]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
11 years agoComplete r251452:
attilio [Thu, 6 Jun 2013 18:19:26 +0000 (18:19 +0000)]
Complete r251452:
Avoid to busy/unbusy a page in cases where there is no need to drop the
vm_obj lock, more nominally when the page is full valid after
vm_page_grab().

Sponsored by: EMC / Isilon storage division
Reviewed by: alc

11 years agoDo not compare the existing mask of a cpuset with a new mask when changing
jhb [Thu, 6 Jun 2013 14:43:19 +0000 (14:43 +0000)]
Do not compare the existing mask of a cpuset with a new mask when changing
the mask of a cpuset.  Also, change the cpuset's mask before updating the
masks of all children.  Previously changing a cpuset's mask first required
setting the mask to a super-set of both the old and new masks and then
changing it a second time to the new mask.

11 years agoAdjust CFLAGS to pick up correct regex.h and posix/regex.h. Note this
jkim [Thu, 6 Jun 2013 11:59:38 +0000 (11:59 +0000)]
Adjust CFLAGS to pick up correct regex.h and posix/regex.h.  Note this
actually reverts r250860 and r250861.

Reported by: gjb, tinderbox

11 years agoDon't busy the page unless we are likely to release the object lock.
alc [Thu, 6 Jun 2013 06:17:20 +0000 (06:17 +0000)]
Don't busy the page unless we are likely to release the object lock.

Reviewed by: kib
Sponsored by: EMC / Isilon Storage Division

11 years ago - Consolidate duplicate code into support functions.
jeff [Wed, 5 Jun 2013 23:53:00 +0000 (23:53 +0000)]
 - Consolidate duplicate code into support functions.
 - Split the bqlock into bqclean and bqdirty locks.
 - Only acquire the wakeup synchronization locks when we cross a
   threshold requiring them.
 - Restructure the way flushbufqueues() targets work so they are more
   smp friendly and sane.

Reviewed by: kib
Discussed with: mckusick, attilio
Sponsored by: EMC / Isilon Storage Division

M    vfs_bio.c

11 years agoDocument the new "-v" option for the gssd daemon.
rmacklem [Wed, 5 Jun 2013 23:28:29 +0000 (23:28 +0000)]
Document the new "-v" option for the gssd daemon.
This is a content change.

MFC after: 2 weeks

11 years agoAdd a "-v" (verbose) option to the gssd daemon, to help
rmacklem [Wed, 5 Jun 2013 23:17:39 +0000 (23:17 +0000)]
Add a "-v" (verbose) option to the gssd daemon, to help
with diagnosis of kerberized NFS mount problems. When set,
messages are sent to syslog() (or fprintf(stderr,...) if
"-d" is also specified) to indicate activity/results of
kgssapi upcalls.

Reviewed by: jhb
MFC after: 2 weeks

11 years agorelease: Allow empty extra distributions.
jilles [Wed, 5 Jun 2013 22:27:49 +0000 (22:27 +0000)]
release: Allow empty extra distributions.

For example, WITHOUT_SHAREDOCS= in src.conf creates an empty doc
distribution.

Submitted by: Kurt Lidl
Tested by: Kurt Lidl
Discussed with: gjb
MFC after: 1 week

11 years agoEnable slow diversity combining for the AR9285.
adrian [Wed, 5 Jun 2013 22:23:13 +0000 (22:23 +0000)]
Enable slow diversity combining for the AR9285.

Now that I understand what's going on - and the RX antenna array maps
to what the receive LNA configuration actually is - I feel comfortable
in enabling this.

If people do have issues with this, there's enough debugging now available
that we have a chance to diagnose it without writing it up as 'weird
crap.'

Tested:

* AR9285 STA w/ diversity combining enabled in EEPROM

TODO:

* (More) testing in hostap mode

11 years agoAs a temporary work-around (read: until there's a nice API for exposing
adrian [Wed, 5 Jun 2013 22:21:13 +0000 (22:21 +0000)]
As a temporary work-around (read: until there's a nice API for exposing
and controlling this form of antenna diversity) - print out the AR9285
antenna diversity configuration at attach time.

This will help track down and diagose if/when people have connectivity
issues on cards (eg if they connect a single antenna to LNA1, yet the
card has RX configured to only occur on LNA2.)

Tested:

* AR9285 w/ antenna diversity enabled in EEPROM;
* AR9285 w/ antenna diversity disabled in EEPROM; mapping only to a
  single antenna (LNA1.)

