]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoAdd a new 'why' argument to kdb_enter(), and a set of constants to use
rwatson [Tue, 25 Dec 2007 17:52:02 +0000 (17:52 +0000)]
Add a new 'why' argument to kdb_enter(), and a set of constants to use
for that argument.  This will allow DDB to detect the broad category of
reason why the debugger has been entered, which it can use for the
purposes of deciding which DDB script to run.

Assign approximate why values to all current consumers of the
kdb_enter() interface.

16 years agoTypos corrected.
schweikh [Tue, 25 Dec 2007 16:36:52 +0000 (16:36 +0000)]
Typos corrected.

16 years agoRename "mbuf_jumbo_pagesize" to "mbuf_jumbo_page". It makes it aligned correctly
wkoszek [Tue, 25 Dec 2007 14:17:16 +0000 (14:17 +0000)]
Rename "mbuf_jumbo_pagesize" to "mbuf_jumbo_page". It makes it aligned correctly
within ddb(4) and in the vmstat(8) output.

This change requires netstat(8) to be recompiled.

Reviewed by: rwatson@
  Tested by: make LINT

16 years agoUse __FBSDID() in the kernel BPF implementation.
rwatson [Tue, 25 Dec 2007 13:24:02 +0000 (13:24 +0000)]
Use __FBSDID() in the kernel BPF implementation.

MFC after: 3 days

16 years agoChange "audit_pipe_preselect" to "audit_pipe_presel" to make it print
wkoszek [Tue, 25 Dec 2007 13:23:19 +0000 (13:23 +0000)]
Change "audit_pipe_preselect" to "audit_pipe_presel" to make it print
with proper alignment in ddb(4) and vmstat(8).

Reviewed by: rwatson@

16 years agoWorkaround p->numbytes overflow, which can result in infinite loop inside
oleg [Tue, 25 Dec 2007 09:36:51 +0000 (09:36 +0000)]
Workaround p->numbytes overflow, which can result in infinite loop inside
dummynet module (prerequisite is using queues with "fat" pipe).

PR: kern/113548

16 years agoSpring forward into the late 1970's by following item 5 in the yacc
imp [Tue, 25 Dec 2007 06:22:33 +0000 (06:22 +0000)]
Spring forward into the late 1970's by following item 5 in the yacc
manual's Appendix D ("Old Features Supported but not Encouraged") in
the Seventh Edition Unix Programmer's Manual (January, 1979) by
retiring the " = {" method of of action specification in favor of a
plain "{".  It is no longer necessary for this bootstrap program to
be compatible with 6th Edition systems.  Some yaccs in the wild do not
support this old syntax any more, and compatibility with those systems
is more important these days (as there are easily 7 orders of magnitude
more of them than real v6 systems today).

Reviewed by: jhb@ and dds@ (the latter gave the reference).

16 years agoAdd the ability to clean up all shared memory segments which are
edwin [Tue, 25 Dec 2007 00:52:24 +0000 (00:52 +0000)]
Add the ability to clean up all shared memory segments which are
unused in one go.

From the original PR:

I've observed that linux apps running under the linuxulator
have a habit of leaving behind shared memory segments which
are unused, but which eventually cause the system to run
out of free segments and these apps will stop working.
ipcrm(1) currently only allows removal of unused message
queues, shared memory segments and semaphores on an individual
basis, or those having a matching (non-zero) key. However
it would often be convenient to just do a complete cleanup
of everything, usually as root.

PR: bin/118292
Submitted by: Callum Gibson <callumgibson@optusnet.com.au>
Not reviewed by: grog@
Approved by: grog@

16 years agoWhen copying multiple files to a directory, make sure that a proper
edwin [Tue, 25 Dec 2007 00:40:32 +0000 (00:40 +0000)]
When copying multiple files to a directory, make sure that a proper
warning is given when the directory doesn't exist.

PR: bin/50656
Submitted by: Edwin Groothuis <edwin@mavetju.org>
Approved by: grog@
Not reviewed by: grog@

16 years agoAdd sysctl mibs for _TSP, _TC1 and _TC2 which is user overridable
ume [Mon, 24 Dec 2007 16:32:14 +0000 (16:32 +0000)]
Add sysctl mibs for _TSP, _TC1 and _TC2 which is user overridable
but is blocked on user_override mib.
Not a few people want to use a passive cooling without their ACPI
BIOS support.

