]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
12 years agoRework sample code which sets SSH_AUTH_SOCK to the user's ssh-agent
eadler [Tue, 15 May 2012 20:15:17 +0000 (20:15 +0000)]
Rework sample code which sets SSH_AUTH_SOCK to the user's ssh-agent
socket path if running

PR: 167932
Submitted by: madpilot (pr)
Submitted by: wblock (final patch)
Reviewed by: If someone else reviewed your modification.
Approved by: cperciva (implicit)
MFC after: 3 days

12 years agozpool_find_import_impl: another /dev/dsk -> /dev fix
avg [Tue, 15 May 2012 17:11:00 +0000 (17:11 +0000)]
zpool_find_import_impl: another /dev/dsk -> /dev fix

This seems to fix zdb -e behavior.

PR: bin/155104
Submitted by: swell.k@gmail.com
MFC after: 2 weeks

12 years agozpool_do_import: use /dev instead of /dev/dsk as a default
avg [Tue, 15 May 2012 17:07:29 +0000 (17:07 +0000)]
zpool_do_import: use /dev instead of /dev/dsk as a default

This affects behavior of zpool import without -d option.

Reported by: Bruce Cran <bruce@cran.org.uk>
Submitted by: Fabian Keil <freebsd-listen@fabiankeil.de>
MFC after: 2 weeks

12 years agoSwitch to a standard 2 clause BSD license (from bsd-style-copyright).
bz [Tue, 15 May 2012 13:23:44 +0000 (13:23 +0000)]
Switch to a standard 2 clause BSD license (from bsd-style-copyright).

Approved by: Myricom Inc. (gallatin)
Approved by: Intel Corporation (jfv)

12 years agoFix typo in function name SDT_PROBE4 and unbreak 4BSD UP.
pluknet [Tue, 15 May 2012 10:58:17 +0000 (10:58 +0000)]
Fix typo in function name SDT_PROBE4 and unbreak 4BSD UP.

12 years agoDo not call bremfree for managed buffers.
gber [Tue, 15 May 2012 09:55:15 +0000 (09:55 +0000)]
Do not call bremfree for managed buffers.

Calling bremfree for these buffers results in panic:
"bremfree: buffer %p not on a queue."

Approved by: kib

12 years agoHandle non-xretry errors the same as xretry errors for now.
adrian [Tue, 15 May 2012 04:55:15 +0000 (04:55 +0000)]
Handle non-xretry errors the same as xretry errors for now.

Although I _should_ handle the other errors in various ways (specifically
errors like FILT), treating them as having transmitted successfully
is completely wrong.  Here, they'd be counted as successful and the BAW
would be advanced.. but the RX side wouldn't have received them.

The specific errors I've been seeing here are HAL_TXERR_FILT.

This patch does fix the issue - I've tested it using -i 0.001 pings
(enough to start aggregation) and now the behaviour is correct:

* The RX side never sees a "moved window" error, and
* The TX side sends BARs as needed, with the RX side correctly handling
  them.

PR: kern/167902

12 years agoImplement the DTrace sched provider. This implementation aims to be
rstone [Tue, 15 May 2012 01:30:25 +0000 (01:30 +0000)]
Implement the DTrace sched provider.  This implementation aims to be
compatible with the sched provider implemented by Solaris and its open-
source derivatives.  Full documentation of the sched provider can be found
on Oracle's DTrace wiki pages.

Note that for compatibility with scripts originally written for Solaris,
serveral probes are defined that will never fire.  These probes are defined
to fire when Solaris-specific features perform certain actions.  As these
features are not present in FreeBSD, the probes can never fire.

Also, I have added a two probes that are not defined in Solaris, lend-pri
and load-change.  These probes have been added to make it possible to
collect schedgraph data with DTrace.

Finally, a few probes are defined in Solaris to take a cpuinfo_t *
argument.  As it was not immediately clear to me how to translate that to
FreeBSD, currently those probes are passed NULL in place of a cpuinfo_t *.

Sponsored by: Sandvine Incorporated
MFC after: 2 weeks

12 years agoMove sh to the correct path for FreeBSD.
gnn [Mon, 14 May 2012 22:02:01 +0000 (22:02 +0000)]
Move sh to the correct path for FreeBSD.