11 years agoUse calloc().
delphij [Wed, 5 Jun 2013 21:56:29 +0000 (21:56 +0000)]
Use calloc().

MFC after: 2 weeks

11 years agoRemove unneeded reference to link.h (sys/link_elf.h).
delphij [Wed, 5 Jun 2013 21:55:20 +0000 (21:55 +0000)]
Remove unneeded reference to link.h (sys/link_elf.h).

MFC after: 2 weeks

11 years agoConnect libgnuregex 2.17 to the build.
jkim [Wed, 5 Jun 2013 21:31:06 +0000 (21:31 +0000)]
Connect libgnuregex 2.17 to the build.

11 years agoFix build for FreeBSD. We do not have alloca.h.
jkim [Wed, 5 Jun 2013 21:28:33 +0000 (21:28 +0000)]
Fix build for FreeBSD.  We do not have alloca.h.

11 years agoAdd GNU regex from glibc 2.17.
jkim [Wed, 5 Jun 2013 21:27:04 +0000 (21:27 +0000)]
Add GNU regex from glibc 2.17.

11 years agoImport couple of important bug fixes from the upstream.
jkim [Wed, 5 Jun 2013 21:12:55 +0000 (21:12 +0000)]
Import couple of important bug fixes from the upstream.

http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commit;h=7e2f0d2
http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commit;h=a445af0

11 years agocxgbe(4): Never install a firmware if hw.cxgbe.fw_install is 0.
np [Wed, 5 Jun 2013 20:57:52 +0000 (20:57 +0000)]
cxgbe(4): Never install a firmware if hw.cxgbe.fw_install is 0.

MFC after: 1 week

11 years agosleep: Explain in a comment why the [EINTR] check is there.
jilles [Wed, 5 Jun 2013 20:15:18 +0000 (20:15 +0000)]
sleep: Explain in a comment why the [EINTR] check is there.

Suggested by: eadler

11 years agosh(1): Document new features in wait builtin.
jilles [Wed, 5 Jun 2013 19:54:28 +0000 (19:54 +0000)]
sh(1): Document new features in wait builtin.

PR: 176916

11 years agoPull in r183297 from upstream llvm trunk:
dim [Wed, 5 Jun 2013 19:46:39 +0000 (19:46 +0000)]
Pull in r183297 from upstream llvm trunk:

  PR15662: Optimized debug info produces out of order function
  parameters

  When a function is inlined we lazily construct the variables
  representing the function's parameters. After that, we add any
  remaining unused parameters.

  If the function doesn't use all the parameters, or uses them out of
  order, then the DWARF would produce them in that order, producing a
  parameter order that doesn't match the source.

  This fix causes us to always keep the arg variables at the start of
  the variable list & in the original order from the source.

Reported by: avg
MFC after: 1 week

11 years agosh: Return status 127 for unknown jobs in wait builtin.
jilles [Wed, 5 Jun 2013 19:40:52 +0000 (19:40 +0000)]
sh: Return status 127 for unknown jobs in wait builtin.

This is required by POSIX, at least for pids that are not known child
processes.

Other problems with job specifications still cause wait to abort with
exit status 2.

PR: 176916

11 years agosh: Allow multiple operands in wait builtin.
jilles [Wed, 5 Jun 2013 19:08:22 +0000 (19:08 +0000)]
sh: Allow multiple operands in wait builtin.

This is only part of the PR; the behaviour for unknown/invalid pids/jobs
remains unchanged (aborts the builtin with status 2).

PR: 176916
Submitted by: Vadim Goncharov

11 years agonew features (mostly for testing netmap capabilities)
luigi [Wed, 5 Jun 2013 17:37:59 +0000 (17:37 +0000)]
new features (mostly for testing netmap capabilities)

+ pkt-gen -f rx now remains active even when traffic stops
  Previous behaviour (exit after 1 second of silence) can be
  restored with the -W option

+ the -X option does a hexdump of the content of a packet (both tx and rx).
  This can be useful to check what goes in and out.

+ the -I option instructs the sender to use indirect buffers
  (not really useful other than to test the kernel module in the
  VALE switch)

11 years ago- fix a bug in the previous commit that was dropping the last packet
luigi [Wed, 5 Jun 2013 17:27:59 +0000 (17:27 +0000)]
- fix a bug in the previous commit that was dropping the last packet
  from each batch flowing on the VALE switch

- feature: add glue for 'indirect' buffers on the sender side:
  if a slot has NS_INDIRECT set, the netmap buffer contains pointer(s)
  to the actual userspace buffers, which are accessed with copyin().
     The feature is not finalised yet, as it will likely need to deal
  with some iovec variant for proper scatter/gather support.
  This will save one copy for clients (e.g. qemu) that cannot
  use the netmap buffer directly.

