]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoThis commit was generated by cvs2svn to compensate for changes in r108968,
Peter Wemm [Wed, 8 Jan 2003 22:12:01 +0000 (22:12 +0000)]
This commit was generated by cvs2svn to compensate for changes in r108968,
which included commits to RCS files with non-trunk default branches.

21 years agoAdd in a patch from tcpdump-current (that will be included in the next
Peter Wemm [Wed, 8 Jan 2003 22:12:01 +0000 (22:12 +0000)]
Add in a patch from tcpdump-current (that will be included in the next
import) to stop a segfault on ia64.  Add a #include to get missing
prototypes, particularly malloc().

Approved by: fenner

21 years agoAdd code to parse the contents of an ACPI MADT table and to dump its
John Baldwin [Wed, 8 Jan 2003 20:46:25 +0000 (20:46 +0000)]
Add code to parse the contents of an ACPI MADT table and to dump its
contents in the form of a comment in the acpidump output.  The output
is not in the prettiest output, but it does work correctly.

Approved by: marcel

21 years agoAdd structure definitions for the various items in a ACPI Multiple
John Baldwin [Wed, 8 Jan 2003 20:45:23 +0000 (20:45 +0000)]
Add structure definitions for the various items in a ACPI Multiple
APIC Descriptor Table (MADT) which has the signature 'ACPI'.

Approved by: marcel

21 years agoBuild a gem module, for sparc64 only for now.
Thomas Moestl [Wed, 8 Jan 2003 20:40:29 +0000 (20:40 +0000)]
Build a gem module, for sparc64 only for now.

21 years agoAdd detach, shutdown, suspend and resume methods. The latter two are
Thomas Moestl [Wed, 8 Jan 2003 20:12:45 +0000 (20:12 +0000)]
Add detach, shutdown, suspend and resume methods. The latter two are
not really tested, but are derived from the original NetBSD version.

21 years agoIn vm_page_alloc(), honor VM_ALLOC_ZERO for system and interrupt class
Alan Cox [Wed, 8 Jan 2003 19:58:42 +0000 (19:58 +0000)]
In vm_page_alloc(), honor VM_ALLOC_ZERO for system and interrupt class
requests when the number of free pages is below the reserved threshold.
Previously, VM_ALLOC_ZERO was only honored when the number of free pages
was above the reserved threshold.  Honoring it in all cases generally
makes sense, does no harm, and simplifies the code.

21 years agoConsistently use spaces in between arguments to strcmp(). Whitespace
John Baldwin [Wed, 8 Jan 2003 19:16:11 +0000 (19:16 +0000)]
Consistently use spaces in between arguments to strcmp().  Whitespace
only.

21 years agoFix module dependancy name.
David E. O'Brien [Wed, 8 Jan 2003 18:53:38 +0000 (18:53 +0000)]
Fix module dependancy name.

PR: 46871
Submitted by: Hartmut Brandt <brandt@fokus.gmd.de>

21 years agoTighten wording of comment.
Garrett Wollman [Wed, 8 Jan 2003 18:52:32 +0000 (18:52 +0000)]
Tighten wording of comment.

Suggested by: gshapiro

21 years agoDo not do manually what sendmail(8) can do better automatically.
Garrett Wollman [Wed, 8 Jan 2003 18:51:45 +0000 (18:51 +0000)]
Do not do manually what sendmail(8) can do better automatically.
Tell sendmail to clean up its own host status cache.
The error condition handling could probably be done better.

21 years agoAdd an hme(4) module.
David E. O'Brien [Wed, 8 Jan 2003 18:45:04 +0000 (18:45 +0000)]
Add an hme(4) module.

PR: 46871
Submitted by: Hartmut Brandt <brandt@fokus.gmd.de>

21 years ago"struct route" is not sufficient. NetBSD PR 18751
Hajimu UMEMOTO [Wed, 8 Jan 2003 17:59:24 +0000 (17:59 +0000)]
"struct route" is not sufficient.  NetBSD PR 18751

Obtained from: KAME
MFC after: 1 days

21 years agoMerge the following from the English version:
Hiroki Sato [Wed, 8 Jan 2003 17:57:41 +0000 (17:57 +0000)]
Merge the following from the English version:

1.8   -> 1.9   early-adopter/article.sgml
1.9   -> 1.10  errata/article.sgml
1.3   -> 1.4   hardware/common/artheader.sgml

Submitted by: Hideyuki KURASHINA <rushani@jp.FreeBSD.org>
References: [doc-jp-work 572]

21 years agoDont attach a Promise chip located behind a i960 bridge/chip.
Søren Schmidt [Wed, 8 Jan 2003 17:44:36 +0000 (17:44 +0000)]
Dont attach a Promise chip located behind a i960 bridge/chip.
This makes it possible to run a Promise SuperTrak SX6000 with
OS type set to "Other" as well as "Linux".

