random: always wake up entropy writers after extraction

commit 489c7fc44b5740d377e8cfdbf0851036e493af00 upstream.

Now that POOL_BITS == POOL_MIN_BITS, we must unconditionally wake up
entropy writers after every extraction. Therefore there's no point of
write_wakeup_threshold, so we can move it to the dustbin of unused
compatibility sysctls. While we're at it, we can fix a small comparison
where we were waking up after <= min rather than < min.

Cc: Theodore Ts'o <tytso@mit.edu>
Suggested-by: Eric Biggers <ebiggers@kernel.org>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: use linear min-entropy accumulation crediting

commit c570449094844527577c5c914140222cb1893e3f upstream.

30e37ec516ae ("random: account for entropy loss due to overwrites")
assumed that adding new entropy to the LFSR pool probabilistically
cancelled out old entropy there, so entropy was credited asymptotically,
approximating Shannon entropy of independent sources (rather than a
stronger min-entropy notion) using 1/8th fractional bits and replacing
a constant 2-2/√𝑒 term (~0.786938) with 3/4 (0.75) to slightly
underestimate it. This wasn't superb, but it was perhaps better than
nothing, so that's what was done. Which entropy specifically was being
cancelled out and how much precisely each time is hard to tell, though
as I showed with the attack code in my previous commit, a motivated
adversary with sufficient information can actually cancel out
everything.

Since we're no longer using an LFSR for entropy accumulation, this
probabilistic cancellation is no longer relevant. Rather, we're now
using a computational hash function as the accumulator and we've
switched to working in the random oracle model, from which we can now
revisit the question of min-entropy accumulation, which is done in
detail in <https://eprint.iacr.org/2019/198>.

Consider a long input bit string that is built by concatenating various
smaller independent input bit strings. Each one of these inputs has a
designated min-entropy, which is what we're passing to
credit_entropy_bits(h). When we pass the concatenation of these to a
random oracle, it means that an adversary trying to receive back the
same reply as us would need to become certain about each part of the
concatenated bit string we passed in, which means becoming certain about
all of those h values. That means we can estimate the accumulation by
simply adding up the h values in calls to credit_entropy_bits(h);
there's no probabilistic cancellation at play like there was said to be
for the LFSR. Incidentally, this is also what other entropy accumulators
based on computational hash functions do as well.

So this commit replaces credit_entropy_bits(h) with essentially `total =
min(POOL_BITS, total + h)`, done with a cmpxchg loop as before.

What if we're wrong and the above is nonsense? It's not, but let's
assume we don't want the actual _behavior_ of the code to change much.
Currently that behavior is not extracting from the input pool until it
has 128 bits of entropy in it. With the old algorithm, we'd hit that
magic 128 number after roughly 256 calls to credit_entropy_bits(1). So,
we can retain more or less the old behavior by waiting to extract from
the input pool until it hits 256 bits of entropy using the new code. For
people concerned about this change, it means that there's not that much
practical behavioral change. And for folks actually trying to model
the behavior rigorously, it means that we have an even higher margin
against attacks.

Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Jean-Philippe Aumasson <jeanphilippe.aumasson@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: simplify entropy debiting

commit 9c07f57869e90140080cfc282cc628d123e27704 upstream.

Our pool is 256 bits, and we only ever use all of it or don't use it at
all, which is decided by whether or not it has at least 128 bits in it.
So we can drastically simplify the accounting and cmpxchg loop to do
exactly this.  While we're at it, we move the minimum bit size into a
constant so it can be shared between the two places where it matters.

The reason we want any of this is for the case in which an attacker has
compromised the current state, and then bruteforces small amounts of
entropy added to it. By demanding a particular minimum amount of entropy
be present before reseeding, we make that bruteforcing difficult.

Note that this rationale no longer includes anything about /dev/random
blocking at the right moment, since /dev/random no longer blocks (except
for at ~boot), but rather uses the crng. In a former life, /dev/random
was different and therefore required a more nuanced account(), but this
is no longer.

Behaviorally, nothing changes here. This is just a simplification of
the code.

Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: use computational hash for entropy extraction

commit 6e8ec2552c7d13991148e551e3325a624d73fac6 upstream.

The current 4096-bit LFSR used for entropy collection had a few
desirable attributes for the context in which it was created. For
example, the state was huge, which meant that /dev/random would be able
to output quite a bit of accumulated entropy before blocking. It was
also, in its time, quite fast at accumulating entropy byte-by-byte,
which matters given the varying contexts in which mix_pool_bytes() is
called. And its diffusion was relatively high, which meant that changes
would ripple across several words of state rather quickly.

However, it also suffers from a few security vulnerabilities. In
particular, inputs learned by an attacker can be undone, but moreover,
if the state of the pool leaks, its contents can be controlled and
entirely zeroed out. I've demonstrated this attack with this SMT2
script, <https://xn--4db.cc/5o9xO8pb>, which Boolector/CaDiCal solves in
a matter of seconds on a single core of my laptop, resulting in little
proof of concept C demonstrators such as <https://xn--4db.cc/jCkvvIaH/c>.

For basically all recent formal models of RNGs, these attacks represent
a significant cryptographic flaw. But how does this manifest
practically? If an attacker has access to the system to such a degree
that he can learn the internal state of the RNG, arguably there are
other lower hanging vulnerabilities -- side-channel, infoleak, or
otherwise -- that might have higher priority. On the other hand, seed
files are frequently used on systems that have a hard time generating
much entropy on their own, and these seed files, being files, often leak
or are duplicated and distributed accidentally, or are even seeded over
the Internet intentionally, where their contents might be recorded or
tampered with. Seen this way, an otherwise quasi-implausible
vulnerability is a bit more practical than initially thought.

Another aspect of the current mix_pool_bytes() function is that, while
its performance was arguably competitive for the time in which it was
created, it's no longer considered so. This patch improves performance
significantly: on a high-end CPU, an i7-11850H, it improves performance
of mix_pool_bytes() by 225%, and on a low-end CPU, a Cortex-A7, it
improves performance by 103%.

This commit replaces the LFSR of mix_pool_bytes() with a straight-
forward cryptographic hash function, BLAKE2s, which is already in use
for pool extraction. Universal hashing with a secret seed was considered
too, something along the lines of <https://eprint.iacr.org/2013/338>,
but the requirement for a secret seed makes for a chicken & egg problem.
Instead we go with a formally proven scheme using a computational hash
function, described in sections 5.1, 6.4, and B.1.8 of
<https://eprint.iacr.org/2019/198>.

BLAKE2s outputs 256 bits, which should give us an appropriate amount of
min-entropy accumulation, and a wide enough margin of collision
resistance against active attacks. mix_pool_bytes() becomes a simple
call to blake2s_update(), for accumulation, while the extraction step
becomes a blake2s_final() to generate a seed, with which we can then do
a HKDF-like or BLAKE2X-like expansion, the first part of which we fold
back as an init key for subsequent blake2s_update()s, and the rest we
produce to the caller. This then is provided to our CRNG like usual. In
that expansion step, we make opportunistic use of 32 bytes of RDRAND
output, just as before. We also always reseed the crng with 32 bytes,
unconditionally, or not at all, rather than sometimes with 16 as before,
as we don't win anything by limiting beyond the 16 byte threshold.

Going for a hash function as an entropy collector is a conservative,
proven approach. The result of all this is a much simpler and much less
bespoke construction than what's there now, which not only plugs a
vulnerability but also improves performance considerably.

Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Jean-Philippe Aumasson <jeanphilippe.aumasson@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: only call crng_finalize_init() for primary_crng

commit 9d5505f1eebeca778074a0260ed077fd85f8792c upstream.

crng_finalize_init() returns instantly if it is called for another pool
than primary_crng. The test whether crng_finalize_init() is still required
can be moved to the relevant caller in crng_reseed(), and
crng_need_final_init can be reset to false if crng_finalize_init() is
called with workqueues ready. Then, no previous callsite will call
crng_finalize_init() unless it is needed, and we can get rid of the
superfluous function parameter.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: access primary_pool directly rather than through pointer

commit ebf7606388732ecf2821ca21087e9446cb4a5b57 upstream.

Both crng_initialize_primary() and crng_init_try_arch_early() are
only called for the primary_pool. Accessing it directly instead of
through a function parameter simplifies the code.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: continually use hwgenerator randomness

commit c321e907aa4803d562d6e70ebed9444ad082f953 upstream.

The rngd kernel thread may sleep indefinitely if the entropy count is
kept above random_write_wakeup_bits by other entropy sources. To make
best use of multiple sources of randomness, mix entropy from hardware
RNGs into the pool at least once within CRNG_RESEED_INTERVAL.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: simplify arithmetic function flow in account()

commit a254a0e4093fce8c832414a83940736067eed515 upstream.

Now that have_bytes is never modified, we can simplify this function.
First, we move the check for negative entropy_count to be first. That
ensures that subsequent reads of this will be non-negative. Then,
have_bytes and ibytes can be folded into their one use site in the
min_t() function.

Suggested-by: Dominik Brodowski <linux@dominikbrodowski.net>
Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: selectively clang-format where it makes sense

commit 248045b8dea5a32ddc0aa44193d6bc70c4b9cd8e upstream.

