]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
9 years agocxgbe(4): Add the ability to dump mailbox commands and replies. It is
np [Tue, 16 Jun 2015 12:36:29 +0000 (12:36 +0000)]
cxgbe(4): Add the ability to dump mailbox commands and replies.  It is
enabled/disabled via bit 0 of adapter->debug_flags (which is available
at dev.t5nex.<n>.debug_flags).

MFC after: 1 week

9 years agofd: make rights a mandatory argument to fget_unlocked
mjg [Tue, 16 Jun 2015 09:52:36 +0000 (09:52 +0000)]
fd: make rights a mandatory argument to fget_unlocked

9 years agofd: don't unnecessary copy capabilities in _fget
mjg [Tue, 16 Jun 2015 09:08:30 +0000 (09:08 +0000)]
fd: don't unnecessary copy capabilities in _fget

9 years agoThe fix for the issue described in the 20150614 sendmail entry has
gshapiro [Tue, 16 Jun 2015 03:03:26 +0000 (03:03 +0000)]
The fix for the issue described in the 20150614 sendmail entry has
been been committed in revision 284436.

MFC after: 1 day

9 years agoThe import of openssl to address the FreeBSD-SA-15:10.openssl security
gshapiro [Tue, 16 Jun 2015 02:58:50 +0000 (02:58 +0000)]
The import of openssl to address the FreeBSD-SA-15:10.openssl security
advisory includes a change which rejects handshakes with DH parameters
below 768 bits.  sendmail releases prior to 8.15.2 (not yet released),
defaulted to a 512 bit DH parameter setting for client connections.
This commit chages that default to 1024 bits.  sendmail 8.15.2, when
released well use a default of 2048 bits.

MFC after: 1 day

9 years agoDeshallify.
pluknet [Mon, 15 Jun 2015 23:30:54 +0000 (23:30 +0000)]
Deshallify.

9 years agoRemove extra blank lines
sjg [Mon, 15 Jun 2015 22:04:29 +0000 (22:04 +0000)]
Remove extra blank lines

9 years agoConvert to LIBADD
bapt [Mon, 15 Jun 2015 21:20:52 +0000 (21:20 +0000)]
Convert to LIBADD

9 years agoRegister libpanel into the available libraries
bapt [Mon, 15 Jun 2015 21:20:21 +0000 (21:20 +0000)]
Register libpanel into the available libraries

9 years agoSkip MAKE_PRINT_VAR_ON_ERROR unless we are doing META_MODE
sjg [Mon, 15 Jun 2015 21:09:18 +0000 (21:09 +0000)]
Skip MAKE_PRINT_VAR_ON_ERROR unless we are doing META_MODE
with all the recursion in normal build it is too much noise.

9 years agoFollow up to r284427: fix NaN mixing for ctanhf too.
tijl [Mon, 15 Jun 2015 20:47:26 +0000 (20:47 +0000)]
Follow up to r284427: fix NaN mixing for ctanhf too.

9 years ago- Change comments to be more consistent with s_ccosh.c and s_csinh.c.
tijl [Mon, 15 Jun 2015 20:40:44 +0000 (20:40 +0000)]
- Change comments to be more consistent with s_ccosh.c and s_csinh.c.
- Fix a case where NaNs were not mixed correctly and signalling NaNs were
  not converted to quiet NaNs.
- Eliminate two negations from ctan(z).

In collaboration with: bde

9 years agoFix some exceptional cases where the sign of the result is unspecified
tijl [Mon, 15 Jun 2015 20:16:53 +0000 (20:16 +0000)]
Fix some exceptional cases where the sign of the result is unspecified
but must still satisfy csinh(conj(z)) == conj(csinh(z)) and csinh(-z) ==
-csinh(z).  This allows eliminating two negations from csin(z).

In collaboration with: bde

9 years agoSkip src component if /usr/src is empty.
delphij [Mon, 15 Jun 2015 20:12:15 +0000 (20:12 +0000)]
Skip src component if /usr/src is empty.

Differential Revision: https://reviews.freebsd.org/D2364
Submitted by: kczekirda
Reviewed by: cperciva, delphij, nwhitehorn, allanjude
MFC after: 2 weeks

9 years agoPay attention to MK_ELFTOOLCHAIN_TOOLS so we build the desired tools.
sjg [Mon, 15 Jun 2015 20:11:15 +0000 (20:11 +0000)]
Pay attention to MK_ELFTOOLCHAIN_TOOLS so we build the desired tools.

9 years agoFix some exceptional cases where the sign of the result is unspecified
tijl [Mon, 15 Jun 2015 20:11:06 +0000 (20:11 +0000)]
Fix some exceptional cases where the sign of the result is unspecified
but must still satisfy ccosh(conj(z)) == conj(ccosh(z)) and ccosh(-z) ==
ccosh(z).

