openbsd
5 months agoMake the kernel compile also when turning on debugging.
mglocker [Fri, 10 May 2024 10:49:10 +0000 (10:49 +0000)]
Make the kernel compile also when turning on debugging.

ok mpi@

5 months agoRegen
claudio [Fri, 10 May 2024 09:21:41 +0000 (09:21 +0000)]
Regen

5 months agoThe ptsignal() race against p_sigmask changes by dosigsuspend() are fixed.
claudio [Fri, 10 May 2024 09:21:01 +0000 (09:21 +0000)]
The ptsignal() race against p_sigmask changes by dosigsuspend() are fixed.
Unlock sigsuspend() and __thrsigdivert() again.

5 months agoCorrect display the details of COMMAND using the ps command from the kernel
asou [Fri, 10 May 2024 06:46:14 +0000 (06:46 +0000)]
Correct display the details of COMMAND using the ps command from the kernel
crash dump.

OK millert@

5 months agoFix broken debugging.
mglocker [Fri, 10 May 2024 06:14:10 +0000 (06:14 +0000)]
Fix broken debugging.

5 months agoAdd missing EC_KEY_free()
tb [Fri, 10 May 2024 05:12:03 +0000 (05:12 +0000)]
Add missing EC_KEY_free()

While eckey_from_explicit_params() frees *out_eckey, eckey_from_object()
and eckey_from_params() do not. These functions are currently all callled
with a NULL *out_eckey, but the latter two would leak if that should ever
change.

ok jsing

5 months agoRemove fixed nonce length information from algorithm2
tb [Fri, 10 May 2024 05:08:05 +0000 (05:08 +0000)]
Remove fixed nonce length information from algorithm2

This information has been part of tls12_key_block_generate() for a while
now. It remained in this table because at that point SSL_CIPHER was still
public. Nothing can access algorithm2 anymore from the outside, so this is
dead weight.

ok jsing

5 months agoInline dsa_builtin_keygen() in DSA_generate_key()
tb [Fri, 10 May 2024 04:53:55 +0000 (04:53 +0000)]
Inline dsa_builtin_keygen() in DSA_generate_key()

ok djm

5 months agomake pf_match_rule() prototype match the function
jsg [Fri, 10 May 2024 03:50:12 +0000 (03:50 +0000)]
make pf_match_rule() prototype match the function

5 months agoMake the openssl_dsa_meth static const
tb [Thu, 9 May 2024 20:57:49 +0000 (20:57 +0000)]
Make the openssl_dsa_meth static const

5 months agoMove openssl_dsa_meth below the methods it uses
tb [Thu, 9 May 2024 20:56:52 +0000 (20:56 +0000)]
Move openssl_dsa_meth below the methods it uses

no functional change

5 months agoMake the DH_METHOD static const
tb [Thu, 9 May 2024 20:43:36 +0000 (20:43 +0000)]
Make the DH_METHOD static const

5 months agoMove public API and DH_METHOD to the bottom of the file
tb [Thu, 9 May 2024 20:40:42 +0000 (20:40 +0000)]
Move public API and DH_METHOD to the bottom of the file

no functional change

5 months agosync the SSL text; ok tb
jmc [Thu, 9 May 2024 17:57:36 +0000 (17:57 +0000)]
sync the SSL text; ok tb

5 months ago- drop ref to non-installed g++(1) page
jmc [Thu, 9 May 2024 17:22:20 +0000 (17:22 +0000)]
- drop ref to non-installed g++(1) page
- drop ref to clang that isn;t particularly helpful and only relevant
to some platforms, as suggested by kettenis

5 months agoEnable ufshci(4) on amd64.
mglocker [Thu, 9 May 2024 17:05:22 +0000 (17:05 +0000)]
Enable ufshci(4) on amd64.

ok kettenis@ deraadt@

5 months agoTiny style tweaks in X509_REQ_add_extension_nid()
tb [Thu, 9 May 2024 14:29:08 +0000 (14:29 +0000)]
Tiny style tweaks in X509_REQ_add_extension_nid()