This is an old driver that has seen a lot of different eras of kernel
coding style. In an effort to make it easier to code for, unify the
coding style around the current norm, by accepting some of -- but
certainly not all of -- the suggestions from clang-format. This should
remove ambiguity in coding style, especially with regards to spacing,
when code is being changed or amended. Consequently it also makes code
review easier on the eyes, following one uniform style rather than
several.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: access input_pool_data directly rather than through pointer

commit 6c0eace6e1499712583b6ee62d95161e8b3449f5 upstream.

This gets rid of another abstraction we no longer need. It would be nice
if we could instead make pool an array rather than a pointer, but the
latent entropy plugin won't be able to do its magic in that case. So
instead we put all accesses to the input pool's actual data through the
input_pool_data array directly.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: cleanup fractional entropy shift constants

commit 18263c4e8e62f7329f38f5eadc568751242ca89c upstream.

The entropy estimator is calculated in terms of 1/8 bits, which means
there are various constants where things are shifted by 3. Move these
into our pool info enum with the other relevant constants. While we're
at it, move an English assertion about sizes into a proper BUILD_BUG_ON
so that the compiler can ensure this invariant.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: prepend remaining pool constants with POOL_

commit b3d51c1f542113342ddfbf6007e38a684b9dbec9 upstream.

The other pool constants are prepended with POOL_, but not these last
ones. Rename them. This will then let us move them into the enum in the
following commit.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: de-duplicate INPUT_POOL constants

commit 5b87adf30f1464477169a1d653e9baf8c012bbfe upstream.

We already had the POOL_* constants, so deduplicate the older INPUT_POOL
ones. As well, fold EXTRACT_SIZE into the poolinfo enum, since it's
related.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: remove unused OUTPUT_POOL constants

commit 0f63702718c91d89c922081ac1e6baeddc2d8b1a upstream.

We no longer have an output pool. Rather, we have just a wakeup bits
threshold for /dev/random reads, presumably so that processes don't
hang. This value, random_write_wakeup_bits, is configurable anyway. So
all the no longer usefully named OUTPUT_POOL constants were doing was
setting a reasonable default for random_write_wakeup_bits. This commit
gets rid of the constants and just puts it all in the default value of
random_write_wakeup_bits.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: rather than entropy_store abstraction, use global

commit 90ed1e67e896cc8040a523f8428fc02f9b164394 upstream.

Originally, the RNG used several pools, so having things abstracted out
over a generic entropy_store object made sense. These days, there's only
one input pool, and then an uneven mix of usage via the abstraction and
usage via &input_pool. Rather than this uneasy mixture, just get rid of
the abstraction entirely and have things always use the global. This
simplifies the code and makes reading it a bit easier.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: remove unused extract_entropy() reserved argument

commit 8b2d953b91e7f60200c24067ab17b77cc7bfd0d4 upstream.

This argument is always set to zero, as a result of us not caring about
keeping a certain amount reserved in the pool these days. So just remove
it and cleanup the function signatures.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: remove incomplete last_data logic

commit a4bfa9b31802c14ff5847123c12b98d5e36b3985 upstream.

There were a few things added under the "if (fips_enabled)" banner,
which never really got completed, and the FIPS people anyway are
choosing a different direction. Rather than keep around this halfbaked
code, get rid of it so that we can focus on a single design of the RNG
rather than two designs.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: cleanup integer types

commit d38bb0853589c939573ea50e9cb64f733e0e273d upstream.

Rather than using the userspace type, __uXX, switch to using uXX. And
rather than using variously chosen `char *` or `unsigned char *`, use
`u8 *` uniformly for things that aren't strings, in the case where we
are doing byte-by-byte traversal.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: cleanup poolinfo abstraction

commit 91ec0fe138f107232cb36bc6112211db37cb5306 upstream.

Now that we're only using one polynomial, we can cleanup its
representation into constants, instead of passing around pointers
dynamically to select different polynomials. This improves the codegen
and makes the code a bit more straightforward.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: fix typo in comments

commit c0a8a61e7abbf66729687ee63659ee25983fbb1e upstream.

s/or/for

Signed-off-by: Schspa Shi <schspa@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: don't reset crng_init_cnt on urandom_read()

commit 6c8e11e08a5b74bb8a5cdd5cbc1e5143df0fba72 upstream.

At the moment, urandom_read() (used for /dev/urandom) resets crng_init_cnt
to zero when it is called at crng_init<2. This is inconsistent: We do it
for /dev/urandom reads, but not for the equivalent
getrandom(GRND_INSECURE).

(And worse, as Jason pointed out, we're only doing this as long as
maxwarn>0.)

crng_init_cnt is only read in crng_fast_load(); it is relevant at
crng_init==0 for determining when to switch to crng_init==1 (and where in
the RNG state array to write).

As far as I understand:

 - crng_init==0 means "we have nothing, we might just be returning the same
   exact numbers on every boot on every machine, we don't even have
   non-cryptographic randomness; we should shove every bit of entropy we
   can get into the RNG immediately"
 - crng_init==1 means "well we have something, it might not be
   cryptographic, but at least we're not gonna return the same data every
   time or whatever, it's probably good enough for TCP and ASLR and stuff;
   we now have time to build up actual cryptographic entropy in the input
   pool"
 - crng_init==2 means "this is supposed to be cryptographically secure now,
   but we'll keep adding more entropy just to be sure".

The current code means that if someone is pulling data from /dev/urandom
fast enough at crng_init==0, we'll keep resetting crng_init_cnt, and we'll
never make forward progress to crng_init==1. It seems to be intended to
prevent an attacker from bruteforcing the contents of small individual RNG
inputs on the way from crng_init==0 to crng_init==1, but that's misguided;
crng_init==1 isn't supposed to provide proper cryptographic security
anyway, RNG users who care about getting secure RNG output have to wait
until crng_init==2.

This code was inconsistent, and it probably made things worse - just get
rid of it.

Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: avoid superfluous call to RDRAND in CRNG extraction

commit 2ee25b6968b1b3c66ffa408de23d023c1bce81cf upstream.

RDRAND is not fast. RDRAND is actually quite slow. We've known this for
a while, which is why functions like get_random_u{32,64} were converted
to use batching of our ChaCha-based CRNG instead.

Yet CRNG extraction still includes a call to RDRAND, in the hot path of
every call to get_random_bytes(), /dev/urandom, and getrandom(2).

This call to RDRAND here seems quite superfluous. CRNG is already
extracting things based on a 256-bit key, based on good entropy, which
is then reseeded periodically, updated, backtrack-mutated, and so
forth. The CRNG extraction construction is something that we're already
relying on to be secure and solid. If it's not, that's a serious
problem, and it's unlikely that mixing in a measly 32 bits from RDRAND
is going to alleviate things.

And in the case where the CRNG doesn't have enough entropy yet, we're
already initializing the ChaCha key row with RDRAND in
crng_init_try_arch_early().

Removing the call to RDRAND improves performance on an i7-11850H by
370%. In other words, the vast majority of the work done by
extract_crng() prior to this commit was devoted to fetching 32 bits of
RDRAND.

Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: early initialization of ChaCha constants

commit 96562f286884e2db89c74215b199a1084b5fb7f7 upstream.

Previously, the ChaCha constants for the primary pool were only
initialized in crng_initialize_primary(), called by rand_initialize().
However, some randomness is actually extracted from the primary pool
beforehand, e.g. by kmem_cache_create(). Therefore, statically
initialize the ChaCha constants for the primary pool.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: <linux-crypto@vger.kernel.org>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: use IS_ENABLED(CONFIG_NUMA) instead of ifdefs

commit 7b87324112df2e1f9b395217361626362dcfb9fb upstream.

Rather than an awkward combination of ifdefs and __maybe_unused, we can
ensure more source gets parsed, regardless of the configuration, by
using IS_ENABLED for the CONFIG_NUMA conditional code. This makes things
cleaner and easier to follow.

I've confirmed that on !CONFIG_NUMA, we don't wind up with excess code
by accident; the generated object file is the same.

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: harmonize "crng init done" messages

commit 161212c7fd1d9069b232785c75492e50941e2ea8 upstream.

We print out "crng init done" for !TRUST_CPU, so we should also print
out the same for TRUST_CPU.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: mix bootloader randomness into pool

commit 57826feeedb63b091f807ba8325d736775d39afd upstream.

If we're trusting bootloader randomness, crng_fast_load() is called by
add_hwgenerator_randomness(), which sets us to crng_init==1. However,
usually it is only called once for an initial 64-byte push, so bootloader
entropy will not mix any bytes into the input pool. So it's conceivable
that crng_init==1 when crng_initialize_primary() is called later, but
then the input pool is empty. When that happens, the crng state key will
be overwritten with extracted output from the empty input pool. That's
bad.

In contrast, if we're not trusting bootloader randomness, we call
crng_slow_load() *and* we call mix_pool_bytes(), so that later
crng_initialize_primary() isn't drawing on nothing.

In order to prevent crng_initialize_primary() from extracting an empty
pool, have the trusted bootloader case mirror that of the untrusted
bootloader case, mixing the input into the pool.

[linux@dominikbrodowski.net: rewrite commit message]
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: do not re-init if crng_reseed completes before primary init

commit 9c3ddde3f811aabbb83778a2a615bf141b4909ef upstream.

