Matthew Dillon [Tue, 13 Nov 2018 19:12:36 +0000 (11:12 -0800)]
kernel - Fix sack NULL pointer dereference
* sack_block_lookup() can get confused when the passed-in sequence
number appears to be less than sblk_start and greater than sblk_end.
This situation can occur when the signed integer delta test has an
overflow due to (sblk_end - seq) overflowing the sign bit verses
(sblk_start - seq).
The result is that sack_block_lookup() can crash on a NULL pointer
indirection.
* Check for the case, complain, and try to allow it. Though I suspect
if the case occurs at all SACK will wind up with a broken list anyway.
* I don't think this case can occur under normal conditions since TCP
buffers do not grow to 2GB+ in size, so the crash we got was triggered
by either an accidently malformed packet or an intentional one.
Matthew Dillon [Tue, 13 Nov 2018 19:10:47 +0000 (11:10 -0800)]
kernel - Fix SMP race in procfs
* procfs was not locking its pid hash table. Implement a per-slot
spinlock. The hash is populated only for pid directories that are
pushed into (e.g. ls /proc/<pid> vs ls /proc).
* Fixes a fairly rare crash that can only occur when procfs is used
heavily.
François Tigeot [Sun, 11 Nov 2018 08:18:03 +0000 (09:18 +0100)]
drm/ttm: Sync ttm_execbuf_util.c with Linux
François Tigeot [Sat, 10 Nov 2018 08:31:20 +0000 (09:31 +0100)]
drm/ttm: Sync ttm_lock.c with Linux
Aaron LI [Sat, 10 Nov 2018 01:08:23 +0000 (09:08 +0800)]
terminfo: Build and install tmux and tmux-256color
François Tigeot [Fri, 9 Nov 2018 17:34:19 +0000 (18:34 +0100)]
drm/linux: Add send_sig()
Aaron LI [Thu, 8 Nov 2018 01:13:04 +0000 (09:13 +0800)]
rc.conf.5: Document ip6addrctl_{policy,verbose} explicitly
Suggested-by: swildner
François Tigeot [Wed, 7 Nov 2018 22:18:51 +0000 (23:18 +0100)]
drm/ttm: Rework ttm_io_prot()
Change it to be as close to the Linux version as possible.
Imre Vadász [Tue, 6 Nov 2018 19:52:36 +0000 (20:52 +0100)]
kernel - Fix typo from
b227f3f50d5dc0f when filling in kpmap header.
Imre Vadász [Mon, 5 Nov 2018 22:14:40 +0000 (23:14 +0100)]
crypto - Avoid including opensslconf.h when using openssl/md5.h in kernel.
Imre Vadász [Mon, 5 Nov 2018 22:13:00 +0000 (23:13 +0100)]
kernel - Make the wdog.h and gpio.h includes conditional in kern_shutdown.c.
Imre Vadász [Sat, 3 Nov 2018 19:51:06 +0000 (20:51 +0100)]
kernel - Fix build when neither INET nor INET6 options are set.
François Tigeot [Tue, 6 Nov 2018 17:32:51 +0000 (18:32 +0100)]
drm/linux: fix pgprot_xxx functions
Sascha Wildner [Tue, 6 Nov 2018 04:28:07 +0000 (05:28 +0100)]
rc.d/Makefile: Break some long lines, sort dntpd.
François Tigeot [Mon, 5 Nov 2018 21:15:18 +0000 (22:15 +0100)]
drm/linux: Fix vmap()
Page protection information was not being used.
Aaron LI [Mon, 5 Nov 2018 20:31:38 +0000 (04:31 +0800)]
rc.subr: Add list_vars() function
Taken-from: FreeBSD
Aaron LI [Mon, 5 Nov 2018 20:25:17 +0000 (04:25 +0800)]
rc.d: Add ip6addrctl (enabled by default with policy AUTO)
Bring in the ip6addrctl rc script from FreeBSD. It is enabled by
default with policy "AUTO", so the IPv4/IP6 preference is configured on
startup. The rc variable "ip6addrctl_policy" is used to adjust the
preference (AUTO, ipv6_prefer, ipv6_prefer).
See the rc.conf(5) man page for more details.
Aaron LI [Mon, 5 Nov 2018 19:22:15 +0000 (03:22 +0800)]
Bring in ip6addrctl(8) from FreeBSD
The ip6addrctl(8) manages the policy table of source and destination
address selection for ougoing IPv4 and IPv6 packets.