Test & assign and use ret instead of rv.

ok jsing

5 months agoStreamline X509_REQ_check_private_key() a bit
tb [Thu, 9 May 2024 14:27:21 +0000 (14:27 +0000)]
Streamline X509_REQ_check_private_key() a bit

Use better variable names, split the success from the error path and
return directly rather than using an ok variable.

ok jsing

5 months agoZap some extra parentheses in X509_REQ_get_pubkey()
tb [Thu, 9 May 2024 14:22:16 +0000 (14:22 +0000)]
Zap some extra parentheses in X509_REQ_get_pubkey()

ok jsing

5 months agoClean up X509_to_X509_REQ()
tb [Thu, 9 May 2024 14:20:57 +0000 (14:20 +0000)]
Clean up X509_to_X509_REQ()

Use better variable names. X509_REQ_new() sets the version to the only
specified version, so there is no point to set it. Extract the subject
name, then assign to make it more obvious that we error happens if the
cert has a missing subject. Switch to X509_get0_pubkey() to avoid some
strange dance with a strangely named variable to adjust the refcount.

ok jsing

5 months agoFurther simplify X509_REQ_get_extensions()
tb [Thu, 9 May 2024 14:00:52 +0000 (14:00 +0000)]
Further simplify X509_REQ_get_extensions()

Instead of inlining a poor version of ASN1_TYPE_unpack_sequence() with
missing error checks, just call the real thing. It's safer and simpler.

ok jsing

5 months agosimplify exit message handling, which was more complicated than
djm [Thu, 9 May 2024 09:46:47 +0000 (09:46 +0000)]
simplify exit message handling, which was more complicated than
it needed to be because of unexpunged ssh1 remnants. ok markus@

5 months agoctime(3) and ctime_r(3) can fail when timestamps are way off.
florian [Thu, 9 May 2024 08:35:40 +0000 (08:35 +0000)]
ctime(3) and ctime_r(3) can fail when timestamps are way off.
Add missing error checks to all calls under sbin/

Input kettenis, millert
OK millert

5 months agoctime(3) and ctime_r(3) can fail when timestamps are way off.
florian [Thu, 9 May 2024 08:35:03 +0000 (08:35 +0000)]
ctime(3) and ctime_r(3) can fail when timestamps are way off.
Add missing error checks to all calls under libexec/

Input kettenis, millert
OK millert

5 months agoBack then I faced intermittent file-system corruptions for which setting
mglocker [Thu, 9 May 2024 08:24:09 +0000 (08:24 +0000)]
Back then I faced intermittent file-system corruptions for which setting
FUA (Force Unit Access) did help.  In the meantime it turned out that
those file-system corruptions were most likely caused by the slot issues.
Now that we fixed the slot management and limited to one slot for now,
remove FUA again, which increases the write performance significantly.

5 months agoSprinkle some more DPRINTFs.
mglocker [Thu, 9 May 2024 08:21:52 +0000 (08:21 +0000)]
Sprinkle some more DPRINTFs.

5 months agoPack hardware descriptor structures.
mglocker [Thu, 9 May 2024 08:20:22 +0000 (08:20 +0000)]
Pack hardware descriptor structures.

5 months agoParse the OCS response value for completed commands, and set error on
mglocker [Thu, 9 May 2024 08:18:20 +0000 (08:18 +0000)]
Parse the OCS response value for completed commands, and set error on
failure.

5 months agoPerform bus DMA synchronization to update the command descriptors.
mglocker [Thu, 9 May 2024 08:16:32 +0000 (08:16 +0000)]
Perform bus DMA synchronization to update the command descriptors.

5 months agoUse 1U for bit operations on 32-bit registers.
mglocker [Thu, 9 May 2024 08:13:57 +0000 (08:13 +0000)]
Use 1U for bit operations on 32-bit registers.

