]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
11 years agoPlug memory leak in dsl_check_snap_cb()
mm [Tue, 19 Mar 2013 07:47:51 +0000 (07:47 +0000)]
Plug memory leak in dsl_check_snap_cb()
This was unnoticed because the function is very rarely used.

MFC after: 3 days

11 years agomdoc: remove superfluous paragraph macro.
joel [Tue, 19 Mar 2013 07:25:58 +0000 (07:25 +0000)]
mdoc: remove superfluous paragraph macro.

11 years agoSeparate the locking macros that are used in the packet flow path
ae [Tue, 19 Mar 2013 06:04:17 +0000 (06:04 +0000)]
Separate the locking macros that are used in the packet flow path
from others. This helps easy switch to use pfil(4) lock.

11 years agoFix style and comments.
ae [Tue, 19 Mar 2013 05:51:47 +0000 (05:51 +0000)]
Fix style and comments.

11 years agoThere are actually two different cases when mlock(2) returns
glebius [Tue, 19 Mar 2013 05:44:25 +0000 (05:44 +0000)]
There are actually two different cases when mlock(2) returns
ENOMEM. Clarify this, taking text from SUS.

Reviewed by: kib

11 years agoFix typo in previous commit: Exit if */dev/dumpdev* does not exist, not if
cperciva [Tue, 19 Mar 2013 05:08:25 +0000 (05:08 +0000)]
Fix typo in previous commit: Exit if */dev/dumpdev* does not exist, not if
*/bin/realpath* does not exist...

Submitted by: markj
Pointy hat to: cperciva

11 years agoIf dumpdev is AUTO but no dump device has been set -- i.e., there is no swap
cperciva [Tue, 19 Mar 2013 04:42:04 +0000 (04:42 +0000)]
If dumpdev is AUTO but no dump device has been set -- i.e., there is no swap
space configured for rc.d/dumpon to designate for dumping -- then exit
silently rather than with a
> realpath: /dev/dumpdev: No such file or directory
error message.

An argument could be made that we should print a (more informative) warning
message; but given that under the same conditions the rc.d/dumpon script will
already print a
> No suitable dump device was found
warning, it seems that printing an additional
> Dump device does not exist.  Savecore not run.
warning would be superfluous.

11 years agoFix the powerpc64 build. MACHINE_CPUARCH is common for powerpc/powerpc64,
jhibbits [Tue, 19 Mar 2013 00:39:02 +0000 (00:39 +0000)]
Fix the powerpc64 build.  MACHINE_CPUARCH is common for powerpc/powerpc64,
not MACHINE_ARCH.

11 years agoDon't hesitate to ask parent to setup IRQ finally.
ray [Mon, 18 Mar 2013 23:51:39 +0000 (23:51 +0000)]
Don't hesitate to ask parent to setup IRQ finally.

Sponsored by: The FreeBSD Foundation

11 years agoAdd bhyve to examples.
neel [Mon, 18 Mar 2013 23:46:14 +0000 (23:46 +0000)]
Add bhyve to examples.

Requested by: alfred, julian
Obtained from: NetApp

11 years agoAllow simplebus to attach to another simplebus.
ray [Mon, 18 Mar 2013 23:41:19 +0000 (23:41 +0000)]
Allow simplebus to attach to another simplebus.

Sponsored by: The FreeBSD Foundation

11 years agoHide "no default resources for" warning under bootverbose. It's ok to use
ray [Mon, 18 Mar 2013 23:38:15 +0000 (23:38 +0000)]
Hide "no default resources for" warning under bootverbose. It's ok to use
optional resources.

Sponsored by: The FreeBSD Foundation

11 years agoAllow simplebus to attach in less strict way, when "simple-bus" listed on not
ray [Mon, 18 Mar 2013 23:35:01 +0000 (23:35 +0000)]
Allow simplebus to attach in less strict way, when "simple-bus" listed on not
first position of compatible property, so simplebus driver can be generic
driver for any bus listed as compatible with "simple-bus".

Sponsored by: The FreeBSD Foundation

11 years agoList TrackPoint device before generic model.
jkim [Mon, 18 Mar 2013 23:31:22 +0000 (23:31 +0000)]
List TrackPoint device before generic model.

