]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
10 years ago- Split transmit queue into one for each type. Apparently there will
hselasky [Wed, 21 May 2014 16:52:55 +0000 (16:52 +0000)]
- Split transmit queue into one for each type. Apparently there will
be a race when using a single active queue for all transmit types.
- Last argument of usb_pause_mtx() is ticks and not milliseconds.
- Remove unused watchdog.
- Remove some unused fields from the RSU softc structure.
- Workaround usbd_transfer_start() recursion from inside of completion
callback.

MFC after: 3 days

10 years agoAdded sysctls / tunables for ZFS dirty data tuning
smh [Wed, 21 May 2014 13:36:04 +0000 (13:36 +0000)]
Added sysctls / tunables for ZFS dirty data tuning

Added the following new sysctls / tunables:
* vfs.zfs.dirty_data_max
* vfs.zfs.dirty_data_max_max
* vfs.zfs.dirty_data_max_percent
* vfs.zfs.dirty_data_sync
* vfs.zfs.delay_min_dirty_percent
* vfs.zfs.delay_scale

PR: kern/189865
MFC after: 2 weeks

10 years agoSync lock(1) on VT_LOCKSWITCH usage with syscons(4), vt(4) and vidcontrol(1).
ray [Wed, 21 May 2014 11:53:15 +0000 (11:53 +0000)]
Sync lock(1) on VT_LOCKSWITCH usage with syscons(4), vt(4) and vidcontrol(1).

Sponsored by: The FreeBSD Foundation

10 years agoFix tty locking.
ray [Wed, 21 May 2014 11:15:38 +0000 (11:15 +0000)]
Fix tty locking.
o Correct expected values for VT_LOCKSWITCH ioctl.
o Check current window for locked state.

Sponsored by: The FreeBSD Foundation

10 years ago- Fix a bug which can make sysctl() fail when -F is specified.
hrs [Wed, 21 May 2014 10:04:51 +0000 (10:04 +0000)]
- Fix a bug which can make sysctl() fail when -F is specified.
- Increase WID_IF_DEFAULT() from 6 to 8 (the default for AF_INET6) because
  we have interfaces with longer names than 6 chars like epairN{a,b}.
- Style fixes.

10 years ago- Replace some constants with macros.
hselasky [Wed, 21 May 2014 09:26:02 +0000 (09:26 +0000)]
- Replace some constants with macros.
- Need to set the pre-fetch memory address when reading the host memory.
- We currently assume that no endianness conversion is needed.

Sponsored by: DARPA, AFRL

10 years agoRemove redundand loop. The inner goto restarts the whole page
kib [Wed, 21 May 2014 08:19:04 +0000 (08:19 +0000)]
Remove redundand loop.  The inner goto restarts the whole page
handling in the situation identical to the loop condition.

Sponsored by: The FreeBSD Foundation
MFC after: 3 days

10 years ago- Configure Rx bulk
kevlo [Wed, 21 May 2014 08:09:44 +0000 (08:09 +0000)]
- Configure Rx bulk
- Announce flow control capability to PHY drivers

10 years agoAvoid starting a USB transfer inside the callback function, right
hselasky [Wed, 21 May 2014 06:23:52 +0000 (06:23 +0000)]
Avoid starting a USB transfer inside the callback function, right
after its been submitted.

MFC after: 3 days

10 years agoDocument VMware-related filesystems additions.
marck [Wed, 21 May 2014 05:27:57 +0000 (05:27 +0000)]
Document VMware-related filesystems additions.

Reviewed by:    jmg
MFC after:      1 week

10 years agoCheck for mismatched vref()/vdrop()
bjk [Wed, 21 May 2014 03:11:27 +0000 (03:11 +0000)]
Check for mismatched vref()/vdrop()

Assert that the hold count has not fallen below the use count, a situation
that would only happen when a vref() (or similar) is erroneously paired
with a vdrop().  This situation has not been observed in the wild, but
could be helpful for someone implementing a new filesystem.

Reviewed by: kib
Approved by: hrs (mentor)

10 years agoOld PCIe implementations cannot allow a DMA transfer to cross a 4GB
scottl [Tue, 20 May 2014 22:43:17 +0000 (22:43 +0000)]
Old PCIe implementations cannot allow a DMA transfer to cross a 4GB
boundary.  This was addressed several years ago by creating a parent
tag hierarchy for the root buses that set the boundary restriction
for appropriate buses and allowed child deviced to inherit it.
Somewhere along the way, this restriction was turned into a case for
marking the tag as a candidate for needing bounce buffers, instead
of just splitting the segment along the boundary line.  This flag
also causes all maps associated with this tag to be non-NULL, which
in turn causes bus_dmamap_sync() to take the slow path of function
pointer indirection to discover that there's no bouncing work to
do.  The end result is a lot of pages set aside in bounce pools
that will never be used, and a slow path for data buffers in nearly
every DMA-capable PCIe device.  For example, our workload at Netflix
was spending nearly 1% of all CPU time going through this slow path.