For example, this utility can be used to configure that IPv4 is
preferred on a dual-stack machine, which by default prefers IPv6.
Simon Arlott [Sat, 3 Nov 2018 20:08:07 +0000 (20:08 +0000)]
libc/mktemp: Pass flags to open()
The flags from the functions that call _gettemp() are never used.
They should be included in the call to open(), otherwise features like
O_CLOEXEC don't work.
aly:
(1) Checked FreeBSD and it behaves the same way as this fix.
(2) Improve the style a bit.
Aaron LI [Sun, 4 Nov 2018 09:55:53 +0000 (17:55 +0800)]
net/if: Call if_ioctl() if the protocol didn't handle the ioctl
This allows for a non-interface socket (like AF_LOCAL which ifconfig(8)
now uses since commit
d4e856128472e2e79172aebf65f19f527ac0010f) to use
a broader class of interface ioctls.
This fixes the regression in ifconfig(8) that is introduced by the above
mentioned commit, which changed ifconfig(8) to use AF_LOCAL socket by
default and thus caused some ioctls failures. For example:
dfly# ifconfig tap0 create
dfly# ifconfig tap0 up
dfly# ifconfig bridge0 create
dfly# ifconfig bridge0 addm tap0
[before]
dfly# ifconfig bridge0
bridge0: flags=8002<BROADCAST,MULTICAST> mtu 1500
ether be:48:6d:6d:a6:1f
groups: bridge
[after]
dfly# ifconfig bridge0
bridge0: flags=8002<BROADCAST,MULTICAST> mtu 1500
ether be:48:6d:6d:a6:1f
priority 32768 hellotime 2 fwddelay 15 maxage 20
member: tap0 flags=3<LEARNING,DISCOVER>
groups: bridge
Do not pass the SIOCSIF{ADDR,BRDADDR,DSTADDR,NETMASK} iotcls to the
drivers because they may assume these ioctls come from an already
privileged layer and thus skip credentials check and input validation.
While there, improve a NULL check.
Taken-from: FreeBSD (r190151, r255442)
Reviewed-by: dillon, sephe
Matthew Dillon [Sun, 4 Nov 2018 06:55:49 +0000 (23:55 -0700)]
boot - Slow down the nfs twiddle
* Slow down the twiddle rate when reading via nfs, it unnecessarily
slows down the boot, by a lot.
François Tigeot [Sat, 3 Nov 2018 14:18:38 +0000 (15:18 +0100)]
drm/ttm: Remove ttm_bo_unreserve_core()
* This function was DragonFly-specific and apparently
isn't even needed anymore
* Sync the changed code with Linux as of
ecff665f5e3f1c6909353e00b9420e45ae23d995
"drm/ttm: make ttm reservation calls behave like reservation calls"
François Tigeot [Sat, 3 Nov 2018 13:07:50 +0000 (14:07 +0100)]
drm: Add linux/rwlock.h
Matthew Dillon [Sat, 3 Nov 2018 00:25:35 +0000 (17:25 -0700)]
vkernel - Fix tty panic
* Fix messed up gettoken/reltoken sequence for the vkernel,
bug was introduced in recent tty work.
Reported-by: tuxillo
Sascha Wildner [Fri, 2 Nov 2018 16:28:53 +0000 (17:28 +0100)]
dmesg(8): Allow seeing backslashes in dmesg output.
This caused confusion already in ACPICA bug reports with pathnames
containing backslashes.
Taken-from: NetBSD
Sascha Wildner [Fri, 2 Nov 2018 16:26:59 +0000 (17:26 +0100)]
libc/vis: Use thread-safe mbrtoc() and wcrtomb().
Taken-from: FreeBSD
Sascha Wildner [Fri, 2 Nov 2018 08:58:43 +0000 (09:58 +0100)]
Sync ACPICA with Intel's version
20181031.
* Fix an Operation Region regression, fixing accurate checking for
region conflicts.
* Add support for 2-byte extended opcodes in the code that continues
parsing during the table load phase.
* Eliminate a no longer used global variable.
For detailed list, please see sys/contrib/dev/acpica/changes.txt.
Sascha Wildner [Fri, 2 Nov 2018 08:04:13 +0000 (09:04 +0100)]
Sync zoneinfo database with tzdata2018g from ftp://ftp.iana.org/tz/releases
* Morocco switched to permanent +01 on 2018-10-27.
* Change HDT to HWT/HPT for WWII-era abbreviations in Hawaii.
For a detailed list of changes, see share/zoneinfo/NEWS.