11 years agoAdd preliminary support for IBM/Lenovo TrackPoint.
jkim [Mon, 18 Mar 2013 23:22:47 +0000 (23:22 +0000)]
Add preliminary support for IBM/Lenovo TrackPoint.

PR: kern/147237 (based on the initial patch for 8.x)
Tested by: glebius (device detection and suspend/resume)
MFC after: 1 month

11 years agoSimplify the assignment of memory to virtual machines by requiring a single
neel [Mon, 18 Mar 2013 22:38:30 +0000 (22:38 +0000)]
Simplify the assignment of memory to virtual machines by requiring a single
command line option "-m <memsize in MB>" to specify the memory size.

Prior to this change the user needed to explicitly specify the amount of
memory allocated below 4G (-m <lowmem>) and the amount above 4G (-M <highmem>).

The "-M" option is no longer supported by 'bhyveload' and 'bhyve'.

The start of the PCI hole is fixed at 3GB and cannot be directly changed
using command line options. However it is still possible to change this in
special circumstances via the 'vm_set_lowmem_limit()' API provided by
libvmmapi.

Submitted by: Dinakar Medavaram (initial version)
Reviewed by: grehan
Obtained from: NetApp

11 years agoReduce stack usage.
pjd [Mon, 18 Mar 2013 21:11:31 +0000 (21:11 +0000)]
Reduce stack usage.

11 years agoCorrect the definition for Exar XR17V258IV: we must use a config_function
rstone [Mon, 18 Mar 2013 19:22:51 +0000 (19:22 +0000)]
Correct the definition for Exar XR17V258IV: we must use a config_function
to specify the offset into the PCI memory spare at which each serial port
will find its registers.  This was already done for other Exar PCI serial
devices; it was accidentally omitted for this specific device.

Sponsored by: Sandvine Incorporated
MFC after: 1 week

11 years agoTweak some comments.
jhb [Mon, 18 Mar 2013 18:04:09 +0000 (18:04 +0000)]
Tweak some comments.

11 years agoPartially revert r195702. Deferring stops is now implemented via a set of
jhb [Mon, 18 Mar 2013 17:23:58 +0000 (17:23 +0000)]
Partially revert r195702.  Deferring stops is now implemented via a set of
calls to toggle TDF_SBDRY rather than passing PBDRY to individual sleep
calls.
- Remove the stop_allowed parameters from cursig() and issignal().
  issignal() checks TDF_SBDRY directly.
- Remove the PBDRY and SLEEPQ_STOP_ON_BDRY flags.

11 years agoo Switch to use physical addresses in rman for FDT.
ray [Mon, 18 Mar 2013 15:18:55 +0000 (15:18 +0000)]
o Switch to use physical addresses in rman for FDT.
o Remove vtophys used to translate virtual address to physical in case rman carry virtual.

Sponsored by: The FreeBSD Foundation

11 years agodo_vfp_vmrs and do_vfp_vmsr should not return anything.
andrew [Mon, 18 Mar 2013 15:14:36 +0000 (15:14 +0000)]
do_vfp_vmrs and do_vfp_vmsr should not return anything.

11 years agoKeep the default AuthorizedKeysFile setting. Although authorized_keys2
des [Mon, 18 Mar 2013 10:50:50 +0000 (10:50 +0000)]
Keep the default AuthorizedKeysFile setting.  Although authorized_keys2
has been deprecated for a while, some people still use it and were
unpleasantly surprised by this change.

I may revert this commit at a later date if I can come up with a way
to give users who still have authorized_keys2 files sufficient advance
warning.

MFC after: ASAP

11 years agoAdd support for the vmsr and vmrs instructions. This supports the system
andrew [Mon, 18 Mar 2013 08:22:35 +0000 (08:22 +0000)]
Add support for the vmsr and vmrs instructions. This supports the system
level version of the instructions. When used in userland the hardware only
allows us to read/write FPSCR.

11 years agoSome ARM vmov similar to 'vmov.f32 s1, s2' will incorrectly have the second
andrew [Mon, 18 Mar 2013 07:41:08 +0000 (07:41 +0000)]
Some ARM vmov similar to 'vmov.f32 s1, s2' will incorrectly have the second
register added to the symbol table by the assembler. On further
investigation it was found the problem was with the my_get_expression
function. This is called by parse_big_immediate.