Reviewed by: njl

16 years agoMove the check for the snp device being already attached after the
kib [Mon, 24 Dec 2007 13:47:16 +0000 (13:47 +0000)]
Move the check for the snp device being already attached after the
fget() call, that is sleeping point, and possibly dropping Giant.

The snp_target == NULL implies the snp_tty == NULL. Remove the code
that is put under snp_target == NULL and snp_tty != NULL clause.

In snpclose(), do the snp_detach() before scheduling the snp device
destruction. Otherwise, after the return from snpclose(), the snp
device is already removed from the snp_list, but tty is still in
snooped state. Any attempt to do i/o on such tty cause panic because
ttytosnp() returns NULL.

Tested by: Peter Holm
MFC after: 1 week

16 years agoAdd Turkish collate
ache [Mon, 24 Dec 2007 10:30:09 +0000 (10:30 +0000)]
Add Turkish collate

PR:             118976
Submitted by:   Ismail YENIGUL <ismail.yenigul@endersys.com.tr>

16 years agoVarious fixes:
marcel [Mon, 24 Dec 2007 01:01:59 +0000 (01:01 +0000)]
Various fixes:
o  BSD disklabels have relative offsets. Even for the BSD in MBR slice
   setup, except when the mbroffset ioctl is supported. Since we don't
   support that ioctl, bsdlabel(8) expects relative offsets. So, when
   reading an existing disklabel, correct for disklabels that mistakenly
   have the mbroffset offsets.
o  Don't take the geometry seriously, because it's untrustworthy. We do
   expect the numbers to be within range. This means that the secperunit
   field will not be computed from secpercyl and ncyls, but simply is
   the mediasize in sectors.
o  Don't enforce partitions to be aligned to track boundaries. The
   default label, constructed by bsdlabel(8), puts partition a at offset
   BBSIZE bytes, which commonly means sector 16.

16 years agoFor the INDEX file to be used on disc1 of the 7.0-RELEASE the longest
kensmith [Mon, 24 Dec 2007 00:43:11 +0000 (00:43 +0000)]
For the INDEX file to be used on disc1 of the 7.0-RELEASE the longest
"build dependencies" field is 5,108 characters which overflows the
length of the junk buffer by a teeny bit.  This whole section needs
much more error checking but for now just completely ignore stuff
we have no interest in instead of copying it to someplace we don't
use in the process.

Insta-MFC probably coming since this is holding up 7.0-RC1...

16 years ago- Fix a typo in comments.
stas [Sun, 23 Dec 2007 23:31:27 +0000 (23:31 +0000)]
- Fix a typo in comments.

MFC after: 1 week
Approved by: cognet

16 years agoOn Christmas lights.
wilko [Sun, 23 Dec 2007 16:31:05 +0000 (16:31 +0000)]
On Christmas lights.

16 years ago- remove code from oroginal file, which is not required on BWCT boards
ticso [Sun, 23 Dec 2007 14:57:35 +0000 (14:57 +0000)]
- remove code from oroginal file, which is not required on BWCT boards
- Be more chatty on startup, since we have enough code space on
  AT91RM9200
- init DS1672 charging
- init USART GPIO

16 years agoBWCT boards uses two different SPI flash chips
ticso [Sun, 23 Dec 2007 14:46:30 +0000 (14:46 +0000)]
BWCT boards uses two different SPI flash chips
check for both status codes

16 years agoRemove trailing whitespace from lines in BPF.
rwatson [Sun, 23 Dec 2007 14:10:33 +0000 (14:10 +0000)]
Remove trailing whitespace from lines in BPF.

MFC after: 3 days

16 years agoRevert previous commit and fix OpenPAM issue properly.
des [Sat, 22 Dec 2007 19:29:32 +0000 (19:29 +0000)]
Revert previous commit and fix OpenPAM issue properly.

16 years agoAdd usage example:
mav [Sat, 22 Dec 2007 13:31:01 +0000 (13:31 +0000)]
Add usage example:
Limit outgoing data rate over fxp0 Ethernet interface to 20Mbit/s
and incoming packet rate to 5000pps.

16 years agoWhen IPSEC fails to allocate policy state for an inpcb, and MAC is in use,
rwatson [Sat, 22 Dec 2007 10:06:11 +0000 (10:06 +0000)]
When IPSEC fails to allocate policy state for an inpcb, and MAC is in use,
free the MAC label on the inpcb before freeing the inpcb.