12 years agoMove sh to the correct location for FreeBS (/bin/sh)
gnn [Mon, 14 May 2012 22:00:24 +0000 (22:00 +0000)]
Move sh to the correct location for FreeBS (/bin/sh)

12 years agoFix execsnoop by changing exece and exec to be FreeBSD's execve.
gnn [Mon, 14 May 2012 21:59:47 +0000 (21:59 +0000)]
Fix execsnoop by changing exece and exec to be FreeBSD's execve.

Reference sh in the correct location (/bin/sh)

12 years agoFix opensnoop for FreeBSD by removing probes with 64 at the end as
gnn [Mon, 14 May 2012 21:58:22 +0000 (21:58 +0000)]
Fix opensnoop for FreeBSD by removing probes with 64 at the end as
these are unnecessary.

Reference sh in the correct location (/bin/sh)

12 years agoStyle fixes.
jpaetzel [Mon, 14 May 2012 18:06:51 +0000 (18:06 +0000)]
Style fixes.

Submitted by: kmoore
Obtained from: PC-BSD
MFC after: 3 days
Sponsored by: iXsystems

12 years agoSet the MBR partition to active when doing a full disk MBR.
jpaetzel [Mon, 14 May 2012 18:03:59 +0000 (18:03 +0000)]
Set the MBR partition to active when doing a full disk MBR.

Submitted by: kmoore
Obtained from: PC-BSD
Sponsored by: iXsystems
MFC after: 3 days

12 years agoMove reset of USB mouse parameters from the USB mouse attach to
hselasky [Mon, 14 May 2012 17:00:32 +0000 (17:00 +0000)]
Move reset of USB mouse parameters from the USB mouse attach to
the USB mouse device open. Protect against multi character
device open. Some other nits.

MFC after: 1 week

12 years agomdoc: Avoid playing tricks with Ns: If Nm is present in the SYNOPSIS section,
joel [Mon, 14 May 2012 16:25:17 +0000 (16:25 +0000)]
mdoc: Avoid playing tricks with Ns: If Nm is present in the SYNOPSIS section,
it will be output on its own line. Ns cancels this effect however. This change
is also consistent with the rest of our manual pages.

12 years agoFix an mdoc(7) formatting nit.
gjb [Mon, 14 May 2012 14:33:08 +0000 (14:33 +0000)]
Fix an mdoc(7) formatting nit.

12 years ago- Eliminate initializations if global variables. Compilers are not
gabor [Mon, 14 May 2012 10:06:49 +0000 (10:06 +0000)]
- Eliminate initializations if global variables.  Compilers are not
  required to optimize these so it may result in larger binary size.

Pointed out by: kib

12 years ago- Update catalogs
gabor [Mon, 14 May 2012 09:55:23 +0000 (09:55 +0000)]
- Update catalogs

12 years ago- Fix build with clang
gabor [Mon, 14 May 2012 09:53:54 +0000 (09:53 +0000)]
- Fix build with clang

12 years agoAdd 3 missing files that was generated but not included with the initial
delphij [Mon, 14 May 2012 09:32:10 +0000 (09:32 +0000)]
Add 3 missing files that was generated but not included with the initial
commit.

Reported by: buganini via IRC
MFC after: 2 weeks

12 years agoMerge from vendor branch: update libpcap to 1.2.1.
delphij [Mon, 14 May 2012 05:12:56 +0000 (05:12 +0000)]
Merge from vendor branch: update libpcap to 1.2.1.

MFC after: 2 weeks

12 years agoSync DLTs with the latest pcap version.
delphij [Mon, 14 May 2012 05:10:41 +0000 (05:10 +0000)]
Sync DLTs with the latest pcap version.

MFC after: 2 weeks

12 years agoAdd missing period at the end of the error message
eadler [Sun, 13 May 2012 23:27:06 +0000 (23:27 +0000)]
Add missing period at the end of the error message

Submitted by: pjd
Approved by: cperciva (implicit)
MFC after: 3 days
X-MFC-With: r235201

12 years agoSupport SCTP_REMOTE_ERROR notification.
tuexen [Sun, 13 May 2012 22:27:54 +0000 (22:27 +0000)]
Support SCTP_REMOTE_ERROR notification.

MFC after: 3 days

12 years agoProvide in the SCTP_SEND_FAILED and SCTP_SEND_FAILED_EVENT notifications
tuexen [Sun, 13 May 2012 19:32:49 +0000 (19:32 +0000)]
Provide in the SCTP_SEND_FAILED and SCTP_SEND_FAILED_EVENT notifications
the correct ssf_error or ssfe_error as required by RFC 6458.

