freebsd.git
6 years agoFollow-up to r325967, which removed /etc/casper, by also removing it
dim [Tue, 12 Dec 2017 22:21:20 +0000 (22:21 +0000)]
Follow-up to r325967, which removed /etc/casper, by also removing it
from BSD.root.dist, so it does not get created again on installworld.

6 years agoMFV r326785: 8880 improve DTrace error checking
markj [Tue, 12 Dec 2017 22:08:34 +0000 (22:08 +0000)]
MFV r326785: 8880 improve DTrace error checking

illumos/illumos-gate@2cf374268f3e1c9e9be6367466b183d27632583a
https://github.com/illumos/illumos-gate/commit/2cf374268f3e1c9e9be6367466b183d27632583a

https://www.illumos.org/issues/8880

Reviewed by: Tim Kordas <tim.kordas@joyent.com>
Reviewed by: Bryan Cantrill <bryan@joyent.com>
Reviewed by: Richard Lowe <richlowe@richlowe.net>
Approved by: Dan McDonald <danmcd@joyent.com>
Author: Jerry Jelinek <jerry.jelinek@joyent.com>

MFC after: 1 week

6 years agoRevert r326792, r326784, r326772, r326712
imp [Tue, 12 Dec 2017 22:06:22 +0000 (22:06 +0000)]
Revert r326792, r326784, r326772, r326712

Something subtle is creating problems for disk access on ubldr. Back
it out unti that can be sorted out.

Sponsored by: Netflix

6 years agoCorrect initialization of pc on powerpc.
markj [Tue, 12 Dec 2017 20:41:11 +0000 (20:41 +0000)]
Correct initialization of pc on powerpc.

PR: 224293
Submitted by: Breno Leitao <breno.leitao@gmail.com>
X-MFC with: r326774
Pointy hat: markj

6 years ago1k objects on the stack are a bad idea. While it's likely safe in this
imp [Tue, 12 Dec 2017 20:22:09 +0000 (20:22 +0000)]
1k objects on the stack are a bad idea. While it's likely safe in this
context, it's also safe to allocate the memory and free it instead.

Noticed by: Eugene Grosbein's script

6 years agoMove buffer size checks outside of the vnode locks.
fsu [Tue, 12 Dec 2017 20:15:57 +0000 (20:15 +0000)]
Move buffer size checks outside of the vnode locks.

Reviewed by:    kib, cem, pfg (mentor)
Approved by:    pfg (mentor)
MFC after:      1 weeks

Differential Revision:    https://reviews.freebsd.org/D13405

6 years agoFix extattr getters in case of neither uio nor buffer was not passed to VOP_*.
fsu [Tue, 12 Dec 2017 20:02:48 +0000 (20:02 +0000)]
Fix extattr getters in case of neither uio nor buffer was not passed to VOP_*.

Approved by:    pfg (mentor)
MFC after:      2 weeks

Differential Revision:    https://reviews.freebsd.org/D13359

6 years agoActually insert the free(d) call missed in r326802.
imp [Tue, 12 Dec 2017 19:45:24 +0000 (19:45 +0000)]
Actually insert the free(d) call missed in r326802.

Noticed by: rpokala@

6 years agoIniailize str so ucs2_to_utf8 won't free stack garbage.
imp [Tue, 12 Dec 2017 19:26:28 +0000 (19:26 +0000)]
Iniailize str so ucs2_to_utf8 won't free stack garbage.

CID: 1381037
Sponsored by: Netflix

6 years agoAdd sanity testing against maximum sane lengths for device paths for
imp [Tue, 12 Dec 2017 19:26:24 +0000 (19:26 +0000)]
Add sanity testing against maximum sane lengths for device paths for
loader and kernel.

CID: 1383608
Sponsored by: Netflix

6 years agoFree load_opt_buf after we're done with it.
imp [Tue, 12 Dec 2017 19:26:19 +0000 (19:26 +0000)]
Free load_opt_buf after we're done with it.

CID: 1383607
Sponsored by: Netflix

6 years agoFix resource leak. Free converted description after printing it.
imp [Tue, 12 Dec 2017 19:26:15 +0000 (19:26 +0000)]
Fix resource leak. Free converted description after printing it.
Also minor style sort of local vars.

CID: 1383606
Sponsored by: Netflix

6 years agoDon't leak new_data.
imp [Tue, 12 Dec 2017 19:26:10 +0000 (19:26 +0000)]
Don't leak new_data.

CID: 1383605
Sponsored by: Netflix

6 years agoCheck return value for set_bootvar and give a good error message.
imp [Tue, 12 Dec 2017 19:25:54 +0000 (19:25 +0000)]
Check return value for set_bootvar and give a good error message.

CID: 1383601
Sponsored by: Netflix

6 years agoless(1): diff reduction vs upstream
asomers [Tue, 12 Dec 2017 17:34:35 +0000 (17:34 +0000)]
less(1): diff reduction vs upstream

No functional change.

