]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
13 years agoMFC
attilio [Mon, 16 May 2011 16:34:03 +0000 (16:34 +0000)]
MFC

13 years agoMFC
attilio [Sun, 15 May 2011 15:47:16 +0000 (15:47 +0000)]
MFC

13 years agoAdd I2C bus driver for the AMD Geode LX series CS5536 Companion
brix [Sun, 15 May 2011 14:01:23 +0000 (14:01 +0000)]
Add I2C bus driver for the AMD Geode LX series CS5536 Companion
Device.

Reviewed by:    jhb (newbus bits only), adrian

13 years agoRecognize the SAB 82532 found in Fujitsu PRIMEPOWER650 and 900.
marius [Sun, 15 May 2011 13:27:38 +0000 (13:27 +0000)]
Recognize the SAB 82532 found in Fujitsu PRIMEPOWER650 and 900.

13 years agoRecognize the eeprom device found in Fujitsu PRIMEPOWER650 and 900.
marius [Sun, 15 May 2011 13:25:26 +0000 (13:25 +0000)]
Recognize the eeprom device found in Fujitsu PRIMEPOWER650 and 900.

13 years agoAdd support for MK48T37.
marius [Sun, 15 May 2011 13:17:08 +0000 (13:17 +0000)]
Add support for MK48T37.

13 years ago- Add 10baseT as an alias for 10baseT/UTP.
marius [Sun, 15 May 2011 12:58:29 +0000 (12:58 +0000)]
- Add 10baseT as an alias for 10baseT/UTP.
- Add shorthand aliases for common media+option combinations as announced
  by miibus(4) so that one can actually supply the media strings found in
  the dmesg output to ifconfig(8).

Obtained from: NetBSD (in principle)
MFC after: 2 weeks

13 years agoWhen setting media always and not just in case of switching to IFM_AUTO
marius [Sun, 15 May 2011 12:51:00 +0000 (12:51 +0000)]
When setting media always and not just in case of switching to IFM_AUTO
clear the options of the current media, i.e. only inherit the instance,
which matches what NetBSD does. Without this it's really non-intuitive
that the following sequence:
ifconfig bge0 media 1000baseT mediaopt full-duplex
ifconfig bge0 media 100baseTX
results in 100baseTX full-duplex to be set or that:
ifconfig bge0 media autoselect mediaopt flowcontrol
ifconfig bge0 media 1000baseT mediaopt full-duplex
tries to set 1000baseT full-duplex with flowcontrol, which isn't suported
und thus fails while the following:
ifconfig re0 media 1000baseT mediaopt flowcontrol,full-duplex
ifconfig re0 media autoselect
just switches to autoselection without flowcontrol.

MFC after: 2 weeks

13 years agoFix a memory leak possible in g_eli_key_allocate() if the key with the
trociny [Sun, 15 May 2011 12:39:30 +0000 (12:39 +0000)]
Fix a memory leak possible in g_eli_key_allocate() if the key with the
same keyno is added while we aren't holding the lock.

Approved by: pjd (mentor)
MFC after: 1 week

13 years agoSimplify the code a bit. For own providers GEOM_PART always provides
ae [Sun, 15 May 2011 11:45:13 +0000 (11:45 +0000)]
Simplify the code a bit. For own providers GEOM_PART always provides
"start" and "end" config attributes.

MFC after: 1 week

13 years agoOnly update the scheduler's byte count table for aggregation queues.
bschmidt [Sun, 15 May 2011 08:09:36 +0000 (08:09 +0000)]
Only update the scheduler's byte count table for aggregation queues.
The other queues, especially the command queue, uses the FIFO mode
which doesn't require the byte count table because queued entries are
processed in order.

Pointed out by: Lucius Windschuh <lwindschuh at googlemail dot com>

13 years agoFix NF calibration breakage introduced by me in a past commit.
adrian [Sun, 15 May 2011 07:59:33 +0000 (07:59 +0000)]
Fix NF calibration breakage introduced by me in a past commit.

Since the returned NF will be -ve, checking for <= 0 is not good
enough. For now, check whether it equals 0 or -1; a future commit
will tidy this mess up and have it return HAL_BOOL instead.