MFC after: 3 days
Submitted by: tanyong <tanyong at ercist dot iscas dot ac dot cn>,
zhouzhouyi

16 years agogive thread0 the tid 100000 and bumpt the others to start at 100001
julian [Sat, 22 Dec 2007 04:56:48 +0000 (04:56 +0000)]
give thread0 the tid 100000 and bumpt the others to start at 100001

MFC after: 1 week

16 years agoMake SCHED_ULE buildable with gcc3.
wkoszek [Fri, 21 Dec 2007 23:30:18 +0000 (23:30 +0000)]
Make SCHED_ULE buildable with gcc3.

Reviewed by: cognet (mentor), jeffr
Approved by: cognet (mentor), jeffr

16 years ago- Remove incomplete sentence fragment
gabor [Fri, 21 Dec 2007 21:48:54 +0000 (21:48 +0000)]
- Remove incomplete sentence fragment

PR: docs/117451
Submitted by: Warren Block <wblock@wonkity.com>
MFC after: 3 days

16 years agoFix/workaround build breakage caused by PAM import
kmacy [Fri, 21 Dec 2007 21:47:19 +0000 (21:47 +0000)]
Fix/workaround build breakage caused by PAM import

struct pam_conv takes a void * for the appdata_ptr  but is being passed
a const char * - explicitly cast away the const

16 years ago- Document PORT_DBDIR
gabor [Fri, 21 Dec 2007 21:44:33 +0000 (21:44 +0000)]
- Document PORT_DBDIR

PR: docs/114731
Submitted by: Matthew Seaman <m.seaman@infracaninophile.co.uk>

16 years ago- Document icase variable of make search, which can be used to control
gabor [Fri, 21 Dec 2007 21:32:12 +0000 (21:32 +0000)]
- Document icase variable of make search, which can be used to control
  case-sensitivity

PR: docs/117593
Submitted by: Nikolay Bachiyski <nbachiyski@developer.bg>

16 years agoClean up some of the pts(4) vs pty(4) stuff in grantpt(3) and friends:
jhb [Fri, 21 Dec 2007 21:26:08 +0000 (21:26 +0000)]
Clean up some of the pts(4) vs pty(4) stuff in grantpt(3) and friends:
- Use PTY* for all pty(4) related constants.
- Use PTMX* for all pts(4) related constants.
- Consistently use _PATH_DEV PTMX rather than "/dev/ptmx".
- Revert 1.7 and properly fix it by using the correct prefix string for
  pts(4) masters.

MFC after: 3 days

16 years agoIn the following scenario:
dougb [Fri, 21 Dec 2007 19:34:26 +0000 (19:34 +0000)]
In the following scenario:
1. Start mergemaster
2. Interrupt it
3. < Somehow the temproot directory disappears >
4. mergemaster -r
Many bad things can happen, especially if the -i option is in use.

Therefore, add a check to make sure it still exists before we start
comparing files.

Brought to my attention by:
PR: bin/40538
Submitted by: Cyrille Lefevre <cyrille.lefevre@laposte.net>

16 years agoMore properly handle links who only have 1 valid IRQ in their bitmask. The
jhb [Fri, 21 Dec 2007 16:53:27 +0000 (16:53 +0000)]
More properly handle links who only have 1 valid IRQ in their bitmask.  The
old code special cased them too early which caused a few differences for
these sort of links relative to other PCI links:

- They were always re-routed via the BIOS call instead of assuming that
  they were already routed if the BIOS had programmed the IRQ into a
  matching device during POST.
- If the BIOS did route that link to a different IRQ that was marked as
  invalid, we trusted the $PIR table rather than the BIOS IRQ.

This change moves the special casing for "unique IRQ" links to only take
that into account when picking an IRQ for an unrouted link so that these
links will now not be routed if the BIOS appears to have routed it already
(some BIOSen have problems with that) and so that if the BIOS uses a
different IRQ than the $PIR, we trust the BIOS routing instead (this is
what we do for all other links as well).

Reported by: Bruce Walter  walter of fortean com
MFC after: 1 week

16 years agoAdjust for OpenPAM Hydrangea.
des [Fri, 21 Dec 2007 12:00:16 +0000 (12:00 +0000)]
Adjust for OpenPAM Hydrangea.