Fix this problem by being more selective about when to set the
COULD_BOUNCE flag.  Only set it when the boundary restriction
exists and the consumer cannot do more than a single DMA segment
at once.  This fixes the case of dynamic buffers (mbufs, bio's)
but doesn't address static buffers allocated from bus_dmamem_alloc().
That case will be addressed in the future.

For those interested, this was discovered thanks to Dtrace Flame
Graphs.

Discussed with: jhb, kib
Obtained from: Netflix, Inc.
MFC after: 3 days

10 years ago- Add myself to committers-src.dot
slm [Tue, 20 May 2014 22:11:52 +0000 (22:11 +0000)]
- Add myself to committers-src.dot

Approved by: ken/scottl (mentors)

10 years agoMove Nx definition to a separate block.
pluknet [Tue, 20 May 2014 21:24:15 +0000 (21:24 +0000)]
Move Nx definition to a separate block.

Noticed by: ru (a while ago)

10 years agoAdd PG_RW check when translating a guest linear to guest physical address.
neel [Tue, 20 May 2014 20:30:28 +0000 (20:30 +0000)]
Add PG_RW check when translating a guest linear to guest physical address.

Set the accessed and dirty bits in the page table entry. If it fails then
restart the page table walk from the beginning. This might happen if another
vcpu modifies the page tables simultaneously.

Reviewed by: alc, kib

10 years agoFix warning messages after r252015
rea [Tue, 20 May 2014 19:55:59 +0000 (19:55 +0000)]
Fix warning messages after r252015

$alias used to hold alias number, but now it carries full variable name,
so messages were tuned to account for that.

Other fixes:
 - eliminate unneeded double spaces;
 - tell user where inet/inet6 keywords are expected to be.
Reviewed by: hrs
MFC after: 1 week

10 years agoAdd ismt(4) driver.
jimharris [Tue, 20 May 2014 19:55:06 +0000 (19:55 +0000)]
Add ismt(4) driver.

ismt(4) supports the SMBus Message Transport controller found on Intel
C2000 series (Avoton) and S1200 series (Briarwood) Atom SoCs.

Sponsored by: Intel

10 years agoUse an intermediate target to associate with _SUBDIR which is marked .MAKE
sjg [Tue, 20 May 2014 18:25:46 +0000 (18:25 +0000)]
Use an intermediate target to associate with _SUBDIR which is marked .MAKE
this allows make -n to do tree walks as expected without
doing anything else (as intended).
Use prefix _sub. to help avoid conflict with any real target.

Reviewed by: imp

10 years ago- The DELAY() should not be used in USB drivers.
hselasky [Tue, 20 May 2014 15:53:17 +0000 (15:53 +0000)]
- The DELAY() should not be used in USB drivers.
- The usb_pause_mtx() function takes ticks and not milliseconds as last argument.

MFC after: 3 days

10 years ago- The DELAY() should not be used in USB drivers.
hselasky [Tue, 20 May 2014 15:47:37 +0000 (15:47 +0000)]
- The DELAY() should not be used in USB drivers.
- The usb_pause_mtx() function takes ticks and not milliseconds as last argument.

MFC after: 3 days

10 years agoOptimise reading of pending interrupt registers. If there are no
hselasky [Tue, 20 May 2014 15:03:23 +0000 (15:03 +0000)]
Optimise reading of pending interrupt registers. If there are no
pending interrupt bits, skip the bit iteration loop.

Reviewed by: ian @

10 years agoAdd IOMMU PCI subclass, found on Tyan S8236 motherboard.
mav [Tue, 20 May 2014 14:39:22 +0000 (14:39 +0000)]
Add IOMMU PCI subclass, found on Tyan S8236 motherboard.

Submitted by: Dmitry Luhtionov <dmitryluhtionov@gmail.com>
MFC after: 2 weeks

10 years agoCorrect some programming details. The layout of the PDTs were
hselasky [Tue, 20 May 2014 14:15:03 +0000 (14:15 +0000)]
Correct some programming details. The layout of the PDTs were
different from what was initially thought. Fix re-programming of
hardware mode register after reset.