A curiosity: on amd64 copyin() appears to be 10-15% faster than pkt_copy()
or bcopy() at least for sizes of 256 and greater.

11 years agoNew install flags were merged to 9-STABLE and will appear in FreeBSD-9.2.
brooks [Wed, 5 Jun 2013 17:02:21 +0000 (17:02 +0000)]
New install flags were merged to 9-STABLE and will appear in FreeBSD-9.2.

Sponsored by: DARPA, AFRL
Submitted by: ru
MFC after: 3 days

11 years agoRelax the vm object locking. Use a read lock.
alc [Wed, 5 Jun 2013 17:00:10 +0000 (17:00 +0000)]
Relax the vm object locking.  Use a read lock.

Sponsored by: EMC / Isilon Storage Division

11 years agoUpdate to bmake-20130604 to fix file descriptor leak.
sjg [Wed, 5 Jun 2013 16:12:50 +0000 (16:12 +0000)]
Update to bmake-20130604 to fix file descriptor leak.

11 years agoRemove part of the NetBSD longjmp code that was not ready to be merged.
andrew [Wed, 5 Jun 2013 07:37:45 +0000 (07:37 +0000)]
Remove part of the NetBSD longjmp code that was not ready to be merged.

11 years agoStyle fixes.
das [Wed, 5 Jun 2013 05:33:01 +0000 (05:33 +0000)]
Style fixes.

Submitted by: bde

11 years ago- Flag sym(4) as supporting unmapped I/O; all necessary conversion actually
marius [Wed, 5 Jun 2013 01:22:59 +0000 (01:22 +0000)]
- Flag sym(4) as supporting unmapped I/O; all necessary conversion actually
  already has been done as part of r246713.
- Revert a part of r251402 in order to appease clang.

11 years agoHandle/mark/nuke unused arguments.
marius [Wed, 5 Jun 2013 01:07:40 +0000 (01:07 +0000)]
Handle/mark/nuke unused arguments.

11 years agoImplement a bit of a hack to store the AR9285/AR9485 RX LNA configuration in
adrian [Wed, 5 Jun 2013 00:45:19 +0000 (00:45 +0000)]
Implement a bit of a hack to store the AR9285/AR9485 RX LNA configuration in
the RX antenna field.

The AR9285/AR9485 use an LNA mixer to determine how to combine the signals
from the two antennas.  This is encoded in the RSSI fields (ctl/ext) for
chain 2.  So, let's use that here.

This maps RX antennas 0->3 to the RX mixer configuration used to
receive a frame.  There's more that can be done but this is good enough
to diagnose if the hardware is doing "odd" things like trying to
receive frames on LNA2 (ie, antenna 2 or "alt" antenna) when there's
only one antenna connected.

Tested:

* AR9285, STA mode

11 years agoAdd a new capability flag to announce that the chip implements LNA mixing
adrian [Wed, 5 Jun 2013 00:42:04 +0000 (00:42 +0000)]
Add a new capability flag to announce that the chip implements LNA mixing
for the RX path.

This is different to the div comb HAL flag, that says it actually
can use this for RX diversity (the "slow" diversity path implemented
but disabled in the AR9285 HAL code.)

Tested:

* AR9285, STA operation

11 years agoDocument the AR9285/AR9485 LNA configuration information that's
adrian [Wed, 5 Jun 2013 00:39:20 +0000 (00:39 +0000)]
Document the AR9285/AR9485 LNA configuration information that's
stored in the ctl/ext RSSI field for chain 2.

Tested:

* AR9285, STA

11 years agoAdd RX STBC statistics.
adrian [Tue, 4 Jun 2013 23:52:57 +0000 (23:52 +0000)]
Add RX STBC statistics.

11 years agoIn vm_object_split(), busy and consequently unbusy the pages only when
attilio [Tue, 4 Jun 2013 22:47:01 +0000 (22:47 +0000)]
In vm_object_split(), busy and consequently unbusy the pages only when
swap_pager_copy() is invoked, otherwise there is no reason to do so.
This will eliminate the necessity to busy pages most of the times.