16 years agoThis commit was generated by cvs2svn to compensate for changes in r174835,
des [Fri, 21 Dec 2007 11:56:21 +0000 (11:56 +0000)]
This commit was generated by cvs2svn to compensate for changes in r174835,
which included commits to RCS files with non-trunk default branches.

16 years agoThis commit was generated by cvs2svn to compensate for changes in r174832,
des [Fri, 21 Dec 2007 11:49:29 +0000 (11:49 +0000)]
This commit was generated by cvs2svn to compensate for changes in r174832,
which included commits to RCS files with non-trunk default branches.

16 years agoVendor import of OpenPAM Hydrangea.
des [Fri, 21 Dec 2007 11:49:29 +0000 (11:49 +0000)]
Vendor import of OpenPAM Hydrangea.

16 years agousbhidctl appears to not report features and write size correctly
imp [Fri, 21 Dec 2007 03:40:36 +0000 (03:40 +0000)]
usbhidctl appears to not report features and write size correctly

This turns out to be due to an argument botch for hid_report_size.
The PR contained patches to fix the argument botch.

Submitted by: Maurice Castro
PR: usb/118915

16 years agoUpdate the comment describing vm_phys_unfree_page().
alc [Fri, 21 Dec 2007 02:44:31 +0000 (02:44 +0000)]
Update the comment describing vm_phys_unfree_page().

16 years agoMinor initialization change to not trigger bogus gcc warning about
imp [Fri, 21 Dec 2007 01:00:04 +0000 (01:00 +0000)]
Minor initialization change to not trigger bogus gcc warning about
passing literal strings to the constructors of string.

16 years agoWarner lets a good one out..
julian [Fri, 21 Dec 2007 00:56:16 +0000 (00:56 +0000)]
Warner lets a good one out..

16 years agoApply missing s/rv/res/g in previous commit.
marcel [Fri, 21 Dec 2007 00:23:23 +0000 (00:23 +0000)]
Apply missing s/rv/res/g in previous commit.

16 years agoModify vm_phys_unfree_page() so that it no longer requires the given
alc [Thu, 20 Dec 2007 22:45:54 +0000 (22:45 +0000)]
Modify vm_phys_unfree_page() so that it no longer requires the given
page to be in the free lists.  Instead, it now returns TRUE if it
removed the page from the free lists and FALSE if the page was not
in the free lists.

This change is required to support superpage reservations.  Specifically,
once reservations are introduced, a cached page can either be in the
free lists or a reservation.

16 years agoMFamd64/ia64/i386: Only set the rman bus tags and handles in
jhb [Thu, 20 Dec 2007 21:42:43 +0000 (21:42 +0000)]
MFamd64/ia64/i386: Only set the rman bus tags and handles in
bus_activate_resource() methods instead of splitting it up between
bus_alloc_resource() and bus_activate_resource().

Glanced at by: marcel

16 years agoAvoid holding the aac_io_lock over copyout.
emaste [Thu, 20 Dec 2007 21:13:58 +0000 (21:13 +0000)]
Avoid holding the aac_io_lock over copyout.

Submitted by: Achim Leubner @ Adaptec.

16 years agoInvoke revoke(2) on the slave pty in the pts(4) case (new_openpty()) to
jhb [Thu, 20 Dec 2007 21:10:06 +0000 (21:10 +0000)]
Invoke revoke(2) on the slave pty in the pts(4) case (new_openpty()) to
kick off any other users on the device line before using it since
openpty(3) is documented to do this.  Note that grantpt(3) does not
call revoke(2), it only adjusts permissions and ownership.

MFC after: 3 days

16 years agoUpdate pkg_version_index to INDEX-8
dougb [Thu, 20 Dec 2007 20:37:22 +0000 (20:37 +0000)]
Update pkg_version_index to INDEX-8

16 years ago- Adjust INDEX file name to INDEX-8 so that it matches the major FreeBSD
gabor [Thu, 20 Dec 2007 20:02:16 +0000 (20:02 +0000)]
- Adjust INDEX file name to INDEX-8 so that it matches the major FreeBSD
  version number, as it is named after that

PR: docs/118053
Submitted by: Michael Plass <mfp49_freebsd@plass-family.net>

16 years agoFix disorder introduced in previous commit.
ru [Thu, 20 Dec 2007 16:40:25 +0000 (16:40 +0000)]
Fix disorder introduced in previous commit.

