openbsd
3 years agoKeep sigalg initialiser order consistent - key type, then hash.
jsing [Sun, 27 Jun 2021 17:45:16 +0000 (17:45 +0000)]
Keep sigalg initialiser order consistent - key type, then hash.

This matches the order that sigalgs are specified in.

ok tb@

3 years agoAdd test coverage for TLSv1.3 client hellos.
jsing [Sun, 27 Jun 2021 17:13:23 +0000 (17:13 +0000)]
Add test coverage for TLSv1.3 client hellos.

This is a little bit clunky due to the number of things that vary (largely
thanks to middlebox compatibility mode, along with the versions and key
share extensions), however it works and can be improved at a later date.

3 years agoAdd test coverage for DTLSv1.2 client hellos.
jsing [Sun, 27 Jun 2021 16:55:46 +0000 (16:55 +0000)]
Add test coverage for DTLSv1.2 client hellos.

3 years agoImprove test coverage for SSL_OP_NO_DTLSv1.
jsing [Sun, 27 Jun 2021 16:54:55 +0000 (16:54 +0000)]
Improve test coverage for SSL_OP_NO_DTLSv1.

3 years agoCorrect handling of SSL_OP_NO_DTLSv1.
jsing [Sun, 27 Jun 2021 16:54:14 +0000 (16:54 +0000)]
Correct handling of SSL_OP_NO_DTLSv1.

When converting to TLS flags, we need to also include SSL_OP_NO_TLSv1,
otherwise the TLS equivalent of SSL_OP_NO_DTLSv1 is TLSv1.0 only, which
does not work so well when we try to switch back to DTLS versions.

3 years agoTeach hexdump() how to identify differing bytes.
jsing [Sun, 27 Jun 2021 16:40:25 +0000 (16:40 +0000)]
Teach hexdump() how to identify differing bytes.

This allows differences between the received data and the test data to be
more readily identified.

3 years agoMore appropriately set cipher_list_len when AES acceleration is available.
jsing [Sun, 27 Jun 2021 16:36:53 +0000 (16:36 +0000)]
More appropriately set cipher_list_len when AES acceleration is available.

3 years agoTweak some data types and sprinkle some const.
jsing [Sun, 27 Jun 2021 16:33:30 +0000 (16:33 +0000)]
Tweak some data types and sprinkle some const.

3 years agoIn addition to 2-byte and 3-byte UTF-8 sequences, correctly identify all
schwarze [Sun, 27 Jun 2021 15:53:33 +0000 (15:53 +0000)]
In addition to 2-byte and 3-byte UTF-8 sequences, correctly identify all
4-byte UTF-8 sequences and not just some of them, to keep them together
and avoid passing them on byte by byte, helping tools like tmux(1).

While here, also do all the range tests with < and > rather than &
for uniformity and readability, and add some comments.

Input and OK jca@ and nicm@.

Soeren at Soeren dash Tempel dot net originally reported the bug
and provided an incomplete patch that was used as a starting point,
and he also tested this final patch.

3 years agoAdd Hart State Management functions. These will be needed to spin up
kettenis [Sun, 27 Jun 2021 15:02:25 +0000 (15:02 +0000)]
Add Hart State Management functions.  These will be needed to spin up
the secondary cores.  From FreeBSD.

ok mlarkin@

3 years agoreuse armv7 installboot for riscv64
jsg [Sun, 27 Jun 2021 04:52:01 +0000 (04:52 +0000)]
reuse armv7 installboot for riscv64
ok deraadt@

3 years agoCreate DMA maps with 64-bit capability when appropriate.
visa [Sun, 27 Jun 2021 04:33:40 +0000 (04:33 +0000)]
Create DMA maps with 64-bit capability when appropriate.

OK kettenis@

3 years agoUse config register to determine if 64-bit DMA is available.
visa [Sun, 27 Jun 2021 04:32:31 +0000 (04:32 +0000)]
Use config register to determine if 64-bit DMA is available.