Fix this by moving the call to parse_big_immediate to the end of the if,
else if, ..., else block.

11 years agoAdd new USB ID.
hselasky [Mon, 18 Mar 2013 07:02:58 +0000 (07:02 +0000)]
Add new USB ID.

PR: usb/177013
MFC after: 1 week

11 years agoAdd FBT for PowerPC DTrace. Also, clean up the DTrace assembly code,
jhibbits [Mon, 18 Mar 2013 05:30:18 +0000 (05:30 +0000)]
Add FBT for PowerPC DTrace.  Also, clean up the DTrace assembly code,
much of which is not necessary for PowerPC.

The FBT module can likely be factored into 3 separate files: common,
intel, and powerpc, rather than duplicating most of the code between
the x86 and PowerPC flavors.

All DTrace modules for PowerPC will be MFC'd together once Fasttrap is
completed.

11 years agor119712 introduced SIS_TYPE_83816 but it was not actually set in
yongari [Mon, 18 Mar 2013 04:46:17 +0000 (04:46 +0000)]
r119712 introduced SIS_TYPE_83816 but it was not actually set in
driver such that checking against the type was always false.
To detect NS DP83816, driver should have checked silicon revision
register for NS controllers. While here, remove SIS_TYPE_83816 to
not make the similar mistake again.

Reported by: Brad Smith ( brad@openbsd )

11 years agoPrint out the current fifo queue depth correctly - not just the max
adrian [Mon, 18 Mar 2013 02:29:57 +0000 (02:29 +0000)]
Print out the current fifo queue depth correctly - not just the max
queue depth.

Silly hat to me.

11 years agoAdd restrict keyword to realpath manpage.
kevlo [Mon, 18 Mar 2013 01:22:28 +0000 (01:22 +0000)]
Add restrict keyword to realpath manpage.

11 years agoDump out information about the RX descriptor free list and FIFO information.
adrian [Mon, 18 Mar 2013 01:12:36 +0000 (01:12 +0000)]
Dump out information about the RX descriptor free list and FIFO information.

11 years agoLog some more information when the RX buffer allocation failed.
adrian [Mon, 18 Mar 2013 01:11:52 +0000 (01:11 +0000)]
Log some more information when the RX buffer allocation failed.

11 years agoSync back vmcontention branch into HEAD:
attilio [Mon, 18 Mar 2013 00:25:02 +0000 (00:25 +0000)]
Sync back vmcontention branch into HEAD:
Replace the per-object resident and cached pages splay tree with a
path-compressed multi-digit radix trie.
Along with this, switch also the x86-specific handling of idle page
tables to using the radix trie.

This change is supposed to do the following:
- Allowing the acquisition of read locking for lookup operations of the
  resident/cached pages collections as the per-vm_page_t splay iterators
  are now removed.
- Increase the scalability of the operations on the page collections.

The radix trie does rely on the consumers locking to ensure atomicity of
its operations.  In order to avoid deadlocks the bisection nodes are
pre-allocated in the UMA zone.  This can be done safely because the
algorithm needs at maximum one new node per insert which means the
maximum number of the desired nodes is the number of available physical
frames themselves.  However, not all the times a new bisection node is
really needed.

The radix trie implements path-compression because UFS indirect blocks
can lead to several objects with a very sparse trie, increasing the number
of levels to usually scan.  It also helps in the nodes pre-fetching by
introducing the single node per-insert property.

This code is not generalized (yet) because of the possible loss of
performance by having much of the sizes in play configurable.
However, efforts to make this code more general and then reusable in
further different consumers might be really done.

The only KPI change is the removal of the function vm_page_splay() which
is now reaped.
The only KBI change, instead, is the removal of the left/right iterators
from struct vm_page, which are now reaped.

Further technical notes broken into mealpieces can be retrieved from the
svn branch:
http://svn.freebsd.org/base/user/attilio/vmcontention/

Sponsored by: EMC / Isilon storage division
In collaboration with: alc, jeff
Tested by: flo, pho, jhb, davide
Tested by: ian (arm)
Tested by: andreast (powerpc)

11 years agoCommit new file FreeBSD tags.
attilio [Sun, 17 Mar 2013 23:53:06 +0000 (23:53 +0000)]
Commit new file FreeBSD tags.

