]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoUpdate for the 'file' 4.19 import.
David E. O'Brien [Thu, 24 May 2007 16:14:38 +0000 (16:14 +0000)]
Update for the 'file' 4.19 import.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r169942,
David E. O'Brien [Thu, 24 May 2007 16:12:23 +0000 (16:12 +0000)]
This commit was generated by cvs2svn to compensate for changes in r169942,
which included commits to RCS files with non-trunk default branches.

17 years agoVirgin import of Christos Zoulas's FILE 4.19.
David E. O'Brien [Thu, 24 May 2007 16:12:23 +0000 (16:12 +0000)]
Virgin import of Christos Zoulas's FILE 4.19.

17 years agohesiod_resolv -> hesiod_resolve
Ruslan Ermilov [Thu, 24 May 2007 13:50:06 +0000 (13:50 +0000)]
hesiod_resolv -> hesiod_resolve

Submitted by: Danny Braniss

17 years agoDNLC_NO_VNODE can't be NULL.
Pawel Jakub Dawidek [Thu, 24 May 2007 13:44:45 +0000 (13:44 +0000)]
DNLC_NO_VNODE can't be NULL.

Reported by: ru

17 years agoFix a formatting error: the list of control packets ends earlier.
Yaroslav Tykhiy [Thu, 24 May 2007 09:43:41 +0000 (09:43 +0000)]
Fix a formatting error: the list of control packets ends earlier.

17 years agoFix the online usage for the "dump" command.
Ruslan Ermilov [Thu, 24 May 2007 09:21:20 +0000 (09:21 +0000)]
Fix the online usage for the "dump" command.

17 years agoInitialize ZFS a bit earlier and block root mounting until
Pawel Jakub Dawidek [Thu, 24 May 2007 07:43:00 +0000 (07:43 +0000)]
Initialize ZFS a bit earlier and block root mounting until
initialization is complete. This fixes some root-on-ZFS
configurations.

Reported by: Bruno Damour <freebsd.ruomad@free.fr>
Tested by: Bruno Damour <freebsd.ruomad@free.fr>

17 years agoAdd sparc64 regression test files.
Diomidis Spinellis [Thu, 24 May 2007 07:00:53 +0000 (07:00 +0000)]
Add sparc64 regression test files.

Submitted by: Carl Johan Gustavsson

17 years agoModify file generation instructions to remove the tty field:
Diomidis Spinellis [Thu, 24 May 2007 06:57:50 +0000 (06:57 +0000)]
Modify file generation instructions to remove the tty field:
the mapping from number to name differs between systems,
and will cause tests to fail.

17 years agoCleanup pax(1) sources a little bit while I poked around in them:
Ralf S. Engelschall [Thu, 24 May 2007 06:44:37 +0000 (06:44 +0000)]
Cleanup pax(1) sources a little bit while I poked around in them:
- remove a superfluous doubled trailing semicolon.
- remove the extra void casts on calls to void-function free(3).

17 years agoFix indentation.
Ralf S. Engelschall [Thu, 24 May 2007 06:01:06 +0000 (06:01 +0000)]
Fix indentation.

17 years agoRemove two superfluous trailing semicolons.
Ralf S. Engelschall [Thu, 24 May 2007 05:58:20 +0000 (05:58 +0000)]
Remove two superfluous trailing semicolons.

17 years agoRemove two unnecessary and useless sub-shell constructs.
Ralf S. Engelschall [Thu, 24 May 2007 05:54:37 +0000 (05:54 +0000)]
Remove two unnecessary and useless sub-shell constructs.

17 years agoUpdate the maintainer id for em driver.
Jack F Vogel [Wed, 23 May 2007 21:47:19 +0000 (21:47 +0000)]
Update the maintainer id for em driver.

17 years agoFix the procedure for generating the test summary files.
Diomidis Spinellis [Wed, 23 May 2007 21:35:45 +0000 (21:35 +0000)]
Fix the procedure for generating the test summary files.
Document the procedure for obtaining an appropriately old version of sa.