21 years agoFor now, set only the resource limits and process priority associated
Robert Watson [Wed, 8 Jan 2003 17:10:11 +0000 (17:10 +0000)]
For now, set only the resource limits and process priority associated
with a class, rather than all aspects of the class when switching
classes for an inetd service.  Because we hard-code /daemon in the
current inetd implementation, using SETALL has unfortunate side-effects
involving the MAC code, and potentially other credential related
settings in the future.  This change maintains the DoS-resistent
aspects of the class behavior, which is all that is promised in the
inetd man page.

A larger set of diffs providing more pluggability and configurability
was deferred for this more simple approach in the short term.

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

21 years agoAdd #include <stdlib.h> so that "s->strings = realloc(...)" doesn't segfault
Peter Wemm [Wed, 8 Jan 2003 17:05:52 +0000 (17:05 +0000)]
Add #include <stdlib.h> so that "s->strings = realloc(...)" doesn't segfault
on ia64.  (This file isn't vendor branched)

21 years agoAdd support for the nVidia nForce2 ATA part.
Søren Schmidt [Wed, 8 Jan 2003 16:51:41 +0000 (16:51 +0000)]
Add support for the nVidia nForce2 ATA part.

Fix support for the nForce1 as well, registers are offset 0x10
against the AMD/VIA parts.

21 years ago- Use cpu_exthigh instead of executing cpuid again to retrieve it for the
John Baldwin [Wed, 8 Jan 2003 16:41:48 +0000 (16:41 +0000)]
- Use cpu_exthigh instead of executing cpuid again to retrieve it for the
  print_AMD_foo() functions.
- Add a brand name table for the brand index provided on Intel CPU's in
  %ebx after cpuid 1.
- For Intel CPUs, if we don't get a processor name from the extended cpuid
  then use the brand index in cpuid_cpuinfo to pick a name from the brand
  table and copy that name into cpu_brand.
- Replace the duplicated code to use the extended cpuid to replace
  cpu_model with the processor name in the AMD and Transmeta sections of
  printcpuinfo() with generic code that replaces cpu_model with
  cpu_brand if cpu_brand is not an empty string.  We also trim leading
  spaces from cpu_brand prior to doing this since at least some processor
  names (notably those of Intel CPUs) have leading spaces in the name.
- Give print_AMD_features() its own private regs[] array since
  printcpuinfo() doesn't use the one it has anymore.

21 years ago- Add a cpu_exthigh variable to hold the highest extended cpuid value
John Baldwin [Wed, 8 Jan 2003 16:35:59 +0000 (16:35 +0000)]
- Add a cpu_exthigh variable to hold the highest extended cpuid value
  returned from cpuid 0x80000000.
- Add a cpu_brand char array to hold the processor name returned by
  cpuid 0x80000002-0x80000004 on AMD, Intel, Transmeta, and possibly
  other CPUs.
- Use cpuid to set cpu_exthigh and read the processor name if it is present
  in identify_cpu().

21 years agoBah, get the test for more than one logical CPU right so we don't bogusly
John Baldwin [Wed, 8 Jan 2003 16:33:03 +0000 (16:33 +0000)]
Bah, get the test for more than one logical CPU right so we don't bogusly
claim a CPU has HT support when it lists 0 or 1 logical CPU's per physical
processor.

21 years agoAdd page queues locking to vunmapbuf().
Peter Grehan [Wed, 8 Jan 2003 12:29:59 +0000 (12:29 +0000)]
Add page queues locking to vunmapbuf().

Obtained from: sparc64
Approved by:  benno

21 years agoSync the i-cache after copying down the interrupt code
Peter Grehan [Wed, 8 Jan 2003 12:27:47 +0000 (12:27 +0000)]
Sync the i-cache after copying down the interrupt code

Approved by:  benno

21 years agoBe more conservative about re-enabling interrupts during trap processing
Peter Grehan [Wed, 8 Jan 2003 12:20:46 +0000 (12:20 +0000)]
Be more conservative about re-enabling interrupts during trap processing
until atomic issues are fully sorted.