Sponsored by: DARPA, AFRL

10 years agoMake sure detach code is executed in all cases. This fixes a panic
hselasky [Tue, 20 May 2014 12:22:53 +0000 (12:22 +0000)]
Make sure detach code is executed in all cases. This fixes a panic
when debugging is enabled.

Reported by: Idwer Vollering <vidwer@gmail.com>
MFC after: 3 days

10 years agoChange comment about HPNDisabled to match the style of other options to
smh [Tue, 20 May 2014 10:28:19 +0000 (10:28 +0000)]
Change comment about HPNDisabled to match the style of other options to
avoid confusion.

Sponsored by: Multiplay

10 years agoWhen exec_new_vmspace() decides that current vmspace cannot be reused
kib [Tue, 20 May 2014 09:19:35 +0000 (09:19 +0000)]
When exec_new_vmspace() decides that current vmspace cannot be reused
on execve(2), it calls vmspace_exec(), which frees the current
vmspace.  The thread executing an exec syscall gets new vmspace
assigned, and old vmspace is freed if only referenced by the current
process.  The free operation includes pmap_release(), which
de-constructs the paging structures used by hardware.

If the calling process is multithreaded, other threads are suspended
in the thread_suspend_check(), and need to be unsuspended and run to
be able to exit on successfull exec.  Now, since the old vmspace is
destroyed, paging structures are invalid, threads are resumed on the
non-existent pmaps (page tables), which leads to triple fault on x86.

To fix, postpone the free of old vmspace until the threads are resumed
and exited.  To avoid modifications to all image activators all of
which use exec_new_vmspace(), memoize the current (old) vmspace in
kern_execve(), and notify it about the need to call vmspace_free()
with a thread-private flag TDP_EXECVMSPC.

http://bugs.debian.org/743141

Reported by: Ivo De Decker <ivo.dedecker@ugent.be> through secteam
Sponsored by: The FreeBSD Foundation
MFC after: 3 days

10 years ago- Include /etc/newsyslog.conf.d/* and /usr/local/etc/newsyslog.conf.d/* by
bdrewery [Tue, 20 May 2014 03:00:20 +0000 (03:00 +0000)]
- Include /etc/newsyslog.conf.d/* and /usr/local/etc/newsyslog.conf.d/* by
  default for newsyslog(8).

  The /usr/local/etc/newsyslog.conf.d will give packages an opportunity to
  install a default configuration to handle their own log files.

MFC after: 2 weeks
Relnotes: yes

10 years agoBump bhyve allocation up to 20 bits to avoid
grehan [Tue, 20 May 2014 02:59:13 +0000 (02:59 +0000)]
Bump bhyve allocation up to 20 bits to avoid
birthday-paradox style address collisions when
bhyve VMs are connected to the same broadcoast
domain and are using pseudo-random allocations.

Reviewed by: gnn
MFC after: 1 week

10 years agoRevert previous change - doesn't cover all cases.
sjg [Mon, 19 May 2014 21:55:47 +0000 (21:55 +0000)]
Revert previous change - doesn't cover all cases.

10 years agoFix tst.ZeroModuleProbes.d.ksh, which was incorrectly modified in r178534.
markj [Mon, 19 May 2014 20:11:55 +0000 (20:11 +0000)]
Fix tst.ZeroModuleProbes.d.ksh, which was incorrectly modified in r178534.
Since "BEGIN" is not the name of a module, the test would just hang.

MFC after: 3 days

10 years agoLet's just use the mib0 partition for our configurations pace. It's
adrian [Mon, 19 May 2014 19:34:44 +0000 (19:34 +0000)]
Let's just use the mib0 partition for our configurations pace. It's
a convenient thing.

tested:

* AP93

10 years ago_SUBDIR is marked .MAKE - since it runs a sub-make.
sjg [Mon, 19 May 2014 19:08:46 +0000 (19:08 +0000)]
_SUBDIR is marked .MAKE - since it runs a sub-make.
Targets thus marked are supposed to run even with -n.
As such they should not do anything except run the sub-make.

Use an intermediate target _* to associate with _SUBDIR and which
depends on installincludes etc so that we get the correct behavior with -n.

Reviewed by: marcel

10 years agoAdd support for decoding the AMD SVM instructions.
jhb [Mon, 19 May 2014 18:07:37 +0000 (18:07 +0000)]
Add support for decoding the AMD SVM instructions.

10 years agoAdd -R to netstat to dump RSS/flow information.
adrian [Mon, 19 May 2014 17:11:43 +0000 (17:11 +0000)]
Add -R to netstat to dump RSS/flow information.

This is intended to help in diagnostics and debugging of NIC and stack
flowid support.