Sponsored by: EMC / Isilon storage division

11 years agoMFC
attilio [Sun, 17 Mar 2013 23:39:52 +0000 (23:39 +0000)]
MFC

11 years agofind: Include nanoseconds when comparing timestamps of files.
jilles [Sun, 17 Mar 2013 22:51:58 +0000 (22:51 +0000)]
find: Include nanoseconds when comparing timestamps of files.

When comparing to the timestamp of a given file using -newer, -Xnewer and
-newerXY (where X and Y are one of m, c, a, B), include nanoseconds in the
comparison.

The primaries that compare a timestamp of a file to a given value (-Xmin,
-Xtime, -newerXt) continue to compare times in whole seconds.

Note that the default value 0 of vfs.timestamp_precision almost always
causes the nanoseconds part to be 0. However, touch -d can set a timestamp
to the microsecond regardless of that sysctl.

MFC after: 1 week

11 years agoFix a couple typos.
alc [Sun, 17 Mar 2013 20:44:09 +0000 (20:44 +0000)]
Fix a couple typos.

Sponsored by: EMC / Isilon Storage Division

11 years agoThe calls to vm_radix_lookup_ge() by vm_reserv_alloc_{contig,page}() can
alc [Sun, 17 Mar 2013 20:40:31 +0000 (20:40 +0000)]
The calls to vm_radix_lookup_ge() by vm_reserv_alloc_{contig,page}() can
be eliminated.  If the calls to vm_radix_lookup_le() return NULL, then
the page at the head of the object's memq must be the page with the least
pindex greater than the specified pindex.

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

11 years agoThe M_ZERO can be eliminated from the uma_zalloc() call in
alc [Sun, 17 Mar 2013 16:49:37 +0000 (16:49 +0000)]
The M_ZERO can be eliminated from the uma_zalloc() call in
vm_radix_node_get() with a small change to vm_radix_reclaim_allnodes_int().
This change further reduced the average number of cycles per
vm_page_insert() call from 532 to 519.

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

11 years agoEliminate an intermediate buffer and some memcpy() operations, and do
ian [Sun, 17 Mar 2013 16:31:09 +0000 (16:31 +0000)]
Eliminate an intermediate buffer and some memcpy() operations, and do
DMA directly to/from the buffers passed in from higher layer drivers.

Reviewed by: gonzo

11 years agoMost allocation of pages to objects proceeds from lower to higher
alc [Sun, 17 Mar 2013 16:23:19 +0000 (16:23 +0000)]
Most allocation of pages to objects proceeds from lower to higher
indices.  Consequentially, vm_page_insert() should use
vm_radix_lookup_le() instead of vm_radix_lookup_ge().  Here's why.  In
the expected case, vm_radix_lookup_le() will quickly find a page less
than the specified key at the same radix node.  In contrast,
vm_radix_lookup_ge() is expected to return NULL, but to do that it must
examine every slot in the radix tree that is greater than the key.

Prior to this change, the average cost of a vm_page_insert() call on my
test machine was 992 cycles.  After this change, the average cost is only
532 cycles, a reduction of 46%.

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

11 years agoSimplify the interface to vm_radix_insert() by eliminating the parameter
alc [Sun, 17 Mar 2013 16:06:03 +0000 (16:06 +0000)]
Simplify the interface to vm_radix_insert() by eliminating the parameter
"index".  The content of a radix tree leaf, or at least its "key", is not
opaque to the other radix tree operations.  Specifically, they know how to
extract the "key" from a leaf.  So, eliminating the parameter "index" isn't
breaking the abstraction.  Moreover, eliminating the parameter "index"
effectively prevents the caller from passing an inconsistent "index" and
leaf to vm_radix_insert().

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

11 years agoFix typo in sysctl description
mm [Sun, 17 Mar 2013 15:53:27 +0000 (15:53 +0000)]
Fix typo in sysctl description

Reported by: Jeremy Chadwick
MFC after: 3 days

11 years agoExpand ambiguous comments some more.
attilio [Sun, 17 Mar 2013 15:27:26 +0000 (15:27 +0000)]
Expand ambiguous comments some more.