13 years agoFix few typos.
mav [Sun, 15 May 2011 01:01:53 +0000 (01:01 +0000)]
Fix few typos.

Submitted by: uqs
MFC after: 1 week

13 years ago- Whitespace fix.
stas [Sun, 15 May 2011 00:46:25 +0000 (00:46 +0000)]
- Whitespace fix.

13 years agoFix yet another inversion in the logic by applying the x86 version of this,
marius [Sat, 14 May 2011 23:20:14 +0000 (23:20 +0000)]
Fix yet another inversion in the logic by applying the x86 version of this,
which avoids CPU_EMPTY() in the first place.
Do I get a beer or something for every inversion I find?

13 years ago- Add symbol versioning to libprocstat.
stas [Sat, 14 May 2011 22:15:38 +0000 (22:15 +0000)]
- Add symbol versioning to libprocstat.

Suggested by: kib

13 years ago- There's no need for nibbletab to be static, it's const however.
marius [Sat, 14 May 2011 20:31:04 +0000 (20:31 +0000)]
- There's no need for nibbletab to be static, it's const however.
- Fix whitespace.

13 years agoFix a longstanding bug where only the first part of the cpumask was
attilio [Sat, 14 May 2011 19:36:12 +0000 (19:36 +0000)]
Fix a longstanding bug where only the first part of the cpumask was
correctly set full.

Submitted by: anonymous

13 years agosysctl that displays the absolute queue id of an rxq.
np [Sat, 14 May 2011 19:27:15 +0000 (19:27 +0000)]
sysctl that displays the absolute queue id of an rxq.

13 years agoMFC
attilio [Sat, 14 May 2011 19:20:13 +0000 (19:20 +0000)]
MFC

13 years agoConvert a file to UTF-8. This only changes a character in the
uqs [Sat, 14 May 2011 19:03:54 +0000 (19:03 +0000)]
Convert a file to UTF-8. This only changes a character in the
copyright section and will not require UTF-8 for fortune(6).

13 years agofortune(6): fix typos and misspellings.
uqs [Sat, 14 May 2011 19:03:45 +0000 (19:03 +0000)]
fortune(6): fix typos and misspellings.

While here, drop words in the spelling files that are no longer used
anywhere.

Speling errors found via: codespell from Lucas De Marchi

13 years agomorse(6): fix typos and space-before-tabs
uqs [Sat, 14 May 2011 19:03:30 +0000 (19:03 +0000)]
morse(6): fix typos and space-before-tabs

13 years agoFix the source address selection for boundall sockets
tuexen [Sat, 14 May 2011 18:22:14 +0000 (18:22 +0000)]
Fix the source address selection for boundall sockets
when sending INITs to a global IPv4 address having
only private IPv4 address.
Allow the usage of a private address and make sure
that no other private address will be used by the
association.
Initial work was done by rrs@.

MFC after: 1 week.

13 years agoSimplify the code here.
attilio [Sat, 14 May 2011 18:22:08 +0000 (18:22 +0000)]
Simplify the code here.

Submitted by: jhb

13 years agoIncrease size of cg_count to allow us to utilize >128 CPUs.
sbruno [Sat, 14 May 2011 17:56:13 +0000 (17:56 +0000)]
Increase size of cg_count  to allow us to utilize >128 CPUs.

Pad cg_count and cg_children to keep the struct aligned

Reviewed by: attilio@

13 years agoEncourage Ben Laurie to finish getting his commit bit by appointing him
cperciva [Sat, 14 May 2011 17:44:12 +0000 (17:44 +0000)]
Encourage Ben Laurie to finish getting his commit bit by appointing him
as the OpenSSL maintainer.

With hat: security-officer
Reviewed by: simon (former maintainer)

13 years agoCurrently we are unable to use capsicum for the primary worker process,
pjd [Sat, 14 May 2011 17:02:03 +0000 (17:02 +0000)]
Currently we are unable to use capsicum for the primary worker process,
because we need to do ioctl(2)s, which are not permitted in the capability
mode. What we do now is to chroot(2) to /var/empty, which restricts access
to file system name space and we drop privileges to hast user and hast
group.

This still allows to access to other name spaces, like list of processes,
network and sysvipc.