17 years agoFreeBSD's namecache works quite well with ZFS, so remove DNLC.
Pawel Jakub Dawidek [Wed, 23 May 2007 21:33:02 +0000 (21:33 +0000)]
FreeBSD's namecache works quite well with ZFS, so remove DNLC.

17 years agoAll objects we create using GFS are directories, so initialize d_type
Pawel Jakub Dawidek [Wed, 23 May 2007 21:27:47 +0000 (21:27 +0000)]
All objects we create using GFS are directories, so initialize d_type
properly, but add XXX comment saying that it can eventually change in
the future.

17 years agoTwo minor fixes, keep old 82542 from using jumbo frames, and add
Jack F Vogel [Wed, 23 May 2007 20:41:20 +0000 (20:41 +0000)]
Two minor fixes, keep old 82542 from using jumbo frames, and add
missing htole64 in encap code.

Reviewed by:Pdeuskar
Approved by:Pdeuskar

17 years agoAdd CWR back into the PRINT_TH_FLAGS list as gcc42 doesn't complain
Andre Oppermann [Wed, 23 May 2007 19:16:21 +0000 (19:16 +0000)]
Add CWR back into the PRINT_TH_FLAGS list as gcc42 doesn't complain
about \8 in a string anymore.

17 years agoIn tcp_log_addrs():
Andre Oppermann [Wed, 23 May 2007 19:07:53 +0000 (19:07 +0000)]
In tcp_log_addrs():
 o add the hex output of the th_flags field to the example log
   line in comments
 o simplify the log line length calculation and make it less
   evil
 o correct the test for the length panic; the line isn't on
   the stack but malloc'ed

17 years agoDon't set lo_name and clobber lo_flags in lock_profile_object_init().
John Baldwin [Wed, 23 May 2007 18:46:54 +0000 (18:46 +0000)]
Don't set lo_name and clobber lo_flags in lock_profile_object_init().
This was just wasteful when this was always called before lock_init()
(which overwrote both fields each time), but when
lock_profile_object_init() was moved into lock_init() the clearing of
lo_flags proved fatal (all locks became spin locks to _sleep(), etc.)

Reported by: kris

17 years agofix comment typo
Sam Leffler [Wed, 23 May 2007 17:28:21 +0000 (17:28 +0000)]
fix comment typo

17 years agoComment that tdsignal() may be entered from the debugger.
Robert Watson [Wed, 23 May 2007 17:27:42 +0000 (17:27 +0000)]
Comment that tdsignal() may be entered from the debugger.

17 years agoInitialize time_lock before calling cpu_initclocks(). This corrects a
Robert Watson [Wed, 23 May 2007 17:27:01 +0000 (17:27 +0000)]
Initialize time_lock before calling cpu_initclocks().  This corrects a
race condition in which hardclock fires before the mutex is initialized
leading to a "corrupt spinlock" panic.

Submitted by: attilio

17 years agoAdd support for "hardware" vlan tag insertion & removal emulation
Andrew Gallatin [Wed, 23 May 2007 16:25:40 +0000 (16:25 +0000)]
Add support for "hardware" vlan tag insertion & removal emulation
in the mxge driver so as to be able to do checksum offload
on vlans.  This is good enough to achieve 10GbE line rate on vlans.

17 years agoFix the dependency for the linux_support.s, explicitely add linux_assym.h.
Konstantin Belousov [Wed, 23 May 2007 15:45:52 +0000 (15:45 +0000)]
Fix the dependency for the linux_support.s, explicitely add linux_assym.h.

Reported by: rwatson
In collaboration with: rdivacky
Sponsored by: Google SoC 2007

17 years agoFix a typo in pcib_alloc_msi{x} which resulted in the
Andrew Gallatin [Wed, 23 May 2007 15:31:00 +0000 (15:31 +0000)]
Fix a typo in pcib_alloc_msi{x} which resulted in the
device's, not the bridge's, softc to be used to check the
PCIB_DISABLE_MSI flag.  This resulted in randomly allowing
or denying MSI interrupts based on whatever value the driver
happened to store at sizeof(device_t) bytes into its softc.

I noticed this when I stopped getting MSI interrupts
after slighly re-arranging mxge's softc yesterday.