Sponsored by: EMC / Isilon storage division
Reviewed by: alc

11 years agosys/dev/mps/mps.c
asomers [Tue, 4 Jun 2013 22:32:33 +0000 (22:32 +0000)]
sys/dev/mps/mps.c
sys/dev/mps/mps_user.c
Fix uninitialized memory reference in mps_read_config_page.  It was
referencing a field (params->hdr.Ext.ExtPageType) that would only be
set when reading an Extended config page.  The symptom was that
MPSIO_READ_CFG_PAGE ioctls would randomly fail with
MPI2_IOCSTATUS_CONFIG_INVALID_PAGE errors.  The solution is to
determine whether an extended or an ordinary config page is requested
by looking at the PageType field, which should be available regardless.

Similarly, mps_user_read_extcfg_header and mps_user_read_extcfg_page,
which call mps_read_config_page, had to be fixed to always set the
PageType field.  They were implicitly assuming that
mps_read_config_page always operated on Extended pages.

Reviewed by: ken
Approved by: ken (mentor)
MFC after: 3 days

11 years agoCAM_DEV_QFREEZE handling should only be done on request submission but
marius [Tue, 4 Jun 2013 20:49:17 +0000 (20:49 +0000)]
CAM_DEV_QFREEZE handling should only be done on request submission but
neither on completion nor by SIM drivers in the first place. This issue
has been revealed by r249466.

Reviewed by: mav
MFC after: 3 days

11 years agoFix build with both INET and INET6 disabled.
jhb [Tue, 4 Jun 2013 20:40:16 +0000 (20:40 +0000)]
Fix build with both INET and INET6 disabled.

11 years agoStart to merge the updated ARM NetBSD setjump/longjmp functions. To begin
andrew [Tue, 4 Jun 2013 19:47:26 +0000 (19:47 +0000)]
Start to merge the updated ARM NetBSD setjump/longjmp functions. To begin
with merge the functions but leave out the code to save/load the VFP
registers as that requires other changes to ensure the VFP is enabled
first.

This removes storing the old fpa registers. These were never fully
supported, and the only user of this code I can find have moved to newer
CPUs which use a VFP.

11 years agoRelax the vm object locking in mac_proc_vm_revoke_recurse(). A read lock
alc [Tue, 4 Jun 2013 17:23:09 +0000 (17:23 +0000)]
Relax the vm object locking in mac_proc_vm_revoke_recurse().  A read lock
suffices in one place.

Sponsored by: EMC / Isilon Storage Division

11 years agoShift Console->Saver menu items to place the default item at the top, making
dteske [Tue, 4 Jun 2013 17:08:26 +0000 (17:08 +0000)]
Shift Console->Saver menu items to place the default item at the top, making
this menu consistent with Console->Repeat, Console->Font, Console->Screenmap
and many others which place the default choice at the top.

NOTE: SVN r249751 changed things so that these menus highlight the active
selection, so the idea that the default choice (now at top) "can be selected
immediately by hitting ENTER" is not possible unless the default choice is
the currently active setting (e.g., on first visit to the menu). However, it
is still warranted to have the default choice at the top of the menu.

PR: bin/169316

11 years agoLessen verbosity of "Exit" menutag items to appear more consistent across
dteske [Tue, 4 Jun 2013 16:48:23 +0000 (16:48 +0000)]
Lessen verbosity of "Exit" menutag items to appear more consistent across
menus from one module to the next.

Replacing "Exit this menu (returning to previous)" with "Exit this menu".

PR: ports/169316

11 years agoAdd myself to committers-src.dot, and note kib is my mentor.
gjb [Tue, 4 Jun 2013 16:02:00 +0000 (16:02 +0000)]
Add myself to committers-src.dot, and note kib is my mentor.

Approved by: kib (mentor)

11 years agoEliminate unnecessary vm object locking from tmpfs_nocacheread().
alc [Tue, 4 Jun 2013 15:40:45 +0000 (15:40 +0000)]
Eliminate unnecessary vm object locking from tmpfs_nocacheread().

11 years agoBuild fix: Only <sys/cdefs.h> should be included before __FBSDID().
jhb [Tue, 4 Jun 2013 15:35:37 +0000 (15:35 +0000)]
Build fix:  Only <sys/cdefs.h> should be included before __FBSDID().
<sys/param.h> needs to be included after any "opt_foo.h" headers so it
sees the same set of defined macros as other headers.

11 years agoImprove r250890, so that we stop processing of a message with zero
glebius [Tue, 4 Jun 2013 11:19:08 +0000 (11:19 +0000)]
Improve r250890, so that we stop processing of a message with zero
descriptors as early as possible, and assert that number of descriptors
is positive in unp_freerights().