MFC after: 3 weeks
Sponsored by: Spectra Logic Corp

6 years agoAddress a possible lost wakeup for gmirror events.
markj [Tue, 12 Dec 2017 17:29:34 +0000 (17:29 +0000)]
Address a possible lost wakeup for gmirror events.

g_mirror_event_send() acquires the I/O queue lock to deliver a wakeup
to the worker thread, and this is done after enqueuing the event.
So it's sufficient to check the event queue before atomically releasing
the queue lock and going to sleep.

MFC after: 1 week
Sponsored by: Dell EMC Isilon

6 years agoGive g_mirror_event_get() a more accurate name.
markj [Tue, 12 Dec 2017 17:25:25 +0000 (17:25 +0000)]
Give g_mirror_event_get() a more accurate name.

MFC after: 1 week
Sponsored by: Dell EMC Isilon

6 years agoDecrement sc_writes when BIO_DELETE requests complete.
markj [Tue, 12 Dec 2017 17:24:30 +0000 (17:24 +0000)]
Decrement sc_writes when BIO_DELETE requests complete.

Otherwise a gmirror that has received a BIO_DELETE request will never be
marked clean (unless sc_writes overflows).

MFC after: 1 week
Sponsored by: Dell EMC Isilon

6 years agoRework alignment handling in __libc_allocate_tls() for Variant I of TLS layout.
mmel [Tue, 12 Dec 2017 11:25:30 +0000 (11:25 +0000)]
Rework alignment handling in __libc_allocate_tls() for Variant I of TLS layout.

There are two versions of variant I of TLS
- ARM and aarch64 uses original version of variant I here TP points to
  start of TCB followed by aligned TLS segment. Both TCB and TLS must
  be aligned to alignment of TLS section. The TCB[0] points to DTV vector
  and DTV values are real addresses (without bias).

- MIPS, PowerPC and RISC-V use modified version of variant I,
  where TP points (with bias) to TLS and TCB immediately precedes TLS
  without any alignment gap. Only TLS should be aligned. The TCB[0]
  points to DTV vector and DTV values are biased by constant value (0x8000)
  from real addresses.

Take all this in account when allocating memory for TLS structures.

MFC after: 1 month
Reviewed by: kib, mizhka
Tested by: mizhka(on mips)
Differential Revision: https://reviews.freebsd.org/D13378

6 years agoAttempt to unbreak buildworld
antoine [Tue, 12 Dec 2017 09:46:53 +0000 (09:46 +0000)]
Attempt to unbreak buildworld

6 years agoClose the correct file descriptor.
delphij [Tue, 12 Dec 2017 06:56:21 +0000 (06:56 +0000)]
Close the correct file descriptor.

MFC after: 2 weeks

6 years agoDecode some PowerPC trap registers
jhibbits [Tue, 12 Dec 2017 03:16:10 +0000 (03:16 +0000)]
Decode some PowerPC trap registers

Decode on Book-E:
* ESR (Exception Syndrome Register)
* MCSR (Machine Check Status Register)

On AIM:
* MSSSR (Memory Subsystem Status Register)

Makes it easier to tell at a glance the type of trap and machine check
conditions now.

6 years agogzip(1): Remove duplicate close()
cem [Tue, 12 Dec 2017 01:19:08 +0000 (01:19 +0000)]
gzip(1): Remove duplicate close()

CID: 1383560
Reported by: Coverity
Sponsored by: Dell EMC Isilon

6 years agoRevert part of 362772. It was causing problems for includes and making
imp [Mon, 11 Dec 2017 23:15:43 +0000 (23:15 +0000)]
Revert part of 362772. It was causing problems for includes and making
the menus disappear.

Sponsored by: Netflix

6 years agoCorrect a typo in syscon driver 'modify' logic
kevans [Mon, 11 Dec 2017 22:55:51 +0000 (22:55 +0000)]
Correct a typo in syscon driver 'modify' logic

Not previously caught because the current consumer (not yet in tree) doesn't
use the 'modify' bits (yet).

Reported by: rpokala

6 years agoPull in r320396 from upstream clang trunk (by Malcolm Parsons):
dim [Mon, 11 Dec 2017 20:04:40 +0000 (20:04 +0000)]
Pull in r320396 from upstream clang trunk (by Malcolm Parsons):

  [Sema] Fix crash in unused-lambda-capture warning for VLAs

  Summary:
  Clang was crashing when diagnosing an unused-lambda-capture for a VLA
  because From.getVariable() is null for the capture of a VLA bound.
  Warning about the VLA bound capture is not helpful, so only warn for
  the VLA itself.

  Fixes: PR35555

  Reviewers: aaron.ballman, dim, rsmith

  Reviewed By: aaron.ballman, dim

  Subscribers: cfe-commits

  Differential Revision: https://reviews.llvm.org/D41016

This fixes a segfault when building recent audio/zynaddsubfx port
versions.