17 years agoRemove duplicate includes.
Olivier Houchard [Wed, 23 May 2007 13:36:02 +0000 (13:36 +0000)]
Remove duplicate includes.

Submitted by:   Cyril Nguyen Huu <cyril ci0 org>

17 years agoRemove duplicate includes.
Olivier Houchard [Wed, 23 May 2007 13:21:57 +0000 (13:21 +0000)]
Remove duplicate includes.

Submitted by:   Cyril Nguyen Huu <cyril ci0 org>

17 years agoNit: avoid shadowing truncate(2) with a local variable.
Dag-Erling Smørgrav [Wed, 23 May 2007 12:09:33 +0000 (12:09 +0000)]
Nit: avoid shadowing truncate(2) with a local variable.

17 years agoEliminate VI_LOCK()/VI_UNLOCK() pair from getattr and close code paths.
Pawel Jakub Dawidek [Wed, 23 May 2007 11:06:09 +0000 (11:06 +0000)]
Eliminate VI_LOCK()/VI_UNLOCK() pair from getattr and close code paths.
It's hard to measure performance improvement on my test machine, but the
change won't degrade performance for sure. I can measure slight improvement
for debugging kernel and it can also be a win for machines where atomic
operation is more expensive.

Reviewed by: kib

17 years agoFix stupid braino in previous commit.
Dag-Erling Smørgrav [Wed, 23 May 2007 10:06:03 +0000 (10:06 +0000)]
Fix stupid braino in previous commit.

17 years agoNo need to force __inline__ of currecord(), as the compiler will usefully
Robert Watson [Wed, 23 May 2007 09:32:30 +0000 (09:32 +0000)]
No need to force __inline__ of currecord(), as the compiler will usefully
inline it when needed already, and the symbol is also required outside of
audit.c.  This silences a new gcc warning on the topic of using __inline__
instead of __inline.

MFC after: 3 days

17 years agoMove futex support code from <arch>/support.s into linux compat directory.
Konstantin Belousov [Wed, 23 May 2007 08:33:06 +0000 (08:33 +0000)]
Move futex support code from <arch>/support.s into linux compat directory.
Implement all futex atomic operations in assembler to not depend on the
fuword() that does not allow to distinguish between -1 and failure return.
Correctly return 0 from atomic operations on success.

In collaboration with: rdivacky
Tested by: Scot Hetzel <swhetzel gmail com>, Milos Vyletel <mvyletel mzm cz>
Sponsored by: Google SoC 2007

17 years agoIf (flags & O_TRUNC), don't truncate the file until we've successfully
Dag-Erling Smørgrav [Wed, 23 May 2007 08:12:34 +0000 (08:12 +0000)]
If (flags & O_TRUNC), don't truncate the file until we've successfully
locked it.

MFC after: 3 weeks

17 years agoAdd some early diagnostics under bootverbose
Kip Macy [Wed, 23 May 2007 05:22:58 +0000 (05:22 +0000)]
Add some early diagnostics under bootverbose
bootverbose is not getting set early enough so hardcode for the moment

17 years agoDo not attempt to load the kernel module when checking if an interface exists.
Andrew Thompson [Wed, 23 May 2007 00:18:44 +0000 (00:18 +0000)]
Do not attempt to load the kernel module when checking if an interface exists.

This would cause pseudo network modules to be reloaded again when trying to
unload the first time if any cloned interfaces exist.

MFC after: 2 weeks

17 years ago- Remove redundant initialization.
Pawel Jakub Dawidek [Tue, 22 May 2007 23:05:48 +0000 (23:05 +0000)]
- Remove redundant initialization.
- Compare pointer with NULL.

17 years agoLock vnode on lookup. This fixes ZIL replay for rmdir/unlink/rename.
Pawel Jakub Dawidek [Tue, 22 May 2007 21:22:25 +0000 (21:22 +0000)]
Lock vnode on lookup. This fixes ZIL replay for rmdir/unlink/rename.