MFC after: 3 days

12 years agoMissed to commit this in r235414.
tuexen [Sun, 13 May 2012 19:25:21 +0000 (19:25 +0000)]
Missed to commit this in r235414.

MFC after: 3 days

12 years agoProvide the error code in SCTP_PEER_ADDR_CHANGE notifications as
tuexen [Sun, 13 May 2012 17:36:04 +0000 (17:36 +0000)]
Provide the error code in SCTP_PEER_ADDR_CHANGE notifications as
specified in RFC 6458.

MFC after: 3 days

12 years agoRemove unused constants.
tuexen [Sun, 13 May 2012 17:17:13 +0000 (17:17 +0000)]
Remove unused constants.

MFC after: 3 days

12 years agoUse ECONNABORTED in cases where the ABORT was sent to the peer.
tuexen [Sun, 13 May 2012 16:56:16 +0000 (16:56 +0000)]
Use ECONNABORTED in cases where the ABORT was sent to the peer.

MFC after: 3 days

12 years agoEnsure the user can read COMM_LOST notifications on 1-to-1 style sockets.
tuexen [Sun, 13 May 2012 16:07:53 +0000 (16:07 +0000)]
Ensure the user can read COMM_LOST notifications on 1-to-1 style sockets.

MFC after: 3 days

12 years agomdoc: avoid unterminated quoted strings.
joel [Sun, 13 May 2012 14:28:00 +0000 (14:28 +0000)]
mdoc: avoid unterminated quoted strings.

12 years agoMinor mdoc nits.
joel [Sun, 13 May 2012 14:16:04 +0000 (14:16 +0000)]
Minor mdoc nits.

12 years agolibpc98: unbreak after zfs-related changes in libi386 code
avg [Sun, 13 May 2012 11:34:05 +0000 (11:34 +0000)]
libpc98: unbreak after zfs-related changes in libi386 code

Pointyhat to: avg
MFC after: 1 month

12 years agozfs boot: cleanup remnants of temporary compat code
avg [Sun, 13 May 2012 10:54:43 +0000 (10:54 +0000)]
zfs boot: cleanup remnants of temporary compat code

MFC after: 1 month

12 years agozfsboottest: update for the recent zfs boot code changes
avg [Sun, 13 May 2012 09:27:08 +0000 (09:27 +0000)]
zfsboottest: update for the recent zfs boot code changes

MFC after: 1 month

12 years agoi386 bootinfo: re-arrange EFI fields for natural alignment and packing
avg [Sun, 13 May 2012 09:25:39 +0000 (09:25 +0000)]
i386 bootinfo: re-arrange EFI fields for natural alignment and packing

Suggested by: bde
MFC after: 2 weeks

12 years agozfs boot code: mark spa_t arguments as const where they are used as such
avg [Sun, 13 May 2012 09:22:18 +0000 (09:22 +0000)]
zfs boot code: mark spa_t arguments as const where they are used as such

MFC after: 1 month

12 years agoRegenerate after r235388.
marius [Sun, 13 May 2012 05:09:36 +0000 (05:09 +0000)]
Regenerate after r235388.

12 years ago- Get rid of debugging support in order to get rid of the V8-specific C
marius [Sun, 13 May 2012 05:08:13 +0000 (05:08 +0000)]
- Get rid of debugging support in order to get rid of the V8-specific C
  compiler frame size used there so this whole thing is V8/V9-agnostic.
- Use 32-bit function alignment as GCC does when using UltraSPARC I or
  higher optimizations.
- Don't waste delay slots when possible.

Unfortunately, this still doesn't make libcompiler_rt a viable replacement
for libgcc on sparc64 though as once installed instead, buildworld times
increase by nearly 60% (which isn't related to these assembler functions).

12 years agoFix config_lazy_lock so that thread caching isn't used for
jasone [Sun, 13 May 2012 01:33:28 +0000 (01:33 +0000)]
Fix config_lazy_lock so that thread caching isn't used for
single-threaded applications.