Eventually this will grow another column (RSS CPU ID) but
that currently isn't cached in the inpcb.

There's also no clean flowtype -> flowtype identifier string.  This is
the mbuf M_HASHTYPE_* values for RSS.

Here's some example output:

adrian@adrian-hackbox:~/work/freebsd/head/src % netstat -Rn | more
Active Internet connections
Proto Recv-Q Send-Q Local Address          Foreign Address           flowid ftype
tcp4       0      0 10.11.1.65.22          10.11.1.64.12409        29041942     2
udp4       0      0 127.0.0.1.123          *.*                     00000000     0
udp6       0      0 fe80::1%lo0.123        *.*                     00000000     0
udp6       0      0 ::1.123                *.*                     00000000     0
udp4       0      0 10.11.1.65.123         *.*                     00000000     0

Tested:

* amd64 system w/ igb NIC; local driver changes to expose RSS flowid in if_igb.

10 years agoAdd ARM_EABI to the list, since arm kernels need it
imp [Mon, 19 May 2014 16:13:40 +0000 (16:13 +0000)]
Add ARM_EABI to the list, since arm kernels need it

# Note: MK_ARM_EABI likely is going to die soon.

10 years agoAdd a topology trace to the g_spoil_event.
ae [Mon, 19 May 2014 16:08:15 +0000 (16:08 +0000)]
Add a topology trace to the g_spoil_event.

MFC after: 1 week

10 years agoWe have two functions from where a geom orphan method could be called:
ae [Mon, 19 May 2014 16:05:42 +0000 (16:05 +0000)]
We have two functions from where a geom orphan method could be called:
g_orphan_register and g_resize_provider_event. Both are called from the
event queue. Also we have GEOM_DEV class, which does deferred destroy
for its consumers via g_dev_destroy (also called from the event queue).
So it is possible, that for some consumers an orphan method will be
called twice. This triggers panic in g_dev_orphan.
Check that consumer isn't already orphaned before call orphan method.

MFC after: 2 weeks

10 years agoSlightly restructure the final loop in rman_reserve_resource_bound().
truckman [Mon, 19 May 2014 04:44:27 +0000 (04:44 +0000)]
Slightly restructure the final loop in rman_reserve_resource_bound().
Replace with the existing loop termination test with a similar
condition from the nested "if" that may terminate the loop a bit
sooner, but still not too early.   This condition can then be removed
from the nested "if".  Relocate an operator to be style(9) compliant.

MFC after: 3 days

10 years agoremove trailing white space...
jmg [Mon, 19 May 2014 04:40:02 +0000 (04:40 +0000)]
remove trailing white space...

fix spelling of unnecessary...

10 years agoAdd PG_U (user/supervisor) checks when translating a guest linear address
neel [Mon, 19 May 2014 03:50:07 +0000 (03:50 +0000)]
Add PG_U (user/supervisor) checks when translating a guest linear address
to a guest physical address.

PG_PS (page size) field is valid only in a PDE or a PDPTE so it is now
checked only in non-terminal paging entries.

Ignore the upper 32-bits of the CR3 for PAE paging.

10 years agoThis is the beta release of the driver for the new
jfv [Mon, 19 May 2014 01:21:02 +0000 (01:21 +0000)]
This is the beta release of the driver for the new
Intel 40G Ethernet Controller XL710 Family. This is
the core driver, a VF driver called i40evf, will be
following soon. Questions or comments to myself or
my co-developer Eric Joyner. Cheers!

10 years agoWhen RSS is enabled and per cpu TCP timers are enabled, do an RSS
adrian [Sun, 18 May 2014 22:39:01 +0000 (22:39 +0000)]
When RSS is enabled and per cpu TCP timers are enabled, do an RSS
lookup for the inp flowid/flowtype to destination CPU.

This only modifies the case where RSS is enabled and the per-cpu tcp
timer option is enabled.  Otherwise the behaviour should be the same
as before.

10 years ago* When copying the flowid from inp -> outbound mbuf, also assign the
adrian [Sun, 18 May 2014 22:37:31 +0000 (22:37 +0000)]
* When copying the flowid from inp -> outbound mbuf, also assign the
  hashtype to to the outbound mbuf as well as the flowid.

* Add in socket options to fetch the hashid, the hashtype and RSS CPU
  ID for a given socket.

10 years agoEnsure that the flowid hashtype is assigned to the inp if the flowid
adrian [Sun, 18 May 2014 22:34:06 +0000 (22:34 +0000)]
Ensure that the flowid hashtype is assigned to the inp if the flowid
is also assigned.