Matthew Dillon [Fri, 2 Nov 2018 02:06:14 +0000 (19:06 -0700)]
boot - Handle tftp block overflows
* Properly truncate the block number when comparing against the
incoming packet. TFTP only implements 16-bit block numbers.
This allows the boot loader to access files larger than 32MB
via TFTP.
* Twiddle every 256 reads instead of every 16.
* Note that TFTP is not really a suitable protocol for loading the
kernel because the ELF loader seeks past the debug info which
can be well over a hundred megabytes in order to get to the syms.
NFS is preferred.
Sascha Wildner [Wed, 31 Oct 2018 09:34:00 +0000 (10:34 +0100)]
Just note that the previous commit was to hammer2 only.
The tmpfs logic turned out to be correct after closer review.
Sascha Wildner [Wed, 31 Oct 2018 09:27:27 +0000 (10:27 +0100)]
kernel/{hammer2,tmpfs}: Fix ENOTDIR/EISDIR logic.
It caused confusion like this:
vm64# touch /usr/obj/t
vm64# rmdir /usr/obj/t
rmdir: /usr/obj/t: Is a directory
vm64#
Likewise unlink(2) on a directory also gives the wrong error
("Not a directory").
This also causes Ruby's unlink() file method to fail.
Reported-by: tuxillo
Matthew Dillon [Wed, 31 Oct 2018 01:50:03 +0000 (18:50 -0700)]
tmpfs - Use atomic_fetchadd_long() to allocate inode numbers
* Since DFly inode numbers are 64 bits wide, tmpfs just increments a
per-mount counter to allocate a new inode.
* We had some old code in tmpfs to allocate inode numbers which wrapped
a tmpfs mount lock around the increment. This is totally unnecessary.
Use atomic_fetchadd_long() instead.
Reported-by: mjg
Matthew Dillon [Wed, 31 Oct 2018 01:48:35 +0000 (18:48 -0700)]
kernel - Fix getsockname() on unnamed AF_LOCAL sockets
* Fix getsockname() on unnamed AF_LOCAL sockets. The socket structure
does not actually bother storing a sockaddr, so sa is returned NULL.
In this situation, just dummy-up an AF_LOCAL sockaddr.
* Don't bother copyout()ing sa_len unless it has changed.
Reported-by: tuxillo
Sascha Wildner [Sun, 28 Oct 2018 16:53:12 +0000 (17:53 +0100)]
bsd-family-tree: Sync with FreeBSD (OpenBSD 6.4).
Imre Vadász [Fri, 26 Oct 2018 20:14:47 +0000 (22:14 +0200)]
sound - Allow /dev/sndstat to be opened multiple times from one process.
* This still isn't great, but at least it's enough to make pulseaudio's
oss module detect the audio devices correctly.
* This now uses the buf->uio_offset, instead of trying to track the read
position ourselves in the driver with sndstat_bufptr.
Sascha Wildner [Sat, 27 Oct 2018 07:45:04 +0000 (09:45 +0200)]
<sys/sysproto.h>: Add back one needed newline and regenerate header.
Sascha Wildner [Sat, 27 Oct 2018 07:39:25 +0000 (09:39 +0200)]
<sys/sysproto.h>: Fix excessive newlines.
Sascha Wildner [Sat, 27 Oct 2018 06:26:44 +0000 (08:26 +0200)]
<sys/wait.h>: Put the enum name under __BSD_VISIBLE.
POSIX only defines idtype_t. In <sys/procctl.h> i just left it because
that whole file isn't covered by the standard anyway.
While here, cleanup whitespace a little bit and clarify DragonFly or
FreeBSD a bit better in a comment.
Matthew Dillon [Sat, 27 Oct 2018 01:33:57 +0000 (18:33 -0700)]
build - refactor rescue and initrd handling
* buildworld now builds the rescue directory tree and initrd
image in /usr/obj.
* installworld will install the rescue and initrd stuff if it
does not exist on the destination. If rescue or initrd stuff
is already installed, it is not overwritten.
* Add 'installrescue' and 'installinitrd' targets. These will
copy the rescue and initrd image built by buildworld.
* The normal 'rescue' and 'initrd' targets are unchanged and
should build (from scratch) and install the rescue and initrd
stuff.
* NOTE: crunchgen is currently hacked to add system paths when
WORLDBUILD is defined, in order to access native binaries
for 'uniq' and a few others, instead of building them in
cross-tools.