If the bootloader supplies sufficient material and crng_reseed() is called
very early on, but not too early that wqs aren't available yet, then we
might transition to crng_init==2 before rand_initialize()'s call to
crng_initialize_primary() made. Then, when crng_initialize_primary() is
called, if we're trusting the CPU's RDRAND instructions, we'll
needlessly reinitialize the RNG and emit a message about it. This is
mostly harmless, as numa_crng_init() will allocate and then free what it
just allocated, and excessive calls to invalidate_batched_entropy()
aren't so harmful. But it is funky and the extra message is confusing,
so avoid the re-initialization all together by checking for crng_init <
2 in crng_initialize_primary(), just as we already do in crng_reseed().

Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: do not sign extend bytes for rotation when mixing

commit 0d9488ffbf2faddebc6bac055bfa6c93b94056a3 upstream.

By using `char` instead of `unsigned char`, certain platforms will sign
extend the byte when `w = rol32(*bytes++, input_rotate)` is called,
meaning that bit 7 is overrepresented when mixing. This isn't a real
problem (unless the mixer itself is already broken) since it's still
invertible, but it's not quite correct either. Fix this by using an
explicit unsigned type.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: use BLAKE2s instead of SHA1 in extraction

commit 9f9eff85a008b095eafc5f4ecbaf5aca689271c1 upstream.

This commit addresses one of the lower hanging fruits of the RNG: its
usage of SHA1.

BLAKE2s is generally faster, and certainly more secure, than SHA1, which
has [1] been [2] really [3] very [4] broken [5]. Additionally, the
current construction in the RNG doesn't use the full SHA1 function, as
specified, and allows overwriting the IV with RDRAND output in an
undocumented way, even in the case when RDRAND isn't set to "trusted",
which means potential malicious IV choices. And its short length means
that keeping only half of it secret when feeding back into the mixer
gives us only 2^80 bits of forward secrecy. In other words, not only is
the choice of hash function dated, but the use of it isn't really great
either.

This commit aims to fix both of these issues while also keeping the
general structure and semantics as close to the original as possible.
Specifically:

   a) Rather than overwriting the hash IV with RDRAND, we put it into
      BLAKE2's documented "salt" and "personal" fields, which were
      specifically created for this type of usage.
   b) Since this function feeds the full hash result back into the
      entropy collector, we only return from it half the length of the
      hash, just as it was done before. This increases the
      construction's forward secrecy from 2^80 to a much more
      comfortable 2^128.
   c) Rather than using the raw "sha1_transform" function alone, we
      instead use the full proper BLAKE2s function, with finalization.

This also has the advantage of supplying 16 bytes at a time rather than
SHA1's 10 bytes, which, in addition to having a faster compression
function to begin with, means faster extraction in general. On an Intel
i7-11850H, this commit makes initial seeding around 131% faster.

BLAKE2s itself has the nice property of internally being based on the
ChaCha permutation, which the RNG is already using for expansion, so
there shouldn't be any issue with newness, funkiness, or surprising CPU
behavior, since it's based on something already in use.

[1] https://eprint.iacr.org/2005/010.pdf
[2] https://www.iacr.org/archive/crypto2005/36210017/36210017.pdf
[3] https://eprint.iacr.org/2015/967.pdf
[4] https://shattered.io/static/shattered.pdf
[5] https://www.usenix.org/system/files/sec20-leurent.pdf

Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Jean-Philippe Aumasson <jeanphilippe.aumasson@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: remove unused irq_flags argument from add_interrupt_randomness()

commit 703f7066f40599c290babdb79dd61319264987e9 upstream.

Since commit
   ee3e00e9e7101 ("random: use registers from interrupted code for CPU's w/o a cycle counter")

the irq_flags argument is no longer used.

Remove unused irq_flags.

Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dexuan Cui <decui@microsoft.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: linux-hyperv@vger.kernel.org
Cc: x86@kernel.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Wei Liu <wei.liu@kernel.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

random: document add_hwgenerator_randomness() with other input functions

commit 2b6c6e3d9ce3aa0e547ac25d60e06fe035cd9f79 upstream.

The section at the top of random.c which documents the input functions
available does not document add_hwgenerator_randomness() which might lead
a reader to overlook it. Add a brief note about it.

Signed-off-by: Mark Brown <broonie@kernel.org>
[Jason: reorganize position of function in doc comment and also document
 add_bootloader_randomness() while we're at it.]
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

lib/crypto: blake2s: avoid indirect calls to compression function for Clang CFI

commit d2a02e3c8bb6b347818518edff5a4b40ff52d6d8 upstream.

blake2s_compress_generic is weakly aliased by blake2s_compress. The
current harness for function selection uses a function pointer, which is
ordinarily inlined and resolved at compile time. But when Clang's CFI is
enabled, CFI still triggers when making an indirect call via a weak
symbol. This seems like a bug in Clang's CFI, as though it's bucketing
weak symbols and strong symbols differently. It also only seems to
trigger when "full LTO" mode is used, rather than "thin LTO".

[    0.000000][    T0] Kernel panic - not syncing: CFI failure (target: blake2s_compress_generic+0x0/0x1444)
[    0.000000][    T0] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.16.0-mainline-06981-g076c855b846e #1
[    0.000000][    T0] Hardware name: MT6873 (DT)
[    0.000000][    T0] Call trace:
[    0.000000][    T0]  dump_backtrace+0xfc/0x1dc
[    0.000000][    T0]  dump_stack_lvl+0xa8/0x11c
[    0.000000][    T0]  panic+0x194/0x464
[    0.000000][    T0]  __cfi_check_fail+0x54/0x58
[    0.000000][    T0]  __cfi_slowpath_diag+0x354/0x4b0
[    0.000000][    T0]  blake2s_update+0x14c/0x178
[    0.000000][    T0]  _extract_entropy+0xf4/0x29c
[    0.000000][    T0]  crng_initialize_primary+0x24/0x94
[    0.000000][    T0]  rand_initialize+0x2c/0x6c
[    0.000000][    T0]  start_kernel+0x2f8/0x65c
[    0.000000][    T0]  __primary_switched+0xc4/0x7be4
[    0.000000][    T0] Rebooting in 5 seconds..

Nonetheless, the function pointer method isn't so terrific anyway, so
this patch replaces it with a simple boolean, which also gets inlined
away. This successfully works around the Clang bug.

In general, I'm not too keen on all of the indirection involved here; it
clearly does more harm than good. Hopefully the whole thing can get
cleaned up down the road when lib/crypto is overhauled more
comprehensively. But for now, we go with a simple bandaid.

Fixes: 6048fdcc5f26 ("lib/crypto: blake2s: include as built-in")
Link: https://github.com/ClangBuiltLinux/linux/issues/1567
Reported-by: Miles Chen <miles.chen@mediatek.com>
Tested-by: Miles Chen <miles.chen@mediatek.com>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: John Stultz <john.stultz@linaro.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

lib/crypto: sha1: re-roll loops to reduce code size

commit 9a1536b093bb5bf60689021275fd24d513bb8db0 upstream.

With SHA-1 no longer being used for anything performance oriented, and
also soon to be phased out entirely, we can make up for the space added
by unrolled BLAKE2s by simply re-rolling SHA-1. Since SHA-1 is so much
more complex, re-rolling it more or less takes care of the code size
added by BLAKE2s. And eventually, hopefully we'll see SHA-1 removed
entirely from most small kernel builds.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

lib/crypto: blake2s: move hmac construction into wireguard

commit d8d83d8ab0a453e17e68b3a3bed1f940c34b8646 upstream.

Basically nobody should use blake2s in an HMAC construction; it already
has a keyed variant. But unfortunately for historical reasons, Noise,
used by WireGuard, uses HKDF quite strictly, which means we have to use
this. Because this really shouldn't be used by others, this commit moves
it into wireguard's noise.c locally, so that kernels that aren't using
WireGuard don't get this superfluous code baked in. On m68k systems,
this shaves off ~314 bytes.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

lib/crypto: blake2s: include as built-in

commit 6048fdcc5f269c7f31d774c295ce59081b36e6f9 upstream.

In preparation for using blake2s in the RNG, we change the way that it
is wired-in to the build system. Instead of using ifdefs to select the
right symbol, we use weak symbols. And because ARM doesn't need the
generic implementation, we make the generic one default only if an arch
library doesn't need it already, and then have arch libraries that do
need it opt-in. So that the arch libraries can remain tristate rather
than bool, we then split the shash part from the glue code.

Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Masahiro Yamada <masahiroy@kernel.org>
Cc: linux-kbuild@vger.kernel.org
Cc: linux-crypto@vger.kernel.org
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

MAINTAINERS: add git tree for random.c

commit 9bafaa9375cbf892033f188d8cb624ae328754b5 upstream.

This is handy not just for humans, but also so that the 0-day bot can
automatically test posted mailing list patches against the right tree.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

MAINTAINERS: co-maintain random.c

commit 58e1100fdc5990b0cc0d4beaf2562a92e621ac7d upstream.

random.c is a bit understaffed, and folks want more prompt reviews. I've
got the crypto background and the interest to do these reviews, and have
authored parts of the file already.

Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ACPI: sysfs: Fix BERT error region memory mapping

commit 1bbc21785b7336619fb6a67f1fff5afdaf229acc upstream.

Currently the sysfs interface maps the BERT error region as "memory"
(through acpi_os_map_memory()) in order to copy the error records into
memory buffers through memory operations (eg memory_read_from_buffer()).

The OS system cannot detect whether the BERT error region is part of
system RAM or it is "device memory" (eg BMC memory) and therefore it
cannot detect which memory attributes the bus to memory support (and
corresponding kernel mapping, unless firmware provides the required
information).