10 years agoAdd a new function to do a CPU ID lookup based on RSS hash information.
adrian [Sun, 18 May 2014 22:32:04 +0000 (22:32 +0000)]
Add a new function to do a CPU ID lookup based on RSS hash information.

This is intended to be used by various places that wish to hash some
information about a TCP/UDP/IP flow but don't necessarily have a
live mbuf to do it with.

Refactor rss_m2cpuid() to use the refactored function.

10 years agoAdd the flowtype to the inpcb.
adrian [Sun, 18 May 2014 22:30:12 +0000 (22:30 +0000)]
Add the flowtype to the inpcb.

The flowid isn't enough to use as part of any RSS related CPU affinity
lookups - the RSS code would like to know what kind of hash it is.

10 years agoAssorted updates to md5.1
bjk [Sun, 18 May 2014 21:16:59 +0000 (21:16 +0000)]
Assorted updates to md5.1

Note that the -c argument's parameter is compared against the digest of
the file, not the file. [1]

Update the "current time" parentheticals for notes about reversing
and colliding the hash functions. [1]

Some general mdoc updates.

PR: docs/188043 [1]
Submitted by: Jamie Landeg-Jones [1]
Approved by: hrs (mentor)
MFC after: 1 week

10 years agoDocument some user-settable make variables in ports.7
bjk [Sun, 18 May 2014 21:05:54 +0000 (21:05 +0000)]
Document some user-settable make variables in ports.7

This is not a comprehensive list, as the variables themselves are spread
out over multiple files, but it is a start.

Add a section to make.conf noting that variables may be set there that
affect ports builds, but refer to ports.7 and elsewhere for the actual
listing; any listing in make.conf.5 would likely become out of date
fairly quickly.

PR: docs/189199
Reviewed by: bdrewery (previous version)
Approved by: hrs (mentor)

10 years agoFix LD_LIBMAP.
kib [Sun, 18 May 2014 17:14:08 +0000 (17:14 +0000)]
Fix LD_LIBMAP.
The r232862 passed the pointer to end of string to lmc_parse() and free().

Submitted by: Wolfgang Jenkner <wjenkner@inode.at>
MFC after: 1 week

10 years agoUpdate the date on the manual page.
gnn [Sun, 18 May 2014 15:31:53 +0000 (15:31 +0000)]
Update the date on the manual page.

Pointed out by: bz

10 years agoSince ipfw nat configures all options in one step, we should set all bits
ae [Sun, 18 May 2014 14:25:19 +0000 (14:25 +0000)]
Since ipfw nat configures all options in one step, we should set all bits
in the mask when calling LibAliasSetMode() to properly clear unneeded
options.

PR: 189655
MFC after: 1 week
Sponsored by: Yandex LLC

10 years agoFix build after recent DWC OTG changes.
hselasky [Sun, 18 May 2014 09:29:00 +0000 (09:29 +0000)]
Fix build after recent DWC OTG changes.

10 years ago- Remove no longer used file. FDT is used to attach device drivers.
hselasky [Sun, 18 May 2014 09:19:13 +0000 (09:19 +0000)]
- Remove no longer used file. FDT is used to attach device drivers.

10 years ago- Add softc pointer argument to FIFO functions as an optimisation.
hselasky [Sun, 18 May 2014 09:13:29 +0000 (09:13 +0000)]
- Add softc pointer argument to FIFO functions as an optimisation.
- Implement support for interrupt filters in the DWC OTG driver, to
reduce the amount of CPU task switching when only feeding the FIFOs.
- Add common spinlock to the USB bus structure.

MFC after: 2 weeks

10 years agoNote my mentors in the committers graph.
allanjude [Sun, 18 May 2014 04:21:12 +0000 (04:21 +0000)]
Note my mentors in the committers graph.
Add my birthday to the calendar.

Approved by: bcr (mentor)

10 years agoAdd support for the 'rdseed' instruction.
jhb [Sun, 18 May 2014 03:57:54 +0000 (03:57 +0000)]
Add support for the 'rdseed' instruction.

10 years agoMake the vmx asm code dtrace-fbt-friendly by
grehan [Sun, 18 May 2014 03:50:17 +0000 (03:50 +0000)]
Make the vmx asm code dtrace-fbt-friendly by
 - inserting frame enter/leave sequences
 - restructuring the vmx_enter_guest routine so that it subsumes
   the vm_exit_guest block, which was the #vmexit RIP and not a
   callable routine.

Reviewed by: neel
MFC after: 3 weeks

10 years agoAdd support for decoding rdrand and rdseed.
jhb [Sat, 17 May 2014 21:10:03 +0000 (21:10 +0000)]
Add support for decoding rdrand and rdseed.