In collaboration with: bde

9 years agoChange -Wl,-r back to -r so gcc doesn't pass --relax to the linker on some
tijl [Mon, 15 Jun 2015 19:48:28 +0000 (19:48 +0000)]
Change -Wl,-r back to -r so gcc doesn't pass --relax to the linker on some
architectures.

PR: 200881

9 years agoRevert r284417 it is not necessary anymore
bapt [Mon, 15 Jun 2015 19:28:07 +0000 (19:28 +0000)]
Revert r284417 it is not necessary anymore

9 years agodo not include src.opts.mk from here
sjg [Mon, 15 Jun 2015 19:23:47 +0000 (19:23 +0000)]
do not include src.opts.mk from here

9 years agoDon't use ${.OBJDIR}/ to qualify target that isn't specified that way
sjg [Mon, 15 Jun 2015 18:43:32 +0000 (18:43 +0000)]
Don't use ${.OBJDIR}/ to qualify target that isn't specified that way

9 years agoEnforce overwritting SHLIBDIR
bapt [Mon, 15 Jun 2015 15:34:20 +0000 (15:34 +0000)]
Enforce overwritting SHLIBDIR

Since METAMODE has been added, sys.mk loads bsd.mkopt.mk which ends load loading
bsd.own.mk which then defines SHLIBDIR before all the Makefile.inc everywhere.

This makes /lib being populated again.

Reported by: many

9 years agottm_vm_page_alloc: use vm_page_alloc for pages without dma32 restriction
avg [Mon, 15 Jun 2015 13:43:23 +0000 (13:43 +0000)]
ttm_vm_page_alloc: use vm_page_alloc for pages without dma32 restriction

This change re-organizes code a little bit to extract common pieces
of ttm_alloc_new_pages() and ttm_get_pages() into dedicated functions.
Also, for requests without address restrictions regular vm_page_alloc()
is used.
Lastly, when vm_page_alloc_contig() fails we call VM_WAIT before calling
vm_pageout_grow_cache() to ensure that there is enough free pages at all.

Reviewed by: kib
MFC after: 15 days

9 years agoFix circular dependency between libzfs and libzfs_core
bapt [Mon, 15 Jun 2015 10:48:48 +0000 (10:48 +0000)]
Fix circular dependency between libzfs and libzfs_core

libzfs_core is the wrapper around kernel ioctls, the ioctl compat code belongs
to it

9 years agoEnsure TESTSDIR is defined before bsd.test.mk is .include'd
ngie [Mon, 15 Jun 2015 07:13:23 +0000 (07:13 +0000)]
Ensure TESTSDIR is defined before bsd.test.mk is .include'd

MFC after: 1 week

9 years agoAdd ioctl.c, linux_syscalls.c, linux32_syscalls.c to beforedepend target.
rodrigc [Mon, 15 Jun 2015 06:48:43 +0000 (06:48 +0000)]
Add ioctl.c, linux_syscalls.c, linux32_syscalls.c to beforedepend target.

These files need to be generated before mkdep is run.

9 years agoUse cpp -I$includedir
rodrigc [Mon, 15 Jun 2015 06:44:22 +0000 (06:44 +0000)]
Use cpp -I$includedir

"cpp -I$1" was expanding to "cpp -Iprint"
which was not the proper directory
of header files.

9 years agoRemove ALLOW_DEPRECATED_ATF_TOOLS/ATFFILE support from atf.test.mk
ngie [Mon, 15 Jun 2015 06:38:59 +0000 (06:38 +0000)]
Remove ALLOW_DEPRECATED_ATF_TOOLS/ATFFILE support from atf.test.mk

The legacy atf tools were removed in atf 0.20

MFC after: 2 weeks

9 years agoAdd a quick (?) note for users who may be having sendmail interoperability issues
gshapiro [Mon, 15 Jun 2015 04:18:29 +0000 (04:18 +0000)]
Add a quick (?) note for users who may be having sendmail interoperability issues
due to the recent (FreeBSD-SA-15:10.openssl) OpenSSL change to reject 512 bit
DH parameters.  Affects 11-CURRENT and 10-STABLE.

9 years agoMFV of 284395,tzdata2015e
edwin [Mon, 15 Jun 2015 00:59:47 +0000 (00:59 +0000)]
MFV of 284395,tzdata2015e

Update to tzdata2015e:

  Changes affecting future time stamps

    Morocco will suspend DST from 2015-06-14 03:00 through 2015-07-19 02:00,
    not 06-13 and 07-18 as we had guessed.  (Thanks to Milamber.)

    Assume Cayman Islands will observe DST starting next year, using US rules.
    Although it isn't guaranteed, it is the most likely.

9 years agoIf MK_META_MODE is unset, assume a value of no.
rodrigc [Sun, 14 Jun 2015 22:36:27 +0000 (22:36 +0000)]
If MK_META_MODE is unset, assume a value of no.