To address that, use jail(2) instead of chroot(2). Using jail(2) will restrict
access to process table, network (we use ip-less jails) and sysvipc (if
security.jail.sysvipc_allowed is turned off). This provides much better
separation.

MFC after: 1 week

13 years agoWhen using capsicum to sanbox, still use other methods first, just in case
pjd [Sat, 14 May 2011 16:55:24 +0000 (16:55 +0000)]
When using capsicum to sanbox, still use other methods first, just in case
one of them have some problems.

13 years agoFix the Merlin 5ghz fast-clock EEPROM fetch to return the correct value.
adrian [Sat, 14 May 2011 15:24:15 +0000 (15:24 +0000)]
Fix the Merlin 5ghz fast-clock EEPROM fetch to return the correct value.

The eeprom Get method should return HAL_OK if fastclock is enabled in the
EEPROM. It was returning the opposite of what it should have.

Submitted by: Matthew Fleming <mdf356@gmail.com>

13 years agoFix the eeprom set API method to return HAL_STATUS.
adrian [Sat, 14 May 2011 15:12:02 +0000 (15:12 +0000)]
Fix the eeprom set API method to return HAL_STATUS.

The code assumed it could return HAL_OK, HAL_EINVAL and other
HAL_STATUS types; so it shouldn't be declared as returning HAL_BOOL.

This commit was brought to you by the Clang compiler.

Submitted by: Matthew Fleming <mdf356@gmail.com>

13 years agoPrefer switching the memory stack from user to kernel *before* switching
marcel [Sat, 14 May 2011 14:55:15 +0000 (14:55 +0000)]
Prefer switching the memory stack from user to kernel *before* switching
the register stack. While the ordering doesn't matter, it creates an
invariant not previously there: the memory stack pointer will always be
larger than the register stack pointer. With this invariant in place,
it's easier to add instrumentation code that detects a stack overflow
because in such a scenario the memory stack pointer and register stack
pointers have crossed each other.

Aside: basic kernel operation needs about half the stack size (~16K)
at most. We have plenty of head room on the kernel stack...

13 years agoSharpening the saw:
marcel [Sat, 14 May 2011 14:47:19 +0000 (14:47 +0000)]
Sharpening the saw:
o   Clobber the register that holds the restart token immediately after
    crossing the restart point. This prevents false positives (i.e. a
    nested exception that we don't know can happen and that is being
    treated as one we know by virtue of a lingering restart token).
o   Now that the bootstrap kernel stack is free, switch onto it and call
    trap() for nested traps that we don't know about. In trap we panic()
    so that we can analyze the condition.

13 years agoOops, fix order of sequence numbers in KASSERT()'s to catch negative
jhb [Sat, 14 May 2011 14:41:40 +0000 (14:41 +0000)]
Oops, fix order of sequence numbers in KASSERT()'s to catch negative
receive windows to match the labels in the panic message.

Submitted by: trociny

13 years agoBe pedantic: mark the pcpu pointer (= register r13) itself as volatile.
marcel [Sat, 14 May 2011 14:40:24 +0000 (14:40 +0000)]
Be pedantic: mark the pcpu pointer (= register r13) itself as volatile.

13 years agoTurn ia64_srlz() and ia64_srlz_i() into defines so that the code is
marcel [Sat, 14 May 2011 14:36:08 +0000 (14:36 +0000)]
Turn ia64_srlz() and ia64_srlz_i() into defines so that the code is
still correct when inlining is disabled.

13 years agoImport initial EEPROM code for Kite (AR9287).
adrian [Sat, 14 May 2011 14:25:15 +0000 (14:25 +0000)]
Import initial EEPROM code for Kite (AR9287).

I've tested this locally and it does indeed read and attach to an AR9287
EEPROM. But a lot more code needs to be ported over to the HAL before
the AR9287 is functional.

I'm importing this separate from the rest of the codebase (and unlinked from
the build for now) in case someone wishes to begin fiddling with porting
the rest of the code over from Linux ath9k.

Obtained from: Linux ath9k

13 years agosh: Add tests for lines that look like heredoc delimiters but are not.
jilles [Sat, 14 May 2011 14:19:30 +0000 (14:19 +0000)]
sh: Add tests for lines that look like heredoc delimiters but are not.