10 years agoCorrect a typo.
rstone [Sat, 17 May 2014 20:39:39 +0000 (20:39 +0000)]
Correct a typo.

MFC after: 1 week

10 years agoMove inclusion of /etc/make.conf and others to old location (this also
imp [Sat, 17 May 2014 20:31:40 +0000 (20:31 +0000)]
Move inclusion of /etc/make.conf and others to old location (this also
moves local.sys.mk). The new location broke adding things to CXXFLAGS
in /etc/src.conf with +=. Move it back until that's sorted out...

10 years agoThe time is not yet ripe to break the lack of dependencies between
imp [Sat, 17 May 2014 20:31:34 +0000 (20:31 +0000)]
The time is not yet ripe to break the lack of dependencies between
src/sys and the rest of the tree for builds.
o eliminate including bsd.mkopts.mk for the moment in kern.opts.mk
o No need to include src.opts.mk at all anymore. The reasons for it
  are now coverted in sys.mk and src.sys.mk.

10 years agoDocument r266209, pmcstat(8) '-l' flag.
gjb [Sat, 17 May 2014 19:45:23 +0000 (19:45 +0000)]
Document r266209, pmcstat(8) '-l' flag.

Sponsored by: The FreeBSD Foundation

10 years agoAllow us to compile the Ti iic driver for both OMAP4 and AM335x.
andrew [Sat, 17 May 2014 18:52:20 +0000 (18:52 +0000)]
Allow us to compile the Ti iic driver for both OMAP4 and AM335x.

MFC after: 1 week

10 years agoADd axge(4) to LINT
lwhsu [Sat, 17 May 2014 18:40:43 +0000 (18:40 +0000)]
ADd axge(4) to LINT

Approved by: markj

10 years agoMove the Ti SoCs to use the ARM platform. This should help allowing a
andrew [Sat, 17 May 2014 18:35:22 +0000 (18:35 +0000)]
Move the Ti SoCs to use the ARM platform. This should help allowing a
single kernel to work on both PandaBoard and BeagleBone.

10 years agoAdd FDT_PLATFORM_DEF2 for when there are multiple platforms needing to use
andrew [Sat, 17 May 2014 18:02:46 +0000 (18:02 +0000)]
Add FDT_PLATFORM_DEF2 for when there are multiple platforms needing to use
the same platform methods.

10 years agoAdd descriptions of the display formats for -e, -l, -r, and -x. Fix a few
jhb [Sat, 17 May 2014 16:12:50 +0000 (16:12 +0000)]
Add descriptions of the display formats for -e, -l, -r, and -x.  Fix a few
typos in indent settings while here.

10 years agoUse 'RESOURCE' instead of the more generic 'TYPE' for the resource name
jhb [Sat, 17 May 2014 16:11:06 +0000 (16:11 +0000)]
Use 'RESOURCE' instead of the more generic 'TYPE' for the resource name
column header when displaying resource usage.  This more closely matches
other procstat displays.

10 years agoAdd -o option to gstat to display "other" operatins (e.g. BIO_FLUSH).
mav [Sat, 17 May 2014 15:26:08 +0000 (15:26 +0000)]
Add -o option to gstat to display "other" operatins (e.g. BIO_FLUSH).

MFC after: 1 week

10 years agoMake GEOM DISK to account also BIO_FLUSH operations.
mav [Sat, 17 May 2014 15:07:00 +0000 (15:07 +0000)]
Make GEOM DISK to account also BIO_FLUSH operations.

10 years agoMark the i.MX51 and i.MX53 boards as compatible with the i.MX51 and i.MX53
andrew [Sat, 17 May 2014 14:57:34 +0000 (14:57 +0000)]
Mark the i.MX51 and i.MX53 boards as compatible with the i.MX51 and i.MX53
respectively.

MFC after: 1 week

10 years agoFix wrong formatting of 0.0.0.0/X table records in ipfw(8).
melifaro [Sat, 17 May 2014 13:45:03 +0000 (13:45 +0000)]
Fix wrong formatting of 0.0.0.0/X table records in ipfw(8).

Add `flags` u16 field to the hole in ipfw_table_xentry structure.
Kernel has been guessing address family for supplied record based
on xent length size.
Userland, however, has been getting fixed-size ipfw_table_xentry structures
guessing address family by checking address by IN6_IS_ADDR_V4COMPAT().

Fix this behavior by providing specific IPFW_TCF_INET flag for IPv4 records.