Reported by: hps
MFC after: 3 days

6 years agoIncrement encap_pad_mbuf_fail when m_dup() fails in padding
shurd [Mon, 11 Dec 2017 20:01:28 +0000 (20:01 +0000)]
Increment encap_pad_mbuf_fail when m_dup() fails in padding

Previously, the counter was only incremented when m_append() failed.  Since
the function can also fail on m_dup() now, increment the counter there as
well.

Sponsored by: Limelight Networks

6 years agoPass the trap frame to fasttrap hooks.
markj [Mon, 11 Dec 2017 19:21:39 +0000 (19:21 +0000)]
Pass the trap frame to fasttrap hooks.

The DTrace fasttrap entry points expect a struct reg containing the
register values of the calling thread. Perform the conversion in
fasttrap rather than in the trap handler: this reduces the number of
ifdefs and avoids wasting stack space for traps that don't involve
DTrace.

MFC after: 2 weeks

6 years agoAdd generic 'syscon' driver
kevans [Mon, 11 Dec 2017 18:04:04 +0000 (18:04 +0000)]
Add generic 'syscon' driver

Upstream dts for allwinner will require a syscon driver, since the emac node
coming in 4.15 will be using xref to /soc/syscon for configuring the emac
clock. Add a generic syscon driver to attach to /soc/syscon for use by
if_awg, providing basic read/write functionality to consumers.

syscon driver will also be used by arm64 at least for A64+H5 emac/if_awg.

Written by: mmel
Reviewed by: manu
Differential Revision: https://reviews.freebsd.org/D13295

6 years agoFix regression with lua import
imp [Mon, 11 Dec 2017 16:18:05 +0000 (16:18 +0000)]
Fix regression with lua import

Don't print when we can't find a file. Copy it instead to the error
buffer. Higher level routines determine if it's appropriate to print
the error message.

Also, remove dead code (labeled bogusly lost functionality) since we
never used that functionality. Remove unused arg from interact() too.

Sponsored by: Netflix

6 years agoUnbreak gcc build by using (void) for functions that take no args.
imp [Mon, 11 Dec 2017 16:17:53 +0000 (16:17 +0000)]
Unbreak gcc build by using (void) for functions that take no args.

Sponsored by: Netflix

6 years agoUse a dedicated counter for inactive queue scans.
markj [Mon, 11 Dec 2017 15:33:24 +0000 (15:33 +0000)]
Use a dedicated counter for inactive queue scans.

The laundry thread keeps track of the number of inactive queue scans
performed by the page daemon, and was previously using the v_pdwakeups
counter to count them. However, in some cases the inactive queue may
be scanned multiple times after a single wakeup, so it's more accurate
to use a dedicated counter.

Reviewed by: alc, kib (previous version)
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D13422

6 years agoRemove hard coded number of lun definition
bapt [Mon, 11 Dec 2017 14:54:42 +0000 (14:54 +0000)]
Remove hard coded number of lun definition

The number of lun exposed is now exposed via sysctl by the kernel.
Use that number in ctlstat instead of the hardcoded version
Add a backward compatibility in case the sysctl(2) request fails.

This also allows ctlstat -l 1118 to actually work when having more than
1024 luns.

Reviewed by: avg, manu (both before the backward compatibility addition)
Approved by: avg, manu (both before the backward compatibility addition)
MFC after: 2 weeks
Sponsored by: Gandi.net
Differential Revision: https://reviews.freebsd.org/D13446

6 years agoFix a comment to be more accurate
imp [Mon, 11 Dec 2017 14:47:23 +0000 (14:47 +0000)]
Fix a comment to be more accurate

6 years agoofed: Define barriers for mips and arm.
kib [Mon, 11 Dec 2017 11:59:45 +0000 (11:59 +0000)]
ofed: Define barriers for mips and arm.

I used the strongest barriers available on the architectures, so if
the future analysis show that it is excessive, the barriers could be
relaxed. Still, it is unlikely that it is meaningful to run IB on 32bit
ARM or current MIPS machines, so the change is to make WITH_OFED to pass
tinderbox.

Sponsored by: Mellanox Technologies
Reviewed by: hselasky
Differential revision: https://reviews.freebsd.org/D13329

6 years agoofed: Remove duplicated symbols from the version file.
kib [Mon, 11 Dec 2017 11:57:46 +0000 (11:57 +0000)]
ofed: Remove duplicated symbols from the version file.

ld.bfd accepts multiple listing of the same symbol in the version script.
lld is stricter and errors out.  Since arm64 and sometimes amd64 use lld,
we should correct this cosmetic issue.

Sponsored by: Mellanox Technologies
Reviewed by: hselasky
Differential revision: https://reviews.freebsd.org/D13329

6 years agoAdd myself to committers-src.dot
chuck [Mon, 11 Dec 2017 04:40:25 +0000 (04:40 +0000)]
Add myself to committers-src.dot

Reviewed by: ken, imp
Approved by: ken (mentor), imp (mentor)
Differential Revision: https://reviews.freebsd.org/D13406

6 years agoi386: Bump KSTACK_PAGES default to match amd64
cem [Mon, 11 Dec 2017 04:32:37 +0000 (04:32 +0000)]
i386: Bump KSTACK_PAGES default to match amd64

Logically, extend r286288 to cover all threads, by default.