Suggested by and OK kettenis@

3 years agomatch on sifive,fu540-c000-gem
jsg [Sun, 27 Jun 2021 01:58:51 +0000 (01:58 +0000)]
match on sifive,fu540-c000-gem

used by the hifive unmatched device tree in mainline linux and u-boot
ok visa@

3 years agomake usage less chatty; ok mlarkin
jmc [Sat, 26 Jun 2021 18:03:45 +0000 (18:03 +0000)]
make usage less chatty; ok mlarkin

3 years agomake SYNOPSIS match usage; ok ajacoutot
jmc [Sat, 26 Jun 2021 18:02:48 +0000 (18:02 +0000)]
make SYNOPSIS match usage; ok ajacoutot

3 years agoFor some reason the riscv64 locore.S ended up with the copyright license
kettenis [Sat, 26 Jun 2021 17:38:40 +0000 (17:38 +0000)]
For some reason the riscv64 locore.S ended up with the copyright license
from the arm64 locore.S.  But the code was clearly copied from FreeBSD's
riscv64 locore.S.  The license is the same, but the author/attribution isn't.
Fix this.

ok deraadt@

3 years agoFix .Xr order. From mandoc -Tlint.
tb [Sat, 26 Jun 2021 17:36:28 +0000 (17:36 +0000)]
Fix .Xr order. From mandoc -Tlint.

3 years agodelete extra explanations in the usage: messages which are described
deraadt [Sat, 26 Jun 2021 15:42:58 +0000 (15:42 +0000)]
delete extra explanations in the usage: messages which are described
far better in the manual pages
ok jmc

3 years agoMake lazy binding work on riscv64.
kettenis [Sat, 26 Jun 2021 14:50:25 +0000 (14:50 +0000)]
Make lazy binding work on riscv64.

prompted by deraadt@

3 years agoBuild ld.so with --march=rv64imac on riscv64 to be absolutely sure that
kettenis [Sat, 26 Jun 2021 14:47:54 +0000 (14:47 +0000)]
Build ld.so with --march=rv64imac on riscv64 to be absolutely sure that
ld.so doesn't use the FP registers.

ok deraadt@

3 years agoUse AFLAGS when building syscall stubs. Drop AINC wich isn't used.
kettenis [Sat, 26 Jun 2021 14:46:48 +0000 (14:46 +0000)]
Use AFLAGS when building syscall stubs.  Drop AINC wich isn't used.

ok deraadt@

3 years agocad: Implement 64-bit DMA mode
visa [Sat, 26 Jun 2021 10:47:59 +0000 (10:47 +0000)]
cad: Implement 64-bit DMA mode

This lets the driver utilize 64-bit DMA on hardware that supports it.

Currently, riscv64 does not constrain DMA-reachable memory to the 32-bit
range. This caused memory errors with cad(4) on machines that have RAM
above 4GB in the physical address space.

Prompted by Mickael Torres

OK kettenis@

3 years agoAdd riscv64 support. From Mickael Torres.
kettenis [Sat, 26 Jun 2021 09:24:51 +0000 (09:24 +0000)]
Add riscv64 support.  From Mickael Torres.

ok matthieu@, jsg@

3 years agoAdd powerpc64 and riscv64 to the list of architectures that have DRM.
kettenis [Sat, 26 Jun 2021 09:23:24 +0000 (09:23 +0000)]
Add powerpc64 and riscv64 to the list of architectures that have DRM.

ok matthieu@, deraadt@, jsg@

3 years agoRevert last change, which is under an #ifdef __linux__ block so no used.
matthieu [Sat, 26 Jun 2021 06:54:00 +0000 (06:54 +0000)]
Revert last change, which is under an #ifdef __linux__ block so no used.

noticed by jsg@

3 years agoriscv64 struct cpu_info has ci_idepth
jsg [Sat, 26 Jun 2021 02:02:47 +0000 (02:02 +0000)]
riscv64 struct cpu_info has ci_idepth

3 years agosync
jsg [Sat, 26 Jun 2021 00:48:28 +0000 (00:48 +0000)]
sync

3 years agoadd /dev/dri/card0 and /dev/dri/renderD128
jsg [Sat, 26 Jun 2021 00:43:28 +0000 (00:43 +0000)]
add /dev/dri/card0 and /dev/dri/renderD128
ok deraadt@

3 years agoadd /dev/dri/
jsg [Sat, 26 Jun 2021 00:38:38 +0000 (00:38 +0000)]
add /dev/dri/
ok deraadt@