Reported by: des

17 years agoFix runtime crash in OpenSSL with "Illegal instruction" by making some
Simon L. B. Nielsen [Tue, 22 May 2007 20:28:19 +0000 (20:28 +0000)]
Fix runtime crash in OpenSSL with "Illegal instruction" by making some
casts a bit less evil.

This was e.g. seen when using portsnap as:

  Fetching snapshot tag from portsnap3.FreeBSD.org... Illegal instruction

Note the patch is slightly different from kan's original patch to
match style in the OpenSSL source files a bit better.

Submitted by: kan
Tested by: many

17 years agoRemove an empty line.
Jung-uk Kim [Tue, 22 May 2007 20:24:07 +0000 (20:24 +0000)]
Remove an empty line.

17 years agoRemove BCM5704S specific tunable (hw.bge.fake_autoneg) and
Jung-uk Kim [Tue, 22 May 2007 19:35:34 +0000 (19:35 +0000)]
Remove BCM5704S specific tunable (hw.bge.fake_autoneg) and
auto-detect the condition.

17 years agoRearrange DMA read/write control register settings based on document snippet
Jung-uk Kim [Tue, 22 May 2007 19:22:58 +0000 (19:22 +0000)]
Rearrange DMA read/write control register settings based on document snippet
provided by davidch via glebius.

PR: kern/96806

17 years agoRemove stale alpha support code.
Jung-uk Kim [Tue, 22 May 2007 19:11:39 +0000 (19:11 +0000)]
Remove stale alpha support code.

17 years agoFix time out check in EEPROM read and move delays to give some settle time.
Jung-uk Kim [Tue, 22 May 2007 18:51:05 +0000 (18:51 +0000)]
Fix time out check in EEPROM read and move delays to give some settle time.

Submitted by: mjacob

17 years agoAdd delays in MI communication register R/W loops.
Jung-uk Kim [Tue, 22 May 2007 18:16:23 +0000 (18:16 +0000)]
Add delays in MI communication register R/W loops.

17 years agoFix typos.
Christian Brueffer [Tue, 22 May 2007 18:14:35 +0000 (18:14 +0000)]
Fix typos.

17 years agoAdd the -n flag for disabling automatic module loading, this will be used by
Andrew Thompson [Tue, 22 May 2007 17:41:09 +0000 (17:41 +0000)]
Add the -n flag for disabling automatic module loading, this will be used by
rc.d to stop it reloading the network module on unload.

17 years agoSome minor cleanups:
Gleb Smirnoff [Tue, 22 May 2007 16:17:31 +0000 (16:17 +0000)]
Some minor cleanups:
- In rt_check() remove the senderr() macro and the "bad" label. They
  used to simplify code, but now aren't.
- Remove extra RT_LOCK_ASSERT() in rt_setgate(). The RT_REMREF macro
  does this.
- In rtfree() convert panics to KASSERTs.
- Strict the routing API: rtfree() should be called only in a case
  when we are completely sure we've got the last reference on the
  rtentry. In all other cases RTFREE_LOCKED() macro should be used.
  If the reference isn't the last one spit out a warning printf.
  Correct the only(?) case for this in rt_check().
- Fix typos in comments.

17 years agomxge cleanups:
Andrew Gallatin [Tue, 22 May 2007 15:57:49 +0000 (15:57 +0000)]
mxge cleanups:

- Remove code to use the special wc_fifo.  It has been disabled by default
  in our other drivers as it actually slows down transmit by a small amount

- Dynamically determine the amount of space required for the rx_done
  ring rather than hardcoding it.

- Compute the number of tx descriptors we are willing to transmit per
  frame as the minimum of 128 or 1/4 the tx ring size.

- Fix a typo in the tx dma tag setup which could lead to unnecessary
  defragging of TSO packets (and potentially even dropping TSO packets
  due to EFBIG being returned).

- Add a counter to keep track of how many times we've needed to
  defragment a frame.  It should always be zero.

- Export new extended f/w counters via sysctl

Sponsored by: Myricom, Inc.

17 years agoReplace "sizeof(struct bucket[cache_size])" with
Colin Percival [Tue, 22 May 2007 15:37:58 +0000 (15:37 +0000)]
Replace "sizeof(struct bucket[cache_size])" with
"cache_size * sizeof(struct bucket)".  The former is valid in C99 but can
confuse earlier compilers, while the latter is a standard idiom which all
C compilers understand.