13 years agonetwork.subr: Use printf(1) builtin for hexprint function.
jilles [Sat, 14 May 2011 12:22:58 +0000 (12:22 +0000)]
network.subr: Use printf(1) builtin for hexprint function.

Now that printf(1) is a shell builtin, there is no need to emulate it
anymore. The external printf(1) is /usr/bin/printf and therefore may not be
available in early boot.

It may be faster to use printf directly but the function is useful for
compatibility.

13 years agoAdd new USB ID's.
hselasky [Sat, 14 May 2011 12:16:09 +0000 (12:16 +0000)]
Add new USB ID's.

Submitted by: Jim Bryant
MFC after: 3 days

13 years agogcc and clang semantics imply certain -mno-* flags when other certain -mno-*
brucec [Sat, 14 May 2011 11:26:00 +0000 (11:26 +0000)]
gcc and clang semantics imply certain -mno-* flags when other certain -mno-*
flags are also specified. This change makes use of this behaviour and removes
unneeded -mno-* flags.

Note that clang does not yet enable AVX support for any CPU. However at some
point in the future it will and since we definitely want to disable it for the
kernel, we might as well add the -mno-avx flag now.

Submitted by: arundel

13 years agoWhen disabling RIFS for Sowl (AR9160) and Howl (AR9130), make sure RIFS
adrian [Sat, 14 May 2011 05:43:33 +0000 (05:43 +0000)]
When disabling RIFS for Sowl (AR9160) and Howl (AR9130), make sure RIFS
is totally disabled.

The Atheros HAL code does this for Sowl/Howl but not for Owl (AR5416) where
RIFS is disabled by default.

This seems to quieten the occasional baseband hang I've been seeing with
the AR9160 in STA mode under constant heavy traffic load.

Obtained from: Atheros

13 years agoRestore BOOT2CFG, accidentally removed in r212938.
imp [Sat, 14 May 2011 04:32:30 +0000 (04:32 +0000)]
Restore BOOT2CFG, accidentally removed in r212938.

PR: 157020

13 years agoMajor fix: when doing open-loop TX power calibration, adjust
adrian [Sat, 14 May 2011 04:17:16 +0000 (04:17 +0000)]
Major fix: when doing open-loop TX power calibration, adjust
the correct CCK rates rather than adjusting the first handful.
This may have affected some AR9280 based NICs.

Minor fix: merlin check update

13 years agoFixes from the Atheros HAL - formatting; update Merlin checks to be consistent.
adrian [Sat, 14 May 2011 04:05:23 +0000 (04:05 +0000)]
Fixes from the Atheros HAL - formatting; update Merlin checks to be consistent.
Nothing functional should change with this commit.

13 years ago- Fix several typos [1]
stas [Sat, 14 May 2011 04:02:56 +0000 (04:02 +0000)]
- Fix several typos [1]
- fuser(1) writes usernames to stderr, not stdout [1]
- Add history section.

Submitted by: Pieter de Goeje <pieter@degoeje.nl> [1]

13 years agoAdd new modifier - "R", when it is specified the path to pid file
sobomax [Sat, 14 May 2011 03:00:55 +0000 (03:00 +0000)]
Add new modifier - "R", when it is specified the path to pid file
will be considered as a path to a binary or a shell script to be executed
after rotation has been completed instead of sending signal to the process
id in that file.

Sponsored by: Sippy Software, Inc.
From the: FreeBSD hacking lounge at BSDCan

13 years agoMFC
attilio [Sat, 14 May 2011 02:28:26 +0000 (02:28 +0000)]
MFC

13 years agoRemove actual files supporting sun4v.
attilio [Sat, 14 May 2011 01:56:48 +0000 (01:56 +0000)]
Remove actual files supporting sun4v.

Approved by: re

13 years agoDisconnect sun4v architecture from the three.
attilio [Sat, 14 May 2011 01:53:38 +0000 (01:53 +0000)]
Disconnect sun4v architecture from the three.

Some files keep the SUN4V tags as a code reference, for the future,
if any rewamped sun4v support wants to be added again.

