup function that will kill the child
Dag-Erling Smørgrav [Wed, 17 Jul 2002 17:44:02 +0000 (17:44 +0000)]
In pam_init_ctx(), register a cleanup function that will kill the child
process if a fatal error occurs.  Deregister it in pam_free_ctx().

22 years agoSort FILES.
Ruslan Ermilov [Wed, 17 Jul 2002 16:35:00 +0000 (16:35 +0000)]
Sort FILES.

22 years agoFix a spelling mistake.
David Malone [Wed, 17 Jul 2002 16:22:24 +0000 (16:22 +0000)]
Fix a spelling mistake.

PR: 40695
Submitted by: Maxim Maximov <mcsi@agava.com>
MFC after: 1 week

22 years agoClean up the syntax WRT semicolons at the end of function-like-macros, and protect...
Mark Murray [Wed, 17 Jul 2002 16:19:37 +0000 (16:19 +0000)]
Clean up the syntax WRT semicolons at the end of function-like-macros, and protect GCCisms from non-GNU compilers and lint.

22 years agoFix installworld.
Ollivier Robert [Wed, 17 Jul 2002 15:35:31 +0000 (15:35 +0000)]
Fix installworld.

Submitted by: Udo Schweigert <Udo.Schweigert@siemens.com>

22 years agoOne external utility was ommitted in the previous delta:
Sheldon Hearn [Wed, 17 Jul 2002 15:34:25 +0000 (15:34 +0000)]
One external utility was ommitted in the previous delta:

alias

22 years agoMFi386: revision 1.524.
Yoshihiro Takahashi [Wed, 17 Jul 2002 12:43:52 +0000 (12:43 +0000)]
MFi386: revision 1.524.

22 years agoMFi386: revision 1.350.
Yoshihiro Takahashi [Wed, 17 Jul 2002 12:42:37 +0000 (12:42 +0000)]
MFi386: revision 1.350.

22 years agodi_createtime -> di_birthtime.
Ollivier Robert [Wed, 17 Jul 2002 10:31:38 +0000 (10:31 +0000)]
di_createtime -> di_birthtime.

Submitted by: Udo Schweigert <Udo.Schweigert@siemens.com>

22 years agoHandle installation of links through bsd.links.mk.
Ruslan Ermilov [Wed, 17 Jul 2002 08:21:50 +0000 (08:21 +0000)]
Handle installation of links through bsd.links.mk.
Removed comments that no longer directly apply here.

22 years agoFix a panic when doing "ipfw add pipe 1 log ..."
Luigi Rizzo [Wed, 17 Jul 2002 07:21:42 +0000 (07:21 +0000)]
Fix a panic when doing "ipfw add pipe 1 log ..."

Also synchronize ip_dummynet.c with the version in RELENG_4 to
ease MFC's.

22 years agoThe following builtins are now available as external commands as well:
Sheldon Hearn [Wed, 17 Jul 2002 07:12:37 +0000 (07:12 +0000)]
The following builtins are now available as external commands as well:

bg
        cd
        command
        fc
        fg
        getopts
        jobs
        read
        umask
        unalias
        wait

22 years agoBetter verbage on hw.pcic sysctls, delete now obsolete comment
Warner Losh [Wed, 17 Jul 2002 06:29:35 +0000 (06:29 +0000)]
Better verbage on hw.pcic sysctls, delete now obsolete comment

22 years agoo Remove workaround that I put in to mask the BadVcc problem.
Warner Losh [Wed, 17 Jul 2002 06:02:07 +0000 (06:02 +0000)]
o Remove workaround that I put in to mask the BadVcc problem.
o Add preliminary support for Cirrus Logic CL-PD6729 using PCI
  interrupts.  To use it you you need to set hw.pcic.pd6729_intr_path
  to 2.  This is allow us to still default to ISA intrrupt path for
  this part (which is found much more often in laptops using ISA IRQs).
  But some PCI cards have this part on them and this should allow them
  to be used.  It is untested on PCI, but it seems to not break the ISA
  case.
o Better sysctl descriptions (I hope).

22 years agoBe more conservative about the address ranges we assign. Some
Warner Losh [Wed, 17 Jul 2002 05:50:06 +0000 (05:50 +0000)]
Be more conservative about the address ranges we assign.  Some
machines don't like the more liberal default, so be more conservative
about what we do by default.

22 years agoGLOB_QUOTE has been retired.
Mike Heffner [Wed, 17 Jul 2002 05:47:49 +0000 (05:47 +0000)]
GLOB_QUOTE has been retired.

22 years agoAdd definitions for the Ext 1 register on cirrus logic PD-6729.
Warner Losh [Wed, 17 Jul 2002 05:43:39 +0000 (05:43 +0000)]
Add definitions for the Ext 1 register on cirrus logic PD-6729.

22 years agoQualify comment on machdep.cpu_idle_hlt. Turning this on on a SMP
Matthew Dillon [Wed, 17 Jul 2002 05:41:43 +0000 (05:41 +0000)]
Qualify comment on machdep.cpu_idle_hlt.  Turning this on on a SMP
machine will result in approximately a 4.2% loss of performance (buildworld)
and approximately a 5% reduction in power consumption (when idle).  Add XXX
note on how to really make hlt work (send an IPI to wakeup HLTed cpus on
a thread-schedule event?  Generate an interrupt somehow?).