This is needed to build HEAD on FreeBSD 10.1,
which has bmake, but does not have
/usr/share/mk/src.opts.mk

Reviewed by: sjg
Reported by: jenkins

9 years agoCorrectly detect the case where the last address is removed.
tuexen [Sun, 14 Jun 2015 22:14:00 +0000 (22:14 +0000)]
Correctly detect the case where the last address is removed.

MFC after: 3 days

9 years agoChange the documentation to reflect where the -V option should be used
bapt [Sun, 14 Jun 2015 21:50:38 +0000 (21:50 +0000)]
Change the documentation to reflect where the -V option should be used

Suggested by: kientzle

9 years agoInvalid pages do not need neither update of the activation count nor
kib [Sun, 14 Jun 2015 20:23:41 +0000 (20:23 +0000)]
Invalid pages do not need neither update of the activation count nor
they coould be dirty.  Move the handling if the invalid pages in the
inactive scan earlier.

Remove some code duplication in the scan by introducing the
'drop_page' label, which centralizes the object and the page unlock.

Suggested and reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

9 years agoFix name of a constant.
tuexen [Sun, 14 Jun 2015 20:16:36 +0000 (20:16 +0000)]
Fix name of a constant.

MFC after: 3 days

9 years agoFix typo in comment.
kib [Sun, 14 Jun 2015 19:19:46 +0000 (19:19 +0000)]
Fix typo in comment.

MFC after: 3 days

9 years agoStop the heartbeat timer when removing a net.
tuexen [Sun, 14 Jun 2015 17:48:44 +0000 (17:48 +0000)]
Stop the heartbeat timer when removing a net.
Thanks to the reporter of
https://code.google.com/p/sctp-refimpl/issues/detail?id=14
for reporting the issue.

MFC after: 3 days

9 years agoLatest clang uses openat(2).
sjg [Sun, 14 Jun 2015 16:31:06 +0000 (16:31 +0000)]
Latest clang uses openat(2).

If the pathname is absolute or dirfd is AT_FDCWD we can
handle it exactly like open(2).
Otherwise we output an A record to indicate that the path of
an open directory needs to be used (earlier in the trace).

Differential Revision: D2810
Reviewed by: jhb
MFC after: a bit

9 years agoTell the compiler that -dc and -r are linker flags.
tijl [Sun, 14 Jun 2015 15:40:17 +0000 (15:40 +0000)]
Tell the compiler that -dc and -r are linker flags.

Reviewed by: adrian

9 years agofd: reduce excessive zeroing on fd close
mjg [Sun, 14 Jun 2015 14:10:05 +0000 (14:10 +0000)]
fd: reduce excessive zeroing on fd close

fde_file as NULL is already an indicator of an unused fd. All other
fields are populated when fp is installed.

9 years agofd: move out actual fp installation to _finstall
mjg [Sun, 14 Jun 2015 14:08:52 +0000 (14:08 +0000)]
fd: move out actual fp installation to _finstall

Use it in fd passing functions as the first step towards fd code cleanup.

9 years agonit: Rename racct_alloc_resource to racct_adjust_resource.
jlh [Sun, 14 Jun 2015 08:33:14 +0000 (08:33 +0000)]
nit: Rename racct_alloc_resource to racct_adjust_resource.

This is more accurate as the amount can be negative.

MFC after: 2 weeks

9 years agoNetBSD commit log:
jlh [Sun, 14 Jun 2015 07:47:18 +0000 (07:47 +0000)]
NetBSD commit log:
  Use a constant array for the MIB. Newer LLVM decided that mib[] warranted
  stack protections, with the obvious crash after the setup was done.
  As a positive side effect, code size shrinks a bit.

I'm not sure why this hasn't bitten us yes, but it is certainly possible and
there are no real drawbacks to this change anyway.

Submitted by: pfg
Obtained from: NetBSD
MFC after: 1 week

9 years agoAs the next step in eliminating PG_CACHE pages, free rather than cache
alc [Sun, 14 Jun 2015 05:23:39 +0000 (05:23 +0000)]
As the next step in eliminating PG_CACHE pages, free rather than cache
pages in vm_pageout_scan().  The reactivation rate of cache pages created
by vm_pageout_scan() is extremely low; typically no more than 0.5% to
2.25% of the pages are ever reactivated.  At the same time, caching pages
is more expensive than freeing them.  For example, in a test with
PostgreSQL, this change reduced the amount of time spent in the inactive
queue scan by 1/6.