3 years agolet pfsync_request_update actually retry when it overfills a packet.
dlg [Fri, 25 Jun 2021 23:48:30 +0000 (23:48 +0000)]
let pfsync_request_update actually retry when it overfills a packet.

a continue in the middle of a do { } while (0) loop is effectively
a break, it doesnt restart the loop.

without the retry, the code leaked update messages which in turn
made pool_destroy in pfsync destroy trip over a kassert cos items
were still out.

found by and fix tested by hrvoje popovski
ok sashan@

3 years agoMove unused eficall.h files to the Attic.
krw [Fri, 25 Jun 2021 20:40:23 +0000 (20:40 +0000)]
Move unused eficall.h files to the Attic.

3 years agoClean up and remove debug prints, and add a few more relevant prints for
patrick [Fri, 25 Jun 2021 19:55:22 +0000 (19:55 +0000)]
Clean up and remove debug prints, and add a few more relevant prints for
when things go wrong.

3 years agobasic radeondrm / X support for riscv64. Ok kettenis@
matthieu [Fri, 25 Jun 2021 19:27:40 +0000 (19:27 +0000)]
basic radeondrm / X support for riscv64. Ok kettenis@

- add wscons devices
- build radeondrm and add MD uvm bits to support it.

3 years agoReplace instances of the magic number '64' with a nice #define
krw [Fri, 25 Jun 2021 19:24:53 +0000 (19:24 +0000)]
Replace instances of the magic number '64' with a nice #define
BLOCKALIGNMENT. This will make it more obvious where this
512-byte block count could/should be converted to a disk sector
count.

No functional change.

3 years agoadd SIZE_MAX. ok kettenis@
matthieu [Fri, 25 Jun 2021 19:22:51 +0000 (19:22 +0000)]
add SIZE_MAX. ok kettenis@

3 years agoMake sure we translate prefetchable mmio space as well.
kettenis [Fri, 25 Jun 2021 18:55:26 +0000 (18:55 +0000)]
Make sure we translate prefetchable mmio space as well.

From Mickael Torres.

3 years ago1) Finish eliminating all uses of EFI_CALL() used in the tree, allowing for the
krw [Fri, 25 Jun 2021 17:49:49 +0000 (17:49 +0000)]
1) Finish eliminating all uses of EFI_CALL() used in the tree, allowing for the
removal of eficall.h files.

2) Allow booting from 4k-byte sector devices.

3) Don't leak memory after successfull i/o.

The end result is that riscv64 efidev.c and efipxe.c are identical to the
arm64/armv7 versions, efirng.c is identical to the amd64/arm64 versions and
efiboot.c has only the arm64 -> riscv64 changes.

ok kettenis@

3 years agoWhile it seems like we can choose any I/O virtual address for peripheral
patrick [Fri, 25 Jun 2021 17:41:22 +0000 (17:41 +0000)]
While it seems like we can choose any I/O virtual address for peripheral
devices, this isn't really the case.  It depends on the bus topology of
how devices are connected.  In the case of PCIe, devices are assigned
addresses (in PCI BARs) from the PCI address spaces.  Now if we take an
address from one of these address spaces for our IOVA, transfers from
from a PCI device to that address will terminate inside of the PCI bus.
This is because from the PCI buses' point-of-view, the address we chose
is part of its address space.  To make sure we don't allocate addresses
from there, reserve the PCI addresses in the IOVA.

Note that smmu(4) currently gives each device its own IOVA.  So the PCI
addresses will be reserved only in IOVA from PCI devices, and only the
addresses concerning the PCI bus it is connected to will be reserved.
All other devices behind an smmu(4) will not have any changes to their
IOVA.

ok kettenis@

3 years agoAllow (w)hole disk allocation for GPT disks. Use fdisk -A when Apple APFS ISC
krw [Fri, 25 Jun 2021 17:27:07 +0000 (17:27 +0000)]
Allow (w)hole disk allocation for GPT disks. Use fdisk -A when Apple APFS ISC
partition is detected. Otherwise the normal big hammer fdisk -ig.