Reviewed by: mjg, pjd, jilles

11 years agoFlag isp(4) as supporting unmapped I/O; all necessary conversion actually
marius [Tue, 4 Jun 2013 11:05:57 +0000 (11:05 +0000)]
Flag isp(4) as supporting unmapped I/O; all necessary conversion actually
already has been done as part of r246713.

Reviewed by: mjacob

11 years agoFix the passing of time on Armada XP.
gber [Tue, 4 Jun 2013 09:33:03 +0000 (09:33 +0000)]
Fix the passing of time on Armada XP.

In order to become independent of Coherency Fabric frequency, configure
Timer and Watchdog to operate in 25MHz mode.

Submitted by: Zbigniew Bodek <zbb@semihalf.com>

11 years agoImplement pmap_copy() for ARMv6/v7.
gber [Tue, 4 Jun 2013 09:21:18 +0000 (09:21 +0000)]
Implement pmap_copy() for ARMv6/v7.

Copy the given range of mappings from the source map to the
destination map, thereby reducing the number of VM faults on fork.

Submitted by:   Zbigniew Bodek <zbb@semihalf.com>
Sponsored by:   The FreeBSD Foundation, Semihalf

11 years agomdoc: convert .Fd to .In, which is much nicer.
joel [Tue, 4 Jun 2013 07:37:06 +0000 (07:37 +0000)]
mdoc: convert .Fd to .In, which is much nicer.

11 years agoMatch the options of the kernel.
obrien [Tue, 4 Jun 2013 06:38:01 +0000 (06:38 +0000)]
Match the options of the kernel.

11 years agoUpdate a comment.
alc [Tue, 4 Jun 2013 05:44:52 +0000 (05:44 +0000)]
Update a comment.

11 years agoAdd man links for expl(3) and expm1l(3).
das [Tue, 4 Jun 2013 05:41:38 +0000 (05:41 +0000)]
Add man links for expl(3) and expm1l(3).

11 years agoUse f_shell_escape() instead of forking to awk. In this case, the
dteske [Tue, 4 Jun 2013 03:47:21 +0000 (03:47 +0000)]
Use f_shell_escape() instead of forking to awk. In this case, the
replacement comes with a great performance increase (as f_shell_escape()
uses the built-in based f_replaceall() which out-performs forking to
awk(1)). This should also improve readability slightly.

11 years agoUse a case-statement instead of multiple (separate even) if-statements.
dteske [Tue, 4 Jun 2013 03:38:16 +0000 (03:38 +0000)]
Use a case-statement instead of multiple (separate even) if-statements.

11 years agoComments.
dteske [Tue, 4 Jun 2013 03:37:35 +0000 (03:37 +0000)]
Comments.

11 years agoSimilar to r251236, r251242, and r251244. Introduce new function-pair to
dteske [Tue, 4 Jun 2013 03:30:44 +0000 (03:30 +0000)]
Similar to r251236, r251242, and r251244. Introduce new function-pair to
dialog(1) API in dialog.subr responsible for retrieving menu-choice data
(for the --menu widget).

Add f_dialog_menuitem_store()/f_dialog_menuitem_fetch() for storing and
retrieving the menuitem (not the tag) from a --menu widget result.

The dialog --menu widget returns the `tag' from the tag/item pair of choices
for a menu list. Because the menu list is often local to the function that
is displaying it, the ability to dereference this into the item is not
possible unless you use a global. This changes things so the function (upon
successful return of dialog) dereferences the tag into the item and stores
it for later retrieval using these functions.

NOTE: Tags are dereferenced into items using f_dialog_menutag2item().

11 years agoAdd the combined (mixed) diversity support capability bit for the
adrian [Tue, 4 Jun 2013 02:56:56 +0000 (02:56 +0000)]
Add the combined (mixed) diversity support capability bit for the
AR9285/AR9485.

11 years agoRelax the object locking in vm_pageout_map_deactivate_pages() and
alc [Tue, 4 Jun 2013 02:28:47 +0000 (02:28 +0000)]
Relax the object locking in vm_pageout_map_deactivate_pages() and
vm_pageout_object_deactivate_pages().  A read lock suffices.

Sponsored by: EMC / Isilon Storage Division

11 years agocxgbe(4): Provide accurate hit count for filters on T5 cards. The
np [Tue, 4 Jun 2013 02:25:25 +0000 (02:25 +0000)]
cxgbe(4): Provide accurate hit count for filters on T5 cards.  The
location within the TCB and the size have both changed.