In particular, we have to use the native 'vnconfig' binary
as building one in crosstools might not be compatible with
the currently running system. buildworld now uses 'vn' to
create the initrd image.
Possibly needs some work.
Matthew Dillon [Sat, 27 Oct 2018 01:17:36 +0000 (18:17 -0700)]
kernel - Refactor kbmux arrays
* Get rid of the generally insane dynamic reallocation code for
these tiny arrays and set a fixed limit of 16 keyboards.
* Add a sanity check in kbd_intr() to deal with potential interrupt
races against installation of a new keyboard.
Matthew Dillon [Sat, 27 Oct 2018 01:08:35 +0000 (18:08 -0700)]
libc - Add kpmap shortcut for gettimeofday()
* If the kern.gettimeofday_quick sysctl is set to 1, libc's
gettimeofday() will use the tick-granular realtime in
the kpmap instead of making a system call. The sysctl
defaults to 0 (off).
* Useful in situations where gettimeofday() interferes with
benchmarks, but otherwise not recommended.
* Running programs react instantly to changes in this sysctl.
Sascha Wildner [Fri, 26 Oct 2018 13:38:02 +0000 (15:38 +0200)]
sys/conf/files: Move vmx a bit up.
Diederik de Groot [Mon, 1 Oct 2018 23:00:05 +0000 (01:00 +0200)]
Virtio_Balloon implementation for DragonFly
- Initial check-in of fbsd virtio_balloon (freebsd/freebsd@1537078)
- Update virtio_balloon man page
- Fix vtballoon_alloc_intrs to deal nicely with number of provided interrupts.
- Use a backoff scheme when allocating vm_pages:
- VM_ALLOC_NORMAL for normal (low memory presure) situation;
- VM_ALLOC_SYSTEM + increased sleep timeout, for high memory presure circumstances.
The latter will allow the system to swap other pages out if necessary.
- Implemented suggestions made by ivadasz
- Added virtio_teardown_intr handling to vtballoon_detach
- Added config_change_intr handling
- Fixed virtio_setup_intr
- Add debug_level sysctl for tracing.
- Add collection of guest memory/pagefault statistics
- Completed/Updated header definitions
- Don't use VM_ALLOC_INTERRUPT. Add better inflate/deflate debug output to thread_sleep function
- Don't collect stats when initialially loaded (Reported by tuxillo).
Sascha Wildner [Fri, 26 Oct 2018 12:42:29 +0000 (14:42 +0200)]
LINT64: Add comments to the virtio devices.
Matthew Dillon [Fri, 26 Oct 2018 02:56:40 +0000 (19:56 -0700)]
build - Fix buildworld from DFly releases prior to 5.0.
* Fix mkmagic to not use the magic.h in /usr/include. It
needs the magic.h generated via libmagic. Since mkmagic
must be built prior to libmagic, we generate a local magic.h
just for mkmagic.
* Fixes buildworld from DFly releases prior to 5.0.
Imre Vadász [Thu, 25 Oct 2018 22:53:08 +0000 (00:53 +0200)]
sound - Fix mmap handling for /dev/dsp* devices.
* The d_mmap function is expected to return a page index, not an address,
the the atop() function for this as usual.
Sascha Wildner [Thu, 25 Oct 2018 17:31:53 +0000 (19:31 +0200)]
<sys/signal.h>: Discard old non-POSIX union members.
The only thing I found that still uses the old names is www/nginx,
a fix for which is in DeltaPorts:
https://github.com/DragonFlyBSD/DeltaPorts/pull/874/commits/
bd7b49f3f6ba39cfc00c7f0ef531358c3382ba1e
While here, fix a small whitespace issue.
Aaron LI [Thu, 25 Oct 2018 12:07:18 +0000 (20:07 +0800)]
rc.conf.5: Remove the obsolete reference to named(8)
Aaron LI [Thu, 25 Oct 2018 05:48:56 +0000 (13:48 +0800)]
Sync /etc/pccard_ether with FreeBSD
Our original 'pccard_ether' became unnecessarily complex as a big part of
the work is already implemented in 'network.subr', so just bring FreeBSD's
version over (with 'checkauto()' ignored as we don't support it).
NOTE: The 'pccard_ether' is still disabled by default on DragonFly.
Aaron LI [Thu, 25 Oct 2018 05:16:47 +0000 (13:16 +0800)]
arp(8): Add option -i to limit operations to one interface
This allows to limit the scope of operation to the ARP entries on a
particular interface, which should be useful on machines with lots of
network interfaces, e.g., inter-VLAN routers.