Only create EFI SYS boot partition on GPT disks that are the ROOTDISK.

ok kettenis@ deraadt@

3 years agoadd linux style memory barriers for risc-v to drm
jsg [Fri, 25 Jun 2021 13:41:09 +0000 (13:41 +0000)]
add linux style memory barriers for risc-v to drm

based on linux operation to rvwmo mapping table in
the rvwmo appendix of the risc-v unprivileged isa spec

ok kettenis@

3 years agoRemove an unused struct.
visa [Fri, 25 Jun 2021 13:29:40 +0000 (13:29 +0000)]
Remove an unused struct.

3 years agouse weaker fences for riscv64 membar
jsg [Fri, 25 Jun 2021 13:25:53 +0000 (13:25 +0000)]
use weaker fences for riscv64 membar

Fences are described in 'RISC-V Unprivileged ISA' syntax is
'fence predecessor,successor'.

"Any combination of device input (I), device output (O), memory reads (R),
and memory writes (W) may be ordered with respect to any combination
of the same."

Previously "fence" was used for membar_* which is short for
"fence iorw,iorw".  Change this to more specific fences based on the
text in membar_sync(9) with store -> w, load -> r.

build test by and ok kettenis@

3 years agoSave quite a bit of space by removing the existence of PTEDs. The
patrick [Fri, 25 Jun 2021 12:40:29 +0000 (12:40 +0000)]
Save quite a bit of space by removing the existence of PTEDs.  The
dynamics of SMMU are a bit different to regular MMU usage, as we do
not need P->V lists or ref/mod emulation (with page access upgrade).
While in the future we might want to save cacheability modes, it is
not necessary right now.  Our PTED construct, which holds that kind
of information, is not needed.  With these gone, we save around 93%
of smmu(4)'s previous memory overhead.

Discussed with drahn@ kettenis@

3 years agoThe network flush code only operates on peerself (like all the other
claudio [Fri, 25 Jun 2021 09:25:48 +0000 (09:25 +0000)]
The network flush code only operates on peerself (like all the other
network commands). Instead of passing the peer as argument to the tree
walker just default to peerself in network_flush_upcall().
OK benno@

3 years agoDo the multiprotocol check first for the IPv4 case. So it is the same
claudio [Fri, 25 Jun 2021 09:23:26 +0000 (09:23 +0000)]
Do the multiprotocol check first for the IPv4 case. So it is the same
everywhere.
OK benno@

3 years agofix decoding of X.509 subject name; from Leif Thuresson via bz3327
djm [Fri, 25 Jun 2021 06:30:22 +0000 (06:30 +0000)]
fix decoding of X.509 subject name; from Leif Thuresson via bz3327
ok markus@

3 years agoUse better language to refer to the user. From l1ving via github
dtucker [Fri, 25 Jun 2021 06:20:39 +0000 (06:20 +0000)]
Use better language to refer to the user.  From l1ving via github
PR#250, ok jmc@

3 years agosync set sizes with latest snapshot
jsg [Fri, 25 Jun 2021 05:22:02 +0000 (05:22 +0000)]
sync set sizes with latest snapshot
initial sizes were from arm64

3 years agosync set sizes with latest snapshot
jsg [Fri, 25 Jun 2021 04:51:52 +0000 (04:51 +0000)]
sync set sizes with latest snapshot
initial sizes were from arm64

3 years agoReplace SIGCHLD/notify_pipe kludge with pselect.
dtucker [Fri, 25 Jun 2021 03:38:17 +0000 (03:38 +0000)]
Replace SIGCHLD/notify_pipe kludge with pselect.

Previously sshd's SIGCHLD handler would wake up select() by writing a
byte to notify_pipe.  We can remove this by blocking SIGCHLD, checking
for child terminations then passing the original signal mask through
to pselect.  This ensures that the pselect will immediately wake up if
a child terminates between wait()ing on them and the pselect.

In -portable, for platforms that do not have pselect the kludge is still
there but is hidden behind a pselect interface.