Reviewed by: marius
Tested by: sbruno
Approved by: re

13 years agoEven though initial calibrations aren't done (yet), add this so we're
adrian [Sat, 14 May 2011 01:41:36 +0000 (01:41 +0000)]
Even though initial calibrations aren't done (yet), add this so we're
consistent with the Atheros HAL.

13 years agoMerge comments about converting directory entries to be more direct and
jhb [Sat, 14 May 2011 01:10:57 +0000 (01:10 +0000)]
Merge comments about converting directory entries to be more direct and
concise.

Inspired by: Gleb Kurtsou

13 years ago- Bump __FreeBSD_version to note the revision when libprocstat(3) and fuser(1)
stas [Fri, 13 May 2011 23:10:55 +0000 (23:10 +0000)]
- Bump __FreeBSD_version to note the revision when libprocstat(3) and fuser(1)
  have been added to the tree.

13 years agoConstify a few string variables. While I'm there, also convert usage() to
delphij [Fri, 13 May 2011 22:35:06 +0000 (22:35 +0000)]
Constify a few string variables.  While I'm there, also convert usage() to
use a prototype.

MFC after: 2 weeks

13 years agoMFC
attilio [Fri, 13 May 2011 20:58:48 +0000 (20:58 +0000)]
MFC

13 years agoFix a brain-o in platform_cpu_mask() by just specifying a possible
attilio [Fri, 13 May 2011 19:56:58 +0000 (19:56 +0000)]
Fix a brain-o in platform_cpu_mask() by just specifying a possible
cpuset_t to be copied, rather than return the array.
I can't rely anymore on this being a simple int/long object.

Reported by: art

13 years agoIf there's no package directory, don't try to install packages from
imp [Fri, 13 May 2011 19:40:02 +0000 (19:40 +0000)]
If there's no package directory, don't try to install packages from
it.  Instead, report that 0 packages are reported.

PR: misc/140436

13 years agoMove the ZERO_REGION_SIZE to a machine-dependent file, as on many
mdf [Fri, 13 May 2011 19:35:01 +0000 (19:35 +0000)]
Move the ZERO_REGION_SIZE to a machine-dependent file, as on many
architectures (i386, for example) the virtual memory space may be
constrained enough that 2MB is a large chunk.  Use 64K for arches
other than amd64 and ia64, with special handling for sparc64 due to
differing hardware.

Also commit the comment changes to kmem_init_zero_region() that I
missed due to not saving the file.  (Darn the unfamiliar development
environment).

Arch maintainers, please feel free to adjust ZERO_REGION_SIZE as you
see fit.

Requested by: alc
MFC after: 1 week
MFC with: r221853

13 years agoUsa a globally visible region of zeros for both /dev/zero and the md
mdf [Fri, 13 May 2011 18:48:00 +0000 (18:48 +0000)]
Usa a globally visible region of zeros for both /dev/zero and the md
device.  There are likely other kernel uses of "blob of zeros" than can
be converted.

Reviewed by: alc
MFC after: 1 week

13 years agoAdd ${} around variable dereference...
imp [Fri, 13 May 2011 18:46:20 +0000 (18:46 +0000)]
Add ${} around variable dereference...

Submitted by: garrett coopper

13 years agoImplement -f to inhibit copying s1 partition out for speed.
imp [Fri, 13 May 2011 18:36:24 +0000 (18:36 +0000)]
Implement -f to inhibit copying s1 partition out for speed.
(-z not taken from PR, it is no longer needed since we basically always
 do that now)

PR: misc/151698
Submitted by: lev@

13 years agoCopy symbolic links as files rather than recreating the links.
imp [Fri, 13 May 2011 18:28:25 +0000 (18:28 +0000)]
Copy symbolic links as files rather than recreating the links.

PR: misc/151697
Submitted by: lev@
MFC after: 2 weeks

13 years agoAdd the cpuset_t conversion for mips.
attilio [Fri, 13 May 2011 16:42:05 +0000 (16:42 +0000)]
Add the cpuset_t conversion for mips.

13 years agoFix the _long() rappresentation on mips by casting the long arguments
attilio [Fri, 13 May 2011 16:39:47 +0000 (16:39 +0000)]
Fix the _long() rappresentation on mips by casting the long arguments
to u_int for all the functions.