The acpi_os_map_memory() arch backend implementation determines the
mapping attributes. On arm64, if the BERT error region is not present in
the EFI memory map, the error region is mapped as device-nGnRnE; this
triggers alignment faults since memcpy unaligned accesses are not
allowed in device-nGnRnE regions.

The ACPI sysfs code cannot therefore map by default the BERT error
region with memory semantics but should use a safer default.

Change the sysfs code to map the BERT error region as MMIO (through
acpi_os_map_iomem()) and use the memcpy_fromio() interface to read the
error region into the kernel buffer.

Link: https://lore.kernel.org/linux-arm-kernel/31ffe8fc-f5ee-2858-26c5-0fd8bdd68702@arm.com
Link: https://lore.kernel.org/linux-acpi/CAJZ5v0g+OVbhuUUDrLUCfX_mVqY_e8ubgLTU98=jfjTeb4t+Pw@mail.gmail.com
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Tested-by: Veronika Kabatova <vkabatov@redhat.com>
Tested-by: Aristeu Rozanski <aris@redhat.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: dann frazier <dann.frazier@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ice: fix crash at allocation failure

Fix a crash in the zero-copy driver that occurs when it fails to
allocate buffers from user-space. This crash can easily be triggered
by a malicious program that does not provide any buffers in the fill
ring for the kernel to use.

Note that this bug does not exist in upstream since the batched buffer
allocation interface got introduced in 5.16 and replaced this code.

Reported-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Tested-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86/mmu: fix NULL pointer dereference on guest INVPCID

commit 9f46c187e2e680ecd9de7983e4d081c3391acc76 upstream.

With shadow paging enabled, the INVPCID instruction results in a call
to kvm_mmu_invpcid_gva.  If INVPCID is executed with CR0.PG=0, the
invlpg callback is not set and the result is a NULL pointer dereference.
Fix it trivially by checking for mmu->invlpg before every call.

There are other possibilities:

- check for CR0.PG, because KVM (like all Intel processors after P5)
  flushes guest TLB on CR0.PG changes so that INVPCID/INVLPG are a
  nop with paging disabled

- check for EFER.LMA, because KVM syncs and flushes when switching
  MMU contexts outside of 64-bit mode

All of these are tricky, go for the simple solution.  This is CVE-2022-1789.

Reported-by: Yongkang Jia <kangel@zju.edu.cn>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[fix conflict due to missing b9e5603c2a3accbadfec570ac501a54431a6bdba]
Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

HID: amd_sfh: Add support for sensor discovery

commit b5d7f43e97dabfa04a4be5ff027ce7da119332be upstream.

Sensor discovery status fails in case of broken sensors or
platform not supported. Hence disable driver on failure
of sensor discovery.

Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Basavaraj Natikar <Basavaraj.Natikar@amd.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: Mario Limonciello <Mario.Limonciello@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.15.43

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mptcp: Do TCP fallback on early DSS checksum failure

commit ae66fb2ba6c3dcaf8b9612b65aa949a1a4bed150 upstream.

RFC 8684 section 3.7 describes several opportunities for a MPTCP
connection to "fall back" to regular TCP early in the connection
process, before it has been confirmed that MPTCP options can be
successfully propagated on all SYN, SYN/ACK, and data packets. If a peer
acknowledges the first received data packet with a regular TCP header
(no MPTCP options), fallback is allowed.

If the recipient of that first data packet finds a MPTCP DSS checksum
error, this provides an opportunity to fail gracefully with a TCP
fallback rather than resetting the connection (as might happen if a
checksum failure were detected later).

This commit modifies the checksum failure code to attempt fallback on
the initial subflow of a MPTCP connection, only if it's a failure in the
first data mapping. In cases where the peer initiates the connection,
requests checksums, is the first to send data, and the peer is sending
incorrect checksums (see
https://github.com/multipath-tcp/mptcp_net-next/issues/275), this allows
the connection to proceed as TCP rather than reset.

Fixes: dd8bcd1768ff ("mptcp: validate the data checksum")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
[mathew.j.martineau: backport: Resolved bitfield conflict in protocol.h]
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.15.42

Link: https://lore.kernel.org/r/20220523165823.492309987@linuxfoundation.org
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Fox Chen <foxhlchen@gmail.com>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Khalid Masum <khalid.masum.92@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

afs: Fix afs_getattr() to refetch file status if callback break occurred

[ Upstream commit 2aeb8c86d49967552394d5e723f87454cb53f501 ]

If a callback break occurs (change notification), afs_getattr() needs to
issue an FS.FetchStatus RPC operation to update the status of the file
being examined by the stat-family of system calls.

Fix afs_getattr() to do this if AFS_VNODE_CB_PROMISED has been cleared
on a vnode by a callback break.  Skip this if AT_STATX_DONT_SYNC is set.

This can be tested by appending to a file on one AFS client and then
using "stat -L" to examine its length on a machine running kafs.  This
can also be watched through tracing on the kafs machine.  The callback
break is seen:

     kworker/1:1-46      [001] .....   978.910812: afs_cb_call: c=0000005f YFSCB.CallBack
     kworker/1:1-46      [001] ...1.   978.910829: afs_cb_break: 100058:23b4c:242d2c2 b=2 s=1 break-cb
     kworker/1:1-46      [001] .....   978.911062: afs_call_done:    c=0000005f ret=0 ab=0 [0000000082994ead]

And then the stat command generated no traffic if unpatched, but with
this change a call to fetch the status can be observed:

            stat-4471    [000] .....   986.744122: afs_make_fs_call: c=000000ab 100058:023b4c:242d2c2 YFS.FetchStatus
            stat-4471    [000] .....   986.745578: afs_call_done:    c=000000ab ret=0 ab=0 [0000000087fc8c84]

Fixes: 08e0e7c82eea ("[AF_RXRPC]: Make the in-kernel AFS filesystem use AF_RXRPC.")
Reported-by: Markus Suvanto <markus.suvanto@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Tested-by: Markus Suvanto <markus.suvanto@gmail.com>
Tested-by: kafs-testing+fedora34_64checkkafs-build-496@auristor.com
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216010
Link: https://lore.kernel.org/r/165308359800.162686.14122417881564420962.stgit@warthog.procyon.org.uk/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i2c: mt7621: fix missing clk_disable_unprepare() on error in mtk_i2c_probe()

[ Upstream commit a2537c98a8a3b57002e54a262d180b9490bc7190 ]

Fix the missing clk_disable_unprepare() before return
from mtk_i2c_probe() in the error handling case.

Fixes: d04913ec5f89 ("i2c: mt7621: Add MediaTek MT7621/7628/7688 I2C driver")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

lockdown: also lock down previous kgdb use

commit eadb2f47a3ced5c64b23b90fd2a3463f63726066 upstream.

KGDB and KDB allow read and write access to kernel memory, and thus
should be restricted during lockdown.  An attacker with access to a
serial port (for example, via a hypervisor console, which some cloud
vendors provide over the network) could trigger the debugger so it is
important that the debugger respect the lockdown mode when/if it is
triggered.

Fix this by integrating lockdown into kdb's existing permissions
mechanism.  Unfortunately kgdb does not have any permissions mechanism
(although it certainly could be added later) so, for now, kgdb is simply
and brutally disabled by immediately exiting the gdb stub without taking
any action.

For lockdowns established early in the boot (e.g. the normal case) then
this should be fine but on systems where kgdb has set breakpoints before
the lockdown is enacted than "bad things" will happen.

CVE: CVE-2022-21499
Co-developed-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mt76: mt7921e: fix possible probe failure after reboot

commit 602cc0c9618a819ab00ea3c9400742a0ca318380 upstream.

It doesn't guarantee the mt7921e gets started with ASPM L0 after each
machine reboot on every platform.

If mt7921e gets started with not ASPM L0, it would be possible that the
driver encounters time to time failure in mt7921_pci_probe, like a
weird chip identifier is read

[  215.514503] mt7921e 0000:05:00.0: ASIC revision: feed0000
[  216.604741] mt7921e: probe of 0000:05:00.0 failed with error -110

or failing to init hardware because the driver is not allowed to access the
register until the device is in ASPM L0 state. So, we call
__mt7921e_mcu_drv_pmctrl in early mt7921_pci_probe to force the device
to bring back to the L0 state for we can safely access registers in any
case.

In the patch, we move all functions from dma.c to pci.c and register mt76
bus operation earilier, that is the __mt7921e_mcu_drv_pmctrl depends on.

Fixes: bf3747ae2e25 ("mt76: mt7921: enable aspm by default")
Reported-by: Kai-Chuan Hsieh <kaichuan.hsieh@canonical.com>
Co-developed-by: Deren Wu <deren.wu@mediatek.com>
Signed-off-by: Deren Wu <deren.wu@mediatek.com>
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dt-bindings: pinctrl: aspeed-g6: remove FWQSPID group

commit a29c96a4053dc3c1d39353b61089882f81c6b23d upstream.

FWQSPID is not a group of FWSPID so remove it.

Fixes: 7488838f2315 ("dt-bindings: pinctrl: aspeed: Document AST2600 pinmux")
Signed-off-by: Jae Hyun Yoo <quic_jaehyoo@quicinc.com>
Acked-by: Rob Herring <robh@kernel.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Link: https://lore.kernel.org/r/20220329173932.2588289-4-quic_jaehyoo@quicinc.com
Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Input: ili210x - fix reset timing

commit e4920d42ce0e9c8aafb7f64b6d9d4ae02161e51e upstream.