The world has largely moved on from i386.  Most FreeBSD users and developers
test on amd64 hardware.  For better or worse, we have written a non-trivial
amount of kernel code that relies on stacks larger than 8 kB, and it "just
works" on amd64, so there has been little incentive to shrink it.

amd64 had its KSTACK_PAGES bumped to 4 back in Peter's initial AMD64 commit,
r114349, in 2003.  Since that time, i386 has limped along on a stack half
the size.  We've even observed the stack overflows years ago, but neglected
to fix the issue; see the 20121223 and 20150728 entries in UPDATING.

If anyone is concerned with this change, I suggest they configure their
AMD64 kernels with KSTACK_PAGES 2 and fix the fallout there first.  Eugene
has identified a list of high stack usage functions in the first PR below.

PR: 219476, 224218
Reported by: eugen@, Shreesh Holla <hshreesh AT yahoo.com>
Relnotes: maybe
Sponsored by: Dell EMC Isilon

6 years agoStart adding commands that I know are good to build up residual
imp [Mon, 11 Dec 2017 02:44:15 +0000 (02:44 +0000)]
Start adding commands that I know are good to build up residual
knowledge about qemu's quirks. Other qemu commands are possible, but
this is the quickest for testing.

Sponsored by: Netflix

6 years agoAdd -I to just run the imaging part of nanobsd so you can test out
imp [Mon, 11 Dec 2017 02:44:10 +0000 (02:44 +0000)]
Add -I to just run the imaging part of nanobsd so you can test out
changes to the packaging part of nanobsd more easily, or experiment
with the image contents w/o regenerating a whole new image tree. This
can save minutes when you don't need to do the installworld /
installkernel, etc.

Sponsored by: Netflix

6 years agoFix the std-x86 layout to produce a bootable system. Due to
imp [Mon, 11 Dec 2017 02:43:56 +0000 (02:43 +0000)]
Fix the std-x86 layout to produce a bootable system. Due to
limitations in mkimg we're still not quite to where I'd like to be
(I'd like to put s3 first on the disk, then s1, but mkimg won't allow
that currently). However, the resulting image now boots with qemu using:

qemu-system-x86_64 -hdd $file -serial telnet::4444,server -nographic

We'll need tweaks to create a specialized /etc/rc.d/growfs that can
create a properly grown image for either the simple or ping-pong
cases, but that will be later. Switched to pure serial console (-h)
instead of video or serial (-P) since that fits this usecase better.

Sponsored by: Netflix

6 years agoWhen building for arm arches, set PKGALIGN to the max cache line size
ian [Sun, 10 Dec 2017 23:06:45 +0000 (23:06 +0000)]
When building for arm arches, set PKGALIGN to the max cache line size
supported by the arch, to meet u-boot's requirement that I/O be done
in cache-aligned chunks.

PR: 223977

6 years agoCorrect r326748, indicating that tgammal(3) is mapped to tgamma(3), not
dim [Sun, 10 Dec 2017 23:03:13 +0000 (23:03 +0000)]
Correct r326748, indicating that tgammal(3) is mapped to tgamma(3), not
to itself.

Noticed by: jilles
MFC after: 3 days

6 years agoSave and restore r9 register in arm ubldr. In old gcc 4.2, r9 was a callee-
ian [Sun, 10 Dec 2017 21:51:27 +0000 (21:51 +0000)]
Save and restore r9 register in arm ubldr.  In old gcc 4.2, r9 was a callee-
saved register, but in arm EABI it may be either callee-saved or dedicated
to some special purpose (such as a TLS pointer).  It appears clang does not
treat it as a callee-saved register (instead using it as another work
register, similar to r12).

Another important side effect of these changes is that saving an extra
register in the push/pop statements keeps the stack aligned to an 8-byte
boundary during the self_reloc() call, as it always should have been.

As stated in the PR...

Essentially the important caller-saved registers are pushed (r0, r1, r9, lr)
before the relocation call, and popped after.  Then r8/r9 are saved as usual
for the syscall trampoline, and lr is stored in r8 (now free) as a
callee-saved value before calling into `main`.

The call to `main` can no longer be a tail call because we must restore r9
especially after main returns (although since we have used r8 to hold lr we
must also restore this).

PR: 224008

6 years agoDo not give up if writing to the chip's control and status registers fails
ian [Sun, 10 Dec 2017 18:55:42 +0000 (18:55 +0000)]
Do not give up if writing to the chip's control and status registers fails
during startup.  When a brand new chip leaves the factory, it is in a
special power-saving mode that disables most functions on the chip to
save battery power.  The chip is stuck in this mode until the first write
to the time registers, which automatically clears the special power-saving
mode and starts the oscillator.

Also, the day-of-week register in this chip counts 1-7, not 0-6, so write
the values accordingly.

These changes are based on the patch submitted by Brian Scott, but I
elimated warnings since this condition is expected, and added some comments,
and so in general blame me for any mistakes.