Reviewed by: art, imp

13 years agoUpdate sticky(7) cross references.
pluknet [Fri, 13 May 2011 16:29:57 +0000 (16:29 +0000)]
Update sticky(7) cross references.

PR: docs/124468
X-MFC with: r218998

13 years agoIn pmap_change_wiring(), use the right argument for pmap_modify_pv().
cognet [Fri, 13 May 2011 15:54:12 +0000 (15:54 +0000)]
In pmap_change_wiring(), use the right argument for pmap_modify_pv().
It only worked because the only consumer calls pmap_change_wiring() to remove
the wiring.

13 years agoNote that the _SWAP operation is supported for all list/queue types.
mdf [Fri, 13 May 2011 15:49:23 +0000 (15:49 +0000)]
Note that the _SWAP operation is supported for all list/queue types.
Also place STAILQ_REMOVE_HEAD in alphabetical order.  Lastly, document
the _SWAP macros.

PR: kern/143033
MFC after: 1 week

13 years agoWhen setting up pc_other_cpus for APs based on pc_allcpu clear pc_cpuid
marius [Fri, 13 May 2011 15:21:31 +0000 (15:21 +0000)]
When setting up pc_other_cpus for APs based on pc_allcpu clear pc_cpuid
in the former rather than the latter.
This gets this branch working on at least Jalapeno-class CPUs.

13 years agoMFC
attilio [Fri, 13 May 2011 15:20:57 +0000 (15:20 +0000)]
MFC

13 years agoAfter rewriting powerpc atomic we decided to commit at the constraint
attilio [Fri, 13 May 2011 15:09:35 +0000 (15:09 +0000)]
After rewriting powerpc atomic we decided to commit at the constraint
that for _ptr operations, when not used directly with uintptr_t, we
needed to manually cast.

Use the cast on the _ptr version, where it actually wasn't (please note
that i386 doesn't get it right, while amd64 doesn't seem to compile
cfi neither in LINT, that is why it doesn't fail).

Reported by: sbruno

13 years agoAllow direct children of PCI-ISA bridges to allocate resources from
brix [Fri, 13 May 2011 15:06:35 +0000 (15:06 +0000)]
Allow direct children of PCI-ISA bridges to allocate resources from
the parent PCI bus.

Heavily inspired by jhb@ and a similar implementation present in
sys/dev/pci/vga_pci.c.

Reviewed by: jhb
Approved by: jhb

13 years agoFix remaining bits that actually weren't converted by mistake.
attilio [Fri, 13 May 2011 14:57:20 +0000 (14:57 +0000)]
Fix remaining bits that actually weren't converted by mistake.

Reported by: sbruno

13 years agoOnly do open loop power control and temperature compensation
adrian [Fri, 13 May 2011 14:33:45 +0000 (14:33 +0000)]
Only do open loop power control and temperature compensation
for the AR9280 based NICs if it's actually enabled.

Some of the OLC code was erroneously called during setup
and calibration. This may have caused some incorrect behaviour.

13 years agoCorrectly use INOUT for the offset/len parameters to vop_allocate. As
mdf [Fri, 13 May 2011 14:29:28 +0000 (14:29 +0000)]
Correctly use INOUT for the offset/len parameters to vop_allocate.  As
far as I can tell this is for documentation only at the moment.

13 years agoRefactor Xen PV code to use new event timers subsystem. That uses one-shot
mav [Fri, 13 May 2011 12:39:37 +0000 (12:39 +0000)]
Refactor Xen PV code to use new event timers subsystem. That uses one-shot
Xen timer and time counter to provide one-shot and periodic time events.

On my tests this reduces idle interruts rate down to about 30Hz, and accor-
ding to Xen VM Manager reduces host CPU load by three times comparing to
the previous periodic 100Hz clock. Also now, when needed, it is possible to
increase HZ rate without useless CPU burning during idle periods.

Now only ia64 and some ARMs left not migrated to the new event timers.

13 years agoRemove duplicate code - add a function which calculates the ratesArray[]
adrian [Fri, 13 May 2011 10:36:38 +0000 (10:36 +0000)]
Remove duplicate code - add a function which calculates the ratesArray[]
table which contains the per-rate target TX power.