Based on other changes for bz#2158, ok djm@

3 years agominimalistic diff to use %ld instead of %d for ptrdiff printing
deraadt [Fri, 25 Jun 2021 01:36:04 +0000 (01:36 +0000)]
minimalistic diff to use %ld instead of %d for ptrdiff printing

3 years agoPull in support from a future clang for __GCC_HAVE_SYNC_COMPARE_AND_SWAP_x
deraadt [Fri, 25 Jun 2021 01:35:13 +0000 (01:35 +0000)]
Pull in support from a future clang for __GCC_HAVE_SYNC_COMPARE_AND_SWAP_x
defines because we need it now
from https://reviews.llvm.org/D91784
ok mlarkin kettenis

3 years agoalarm(3): remove superfluous pointer
cheloha [Thu, 24 Jun 2021 22:43:31 +0000 (22:43 +0000)]
alarm(3): remove superfluous pointer

The pointer `itp' doesn't serve any purpose here, remove it.

Since we're changing these lines, we may as well rename `it' to `itv'
to match the existing `oitv'.

Thread: https://marc.info/?l=openbsd-tech&m=162380665115598&w=2

ok millert@

3 years agotrim usage to match the man page;
jmc [Thu, 24 Jun 2021 21:11:40 +0000 (21:11 +0000)]
trim usage to match the man page;
remove -DSEEALSO, as suggested by millert

ok millert

3 years agoAdd Buttonville and Peterborough (ON), both I've flown into.
ian [Thu, 24 Jun 2021 18:40:59 +0000 (18:40 +0000)]
Add Buttonville and Peterborough (ON), both I've flown into.

3 years agoUpdate the name of RNO (name changed in 1994), also add:
mlarkin [Thu, 24 Jun 2021 18:05:02 +0000 (18:05 +0000)]
Update the name of RNO (name changed in 1994), also add:

MEV - Minden-Tahoe airport, Minden, Nevada, USA
CXP - Carson airport, Carson City, Nevada, USA
TKF - Truckee Tahoe airport, California, USA

I have landed at all three.

3 years agoadd some aarch64 bits missed in Makefile.in 1.6
jsg [Thu, 24 Jun 2021 13:27:45 +0000 (13:27 +0000)]
add some aarch64 bits missed in Makefile.in 1.6

ok deraadt@ drahn@

3 years agoSimplify the multiprotocol handling by moving the while loops out of the
claudio [Thu, 24 Jun 2021 13:03:31 +0000 (13:03 +0000)]
Simplify the multiprotocol handling by moving the while loops out of the
switch statement. This way common code is referenced only once.
OK sthen@

3 years agoaspath_deflate() did free the passed in data but since the way aspaths
claudio [Thu, 24 Jun 2021 10:04:05 +0000 (10:04 +0000)]
aspath_deflate() did free the passed in data but since the way aspaths
are processed in the Adj-RIB-Out this is no longer needed since the passed
in pointer is still referenced and is not allowed to be freed.
Adjust the mrt code similar to how up_generate_attr() uses aspath_deflate().
OK sthen@

3 years agoAdd support for the 64-bit prefetchable memory window.
kettenis [Thu, 24 Jun 2021 09:34:17 +0000 (09:34 +0000)]
Add support for the 64-bit prefetchable memory window.

ok patrick@

3 years agoFix add-path capability encoding, the length was not correctly calculated
claudio [Thu, 24 Jun 2021 09:26:18 +0000 (09:26 +0000)]
Fix add-path capability encoding, the length was not correctly calculated
because it included two extra bytes (copy-paste error from graceful restart).

3 years agounveil: cleanup code. no intented functional change.
semarie [Thu, 24 Jun 2021 07:21:59 +0000 (07:21 +0000)]
unveil: cleanup code. no intented functional change.

return early for simple conditions instead of using navigating inside
if-branches.

with and ok claudio@

3 years agorepair missing dependencies against bfd.h for riscv64
deraadt [Thu, 24 Jun 2021 05:41:43 +0000 (05:41 +0000)]
repair missing dependencies against bfd.h for riscv64
ok jsg drahn

3 years agosync
deraadt [Wed, 23 Jun 2021 23:57:43 +0000 (23:57 +0000)]
sync

3 years agoAdjust test. You're not supposed to change errno in a signal handler and
kettenis [Wed, 23 Jun 2021 22:39:31 +0000 (22:39 +0000)]
Adjust test.  You're not supposed to change errno in a signal handler and
count on it being observable in the normal program flow after the signal
handler returns.  Such code would break code that sets errno to 0 and
looks at its value later.  With the recent futex(2) changes this particular
aspect of the test no longer passed.

ok deraadt@, bluhm@

3 years agoThe first page of the I/O virtual address space is reserved so that
patrick [Wed, 23 Jun 2021 19:46:13 +0000 (19:46 +0000)]
The first page of the I/O virtual address space is reserved so that
is easier to spot misconfiguration or wrong behaviour where NULL is
used as address.  Right now that page is not part of the IOVA at all,
so when we reserve regions, like PCI I/O space, which can cover that
page as well, extent(9) will panic.  Instead, include it in the IOVA
but reserve it right away.  This way that page can be reserved twice.