Approved by:  benno

21 years agoFix incorrect error returns and sign-extension.
Peter Grehan [Wed, 8 Jan 2003 12:19:37 +0000 (12:19 +0000)]
Fix incorrect error returns and sign-extension.

Approved by:  benno

21 years agoFetch the initial time from the rtc OpenFirmware node. This is a short-term
Peter Grehan [Wed, 8 Jan 2003 12:17:44 +0000 (12:17 +0000)]
Fetch the initial time from the rtc OpenFirmware node. This is a short-term
measure until the rtc h/w driver is written, and it's a lot better
than having "jan 1 1970" on filesys times.

Approved by:  benno

21 years agoRemove obsolete NFS_ROOT conditional.
Peter Grehan [Wed, 8 Jan 2003 12:15:54 +0000 (12:15 +0000)]
Remove obsolete NFS_ROOT conditional.

Approved by:  benno

21 years agoImplement bus_dmamap_load_mbuf/bus_dmamap_load_uio.
Peter Grehan [Wed, 8 Jan 2003 11:46:53 +0000 (11:46 +0000)]
Implement bus_dmamap_load_mbuf/bus_dmamap_load_uio.
Tested load_mbuf with GEM ethernet driver.

Submitted by: Hiten Pandya <hiten@unixdaemons.com> (modified by grehan)

21 years ago- define HAS_STREAM_METHODS correctly
Peter Grehan [Wed, 8 Jan 2003 11:43:36 +0000 (11:43 +0000)]
- define HAS_STREAM_METHODS correctly
- dmamap_load_mbuf/load_uio prototypes

Submitted partly by:  Hiten Pandya <hiten@unixdaemons.com>

21 years agoo Refer to "Network Associates Laboratories" instead of "NAI Labs" or
Chris Costello [Wed, 8 Jan 2003 11:06:22 +0000 (11:06 +0000)]
o Refer to "Network Associates Laboratories" instead of "NAI Labs" or
  "Network Associates Labs" in the copyright notice.
o Remove clause #3 in the license terms.
o Remove the line break from my name.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoInclude a proper reference to the Developers' Handbook in the
Chris Costello [Wed, 8 Jan 2003 10:53:37 +0000 (10:53 +0000)]
Include a proper reference to the Developers' Handbook in the
SEE ALSO section.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoDo not strip CTL* escapes from redirection filenames in exptilde(),
Tim J. Robbins [Wed, 8 Jan 2003 10:50:08 +0000 (10:50 +0000)]
Do not strip CTL* escapes from redirection filenames in exptilde(),
expari(), expbackq() and evalvar(). Similar to revision 1.39.
Patch from Tor Egge.

PR: 45349
MFC after: 2 weeks

21 years agoCross-reference mac_lomac.4
Chris Costello [Wed, 8 Jan 2003 10:47:18 +0000 (10:47 +0000)]
Cross-reference mac_lomac.4

Sponsored by: DARPA, Network Associates Laboratories

21 years agoDocument the LOMAC security policy.
Chris Costello [Wed, 8 Jan 2003 10:30:00 +0000 (10:30 +0000)]
Document the LOMAC security policy.
This man page was co-written by Brian Feldman <green>.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoAdd code that works around the problem that the older Promise
Søren Schmidt [Wed, 8 Jan 2003 10:03:31 +0000 (10:03 +0000)]
Add code that works around the problem that the older Promise
controllers (ultra/fasttrak-66/100) fails on 48bit accesses.

21 years agoAdd a declaration that malloc() returns a pointer, not an int. Otherwise
Peter Wemm [Wed, 8 Jan 2003 07:02:07 +0000 (07:02 +0000)]
Add a declaration that malloc() returns a pointer, not an int.  Otherwise
this segfaults on ia64.  (Cannot cast an int to a pointer).
The rest of the code does it this way rather than use system includes....

21 years agoChange this:
Peter Wemm [Wed, 8 Jan 2003 06:43:27 +0000 (06:43 +0000)]
Change this:
#include <strings.h>
...
  foo = (char *)strdup(...);
To:
#include <string.h>
  foo = strdup(...);
because the former segfaults on an ia64 since there is no prototype
for strdup() in strings.h.  Converting an "int" to a pointer is fatal.