This code is shared between the v14 eeprom board setup (AR5416, AR9160,
AR9280) and will also be used by the upcoming Kite (AR9287) support.

13 years agoSome diversity changes relating to AR9285.
adrian [Fri, 13 May 2011 09:57:12 +0000 (09:57 +0000)]
Some diversity changes relating to AR9285.

* grab the main, alt and selected LNA config
* add some optional / disabled logging code
* add a check to reject packets with an invalid main rssi too,
  in case the alt is the active receive chain and main is -ve.

Note: The software-controlled combined diversity code is still disabled.

13 years agoFixed an embedded shell script.
ru [Fri, 13 May 2011 09:55:48 +0000 (09:55 +0000)]
Fixed an embedded shell script.

Reviewed by: sobomax

13 years agoMark all socket and file descriptors close-on-exec.
des [Fri, 13 May 2011 07:21:41 +0000 (07:21 +0000)]
Mark all socket and file descriptors close-on-exec.

PR: bin/151866
MFC after: 3 weeks

13 years agoUse a name instead of a magic number for kern_yield(9) when the priority
mdf [Fri, 13 May 2011 05:27:58 +0000 (05:27 +0000)]
Use a name instead of a magic number for kern_yield(9) when the priority
should not change.  Fetch the td_user_pri under the thread lock.  This
is probably not necessary but a magic number also seems preferable to
knowing the implementation details here.

Requested by: Jason Behmer < jason DOT behmer AT isilon DOT com >

13 years agoFix msleep() usage in Xen balloon driver to not wake up on every HZ tick.
mav [Fri, 13 May 2011 03:40:16 +0000 (03:40 +0000)]
Fix msleep() usage in Xen balloon driver to not wake up on every HZ tick.

13 years ago- Use bus_describe_intr() to describe interrupt usage.
davidch [Thu, 12 May 2011 23:26:53 +0000 (23:26 +0000)]
- Use bus_describe_intr() to describe interrupt usage.
- Use bus_bind_intr() to bind interrupt to a CPU when RSS/TSS is used.
- Use M_DONTWAIT for RSS/TSS buffer allocation.
- Add statistic to track max DRBR queue depth.
- Fix problem in bxe_change_mtu() which referenced the old MTU size
  in a debug print statement.

MFC after: Two weeks

13 years agoFix pSeries/PAPR boot CD construction after actual testing. It seems that
nwhitehorn [Thu, 12 May 2011 22:31:13 +0000 (22:31 +0000)]
Fix pSeries/PAPR boot CD construction after actual testing. It seems that
IBM SLOF is (a) case-sensitive and (b) accepts only booting files in
certain subdirectories of the CD.

13 years ago- Don't try to build NWFS support module if NCP/IPX is disabled in the build.
stas [Thu, 12 May 2011 22:21:39 +0000 (22:21 +0000)]
- Don't try to build NWFS support module if NCP/IPX is disabled in the build.
- Rename ZFS definition to LIBPROCSTAT_ZFS to be consistent with NWFS and to
  prevent possible collisions.

Reported by: many

13 years agoBuilds cleanly at the default WARNS level (WARNS=6).
des [Thu, 12 May 2011 21:30:46 +0000 (21:30 +0000)]
Builds cleanly at the default WARNS level (WARNS=6).

13 years agoIncrease WARNS to 4.
des [Thu, 12 May 2011 21:26:42 +0000 (21:26 +0000)]
Increase WARNS to 4.

13 years agoMechanical whitespace cleanup.
des [Thu, 12 May 2011 21:18:55 +0000 (21:18 +0000)]
Mechanical whitespace cleanup.

13 years agoIncrease WARNS to 3.
des [Thu, 12 May 2011 21:12:24 +0000 (21:12 +0000)]
Increase WARNS to 3.

13 years agouintmax_t is a better type here...
imp [Thu, 12 May 2011 19:10:00 +0000 (19:10 +0000)]
uintmax_t is a better type here...