PR: 223642

6 years agowc(1): Restore regular file char count fast path
cem [Sun, 10 Dec 2017 17:56:03 +0000 (17:56 +0000)]
wc(1): Restore regular file char count fast path

fstat(2) is going to be a lot faster than reading all of the bytes in a
file, if we just need a character count for a regular file.  This fast path
was accidentally broken in r326736.

PR: 224160
Reported by: bde
Sponsored by: Dell EMC Isilon

6 years agoDocument the existence and precision of the remaining long double
dim [Sun, 10 Dec 2017 16:42:59 +0000 (16:42 +0000)]
Document the existence and precision of the remaining long double
functions for which an imprecise stub implementation was added in
r255294, namely powl(3) and tgammal(3).

Submitted by: Steve Kargl
MFC after: 3 days

6 years agoRelax too restrictive assert.
mmel [Sun, 10 Dec 2017 09:17:21 +0000 (09:17 +0000)]
Relax too restrictive assert.

The problem has been reported to upstream and similar change will
be included in next jemalloc release.

Submitted by: David Goldblatt <davidtgoldblatt@gmail.com>
MFC after:  2 weeks

6 years agoRevert r326740. I committed wrong diff.
mmel [Sun, 10 Dec 2017 09:15:37 +0000 (09:15 +0000)]
Revert r326740. I committed wrong diff.

6 years agoRelax too restrictive assert.
mmel [Sun, 10 Dec 2017 08:55:48 +0000 (08:55 +0000)]
Relax too restrictive assert.

The problem has been reported to upstream and similar change will
be included in next jemalloc release.

Submitted by: David Goldblatt <davidtgoldblatt@gmail.com>
MFC after:  2 weeks

6 years agoRetrieve the page outside of holding locks
jhibbits [Sun, 10 Dec 2017 04:43:27 +0000 (04:43 +0000)]
Retrieve the page outside of holding locks

pmap_track_page() only works with physical memory pages, which have a
constant vm_page_t address.  Microoptimize pmap_track_page() to perform one
less operation under the lock.

6 years agopw(8): correct expiration period handling and command line overrides
eugen [Sat, 9 Dec 2017 23:34:00 +0000 (23:34 +0000)]
pw(8): correct expiration period handling and command line overrides
to preconfigured values for -e, -p and -w flags.

Use non-negative symbols instead of magic values
in passwd_val/pw_password functions.

PR: 223431
Submitted by: Yuri Pankov (in part, patch for the manual)
Reported by: mav (mentor)
MFC after: 3 days
Relnotes: yes

6 years ago[net80211] add a method for checking if a VAP WME AC has a NOACK policy or not.
adrian [Sat, 9 Dec 2017 23:16:02 +0000 (23:16 +0000)]
[net80211] add a method for checking if a VAP WME AC has a NOACK policy or not.

A subsequent set of commits will introduce this instead of a whole lot of
gymnastics to check the WME category.

6 years agowc(1): Extend non-controversial optimizations to '-c' mode
cem [Sat, 9 Dec 2017 21:55:19 +0000 (21:55 +0000)]
wc(1): Extend non-controversial optimizations to '-c' mode

wc(1)'s slow path for counting words or multibyte characters requires
conversion of the 8-bit input stream to wide characters.  However, a faster
path can be used for counting only lines ('-l' -- newlines have the same
representation in all supported encodings) or bytes ('-c').

The existing line count optimization was not used if the input was the
implicit stdin.  Additionally, it wasn't used if only byte counting was
requested.  This change expands the fast path to both of these scenarios.

Expanding the buffer size from 64 kB helps reduce the number of read(2)
calls needed, but exactly what impact that change has and what size to
expand the buffer to are still under discussion.

PR: 224160
Tested by: wosch (earlier version)
Sponsored by: Dell EMC Isilon

6 years agoDefine macros used by the pNFS server code.
rmacklem [Sat, 9 Dec 2017 21:04:56 +0000 (21:04 +0000)]
Define macros used by the pNFS server code.

This commit defines some macros used by the pNFS server code.
They will not be used until the main pNFS server code merge occurs,
which will probably be in April 2018.

6 years agogifconfig_gif0 no longer works, document replacement
karels [Sat, 9 Dec 2017 17:27:36 +0000 (17:27 +0000)]
gifconfig_gif0 no longer works, document replacement

rc.conf(5) documents the gifconfig_<interface> keyword, which is
no longer implemented. Document the replacement, which works with
cloned_interfaces as well.

Reviewed by: dab
Group Reviwers: manpages
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D13130

6 years agoImprove options and error handling.
zeising [Sat, 9 Dec 2017 15:59:10 +0000 (15:59 +0000)]
Improve options and error handling.

Improve options handling and error out if multiple mutually exclusive
options are passed to acpiconf.  Switch from using atoi() to strtol() for
argument parsing, and add error checking and handling, instead of blindly
trusting that the integer conversion is OK.
Cange err() to errx() in once case, the errno value was garbage there.