21 years agoAdd #include "diff.h" in order to get the prototype for xmalloc().
Peter Wemm [Wed, 8 Jan 2003 06:40:35 +0000 (06:40 +0000)]
Add #include "diff.h" in order to get the prototype for xmalloc().
The corresponding warning has been ignored for ages:
  prepend_args.c:75: warning: initialization makes pointer from integer
  without a cast
This is fatal on ia64.  You cannot convert an implicit int return from
an undeclared function to a pointer as you've lost the upper 32 bits.
On ia64, the warning is "pointer from different sized integer".

(This file is not vendor branched)

21 years agoDestroy the management device when detaching the driver.
Scott Long [Wed, 8 Jan 2003 06:30:34 +0000 (06:30 +0000)]
Destroy the management device when detaching the driver.

21 years ago- Add acpi module binding.
Matthew N. Dodd [Wed, 8 Jan 2003 06:24:25 +0000 (06:24 +0000)]
- Add acpi module binding.
- Restore speed and format settings on resume for CS423x and
  CS423x-PCI devices.

21 years agoAdd (somewhat overdue-ly) lmc(4) and ida(4) devices to the hardware
Bruce A. Mah [Wed, 8 Jan 2003 05:50:37 +0000 (05:50 +0000)]
Add (somewhat overdue-ly) lmc(4) and ida(4) devices to the hardware
list.

Also add an entry for lmc(4) to the release notes, since this is a
new driver for 5.0 and CURRENT.

21 years agoRetry to get full device descriptor, this let my slow CD Tower device work.
David Xu [Wed, 8 Jan 2003 05:03:35 +0000 (05:03 +0000)]
Retry to get full device descriptor, this let my slow CD Tower device work.

21 years agoSome KSE syscalls are MPSAFE.
David Xu [Wed, 8 Jan 2003 04:57:53 +0000 (04:57 +0000)]
Some KSE syscalls are MPSAFE.

21 years agoUse bus_space_subregion to add offsets to bus handles instead of doing
Jake Burkholder [Wed, 8 Jan 2003 04:30:30 +0000 (04:30 +0000)]
Use bus_space_subregion to add offsets to bus handles instead of doing
it manually.

21 years agoImplement bus_space_subregion.
Jake Burkholder [Wed, 8 Jan 2003 04:29:00 +0000 (04:29 +0000)]
Implement bus_space_subregion.

21 years agoEnumerate logical hyperthread CPUs manually if they aren't already listed
John Baldwin [Wed, 8 Jan 2003 01:33:18 +0000 (01:33 +0000)]
Enumerate logical hyperthread CPUs manually if they aren't already listed
in the mptable.  The way this works is that we determine if the system
has hyperthreading and how many logical CPU's should be in each physical
CPU by using the information returned by cpuid.  During the first pass of
the mptable, we build a bitmask of the APIC IDs of the CPUs listed in the
mptable.  We then scan that bitmask to see if the CPUs are already listed
by the mptable, or if there are any APIC IDs already in use that would
conflict with the APIC IDs of the logical CPUs.  If that test succeeds,
then we fixup the count of application processors.  Later on during the
second pass of the mptable we create fake processor entries for logical
CPUs and add them to the system.

We only need this type of fixup hack when using the mptable to enumerate
CPUs.  The ACPI MADT table properly enumerates all logical CPUs.

21 years agoIf the boot processor supports hyperthreading and contains more than one
John Baldwin [Wed, 8 Jan 2003 01:23:16 +0000 (01:23 +0000)]
If the boot processor supports hyperthreading and contains more than one
logical CPU, display the number of logical CPUs per physical processor
underneath the list of CPU features.

21 years agoAdd a cpuid_cpuinfo variable to hold the results of %ebx from cpuid with
John Baldwin [Wed, 8 Jan 2003 01:20:05 +0000 (01:20 +0000)]
Add a cpuid_cpuinfo variable to hold the results of %ebx from cpuid with
%eax of 1 and set it in identify_cpu().

21 years agoNo need to support UL_GETSIZE and UL_SETSIZE anymore; they are about to be
Tim J. Robbins [Wed, 8 Jan 2003 01:18:13 +0000 (01:18 +0000)]
No need to support UL_GETSIZE and UL_SETSIZE anymore; they are about to be
renamed in RELENG_5_0.