Approved by: kientzle

17 years agoHook mxge_lro.c to the build so that the if_mxge.ko module
Andrew Gallatin [Tue, 22 May 2007 14:38:17 +0000 (14:38 +0000)]
Hook mxge_lro.c to the build so that the if_mxge.ko module
will load.

17 years agobackout filter of Nil UUID as the boot loader code already filters out Nil UUIDs...
Ralf S. Engelschall [Tue, 22 May 2007 13:53:59 +0000 (13:53 +0000)]
backout filter of Nil UUID as the boot loader code already filters out Nil UUIDs (see src/sys/boot/i386/libi386/smbios.c:smbios_setuuid for details)

17 years agoAdd support for setmode and settarget messages.
Alexander Motin [Tue, 22 May 2007 12:23:39 +0000 (12:23 +0000)]
Add support for setmode and settarget messages.

Approved by: glebius (mentor)

17 years agoAllow node to bypass traffic while no alias address defined.
Alexander Motin [Tue, 22 May 2007 12:14:43 +0000 (12:14 +0000)]
Allow node to bypass traffic while no alias address defined.

Approved by: glebius (mentor)

17 years agoReturn back NETGRAPH_MPPC_COMPRESSION depends.
Alexander Motin [Tue, 22 May 2007 12:00:31 +0000 (12:00 +0000)]
Return back NETGRAPH_MPPC_COMPRESSION depends.

Reviewed by: ru, julian
Approved by: glebius (mentor)

17 years agoRemove the restore and chroot command to avoid disk-full error.
Yoshihiro Takahashi [Tue, 22 May 2007 11:32:55 +0000 (11:32 +0000)]
Remove the restore and chroot command to avoid disk-full error.

17 years agoSet .PATH before referring to the corresponding sources.
Diomidis Spinellis [Tue, 22 May 2007 10:49:42 +0000 (10:49 +0000)]
Set .PATH before referring to the corresponding sources.

17 years agoDon't forget to clean generated ssp.h.
Ruslan Ermilov [Tue, 22 May 2007 10:40:58 +0000 (10:40 +0000)]
Don't forget to clean generated ssp.h.

17 years agoDon't want a stray systrace_args.c in objdir.
Ruslan Ermilov [Tue, 22 May 2007 10:36:12 +0000 (10:36 +0000)]
Don't want a stray systrace_args.c in objdir.

17 years agoRemove the ugly csh(1) based UUID lower-case translation hack from
Ralf S. Engelschall [Tue, 22 May 2007 10:22:24 +0000 (10:22 +0000)]
Remove the ugly csh(1) based UUID lower-case translation hack from
/etc/rc.d/hostid now that we switched the origin of the UUID (variable
smbios.system.uuid as provided by the i386 BIOS code) to already provide
a standard conforming lower-case UUID text representation.

17 years agoIncrease precision of time values in the process accounting
Diomidis Spinellis [Tue, 22 May 2007 06:51:38 +0000 (06:51 +0000)]
Increase precision of time values in the process accounting
structure, while maintaining backward compatibility with legacy
file and record formats.

17 years agoRegression tests for sparc64 record formats.
Diomidis Spinellis [Tue, 22 May 2007 06:34:20 +0000 (06:34 +0000)]
Regression tests for sparc64 record formats.

Submitted by: Carl Johan Gustavsson

17 years agoRegression tests for amd64 file and record formats.
Diomidis Spinellis [Tue, 22 May 2007 06:30:48 +0000 (06:30 +0000)]
Regression tests for amd64 file and record formats.

Submitted by: Larry Rosenman (who also provided access to a testing host)

17 years agoRegression tests for the new and legacy process accounting formats.
Diomidis Spinellis [Tue, 22 May 2007 06:21:20 +0000 (06:21 +0000)]
Regression tests for the new and legacy process accounting formats.