3 years agohelp the debugger look in ports for external parts like PadWalker
espie [Wed, 23 Jun 2021 16:51:15 +0000 (16:51 +0000)]
help the debugger look in ports for external parts like PadWalker
and Readline.

feedback and okay afresh1@

3 years agortsock: revert from timeout_set_flags(9) to timeout_set_proc(9); ok mvs@
cheloha [Wed, 23 Jun 2021 16:10:45 +0000 (16:10 +0000)]
rtsock: revert from timeout_set_flags(9) to timeout_set_proc(9); ok mvs@

3 years agotitmp(4)
kettenis [Wed, 23 Jun 2021 15:32:40 +0000 (15:32 +0000)]
titmp(4)

3 years agoEnable titmp(4).
kettenis [Wed, 23 Jun 2021 15:26:10 +0000 (15:26 +0000)]
Enable titmp(4).

ok deraadt@

3 years agoAdd titmp(4), a driver for the TI TMP451 temperature sensor.
kettenis [Wed, 23 Jun 2021 15:25:39 +0000 (15:25 +0000)]
Add titmp(4), a driver for the TI TMP451 temperature sensor.

ok deraadt@

3 years agoadb(4/macppc): fix adb_cuda_tickle() prototype
cheloha [Wed, 23 Jun 2021 14:12:59 +0000 (14:12 +0000)]
adb(4/macppc): fix adb_cuda_tickle() prototype

Timeout callback functions are of type void (*)(void *).
adb_cuda_tickle() needs a void pointer for a first parameter.

ok mpi@

3 years agoIn unveil_add_vnode() refactor code around the indexes i and j. In one
claudio [Wed, 23 Jun 2021 14:09:01 +0000 (14:09 +0000)]
In unveil_add_vnode() refactor code around the indexes i and j. In one
place the wrong index is used resulting in re-evaluating all unveil nodes.
Also loop over over all but the last (just added vnode) -- again there is
no need to re-evaluate the cover of the just added unveil.
OK anton@ semarie@

3 years agoMake sure the bus is idle before starting a transfer.
kettenis [Wed, 23 Jun 2021 13:39:12 +0000 (13:39 +0000)]
Make sure the bus is idle before starting a transfer.

ok deraadt@

3 years agoThe value of -l should be treated as a 512-byte block count.
krw [Wed, 23 Jun 2021 13:07:13 +0000 (13:07 +0000)]
The value of -l should be treated as a 512-byte block count.

Tweak man page.

3 years agoUse print_host() to log destination, netmask and gateway. Add pretty
tobhe [Wed, 23 Jun 2021 12:21:23 +0000 (12:21 +0000)]
Use print_host() to log destination, netmask and gateway.  Add pretty
printing for route flags.

ok markus@

3 years agoFactor out vroute_addr().
tobhe [Wed, 23 Jun 2021 12:11:40 +0000 (12:11 +0000)]
Factor out vroute_addr().

ok markus@

3 years agobtrace(8): init and update timespec for BEGIN/END event
dv [Wed, 23 Jun 2021 11:24:01 +0000 (11:24 +0000)]
btrace(8): init and update timespec for BEGIN/END event

BEGIN and END use a fake dt(4) event, so in order to use the nsecs
var or time() it needs a timespec set. Init for BEGIN and update
at END.

ok mpi@

3 years agoGarbage collect prototoype for ssl_parse_serverhello_tlsext() which
tb [Wed, 23 Jun 2021 11:12:33 +0000 (11:12 +0000)]
Garbage collect prototoype for ssl_parse_serverhello_tlsext() which
was removed in t1_lib.c r1.141.