Requested by: alc

11 years agoRemove negative name cache entry pointing to the target name, which
kib [Sun, 17 Mar 2013 15:11:37 +0000 (15:11 +0000)]
Remove negative name cache entry pointing to the target name, which
could be instantiated while tdvp was unlocked.

Reported by: Rick Miller <vmiller at hostileadmin com>
Tested by: pho
MFC after: 1 week

11 years agoIn m_align() add assertions that mbuf is virgin, similar to assertions
glebius [Sun, 17 Mar 2013 07:41:14 +0000 (07:41 +0000)]
In m_align() add assertions that mbuf is virgin, similar to assertions
in M_ALIGN(), MH_ALIGN, MEXT_ALIGN() macros.

11 years agoAdd MEXT_ALIGN() macro, similar to M_ALIGN() and MH_ALIGN(), but for
glebius [Sun, 17 Mar 2013 07:39:45 +0000 (07:39 +0000)]
Add MEXT_ALIGN() macro, similar to M_ALIGN() and MH_ALIGN(), but for
mbufs with external buffer.

11 years agoIn m_megapullup() instead of reserving some space at the end of packet,
glebius [Sun, 17 Mar 2013 07:37:10 +0000 (07:37 +0000)]
In m_megapullup() instead of reserving some space at the end of packet,
m_align() it, reserving space to prepend data.

Reviewed by: mav

11 years agoFix a typo in a comment.
rpaulo [Sun, 17 Mar 2013 07:28:17 +0000 (07:28 +0000)]
Fix a typo in a comment.

11 years agoRemove EOL whitespace accidentally introduced in r248393.
joel [Sun, 17 Mar 2013 06:57:25 +0000 (06:57 +0000)]
Remove EOL whitespace accidentally introduced in r248393.

11 years agoMove example to EXAMPLES.
joel [Sun, 17 Mar 2013 06:54:04 +0000 (06:54 +0000)]
Move example to EXAMPLES.

11 years agoMove example to EXAMPLES.
joel [Sun, 17 Mar 2013 06:36:04 +0000 (06:36 +0000)]
Move example to EXAMPLES.

11 years agoIn the uart module build ofw_bus_if.h on arm along with sparc64 as LINT
andrew [Sun, 17 Mar 2013 06:33:49 +0000 (06:33 +0000)]
In the uart module build ofw_bus_if.h on arm along with sparc64 as LINT
fails when built locally without it.

11 years agoAdd a macro that gets the physical address of a memory mapped device
ian [Sun, 17 Mar 2013 03:04:43 +0000 (03:04 +0000)]
Add a macro that gets the physical address of a memory mapped device
register from a bus space resource.

Note that this macro is just for ARM, and is intended to have a short
lifespan.  The DMA engines in some SoCs need the physical address of a
memory-mapped device register as one of the arguments for the transfer.
Several scattered ad-hoc solutions have been converted to use this macro,
which now also serves to mark the places where a more complete fix needs
to be applied (after that fix has been designed).

11 years agoLink libgcc_s against compiler-rt on ARM EABI. This allows us to use all of
andrew [Sun, 17 Mar 2013 01:03:41 +0000 (01:03 +0000)]
Link libgcc_s against compiler-rt on ARM EABI. This allows us to use all of
the symbols in compiler-rt, including the ones not available in the old
libgcc. This fixes the build with clang which generates calls to funstions
that are missing from libgcc_s.

11 years agoThe -mno-apcs-frame argument is unavaliable on clang, also ignore it there.
andrew [Sun, 17 Mar 2013 00:56:17 +0000 (00:56 +0000)]
The -mno-apcs-frame argument is unavaliable on clang, also ignore it there.

11 years agoRequire CAP_SEEK if both O_APPEND and O_TRUNC flags are absent.
pjd [Sat, 16 Mar 2013 23:19:13 +0000 (23:19 +0000)]
Require CAP_SEEK if both O_APPEND and O_TRUNC flags are absent.

In other words we don't require CAP_SEEK if either O_APPEND or O_TRUNC flag is
given, because O_APPEND doesn't allow to overwrite existing data and O_TRUNC
requires CAP_FTRUNCATE already.