According to Ilitek "231x & ILI251x Programming Guide" Version: 2.30
"2.1. Power Sequence", "T4 Chip Reset and discharge time" is minimum
10ms and "T2 Chip initial time" is maximum 150ms. Adjust the reset
timings such that T4 is 12ms and T2 is 160ms to fit those figures.

This prevents sporadic touch controller start up failures when some
systems with at least ILI251x controller boot, without this patch
the systems sometimes fail to communicate with the touch controller.

Fixes: 201f3c803544c ("Input: ili210x - add reset GPIO support")
Signed-off-by: Marek Vasut <marex@denx.de>
Link: https://lore.kernel.org/r/20220518204901.93534-1-marex@denx.de
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: Enable repeat tlbi workaround on KRYO4XX gold CPUs

[ Upstream commit 51f559d66527e238f9a5f82027bff499784d4eac ]

Add KRYO4XX gold/big cores to the list of CPUs that need the
repeat TLBI workaround. Apply this to the affected
KRYO4XX cores (rcpe to rfpe).

The variant and revision bits are implementation defined and are
different from the their Cortex CPU counterparts on which they are
based on, i.e., (r0p0 to r3p0) is equivalent to (rcpe to rfpe).

Signed-off-by: Shreyas K K <quic_shrekk@quicinc.com>
Reviewed-by: Sai Prakash Ranjan <quic_saipraka@quicinc.com>
Link: https://lore.kernel.org/r/20220512110134.12179-1-quic_shrekk@quicinc.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: atlantic: verify hw_head_ lies within TX buffer ring

[ Upstream commit 2120b7f4d128433ad8c5f503a9584deba0684901 ]

Bounds check hw_head index provided by NIC to verify it lies
within the TX buffer ring.

Reported-by: Aashay Shringarpure <aashay@google.com>
Reported-by: Yi Chou <yich@google.com>
Reported-by: Shervin Oloumi <enlightened@google.com>
Signed-off-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: atlantic: add check for MAX_SKB_FRAGS

[ Upstream commit 6aecbba12b5c90b26dc062af3b9de8c4b3a2f19f ]

Enforce that the CPU can not get stuck in an infinite loop.

Reported-by: Aashay Shringarpure <aashay@google.com>
Reported-by: Yi Chou <yich@google.com>
Reported-by: Shervin Oloumi <enlightened@google.com>
Signed-off-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: atlantic: reduce scope of is_rsc_complete

[ Upstream commit 79784d77ebbd3ec516b7a5ce555d979fb7946202 ]

Don't defer handling the err case outside the loop. That's pointless.

And since is_rsc_complete is only used inside this loop, declare
it inside the loop to reduce it's scope.

Signed-off-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: atlantic: fix "frag[0] not initialized"

[ Upstream commit 62e0ae0f4020250f961cf8d0103a4621be74e077 ]

In aq_ring_rx_clean(), if buff->is_eop is not set AND
buff->len < AQ_CFG_RX_HDR_SIZE, then hdr_len remains equal to
buff->len and skb_add_rx_frag(xxx, *0*, ...) is not called.

The loop following this code starts calling skb_add_rx_frag() starting
with i=1 and thus frag[0] is never initialized. Since i is initialized
to zero at the top of the primary loop, we can just reference and
post-increment i instead of hardcoding the 0 when calling
skb_add_rx_frag() the first time.

Reported-by: Aashay Shringarpure <aashay@google.com>
Reported-by: Yi Chou <yich@google.com>
Reported-by: Shervin Oloumi <enlightened@google.com>
Signed-off-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: stmmac: fix missing pci_disable_device() on error in stmmac_pci_probe()

[ Upstream commit 0807ce0b010418a191e0e4009803b2d74c3245d5 ]

Switch to using pcim_enable_device() to avoid missing pci_disable_device().

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Link: https://lore.kernel.org/r/20220510031316.1780409-1-yangyingliang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ethernet: tulip: fix missing pci_disable_device() on error in tulip_init_one()

[ Upstream commit 51ca86b4c9c7c75f5630fa0dbe5f8f0bd98e3c3e ]

Fix the missing pci_disable_device() before return
from tulip_init_one() in the error handling case.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Link: https://lore.kernel.org/r/20220506094250.3630615-1-yangyingliang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

nl80211: fix locking in nl80211_set_tx_bitrate_mask()

[ Upstream commit f971e1887fdb3ab500c9bebf4b98f62d49a20655 ]

This accesses the wdev's chandef etc., so cannot safely
be used without holding the lock.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20220506102136.06b7205419e6.I2a87c05fbd8bc5e565e84d190d4cfd2e92695a90@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: fix wrong network header length

[ Upstream commit cf3ab8d4a797960b4be20565abb3bcd227b18a68 ]

When clatd starts with ebpf offloaing, and NETIF_F_GRO_FRAGLIST is enable,
several skbs are gathered in skb_shinfo(skb)->frag_list. The first skb's
ipv6 header will be changed to ipv4 after bpf_skb_proto_6_to_4,
network_header\transport_header\mac_header have been updated as ipv4 acts,
but other skbs in frag_list didnot update anything, just ipv6 packets.

udp_queue_rcv_skb will call skb_segment_list to traverse other skbs in
frag_list and make sure right udp payload is delivered to user space.
Unfortunately, other skbs in frag_list who are still ipv6 packets are
updated like the first skb and will have wrong transport header length.

e.g.before bpf_skb_proto_6_to_4,the first skb and other skbs in frag_list
has the same network_header(24)& transport_header(64), after
bpf_skb_proto_6_to_4, ipv6 protocol has been changed to ipv4, the first
skb's network_header is 44,transport_header is 64, other skbs in frag_list
didnot change.After skb_segment_list, the other skbs in frag_list has
different network_header(24) and transport_header(44), so there will be 20
bytes different from original,that is difference between ipv6 header and
ipv4 header. Just change transport_header to be the same with original.

Actually, there are two solutions to fix it, one is traversing all skbs
and changing every skb header in bpf_skb_proto_6_to_4, the other is
modifying frag_list skb's header in skb_segment_list. Considering
efficiency, adopt the second one--- when the first skb and other skbs in
frag_list has different network_header length, restore them to make sure
right udp payload is delivered to user space.

Signed-off-by: Lina Wang <lina.wang@mediatek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

fbdev: Prevent possible use-after-free in fb_release()

[ Upstream commit 89bfd4017e58faaf70411555e7f508495114e90b ]

Most fbdev drivers have issues with the fb_info lifetime, because call to
framebuffer_release() from their driver's .remove callback, rather than
doing from fbops.fb_destroy callback.

Doing that will destroy the fb_info too early, while references to it may
still exist, leading to a use-after-free error.

To prevent this, check the fb_info reference counter when attempting to
kfree the data structure in framebuffer_release(). That will leak it but
at least will prevent the mentioned error.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20220505220413.365977-1-javierm@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

Revert "fbdev: Make fb_release() return -ENODEV if fbdev was unregistered"

[ Upstream commit 135332f34ba2662bc1e32b5c612e06a8cc41a053 ]

This reverts commit aafa025c76dcc7d1a8c8f0bdefcbe4eb480b2f6a. That commit
attempted to fix a NULL pointer dereference, caused by the struct fb_info
associated with a framebuffer device to not longer be valid when the file
descriptor was closed.

The issue was exposed by commit 27599aacbaef ("fbdev: Hot-unplug firmware
fb devices on forced removal"), which added a new path that goes through
the struct device removal instead of directly unregistering the fb.

Most fbdev drivers have issues with the fb_info lifetime, because call to
framebuffer_release() from their driver's .remove callback, rather than
doing from fbops.fb_destroy callback. This meant that due to this switch,
the fb_info was now destroyed too early, while references still existed,
while before it was simply leaked.

The patch we're reverting here reinstated that leak, hence "fixed" the
regression. But the proper solution is to fix the drivers to not release
the fb_info too soon.

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20220504115917.758787-1-javierm@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: add ping test with ping_group_range tuned

[ Upstream commit e71b7f1f44d3d88c677769c85ef0171caf9fc89f ]

The 'ping' utility is able to manage two kind of sockets (raw or icmp),
depending on the sysctl ping_group_range. By default, ping_group_range is
set to '1 0', which forces ping to use an ip raw socket.

Let's replay the ping tests by allowing 'ping' to use the ip icmp socket.
After the previous patch, ipv4 tests results are the same with both kinds
of socket. For ipv6, there are a lot a new failures (the previous patch
fixes only two cases).