Proposed by dlg@

5 months agoDon't do math on KVA to get the required slot offset since that could
mglocker [Thu, 9 May 2024 08:12:22 +0000 (08:12 +0000)]
Don't do math on KVA to get the required slot offset since that could
cause invalid pointers depending on the compiler interpretation of
(void *).  Instead work with the structure pointer itself.

Proposed by dlg@

5 months agoDon't use the task id for UPIU commands. We don't use task management
mglocker [Thu, 9 May 2024 08:09:17 +0000 (08:09 +0000)]
Don't use the task id for UPIU commands.  We don't use task management
commands yet.

5 months agoDon't schedule interrupt aggregation when commands are still in-progress.
mglocker [Thu, 9 May 2024 08:06:42 +0000 (08:06 +0000)]
Don't schedule interrupt aggregation when commands are still in-progress.
As of the documentation:

"NOTE Write operations to IACTH and IATOVAL are only allowed when no
commands are outstanding."

Instead we only schedule interrupt aggregation at the start of the
SCSI command call, when all commands have completed.

5 months agoPrevent that scheduling of new commands is interfering with processing
mglocker [Thu, 9 May 2024 08:04:48 +0000 (08:04 +0000)]
Prevent that scheduling of new commands is interfering with processing
of completed commands with a command mutex.

5 months agoDon't relay on the doorbell register to track our slots. As of the
mglocker [Thu, 9 May 2024 08:02:59 +0000 (08:02 +0000)]
Don't relay on the doorbell register to track our slots.  As of the
documentation:

"UTRLDBR is a volatile register; software should only use its value to
determine commands that have completed, not to determine which commands
have previously been issued."

Instead we use the CCB structure to track our slots, as proposed by dlg@.

CAVEAT: Since using more than one slot is currently causing OCS errors,
we limit the slots to one until we can find a solution.

5 months agossl_ciph.c: unwrap a line
tb [Thu, 9 May 2024 07:55:48 +0000 (07:55 +0000)]
ssl_ciph.c: unwrap a line

5 months agoRemove leftover logic of SSL2 support
tb [Thu, 9 May 2024 07:47:50 +0000 (07:47 +0000)]
Remove leftover logic of SSL2 support

SSL2_CF_8_BYTE_ENC was set by things such as RC4_64_WITH_MD5, which fell
victim to tedu's axe a decade ago. Zap that.

ok jsing

5 months agoPlug a "leak" in ssl_security_group()
tb [Thu, 9 May 2024 07:12:03 +0000 (07:12 +0000)]
Plug a "leak" in ssl_security_group()

The way the CBB API is used, CBB_add_u16() and CBB_finish() can't actually
fail here, but if they could, cbb->base would leak. Rewrite this code with
the proper idioms to make it look right.

ok jsing

5 months agoAlign RSA and EC key generation with each other
tb [Thu, 9 May 2024 06:08:11 +0000 (06:08 +0000)]
Align RSA and EC key generation with each other

Being two different cryptographic primitives, it is clear that there must
be some differences between RSA and EC keygen, but they don't have to be
entirely different. We need to set the key type, RSA needs a bit size and
ECDSA needs a curve. That's all the differences there need to be.

Garbage collect a few useless elses and avoid two exit labels paths where
one would do just fine.

As another small bonus, this file no longer uses "deprecated API", so the
portable fork can get rid of an ugly openssl 3 patch if they want to.

ok florian

5 months agoSuppress cache-info dmesg lines when they are identical to the
guenther [Wed, 8 May 2024 18:00:55 +0000 (18:00 +0000)]
Suppress cache-info dmesg lines when they are identical to the
previous cpu.

testing on hybrid box by jmatthew@
suggestions from kettenis@
ok deraadt@

5 months agoixl(4): force mss of tso packets in hardware supported range.
jan [Wed, 8 May 2024 17:52:11 +0000 (17:52 +0000)]
ixl(4): force mss of tso packets in hardware supported range.

ok bluhm@

5 months agofix line wrapping in function definition
tb [Wed, 8 May 2024 16:35:05 +0000 (16:35 +0000)]
fix line wrapping in function definition

5 months ago- for pwraction, point to acpibtn(4)
jmc [Wed, 8 May 2024 15:30:26 +0000 (15:30 +0000)]
- for pwraction, point to acpibtn(4)
- for lidaction, document the value 0
- for lidaction, adjust the description to a format similar
to that of pwraction

ok kettenis deraadt

5 months agoAdd more regress coverage for lhash.
jsing [Wed, 8 May 2024 15:13:23 +0000 (15:13 +0000)]
Add more regress coverage for lhash.