Reviewed by: emaste
Approved by: emaste
Differential Revision: D13430

6 years agoFix the act_scan_laundry_weight mechanism.
markj [Sat, 9 Dec 2017 15:47:26 +0000 (15:47 +0000)]
Fix the act_scan_laundry_weight mechanism.

r292392 modified the active queue scan to weigh clean pages differently
from dirty pages when attempting to meet the inactive queue target. When
r306706 was merged into the PQ_LAUNDRY branch, this mechanism was
broken. Fix it by scalaing the correct page shortage variable.

Reviewed by: alc, kib
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D13423

6 years agoProvide a sysctl to force synchronous initialization of inode blocks.
markj [Sat, 9 Dec 2017 15:44:30 +0000 (15:44 +0000)]
Provide a sysctl to force synchronous initialization of inode blocks.

FFS performs asynchronous inode initialization, using a barrier write
to ensure that the inode block is written before the corresponding
cylinder group header update. Some GEOMs do not appear to handle
BIO_ORDERED correctly, meaning that the barrier write may not work as
intended. The sysctl allows one to work around this problem at the
cost of expensive file creation on new filesystems. The default
behaviour is unchanged.

Reviewed by: kib, mckusick
MFC after: 1 weeks
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D13428

6 years agoMove the DIAGNOSTIC check for lost iSCSI PDUs from icl_conn_close()
trasz [Sat, 9 Dec 2017 15:34:40 +0000 (15:34 +0000)]
Move the DIAGNOSTIC check for lost iSCSI PDUs from icl_conn_close()
to icl_conn_free().  It's perfectly valid for the counter to be non-zero
in the former.

MFC after: 2 weeks
Sponsored by: playkey.net

6 years agoarc lint: Explain linting explain
eadler [Sat, 9 Dec 2017 08:21:29 +0000 (08:21 +0000)]
arc lint: Explain linting explain

shell scripts in scripts don't need
to be chmod +x to work. In fact most are not.
Of the tests I found from a simple search:
65 are chmod +x
84 are chmod -x

simply disable the check for test shell scripts.

6 years agoIndent multiple device path entries correctly.
imp [Sat, 9 Dec 2017 07:44:00 +0000 (07:44 +0000)]
Indent multiple device path entries correctly.

Sponsored by: Netflix

6 years agoRemove vestiges of -d and -p commands.
imp [Sat, 9 Dec 2017 07:43:55 +0000 (07:43 +0000)]
Remove vestiges of -d and -p commands.
Fix two core dumps when optional data isn't specified.

Sponsored by: Netflix

6 years agoForgotten in 326725
imp [Sat, 9 Dec 2017 06:57:19 +0000 (06:57 +0000)]
Forgotten in 326725

Release Notes: Yes

6 years agoImport Netflix's efibootmgr to help manage UEFI boot variables
imp [Sat, 9 Dec 2017 06:52:20 +0000 (06:52 +0000)]
Import Netflix's efibootmgr to help manage UEFI boot variables

efibootmgr manages the UEFI BootXXXX variables that implement the UEFI
Boot Manager protocol defined in the UEFI standards. It is modeled
after the Linux program of the same name with a mostly compatible set
of command line options. Since there's a fair amount of OS specifioc
code due to differeing names and methods of doing things, the
compatibility isn't 100%.

Basic functionality is implemented, though the more advanced next boot
functionality that's been defined elsewhere is unimplemented.

Submitted by: Matt Williams (with unix / efi path xlate by me)
Sponsored by: Netflix

6 years agoAdd two dates derived from looking at archival source tarballs in the
imp [Sat, 9 Dec 2017 05:05:25 +0000 (05:05 +0000)]
Add two dates derived from looking at archival source tarballs in the
TUHS collection for 7th Edition and V32.

6 years agoThis path belongs in ficl/Makefile, not the common defines for users
imp [Fri, 8 Dec 2017 22:19:41 +0000 (22:19 +0000)]
This path belongs in ficl/Makefile, not the common defines for users
of ficl.

Sponsored by: Netflix

6 years agoRemove basename_r(3).
ed [Fri, 8 Dec 2017 22:06:18 +0000 (22:06 +0000)]
Remove basename_r(3).

Now that the POSIX working group is going to require that basename(3)
and dirname(3) are thread-safe in future revisions of the standard,
there is even less of a need to provide basename_r(3). Remove this
function to prevent people from writing code that only builds on
FreeBSD and Bionic.

Removing this function seems to break exactly one port: sbruno@'s
qemu-user-static. I will send him a pull request on GitHub in a bit.
__FreeBSD_version will not be bumped, as any value from 2017 can be used
to test for the presence of a thread-safe basename(3)/dirname(3).

PR: https://bugs.freebsd.org/224016

6 years agoDon't automatically create objdir here. We're just after the
imp [Fri, 8 Dec 2017 21:59:50 +0000 (21:59 +0000)]
Don't automatically create objdir here. We're just after the
TMPPATH. We'll make the paths later.