3 years agoaugment the global pf state list with its own locks.
dlg [Wed, 23 Jun 2021 06:53:51 +0000 (06:53 +0000)]
augment the global pf state list with its own locks.

before this, things that iterated over the global list of pf states
had to take the net, pf, or pf state locks. in particular, the
ioctls that dump the state table took the net and pf state locks
before iterating over the states and using copyout to export them
to userland. when we tried replacing the use rwlocks with mutexes
under the pf locks, this blew up because you can't sleep when holding
a mutex and there's a sleeping lock used inside copyout.

this diff introduces two locks around the global state list: a mutex
that protects the head and tail of the list, and an rwlock that
protects the links between elements in the list. inserts on the
state list only occur during packet handling and can be done by
taking the mutex and putting the state on the tail before releasing
the mutex. iterating over states is only done from thread/process
contexts, so we can take a read lock, then the mutex to get a
snapshot of the head and tail pointers, and then keep the read lock
to iterate between the head and tail points. because it's a read
lock we can then take other sleeping locks (eg, the one inside
copyout) without (further) gymnastics. the pf state purge code takes
the rwlock exclusively and the mutex to remove elements from the
list.

this allows the ioctls and purge code to loop over the list
concurrently and largely without blocking the creation of states
when pf is processing packets.

pfsync also iterates over the state list when doing bulk sends,
which the state purge code needs to be careful around.

ok sashan@

3 years agopf_purge_expired_states can check the time once instead of for every state.
dlg [Wed, 23 Jun 2021 05:51:27 +0000 (05:51 +0000)]
pf_purge_expired_states can check the time once instead of for every state.

3 years agopfsync_undefer_notify needs to be careful before dereferecing state keys.
dlg [Wed, 23 Jun 2021 05:43:53 +0000 (05:43 +0000)]
pfsync_undefer_notify needs to be careful before dereferecing state keys.

pfsync_undefer_notify uses the state keys to look up the address
family, which is used to figure out if it should call ipv4 or ipv6
functions. however, the pf state purge code can unlink a state from
the trees (ie, the state keys get removed) while the pfsync defer
code is holding a reference to it and expects to be able to send
the deferred packet in the future. we can test if the state keys
are set by checking if the timeout state is PFTM_UNLINK or not.

this currently relies on both pf_remove_state and pfsync_undefer_notify
being called with the NET_LOCK held. this probably needs to be
rethought later but is good enough for now.

found the hard way on a production firewall at work.

3 years agorework pf_state_expires to avoid confusion around state->timeout.
dlg [Wed, 23 Jun 2021 04:16:32 +0000 (04:16 +0000)]
rework pf_state_expires to avoid confusion around state->timeout.

im going to make it so pf_purge_expired_states() can gather states
largely without sharing a lock with pfsync or actual packet processing
in pf. if pf or pfsync unlink a state while pf_purge_expired_states
is looking at it, we can race with some checks and fall over a
KASSERT.

i'm fixing this by having the caller of pf_state_expires read
state->timeout first, do it's checks, and then pass the value as
an argument into pf_state_expires. this means there's a consistent
view of the state->timeout variable across all the checks that
pf_purge_expired_states in particular does. if pf/pfsync does change
the timeout while pf_purge_expired_states is looking at it, the
worst thing that happens is that it doesn't get picked as a candidate
for purging in this pass and will have to wait for the next sweep.

ok sashan@ as part of a bigger diff

3 years agosubstantially shorten the usage() message
deraadt [Wed, 23 Jun 2021 02:53:44 +0000 (02:53 +0000)]
substantially shorten the usage() message
pieces by jmc, pieces by me

3 years agoreduce verbosity in usage and make it match SYNOPSIS;
jmc [Tue, 22 Jun 2021 20:19:28 +0000 (20:19 +0000)]
reduce verbosity in usage and make it match SYNOPSIS;
i reformatted it to match 80w, and removed two functions that
had been used to populate usage: getdistoptlist and msgprusage;

ok millert

3 years agoreduce verbosity in usage and make it match SYNOPSIS;
jmc [Tue, 22 Jun 2021 20:16:36 +0000 (20:16 +0000)]
reduce verbosity in usage and make it match SYNOPSIS;
ok millert

3 years agoreduce verbosity in usage and make it match SYNOPSIS;
jmc [Tue, 22 Jun 2021 20:14:25 +0000 (20:14 +0000)]
reduce verbosity in usage and make it match SYNOPSIS;
ok millert