MFC after: 1 week

11 years agoFix a code typo (missing `eval') which lead to improperly-sized menu when
dteske [Tue, 4 Jun 2013 00:42:31 +0000 (00:42 +0000)]
Fix a code typo (missing `eval') which lead to improperly-sized menu when
displaying packages (menu was always maximum height). The problem was
introduced by SVN r251232.

11 years agoPad the PCPU MD struct, to satisfy an assert added with the projects/counters
jhibbits [Tue, 4 Jun 2013 00:40:26 +0000 (00:40 +0000)]
Pad the PCPU MD struct, to satisfy an assert added with the projects/counters
branch import.

PR: ports/179173,ports/179164

11 years agoFix menu accenting for the package category list. The single-space indent
dteske [Tue, 4 Jun 2013 00:38:19 +0000 (00:38 +0000)]
Fix menu accenting for the package category list. The single-space indent
for duplicate initial-characters was removed for making comparisons but was
not put back.

11 years agoWhitespace and comments.
dteske [Tue, 4 Jun 2013 00:35:57 +0000 (00:35 +0000)]
Whitespace and comments.

11 years agoImprove debugging for f_dialog_*_constrain() to print constrained rather
dteske [Tue, 4 Jun 2013 00:30:05 +0000 (00:30 +0000)]
Improve debugging for f_dialog_*_constrain() to print constrained rather
than unconstrained values.

11 years agobmake does not like universe_${target}_prologue in the .ORDER
sjg [Mon, 3 Jun 2013 23:51:46 +0000 (23:51 +0000)]
bmake does not like universe_${target}_prologue in the .ORDER
when there is an explicit dependency for it anyway.
Remove it so both bmake and fmake do the same thing.

Reviewed by: OBrien

11 years agoFixing a typo.
hiren [Mon, 3 Jun 2013 22:22:53 +0000 (22:22 +0000)]
Fixing a typo.

Approved by: sbruno (mentor)
MFC after: 3 days

11 years agoFix a typo: XPORT_SPI should be tested against transport, nor protocol.
delphij [Mon, 3 Jun 2013 21:52:19 +0000 (21:52 +0000)]
Fix a typo: XPORT_SPI should be tested against transport, nor protocol.

Submitted by: Sascha Wildner <swildner dragonflybsd org>
Reviewed by: mjacob
MFC after: 2 weeks

11 years agoAdd ATOMIC_*_LOCK_FREE constants.
ed [Mon, 3 Jun 2013 20:52:20 +0000 (20:52 +0000)]
Add ATOMIC_*_LOCK_FREE constants.

These constants are part of the C standard. Both Clang and GCC seem to
export these constants under the name __GCC_ATOMIC_*_LOCK_FREE.

11 years agoext2fs: space vs tab.
pfg [Mon, 3 Jun 2013 20:33:05 +0000 (20:33 +0000)]
ext2fs: space vs tab.

Obtained from: Christoph Mallon
MFC after: 3 days

11 years agoChange a comma to a semicolon.
kargl [Mon, 3 Jun 2013 20:09:22 +0000 (20:09 +0000)]
Change a comma to a semicolon.

Remove a blank line that crept into the declarations.

Fix a comment to show a sign on a NaN.

11 years agoext2fs: Small cosmetic fixes.
pfg [Mon, 3 Jun 2013 20:02:45 +0000 (20:02 +0000)]
ext2fs: Small cosmetic fixes.

Make a long macro readable and sort a header.

Obtained from: Christoph Mallon
MFC after: 3 days

11 years agold80 and ld128 implementations of expm1l(). This code started life
kargl [Mon, 3 Jun 2013 19:51:32 +0000 (19:51 +0000)]
ld80 and ld128 implementations of expm1l().  This code started life
as a fairly faithful implementation of the algorithm found in

PTP Tang, "Table-driven implementation of the Expm1 function
in IEEE floating-point arithmetic," ACM Trans. Math. Soft., 18,
211-222 (1992).

Over the last 18-24 months, the code has under gone significant
optimization and testing.

Reviewed by: bde
Obtained from: bde (most of the optimizations)

11 years agoFix the order of TX shutdown and reset.
adrian [Mon, 3 Jun 2013 19:39:37 +0000 (19:39 +0000)]
Fix the order of TX shutdown and reset.

* Grab the reset lock first, so any subsequent interrupt, TX, RX work
  will fail

* Then shut down interrupts

* Then wait for TX/RX to finish running

At this point no further work will be running, so it's safe to do the
reset path code.