5 months agofix iwx(4) monitor mode
stsp [Wed, 8 May 2024 14:03:54 +0000 (14:03 +0000)]
fix iwx(4) monitor mode

Monitor mode was broken by a recent firmware update. Two tweaks make
it work again:

1) The firmware does not like us sending the power-mode command while
in monitor mode and will panic, so simply don't do that.

2) We no longer add two queues while in monitor mode, just one queue
for frame injection. This queue's index will be 1, not 2. Make the
driver expect the correct index to prevent an error when monitor mode
is entered.

tested by jmc@ and myself on iwx ax200

5 months agodisable the regular ieee80211_encap() Tx path in monitor mode
stsp [Wed, 8 May 2024 14:02:59 +0000 (14:02 +0000)]
disable the regular ieee80211_encap() Tx path in monitor mode

Frames injected from user space carry the DLT_IEEE802_11_RADIO
mbuf tag, and are handled as a special case. Do not fall back
to regular encapsulation while we are in monitor mode and the
frame injected by userspace is found to be invalid.

This fixes an issue when iwx(4) runs in monitor mode with addresses
configured on the interface and leaving 11n/11ac mode directly for
monitor mode. In this case, traffic generated by userspace or the
kernel (such as ICMPv6) would trigger Tx attempts, which in turn
would trigger an attempt to set up a block ACK agreement and then
cause a firmware panic.

This points at a related issue where interface configuration state
is not properly cleaned up while switching into monitor mode.
The 11n/11ac interface config should ideally be cleared completely,
preventing block ack from being initiated.

But preventing the stack from trying to send frames down the regular
Tx path in monitor mode is a good idea in general because drivers may
not handle this very well for various reasons, block ack being just one.

tested by jmc@ and myself on iwx ax200

5 months agoRework how action SIG_HOLD is handled in ptsignal.
claudio [Wed, 8 May 2024 13:05:33 +0000 (13:05 +0000)]
Rework how action SIG_HOLD is handled in ptsignal.

Since we want to unlock sigsuspend, ptsignal needs to double check in the
SSLEEP case that the signal being delivered is still masked or unmasked.
Remove the early return for action SIG_HOLD so that the SSLEEP case can
properly recheck the sigmask.

On top of this update siglist only in one place at the end of ptsignal
this now includes the clearing of signals for the SA_CONT and SA_STOP
cases.

OK mpi@

5 months agoFix route leak in ip input.
bluhm [Wed, 8 May 2024 13:01:30 +0000 (13:01 +0000)]
Fix route leak in ip input.

In previous commit when refactoring the route cache, a rtfree() has
been forgotten.  For each forwarded packet the reference counter
of the route entry was increased.  This eventually leads to an
integer overflow and triggers kassert.

reported by and OK jan@

5 months agoAvoid OpenSSL SSL repetitions
tb [Wed, 8 May 2024 09:41:33 +0000 (09:41 +0000)]
Avoid OpenSSL SSL repetitions

with the help of jmc

5 months agoavoid various repetitions in the descriptions of libcrypto and libssl
tb [Wed, 8 May 2024 08:24:23 +0000 (08:24 +0000)]
avoid various repetitions in the descriptions of libcrypto and libssl

with/ok jmc

5 months agoSimplify X509_REQ_get_extensions()
tb [Wed, 8 May 2024 08:20:08 +0000 (08:20 +0000)]
Simplify X509_REQ_get_extensions()

Now that we know the two OIDs we need to look for when checking for the
extension list attribute in a certification request, we can simplify this
quite a bit. There is one change of behavior. Attribute value sets are not
supposed to be empty and it makes no sense to return an empty stack of
extensions in that case, return NULL instead, matching BoringSSL.