21 years ago- Fix the name of the hyperthreading cpuid feature flag to be HTT instead
John Baldwin [Wed, 8 Jan 2003 01:15:26 +0000 (01:15 +0000)]
- Fix the name of the hyperthreading cpuid feature flag to be HTT instead
  of HHT.
- Document fields returned in %ebx by a cpuid with %eax of 1.

21 years agoMove the MOD_SHUTDOWN event from shutdown_post_sync to shutdown_final,
Peter Wemm [Tue, 7 Jan 2003 22:24:13 +0000 (22:24 +0000)]
Move the MOD_SHUTDOWN event from shutdown_post_sync to shutdown_final,
so that entities that want to use the post_sync hook to write stuff
to devices and other tidy-up can do so before the device tree is
shot down.  eg: da doing a SYNC_CACHE etc.  This should get crashdumps
working on mpt devices again, and stops the ia64 boxes locking up
on regular shutdown when da tries to issue the scsi commands to mpt.

Obtained from:  njl, gibbs

21 years agopthread_attr_get_np() now takes 'pthread_t' (not 'pthread_t *')
Max Khon [Tue, 7 Jan 2003 21:43:30 +0000 (21:43 +0000)]
pthread_attr_get_np() now takes 'pthread_t' (not 'pthread_t *')
to be consistent with other pthread_XXX functions

21 years agoIn vn_open(), unset ndp->ni_vp when returning failure so that code
Brian Feldman [Tue, 7 Jan 2003 20:59:55 +0000 (20:59 +0000)]
In vn_open(), unset ndp->ni_vp when returning failure so that code
which expects it to be NULL unless the return value was 0 will work.

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

21 years agoUse copyout to access user memory.
Alfred Perlstein [Tue, 7 Jan 2003 20:10:04 +0000 (20:10 +0000)]
Use copyout to access user memory.

Submittted by: pho
MFC After: 2 days

21 years agoMake bogus_offset local to bufinit().
Alan Cox [Tue, 7 Jan 2003 19:55:08 +0000 (19:55 +0000)]
Make bogus_offset local to bufinit().

21 years agoThe README refers to a LICENSE file, so add that file too.
Prafulla Deuskar [Tue, 7 Jan 2003 19:36:00 +0000 (19:36 +0000)]
The README refers to a LICENSE file, so add that file too.
BTW the license is also embedded in the source files.

MFC after: 1 day

21 years agoFix warnings & errors caused by my last commit.
Poul-Henning Kamp [Tue, 7 Jan 2003 19:09:10 +0000 (19:09 +0000)]
Fix warnings & errors caused by my last commit.

21 years agoThis patch fixes a problem caused by applications that rapidly and
Kirk McKusick [Tue, 7 Jan 2003 18:23:50 +0000 (18:23 +0000)]
This patch fixes a problem caused by applications that rapidly and
repeatedly truncate the same file. Each time the file is truncated,
a buffer is grabbed to store the indirect block numbers that need
to be freed. Those blocks cannot be freed until the inode claiming
them is written to disk. Thus, the number of buffers being held by
soft updates explodes and in extreme cases can run the kernel out
of buffers. The problem can be avoided by doing an fsync on the
file every debug.maxindirdep truncates (currently defaulted to 50).
The fsync causes the inode to be written so that the held buffers
can be freed. The check for excessive buffers is checked as part
of the existing hook for excessive dependencies (softdep_slowdown)
in the truncate code.

Reported by: David Schultz <dschultz@uclink.Berkeley.EDU>
Sponsored by:   DARPA & NAI Labs.
MFC after: 3 weeks

21 years agoCast the integer read as the first argument for %b to an unsigned integer
John Baldwin [Tue, 7 Jan 2003 18:17:18 +0000 (18:17 +0000)]
Cast the integer read as the first argument for %b to an unsigned integer
so it's value is not sign extended when assigned to the uintmax_t variable
used internally by printf.  For example, if bit 31 is set in the cpuid
feature word, then %b would print out the initial value as a 16 character
hexadecimal value.  Now it only prints out an 8 character value.

Reviewed by: bde

21 years agoNew release note: SA-02:44.
Bruce A. Mah [Tue, 7 Jan 2003 17:35:04 +0000 (17:35 +0000)]
New release note:  SA-02:44.

21 years agoAdd support for OPIE and PAM.
David E. O'Brien [Tue, 7 Jan 2003 17:31:23 +0000 (17:31 +0000)]
Add support for OPIE and PAM.