PR: kern/179232

11 years agoFix two comments that got lost in the disentanglement of the larger diff.
kargl [Mon, 3 Jun 2013 19:29:03 +0000 (19:29 +0000)]
Fix two comments that got lost in the disentanglement of the larger diff.

11 years agoFix receive on the AR9285 (Kite) with only one antenna connected.
adrian [Mon, 3 Jun 2013 19:14:29 +0000 (19:14 +0000)]
Fix receive on the AR9285 (Kite) with only one antenna connected.

The main problem here is that fast and driver RX diversity isn't actually
configured; I need to figure out why that is.  That said, this makes
the single-antenna connected AR9285 and AR2427 (AR9285 w/ no 11n) work
correctly.

PR: kern/179269

11 years agold80/s_expl.c:
kargl [Mon, 3 Jun 2013 19:13:44 +0000 (19:13 +0000)]
ld80/s_expl.c:

* Use integral numerical constants, and let the compiler do the
  conversion to long double.

ld128/s_expl.c:

* Use integral numerical constants, and let the compiler do the
  conversion to long double.
* Use the ENTERI/RETURNI macros, which are no-ops on ld128.  This
  however makes the ld80 and ld128 identical.

Reviewed by: bde (as part of larger diff)

11 years agoMicro-optimization: move the unary mius operator to operate
kargl [Mon, 3 Jun 2013 18:57:35 +0000 (18:57 +0000)]
Micro-optimization: move the unary mius operator to operate
on a literal constant.

Obtained from: bde

11 years agoAdd a comment to note that bde supplied most, if not all,
kargl [Mon, 3 Jun 2013 18:53:40 +0000 (18:53 +0000)]
Add a comment to note that bde supplied most, if not all,
of the optimizations.

11 years agoext2fs: Update Block Group Descriptor struct.
pfg [Mon, 3 Jun 2013 18:52:14 +0000 (18:52 +0000)]
ext2fs: Update Block Group Descriptor struct.

Uncover some, previously reserved, fields that are used by Ext4.
These are currently unused but it is good to have them for future
reference.

Reviewed by: bde
MFC after: 3 days

11 years agold80/s_expl.c:
kargl [Mon, 3 Jun 2013 18:51:34 +0000 (18:51 +0000)]
ld80/s_expl.c:

* In the special case x = -Inf or -NaN, use a micro-optimization
  to eliminate the need to access u.xbits.man.

* Fix an off-by-one for small arguments |x| < 0x1p-65.

ld128/s_expl.c:

* In the special case x = -Inf or -NaN, use a micro-optimization
  to eliminate the need to access u.xbits.manh and u.xbits.manl.

* Fix an off-by-one for small arguments |x| < 0x1p-114.

Obtained from: bde

11 years agold80/s_expl.c:
kargl [Mon, 3 Jun 2013 18:40:00 +0000 (18:40 +0000)]
ld80/s_expl.c:

* Update the evaluation of the polynomial.  This allows the removal
  of the now unused variables t23 and t45.

ld128/s_expl.c:

* Update the evaluation of the polynomial and the intermediate
  result t.  This update allows several numerical constants to be
  written as double rather than long double constants.   Update
  the constants as appropriate.

Obtained from: bde

11 years agoRename a few P2, P3, ... coefficients to A2, A3, ... missed in
kargl [Mon, 3 Jun 2013 18:18:08 +0000 (18:18 +0000)]
Rename a few P2, P3, ... coefficients to A2, A3, ... missed in
my previous commit.

11 years agoUpdate a comment to reflect that we are using an endpoint of
kargl [Mon, 3 Jun 2013 18:14:18 +0000 (18:14 +0000)]
Update a comment to reflect that we are using an endpoint of
an interval instead of a midpoint.

11 years agoAdd a u suffix to the IEEEl2bits unions o_threshold and u_threshold,
kargl [Mon, 3 Jun 2013 18:07:04 +0000 (18:07 +0000)]
Add a u suffix to the IEEEl2bits unions o_threshold and u_threshold,
and use macros to access the e component of the unions.  This allows
the portions of the code in ld80 to be identical to the ld128 code.

Obtained from: bde

11 years agoIntroduce the macro LOG2_INTERVAL, which is log2(number of intervals).
kargl [Mon, 3 Jun 2013 17:51:08 +0000 (17:51 +0000)]
Introduce the macro LOG2_INTERVAL, which is log2(number of intervals).
Use the macroi as a micro-optimization to convert a subtraction and
division to a shift.