12 years agoFix two cases in the new NFS server where a tsleep() is
rmacklem [Sat, 12 May 2012 22:20:55 +0000 (22:20 +0000)]
Fix two cases in the new NFS server where a tsleep() is
used, when the code should actually protect the tested
variable with a mutex. Since the tsleep()s had a 10sec
timeout, the race would have only delayed the allocation
of a new clientid for a client. The sleeps will also
rarely occur, since having a callback in progress when
a client acquires a new clientid, is unlikely.
in practice, since having a callback in progress when
a fresh clientid is being acquired by a client is unlikely.

MFC after: 1 month

12 years agoImport dtracetoolkit into cddl/contrib
gnn [Sat, 12 May 2012 21:25:48 +0000 (21:25 +0000)]
Import dtracetoolkit into cddl/contrib

12 years agoRevert previous failed cp.
gnn [Sat, 12 May 2012 21:23:31 +0000 (21:23 +0000)]
Revert previous failed cp.

12 years agoImport dtracetoolkit into cddl/contrib
gnn [Sat, 12 May 2012 21:15:21 +0000 (21:15 +0000)]
Import dtracetoolkit into cddl/contrib

12 years agoSetup the CPU port and broadcast map on the AR7240, rather than
adrian [Sat, 12 May 2012 21:08:07 +0000 (21:08 +0000)]
Setup the CPU port and broadcast map on the AR7240, rather than
depending upon the bootloader initialising it.

The aim is to eventually support a full switch set and reinitialisation
rather than relying on a consistent bootloader setup.

Remove the port flood config from arswitch.c, it's not yet used and
it's totally incorrect.

Whilst I'm here, also add in a comment describing why the full switch
reset is disabled.

Obtained from: Linux (OpenWRT) - Values

12 years agoRemove leading .Ns macro to silence mdoc(7) warning.
gjb [Sat, 12 May 2012 20:53:00 +0000 (20:53 +0000)]
Remove leading .Ns macro to silence mdoc(7) warning.

12 years agoAdd new pager type, OBJT_MGTDEVICE. It provides the device pager
kib [Sat, 12 May 2012 20:49:58 +0000 (20:49 +0000)]
Add new pager type, OBJT_MGTDEVICE. It provides the device pager
which carries fictitous managed pages. In particular, the consumers of
the new object type can remove all mappings of the device page with
pmap_remove_all().

The range of physical addresses used for fake page allocation shall be
registered with vm_phys_fictitious_reg_range() interface to allow the
PHYS_TO_VM_PAGE() to work in pmap.

Most likely, only i386 and amd64 pmaps can handle fictitious managed
pages right now.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agomdoc(7) markup cleanup.
gjb [Sat, 12 May 2012 20:46:19 +0000 (20:46 +0000)]
mdoc(7) markup cleanup.

12 years agoAdd a facility to register a range of physical addresses to be used
kib [Sat, 12 May 2012 20:42:56 +0000 (20:42 +0000)]
Add a facility to register a range of physical addresses to be used
for allocation of fictitious pages, for which PHYS_TO_VM_PAGE()
returns proper fictitious vm_page_t. The range should be de-registered
after consumer stopped using it.

De-inline the PHYS_TO_VM_PAGE() since it now carries code to iterate
over registered ranges.

A hash container might be developed instead of range registration
interface, and fake pages could be put automatically into the hash,
were PHYS_TO_VM_PAGE() could look them up later. This should be
considered before the MFC of the commit is done.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agoFix mdoc(7) markup.
gjb [Sat, 12 May 2012 20:41:40 +0000 (20:41 +0000)]
Fix mdoc(7) markup.

12 years ago.. oops, include setting the MTU.
adrian [Sat, 12 May 2012 20:40:13 +0000 (20:40 +0000)]
.. oops, include setting the MTU.

12 years agoDocument what the flood register setting does.
adrian [Sat, 12 May 2012 20:39:41 +0000 (20:39 +0000)]
Document what the flood register setting does.

12 years agoAdd the remaining scripts from the DTraceToolkit, version 0.99, to the
gnn [Sat, 12 May 2012 20:38:18 +0000 (20:38 +0000)]
Add the remaining scripts from the DTraceToolkit, version 0.99, to the
vendor tree.

http://www.brendangregg.com/dtrace.html#DTraceToolkit

12 years ago* Add in the AR7240 global control field for setting the maximum frame
adrian [Sat, 12 May 2012 20:36:12 +0000 (20:36 +0000)]
* Add in the AR7240 global control field for setting the maximum frame
  size for the AR7240.

* Include SM/MS macros, thanks to ath_hal(4).