Submitted by: mikeh (reworked by me)

21 years agoUse manpage entities for ciss(4), iir(4), and trm(4).
Bruce A. Mah [Tue, 7 Jan 2003 16:55:03 +0000 (16:55 +0000)]
Use manpage entities for ciss(4), iir(4), and trm(4).

21 years agoNew release notes: fwcontrol(8) (+MFC), swapoff(8), xargs(1) -P,
Bruce A. Mah [Tue, 7 Jan 2003 16:48:23 +0000 (16:48 +0000)]
New release notes:  fwcontrol(8) (+MFC), swapoff(8), xargs(1) -P,
bzip2recover (+MFC).

Modified release notes:  compat4x update, awk 20021213 update,
OpenPam Daffodil.

MFCs noted:  cvs-1.11.2.1 20021201 snapshot.

Other changes:  Capitalize "FireWire" correctly.  Use manpage entities for
ciss(4), ida(4), iir(4), trm(4).

21 years agoCorrect a couple small typos.
Chris Costello [Tue, 7 Jan 2003 13:18:21 +0000 (13:18 +0000)]
Correct a couple small typos.

Submitted by: Attila Nagy <bra@fsn.hu>

21 years ago- Fix print format for FreeBSD-4
Hidetoshi Shimokawa [Tue, 7 Jan 2003 12:58:08 +0000 (12:58 +0000)]
- Fix print format for FreeBSD-4
- Reduce debug message.

21 years agoAdd an entry for libugidfw.
Chris Costello [Tue, 7 Jan 2003 11:30:53 +0000 (11:30 +0000)]
Add an entry for libugidfw.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoDocument the file system firewall interface library functions.
Chris Costello [Tue, 7 Jan 2003 11:23:43 +0000 (11:23 +0000)]
Document the file system firewall interface library functions.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoTeach sysinstall about rpcbind, rpc.lockd, and rpc.statd. As an added
Scott Long [Tue, 7 Jan 2003 07:46:50 +0000 (07:46 +0000)]
Teach sysinstall about rpcbind, rpc.lockd, and rpc.statd.  As an added
bonus, rpcbind will be enabled automatically if rpc.lockd, rpc.statd, amd,
NFS Server, or NIS is enabled.

21 years agoClear some KSE fields after kse mode was turned off.
David Xu [Tue, 7 Jan 2003 06:56:43 +0000 (06:56 +0000)]
Clear some KSE fields after kse mode was turned off.

21 years ago#include "namespace.h" to get a prototype for _err().
Tim J. Robbins [Tue, 7 Jan 2003 06:55:58 +0000 (06:55 +0000)]
#include "namespace.h" to get a prototype for _err().

21 years agoRemove an unused variable: mbresult.
Tim J. Robbins [Tue, 7 Jan 2003 06:20:47 +0000 (06:20 +0000)]
Remove an unused variable: mbresult.

21 years ago#include <stdlib.h> for free()'s prototype.
Tim J. Robbins [Tue, 7 Jan 2003 06:17:13 +0000 (06:17 +0000)]
#include <stdlib.h> for free()'s prototype.

21 years agoUse hidden names (_close, _dup2, _waitpid, etc.) where appropriate.
Tim J. Robbins [Tue, 7 Jan 2003 06:07:56 +0000 (06:07 +0000)]
Use hidden names (_close, _dup2, _waitpid, etc.) where appropriate.

21 years agoAdd waitpid to the list of hidden names for use by wordexp.c and grantpt.c.
Tim J. Robbins [Tue, 7 Jan 2003 06:05:35 +0000 (06:05 +0000)]
Add waitpid to the list of hidden names for use by wordexp.c and grantpt.c.

21 years agoForgot to call setrunnable() for un-idled thread.
David Xu [Tue, 7 Jan 2003 06:04:33 +0000 (06:04 +0000)]
Forgot to call setrunnable() for un-idled thread.

21 years agoCheck signals for idled threads.
David Xu [Tue, 7 Jan 2003 05:56:38 +0000 (05:56 +0000)]
Check signals for idled threads.

21 years agoTypo fixes.
Giorgos Keramidas [Tue, 7 Jan 2003 05:06:27 +0000 (05:06 +0000)]
Typo fixes.