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

nl80211: validate S1G channel width

[ Upstream commit 5d087aa759eb82b8208411913f6c2158bd85abc0 ]

Validate the S1G channel width input by user to ensure it matches
that of the requested channel

Signed-off-by: Kieran Frewen <kieran.frewen@morsemicro.com>
Signed-off-by: Bassem Dawood <bassem@morsemicro.com>
Link: https://lore.kernel.org/r/20220420041321.3788789-2-kieran.frewen@morsemicro.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mac80211: fix rx reordering with non explicit / psmp ack policy

[ Upstream commit 5e469ed9764d4722c59562da13120bd2dc6834c5 ]

When the QoS ack policy was set to non explicit / psmp ack, frames are treated
as not being part of a BA session, which causes extra latency on reordering.
Fix this by only bypassing reordering for packets with no-ack policy

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20220420105038.36443-1-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: qla2xxx: Fix missed DMA unmap for aborted commands

[ Upstream commit 26f9ce53817a8fd84b69a73473a7de852a24c897 ]

Aborting commands that have already been sent to the firmware can
cause BUG in qlt_free_cmd(): BUG_ON(cmd->sg_mapped)

For instance:

 - Command passes rdx_to_xfer state, maps sgl, sends to the firmware

 - Reset occurs, qla2xxx performs ISP error recovery, aborts the command

 - Target stack calls qlt_abort_cmd() and then qlt_free_cmd()

 - BUG_ON(cmd->sg_mapped) in qlt_free_cmd() occurs because sgl was not
   unmapped

Thus, unmap sgl in qlt_abort_cmd() for commands with the aborted flag set.

Link: https://lore.kernel.org/r/AS8PR10MB4952D545F84B6B1DFD39EC1E9DEE9@AS8PR10MB4952.EURPRD10.PROD.OUTLOOK.COM
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Gleb Chesnokov <Chesnokov.G@raidix.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: scsi_dh_alua: Properly handle the ALUA transitioning state

[ Upstream commit 6056a92ceb2a7705d61df7ec5370548e96aee258 ]

The handling of the ALUA transitioning state is currently broken. When a
target goes into this state, it is expected that the target is allowed to
stay in this state for the implicit transition timeout without a path
failure. The handler has this logic, but it gets skipped currently.

When the target transitions, there is in-flight I/O from the initiator. The
first of these responses from the target will be a unit attention letting
the initiator know that the ALUA state has changed.  The remaining
in-flight I/Os, before the initiator finds out that the portal state has
changed, will return not ready, ALUA state is transitioning. The portal
state will change to SCSI_ACCESS_STATE_TRANSITIONING. This will lead to all
new I/O immediately failing the path unexpectedly. The path failure happens
in less than a second instead of the expected successes until the
transition timer is exceeded.

Allow I/Os to continue while the path is in the ALUA transitioning
state. The handler already takes care of a target that stays in the
transitioning state for too long by changing the state to ALUA state
standby once the transition timeout is exceeded at which point the path
will fail.

Link: https://lore.kernel.org/r/CAHZQxy+4sTPz9+pY3=7VJH+CLUJsDct81KtnR2be8ycN5mhqTg@mail.gmail.com
Reviewed-by: Hannes Reinecke <hare@suse.de>
Acked-by: Krishna Kant <krishna.kant@purestorage.com>
Acked-by: Seamus Connor <sconnor@purestorage.com>
Signed-off-by: Brian Bunker <brian@purestorage.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf bench numa: Address compiler error on s390

[ Upstream commit f8ac1c478424a9a14669b8cef7389b1e14e5229d ]

The compilation on s390 results in this error:

  # make DEBUG=y bench/numa.o
  ...
  bench/numa.c: In function ‘__bench_numa’:
  bench/numa.c:1749:81: error: ‘%d’ directive output may be truncated
              writing between 1 and 11 bytes into a region of size between
              10 and 20 [-Werror=format-truncation=]
  1749 |        snprintf(tname, sizeof(tname), "process%d:thread%d", p, t);
                                                               ^~
  ...
  bench/numa.c:1749:64: note: directive argument in the range
                 [-2147483647, 2147483646]
  ...
  #

The maximum length of the %d replacement is 11 characters because of the
negative sign.  Therefore extend the array by two more characters.

Output after:

  # make  DEBUG=y bench/numa.o > /dev/null 2>&1; ll bench/numa.o
  -rw-r--r-- 1 root root 418320 May 19 09:11 bench/numa.o
  #

Fixes: 3aff8ba0a4c9c919 ("perf bench numa: Avoid possible truncation when using snprintf()")
Suggested-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Sumanth Korikkar <sumanthk@linux.ibm.com>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Link: https://lore.kernel.org/r/20220520081158.2990006-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf regs x86: Fix arch__intr_reg_mask() for the hybrid platform

[ Upstream commit 01b28e4a58152e8906eeb5f1b55a0c404c48c7c8 ]

The X86 specific arch__intr_reg_mask() is to check whether the kernel
and hardware can collect XMM registers. But it doesn't work on some
hybrid platform.

Without the patch on ADL-N:

  $ perf record -I?
  available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10
  R11 R12 R13 R14 R15

The config of the test event doesn't contain the PMU information. The
kernel may fail to initialize it on the correct hybrid PMU and return
the wrong non-supported information.

Add the PMU information into the config for the hybrid platform. The
same register set is supported among different hybrid PMUs. Checking
the first available one is good enough.

With the patch on ADL-N:

  $ perf record -I?
  available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10
  R11 R12 R13 R14 R15 XMM0 XMM1 XMM2 XMM3 XMM4 XMM5 XMM6 XMM7 XMM8 XMM9
  XMM10 XMM11 XMM12 XMM13 XMM14 XMM15

Fixes: 6466ec14aaf44ff1 ("perf regs x86: Add X86 specific arch__intr_reg_mask()")
Reported-by: Ammy Yi <ammy.yi@intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Acked-by: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Link: https://lore.kernel.org/r/20220518145125.1494156-1-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpio: mvebu/pwm: Refuse requests with inverted polarity

[ Upstream commit 3ecb10175b1f776f076553c24e2689e42953fef5 ]

The driver doesn't take struct pwm_state::polarity into account when
configuring the hardware, so refuse requests for inverted polarity.

Fixes: 757642f9a584 ("gpio: mvebu: Add limited PWM support")
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpio: gpio-vf610: do not touch other bits when set the target bit

[ Upstream commit 9bf3ac466faa83d51a8fe9212131701e58fdef74 ]

For gpio controller contain register PDDR, when set one target bit,
current logic will clear all other bits, this is wrong. Use operator
'|=' to fix it.

Fixes: 659d8a62311f ("gpio: vf610: add imx7ulp support")
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf build: Fix check for btf__load_from_kernel_by_id() in libbpf

[ Upstream commit 0ae065a5d265bc5ada13e350015458e0c5e5c351 ]

Avi Kivity reported a problem where the __weak
btf__load_from_kernel_by_id() in tools/perf/util/bpf-event.c was being
used and it called btf__get_from_id() in tools/lib/bpf/btf.c that in
turn called back to btf__load_from_kernel_by_id(), resulting in an
endless loop.

Fix this by adding a feature test to check if
btf__load_from_kernel_by_id() is available when building perf with
LIBBPF_DYNAMIC=1, and if not then provide the fallback to the old
btf__get_from_id(), that doesn't call back to btf__load_from_kernel_by_id()
since at that time it didn't exist at all.

Tested on Fedora 35 where we have libbpf-devel 0.4.0 with LIBBPF_DYNAMIC
where we don't have btf__load_from_kernel_by_id() and thus its feature
test fail, not defining HAVE_LIBBPF_BTF__LOAD_FROM_KERNEL_BY_ID:

  $ cat /tmp/build/perf-urgent/feature/test-libbpf-btf__load_from_kernel_by_id.make.output
  test-libbpf-btf__load_from_kernel_by_id.c: In function ‘main’:
  test-libbpf-btf__load_from_kernel_by_id.c:6:16: error: implicit declaration of function ‘btf__load_from_kernel_by_id’ [-Werror=implicit-function-declaration]
      6 |         return btf__load_from_kernel_by_id(20151128, NULL);
        |                ^~~~~~~~~~~~~~~~~~~~~~~~~~~
  cc1: all warnings being treated as errors
  $

  $ nm /tmp/build/perf-urgent/perf | grep btf__load_from_kernel_by_id
  00000000005ba180 T btf__load_from_kernel_by_id
  $

  $ objdump --disassemble=btf__load_from_kernel_by_id -S /tmp/build/perf-urgent/perf

  /tmp/build/perf-urgent/perf:     file format elf64-x86-64
  <SNIP>
  00000000005ba180 <btf__load_from_kernel_by_id>:
  #include "record.h"
  #include "util/synthetic-events.h"

  #ifndef HAVE_LIBBPF_BTF__LOAD_FROM_KERNEL_BY_ID
  struct btf *btf__load_from_kernel_by_id(__u32 id)
  {
    5ba180: 55                    push   %rbp
    5ba181: 48 89 e5              mov    %rsp,%rbp
    5ba184: 48 83 ec 10           sub    $0x10,%rsp
    5ba188: 64 48 8b 04 25 28 00  mov    %fs:0x28,%rax
    5ba18f: 00 00
    5ba191: 48 89 45 f8           mov    %rax,-0x8(%rbp)
    5ba195: 31 c0                 xor    %eax,%eax
         struct btf *btf;
  #pragma GCC diagnostic push
  #pragma GCC diagnostic ignored "-Wdeprecated-declarations"
         int err = btf__get_from_id(id, &btf);
    5ba197: 48 8d 75 f0           lea    -0x10(%rbp),%rsi
    5ba19b: e8 a0 57 e5 ff        call   40f940 <btf__get_from_id@plt>
    5ba1a0: 89 c2                 mov    %eax,%edx
  #pragma GCC diagnostic pop

         return err ? ERR_PTR(err) : btf;
    5ba1a2: 48 98                 cltq
    5ba1a4: 85 d2                 test   %edx,%edx
    5ba1a6: 48 0f 44 45 f0        cmove  -0x10(%rbp),%rax
  }
  <SNIP>

Fixes: 218e7b775d368f38 ("perf bpf: Provide a weak btf__load_from_kernel_by_id() for older libbpf versions")
Reported-by: Avi Kivity <avi@scylladb.com>
Link: https://lore.kernel.org/linux-perf-users/f0add43b-3de5-20c5-22c4-70aff4af959f@scylladb.com
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/linux-perf-users/YobjjFOblY4Xvwo7@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: ufs: core: Fix referencing invalid rsp field

[ Upstream commit d5d92b64408443e113b9742f8f1c35278910dd4d ]

Fix referencing sense data when it is invalid. When the length of the data
segment is 0, there is no valid information in the rsp field, so
ufshpb_rsp_upiu() is returned without additional operation.