* This field is for normal packets, VLAN and other headers are added to
  this by the switch device.

* Set the MTU to 1536, to match what is done in Linux.  Use the SM
  macro to write this field.

Obtained from: Atheros (AR7240 datasheet), Linux OpenWRT (MTU default)

12 years agoSplit the code from vm_page_getfake() to initialize the fake page struct
kib [Sat, 12 May 2012 20:34:22 +0000 (20:34 +0000)]
Split the code from vm_page_getfake() to initialize the fake page struct
vm_page into new interface vm_page_initfake(). Handle the case of fake
page re-initialization with changed memattr.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agoAssert that the page passed to vm_page_putfake() is unmanaged.
kib [Sat, 12 May 2012 20:27:51 +0000 (20:27 +0000)]
Assert that the page passed to vm_page_putfake() is unmanaged.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agosparc64/zfs boot: take advantage of new libzfsboot capabilities
avg [Sat, 12 May 2012 20:27:33 +0000 (20:27 +0000)]
sparc64/zfs boot: take advantage of new libzfsboot capabilities

Also drop the now unneeded compatibility shims.

Tested by: marius
MFC after: 1 month

12 years agoMinor mdoc nits.
joel [Sat, 12 May 2012 20:27:13 +0000 (20:27 +0000)]
Minor mdoc nits.

12 years agoAssert that fictitious or unmanaged pages do not appear on
kib [Sat, 12 May 2012 20:24:46 +0000 (20:24 +0000)]
Assert that fictitious or unmanaged pages do not appear on
active/inactive lists.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agozfs boot code: use %j and uintmax_t instead %ll and uint64_t in printfs
avg [Sat, 12 May 2012 20:23:30 +0000 (20:23 +0000)]
zfs boot code: use %j and uintmax_t instead %ll and uint64_t in printfs

This is to silence warnings that result from different definitions of
uint64_t on different architectures, specifically i386 and sparc64.

MFC after: 1 month

12 years agoProvide in the association change notification the received ABORT chunk
tuexen [Sat, 12 May 2012 20:11:35 +0000 (20:11 +0000)]
Provide in the association change notification the received ABORT chunk
if case of SCTP_COMM_LOST or SCTP_CANT_STR_ASSOC as required by RFC 6458.

MFC after: 3 days

12 years agoCommit the change forgotten in r235356.
kib [Sat, 12 May 2012 20:10:18 +0000 (20:10 +0000)]
Commit the change forgotten in r235356.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agoSet formatting width and offset in DB_COMMAND(9).
gjb [Sat, 12 May 2012 20:08:16 +0000 (20:08 +0000)]
Set formatting width and offset in DB_COMMAND(9).

12 years agoFix an mdoc(7) nit.
gjb [Sat, 12 May 2012 20:06:17 +0000 (20:06 +0000)]
Fix an mdoc(7) nit.

12 years agoMake the vm_page_array_size long. Remove redundand zero initialization
kib [Sat, 12 May 2012 20:03:06 +0000 (20:03 +0000)]
Make the vm_page_array_size long. Remove redundand zero initialization
for vm_page_array_size and nearby variablees.

Sponsored by: The FreeBSD Foundation
Reviewed by: alc
MFC after: 1 month

12 years agoRemove end of line whitespace.
joel [Sat, 12 May 2012 19:59:37 +0000 (19:59 +0000)]
Remove end of line whitespace.

12 years agoMinor mdoc nit.
joel [Sat, 12 May 2012 19:53:44 +0000 (19:53 +0000)]
Minor mdoc nit.

12 years agoRemove duplicate config bits.
adrian [Sat, 12 May 2012 19:43:22 +0000 (19:43 +0000)]
Remove duplicate config bits.

Submitted by: juli

12 years agoAdd glue/support for the SAM9XE512-based Ethernut 5 boards. Currently,
marius [Sat, 12 May 2012 18:11:26 +0000 (18:11 +0000)]
Add glue/support for the SAM9XE512-based Ethernut 5 boards. Currently,
all integrated and on-board peripherals except the DataFlash (at91_spi(4)
and at45d(4) still need to be unb0rken) and NAND Flash (missing NAND
framework) are working.
AFAICT, this makes FreeBSD the first operating system besides Nut/OS
supporting Ethernut 5 out of tree.