16 years agoMinistat was repocopied to src/usr.bin/ministat
phk [Thu, 20 Dec 2007 10:18:15 +0000 (10:18 +0000)]
Ministat was repocopied to src/usr.bin/ministat

16 years agoPull ministat into the installed system and write it a man-page.
phk [Thu, 20 Dec 2007 10:17:07 +0000 (10:17 +0000)]
Pull ministat into the installed system and write it a man-page.

(Repocopied from src/tools/tools/ministat)

16 years agoBy definition promiscuous mode should see all unicast frames as well
yongari [Thu, 20 Dec 2007 07:26:20 +0000 (07:26 +0000)]
By definition promiscuous mode should see all unicast frames as well
as multicast/broadcast frames. Previously re(4) ignored multicast
frames in promiscuous mode. The RTL8169 datasheet was not clear
how it handles multicast frames in promiscuous mode.

PR: kern/118572
MFC after: 3 days

16 years agocall underscore version of pthread_cleanup_pop instead.
davidxu [Thu, 20 Dec 2007 04:40:12 +0000 (04:40 +0000)]
call underscore version of pthread_cleanup_pop instead.

16 years agoRemove vfork() overloading, it is no longer needed.
davidxu [Thu, 20 Dec 2007 04:32:28 +0000 (04:32 +0000)]
Remove vfork() overloading, it is no longer needed.

16 years agoGrumble. DO declare logbl(), DON'T declare logl() just yet.
das [Thu, 20 Dec 2007 03:16:55 +0000 (03:16 +0000)]
Grumble. DO declare logbl(), DON'T declare logl() just yet.
bde is going to commit logl() Real Soon Now.
I'm just trying to slow him down with merge conflicts.

Noticed by: bde

16 years agoenable wired driver support
sam [Thu, 20 Dec 2007 00:52:02 +0000 (00:52 +0000)]
enable wired driver support

Submitted by: "Paul B. Mahol" <onemda@gmail.com>
MFC after: 1 week

16 years agoIn openprom_ioctl() ensure appropriate permissions and that data isn't
marius [Thu, 20 Dec 2007 00:31:04 +0000 (00:31 +0000)]
In openprom_ioctl() ensure appropriate permissions and that data isn't
NULL and doesn't point to a NULL pointer before dereferencing it. This
fixes a panic triggered by Xorg 7.3.

Reported and tested by: Bill Green
MFC after: 3 days

16 years agoRemove the declaration of logl(). The relevant bits haven't been
das [Thu, 20 Dec 2007 00:06:33 +0000 (00:06 +0000)]
Remove the declaration of logl(). The relevant bits haven't been
committed yet, but the declaration leaked in when I added nan() and
friends.

Reported by: pav

16 years agoDescribe new NG_CAR_COUNT_PACKETS option.
mav [Wed, 19 Dec 2007 23:12:37 +0000 (23:12 +0000)]
Describe new NG_CAR_COUNT_PACKETS option.

16 years agoCorrect one half of a loop continuation condition in vm_phys_unfree_page().
alc [Wed, 19 Dec 2007 23:09:45 +0000 (23:09 +0000)]
Correct one half of a loop continuation condition in vm_phys_unfree_page().
At present, this error is inconsequential; the other half of the loop
continuation condition is sufficient to achieve correct execution.

16 years agoAdd option to set packets per second limits instead of default
mav [Wed, 19 Dec 2007 22:50:14 +0000 (22:50 +0000)]
Add option to set packets per second limits instead of default
bits per second ones.

16 years agoWhen devclass_get_maxunit is passed a NULL, return -1 to indicate that
imp [Wed, 19 Dec 2007 22:05:07 +0000 (22:05 +0000)]
When devclass_get_maxunit is passed a NULL, return -1 to indicate that
there's nothing allocated at all yet.

16 years agoUse .An -nosplit in the AUTHORS section of sf_buf(9) to avoid an undesired
rwatson [Wed, 19 Dec 2007 20:34:09 +0000 (20:34 +0000)]
Use .An -nosplit in the AUTHORS section of sf_buf(9) to avoid an undesired
line break.

MFC after: 3 days

16 years agoo NetBSD 4.0 added.
maxim [Wed, 19 Dec 2007 18:19:53 +0000 (18:19 +0000)]
o NetBSD 4.0 added.