Link: https://lore.kernel.org/r/252651381.41652940482659.JavaMail.epsvc@epcpadp4
Fixes: 4b5f49079c52 ("scsi: ufs: ufshpb: L2P map management for HPB read")
Acked-by: Avri Altman <avri.altman@wdc.com>
Signed-off-by: Daejun Park <daejun7.park@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

riscv: dts: sifive: fu540-c000: align dma node name with dtschema

[ Upstream commit b17410182b6f98191fbf7f42d3b4a78512769d29 ]

Fixes dtbs_check warnings like:

  dma@3000000: $nodename:0: 'dma@3000000' does not match '^dma-controller(@.*)?$'

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Link: https://lore.kernel.org/r/20220407193856.18223-1-krzysztof.kozlowski@linaro.org
Fixes: c5ab54e9945b ("riscv: dts: add support for PDMA device of HiFive Unleashed Rev A00")
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: bridge: Clear offload_fwd_mark when passing frame up bridge interface.

[ Upstream commit fbb3abdf2223cd0dfc07de85fe5a43ba7f435bdf ]

It is possible to stack bridges on top of each other. Consider the
following which makes use of an Ethernet switch:

       br1
     /    \
    /      \
   /        \
 br0.11    wlan0
   |
   br0
 /  |  \
p1  p2  p3

br0 is offloaded to the switch. Above br0 is a vlan interface, for
vlan 11. This vlan interface is then a slave of br1. br1 also has a
wireless interface as a slave. This setup trunks wireless lan traffic
over the copper network inside a VLAN.

A frame received on p1 which is passed up to the bridge has the
skb->offload_fwd_mark flag set to true, indicating that the switch has
dealt with forwarding the frame out ports p2 and p3 as needed. This
flag instructs the software bridge it does not need to pass the frame
back down again. However, the flag is not getting reset when the frame
is passed upwards. As a result br1 sees the flag, wrongly interprets
it, and fails to forward the frame to wlan0.

When passing a frame upwards, clear the flag. This is the Rx
equivalent of br_switchdev_frame_unmark() in br_dev_xmit().

Fixes: f1c2eddf4cb6 ("bridge: switchdev: Use an helper to clear forward mark")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://lore.kernel.org/r/20220518005840.771575-1-andrew@lunn.ch
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: flowtable: move dst_check to packet path

[ Upstream commit 2738d9d963bd1f06d5114c2b4fa5771a95703991 ]

Fixes sporadic IPv6 packet loss when flow offloading is enabled.

IPv6 route GC and flowtable GC are not synchronized.
When dst_cache becomes stale and a packet passes through the flow before
the flowtable GC teardowns it, the packet can be dropped.
So, it is necessary to check dst every time in packet path.

Fixes: 227e1e4d0d6c ("netfilter: nf_flowtable: skip device lookup from interface index")
Signed-off-by: Ritaro Takenaka <ritarot634@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: flowtable: pass flowtable to nf_flow_table_iterate()

[ Upstream commit 217cff36e885627c41a14e803fc44f9cbc945767 ]

The flowtable object is already passed as argument to
nf_flow_table_iterate(), do use not data pointer to pass flowtable.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: flowtable: fix TCP flow teardown

[ Upstream commit e5eaac2beb54f0a16ff851125082d9faeb475572 ]

This patch addresses three possible problems:

1. ct gc may race to undo the timeout adjustment of the packet path, leaving
   the conntrack entry in place with the internal offload timeout (one day).

2. ct gc removes the ct because the IPS_OFFLOAD_BIT is not set and the CLOSE
   timeout is reached before the flow offload del.

3. tcp ct is always set to ESTABLISHED with a very long timeout
   in flow offload teardown/delete even though the state might be already
   CLOSED. Also as a remark we cannot assume that the FIN or RST packet
   is hitting flow table teardown as the packet might get bumped to the
   slow path in nftables.

This patch resets IPS_OFFLOAD_BIT from flow_offload_teardown(), so
conntrack handles the tcp rst/fin packet which triggers the CLOSE/FIN
state transition.

Moreover, teturn the connection's ownership to conntrack upon teardown
by clearing the offload flag and fixing the established timeout value.
The flow table GC thread will asynchonrnously free the flow table and
hardware offload entries.

Before this patch, the IPS_OFFLOAD_BIT remained set for expired flows on
which is also misleading since the flow is back to classic conntrack
path.

If nf_ct_delete() removes the entry from the conntrack table, then it
calls nf_ct_put() which decrements the refcnt. This is not a problem
because the flowtable holds a reference to the conntrack object from
flow_offload_alloc() path which is released via flow_offload_free().

This patch also updates nft_flow_offload to skip packets in SYN_RECV
state. Since we might miss or bump packets to slow path, we do not know
what will happen there while we are still in SYN_RECV, this patch
postpones offload up to the next packet which also aligns to the
existing behaviour in tc-ct.

flow_offload_teardown() does not reset the existing tcp state from
flow_offload_fixup_tcp() to ESTABLISHED anymore, packets bump to slow
path might have already update the state to CLOSE/FIN.

Joint work with Oz and Sven.

Fixes: 1e5b2471bcc4 ("netfilter: nf_flow_table: teardown flow timeout race")
Signed-off-by: Oz Shlomo <ozsh@nvidia.com>
Signed-off-by: Sven Auhagen <sven.auhagen@voleatech.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

igb: skip phy status check where unavailable

[ Upstream commit 942d2ad5d2e0df758a645ddfadffde2795322728 ]

igb_read_phy_reg() will silently return, leaving phy_data untouched, if
hw->ops.read_reg isn't set. Depending on the uninitialized value of
phy_data, this led to the phy status check either succeeding immediately
or looping continuously for 2 seconds before emitting a noisy err-level
timeout. This message went out to the console even though there was no
actual problem.

Instead, first check if there is read_reg function pointer. If not,
proceed without trying to check the phy status register.

Fixes: b72f3f72005d ("igb: When GbE link up, wait for Remote receiver status condition")
Signed-off-by: Kevin Mitchell <kevmitch@arista.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mptcp: fix checksum byte order

[ Upstream commit ba2c89e0ea74a904d5231643245753d77422e7f5 ]

The MPTCP code typecasts the checksum value to u16 and
then converts it to big endian while storing the value into
the MPTCP option.

As a result, the wire encoding for little endian host is
wrong, and that causes interoperabilty interoperability
issues with other implementation or host with different endianness.

Address the issue writing in the packet the unmodified __sum16 value.

MPTCP checksum is disabled by default, interoperating with systems
with bad mptcp-level csum encoding should cause fallback to TCP.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/275
Fixes: c5b39e26d003 ("mptcp: send out checksum for DSS")
Fixes: 390b95a5fb84 ("mptcp: receive checksum for DSS")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mptcp: reuse __mptcp_make_csum in validate_data_csum

[ Upstream commit 8401e87f5a36d370cbf1e9d4ba602a553ce9324a ]

This patch reused __mptcp_make_csum() in validate_data_csum() instead of
open-coding.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mptcp: change the parameter of __mptcp_make_csum

[ Upstream commit c312ee219100e86143a1d3cc10b367bc43a0e0b8 ]

This patch changed the type of the last parameter of __mptcp_make_csum()
from __sum16 to __wsum. And export this function in protocol.h.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: 9197/1: spectre-bhb: fix loop8 sequence for Thumb2

[ Upstream commit 3cfb3019979666bdf33a1010147363cf05e0f17b ]

In Thumb2, 'b . + 4' produces a branch instruction that uses a narrow
encoding, and so it does not jump to the following instruction as
expected. So use W(b) instead.

Fixes: 6c7cb60bff7a ("ARM: fix Thumb2 regression with Spectre BHB")
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: 9196/1: spectre-bhb: enable for Cortex-A15

[ Upstream commit 0dc14aa94ccd8ba35eb17a0f9b123d1566efd39e ]

The Spectre-BHB mitigations were inadvertently left disabled for
Cortex-A15, due to the fact that cpu_v7_bugs_init() is not called in
that case. So fix that.

Fixes: b9baf5c8c5c3 ("ARM: Spectre-BHB workaround")
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: af_key: add check for pfkey_broadcast in function pfkey_process

[ Upstream commit 4dc2a5a8f6754492180741facf2a8787f2c415d7 ]

If skb_clone() returns null pointer, pfkey_broadcast() will
return error.
Therefore, it should be better to check the return value of
pfkey_broadcast() and return error if fails.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Properly block LRO when XDP is enabled

[ Upstream commit cf6e34c8c22fba66bd21244b95ea47e235f68974 ]

LRO is incompatible and mutually exclusive with XDP. However, the needed
checks are only made when enabling XDP. If LRO is enabled when XDP is
already active, the command will succeed, and XDP will be skipped in the
data path, although still enabled.