12 years agoDocuement requirement to alter some sysctls when using igb(4) with
gjb [Sat, 12 May 2012 18:10:27 +0000 (18:10 +0000)]
Docuement requirement to alter some sysctls when using igb(4) with
jumbo frames.

PR: 153738
Submitted by: Tom Judge (tom!tomjudge.com)
Discussed with: jfv (in part, a long time ago...)
MFC after: 1 week

12 years agoRevert this - I disabled it whilst hwpmc is/was broken.
adrian [Sat, 12 May 2012 17:42:22 +0000 (17:42 +0000)]
Revert this - I disabled it whilst hwpmc is/was broken.

12 years agoFlip on WN1043ND switch PHY support.
adrian [Sat, 12 May 2012 17:41:42 +0000 (17:41 +0000)]
Flip on WN1043ND switch PHY support.

* Add the i2c bitbang bus;
* Add the etherswitch/rtl8366rb drivers;
* "fix" the USB GPIO configuration so USB actually works.

Submitted by: Stefan Bethke <stb@lassitu.de>

12 years agomdoc: use Po and Pc macros instead of parens. Also avoid starting a line
joel [Sat, 12 May 2012 17:14:55 +0000 (17:14 +0000)]
mdoc: use Po and Pc macros instead of parens. Also avoid starting a line
with Ns.

12 years agoadd a zfs spa_t change missed in r235329
avg [Sat, 12 May 2012 16:36:54 +0000 (16:36 +0000)]
add a zfs spa_t change missed in r235329

sys/cddl/boot is obviously not under sys/boot...

Pointed out by: Jan Beich <jbeich@tormail.org>
Pointyhat to: avg
MFC after: 1 month

12 years agoRegenerate src.conf(5) for mdoc(7) and typo corrections.
gjb [Sat, 12 May 2012 16:12:36 +0000 (16:12 +0000)]
Regenerate src.conf(5) for mdoc(7) and typo corrections.

PR: 167804
Submitted by: Nobuyuki Koganemaru (kogane!jp.freebsd.org)
MFC after: 3 days

12 years agoRemove tab from kernel configuration option. This is consistent with the rest
joel [Sat, 12 May 2012 16:08:05 +0000 (16:08 +0000)]
Remove tab from kernel configuration option. This is consistent with the rest
of our manual pages.

12 years agoAdd opt_wlan.h includes, so IEEE80211_* configuration changes are
adrian [Sat, 12 May 2012 15:11:53 +0000 (15:11 +0000)]
Add opt_wlan.h includes, so IEEE80211_* configuration changes are
correctly picked up.

Noticed by: Justin Hibbits, whilst debugging @ BSDCan

12 years agoGeneral mdoc(7) and typo fixes.
gjb [Sat, 12 May 2012 15:08:22 +0000 (15:08 +0000)]
General mdoc(7) and typo fixes.

PR: 167804
Submitted by: Nobuyuki Koganemaru (kogane!jp.freebsd.org)
MFC after: 3 days

12 years agomdoc: remove redundant paragraph macro.
joel [Sat, 12 May 2012 14:46:49 +0000 (14:46 +0000)]
mdoc: remove redundant paragraph macro.

12 years agomdoc: remove redundant Pp and end a display block with Ed.
joel [Sat, 12 May 2012 14:43:52 +0000 (14:43 +0000)]
mdoc: remove redundant Pp and end a display block with Ed.

12 years agoConvert the if_vr(4) driver model to the interrupt filter model and use
rpaulo [Sat, 12 May 2012 14:37:25 +0000 (14:37 +0000)]
Convert the if_vr(4) driver model to the interrupt filter model and use
a taskqueue.

This gives a 16% performance improvement under high load on slow systems,
especially when vr shares an interrupt with another device, which is
common with the Alix x86 boards.
Contrary to the other devices, I left the interrupt processing for loop
in because there was no significant difference in performance and this
should avoid enqueuing more taskqueues unnecessarily.
We also decided to move the vr_start_locked() call inside the for loop
because we found out that it helps performance since TCP ACKs now have a
chance to go out quicker.

Reviewed by: yongari (older version, same idea)
Discussed with: yongari, jhb

12 years agoAdd two functions xpt_batch_start() and xpt_batch_done() to the CAM SIM KPI
mav [Sat, 12 May 2012 13:55:36 +0000 (13:55 +0000)]
Add two functions xpt_batch_start() and xpt_batch_done() to the CAM SIM KPI
to allow drivers to handle request completion directly without passing
them to the CAM SWI thread removing extra context switch.
Modify all ATA/SATA drivers to use them.