17 years agorestore interrupts to working order after INTR_THREAD changes
Kip Macy [Tue, 22 May 2007 06:17:55 +0000 (06:17 +0000)]
restore interrupts to working order after INTR_THREAD changes

- ithread_wrapper was being treated as a wrapper for fast interrupts when
  in fact it was intended for ithread interrupts

17 years agoDocument the .in .out file naming conventions adopted in commit_prep.pl's
Diomidis Spinellis [Tue, 22 May 2007 06:11:44 +0000 (06:11 +0000)]
Document the .in .out file naming conventions adopted in commit_prep.pl's
exclusions list for header checking.

17 years agoTest the integer-arithmetic float conversions encode_timeval and encode_long
Diomidis Spinellis [Tue, 22 May 2007 05:52:04 +0000 (05:52 +0000)]
Test the integer-arithmetic float conversions encode_timeval and encode_long
in kern_acct.c.

17 years agoBackout 1.5 as requested by deischen
Jonathan Chen [Tue, 22 May 2007 05:28:40 +0000 (05:28 +0000)]
Backout 1.5 as requested by deischen

17 years agoEliminate the reactivation of cached pages in vm_fault_prefault() and
Alan Cox [Tue, 22 May 2007 04:45:59 +0000 (04:45 +0000)]
Eliminate the reactivation of cached pages in vm_fault_prefault() and
vm_map_pmap_enter() unless the caller is madvise(MADV_WILLNEED).  With
the exception of calls to vm_map_pmap_enter() from
madvise(MADV_WILLNEED), vm_fault_prefault() and vm_map_pmap_enter()
are both used to create speculative mappings.  Thus, always
reactivating cached pages is a mistake.  In principle, cached pages
should only be reactivated by an actual access.  Otherwise, the
following misbehavior can occur.  On a hard fault for a text page the
clustering algorithm fetches not only the required page but also
several of the adjacent pages.  Now, suppose that one or more of the
adjacent pages are never accessed.  Ultimately, these unused pages
become cached pages through the efforts of the page daemon.  However,
the next activation of the executable reactivates and maps these
unused pages.  Consequently, they are never replaced.  In effect, they
become pinned in memory.

17 years agoIn the error handling path, don't call close(fd) if the error we're
Colin Percival [Tue, 22 May 2007 04:21:00 +0000 (04:21 +0000)]
In the error handling path, don't call close(fd) if the error we're
handling is that fd = open(foo) is -1.  This bug is harmless since
close(-1) just returns an error (which the code ignores).

Found by: Coverity Prevent(tm)
CID: 1503 (in userland test run)

17 years ago__cleanup() is needed for ports/devel/valgrind, export it.
Jonathan Chen [Tue, 22 May 2007 03:03:28 +0000 (03:03 +0000)]
__cleanup() is needed for ports/devel/valgrind, export it.

17 years agoAllow FreeBSD's native ELF image activators to execute shared libraries the
Alexander Kabaev [Tue, 22 May 2007 02:22:58 +0000 (02:22 +0000)]
Allow FreeBSD's native ELF image activators to execute shared libraries the
same way it was enabled for Linux binares in linuxulator.

This allows binaries built with -pie. Many ports auto-detect -fPIE support
in GCC 4.2 and build binaries FreeBSD was unable to run.

17 years agoRemove duplicate ' 2' manpage section number.
Giorgos Keramidas [Mon, 21 May 2007 20:33:51 +0000 (20:33 +0000)]
Remove duplicate ' 2' manpage section number.

Submitted by:  Rui Paulo
MFC after:  3 days

17 years agoFrom OpenBSD, rev. 1.379
Daniel Hartmeier [Mon, 21 May 2007 20:12:35 +0000 (20:12 +0000)]
From OpenBSD, rev. 1.379
Document how 'allow-opts' applies to routing headers in IPv6.

MFC after: 1 week
Discussed with: mlaier

17 years agoFrom OpenBSD, rev. 1.524, 1.528, 1.529
Daniel Hartmeier [Mon, 21 May 2007 20:08:59 +0000 (20:08 +0000)]
From OpenBSD, rev. 1.524, 1.528, 1.529
Deal with IPv6 routing headers (see FreeBSD-SA-07:03.ipv6 for background)
Block IPv6 packets with routing headers by default, unless 'allow-opts'
is specified. Block RH0 unconditionally. Deal with ip6_plen 0.