Sponsored by: The FreeBSD Foundation

11 years agoUpdate the tests now that absence of the O_APPEND flag requires CAP_SEEK
pjd [Sat, 16 Mar 2013 23:13:49 +0000 (23:13 +0000)]
Update the tests now that absence of the O_APPEND flag requires CAP_SEEK
capability. Add some more tests.

Sponsored by: The FreeBSD Foundation

11 years agoAdd __aeabi_memset to libkern, implemented using memset, as clang may
andrew [Sat, 16 Mar 2013 23:11:55 +0000 (23:11 +0000)]
Add __aeabi_memset to libkern, implemented using memset, as clang may
generate calls to it.

11 years agoThe mode argument for open(2)/openat(2) only makes sense if the O_CREAT flag
pjd [Sat, 16 Mar 2013 23:10:40 +0000 (23:10 +0000)]
The mode argument for open(2)/openat(2) only makes sense if the O_CREAT flag
was given.

Sponsored by: The FreeBSD Foundation

11 years agoAdd a couple of examples.
joel [Sat, 16 Mar 2013 22:53:38 +0000 (22:53 +0000)]
Add a couple of examples.

Obtained from: OpenBSD

11 years agoFix the '-Wtautological-compare' warning emitted by clang for comparing the
neel [Sat, 16 Mar 2013 22:53:05 +0000 (22:53 +0000)]
Fix the '-Wtautological-compare' warning emitted by clang for comparing the
unsigned enum type with a negative value.

Obtained from: NetApp

11 years agoAdd a note to the HISTORY section about lchflags(2) being introduced in
pjd [Sat, 16 Mar 2013 22:44:14 +0000 (22:44 +0000)]
Add a note to the HISTORY section about lchflags(2) being introduced in
FreeBSD 5.0.

11 years agoAllow vmm stats to be specific to the underlying hardware assist technology.
neel [Sat, 16 Mar 2013 22:40:20 +0000 (22:40 +0000)]
Allow vmm stats to be specific to the underlying hardware assist technology.
This can be done by using the new macros VMM_STAT_INTEL() and VMM_STAT_AMD().
Statistic counters that are common across the two are defined using VMM_STAT().

Suggested by: Anish Gupta
Discussed with: grehan
Obtained from: NetApp

11 years agoFix version in the .Fx macro.
pluknet [Sat, 16 Mar 2013 22:37:56 +0000 (22:37 +0000)]
Fix version in the .Fx macro.

Reported by: <deeptech71@gmail.com>

11 years agoStyle: Whitespace fixes.
pjd [Sat, 16 Mar 2013 22:37:30 +0000 (22:37 +0000)]
Style: Whitespace fixes.

11 years agoStyle: Remove redundant space.
pjd [Sat, 16 Mar 2013 22:36:24 +0000 (22:36 +0000)]
Style: Remove redundant space.

11 years agoCross-reference gvinum(8) instead of vinum(8).
joel [Sat, 16 Mar 2013 22:02:47 +0000 (22:02 +0000)]
Cross-reference gvinum(8) instead of vinum(8).

11 years agoRemove reference to vinum(4). The manual page was removed in r248370.
joel [Sat, 16 Mar 2013 21:50:06 +0000 (21:50 +0000)]
Remove reference to vinum(4). The manual page was removed in r248370.

11 years agoHide version string under verbose.
joel [Sat, 16 Mar 2013 17:57:00 +0000 (17:57 +0000)]
Hide version string under verbose.

Approved by: mav

11 years ago- Replace compat macros with function calls.
glebius [Sat, 16 Mar 2013 08:58:28 +0000 (08:58 +0000)]
- Replace compat macros with function calls.

11 years ago- Replace compat macros with function calls.
glebius [Sat, 16 Mar 2013 08:57:36 +0000 (08:57 +0000)]
- Replace compat macros with function calls.
- Remove superfluous cleaning of m_len after allocating.

Sponsored by: Nginx, Inc.

11 years agoContrary to what the deleted comment said, the m_move_pkthdr()
glebius [Sat, 16 Mar 2013 08:55:21 +0000 (08:55 +0000)]
Contrary to what the deleted comment said, the m_move_pkthdr()
will not smash the M_EXT and data pointer, so it is safe to
pass an mbuf with external storage procuded by m_getcl() to
m_move_pkthdr().

Reviewed by: andre
Sponsored by: Nginx, Inc.