Reviewed by: gibbs, ken
MFC after: 2 weeks

12 years agoPR# 165923 reported intermittent write failures for dirty
rmacklem [Sat, 12 May 2012 12:02:51 +0000 (12:02 +0000)]
PR# 165923 reported intermittent write failures for dirty
memory mapped pages being written back on an NFS mount.
Since any thread can call VOP_PUTPAGES() to write back a
dirty page, the credentials of that thread may not have
write access to the file on an NFS server. (Often the uid
is 0, which may be mapped to "nobody" in the NFS server.)
Although there is no completely correct fix for this
(NFS servers check access on every write RPC instead of at
open/mmap time), this patch avoids the common cases by
holding onto a credential that recently opened the file
for writing and uses that credential for the write RPCs
being done by VOP_PUTPAGES() for both NFS clients.

Tested by: Joel Ray Holveck (joelh at juniper.net)
PR: kern/165923
Reviewed by: kib
MFC after: 2 weeks

12 years agoAdd more obsolete files.
antoine [Sat, 12 May 2012 10:53:49 +0000 (10:53 +0000)]
Add more obsolete files.

12 years agozfs boot: try to set vfs.root.mountfrom from currdev as a fallback
avg [Sat, 12 May 2012 09:07:41 +0000 (09:07 +0000)]
zfs boot: try to set vfs.root.mountfrom from currdev as a fallback

This way with the new zfsloader there is no need to explicitly set zfs
root filesystem either via vfs.root.mountfrom or fstab.
It should be automatically picked up from currdev which is by default
is set from bootfs.

Tested by: Florian Wagner <florian@wagner-flo.net> (x86)
MFC after: 1 month

12 years agozfsboot/zfsloader: support accessing filesystems within a pool
avg [Sat, 12 May 2012 09:03:30 +0000 (09:03 +0000)]
zfsboot/zfsloader: support accessing filesystems within a pool

In zfs loader zfs device name format now is "zfs:pool/fs",
fully qualified file path is "zfs:pool/fs:/path/to/file"
loader allows accessing files from various pools and filesystems as well
as changing currdev to a different pool/filesystem.

zfsboot accepts kernel/loader name in a format pool:fs:path/to/file or,
as before, pool:path/to/file; in the latter case a default filesystem
is used (pool root or bootfs).  zfsboot passes guids of the selected
pool and dataset to zfsloader to be used as its defaults.

zfs support should be architecture independent and is provided
in a separate library, but architectures wishing to use this zfs support
still have to provide some glue code and their devdesc should be
compatible with zfs_devdesc.
arch_zfs_probe method is used to discover all disk devices that may
be part of ZFS pool(s).

libi386 unconditionally includes zfs support, but some zfs-specific
functions are stubbed out as weak symbols.  The strong definitions
are provided in libzfsboot.
This change mean that the size of i386_devspec becomes larger
to match zfs_devspec.

Backward-compatibility shims are provided for recently added sparc64
zfs boot support.  Currently that architecture still works the old
way and does not support the new features.

TODO:
- clear up pool root filesystem vs pool bootfs filesystem distinction
- update sparc64 support
- set vfs.root.mountfrom based on currdev (for zfs)

Mid-future TODO:
- loader sub-menu for selecting alternative boot environment

Distant future TODO:
- support accessing snapshots, using a snapshot as readonly root

Reviewed by: marius (sparc64),
Gavin Mu <gavin.mu@gmail.com> (sparc64)
Tested by: Florian Wagner <florian@wagner-flo.net> (x86),
marius (sparc64)
No objections: fs@, hackers@
MFC after: 1 month

12 years agomdoc: end list context with El.
joel [Sat, 12 May 2012 07:52:45 +0000 (07:52 +0000)]
mdoc: end list context with El.

12 years agomdoc: list items should be specified with the It macro.
joel [Sat, 12 May 2012 07:01:21 +0000 (07:01 +0000)]
mdoc: list items should be specified with the It macro.

12 years ago* Remove the AR7240 register defines and reuse the AR8x16 defines.
adrian [Sat, 12 May 2012 06:24:21 +0000 (06:24 +0000)]
* Remove the AR7240 register defines and reuse the AR8x16 defines.
* Include a new register define to represent "disable port mirroring
  to CPU port".