MFC after: 1 week
Discussed with: mlaier

17 years agoUse lower cases for UUID string to conform RFC4122 and ISO/IEC-9834-8:2005.
Jung-uk Kim [Mon, 21 May 2007 18:48:18 +0000 (18:48 +0000)]
Use lower cases for UUID string to conform RFC4122 and ISO/IEC-9834-8:2005.

17 years agoForgot to add hptiop(4) here.
Christian Brueffer [Mon, 21 May 2007 18:35:40 +0000 (18:35 +0000)]
Forgot to add hptiop(4) here.

17 years agoImprove mxge receive performance:
Andrew Gallatin [Mon, 21 May 2007 18:32:27 +0000 (18:32 +0000)]
Improve mxge receive performance:

- Update to the latest (1.4.18) f/w.  This f/w introduces a new
  receive mode which allows us to use FreeBSD's physically discontinuous
  MJUM9BYTES clusters.

- Switch the driver from chaining MJUMPAGESIZE clusters to using
  MJUM9BYTES clusters to avoid mbuf chaining overheads.  Due to this
  change, people running obsolete f/w images will be limited to an MTU of
  PAGE_SIZE - 16.

- Add (disabled by default) support for Large Receive Offload.

Sponsored by: Myricom, Inc.

17 years agoMake pointer argument to kread_string() const since the kernel structure
Robert Watson [Mon, 21 May 2007 18:16:04 +0000 (18:16 +0000)]
Make pointer argument to kread_string() const since the kernel structure
field is const, and then employ __DECONST before getting into the kvm
code.  This eliminates a gcc 4.2 warning about losing constification.

__DECONST advice from: sam

17 years agoFix markup.
Ruslan Ermilov [Mon, 21 May 2007 15:29:42 +0000 (15:29 +0000)]
Fix markup.

17 years agoForce the alignment of the chars arrays, as they are casted later to
Olivier Houchard [Mon, 21 May 2007 14:38:45 +0000 (14:38 +0000)]
Force the alignment of the chars arrays, as they are casted later to
structs.
gcc 4.2 doesn't do it by default, and that results in unaligned access on
arm.

17 years agoForce the alignment of the chars arrays, as they are casted later to
Olivier Houchard [Mon, 21 May 2007 14:38:20 +0000 (14:38 +0000)]
Force the alignment of the chars arrays, as they are casted later to
structs.
gcc 4.2 doesn't do it by default, and that results in unaligned access on
arm.

Reviewed by: gnn, imp

17 years agoShort name of kernel audit worker thread from "audit_worker" to "audit".
Robert Watson [Mon, 21 May 2007 14:08:26 +0000 (14:08 +0000)]
Short name of kernel audit worker thread from "audit_worker" to "audit".

MFC after: 3 days

17 years agoBump __FreeBSDversion for shared library version bump.
Xin LI [Mon, 21 May 2007 14:03:28 +0000 (14:03 +0000)]
Bump __FreeBSDversion for shared library version bump.
This makes it possible for our ports infrastructure
to determine whether it is appropriate to install
compat6x in the future.

17 years agoMake sure fwsid is not NULL.
Hidetoshi Shimokawa [Mon, 21 May 2007 12:17:54 +0000 (12:17 +0000)]
Make sure fwsid is not NULL.

MFC after: 3 days

17 years agoBack out the addition of -static to LDFLAGS; I'm pretty confident
Ruslan Ermilov [Mon, 21 May 2007 12:04:16 +0000 (12:04 +0000)]
Back out the addition of -static to LDFLAGS; I'm pretty confident
that the build failure was caused by a computer/sources date/time
mismatch that caused GCC tools to be mistakenly rebuilt again at
an inappropriate time during buildworld, re-linking them against
new libraries instead of host's installed libraries and thus making
them not runnable by the host.  Normally they are only built in
the early stage of buildworld (build-tools) that links them against
shared libraries of the host, but if either the system clock or
modification date/time on source files is set incorrectly, make(1)
can be foolished into thinking that tools are stale and will rebuild
them again, now in the "target" environment which is not suitable
for building helper apps that are to be run during buildworld.