11 years agoBelatedly remove the vinum(4) manual page. The vinum
glebius [Sat, 16 Mar 2013 08:51:47 +0000 (08:51 +0000)]
Belatedly remove the vinum(4) manual page. The vinum
manager is absent in FreeBSD since 6.0-RELEASE.

Reviewed by: joel

11 years agoChange the type of 'ndesc' from 'int' to 'uint16_t' so that descriptor index
neel [Sat, 16 Mar 2013 05:40:29 +0000 (05:40 +0000)]
Change the type of 'ndesc' from 'int' to 'uint16_t' so that descriptor index
wraparound is handled correctly.

The gory details are available here:
http://lists.freebsd.org/pipermail/freebsd-virtualization/2013-March/001119.html

This fixes a regression introduced in r247871.

Pointed out by: Bruce Evans, Chris Torek

11 years agoAdd END to ARM libkern assembly functions
andrew [Sat, 16 Mar 2013 04:08:01 +0000 (04:08 +0000)]
Add END to ARM libkern assembly functions

11 years agoMove the __aeabi_unwind_cpp_pr{0,1,2} functions to libkern so they can be
andrew [Sat, 16 Mar 2013 04:06:49 +0000 (04:06 +0000)]
Move the __aeabi_unwind_cpp_pr{0,1,2} functions to libkern so they can be
referenced in a non-debug kernel.

11 years agoThe compiler argument -mno-apcs-frame has no meaning when using EABI as we
andrew [Sat, 16 Mar 2013 03:57:46 +0000 (03:57 +0000)]
The compiler argument -mno-apcs-frame has no meaning when using EABI as we
will use aapcs frames, not apcs frames.

11 years agoImplement the required but unused __aeabi_unwind_cpp_* functions in the
andrew [Sat, 16 Mar 2013 03:50:27 +0000 (03:50 +0000)]
Implement the required but unused __aeabi_unwind_cpp_* functions in the
trampoline kernel.

11 years agoFix the indentation for a few commands that were missed or incorrectly
andrew [Sat, 16 Mar 2013 03:21:25 +0000 (03:21 +0000)]
Fix the indentation for a few commands that were missed or incorrectly
indented in r248362.

11 years agoAdjust the indentation of the trampoline compilation to make the commands
andrew [Sat, 16 Mar 2013 03:15:24 +0000 (03:15 +0000)]
Adjust the indentation of the trampoline compilation to make the commands
easier to follow.

11 years agoAdd an END macro to ARM. This is mostly used to tell gas where the bounds
andrew [Sat, 16 Mar 2013 02:48:49 +0000 (02:48 +0000)]
Add an END macro to ARM. This is mostly used to tell gas where the bounds
of the functions are when creating the EABI unwind tables.

11 years agoSort syscalls properly.
pjd [Fri, 15 Mar 2013 23:00:13 +0000 (23:00 +0000)]
Sort syscalls properly.

11 years agosh: Recognize "--" and explicitly reject options in wait builtin.
jilles [Fri, 15 Mar 2013 20:29:31 +0000 (20:29 +0000)]
sh: Recognize "--" and explicitly reject options in wait builtin.

If syntactically invalid job identifiers are to be taken as jobs that exited
with status 127, this should not apply to options, so that we can add
options later if need be.

11 years agoWhy'd I keep this here? remove it entirely now.
adrian [Fri, 15 Mar 2013 20:22:20 +0000 (20:22 +0000)]
Why'd I keep this here? remove it entirely now.

11 years agoAdd a few examples.
joel [Fri, 15 Mar 2013 20:12:54 +0000 (20:12 +0000)]
Add a few examples.

Obtained from: OpenBSD

11 years agoFix two bugs:
adrian [Fri, 15 Mar 2013 20:00:08 +0000 (20:00 +0000)]
Fix two bugs:

* when pulling frames off of the TID queue, the ATH_TID_REMOVE()
  macro decrements the axq_depth field.  So don't do it twice.

* in ath_tx_comp_cleanup_aggr(), bf wasn't being reset to bf_first
  before walking the buffer list to complete buffers; so those buffers
  will leak.

