When more than one thread enters ufsdirhash_create() for the same
directory and the inode dirhash is instantiated, but the dirhash' hash
is not, all of them lock the dirhash shared and then try to upgrade.
Since there are several threads owning the lock shared, upgrade fails
and the same attempt is repeated, ad infinitum.
To break the lockstep, lock the dirhash in exclusive mode after the
failed try-upgrade.
Bruce Evans [Thu, 7 Dec 2017 07:55:38 +0000 (07:55 +0000)]
Move instantiation of msgbufp from 9 MD files to subr_prf.c.
This variable should be pure MI except possibly for reading it in MD
dump routines. Its initialization was pure MD in 4.4BSD, but FreeBSD
changed this in r36441 in 1998. There were many imperfections in
r36441. This commit fixes only a small one, to simplify fixing the
others 1 arch at a time. (r47678 added support for
special/early/multiple message buffer initialization which I want in
a more general form, but this was too fragile to use because hacking
on the msgbufp global corrupted it, and was only used for 5 hours in
-current...)
Alan Somers [Thu, 7 Dec 2017 02:08:55 +0000 (02:08 +0000)]
dc(1): fix modulo operations with fractional inputs
Our dc(1) has never correctly calculated remainders with fractional inputs.
Both bmod and bdivmod seem to have copy/pasted code from bdiv, which results
in the remainder having the wrong output scale.
Warner Losh [Wed, 6 Dec 2017 23:05:15 +0000 (23:05 +0000)]
Now that cam_periph_runccb() can be called from situations where the
kernel scheduler is stopped, replace the by hand calling of
xpt_polled_action() with it.
Gleb Smirnoff [Wed, 6 Dec 2017 23:03:34 +0000 (23:03 +0000)]
Garbage collect IFCAP_POLLING_NOCOUNT. It wasn't used since very
beginning of polling(4). The module always ignored return value
from driver polling handler.
For mails which has a body not respecting RFC2822 (which often happen with
crontabs) try to split by words finding the last space before 1000's character
If no spaces are found then consider the mail to be malformed anyway
Alan Somers [Wed, 6 Dec 2017 22:06:48 +0000 (22:06 +0000)]
Optimize telldir(3)
Currently each call to telldir() requires a malloc and adds an entry to a
linked list which must be traversed on future telldir(), seekdir(),
closedir(), and readdir() calls. Applications that call telldir() for every
directory entry incur O(n^2) behavior in readdir() and O(n) in telldir() and
closedir().
This optimization eliminates the malloc() and linked list in most cases by
packing the relevant information into a single long. On 64-bit architectures
msdosfs, NFS, tmpfs, UFS, and ZFS can all use the packed representation. On
32-bit architectures msdosfs, NFS, and UFS can use the packed
representation, but ZFS and tmpfs can only use it for about the first 128
files per directory. Memory savings is about 50 bytes per telldir(3) call.
Speedup for telldir()-heavy directory traversals is about 20-30x for one
million files per directory.
The function fwscanf() return value is wrong when encountering an early
matching failure.
According to the Open Group documentation for fwscanf:
"Upon successful completion, these functions shall return the number of
successfully matched and assigned input items; this number can be zero in
the event of an early matching failure."
Without this change, fwscanf would return EOF in the case of an early
matching failure, instead of the proper return value of 0.
This change aligns fwscanf(3) with the implementation in fscanf(3).
Alan Somers [Wed, 6 Dec 2017 17:01:25 +0000 (17:01 +0000)]
cam: fix sign-extension error in adagetparams
adagetparams contains a sign-extension error that will cause the sector
count to be incorrectly calculated for ATA disks of >=1TiB that still use
CHS addressing. Disks using LBA48 addressing are unaffected.
Revert local changes made to make zstd(1) frontend behave like gzip(1) and friends
This change was made to allow zstd(1) to be a dropin replacement for gzip(1) and
friends, allowing easy integration, in particular with newsyslog(8). At the
price of having a zstd(1) command which by default behaves differently than what
upstream default, confusing users.
newsyslog(8) has been adapted to now be more flexible in what it accepts as
compression program, so we can switch back zstd(1) to its default behaviour
Justin Hibbits [Wed, 6 Dec 2017 02:05:21 +0000 (02:05 +0000)]
Use unsigned intptr_t type for framebuffer addresses
Summary:
Some architectures (powerpc Book-E) have a vm_paddr_t larger than intptr_t.
Casting from the intptr_t to vm_paddr_t causes sign extension, leading to a
potentially invalid address.
This was seen when running X on a PowerPC P1022 machine, which mapped the
backing framebuffer at 0xc1800000. When mmap()d by X, this yielded an invalid
address of 0xffffffffc1800000, or, as the hardware would see it, 0xfc1800000.