PR: docs/46815
Submitted by: SUZUKI Koichi <koich@cac.co.jp>

21 years agoWe've had something like this file since 1992, so therefore this file
Warner Losh [Tue, 7 Jan 2003 04:37:40 +0000 (04:37 +0000)]
We've had something like this file since 1992, so therefore this file
is a derivitive work, so the copyright notice in the file should say
1992-2003.

21 years agoDon't call fw_bus_probe() twice when successive self-id packets received.
Hidetoshi Shimokawa [Tue, 7 Jan 2003 04:26:45 +0000 (04:26 +0000)]
Don't call fw_bus_probe() twice when successive self-id packets received.
This change should stabilize SBP-II device probing.

21 years agoPut the ASIC revision into the device description string so it appears
John Polstra [Tue, 7 Jan 2003 01:33:03 +0000 (01:33 +0000)]
Put the ASIC revision into the device description string so it appears
in the dmesg output.

MFC after: 3 days

21 years agoo Only try to recopy the mbuf into an mbuf cluster if
Maxime Henrion [Tue, 7 Jan 2003 01:01:20 +0000 (01:01 +0000)]
o Only try to recopy the mbuf into an mbuf cluster if
  bus_dmamap_load_mbuf() returned EFBIG.
o Fix mbuf leaks in an error (rare) code path.
o Reuse the TX descriptor if xl_encap() failed instead of
  just picking the next one.
o Better error messages.

21 years agoDon't reference the maxlen and flags fields in the ring control blocks
John Polstra [Mon, 6 Jan 2003 23:46:47 +0000 (23:46 +0000)]
Don't reference the maxlen and flags fields in the ring control blocks
as separate 16-bit entities.  Some of the ring control blocks are
in NIC memory, so they must be referenced using 32-bit accesses.
Smaller accesses have been observed to fail under some conditions.
This caused the rings to be set up wrong, leading to writes by the
card outside of the intended bounds of the rings.  This problem was
diagnosed by Michael Barthelow.  Don Bowman submitted a patch which
fixed the problem using a slightly different approach.

Reference ring control blocks in NIC memory using a pointer to
volatile.

Parenthesize the BGE_HOSTADDR macro definition properly.

MFC after: 3 days

21 years agoRemove redundant documenation.
Tom Rhodes [Mon, 6 Jan 2003 23:20:55 +0000 (23:20 +0000)]
Remove redundant documenation.

PR: 46253
Submitted by: Jeroen Ruigrok van der Werven <asmodai@wxs.nl>

21 years agoCorrect order of alignment in EXAMPLES.
Tom Rhodes [Mon, 6 Jan 2003 22:52:57 +0000 (22:52 +0000)]
Correct order of alignment in EXAMPLES.

PR: 42760
Submitted by: Christian Brueffer <chris@unixpages.org>

21 years agoHappy new year 2003 !
Sebastien Gioria [Mon, 6 Jan 2003 22:17:27 +0000 (22:17 +0000)]
Happy new year 2003 !

21 years agoConvert to use bus_dmamap_load_mbuf(); decouple software TX descriptors
Thomas Moestl [Mon, 6 Jan 2003 22:12:57 +0000 (22:12 +0000)]
Convert to use bus_dmamap_load_mbuf(); decouple software TX descriptors
from the hardware descriptors to avoid the overhead of having a DMA
map for each of them. Bump the number of hardware descriptors to 128,
and use half as many software descriptors for now.
Some minor cleanups.

21 years agomore crypto test stuff:
Sam Leffler [Mon, 6 Jan 2003 22:11:56 +0000 (22:11 +0000)]
more crypto test stuff:

o cryptotest can now run multiple threads with -t option
o cryptotest can now "profile" time spent doing symmetric ops with -p
o cryptostats dumps the crypto statistics block
o cryptokeystat is an openbsd app that tests public key ops

21 years ago- Convert to use bus_dmamap_load_mbuf().
Thomas Moestl [Mon, 6 Jan 2003 22:09:01 +0000 (22:09 +0000)]
- Convert to use bus_dmamap_load_mbuf().
- remove DPRINTF(), there is a CTR*() for any of them, and KTR is
  far more useful to debug this driver.
- some cleanups; remove some unused code and definitions.