16 years agoRedefine bus_space_tag_t on PowerPC from a 32-bit integral to
marcel [Wed, 19 Dec 2007 18:00:50 +0000 (18:00 +0000)]
Redefine bus_space_tag_t on PowerPC from a 32-bit integral to
a pointer to struct bus_space. The structure contains function
pointers that do the actual bus space access.

The reason for this change is that previously all bus space
accesses were little endian (i.e. had an explicit byte-swap
for multi-byte accesses), because all busses on Macs are little
endian.
The upcoming support for Book E, and in particular the E500
core, requires support for big-endian busses because all
embedded peripherals are in the native byte-order.

With this change, there's no distinction between I/O port
space and memory mapped I/O. PowerPC doesn't have I/O port
space. Busses assign tags based on the byte-order only.
For that purpose, two global structures exist (bs_be_tag and
bs_le_tag), of which the address can be taken to get a valid
tag.

Obtained from: Juniper, Semihalf

16 years agoActually program the interrupt controller for priorities. As we
imp [Wed, 19 Dec 2007 17:34:17 +0000 (17:34 +0000)]
Actually program the interrupt controller for priorities.  As we
support more AT91 platforms, we'll need to move this into some
platform init routine.

16 years agoAdd Dell's sub-vendor id to identify PERC6 RAID controller which has LSI
ambrisko [Wed, 19 Dec 2007 17:23:47 +0000 (17:23 +0000)]
Add Dell's sub-vendor id to identify PERC6 RAID controller which has LSI
vendor id's.

16 years agoI incorrectly assumed the log buffer started from the beginning when it
ambrisko [Wed, 19 Dec 2007 17:22:07 +0000 (17:22 +0000)]
I incorrectly assumed the log buffer started from the beginning when it
is actually a circular log.  Deal with it rolling around.  Fortunately,
the log area is big and I haven't seen any roll over yet.  Update and
get rid of the obsolete comment.

16 years agoAllow negative values to be specified in the loader.
ambrisko [Wed, 19 Dec 2007 17:06:32 +0000 (17:06 +0000)]
Allow negative values to be specified in the loader.

16 years agoThe default of 3 retransmits corresponds to a timeout value of
ru [Wed, 19 Dec 2007 16:58:54 +0000 (16:58 +0000)]
The default of 3 retransmits corresponds to a timeout value of
45 seconds, not 15.

16 years agoFix bugs in the TCP syncache timeout code. including:
ru [Wed, 19 Dec 2007 16:56:28 +0000 (16:56 +0000)]
Fix bugs in the TCP syncache timeout code. including:

When system ticks are positive, for entries in the cache
bucket, syncache_timer() ran on every tick (doing nothing
useful) instead of the supposed 3, 6, 12, and 24 seconds
later (when it's time to retransmit SYN,ACK).

When ticks are negative, syncache_timer() was scheduled
for the too far future (up to ~25 days on systems with
HZ=1000), no SYN,ACK retransmits were attempted at all,
and syncache entries added in that period that correspond
to non-established connections stay there forever.

Only HEAD and RELENG_7 are affected.

Reviewed by: silby, kmacy (earlier version)
Submitted by: Maxim Dounin, ru

16 years agoDo not recursively acquire aac_io_lock in aac_ioctl_event.
emaste [Wed, 19 Dec 2007 14:33:40 +0000 (14:33 +0000)]
Do not recursively acquire aac_io_lock in aac_ioctl_event.

MFC After: 3 days

16 years agoUpdate the number of reported tests to keep prove(1) happy.
dds [Wed, 19 Dec 2007 11:52:33 +0000 (11:52 +0000)]
Update the number of reported tests to keep prove(1) happy.

16 years agoCorrect the "move file from directory to existing directory" test for
dds [Wed, 19 Dec 2007 11:00:22 +0000 (11:00 +0000)]
Correct the "move file from directory to existing directory" test for
cross-device moves.

16 years agoAdd missing section number.
brueffer [Wed, 19 Dec 2007 08:01:24 +0000 (08:01 +0000)]
Add missing section number.

16 years agoEliminate redundant code from vm_page_startup().
alc [Wed, 19 Dec 2007 05:47:50 +0000 (05:47 +0000)]
Eliminate redundant code from vm_page_startup().

16 years agoRemove extraneous debug statements.
kmacy [Wed, 19 Dec 2007 05:17:40 +0000 (05:17 +0000)]
Remove extraneous debug statements.