Differential Revision: https://reviews.freebsd.org/D2805
Reviewed by: kib
Sponsored by: EMC / Isilon Storage Division

9 years agoNO_WARNS when building for host
sjg [Sun, 14 Jun 2015 03:34:09 +0000 (03:34 +0000)]
NO_WARNS when building for host

9 years agoAvoid circular dependency with nm
sjg [Sun, 14 Jun 2015 03:33:27 +0000 (03:33 +0000)]
Avoid circular dependency with nm

9 years agoNeed to provide -I for meta mode
sjg [Sun, 14 Jun 2015 03:32:49 +0000 (03:32 +0000)]
Need to provide -I for meta mode

9 years agoNeeds libxo
sjg [Sun, 14 Jun 2015 03:31:23 +0000 (03:31 +0000)]
Needs libxo

9 years agoCreate proper targets for linux*syscalls.c
sjg [Sun, 14 Jun 2015 03:30:39 +0000 (03:30 +0000)]
Create proper targets for linux*syscalls.c

9 years agoWe cannot build makewhatis and mandoc
sjg [Sun, 14 Jun 2015 03:29:24 +0000 (03:29 +0000)]
We cannot build makewhatis and mandoc

9 years agoAvoid trailing white-space
sjg [Sun, 14 Jun 2015 03:28:48 +0000 (03:28 +0000)]
Avoid trailing white-space

9 years agoWatch out for bin/cat in host stage tree
sjg [Sun, 14 Jun 2015 03:28:25 +0000 (03:28 +0000)]
Watch out for bin/cat in host stage tree

9 years agoAvoid anoying warning
sjg [Sun, 14 Jun 2015 03:27:22 +0000 (03:27 +0000)]
Avoid anoying warning

9 years agoFix typo - this upsets llvm/gcc-4.9, but not gcc-4.2 for some reason.
adrian [Sun, 14 Jun 2015 00:31:22 +0000 (00:31 +0000)]
Fix typo - this upsets llvm/gcc-4.9, but not gcc-4.2 for some reason.

9 years agoFix incorrect library path.
adrian [Sat, 13 Jun 2015 22:36:42 +0000 (22:36 +0000)]
Fix incorrect library path.

9 years agoFix previous commit (r284357)
ngie [Sat, 13 Jun 2015 22:29:43 +0000 (22:29 +0000)]
Fix previous commit (r284357)

I forgot to convert the && to a ||

Pointyhat to: ngie
X-MFC with: r283678, r284336, r284357

9 years agoFix inverted check by skipping over the model-specific checks if the maker
ngie [Sat, 13 Jun 2015 22:27:59 +0000 (22:27 +0000)]
Fix inverted check by skipping over the model-specific checks if the maker
or product is NULL, not if they are both not NULL

Reported by: araujo, kib
X-MFC with: r283678, r284336
Pointyhat to: allanjude

9 years agoFix up crunchgen binary generation to work with external cross-build
adrian [Sat, 13 Jun 2015 22:01:21 +0000 (22:01 +0000)]
Fix up crunchgen binary generation to work with external cross-build
tools.

* Allow STRIP to be overridden by the environment
* Use CC to tie things together, not LD

Tested:

* i386, mips32

Submitted by: kan

9 years agoWhen using -L the code skips a socket if the local or foreign
tuexen [Sat, 13 Jun 2015 20:15:44 +0000 (20:15 +0000)]
When using -L the code skips a socket if the local or foreign
address is loopback. So it is shown if both are not loopback.
The man page says that it is shown if the local or foreign
address is not loopback. Change the man page to reflect the
code.

MFC after: 3 days

9 years agoAdd support for SCTP.
tuexen [Sat, 13 Jun 2015 20:11:02 +0000 (20:11 +0000)]
Add support for SCTP.

MFC after: 3 days

9 years agoAllow more than one local or remote address per socket. This is needed to
tuexen [Sat, 13 Jun 2015 20:05:20 +0000 (20:05 +0000)]
Allow more than one local or remote address per socket. This is needed to
support SCTP (and MPTCP in the future). No functional change for existing
protocols.

MFC after: 3 days

9 years agoAllow more than one socket entry for a file descriptor. This is needed
tuexen [Sat, 13 Jun 2015 20:01:54 +0000 (20:01 +0000)]
Allow more than one socket entry for a file descriptor. This is needed
for supporting 1-to-many style SCTP sockets. For other sochets, there
is no functional change.

MFC after: 3 days

9 years agoSimplify printaddr(). No functional change.
tuexen [Sat, 13 Jun 2015 19:56:04 +0000 (19:56 +0000)]
Simplify printaddr(). No functional change.