This new option will also be used in /etc/pccard_ether.
Taken from FreeBSD (r117729, r154162) with modifications.
Sascha Wildner [Thu, 25 Oct 2018 08:27:37 +0000 (10:27 +0200)]
<sys/eventhandler.h>: Move the include inside the guard.
Sascha Wildner [Thu, 25 Oct 2018 07:54:46 +0000 (09:54 +0200)]
<sys/stdint.h>: Move the includes inside the guard.
Sascha Wildner [Wed, 24 Oct 2018 18:27:06 +0000 (20:27 +0200)]
Sync zoneinfo database with tzdata2018f from ftp://ftp.iana.org/tz/releases
* Volgograd moves from +03 to +04 on 2018-10-28.
* Fiji ends DST 2019-01-13, not 2019-01-20.
* Most of Chile changes DST dates, effective 2019-04-06.
* Use PST and PDT for Philippine time.
* Various adjustments to past timestamps.
* Miscellaneous fixes and adjustments.
For a detailed list of changes, see share/zoneinfo/NEWS.
Sascha Wildner [Mon, 22 Oct 2018 18:12:22 +0000 (20:12 +0200)]
Update the pciconf(8) database.
October 13, 2018 snapshot from https://pci-ids.ucw.cz
Matthew Dillon [Sat, 20 Oct 2018 03:58:11 +0000 (20:58 -0700)]
kernel - Remove two debug messages
* Remove the debug messages for two cases which have been validated
to occur under synth loads, thus these code paths have been tested.
Matthew Dillon [Fri, 19 Oct 2018 23:24:40 +0000 (16:24 -0700)]
nrelease - Use built boot blocks instead of system boot blocks
* Adjust the nrelease Makefile to install the boot blocks from
the nrelease build into the image instead of the boot blocks
from the running system.
Submitted-by: NI33 (irc)
Aaron LI [Thu, 18 Oct 2018 17:23:51 +0000 (01:23 +0800)]
openresolv: Bind to system as resolvconf(8)
The resolvconf(8) will be used by dhcpcd(8), which will be imported
later to provide out-of-box IPv6 DHCP support, replacing the IPv4-only
dhclient(8) in base and the rtsold(8).
Thanks to swildner for reviewing and providing suggestions.
Aaron LI [Thu, 18 Oct 2018 17:11:18 +0000 (01:11 +0800)]
openresolv: Add READMEs
Aaron LI [Fri, 19 Oct 2018 15:59:51 +0000 (23:59 +0800)]
Merge branch 'vendor/OPENRESOLV'
Aaron LI [Thu, 18 Oct 2018 17:00:16 +0000 (01:00 +0800)]
Import openresolv 3.9.0
Version: 3.9.0
Date: 2016-Dec-30
Homepage: https://roy.marples.name/projects/openresolv
François Tigeot [Mon, 15 Oct 2018 20:17:54 +0000 (22:17 +0200)]
drm/linux: Fix potential ww mutex induced panics
ww mutex locking functions can be called with a NULL ctx and were
happily setting both lock->acquired to non-zero and lock->ctx to
NULL, a combination which other parts of our ww mutex implementation
assume is not possible.
Obtained-from: OpenBSD
Matthew Dillon [Mon, 15 Oct 2018 03:09:47 +0000 (20:09 -0700)]
kernel - Fix NUMA contention due to assymetric memory
* Fix NUMA contention in situations where memory is associated
with CPU cores assymetrically. In particular, with the 2990WX,
half the cores will have no memory associated with them.
* This was forcing DFly to allocate memory from queues belonging to
other nearby cores, causing unnecessary SMP contention, as well
as burn extra time iterating queues.
* Fix by calculating the average number of free pages per-core,
and then adjust any VM page queue with pages less than the average
by stealing pages from queues with greater than the average.
We use a simple iterator to steal pages, so the CPUs with less
(or zero) direct-attached memory will operate more UMA-like
(just on 4K boundaries instead of 256-1024 byte boundaries).
* Tested with a 64-thread concurrent compile test. systat -pv 1
showed all remaining contention disappear. Literally, *ZERO*
contention when we run the test with each thread in its own jail
with no shared resources.
* NOTE! This fix is specific to asymetric NUMA configurations
which are fairly rare in the wild and will not speed up more
conventional systems.