Reviewed By: ray
Differential Revision: https://reviews.freebsd.org/D13332
Warner Losh [Tue, 5 Dec 2017 23:02:31 +0000 (23:02 +0000)]
Remove stray cam_periph_async call. It's called twice this way. While
currently harmless for AC_UNIT_ATTENTION event (cam_periph_async does
nothing with them), it's still in error because if it were to start in
the future, it would be done twice.
Warner Losh [Tue, 5 Dec 2017 22:24:20 +0000 (22:24 +0000)]
Since this is contrib code, create an upstreamable version of my
change. Now on FreeBSD and NetBSD if _STANDALONE is defined, we
include the kernel version with alloances for the quirky differences
between the two.
Kyle Evans [Tue, 5 Dec 2017 22:05:10 +0000 (22:05 +0000)]
dtb/allwinner: Restore a83t/BananaPi-M3 DTS after r342822
sinovoip-bpi-m3.dts was disconnected from the build in r324822. Since then,
a CCU driver has been added and several other changes have been made to
make us compatible with upstream DTS for this board.
Add links for older DTB that might be used: our u-boot port was expecting
sinovoip-bpi-m3.dtb up until ports r455629, and our u-boot will not be
switching to the upstream name (sun8i-a83t-bananapi-m3) quite yet.
Kyle Evans [Tue, 5 Dec 2017 21:40:52 +0000 (21:40 +0000)]
a10_gpio: Don't do read/set dance if pin is already configured for output
This fixes some regulator issues with a83t/BananaPi-M3; the pin value was
getting clobbered as we reconfigured the pin when initializing the
regulator.
Warner Losh [Tue, 5 Dec 2017 21:38:24 +0000 (21:38 +0000)]
Now that we offer a semi-sane standards-ish set of #include files in
the stand environment that's safe to use (and insulated from whatever
build env you might normally have), stop hacking the bzlib and zlib
sources with sed. There's no longer any need.
Warner Losh [Tue, 5 Dec 2017 21:38:19 +0000 (21:38 +0000)]
Stop building with the standard system headers.
Building with the standard system headers isn't a perfect match to the
stand environment. Instead, copy over the files we know are safe to
use and constrain what else is used. We use -nostdinc to achieve this.
This also fixes issues with building 32-bit libraries on amd64
sometimes pulling in the wrong cpufunc.h giving an error now that we
stop on errors. It will also enable an easier transition to lua boot.
Warner Losh [Tue, 5 Dec 2017 21:38:14 +0000 (21:38 +0000)]
Don't inherit CFLAGS. This a specialized test program, and can be
built with mostly default flags. Do so in anticipation of the rest of
stand not building with system headers.
Warner Losh [Tue, 5 Dec 2017 21:38:04 +0000 (21:38 +0000)]
Prefer stdint.h to inttypes.h since the added prototypes form the
latter aren't used. Prefer sys/link_elf.h to link.h so we're only
dependent on the kernel tree. The default installation of link.h just
includes this file, and any benefit from that is outweighed by the
hassle it causes. This reduces the footprint of files needed from the
system includes (or sysroot in buildworld).
Warner Losh [Tue, 5 Dec 2017 21:37:59 +0000 (21:37 +0000)]
Make sure we include the right path for skein.h, as well only include
the ZFS flags for zfs_modules.c. This keeps us from pulling from the
system or sysroot during buildworld.
Bryan Drewery [Tue, 5 Dec 2017 21:36:08 +0000 (21:36 +0000)]
native-xtools: Ensure GCC files are cleaned up.
Because we force enable MK_GCC when building we need to also force
enable it for the cleaning phase. Otherwise the NXB_TARGET files
are found in the next build's kernel-toolchain phase and cause
an error.
Bryan Drewery [Tue, 5 Dec 2017 21:30:22 +0000 (21:30 +0000)]
Deal with bmake-20170301 no longer resolving -C like it used to.
Several checks assume .CURDIR is resolved, such as for determining RELDIR from
SRCTOP/.CURDIR. If -C is used then the path is no longer resolved like it was
before which is problematic for symlinked source trees. A similar change was
also made to ports post bmake-20170301.
This fixes 'make -C <symlinked path> buildworld' using the wrong OBJDIR.
Bryan Drewery [Tue, 5 Dec 2017 21:29:47 +0000 (21:29 +0000)]
AUTO_OBJ: For all top-level targets enforce using an OBJDIR.
This will cause an error if the wanted OBJDIR is not writable. Previously it
would cause the files to generate to the source tree. This was too obscure and
things like buildworld really expect a proper OBJDIR layout.
Stephen Hurd [Tue, 5 Dec 2017 21:00:31 +0000 (21:00 +0000)]
iflib: Support to padding Ethernet frames to a min size
Some bnxt devices do not correctly send frames smaller than
52 bytes (without CRC), so add a quirk that will pad frames to an
arbitrary size before passing off to the encap routine.