PR: bin/189471
Submitted by: Dennis Yusupoff <dyr@smartspb.net>
MFC after: 2 weeks

10 years agoProvide compatibility #define after r265408.
glebius [Sat, 17 May 2014 12:33:27 +0000 (12:33 +0000)]
Provide compatibility #define after r265408.

Suggested by: truckman

10 years agoo In pf_normalize_ip() we don't need mtag in
glebius [Sat, 17 May 2014 12:30:27 +0000 (12:30 +0000)]
o In pf_normalize_ip() we don't need mtag in
  !(PFRULE_FRAGCROP|PFRULE_FRAGDROP) case.
o In the (PFRULE_FRAGCROP|PFRULE_FRAGDROP) case we should allocate mtag
  if we don't find any.

Tested by: Ian FREISLICH <ianf cloudseed.co.za>

10 years agoFix a comment s/initarm_/platform_/
andrew [Sat, 17 May 2014 11:29:44 +0000 (11:29 +0000)]
Fix a comment s/initarm_/platform_/

10 years agoAdd the start of the ARM platform code. This is based on the PowerPC
andrew [Sat, 17 May 2014 11:27:36 +0000 (11:27 +0000)]
Add the start of the ARM platform code. This is based on the PowerPC
platform code, it is expected these will be merged in the future when the
ARM code is more complete.

Until more boards can be tested only use this with the Raspberry Pi and
rrename the functions on the other SoCs.

Reviewed by: ian@

10 years agoUpdate example portion of comment to coincide with r264840 changes.
dteske [Sat, 17 May 2014 04:29:14 +0000 (04:29 +0000)]
Update example portion of comment to coincide with r264840 changes.

MFC after: 3 days

10 years agoCorrect some minor nits in the per-thread signal format description such
jhb [Sat, 17 May 2014 04:06:54 +0000 (04:06 +0000)]
Correct some minor nits in the per-thread signal format description such
as missing posessives and misordering of fields.

MFC after: 1 week

10 years ago- Add support for dumping current resource usage for processes via a new -r
jhb [Sat, 17 May 2014 03:54:51 +0000 (03:54 +0000)]
- Add support for dumping current resource usage for processes via a new -r
  flag to procstat.
- Add an -H flag to request information about threads rather than processes
  when dumping statistics.  Currently it is only used for -r to display
  resource usage for individual threads instead of the entire process.

Reviewed by: kib (older version without -H)
MFC after: 1 month

10 years agoLook for root certificates in /usr/local/etc/ssl before /etc/ssl.
des [Sat, 17 May 2014 03:39:56 +0000 (03:39 +0000)]
Look for root certificates in /usr/local/etc/ssl before /etc/ssl.

MFH: 1 week

10 years agoCorrect documentation of the limit on how much memory can be mlock()ed
bjk [Sat, 17 May 2014 03:05:52 +0000 (03:05 +0000)]
Correct documentation of the limit on how much memory can be mlock()ed

vm.max_wired is a system-wide limit, not per-process.  Reword the
section to make this more clear.

PR: docs/189214
Submitted by: Lawrence Chen (original text)
Approved by: hrs (mentor)

10 years agoBump .Dd, missed in r266206
bjk [Sat, 17 May 2014 02:53:10 +0000 (02:53 +0000)]
Bump .Dd, missed in r266206

Approved by: hrs (mentor, implicit)

10 years agoClear the data buffer length field when freeing a command structure so that
jhb [Sat, 17 May 2014 02:45:04 +0000 (02:45 +0000)]
Clear the data buffer length field when freeing a command structure so that
it doesn't leak through when the command structure is reused for a user
command without a data buffer.

PR: amd64/189668
Tested by: Pete Long <pete@nrth.org>
MFC after: 1 week

10 years agoRead in SRCCONF early and consistently, if src.sys.mk exists, which is
imp [Sat, 17 May 2014 01:47:06 +0000 (01:47 +0000)]
Read in SRCCONF early and consistently, if src.sys.mk exists, which is
should for all normal builds. Read /etc/make.conf earlier than before,
but consistently before SRCCONF and local.sys.mk.

10 years agoReserve IP_FLOWID, IP_FLOWTYPE, IP_RSSCPUID socket option IDs for
adrian [Sat, 17 May 2014 00:09:12 +0000 (00:09 +0000)]
Reserve IP_FLOWID, IP_FLOWTYPE, IP_RSSCPUID socket option IDs for
near-term future use.

These are intended to fetch the current flow id, flow hash type
(M_HASHTYPE_* from the sys/mbuf.h) and if RSS is enabled, the
RSS destined CPU ID for the receive path.