* Before and after timings on the 2990WX.
cd /tmp/src
time make -j 128 nativekernel NO_MODULES=TRUE > /dev/null
BEFORE
703.915u 167.605s 0:49.97 1744.0% 9993+749k 22188+8io 216pf+0w
699.550u 171.148s 0:50.87 1711.5% 9994+749k 21066+8io 150pf+0w
AFTER
678.406u 108.857s 0:45.66 1724.1% 10105+757k 22188+8io 216pf+0w
674.805u 115.256s 0:46.67 1692.8% 10077+755k 21066+8io 150pf+0w
This is a 4.2 second difference on the second run, an over 8%
improvement which is nothing to sneeze at.
François Tigeot [Fri, 12 Oct 2018 20:50:58 +0000 (22:50 +0200)]
drm/radeon: add missing ttm_eu_backoff_reservation to radeon_bo_list_validate
François Tigeot [Thu, 11 Oct 2018 20:40:30 +0000 (22:40 +0200)]
drm/radeon: Finish replacing task queues by Linux work queues
François Tigeot [Thu, 11 Oct 2018 20:38:26 +0000 (22:38 +0200)]
drm/radeon: Reduce differences with Linux 3.18
No functional changes
François Tigeot [Thu, 11 Oct 2018 20:26:42 +0000 (22:26 +0200)]
drm/linux: Add a few headers
Sascha Wildner [Thu, 11 Oct 2018 13:19:43 +0000 (15:19 +0200)]
kernel: Clean up syscalls.master a bit more.
All this and also
f29ea48f7a62b011be75119d46acd81bee2dd8ee could have
been cleaned up in
d6e8ab2d2508bde32ad7b7daaf86d44ec7330986.
While here, also remove some obsolete comments.
Sascha Wildner [Thu, 11 Oct 2018 13:10:05 +0000 (15:10 +0200)]
<vm/vm_extern.h>: Remove some dead code.
Sascha Wildner [Thu, 11 Oct 2018 12:48:08 +0000 (14:48 +0200)]
syscalls.master: Remove a leftover semicolon.
Sascha Wildner [Thu, 11 Oct 2018 12:44:46 +0000 (14:44 +0200)]
kernel/libc: Remove sigstack() remains.
There is no sigstack() syscall anymore in DragonFly, besides being
obsolete, so remove some remains like the manual page and prototype,
and do some related cleanup.
François Tigeot [Sun, 7 Oct 2018 12:47:03 +0000 (14:47 +0200)]
drm: Add linux/interval_tree.h
François Tigeot [Sun, 7 Oct 2018 12:27:27 +0000 (14:27 +0200)]
drm/linux: Add rcu_dereference_protected()
Sascha Wildner [Sat, 6 Oct 2018 07:29:28 +0000 (09:29 +0200)]
kernel: Add further support for fixed-width integer type sysctls.
U32 support was added in
c48b2bac286ed1aaa6b974adfbc844e82bc65f38.
QUAD and UQUAD are now aliases of S64 and U64.
Taken-from: FreeBSD
Matthew Dillon [Sat, 6 Oct 2018 04:32:48 +0000 (21:32 -0700)]
kernel - Refactor tty clist code (2)
* Fix initialization of high water mark, the last commit accidently
removed it.
* Fixes issue with slow console output (emacs, alt-x "shell", then
'ls' or whatever).
Reported-by: tuxillo
Sascha Wildner [Fri, 5 Oct 2018 06:32:18 +0000 (08:32 +0200)]
Remove <sys/clist.h> via 'make upgrade'.
Matthew Dillon [Thu, 4 Oct 2018 22:28:11 +0000 (15:28 -0700)]
kernel - Fix missing tokens in killalllwps()
* There appears to be at least one code path where killalllwps()
is being called without the necessary tokens held.
* Just have the routine itself obtain the necessary tokens.
* Might be responsible for extremely rare core-dump stop/wait
stalls.
Matthew Dillon [Thu, 4 Oct 2018 21:37:12 +0000 (14:37 -0700)]
kernel - Refactor tty clist code
* Remove all the old cruft, completely rewrite the clist code to use
a single linear buffer and a FIFO mechanism.
* The linear buffer just uses 16-bit elements in order to record
TTY_QUOTE along with the character.
* Fixes bug in last commit (lack of global locks around global clist
caches) by removing the cache entirely.
Matthew Dillon [Thu, 4 Oct 2018 17:22:35 +0000 (10:22 -0700)]
kernel - Refactor tty_token, fix SMP performance issues
* Remove most uses of tty_token in favor of per-tty tp->t_token.
This is particularly important for removing bottlenecks related to PTYs,
which are used all over the place. tty_token remains in a few places
managing overall registration and global list manipulation.