3 years agoreduce verbosity in usage and make it match SYNOPSIS;
jmc [Tue, 22 Jun 2021 20:12:17 +0000 (20:12 +0000)]
reduce verbosity in usage and make it match SYNOPSIS;
ok millert

3 years agozap wonky commas;
jmc [Tue, 22 Jun 2021 20:01:19 +0000 (20:01 +0000)]
zap wonky commas;

3 years agoReduce verbosity in tset/reset usage()
tb [Tue, 22 Jun 2021 18:33:48 +0000 (18:33 +0000)]
Reduce verbosity in tset/reset usage()

from jmc, ok kn

3 years agoHoist initialization of _nc_progname to start of main()
tb [Tue, 22 Jun 2021 18:32:24 +0000 (18:32 +0000)]
Hoist initialization of _nc_progname to start of main()

This way early calls to err(), failed() and usage() show "reset"
instead of "tset" if the program was invoked as reset(1).

Issue noted by jmc
ok kn

3 years agoClarify tls_config_set_*_file() file I/O semantics
kn [Tue, 22 Jun 2021 17:59:48 +0000 (17:59 +0000)]
Clarify tls_config_set_*_file() file I/O semantics

tls_config_set_*_file(3) do not just set the file paths like
tls_config_set_*_path(3) do, they do load the given file(s) into memory
directly using tls_config_load_file().

This distinction is important because it means a later tls_connect(3)
will not do any file I/O (at least wrt. those files), which is relevant when
for example pleding without "[rwc]path" after loading files into memory and
before doing tls_connect(3).

The manual's current wording made me use the following due to above way of
pledging a program:
tls_load_file()
tls_config_set_ca_mem()
tls_unload_file()

While in fact a single tls_config_set_ca_file() call does the same.

tls_config.c r1.26 (Aug 2016) change the code but forgot to amend the manual
as noted by tb, thanks.

Feedback OK tb

3 years agoAdjust expected output after the change to show more proper info in Adj-RIB-Out
claudio [Tue, 22 Jun 2021 17:57:07 +0000 (17:57 +0000)]
Adjust expected output after the change to show more proper info in Adj-RIB-Out

3 years agobwfm(4) on PCI isn't really MPSAFE, and I'm not sure how this flag
patrick [Tue, 22 Jun 2021 16:36:59 +0000 (16:36 +0000)]
bwfm(4) on PCI isn't really MPSAFE, and I'm not sure how this flag
even got there in the first place.  I've been wondering why I have
seen a bit of mbuf corruption here and there since I put the bwfm(4)
M.2 PCIe card into my arm64 machine.  Well, duh.

3 years agoPp not required before list items;
jmc [Tue, 22 Jun 2021 15:01:55 +0000 (15:01 +0000)]
Pp not required before list items;

3 years agoreduce verbosity of usage(), making it match SYNOPSIS;
jmc [Tue, 22 Jun 2021 14:52:33 +0000 (14:52 +0000)]
reduce verbosity of usage(), making it match SYNOPSIS;
ok kettenis

3 years agoreduce verbosity of usage(); keep it the same as SYNOPSIS, wrapped at 80w.
jmc [Tue, 22 Jun 2021 14:51:29 +0000 (14:51 +0000)]
reduce verbosity of usage(); keep it the same as SYNOPSIS, wrapped at 80w.
lowercase for "usage", and add -e, which was missing;

ok kettenis

3 years agoThe values for blocks and offset in -b should be treated as 512-byte block
krw [Tue, 22 Jun 2021 14:01:58 +0000 (14:01 +0000)]
The values for blocks and offset in -b should be treated as 512-byte block
counts.

Tweak man page accordingly.

Requested by deraadt@ and kettenis@.

3 years agoFix uninitialized variables introduced in rev 1.361
jca [Mon, 21 Jun 2021 22:09:14 +0000 (22:09 +0000)]
Fix uninitialized variables introduced in rev 1.361

Thankfully clang elided the code in an almost harmless way (at least on
amd64 GENERIC.MP).  Spotted by chance when building kernels
with -Wno-error=uninitialized.

ok dlg@ sashan@ bluhm@