10 years agoFix spelling mistake in comment.
gavin [Fri, 16 May 2014 21:20:13 +0000 (21:20 +0000)]
Fix spelling mistake in comment.

Spotted during: http://www.bsdcan.org/2014/schedule/events/484.en.html

10 years agoRemove some unused variables.
brueffer [Fri, 16 May 2014 21:19:17 +0000 (21:19 +0000)]
Remove some unused variables.

Found with: Clang Static Analyzer
MFC after: 2 weeks

10 years agoFix an issue in range specification handling when a "-foo" is specified in
hrs [Fri, 16 May 2014 18:44:23 +0000 (18:44 +0000)]
Fix an issue in range specification handling when a "-foo" is specified in
ifconfig_IF_aliasN.

10 years agoAdd definitions for more structured extended features as well as
jhb [Fri, 16 May 2014 17:45:09 +0000 (17:45 +0000)]
Add definitions for more structured extended features as well as
XSAVE Extended Features for AVX512 and MPX (Memory Protection Extensions).

Obtained from: Intel's Instruction Set Extensions Programming Reference
                (March 2014)

10 years agoEnable host controller interrupts.
hselasky [Fri, 16 May 2014 16:36:07 +0000 (16:36 +0000)]
Enable host controller interrupts.

Sponsored by: DARPA, AFRL

10 years agoUSB endpoints are almost always single-digits, fix the path in the man
gavin [Fri, 16 May 2014 16:28:09 +0000 (16:28 +0000)]
USB endpoints are almost always single-digits, fix the path in the man
page to be clearer.

PR: docs/175560
Submitted by: Andreas Gustafsson <gson gson.org>
MFC after: 3 days

10 years agoRemove old files.
hselasky [Fri, 16 May 2014 15:53:47 +0000 (15:53 +0000)]
Remove old files.

Sponsored by: DARPA, AFRL

10 years agoCancel DAD for an ifa when the ifp has ND6_IFF_IFDISABLED as early as
hrs [Fri, 16 May 2014 15:53:31 +0000 (15:53 +0000)]
Cancel DAD for an ifa when the ifp has ND6_IFF_IFDISABLED as early as
possible and do not clear IN6_IFF_TENTATIVE.  If IFDISABLED was accidentally
set after a DAD started, TENTATIVE could be cleared because no NA was
received due to IFDISABLED, and as a result it could prevent DAD when
manually clearing IFDISABLED after that.

10 years agoFix a compile warning about unused variable.
hselasky [Fri, 16 May 2014 15:53:14 +0000 (15:53 +0000)]
Fix a compile warning about unused variable.

Sponsored by: DARPA, AFRL

10 years agoRename "saf1761_dci_xxx" into "saf1761_otg_xxx" to reflect that this
hselasky [Fri, 16 May 2014 15:50:21 +0000 (15:50 +0000)]
Rename "saf1761_dci_xxx" into "saf1761_otg_xxx" to reflect that this
driver supports both host and device side mode.

Sponsored by: DARPA, AFRL

10 years agoImplement basic support for the USB host controller found in the
hselasky [Fri, 16 May 2014 15:41:55 +0000 (15:41 +0000)]
Implement basic support for the USB host controller found in the
SAF1761 chip, supporting BULK and CONTROL endpoints. This code is not
yet tested.

Sponsored by: DARPA, AFRL

10 years agoDon't cast away the const qualifier. Fixes gcc build.
des [Fri, 16 May 2014 15:32:16 +0000 (15:32 +0000)]
Don't cast away the const qualifier.  Fixes gcc build.

10 years agoMove configuration of IPv6 NDP flags to a point before handling ifconfig_IF.
hrs [Fri, 16 May 2014 14:48:21 +0000 (14:48 +0000)]
Move configuration of IPv6 NDP flags to a point before handling ifconfig_IF.
This fixes a race that a non-IPv4 interface can get an EUI64 LLA even if it
has IFDISABLED nd6 flag at boot time.

10 years ago- Allow the SAF1761 driver to attach to the root HUB USB driver.
hselasky [Fri, 16 May 2014 10:37:25 +0000 (10:37 +0000)]
- Allow the SAF1761 driver to attach to the root HUB USB driver.

Sponsored by: DARPA, AFRL

10 years ago- Add flattended device tree probe-, attach- and detach code for the
hselasky [Fri, 16 May 2014 10:35:21 +0000 (10:35 +0000)]
- Add flattended device tree probe-, attach- and detach code for the
SAF1761 driver, compatible to existing Linux based FDT tables for the
same hardware.

Sponsored by: DARPA, AFRL