* tty structures are now required to be persistent. Implement a sepearate
ttyinit() function. Continue to allow ttyregister() and ttyunregister()
calls, but these no longer presume destruction of the structure.
* Refactor ttymalloc() to take a **tty pointer and interlock allocations.
Allocations are intended to be one-time. ttymalloc() only requires the
tty_token for initial allocations.
* Remove all critical section use that was combined with tty_token and
tp->t_token. Leave only the tokens. The critical sections were
hold-overs going all the way back to pre-SMP days.
* syscons now gets its own token, vga_token. The ISA VGA code and
the framebuffer code also now use this token instead of tty_token.
* The keyboard subsystem now uses kbd_token instead of tty_token.
* A few remaining serial-like devices (snp, nmdm) also get their own
tokens, as well as use the now required tp->t_token.
* Remove use of tty_token in the session management code. This fixes
a niggling performance path since sessions almost universally go
hand-in-hand with fork/exec/exit sequences. Instead we use the
already-existing per-hash session token.
Imre Vadász [Thu, 4 Oct 2018 19:08:28 +0000 (21:08 +0200)]
ichwd - Fix typo in TCO2_CNT register #define.
* The TCO2_CNT isn't actually used in the driver at the moment.
Sascha Wildner [Thu, 4 Oct 2018 16:44:36 +0000 (18:44 +0200)]
ioctl.9: nwfs(4) is no longer in our tree.
It was removed in
6f25d5554e6df42aa5235bf7e3e8e1cc9da25eec.
Sascha Wildner [Thu, 4 Oct 2018 15:51:01 +0000 (17:51 +0200)]
kernel: Remove some old serial drivers.
Namely digi(4), rp(4) and si(4) (the latter along with its userland
utility sicontrol(8)). It is unlikely that these work properly any
more.
In preparation for the upcoming tty_token work by dillon.
Requested-by: dillon
Sascha Wildner [Thu, 4 Oct 2018 06:44:53 +0000 (08:44 +0200)]
kernel/lwkt_token: Fix indent.
Matthew Dillon [Thu, 4 Oct 2018 04:17:51 +0000 (21:17 -0700)]
kernel - Change pool token hash
* Change the pool token hash to something less stupid.
* Improve lwkt.tokens_debug_output to also output the address of
the actual token instead of just its name. This will help us
track down pool token collisions a bit better.
Imre Vadász [Wed, 3 Oct 2018 20:56:19 +0000 (22:56 +0200)]
ichsmb - Add additional PCI IDs to ichsmb.
* This moves from a long switch statement to a lookup table.
Taken-From: FreeBSD
Sascha Wildner [Wed, 3 Oct 2018 16:43:21 +0000 (18:43 +0200)]
Sync ACPICA with Intel's version
20181003:
* Fix a regression that could cause the compiler to fault.
* New remark for excessively long Sleep()s.
For detailed list, please see sys/contrib/dev/acpica/changes.txt.
Matthew Dillon [Tue, 2 Oct 2018 04:12:28 +0000 (21:12 -0700)]
kernel - Remove duplicate TRIM, only trim with the 'trim' mount opt
* ffs_blkfree_cg() was improperly issuing a synchronous
VOP_FREEBLKS() on the underlying device. This issues a
BUF_CMD_FREEBLKS stategy op to the underlying device,
which is executed unconditionally. This basically runs
an unconditional TRIM whether the 'trim' mount flag is
specified or not.
Remove the VOP_FREEBLKS() call.
* For softupdates operation, ffs_blkfree() handles the
'trim' mount flag by issuing a BUF_CMD_FREEBLKS and
sequencing the call to ffs_blkfree_cg() when it completes.
When 'trim' was enabled, *two* TRIM operations were being
executed on the underlying device, and prior to our fix,
if 'trim' was not enabled, *one* TRIM operation would
still be executed instead of zero.
* In many situations... possibly even most situations,
trim operations seriously reduce performance due to
being serialized by AHCI or by the target device. It
is not as useful as people often think it should be
on normal filesystems.
* The removal of the unconditional TRIM significantly improves
UFS performance, meaning primarily installkernel's
since DragonFly doesn't use UFS for its main filesystem by
default any more.
* The 'trim' mount option for UFS will still work as advertised
when coupled with softupdates.
Imre Vadász [Mon, 1 Oct 2018 19:14:01 +0000 (21:14 +0200)]
nmalloc - Try bigcache in posix_memalign() for big PAGE_SIZE aligned allocs.