Sponsored by: Netflix

6 years agoRemove local here. It was a cut and paste error.
imp [Fri, 8 Dec 2017 21:59:42 +0000 (21:59 +0000)]
Remove local here. It was a cut and paste error.

Sponsored by: Netflix

6 years agoCorrectly define the unordered_map namespace in ofed/libibnetdisc .
hselasky [Fri, 8 Dec 2017 21:01:09 +0000 (21:01 +0000)]
Correctly define the unordered_map namespace in ofed/libibnetdisc .

This should fix ofed/libibnetdisc compilation with C-compilers
different from clang and GCC v4.2.1.

Submitted by: kib
Sponsored by: Mellanox Technologies

6 years agoboot1.c needs EFI_ZFS_BOOT too, so add it globally. Otherwise we'll
imp [Fri, 8 Dec 2017 19:57:26 +0000 (19:57 +0000)]
boot1.c needs EFI_ZFS_BOOT too, so add it globally. Otherwise we'll
not be able to actually read ZFS partitions.

Submitted by: kevans@

6 years agoBuild WITHOUT_FORTH too, at least for amd64 and i386
imp [Fri, 8 Dec 2017 19:57:21 +0000 (19:57 +0000)]
Build WITHOUT_FORTH too, at least for amd64 and i386

Sponsored by: Netflix

6 years agoCreate interp class.
imp [Fri, 8 Dec 2017 19:57:16 +0000 (19:57 +0000)]
Create interp class.

Create an interp class. Use it to separate out the different types of
interpreters: forth and simple with function pointers rather than
via #ifdefs.

Obtained from: lua boot loader project
    (via https://bsdimp@github.com/bsdimp/freebsd.git lua-bootloader)
Sponsored by: Netflix

6 years agoConst poison a couple of interfaces.
imp [Fri, 8 Dec 2017 19:57:11 +0000 (19:57 +0000)]
Const poison a couple of interfaces.

Obtained from: lua boot project
Sponsored by: Netflix

6 years agoPut the files we're copying over into a few variables and add them to
imp [Fri, 8 Dec 2017 19:57:06 +0000 (19:57 +0000)]
Put the files we're copying over into a few variables and add them to
CLEANDIRS and CLEANFILES so make clean removes any divots.

Sponsored by: Netflix

6 years agoProvide implementations for iscntrl, ispunct and isgraph.
imp [Fri, 8 Dec 2017 19:57:02 +0000 (19:57 +0000)]
Provide implementations for iscntrl, ispunct and isgraph.

Sponsored by: Netflix

6 years agoRemove _KERNEL hack now that errno.h does the right thing when
imp [Fri, 8 Dec 2017 19:56:57 +0000 (19:56 +0000)]
Remove _KERNEL hack now that errno.h does the right thing when
_STANDALONE is defined.

Sponsored By: Netflix

6 years agoAdd partial support signal.h functioanlity. Pull in machine/signal.h
imp [Fri, 8 Dec 2017 19:56:35 +0000 (19:56 +0000)]
Add partial support signal.h functioanlity. Pull in machine/signal.h
to define sig_atomic_t.

Sponsored by: Netflix

6 years agoFree mbuf chain when m_dup fails
shurd [Fri, 8 Dec 2017 19:50:06 +0000 (19:50 +0000)]
Free mbuf chain when m_dup fails

Fix memory leak where mbuf chain wasn't free()d if iflib_ether_pad()
has a failure in m_dup().

Reported by: "Ryan Stone" <rysto32@gmail.com>
Sponsored by: Limelight Networks

6 years agoActually add the -x setenv test Makefile, missed in r326499.
markj [Fri, 8 Dec 2017 19:26:25 +0000 (19:26 +0000)]
Actually add the -x setenv test Makefile, missed in r326499.

X-MFC with: r326499

6 years agoHandle read-only mbufs in iflib ether pad function
shurd [Fri, 8 Dec 2017 18:43:31 +0000 (18:43 +0000)]
Handle read-only mbufs in iflib ether pad function

If ethernet padding is enabled, and a read-only mbuf is passed,
it would modify the mbuf using m_append(). Instead, call m_dup() and
append to the new packet.

Reported by: Pyun YongHyeon
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D13414

6 years agoSPDX: more ISC-related files.
pfg [Fri, 8 Dec 2017 17:52:53 +0000 (17:52 +0000)]
SPDX: more ISC-related files.

6 years agoat(1): annotate some intended switch-case fallthroughs
asomers [Fri, 8 Dec 2017 17:15:20 +0000 (17:15 +0000)]
at(1): annotate some intended switch-case fallthroughs

Reported by: Coverity
CID: 1008191
MFC after: 3 weeks
Sponsored by: Spectra Logic Corp

6 years agoSPDX: license IDs for some ISC-related files.
pfg [Fri, 8 Dec 2017 15:57:29 +0000 (15:57 +0000)]
SPDX: license IDs for some ISC-related files.