OK'ed by: kan

17 years agoMFi386: revision 1.652
Yoshihiro Takahashi [Mon, 21 May 2007 11:57:36 +0000 (11:57 +0000)]
MFi386: revision 1.652

   - Move GDT/LDT locking into a seperate spinlock, removing the global
     scheduler lock from this responsibility.

17 years agoCleanup style by consistently using braces around variable expansion and
Ralf S. Engelschall [Mon, 21 May 2007 11:57:01 +0000 (11:57 +0000)]
Cleanup style by consistently using braces around variable expansion and
apply an addition from Andrew Thompson <thompsa> for filtering out the
special "Nil" UUID (all zeros) which would be a useless host UUID.

17 years agoAdjust UUID lower-case translation from straight-forward tr(1)
Ralf S. Engelschall [Mon, 21 May 2007 11:44:13 +0000 (11:44 +0000)]
Adjust UUID lower-case translation from straight-forward tr(1)
usage to an equivalent csh(1) usage as tr(1) stays in /usr/bin and
/etc/rc.d/hostid has just the root filesystem (and this way mainly the
tools in /bin) available.

I've chosen csh(1) here as the string manipulation tools available in
/bin is extremely limited and the (only) alternative ed(1) usage would
have been a lot more complicated or even might require a temporary file.

17 years agoPerhaps a better fix- make the called function do the (possible never reached)
Matt Jacob [Mon, 21 May 2007 09:26:47 +0000 (09:26 +0000)]
Perhaps a better fix- make the called function do the (possible never reached)
initialization. Do a token amount of style cleanup.

Poked by: bde

17 years agoStyle: remove redundant parentheses.
Ruslan Ermilov [Mon, 21 May 2007 09:01:23 +0000 (09:01 +0000)]
Style: remove redundant parentheses.

17 years agoRemove the libmytinfo_p.a link as well.
Ruslan Ermilov [Mon, 21 May 2007 08:53:26 +0000 (08:53 +0000)]
Remove the libmytinfo_p.a link as well.

17 years agoGCC doesn't segfault anymore while building world with -mcpu=xscale, so use it.
Olivier Houchard [Mon, 21 May 2007 08:39:44 +0000 (08:39 +0000)]
GCC doesn't segfault anymore while building world with -mcpu=xscale, so use it.

17 years agoFreeBSD expects VFP, so use that as the default FPU.
Olivier Houchard [Mon, 21 May 2007 08:38:42 +0000 (08:38 +0000)]
FreeBSD expects VFP, so use that as the default FPU.

Reviewed by: kan

17 years agoThe standardized textual representation of UUIDs according to RFC 4122
Ralf S. Engelschall [Mon, 21 May 2007 08:22:43 +0000 (08:22 +0000)]
The standardized textual representation of UUIDs according to RFC 4122
and ISO/IEC-9834-8:2005 is with LOWER-CASE hexadecimal characters only,
so translate the (usually upper-case and this way not conforming)
representation of the BIOS UUID when reading it. Also be more strict
about the valid characters in the textual representation by checking for
just the hexadecimal characters.

17 years agoDon't remove boot_i386.8 on amd64.
Xin LI [Mon, 21 May 2007 07:14:01 +0000 (07:14 +0000)]
Don't remove boot_i386.8 on amd64.

17 years agoReflect shared library bump after symbol versioning is enabled by
Xin LI [Mon, 21 May 2007 07:00:29 +0000 (07:00 +0000)]
Reflect shared library bump after symbol versioning is enabled by
default.

17 years agoWhen speculation fails (as determined by the chk instruction) the
Marcel Moolenaar [Mon, 21 May 2007 05:11:43 +0000 (05:11 +0000)]
When speculation fails (as determined by the chk instruction) the
processor is to jump to recovery code. This branching behaviour
may not be implemented by the processor and a Speculative Operation
fault is raised. The OS is responsible to emulate the branch.
Implement this, because GCC 4.2 uses advanced loads regularly.