Obtained from: Patrick Kelsey <kelsey@ieee.org>

12 years agoNow that there's a hint for it, add a "I'm an AR7240 switch!" hint.
adrian [Sat, 12 May 2012 05:27:14 +0000 (05:27 +0000)]
Now that there's a hint for it, add a "I'm an AR7240 switch!" hint.

12 years agoFurther arswitch work:
adrian [Sat, 12 May 2012 05:26:49 +0000 (05:26 +0000)]
Further arswitch work:

* Add in the AR724x support.  It probes the same as an AR8216/AR8316, so
  just add in a hint to force the probe success rather than auto-detecting
  it.

* Add in the missing entries from conf/files, lacking in the previous
  commit.

The register values and CPU port / mirror port initialisation value was
obtained from Linux OpenWRT ag71xx_ar7240.c.

The DELAY(1000) to let things settle is my local workaround.  For some
reason, PHY4 doesn't seem to probe very reliably without it.  It's quite
possible that we're missing some MDIO bus initialisation code in if_arge
for the AR724x case.  As I dislike DELAY() workarounds in general, it's
definitely worth trying to figure out why this is the case.

Tested on: AP93 (AR7240) reference design

Obtained from: Linux OpenWRT

12 years agoImport jemalloc 3.0.0. This fixes memory zeroing bugs that manifested as
jasone [Sat, 12 May 2012 05:26:49 +0000 (05:26 +0000)]
Import jemalloc 3.0.0.  This fixes memory zeroing bugs that manifested as
jemalloc assertion failures for debug builds, or as calloc() sometimes
returning non-zeroed memory for production builds.

12 years agoPointy hat to sleep deprived committer.
sbruno [Sat, 12 May 2012 05:05:22 +0000 (05:05 +0000)]
Pointy hat to sleep deprived committer.

Use a *real* variable type instead of one I made up.

Background Music:  Queen -- Bicycle Race

Reviewed by:    BSDCAN 2012 Hacker Lounge Audience

12 years agoAdd in the AP93 configuration file.
adrian [Sat, 12 May 2012 04:52:59 +0000 (04:52 +0000)]
Add in the AP93 configuration file.

The AP93 has:

* AR7240 - mips24k processor with integrated 10/100 switch and
  various other peripherals;
* AR9283 - 2x2 2.4GHz 802.11n (with calibration data in flash);
* 64MB RAM;
* 16MB SPI flash.

The switch code detects as an AR8216 at the present moment, which isn't
_entirely_ strictly true.  However, the MII/MDIO routing in AP93.hints
works - the arge0 MAC connects to PHY4 in the switch, but via the
switch internal MDIO bus.  The switch connects to arge0's MDIO bus,
but only to export the switch registers.

Thanks to stb and ray for the switch work, and ray for helping determine
what the correct switch hints should be for this thing.

12 years agoRemove duplicate words in mdoc(7) pages.
gjb [Sat, 12 May 2012 03:46:43 +0000 (03:46 +0000)]
Remove duplicate words in mdoc(7) pages.

PR: 167810
Submitted by: Bryan Drewery {bryan!shatow%net} (hackers lounge)
Found with: textproc/igor
MFC after: 3 days

12 years agoFix inappropriate data type for two bus_dmamap_t variables that were causing
sbruno [Sat, 12 May 2012 03:30:50 +0000 (03:30 +0000)]
Fix inappropriate data type for two bus_dmamap_t variables that were causing
PAE to insta-panic on startup.  Remove one unused variable that was
commented out.

Reviewed by: ambrisko@
Obtained from: jhb@ peter@ bz@ and countless others during BSDCAN
MFC after: 3 days

12 years agoGeneral mdoc(7) and typo fixes.
gjb [Sat, 12 May 2012 03:25:46 +0000 (03:25 +0000)]
General mdoc(7) and typo fixes.

PR: 167776
Submitted by: Nobuyuki Koganemaru (kogane!jp.freebsd.org)
MFC after: 3 days

12 years agoI forgot to bump the manpage date.
mdf [Sat, 12 May 2012 02:49:40 +0000 (02:49 +0000)]
I forgot to bump the manpage date.

Reminded by: Garrett Cooper
MFC after: 3 days
X-MFC-with: r235297

12 years agoBump .Dd
gjb [Sat, 12 May 2012 00:30:54 +0000 (00:30 +0000)]
Bump .Dd