Ed Maste [Tue, 5 Dec 2017 20:19:13 +0000 (20:19 +0000)]
use @@@ instead of @@ in __sym_default
Using
.symver foo,foo@@VER
causes foo and foo@@VER to be output to the .o file. This requires foo
to be weak since the linker handles foo@@VER as foo.
Using
.symver foo,foo@@@VER
causes just foo@@ver to be output and avoid the need for making foo
weak. It also reduces the constraint on how exactly a linker has to
handle foo and foo@@VER being present.
Gleb Smirnoff [Tue, 5 Dec 2017 19:54:55 +0000 (19:54 +0000)]
When parsing remote messages, require them to have standard timestamp
field, and support properly parse out the hostname as described by RFC3164,
which wasn't done before. However, don't discard message if it doesn't
have hostname, for compatibility.
Enable logging of the message supplied hostname instead of real hostname
with -H switch.
Eric Joyner [Tue, 5 Dec 2017 18:42:07 +0000 (18:42 +0000)]
ifconfig(8): Display extended compliance code string for SFP transceivers
- Updates tables in affected files with new entries from newer spec
revisions of SFF-8472, SFF-8024, and SFF-8636
- Change ifconfig to read and display the extended compliance code for
SFP media if the extended compliance code is not 0. This was being displayed
for QSFP transceivers only, but SFP28 media uses this to report 25G
capability.
Bryan Drewery [Tue, 5 Dec 2017 17:23:33 +0000 (17:23 +0000)]
Fix cyclic dependency after r326552.
The OBJS_DEPEND_GUESS mechanism was making vmx_genassym.o depend
on all headers along with vmx_assym.h, though vmx_assym.h depends
on having vmx_genassym.o present to generate. Moving the headers
to DPSRCS is enough to resolve the issue as they will no longer
be implicit dependencies for all objects. Because of this we
need explicit OBJS_DEPEND_GUESS entries to ensure the headers
are generated when needed for the *_support.o files that need
them.
Colin Percival [Tue, 5 Dec 2017 09:22:14 +0000 (09:22 +0000)]
Make EC2 instances use Amazon's NTP service for time synchronization.
Since Amazon provides NTP servers within their network, this should
be far superior to using the default NTP pools; and since the service
is provided by Amazon there's very little risk in enabling it by
default. (If someone is able to compromise Amazon's NTP servers and
exploit them to attack EC2 instances, they would almost certainly be
able to compromise EC2 instances even without ntpd running...)
MFC after: 1 week
Relnotes: EC2 instances now keep their clocks synchronized using
the Amazon Time Sync Service (aka. NTP).
Colin Percival [Tue, 5 Dec 2017 09:08:48 +0000 (09:08 +0000)]
Resurrect r321659: Turn off ChallengeResponseAuthentication for EC2 AMIs.
EC2 instances are normally launched with an SSH public key specified,
which is then used for logging in (by default, as 'ec2-user'). Having
ChallengeResponseAuthentication enabled (as FreeBSD's default sshd_config
does) has no functional effect in a new EC2 instance, since you can't log
in using a password until a password has been set -- but having this
enabled results in alerts from automated scanning tools which can detect
that sshd advertises support for keyboard-interactive logins (since they
can't detect that accounts have no password set).
EC2 users who want to use passwords to log in to their instances will need
to set 'ChallengeResponseAuthentication yes' in FreeBSD 12.0 and later.
Discussed with: gjb, gtetlow, emaste, des
Requested by: Amazon
X-MFC: No
Relnotes: ChallengeResponseAuthentication is turned off by default in
Amazon EC2 AMIs.
Correctly prefix the infiniband include directory for buildworld. This fixes
the OFED buildworld target, WITH_OFED=YES, when the include files are not
already installed locally, but only in the temporary object directory.
Alan Somers [Tue, 5 Dec 2017 04:22:35 +0000 (04:22 +0000)]
dc(1): fix input of non-decimal fractional numbers
Inputting fractional non-decimal numbers has never worked correctly in our
OpenBSD-derived dc(1). It truncates the input to a number of decimal places
equal to the number of hexadecimal (or whatever base) places given on the
input. That's unacceptable, because many numbers require more precision to
represent in base 10 than in their original bases.
Fix this bug by using as many decimal places as needed to represent the
input, up to the maximum of the global scale factor.
This has one mildly surprising side effect: the scale of a number entered in
non-decimal mode will no longer necessarily equal the number of hexadecimal
(or whatever base) places given on the input. I think that's an acceptable
behavior change, given that inputting fractional non-decimal numbers never
worked in the first place, and the man page doesn't specify whether trailing
zeros on the input should affect a number's scale.