6 years agoSet the io width when using an ACPI uart. Previously it would only ever be
andrew [Fri, 8 Dec 2017 10:05:23 +0000 (10:05 +0000)]
Set the io width when using an ACPI uart. Previously it would only ever be
set when finding the uart from the device tree.

Sponsored by: DARPA, AFRL

6 years agoRemove PTE VA mappings for tracked pages in 64-bit mode
jhibbits [Fri, 8 Dec 2017 03:49:53 +0000 (03:49 +0000)]
Remove PTE VA mappings for tracked pages in 64-bit mode

This was done in 32-bit mode, but not duplicated when 64-bit mode was
brought in.  Without this, stale mappings can be left, leading to odd
crashes when the wrong VA is checked in XX_PhysToVirt() (dpaa(4)).

6 years agoSupport mounted boot partitions in the installer. This allows the platform
nwhitehorn [Fri, 8 Dec 2017 00:57:13 +0000 (00:57 +0000)]
Support mounted boot partitions in the installer. This allows the platform
layer, for example, to specify that the EFI boot partition should be
mounted at /efi and formatted normally with newfs_msdos rather than
splatted to from /boot/boot1.efifat.

This commit adds only the API for this; actual platform use will come later.

6 years agoSeparate out send buffer autoscaling code into function, so that
glebius [Thu, 7 Dec 2017 22:36:58 +0000 (22:36 +0000)]
Separate out send buffer autoscaling code into function, so that
alternative TCP stacks may reuse it instead of pasting.

Obtained from: Netflix

6 years agoRetire SCTP_WITH_NO_CSUM option.
tuexen [Thu, 7 Dec 2017 22:19:08 +0000 (22:19 +0000)]
Retire SCTP_WITH_NO_CSUM option.

This option was used in the early days to allow performance measurements
extrapolating the use of SCTP checksum offloading. Since this feature
is now available, get rid of this option.
This also un-breaks the LINT kernel. Thanks to markj@ for making me
aware of the problem.

6 years agoAvoid setting -Wno-tentative-definition-incomplete-type with gcc.
markj [Thu, 7 Dec 2017 22:11:23 +0000 (22:11 +0000)]
Avoid setting -Wno-tentative-definition-incomplete-type with gcc.

No version of gcc that I've tried accepts this flag.

Reviewed by: imp
Differential Revision: https://reviews.freebsd.org/D13415

6 years agoRemove an unused incude from lib/msun/bsdsrc/b_log.c.
dim [Thu, 7 Dec 2017 20:41:23 +0000 (20:41 +0000)]
Remove an unused incude from lib/msun/bsdsrc/b_log.c.

Submitted by: Steve Kargl
MFC after: 3 days

6 years agoRemove the sentence in math(3) about some long double math functions not
dim [Thu, 7 Dec 2017 20:38:37 +0000 (20:38 +0000)]
Remove the sentence in math(3) about some long double math functions not
being available.

Submitted by: Steve Kargl
MFC after: 3 days

6 years agogeom_raid (RAID5): do not lose bp->bio_error, keep it in pbp->bio_error
eugen [Thu, 7 Dec 2017 20:09:17 +0000 (20:09 +0000)]
geom_raid (RAID5): do not lose bp->bio_error, keep it in pbp->bio_error
and return it by passing to g_raid_iodone()

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

6 years agoFix the reldoc target in release/Makefile following doc commit r51264.
gjb [Thu, 7 Dec 2017 20:00:19 +0000 (20:00 +0000)]
Fix the reldoc target in release/Makefile following doc commit r51264.

Sponsored by: The FreeBSD Foundation

6 years agomlx4: Remove redundant declarations to fix GCC build
cem [Thu, 7 Dec 2017 19:57:51 +0000 (19:57 +0000)]
mlx4: Remove redundant declarations to fix GCC build

These were made redundant in r325841.

Reviewed by: hselasky
MFC after: 1 week (hselasky will MFC)
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D13401

6 years agoUse correct field in the description for the lock after r319722.
bz [Thu, 7 Dec 2017 19:40:46 +0000 (19:40 +0000)]
Use correct field in the description for the lock after r319722.

Reviewed by: glebius
Sponsored by: iXsystems, Inc.

6 years agoFix the UMA reclaim worker after r326347.
markj [Thu, 7 Dec 2017 19:38:09 +0000 (19:38 +0000)]
Fix the UMA reclaim worker after r326347.

atomic_set_*() sets a bit in the target memory location, so
atomic_set_int(&uma_reclaim_needed, 0) does not do what it looks like
it does.

PR: 224080
Reviewed by: jeff, kib
Differential Revision: https://reviews.freebsd.org/D13412

6 years agoMerge OpenSSL 1.0.2n.
jkim [Thu, 7 Dec 2017 18:02:57 +0000 (18:02 +0000)]
Merge OpenSSL 1.0.2n.

6 years agoImport OpenSSL 1.0.2n.
jkim [Thu, 7 Dec 2017 17:37:15 +0000 (17:37 +0000)]
Import OpenSSL 1.0.2n.