* This improves performance (by reducing the amount of mmap() calls required
and avoiding some amount of page faults) for applications which frequently
call posix_memalign() with a 4096 byte alignment for allocations bigger
than 8KB. Allocations of 8KB and smaller are efficiently handled by the
SLAB allocator already.
Matthew Dillon [Mon, 1 Oct 2018 18:43:13 +0000 (11:43 -0700)]
nmalloc - Cache 4K and 8K allocations in posix_memalign()
* Adjust posix_memalign() to use the slab allocator for 4K and 8K
allocations same as malloc() does, to improve performance for
programs using posix_memalign().
Matthew Dillon [Mon, 1 Oct 2018 18:40:23 +0000 (11:40 -0700)]
pciconf - Enhance output
* Add more registers defs from FreeBSD to pcireg.h
* Bring expanded pciconf code in form FreeBSD. This code
displays more configuration data for bridges and control
registers, and fixes a broken bus error reporting conditional.
* Also normalize SLT -> SLOT naming in pcireg.h
Aaron LI [Mon, 1 Oct 2018 00:51:28 +0000 (08:51 +0800)]
Bump __DragonFly_version for adding wait6() and waitid()
Matthew Dillon [Sun, 30 Sep 2018 19:45:14 +0000 (12:45 -0700)]
system - Add wait6(), waitid(), and si_pid/si_uid siginfo support
* Add the wait6() system call (header definitions taken from FreeBSD).
This required rearranging kern_wait() a bit. In particular, we now
maintain a hold count of 1 on the process during processing instead
of releasing the hold count early.
* Add waitid() to libc (waitid.c taken from FreeBSD).
* Adjust manual pages (taken from FreeBSD).
* Add siginfo si_pid and si_uid support. This basically allows a process
taking a signal to determine where the signal came from. The fields
already existed in siginfo but were not implemented.
Implemented using a non-queued per-process array of signal numbers.
The last originator sending any given signal is recorded and passed
through to userland in the siginfo.
* Fixes the 'lightdm' X display manager. lightdm relies on si_pid
support. In addition, note that avoiding long lightdm related
latencies and timeouts require a softlink from libmozjs-52.so to
libmozjs-52.so.0 (must be addressed in dports, not addressed in this
commit).
Loosely-taken-from: FreeBSD (wait6, waitid support only)
Reviewed-by: swildner
Sascha Wildner [Sun, 30 Sep 2018 14:59:53 +0000 (16:59 +0200)]
kernel/virtio: Remove some last traces of virtio_if.m.
It was removed in
b817dce2e30a257b86b253a698e63a418758d76d.
Sascha Wildner [Sun, 30 Sep 2018 14:34:15 +0000 (16:34 +0200)]
kernel: Bring in a port of FreeBSD's virtio_random(4) driver.
Tested-by: Diederik de Groot <info@talon.nl>
Sascha Wildner [Sat, 29 Sep 2018 21:55:11 +0000 (23:55 +0200)]
libstand/hammer1: One more -Wundef fix.
Sascha Wildner [Sat, 29 Sep 2018 21:11:49 +0000 (23:11 +0200)]
libcrypt: Fix -Wundef (no functional change).
François Tigeot [Sat, 29 Sep 2018 19:48:35 +0000 (21:48 +0200)]
drm/radeon: drop btc_get_max_clock_from_voltage_dependency_table
François Tigeot [Sat, 29 Sep 2018 19:33:40 +0000 (21:33 +0200)]
drm/radeon/dpm: drop clk/voltage dependency filters for SI
François Tigeot [Sat, 29 Sep 2018 19:22:05 +0000 (21:22 +0200)]
drm/radeon: Pass dma-buf as argument to gem_prime_import_sg_table
Sascha Wildner [Sat, 29 Sep 2018 12:58:14 +0000 (14:58 +0200)]
Fix the __BSD_VISIBLE check in a few headers.
#if must be used, not #ifdef.
Sascha Wildner [Sat, 29 Sep 2018 11:16:27 +0000 (13:16 +0200)]
<sys/random.h>: Further cleanup.
* Remove inclusion of <sys/interrupt.h> which is no longer needed since
7e071e7abb9f0b54413dda805a4239f7b0eb455e. It can now be moved to
kern_memio.c.
* Remove forward declaration of struct thread which is no longer needed
since
fef8985e09e16aeee4a2014876e2e4b758656ac2.
* Fix a typo in a comment (Reported-by: aly).