This removes last use of ext_nids and ext_nid_list[], so these two bits
of unprotected global mutable state can now join the party in the attic.

ok jsing

5 months agoSimplify X509_REQ_extension_nid()
tb [Wed, 8 May 2024 08:11:50 +0000 (08:11 +0000)]
Simplify X509_REQ_extension_nid()

Now that the global ext_nids[] array can no longer be modified by the
application, we can simplify this by returning the two possible NIDs
that we accept in the extension list attribute in PKCS#10 certification
requests.

The year is 2024. This API is entirely unused by the ecosystem. Well not
entirely! One small village of indomitable rare API use still holds out
against the cleansers. You may have guessed it: security/xca.

ok jsing

5 months agoDefang X509_REQ_{s,g}et_extension_nids()
tb [Wed, 8 May 2024 07:55:10 +0000 (07:55 +0000)]
Defang X509_REQ_{s,g}et_extension_nids()

These fiddle with unprotected global state, so aren't thread safe and
of course there was no good reason to have this API in the first place.
Nothing uses it, so it becomes a noop and will be removed in the next
major bump.

ok jsing

5 months agosymbols test: drop headers that don't define any symbols
tb [Wed, 8 May 2024 06:54:43 +0000 (06:54 +0000)]
symbols test: drop headers that don't define any symbols

5 months agoavoid memcpy(malloc(0), ..., 0), which is not portable.
djm [Tue, 7 May 2024 23:40:53 +0000 (23:40 +0000)]
avoid memcpy(malloc(0), ..., 0), which is not portable.
ok florian@

5 months agoopenssl: toolkit implementing the TLS v1 protocol is weird
tb [Tue, 7 May 2024 21:00:18 +0000 (21:00 +0000)]
openssl: toolkit implementing the TLS v1 protocol is weird

Well, it's a toolkit alright, and a terrible one at that, but TLS v1
(which is this beloved toolkit's name for TLS v1.0) is a thing firmly
from the past, so drop the v1.

5 months agoPEM_read_bio_PrivateKey: fix grammar
tb [Tue, 7 May 2024 20:40:07 +0000 (20:40 +0000)]
PEM_read_bio_PrivateKey: fix grammar

This old [...] routines use [...] -> These old [...] routines [...]

5 months agoAdditional check for TSO packets with 0 MSS.
jan [Tue, 7 May 2024 18:35:23 +0000 (18:35 +0000)]
Additional check for TSO packets with 0 MSS.

Tested by bluhm

ok bluhm@

5 months agorw_enter() with RW_NOSLEEP returns EBUSY and not the expected EWOULDBLOCK
claudio [Tue, 7 May 2024 15:54:23 +0000 (15:54 +0000)]
rw_enter() with RW_NOSLEEP returns EBUSY and not the expected EWOULDBLOCK

This fixes random gmake failures during ports builds caused by:
   gmake[2]: *** read jobs pipe: Device busy.  Stop.
Fix verified by tb@ on his bulk build box
OK mvs@ tb@

6 months agoClear MNT_LOCAL flag on FUSE file system. It can be local or remote, but
mvs [Tue, 7 May 2024 14:27:11 +0000 (14:27 +0000)]
Clear MNT_LOCAL flag on FUSE file system. It can be local or remote, but
kernel can't tell the difference.

From Kirill A. Korinsky

ok claudio mpi

6 months agodrop the MD byte-swap micro-optimizations on clang architectures
naddy [Tue, 7 May 2024 14:26:48 +0000 (14:26 +0000)]
drop the MD byte-swap micro-optimizations on clang architectures

The compiler already translates the generic code into arithmetic
byte-swap instructions or byte-swapping memory load and store
instructions if available on an architecture.

ok deraadt@ guenther@

6 months agoReorder functions and drop static function prototypes.
jsing [Tue, 7 May 2024 13:40:42 +0000 (13:40 +0000)]
Reorder functions and drop static function prototypes.