MFC after: 3 days

9 years agoremoved extra keyword
sjg [Sat, 13 Jun 2015 19:53:24 +0000 (19:53 +0000)]
removed extra keyword

9 years agoWe no longer need CSU_DIR.i386
sjg [Sat, 13 Jun 2015 19:52:35 +0000 (19:52 +0000)]
We no longer need CSU_DIR.i386
and until there is an alternative we need MK_INSTALL_AS_USER=yes

9 years agoFix panic when adding vtnet interfaces to a bridge
kp [Sat, 13 Jun 2015 19:39:21 +0000 (19:39 +0000)]
Fix panic when adding vtnet interfaces to a bridge

vtnet interfaces are always in promiscuous mode (at least if the
VIRTIO_NET_F_CTRL_RX feature is not negotiated with the host).  if_promisc() on
a vtnet interface returned ENOTSUP although it has IFF_PROMISC set. This
confused the bridge code. Instead we now accept all enable/disable promiscuous
commands (and always keep IFF_PROMISC set).

There are also two issues with the if_bridge error handling.

If if_promisc() fails it uses bridge_delete_member() to clean up. This tries to
disable promiscuous mode on the interface. That runs into an assert, because
promiscuous mode was never set in the first place. (That's the panic reported in
PR 200210.)
We can only unset promiscuous mode if the interface actually is promiscuous.
This goes against the reference counting done by if_promisc(), but only the
first/last if_promic() calls can actually fail, so this is safe.

A second issue is a double free of bif. It's already freed by
bridge_delete_member().

PR: 200210
Differential Revision: https://reviews.freebsd.org/D2804
Reviewed by: philip (mentor)

9 years agoFix the following clang 3.7.0 warnings in lib/libfetch/http.c:
dim [Sat, 13 Jun 2015 19:26:48 +0000 (19:26 +0000)]
Fix the following clang 3.7.0 warnings in lib/libfetch/http.c:

    lib/libfetch/http.c:1628:26: error: address of array 'purl->user'
    will always evaluate to 'true' [-Werror,-Wpointer-bool-conversion]
                                    aparams.user = purl->user ?
                                                   ~~~~~~^~~~ ~
    lib/libfetch/http.c:1630:30: error: address of array 'purl->pwd'
    will always evaluate to 'true' [-Werror,-Wpointer-bool-conversion]
                                    aparams.password = purl->pwd?
                                                       ~~~~~~^~~~
    lib/libfetch/http.c:1657:25: error: address of array 'url->user'
    will always evaluate to 'true' [-Werror,-Wpointer-bool-conversion]
                                    aparams.user = url->user ?
                                                   ~~~~~^~~~ ~
    lib/libfetch/http.c:1659:29: error: address of array 'url->pwd'
    will always evaluate to 'true' [-Werror,-Wpointer-bool-conversion]
                                    aparams.password = url->pwd ?
                                                       ~~~~~^~~ ~
    lib/libfetch/http.c:1669:25: error: address of array 'url->user'
    will always evaluate to 'true' [-Werror,-Wpointer-bool-conversion]
                                    aparams.user = url->user ?
                                                   ~~~~~^~~~ ~
    lib/libfetch/http.c:1671:29: error: address of array 'url->pwd'
    will always evaluate to 'true' [-Werror,-Wpointer-bool-conversion]
                                    aparams.password = url->pwd ?
                                                       ~~~~~^~~ ~

Since url->user and url->pwd are arrays, they can never be NULL, so the
checks can be removed.

Reviewed by: bapt
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D2673

9 years agoAdd META_MODE support.
sjg [Sat, 13 Jun 2015 19:20:56 +0000 (19:20 +0000)]
Add META_MODE support.

Off by default, build behaves normally.
WITH_META_MODE we get auto objdir creation, the ability to
start build from anywhere in the tree.

Still need to add real targets under targets/ to build packages.

Differential Revision:       D2796
Reviewed by: brooks imp

9 years agoFix typo when deregistering the VLAN unconfig event handler
bryanv [Sat, 13 Jun 2015 16:13:31 +0000 (16:13 +0000)]
Fix typo when deregistering the VLAN unconfig event handler

Submitted by: Masao Uebayashi <uebayasi@tombiinc.com>
MFC after: 3 days

9 years agoUse ${CRUNCHGEN} rather than bare word.
sjg [Sat, 13 Jun 2015 15:36:13 +0000 (15:36 +0000)]
Use ${CRUNCHGEN} rather than bare word.

Reviewed by: brooks

9 years agoUnbreak libxo's handling of characters not representable in current locale
kan [Sat, 13 Jun 2015 15:14:39 +0000 (15:14 +0000)]
Unbreak libxo's handling of characters not representable in current locale

The xo_format_string_direct function loops forever never advancing the
processed string pointer when it encounters a character that makes
mbrtowc fail. Make it emit '?' character instead, as it seems this is
what the code intent was, sans bugs.

Differential Revision: https://reviews.freebsd.org/D2802
Reviewed by: marcel

9 years agoelfcopy: Handle objects without a ".shstrtab" section string table
emaste [Sat, 13 Jun 2015 14:24:31 +0000 (14:24 +0000)]
elfcopy: Handle objects without a ".shstrtab" section string table

As of LLVM revision 238073, LLVM stores symbols and section names in
the same string table.  From the upstream commit mesage:

  With the scheme of naming sections like ".text.foo" where foo is a
  symbol, there is a big potential saving in using a single one.