11 years agoAdd support for Exar XR17V358 8-port serial device to puc(4)
rstone [Fri, 15 Mar 2013 19:58:44 +0000 (19:58 +0000)]
Add support for Exar XR17V358 8-port serial device to puc(4)

Reviewed by: marius
Sponsored by: Sandvine Inc.
MFC after: 1 week

11 years agoSeparate the copyright lines and the informational block by a blank line.
kib [Fri, 15 Mar 2013 14:01:37 +0000 (14:01 +0000)]
Separate the copyright lines and the informational block by a blank line.

Requested by: joel
MFC after: 2 weeks

11 years ago- Use m_getcl() instead of hand allocating.
glebius [Fri, 15 Mar 2013 13:48:53 +0000 (13:48 +0000)]
- Use m_getcl() instead of hand allocating.
- Do not calculate constant length values at run time,
  CTASSERT() their sanity.
- Remove superfluous cleaning of mbuf fields after allocation.
- Replace compat macros with function calls.

Sponsored by: Nginx, Inc.

11 years agoWe can, and should use M_WAITOK here.
glebius [Fri, 15 Mar 2013 13:10:06 +0000 (13:10 +0000)]
We can, and should use M_WAITOK here.

Sponsored by: Nginx, Inc.

11 years agoAdd my copyright for the 2012 year work, in particular vn_io_fault()
kib [Fri, 15 Mar 2013 12:57:30 +0000 (12:57 +0000)]
Add my copyright for the 2012 year work, in particular vn_io_fault()
and f_offset locking.  Add required Foundation notice for r248319.

Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

11 years agoUse m_get/m_gethdr instead of compat macros.
glebius [Fri, 15 Mar 2013 12:55:30 +0000 (12:55 +0000)]
Use m_get/m_gethdr instead of compat macros.

Sponsored by: Nginx, Inc.

11 years ago- Use m_getcl() instead of hand allocating.
glebius [Fri, 15 Mar 2013 12:53:53 +0000 (12:53 +0000)]
- Use m_getcl() instead of hand allocating.

Sponsored by: Nginx, Inc.

11 years ago- Use m_getcl() instead of hand allocating.
glebius [Fri, 15 Mar 2013 12:52:59 +0000 (12:52 +0000)]
- Use m_getcl() instead of hand allocating.
- Convert panic() to KASSERT.
- Remove superfluous cleaning of mbuf fields after allocation.
- Add comment on possible use of m_get2() here.

Sponsored by: Nginx, Inc.

11 years ago- Use m_getcl() instead of hand allocating.
glebius [Fri, 15 Mar 2013 12:50:29 +0000 (12:50 +0000)]
- Use m_getcl() instead of hand allocating.
- Use m_get()/m_gethdr() instead of macros.
- Remove superfluous cleaning of mbuf fields after allocation.

Sponsored by: Nginx, Inc.

11 years agoUse m_getcl() instead of hand made allocation.
glebius [Fri, 15 Mar 2013 12:33:23 +0000 (12:33 +0000)]
Use m_getcl() instead of hand made allocation.

Sponsored by: Nginx, Inc.

11 years agoImplement the helper function vn_io_fault_pgmove(), intended to use by
kib [Fri, 15 Mar 2013 11:16:12 +0000 (11:16 +0000)]
Implement the helper function vn_io_fault_pgmove(), intended to use by
the filesystem VOP_READ() and VOP_WRITE() implementations in the same
way as vn_io_fault_uiomove() over the unmapped buffers.  Helper
provides the convenient wrapper over the pmap_copy_pages() for struct
uio consumers, taking care of the TDP_UIOHELD situations.

Sponsored by: The FreeBSD Foundation
Tested by: pho
MFC after: 2 weeks

11 years agoUse m_get() and m_getcl() instead of compat macros.
glebius [Fri, 15 Mar 2013 10:21:18 +0000 (10:21 +0000)]
Use m_get() and m_getcl() instead of compat macros.

11 years agoUse m_get2() + m_align() instead of hand made key_alloc_mbuf(). Code
glebius [Fri, 15 Mar 2013 10:20:15 +0000 (10:20 +0000)]
Use m_get2() + m_align() instead of hand made key_alloc_mbuf(). Code
examination shows, that although key_alloc_mbuf() could return chains,
the callers never use chains, so m_get2() should suffice.

Sponsored by: Nginx, Inc.