Noticed by: Andrey Chernov

16 years agoAdd missing dependencies on Makefile (fixes NO_CLEAN builds).
ru [Wed, 19 Dec 2007 05:10:07 +0000 (05:10 +0000)]
Add missing dependencies on Makefile (fixes NO_CLEAN builds).

16 years agoReduce lock contention for simple cases.
imp [Wed, 19 Dec 2007 04:30:10 +0000 (04:30 +0000)]
Reduce lock contention for simple cases.

# this really should be done with pthread_once, but I've debugged this code.

Reviewed by: arch@

16 years agoAdd note about other systems.
imp [Wed, 19 Dec 2007 03:33:13 +0000 (03:33 +0000)]
Add note about other systems.

16 years agoCorrect typo in usage message.
imp [Wed, 19 Dec 2007 03:31:44 +0000 (03:31 +0000)]
Correct typo in usage message.

16 years agoGo ahead and install usbdevs. This will be used by the usbdevs device
imp [Wed, 19 Dec 2007 01:28:17 +0000 (01:28 +0000)]
Go ahead and install usbdevs.  This will be used by the usbdevs device
to print more information about some devices.

16 years agoAdd verbage on some of the rules for a 'filter' vs an 'ithread'
imp [Wed, 19 Dec 2007 01:26:34 +0000 (01:26 +0000)]
Add verbage on some of the rules for a 'filter' vs an 'ithread'
interrupt handler.

16 years agoAs several people pointed out, I did all the ctype casts the wrong
des [Wed, 19 Dec 2007 00:26:36 +0000 (00:26 +0000)]
As several people pointed out, I did all the ctype casts the wrong
way (not for the first time...)

Noticed by: bde, ru ++
MFC after: 1 week

16 years agoMove all the xprintf-related symbols to FBSDprivate_1.0.
das [Tue, 18 Dec 2007 23:49:05 +0000 (23:49 +0000)]
Move all the xprintf-related symbols to FBSDprivate_1.0.

Discussed with: deischen, kan, phk

16 years agoSince nan() is supposed to work the same as strtod("nan(...)", NULL),
das [Tue, 18 Dec 2007 23:46:32 +0000 (23:46 +0000)]
Since nan() is supposed to work the same as strtod("nan(...)", NULL),
my original implementation made both use the same code. Unfortunately,
this meant libm depended on a vendor header at compile time and previously-
unexposed vendor bits in libc at runtime.

Hence, I just wrote my own version of the relevant vendor routine. As it
turns out, mine has a factor of 8 fewer of lines of code, and is a bit more
readable anyway. The strtod() and *scanf() routines still use vendor code.

Reviewed by: bde

16 years agoDon't overload tcp_usrreqs unless the kernel doesn't provide offload support.
kmacy [Tue, 18 Dec 2007 23:00:25 +0000 (23:00 +0000)]
Don't overload tcp_usrreqs unless the kernel doesn't provide offload support.

16 years agoIncorporate TCP offload hooks in to core TCP code.
kmacy [Tue, 18 Dec 2007 22:59:07 +0000 (22:59 +0000)]
Incorporate TCP offload hooks in to core TCP code.
  - Rename output routines tcp_gen_* -> tcp_output_*.
  - Rename notification routines that turn in to no-ops in the absence of TOE
    from tcp_gen_* -> tcp_offload_*.
  - Fix some minor comment nits.
  - Add a /* FALLTHROUGH */

Reviewed by: Sam Leffler, Robert Watson, and Mike Silbersack

16 years agoBe more exact with sigaction SA_SIGINFO handling.
obrien [Tue, 18 Dec 2007 20:39:13 +0000 (20:39 +0000)]
Be more exact with sigaction SA_SIGINFO handling.

Reviewed by: marcel

16 years agoFix printing of the number of syncache entries added.
ru [Tue, 18 Dec 2007 12:07:10 +0000 (12:07 +0000)]
Fix printing of the number of syncache entries added.

16 years agoDocument NO_PROXY / no_proxy.
des [Tue, 18 Dec 2007 11:03:26 +0000 (11:03 +0000)]
Document NO_PROXY / no_proxy.

MFC after: 3 weeks

16 years agoAdd support for the NO_PROXY / no_proxy environment variable as used by
des [Tue, 18 Dec 2007 11:03:07 +0000 (11:03 +0000)]
Add support for the NO_PROXY / no_proxy environment variable as used by
lynx, curl etc.  Note that this patch differs significantly from that
in the PR, as the submitter refined it after submitting the PR.