This is a cherry-pick of ELF Tool Chain revision 3225.

Sponsored by: The FreeBSD Foundation

9 years agoAdd if_vlan, ipfw{,_nat}.ko, and libalias to the AR934X kernel config
allanjude [Sat, 13 Jun 2015 06:09:00 +0000 (06:09 +0000)]
Add if_vlan, ipfw{,_nat}.ko, and libalias to the AR934X kernel config

This makes the TP-Link WDR3600 routers more useful

Differential Revision: https://reviews.freebsd.org/D2780
Approved by: adrian
Sponsored by: ScaleEngine Inc.

9 years agoacpi_ibm.ko panics if SMBIOS information is not available
allanjude [Sat, 13 Jun 2015 05:55:26 +0000 (05:55 +0000)]
acpi_ibm.ko panics if SMBIOS information is not available

Add a check for NULL before strcmp on smbios information incase it is not populated

Differential Revision: https://reviews.freebsd.org/D2750
Reviewed by: ngie, jhb
Approved by: rpaulo
Sponsored by: ScaleEngine Inc.

9 years agoPHY LOCK acquires the hardware lock via bxe_acquire_phy_lock() and releases it via...
davidcs [Sat, 13 Jun 2015 01:28:19 +0000 (01:28 +0000)]
PHY LOCK acquires the hardware lock via bxe_acquire_phy_lock() and releases it via bxe_release_phy_lock(). It was simply acquiring a mutex earlier which can cause the PHY to use bogus values. Fixes intermittent link failures.

bxe_ioctl() completes all functions within its context as opposed to a taskqueue earlier.

bxe_handle_rx_mode_tq() no longer required. bxe_set_rx_mode() handles the functionality within its context

Submitted by:gary.zambrano@qlogic.com
MFC after:5 days

9 years agoReduce the arm/armv6 image size from 1G to 480M.
gjb [Fri, 12 Jun 2015 19:42:27 +0000 (19:42 +0000)]
Reduce the arm/armv6 image size from 1G to 480M.

Since the images are effectively mostly zeros at 1G,
reduce the size to allow installation on smaller SD
cards, such as 512Mb.

While here, stop writing the /boot.txt file on the
WANDBOARD, which isn't used anyway.

Discussed with: imp
MFC after: 3 days
Sponsored by: The FreeBSD Foundation

9 years agoFix the reporting of the PMTUD state for specific paths.
tuexen [Fri, 12 Jun 2015 18:59:29 +0000 (18:59 +0000)]
Fix the reporting of the PMTUD state for specific paths.

MFC after: 3 days

9 years agoCode cleanup.
tuexen [Fri, 12 Jun 2015 17:20:09 +0000 (17:20 +0000)]
Code cleanup.

MFC after: 3 days

9 years agoMerge OpenSSL 1.0.1o.
jkim [Fri, 12 Jun 2015 16:48:26 +0000 (16:48 +0000)]
Merge OpenSSL 1.0.1o.

9 years agoIn case of an output error, continue with the next net, don't try to
tuexen [Fri, 12 Jun 2015 16:01:41 +0000 (16:01 +0000)]
In case of an output error, continue with the next net, don't try to
continue sending on the same net.

This fixes a bug where an invalid mbuf chain was constructed, if a
full size frame of control chunks should be sent and there is a
output error.

Based on a discussion with rrs@, change move to the next net. This fixes
the bug and improves the behaviour.

Thanks to Irene Ruengeler for spending a lot of time in narrowing this
problem down.
MFC after: 3 days

9 years agoReport the values of x86 segment registers to remote debuggers.
jhb [Fri, 12 Jun 2015 15:14:08 +0000 (15:14 +0000)]
Report the values of x86 segment registers to remote debuggers.

While here, also report %eflags from the i386 trapframe.

Differential Revision: https://reviews.freebsd.org/D2743
Reviewed by: kib
Obtained from: 1 month

9 years agoEnsure that the upper 16 bits of segment registers manually saved in
jhb [Fri, 12 Jun 2015 15:06:17 +0000 (15:06 +0000)]
Ensure that the upper 16 bits of segment registers manually saved in
trapframes are cleared by explicitly pushing a zero and then moving
the segment register into the low 16 bits.  Certain Intel processors
treat a push of a segment register as a move of the segment register
into the low 16 bits leaving the upper 16 bits of the word in the
stack unchanged.

Reviewed by: kib
MFC after: 1 month

9 years agoAdd a missing word, should have been in r284290.
brueffer [Fri, 12 Jun 2015 15:03:59 +0000 (15:03 +0000)]
Add a missing word, should have been in r284290.