No functional change.

6 months agoadd smtpd-tables.7 for real; spotted by tb, thanks!
op [Tue, 7 May 2024 12:13:43 +0000 (12:13 +0000)]
add smtpd-tables.7 for real; spotted by tb, thanks!

6 months agochange the smtpd table protocol
op [Tue, 7 May 2024 12:10:06 +0000 (12:10 +0000)]
change the smtpd table protocol

Using imsg for the "proc" table (external programs) has proven quite
painful in practice since a lot of smtpd internals (structs, enums,
etc..) have to be kept in sync with the various tables implementations.

Instead, a filter-like protocol for tables decouples the implementations
and allows to write and test tables easily.

The new text-based transport protocol is documented in the (added)
smtpd-tables(7) manpage.

The old imsg protocol is no longer supported and existing tables have to
be converted.  In particular, users of opensmtpd-extras tables will need
install the new opensmtpd-table-* packages.

With lots of suggestions and improvements from gilles and a tweak
from Philipp (philipp+openbsd [at] bureaucracy [dot] de), thanks!

ok gilles

6 months agoIn Rev 1.296 the update of the siglist was moved to the end of ptsignal().
claudio [Tue, 7 May 2024 10:46:35 +0000 (10:46 +0000)]
In Rev 1.296 the update of the siglist was moved to the end of ptsignal().
One atomic_clearbits_int() hiding in SSTOP was missed when converting all
the exceptions that cleared the siglist again. Instead of clearing the bits
the mask needs to be set to 0 so that it is properly ignored.
OK mpi@

6 months agosync
op [Tue, 7 May 2024 09:10:58 +0000 (09:10 +0000)]
sync

6 months agoinstall smtpd-tables(7)
op [Tue, 7 May 2024 09:10:33 +0000 (09:10 +0000)]
install smtpd-tables(7)

6 months agosync
deraadt [Tue, 7 May 2024 00:44:02 +0000 (00:44 +0000)]
sync

6 months agoremove SSH1 leftovers
tobias [Mon, 6 May 2024 19:26:17 +0000 (19:26 +0000)]
remove SSH1 leftovers

Authored with Space Meyer <git at the-space dot agency>

ok djm

6 months agovmstat/dkstats.c: remove obfuscatory timerset() macro
cheloha [Mon, 6 May 2024 16:54:22 +0000 (16:54 +0000)]
vmstat/dkstats.c: remove obfuscatory timerset() macro

Using a macro to perform a plain struct assignment obfuscates the
obvious.  Delete the timerset() macro.

While here, remove some superfluous timerclear(3) calls.

Thread: https://marc.info/?l=openbsd-tech&m=171346446031940&w=2

ok millert@

6 months agotty(1): exit with status 2 if unveil(2) or pledge(2) fail
cheloha [Mon, 6 May 2024 16:49:46 +0000 (16:49 +0000)]
tty(1): exit with status 2 if unveil(2) or pledge(2) fail

tty(1) returns >1 on error, not 1.

Thread: https://marc.info/?l=openbsd-tech&m=171494082629111&w=2

ok deraadt@ guenther@

6 months agoFix function wrapping.
jsing [Mon, 6 May 2024 14:38:20 +0000 (14:38 +0000)]
Fix function wrapping.

6 months agoEnable lhash regress.
jsing [Mon, 6 May 2024 14:37:26 +0000 (14:37 +0000)]
Enable lhash regress.

6 months agoGuard call to contract() from doall_util_fn().
jsing [Mon, 6 May 2024 14:36:05 +0000 (14:36 +0000)]
Guard call to contract() from doall_util_fn().

It is not safe to unconditionally call contract() - when called repeatedly
it will shrink the bucket array to zero and then attempt to access that
allocation on the next call. Use the same guard that is used in
lh_delete().