Obtained from: bde

11 years ago- Fix a couple of inverted panic messages for shared/exclusive mismatches
jhb [Mon, 3 Jun 2013 17:41:11 +0000 (17:41 +0000)]
- Fix a couple of inverted panic messages for shared/exclusive mismatches
  of a lock within a single thread.
- Fix handling of interlocks in WITNESS by properly requiring the interlock
  to be held exactly once if it is specified.

11 years agoWhitespace.
kargl [Mon, 3 Jun 2013 17:40:52 +0000 (17:40 +0000)]
Whitespace.

11 years agoAssert that interrupts are enabled in the trap handlers on x86 before
kib [Mon, 3 Jun 2013 17:40:05 +0000 (17:40 +0000)]
Assert that interrupts are enabled in the trap handlers on x86 before
calling generic code to deliver signals.

Discussed with: bde
Tested by: pho
MFC after: 1 week

11 years ago- Handle the recursed/not recursed flags with RA_RLOCKED in rw_assert().
jhb [Mon, 3 Jun 2013 17:38:57 +0000 (17:38 +0000)]
- Handle the recursed/not recursed flags with RA_RLOCKED in rw_assert().
- Tweak a panic message.

11 years agoBe more generous when donating the current thread time to the owner of
kib [Mon, 3 Jun 2013 17:36:43 +0000 (17:36 +0000)]
Be more generous when donating the current thread time to the owner of
the vnode lock while iterating over the free vnode list.  Instead of
yielding, pause for 1 tick.  The change is reported to help in some
virtualized environments.

Submitted by: Roger Pau Monn? <roger.pau@citrix.com>
Discussed with: jilles
Tested by: pho
MFC after: 2 weeks

11 years ago* Rename the polynomial coefficients from P2, P3, ... to A2, A3, ....
kargl [Mon, 3 Jun 2013 17:36:26 +0000 (17:36 +0000)]
* Rename the polynomial coefficients from P2, P3, ... to A2, A3, ....
  The names now coincide with the name used in PTP Tang's paper.

* Rename the variable from s to tbl to better reflect that
  this is a table, and to be consistent with the naming scheme
  in s_exp2l.c

Reviewed by: bde (as part of larger diff)

11 years agoRemove irrelevant comments.
kib [Mon, 3 Jun 2013 17:30:40 +0000 (17:30 +0000)]
Remove irrelevant comments.

Discussed with: alc
MFC after: 3 days

11 years agocxgbe(4): t4fw_cfg must be explicitly loaded if the driver is being
np [Mon, 3 Jun 2013 17:30:21 +0000 (17:30 +0000)]
cxgbe(4): t4fw_cfg must be explicitly loaded if the driver is being
loaded via loader.conf.

Submitted by: jwd@
MFC after: 3 days

11 years ago* Style(9). Start non-Copyright fancy formatted comments with /**.
kargl [Mon, 3 Jun 2013 17:24:46 +0000 (17:24 +0000)]
* Style(9).   Start non-Copyright fancy formatted comments with /**.

Reviewed by: bde (as part of larger diff)

11 years agold80/s_expl.c:
kargl [Mon, 3 Jun 2013 17:21:43 +0000 (17:21 +0000)]
ld80/s_expl.c:

* Update Copyright years to include 2013.

ld128/s_expl.c:

* Correct and update Copyright years.  This code originated from
  the ld80 version, so it should reflect the same time period.

Reviewed by: bde (as part of larger diff)

11 years agoAdd libiconv based versions of *c16*() and *c32*().
ed [Mon, 3 Jun 2013 17:17:56 +0000 (17:17 +0000)]
Add libiconv based versions of *c16*() and *c32*().

I initially thought wchar_t was locale independent, but this seems to be
only the case on Linux. This means that we cannot depend on the *wc*()
routines to implement *c16*() and *c32*(). Instead, use the Citrus
libiconv that is part of libc.

I'll see if there is anything I can do to make the existing functions
somewhat useful in case the system is built without libiconv in the
nearby future. If not, I'll simply remove the broken implementations.

Reviewed by: jilles, gabor

11 years agoUpdate jemalloc to version 3.4.0.
jasone [Mon, 3 Jun 2013 14:36:28 +0000 (14:36 +0000)]
Update jemalloc to version 3.4.0.

11 years agoSwitch to 2-clause license
emaste [Mon, 3 Jun 2013 13:11:48 +0000 (13:11 +0000)]
Switch to 2-clause license

Approved by: luigi@