Submitted by: Fabian Keil

9 years agoFree the segment objects properly.
marcel [Fri, 12 Jun 2015 14:53:56 +0000 (14:53 +0000)]
Free the segment objects properly.

9 years agoGet rid of some more NO_OBJs
sjg [Fri, 12 Jun 2015 14:30:27 +0000 (14:30 +0000)]
Get rid of some more NO_OBJs

Differential Revision:       D2748

9 years agoUnbreak mouse on resume on Thinkpads when hw.psm.trackpoint_support=0,
glebius [Fri, 12 Jun 2015 13:57:04 +0000 (13:57 +0000)]
Unbreak mouse on resume on Thinkpads when hw.psm.trackpoint_support=0,
which is default.  It was broken in r281441.

It appears that set_trackpoint_parameters() call on resume disables the
mouse.  So, we need not call it on resume if hw.psm.trackpoint_support=0.

The problem is that the probe functions are used both for probing and
for reiniting on resume. And the absense of the softc parameter is used
as a mark to distinguish reinit and probe, which is quite ugly. At the
same time the softc parameter is needed to call set_trackpoint_parameters().

o Change the arguments of probefunc_t to always supply the softc, and
  use additional enum argument to tell probing from initing.
o Don't call set_trackpoint_parameters() from global doinitialize(),
  instead call it from the enable_trackpoint() only.
o In enable_synaptics() call enable_trackpoint() in both probe and
  reinit cases.

Together with:  Jan Kokemüller <jan.kokemueller gmail.com>

9 years agoRename ECAM PCI driver file.
br [Fri, 12 Jun 2015 13:54:25 +0000 (13:54 +0000)]
Rename ECAM PCI driver file.

Requested by: imp

9 years agoAdd generic ECAM PCI device driver found in Gem5 simulator.
br [Fri, 12 Jun 2015 13:16:50 +0000 (13:16 +0000)]
Add generic ECAM PCI device driver found in Gem5 simulator.
Work based on Cavium Thunder PCIe driver by Semihalf.

Reviewed by: andrew, jhb
Sponsored by: HEIF5
Differential Revision: https://reviews.freebsd.org/D2386

9 years agoA miss from r284310. vm_pager_get_pages() updates the array, so there is
glebius [Fri, 12 Jun 2015 13:15:14 +0000 (13:15 +0000)]
A miss from r284310. vm_pager_get_pages() updates the array, so there is
no need for vm_page_lookup().

9 years agoWe need to handle 64-bit BARs ourselves to avoid that the
marcel [Fri, 12 Jun 2015 12:27:10 +0000 (12:27 +0000)]
We need to handle 64-bit BARs ourselves to avoid that the
PCI infrastructure instantiates a non-existent resource.
This has BARs suddenly show up with pciconf(8) under
VMware as well.  Now that we read the BAR ourselves, ask
for the correct resource type.

9 years agoMake KPI of vm_pager_get_pages() more strict: if a pager changes a page
glebius [Fri, 12 Jun 2015 11:32:20 +0000 (11:32 +0000)]
Make KPI of vm_pager_get_pages() more strict: if a pager changes a page
in the requested array, then it is responsible for disposition of previous
page and is responsible for updating the entry in the requested array.
Now consumers of KPI do not need to re-lookup the pages after call to
vm_pager_get_pages().

Reviewed by: kib
Sponsored by: Netflix
Sponsored by: Nginx, Inc.

9 years agozfs clone should not mount the clone if canmount == noauto
avg [Fri, 12 Jun 2015 11:21:35 +0000 (11:21 +0000)]
zfs clone should not mount the clone if canmount == noauto

Creation of a new filesystem does not imply an intent to mount it.

Since canmount property is not inherited and its default value is 'on',
the only scenario where this matters is zfs clone -o canmount=noauto.
zfs create -o canmount=noauto already does not mount the new filesystem.

Also see:
https://www.illumos.org/issues/5984
https://reviews.csiden.org/r/228/
https://github.com/FransUrbo/zfs/commit/dd0e0e69f5b1c83bf2895ac00a0b83af77473175
https://github.com/zfsonlinux/zfs/issues/2241