PR: 110388
Submitted by: Alexander Pohoyda <alexander.pohoyda@gmx.net>
MFC after: 3 weeks

16 years agoOld patch I had lying around: correctly cast the argument to is*().
des [Tue, 18 Dec 2007 10:41:12 +0000 (10:41 +0000)]
Old patch I had lying around: correctly cast the argument to is*().
IWBNI gcc could warn about this the way it warns about printf() abuse.

MFC after: 1 week

16 years agoStylify ipcs/ipcs.c and ipcrm/ipcrm in preparation of the upcoming
edwin [Tue, 18 Dec 2007 09:39:47 +0000 (09:39 +0000)]
Stylify ipcs/ipcs.c and ipcrm/ipcrm in preparation of the upcoming
changes as proposed in bin/118292.

Feel free to mention any I have missed, there is much to learn with
regarding to style(9).

Approved by: grog@

16 years agoSimplify the error handling and use the dereferenced sc->sc_ifp pointer.
thompsa [Tue, 18 Dec 2007 09:13:04 +0000 (09:13 +0000)]
Simplify the error handling and use the dereferenced sc->sc_ifp pointer.

16 years agoPortability improvements to investigate behavior of other OSs.
dds [Tue, 18 Dec 2007 08:53:04 +0000 (08:53 +0000)]
Portability improvements to investigate behavior of other OSs.
Now works under Solaris and Linux.

16 years agoAdd more tests. All rename(2)-based tests now succeed.
dds [Tue, 18 Dec 2007 08:49:47 +0000 (08:49 +0000)]
Add more tests.  All rename(2)-based tests now succeed.
The performance of the cross-device equivalents is under investigation.

16 years agoWhen the bridge has an address and a packet comes in for it then drop it if the
thompsa [Tue, 18 Dec 2007 07:04:50 +0000 (07:04 +0000)]
When the bridge has an address and a packet comes in for it then drop it if the
link has been marked discarding by Spanning Tree. This would cause the bridge
to see duplicate packets to itself even if STP has correctly calculated the
topology and blocked redundant links.

Reported by: trasz
Tested by: trasz
MFC after: 3 days

16 years agoUse fixed point integer math instead of floating point math when
jasone [Tue, 18 Dec 2007 05:27:57 +0000 (05:27 +0000)]
Use fixed point integer math instead of floating point math when
calculating run sizes.  Use of the floating point unit was a potential
pessimization to context switching for applications that do not otherwise
use floating point math. [1]

Reformat cpp macro-related comments to improve consistency.

Submitted by: das

16 years ago - Use the macro to check the port status has it will also test if its
thompsa [Tue, 18 Dec 2007 02:12:03 +0000 (02:12 +0000)]
 - Use the macro to check the port status has it will also test if its
   administratively down (!IFF_UP)
 - Use the same parameters to lagg_link_active() to get the backup port as in
   the output path, this didnt actually matter in practice as sc_primary is
   always the first on the port list.

MFC after: 3 days

16 years agoFix logical bug in the bzip2 reading code, which results in bogus EIO
sobomax [Tue, 18 Dec 2007 01:50:49 +0000 (01:50 +0000)]
Fix logical bug in the bzip2 reading code, which results in bogus EIO
returned on a perfectly valid bzip2 stream whose decompressed size
is multiple of read-ahead buffer size. Reproduce the problem is easy:
create some power-of-two sized file (truncate -s 1m file will do),
bzip2 it and try to load it as md_image from loader. See how it fails.

The bug doesn't affect gzip code (which most of bzip2-reading code was
copied from) probably due to the fact that libgzip doesn't report
Z_STREAM_END with the last block, but requires extra call to inflate()
to retrieve it and has some extra data in the input stream at that time.
However, apply similar fix to gzipfs.c just in the case the API will
change in the future to do what bzip2 code does.

Add some ifdef'ed code to enable testing bzipfs.c from witin normal
FreeBSD environment as opposed to the restricted loader one, so that
one can use gdb and whatnot.

Sponsored by: Sippy Software, Inc., http://www.sippysoft.com/
MFC in: 7 days

16 years ago- sctp-iterator should run at PI_NET priority ...not 0.
rrs [Tue, 18 Dec 2007 01:24:15 +0000 (01:24 +0000)]
- sctp-iterator should run at PI_NET priority ...not 0.

MFC after: 1 week