Submitted by: avg@

13 years agoAdd initial BCM5719 support. TSO and jumbo frame was intentionally
yongari [Thu, 12 May 2011 17:15:57 +0000 (17:15 +0000)]
Add initial BCM5719 support. TSO and jumbo frame was intentionally
disabled for BCM5719 A0 revision due to known hardware errata.
Many thanks to Broadcom for continuing support of FreeBSD.

Submitted by: Geans Pin at Broadcom

13 years agoExplicitly clear 1000baseT control register for F1 PHY used in
yongari [Thu, 12 May 2011 17:11:31 +0000 (17:11 +0000)]
Explicitly clear 1000baseT control register for F1 PHY used in
AR8132 FastEthernet controller. The PHY has no ability to
establish a gigabit link. Previously only link parters which
support down-shifting was able to establish link.
This change should fix a long standing link establishment issue of
AR8132.

PR: kern/156935
MFC after: 1 week

13 years agoFix printf int mismatch on 32-bit architectures...
imp [Thu, 12 May 2011 17:02:56 +0000 (17:02 +0000)]
Fix printf int mismatch on 32-bit architectures...

13 years agoAdd (somewhat speculative) bootable CD support for PAPR/pSeries-type
nwhitehorn [Thu, 12 May 2011 15:03:17 +0000 (15:03 +0000)]
Add (somewhat speculative) bootable CD support for PAPR/pSeries-type
systems. In principle, FreeBSD should run on at least some of these
already, and support for the remainder will hopefully show up eventually,
so add this while I'm thinking about it.

13 years agoRemove some hacks to handle strange behavior of LXT 970 PHYs now better
nwhitehorn [Thu, 12 May 2011 14:27:28 +0000 (14:27 +0000)]
Remove some hacks to handle strange behavior of LXT 970 PHYs now better
handled in miibus after r221812. Thanks to marius@ for piecing this
together!

13 years agoSome PHYs like the Level One LXT970 optionally can default to isolation
marius [Thu, 12 May 2011 14:16:07 +0000 (14:16 +0000)]
Some PHYs like the Level One LXT970 optionally can default to isolation
after rest, in which case we may need to deisolate it.
Tested by: nwhitehorn

MFC after 1 week

13 years agoNow that the devices with functioning ps-poll hardware support have
adrian [Thu, 12 May 2011 14:03:29 +0000 (14:03 +0000)]
Now that the devices with functioning ps-poll hardware support have
been enumerated (merlin and later), flick this on.

13 years agoMFC
attilio [Thu, 12 May 2011 14:01:40 +0000 (14:01 +0000)]
MFC

13 years agofix build on 32-bit platforms for r221803
avg [Thu, 12 May 2011 12:18:01 +0000 (12:18 +0000)]
fix build on 32-bit platforms for r221803

Casting a pointer to a wide integer is probably not that bad, but I am
still guilty of not testing this.

Pointyhat to: avg
MFC after: 1 week
X-MFC with: r221803

13 years ago- Do no try to drop a NULL filedesc pointer.
stas [Thu, 12 May 2011 10:56:33 +0000 (10:56 +0000)]
- Do no try to drop a NULL filedesc pointer.

13 years ago- Commit work from libprocstat project. These patches add support for runtime
stas [Thu, 12 May 2011 10:11:39 +0000 (10:11 +0000)]
- Commit work from libprocstat project.  These patches add support for runtime
  file and processes information retrieval from the running kernel via sysctl
  in the form of new library, libprocstat.  The library also supports KVM backend
  for analyzing memory crash dumps.  Both procstat(1) and fstat(1) utilities have
  been modified to take advantage of the library (as the bonus point the fstat(1)
  utility no longer need superuser privileges to operate), and the procstat(1)
  utility is now able to display information from memory dumps as well.

  The newly introduced fuser(1) utility also uses this library and able to operate
  via sysctl and kvm backends.

  The library is by no means complete (e.g. KVM backend is missing vnode name
  resolution routines, and there're no manpages for the library itself) so I
  plan to improve it further.  I'm commiting it so it will get wider exposure
  and review.

  We won't be able to MFC this work as it relies on changes in HEAD, which
  was introduced some time ago, that break kernel ABI.  OTOH we may be able
  to merge the library with KVM backend if we really need it there.

Discussed with: rwatson