21 years agoChange the iommu code to be able to handle more than one DVMA area per
Thomas Moestl [Mon, 6 Jan 2003 21:59:54 +0000 (21:59 +0000)]
Change the iommu code to be able to handle more than one DVMA area per
map. Use this new feature to implement iommu_dvmamap_load_mbuf() and
iommu_dvmamap_load_uio() functions in terms of a new helper function,
iommu_dvmamap_load_buffer(). Reimplement the iommu_dvmamap_load()
to use it, too.
This requires some changes to the map format; in addition to that,
remove unused or redundant members.
Add SBus and Psycho wrappers for the new functions, and make them
available through the respective DMA tags.

21 years agoUpdate release documentation copyright dates.
Bruce A. Mah [Mon, 6 Jan 2003 21:59:16 +0000 (21:59 +0000)]
Update release documentation copyright dates.

21 years agoMajor update to the refuse file.
Tom Rhodes [Mon, 6 Jan 2003 21:56:10 +0000 (21:56 +0000)]
Major update to the refuse file.

PR: 44164
Submitted by: Sergei Kolobov <sergei@kolobov.com>

21 years agodon't reference a pkthdr after M_MOVE_PKTHDR has "remove it"; instead
Sam Leffler [Mon, 6 Jan 2003 21:33:54 +0000 (21:33 +0000)]
don't reference a pkthdr after M_MOVE_PKTHDR has "remove it"; instead
reference the pkthdr now in the destination of the move

Sponsored by: Vernier Networks

21 years agopurge extraneous clears of M_PKTHDR since M_MOVE_PKTHDR does this already
Sam Leffler [Mon, 6 Jan 2003 21:29:27 +0000 (21:29 +0000)]
purge extraneous clears of M_PKTHDR since M_MOVE_PKTHDR does this already

21 years agofix memory allocation problems and collateral damage:
Sam Leffler [Mon, 6 Jan 2003 21:23:06 +0000 (21:23 +0000)]
fix memory allocation problems and collateral damage:

o create a separate tag for each object allocated with bus_dmamem_alloc so
  the tag's maxsize is setup appropriately; this reduces memory allocation
  for the queue descriptors from 16M to what it should be and also fixes
  memory allocation for public key operands
o release bus dma resources on detach so module usage doesn't leak
o remove public key op disable now that bus dma memory allocation is fixed
o collect attach error handling in one place

Sponsored by: Vernier Networks

21 years ago- remove the unused parent DMA tag argument from
Thomas Moestl [Mon, 6 Jan 2003 20:54:07 +0000 (20:54 +0000)]
- remove the unused parent DMA tag argument from
  _nexus_dmamap_load_buffer()
- implement nexus_dmamap_load() in terms of _nexus_dmamap_load_buffer().
  Note that this is untested, as this code is not currently used (but
  might be later for UPA devices).
- move BUS_DMAMAP_NSEGS to bus_private.h
- disable the ecache flushing in nexus_dmamap_sync(); it should not be
  needed, although the docs are not entirely clear on that.

21 years agogeom_bsd need -lmd to link now.
Poul-Henning Kamp [Mon, 6 Jan 2003 20:11:32 +0000 (20:11 +0000)]
geom_bsd need -lmd to link now.

21 years agoBSD disklabels expose the controling label though the 'c' partition, and
Poul-Henning Kamp [Mon, 6 Jan 2003 20:10:41 +0000 (20:10 +0000)]
BSD disklabels expose the controling label though the 'c' partition, and
some trick is necessary to prevent further BSD geoms from attaching to
that.  Our old trick was to make sure we don't attach to a geom from
the "BSD" class, but this doesn't work if an intermediary geom obscures
this fact.  Instead, calculate the MD5 checksum of the label we target
and ask if anybody below us loves that label.  If they do we don't.

Coded by:       gordon.

21 years agoIn userland case include <errno.h>, not <err.h>. This is needed to make
Poul-Henning Kamp [Mon, 6 Jan 2003 20:05:45 +0000 (20:05 +0000)]
In userland case include <errno.h>, not <err.h>.  This is needed to make
the src/tools/regression/geom stuff compile.

21 years agoBump the IOMMU TSB size to 32kB, to match the default size on PCI
Thomas Moestl [Mon, 6 Jan 2003 19:48:31 +0000 (19:48 +0000)]
Bump the IOMMU TSB size to 32kB, to match the default size on PCI
systems.

21 years agoPrefix the members of struct bus_space_tag and struct bus_dma_tag with
Thomas Moestl [Mon, 6 Jan 2003 19:43:10 +0000 (19:43 +0000)]
Prefix the members of struct bus_space_tag and struct bus_dma_tag with
a uniqifier. No functional changes.