Issue found when investigating haproxy crashes reported by wizard-it on
GitHub.

ok tb@

6 months agoProvide initial regress for lhash.
jsing [Mon, 6 May 2024 14:31:25 +0000 (14:31 +0000)]
Provide initial regress for lhash.

For now, this is very limited and only tests calling lh_doall_arg()
multiple times on an empty linked hash. This process currently triggers
a SIGSEGV, which will be soon fixed.

6 months agoAdd support for Tx checksum offloading to dwqe(4).
stsp [Mon, 6 May 2024 09:54:38 +0000 (09:54 +0000)]
Add support for Tx checksum offloading to dwqe(4).

Also improve our macros documenting Tx descriptor formats.

Tested:
RK3568: kettenis, jmatthew
Elkhart Lake: stsp

ok kettenis@, jmatthew@

6 months agosndiod: Ignore server.device settings to a non-working device
ratchov [Mon, 6 May 2024 05:37:26 +0000 (05:37 +0000)]
sndiod: Ignore server.device settings to a non-working device

6 months agomatch new Realtek RTL8192EU id; from Kyle Markley
jsg [Mon, 6 May 2024 05:02:25 +0000 (05:02 +0000)]
match new Realtek RTL8192EU id; from Kyle Markley

6 months agoregen
jsg [Mon, 6 May 2024 05:00:52 +0000 (05:00 +0000)]
regen

6 months agoanother Realtek RTL8192EU id; from Kyle Markley
jsg [Mon, 6 May 2024 05:00:19 +0000 (05:00 +0000)]
another Realtek RTL8192EU id; from Kyle Markley

6 months agoexpose hardware counters as kstats.
dlg [Mon, 6 May 2024 04:25:52 +0000 (04:25 +0000)]
expose hardware counters as kstats.

igc(4) counters are read to clear like em(4) counters, so this code
looks very much like em(4) but with less quirks so far.

ok bluhm@ bket@

6 months agodrm/amdgpu: Fix leak when GPU memory allocation fails
jsg [Mon, 6 May 2024 02:12:36 +0000 (02:12 +0000)]
drm/amdgpu: Fix leak when GPU memory allocation fails

From Mukul Joshi
d7d7284c2a7c67ea5f07eb28d8c2ea0f3ee4b86b in linux-6.6.y/6.6.30
25e9227c6afd200bed6774c866980b8e36d033af in mainline linux

6 months agodrm/amdgpu: Assign correct bits for SDMA HDP flush
jsg [Mon, 6 May 2024 02:09:51 +0000 (02:09 +0000)]
drm/amdgpu: Assign correct bits for SDMA HDP flush

From Lijo Lazar
65356a1cc49657b2fc833489d226053fa601334a in linux-6.6.y/6.6.30
aebd3eb9d3ae017e6260043f6bcace2f5ef60694 in mainline linux

6 months agodrm/amdgpu/sdma5.2: use legacy HDP flush for SDMA2/3
jsg [Mon, 6 May 2024 02:07:15 +0000 (02:07 +0000)]
drm/amdgpu/sdma5.2: use legacy HDP flush for SDMA2/3

From Alex Deucher
b33d7aaa2d3d1bef20fd5d9ab0d1776376fc98b6 in linux-6.6.y/6.6.30
9792b7cc18aaa0c2acae6af5d0acf249bcb1ab0d in mainline linux

6 months agodrm/ttm: stop pooling cached NUMA pages v2
jsg [Mon, 6 May 2024 02:05:30 +0000 (02:05 +0000)]
drm/ttm: stop pooling cached NUMA pages v2

From Christian Koenig
ab65b0cf0324b6dae89cfc9c7d9625e591b24e43 in linux-6.6.y/6.6.30
b6976f323a8687cc0d55bc92c2086fd934324ed5 in mainline linux