Reviewed by: mahrens
MFC after: 8 days
Sponsored by: ClusterHQ

9 years agoMFV r284042: 1778 Assertion failed: rn->rn_nozpool == B_FALSE, file
avg [Fri, 12 Jun 2015 11:16:43 +0000 (11:16 +0000)]
MFV r284042: 1778 Assertion failed: rn->rn_nozpool == B_FALSE, file
../common/libzfs_import.c, line 1077, function zpool_open_func

illumos/illumos-gate@bd0f709169e67f4bd34526e186a7c34f595f0d9b

Author: Andrew Stormont <andyjstormont@gmail.com>
MFC after: 13 days

9 years agoMFV r284036: 5961 Fix stack overflow in zfs_create_fs
avg [Fri, 12 Jun 2015 11:10:49 +0000 (11:10 +0000)]
MFV r284036: 5961 Fix stack overflow in zfs_create_fs

illumos/illumos-gate@c701fde6911c957e71b37aac4daf672bd828f4d7

Author: glebius
MFC after: 11 days

9 years agoMFV r284030: 5818 zfs {ref}compressratio is incorrect with 4k sector size
avg [Fri, 12 Jun 2015 10:57:05 +0000 (10:57 +0000)]
MFV r284030: 5818 zfs {ref}compressratio is incorrect with 4k sector size

illumos/illumos-gate@81cd5c555f505484180a62ca5a2fbb00d70c57d6

Author: Matthew Ahrens <mahrens@delphix.com>
MFC after: 17 days

9 years agoMFV r283534: 5515 dataset user hold doesn't reject empty tags
avg [Fri, 12 Jun 2015 10:52:53 +0000 (10:52 +0000)]
MFV r283534: 5515 dataset user hold doesn't reject empty tags

illumos/illumos-gate@752fd8dabccac68d6d09f82f3bf3561e055e400b

Author: Josef 'Jeff' Sipek <josef.sipek@nexenta.com>
MFC after: 10 days

9 years agoMFV r284040: check that datasets are snapshots
avg [Fri, 12 Jun 2015 10:41:24 +0000 (10:41 +0000)]
MFV r284040: check that datasets are snapshots

5946 zfs_ioc_space_snaps must check that firstsnap and lastsnap refer to snapshots
5945 zfs_ioc_send_space must ensure that fromsnap refers to a snapshot
Reviewed by: Steven Hartland <killing@multiplay.co.uk>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Gordon Ross <gordon.ross@nexenta.com>

illumos/illumos-gate@24218bebb460e4015fac2c9f2cec1902eddbcd7b

Note that the upstream commit is modified during MFV: in the upstream
the check is done by inspecting ds_is_snapshot field while in FreeBSD
we call dsl_dataset_is_snapshot().
This is because illumos/illumos-gate@bc9014e6a81272073b9854d9f65dd59e18d18c35
(r277428 in vendor-sys/illumos) is not MFV-ed yet.

MFC after: 10 days

9 years agoseveral lockstat improvements
avg [Fri, 12 Jun 2015 10:01:24 +0000 (10:01 +0000)]
several lockstat improvements

0. For spin events report time spent spinning, not a loop count.
While loop count is much easier and cheaper to obtain it is hard
to reason about the reported numbers, espcially for adaptive locks
where both spinning and sleeping can happen.
So, it's better to compare apples and apples.

1. Teach lockstat about FreeBSD rw locks.
This is done in part by changing the corresponding probes
and in part by changing what probes lockstat should expect.

2. Teach lockstat that rw locks are adaptive and can spin on FreeBSD.

3. Report lock acquisition events for successful rw try-lock operations.

4. Teach lockstat about FreeBSD sx locks.
Reporting of events for those locks completely mirrors
rw locks.

5. Report spin and block events before acquisition event.
This is behavior documented for the upstream, so it makes sense to stick
to it.  Note that because of FreeBSD adaptive lock implementations
both the spin and block events may be reported for the same acquisition
while the upstream reports only one of them.

Differential Revision: https://reviews.freebsd.org/D2727
Reviewed by: markj
MFC after: 17 days
Relnotes: yes
Sponsored by: ClusterHQ

9 years agoxen-blk{front/back}: remove broken FreeBSD extensions
royger [Fri, 12 Jun 2015 07:50:34 +0000 (07:50 +0000)]
xen-blk{front/back}: remove broken FreeBSD extensions

The FreeBSD extension adds a new request type, called blkif_segment_block
which has a size of 112bytes for both i386 and amd64. This is fine on
amd64, since requests have a size of 112B there also. But this is not true
for i386, where requests have a size of 108B. So on i386 we basically
overrun the ring slot when queuing a request of type blkif_segment_block_t,
which is very bad.

Remove this extension (including a cleanup of the public blkif.h header
file) from blkfront and blkback.

Sponsored by: Citrix Systems R&D
Tested-by: cperciva
9 years agoussreq: use saved fdp pointer insted of td->td_proc->p_fd
mjg [Fri, 12 Jun 2015 06:28:22 +0000 (06:28 +0000)]
ussreq: use saved fdp pointer insted of td->td_proc->p_fd

No functional changes.

9 years agoReorganize targets/pseudo/userland
sjg [Fri, 12 Jun 2015 03:26:05 +0000 (03:26 +0000)]
Reorganize targets/pseudo/userland

This target mainly exists to help check things build.
Eventually targets that build real packages or images will be more
important.
As such move the pseudo targets that only exist for the benefit of
'userland' to under it.