6 months agomm, treewide: introduce NR_PAGE_ORDERS
jsg [Mon, 6 May 2024 01:59:03 +0000 (01:59 +0000)]
mm, treewide: introduce NR_PAGE_ORDERS

From Kirill A. Shutemov
ded1ffea52132e58eaaa7d4ea39477f911796a40 in linux-6.6.y/6.6.30
fd37721803c6e73619108f76ad2e12a9aa5fafaf in mainline linux

6 months agodrm/amdgpu: fix visible VRAM handling during faults
jsg [Mon, 6 May 2024 01:53:26 +0000 (01:53 +0000)]
drm/amdgpu: fix visible VRAM handling during faults

From Christian Koenig
4c5eaf0cad27a66c4788e0603b9f7a68df83a947 in linux-6.6.y/6.6.30
a6ff969fe9cbf369e3cd0ac54261fec1122682ec in mainline linux

6 months agodrm/amdgpu: add shared fdinfo stats
jsg [Mon, 6 May 2024 01:50:15 +0000 (01:50 +0000)]
drm/amdgpu: add shared fdinfo stats

From Alex Deucher
f85a55fb87c2ee58e957b9c828aa70306a759d8d in linux-6.6.y/6.6.30
ba1a58d5b907bdf1814f8f57434aebc86233430f in mainline linux

6 months agodrm: add drm_gem_object_is_shared_for_memory_stats() helper
jsg [Mon, 6 May 2024 01:46:20 +0000 (01:46 +0000)]
drm: add drm_gem_object_is_shared_for_memory_stats() helper

From Alex Deucher
a4ae24cd04a8e2c1f61586a2e129b1cbd44a78db in linux-6.6.y/6.6.30
b31f5eba32ae8cc28e7cfa5a55ec8670d8c718e2 in mainline linux

6 months agoadd NR_PAGE_ORDERS for 6.6.30 drm
jsg [Mon, 6 May 2024 01:42:05 +0000 (01:42 +0000)]
add NR_PAGE_ORDERS for 6.6.30 drm

6 months agostart documenting ioctls
kn [Sun, 5 May 2024 19:13:13 +0000 (19:13 +0000)]
start documenting ioctls

6 months agoremove unused 0-sized softdep files
jsg [Sun, 5 May 2024 13:33:04 +0000 (13:33 +0000)]
remove unused 0-sized softdep files
ok tb@

6 months agoDocument "ls rogue" exit status.
ajacoutot [Sun, 5 May 2024 08:16:45 +0000 (08:16 +0000)]
Document "ls rogue" exit status.

6 months agoStop building with and documenting FFS_SOFTUPDATES.
jsg [Sun, 5 May 2024 07:26:58 +0000 (07:26 +0000)]
Stop building with and documenting FFS_SOFTUPDATES.
Code that used it was removed in February.

6 months agoremove unneeded includes
jsg [Sun, 5 May 2024 06:14:37 +0000 (06:14 +0000)]
remove unneeded includes

6 months agoadd upstream change to fix the build
jsg [Sun, 5 May 2024 02:55:34 +0000 (02:55 +0000)]
add upstream change to fix the build
ok tb@ deraadt@

6 months agoUpdate awk to the May 4, 2024 version.
millert [Sat, 4 May 2024 22:59:21 +0000 (22:59 +0000)]
Update awk to the May 4, 2024 version.
Fixes a use-after-free bug with ARGV for "delete ARGV".

6 months agodma_free(9) just once in error case
kn [Sat, 4 May 2024 16:40:38 +0000 (16:40 +0000)]
dma_free(9) just once in error case

buf is free'd and NULL, don't do it again to avoid a panic.

OK krw

6 months agoAdd TSO capabilites.
mbuhl [Sat, 4 May 2024 13:35:26 +0000 (13:35 +0000)]
Add TSO capabilites.
Comments, suggestions and testing from bket@, jan@, and bluhm@

6 months agoremove prototypes for removed functions
jsg [Sat, 4 May 2024 11:25:24 +0000 (11:25 +0000)]
remove prototypes for removed functions