This commit fixes the bug by checking the XDP status in
mlx5e_fix_features and disabling LRO if XDP is enabled.

Fixes: 86994156c736 ("net/mlx5e: XDP fast RX drop bpf programs support")
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: DR, Fix missing flow_source when creating multi-destination FW table

[ Upstream commit 2c5fc6cd269ad3476da99dad02521d2af4a8e906 ]

In order to support multiple destination FTEs with SW steering
FW table is created with single FTE with multiple actions and
SW steering rule forward to it. When creating this table, flow
source isn't set according to the original FTE.

Fix this by passing the original FTE flow source to the created
FW table.

Fixes: 34583beea4b7 ("net/mlx5: DR, Create multi-destination table for SW-steering use")
Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

NFC: nci: fix sleep in atomic context bugs caused by nci_skb_alloc

[ Upstream commit 23dd4581350d4ffa23d58976ec46408f8f4c1e16 ]

There are sleep in atomic context bugs when the request to secure
element of st-nci is timeout. The root cause is that nci_skb_alloc
with GFP_KERNEL parameter is called in st_nci_se_wt_timeout which is
a timer handler. The call paths that could trigger bugs are shown below:

    (interrupt context 1)
st_nci_se_wt_timeout
  nci_hci_send_event
    nci_hci_send_data
      nci_skb_alloc(..., GFP_KERNEL) //may sleep

   (interrupt context 2)
st_nci_se_wt_timeout
  nci_hci_send_event
    nci_hci_send_data
      nci_send_data
        nci_queue_tx_data_frags
          nci_skb_alloc(..., GFP_KERNEL) //may sleep

This patch changes allocation mode of nci_skb_alloc from GFP_KERNEL to
GFP_ATOMIC in order to prevent atomic context sleeping. The GFP_ATOMIC
flag makes memory allocation operation could be used in atomic context.

Fixes: ed06aeefdac3 ("nfc: st-nci: Rename st21nfcb to st-nci")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20220517012530.75714-1-duoming@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/qla3xxx: Fix a test in ql_reset_work()

[ Upstream commit 5361448e45fac6fb96738df748229432a62d78b6 ]

test_bit() tests if one bit is set or not.
Here the logic seems to check of bit QL_RESET_PER_SCSI (i.e. 4) OR bit
QL_RESET_START (i.e. 3) is set.

In fact, it checks if bit 7 (4 | 3 = 7) is set, that is to say
QL_ADAPTER_UP.

This looks harmless, because this bit is likely be set, and when the
ql_reset_work() delayed work is scheduled in ql3xxx_isr() (the only place
that schedule this work), QL_RESET_START or QL_RESET_PER_SCSI is set.

This has been spotted by smatch.

Fixes: 5a4faa873782 ("[PATCH] qla3xxx NIC driver")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/80e73e33f390001d9c0140ffa9baddf6466a41a2.1652637337.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: at91: generated: consider range when calculating best rate

[ Upstream commit d0031e6fbed955ff8d5f5bbc8fe7382482559cec ]

clk_generated_best_diff() helps in finding the parent and the divisor to
compute a rate closest to the required one. However, it doesn't take into
account the request's range for the new rate. Make sure the new rate
is within the required range.

Fixes: 8a8f4bf0c480 ("clk: at91: clk-generated: create function to find best_diff")
Signed-off-by: Codrin Ciubotariu <codrin.ciubotariu@microchip.com>
Link: https://lore.kernel.org/r/20220413071318.244912-1-codrin.ciubotariu@microchip.com
Reviewed-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: Fix interrupt moderation settings getting cleared

[ Upstream commit bf13502ed5f941b0777b3fd1e24dac5d93f3886c ]

Adaptive-rx and Adaptive-tx are interrupt moderation settings
that can be enabled/disabled using ethtool:
ethtool -C ethX adaptive-rx on/off adaptive-tx on/off

Unfortunately those settings are getting cleared after
changing number of queues, or in ethtool world 'channels':
ethtool -L ethX rx 1 tx 1

Clearing was happening due to introduction of bit fields
in ice_ring_container struct. This way only itr_setting
bits were rebuilt during ice_vsi_rebuild_set_coalesce().

Introduce an anonymous struct of bitfields and create a
union to refer to them as a single variable.
This way variable can be easily saved and restored.

Fixes: 61dc79ced7aa ("ice: Restore interrupt throttle settings after VSI rebuild")
Signed-off-by: Michal Wilczynski <michal.wilczynski@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: move ice_container_type onto ice_ring_container

[ Upstream commit dc23715cf30a9acb808f5b08962877c390d3e6ea ]

Currently ice_container_type is scoped only for ice_ethtool.c. Next
commit that will split the ice_ring struct onto Rx/Tx specific ring
structs is going to also modify the type of linked list of rings that is
within ice_ring_container. Therefore, the functions that are taking the
ice_ring_container as an input argument will need to be aware of a ring
type that will be looked up.

Embed ice_container_type within ice_ring_container and initialize it
properly when allocating the q_vectors.

Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: fix possible under reporting of ethtool Tx and Rx statistics

[ Upstream commit 31b6298fd8e29effe9ed6b77351ac5969be56ce0 ]

The hardware statistics counters are not cleared during resets so the
drivers first access is to initialize the baseline and then subsequent
reads are for reporting the counters. The statistics counters are read
during the watchdog subtask when the interface is up. If the baseline
is not initialized before the interface is up, then there can be a brief
window in which some traffic can be transmitted/received before the
initial baseline reading takes place.

Directly initialize ethtool statistics in driver open so the baseline will
be initialized when the interface is up, and any dropped packets
incremented before the interface is up won't be reported.

Fixes: 28dc1b86f8ea9 ("ice: ignore dropped packets during init")
Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: fix crash when writing timestamp on RX rings

[ Upstream commit 4503cc7fdf9a84cd631b0cb8ecb3c9b1bdbf3594 ]

Do not allow to write timestamps on RX rings if PF is being configured.
When PF is being configured RX rings can be freed or rebuilt. If at the
same time timestamps are updated, the kernel will crash by dereferencing
null RX ring pointer.

PID: 1449   TASK: ff187d28ed658040  CPU: 34  COMMAND: "ice-ptp-0000:51"
 #0 [ff1966a94a713bb0] machine_kexec at ffffffff9d05a0be
 #1 [ff1966a94a713c08] __crash_kexec at ffffffff9d192e9d
 #2 [ff1966a94a713cd0] crash_kexec at ffffffff9d1941bd
 #3 [ff1966a94a713ce8] oops_end at ffffffff9d01bd54
 #4 [ff1966a94a713d08] no_context at ffffffff9d06bda4
 #5 [ff1966a94a713d60] __bad_area_nosemaphore at ffffffff9d06c10c
 #6 [ff1966a94a713da8] do_page_fault at ffffffff9d06cae4
 #7 [ff1966a94a713de0] page_fault at ffffffff9da0107e
    [exception RIP: ice_ptp_update_cached_phctime+91]
    RIP: ffffffffc076db8b  RSP: ff1966a94a713e98  RFLAGS: 00010246
    RAX: 16e3db9c6b7ccae4  RBX: ff187d269dd3c180  RCX: ff187d269cd4d018
    RDX: 0000000000000000  RSI: 0000000000000000  RDI: 0000000000000000
    RBP: ff187d269cfcc644   R8: ff187d339b9641b0   R9: 0000000000000000
    R10: 0000000000000002  R11: 0000000000000000  R12: ff187d269cfcc648
    R13: ffffffff9f128784  R14: ffffffff9d101b70  R15: ff187d269cfcc640
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #8 [ff1966a94a713ea0] ice_ptp_periodic_work at ffffffffc076dbef [ice]
 #9 [ff1966a94a713ee0] kthread_worker_fn at ffffffff9d101c1b
 #10 [ff1966a94a713f10] kthread at ffffffff9d101b4d
 #11 [ff1966a94a713f50] ret_from_fork at ffffffff9da0023f

Fixes: 77a781155a65 ("ice: enable receive hardware timestamping")
Signed-off-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Reviewed-by: Michal Schmidt <mschmidt@redhat.com>
Tested-by: Dave Cain <dcain@redhat.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: vmxnet3: fix possible NULL pointer dereference in vmxnet3_rq_cleanup()

[ Upstream commit edf410cb74dc612fd47ef5be319c5a0bcd6e6ccd ]

In vmxnet3_rq_create(), when dma_alloc_coherent() fails,
vmxnet3_rq_destroy() is called. It sets rq->rx_ring[i].base to NULL. Then
vmxnet3_rq_create() returns an error to its callers mxnet3_rq_create_all()
-> vmxnet3_change_mtu(). Then vmxnet3_change_mtu() calls
vmxnet3_force_close() -> dev_close() in error handling code. And the driver
calls vmxnet3_close() -> vmxnet3_quiesce_dev() -> vmxnet3_rq_cleanup_all()
-> vmxnet3_rq_cleanup(). In vmxnet3_rq_cleanup(),
rq->rx_ring[ring_idx].base is accessed, but this variable is NULL, causing
a NULL pointer dereference.

To fix this possible bug, an if statement is added to check whether
rq->rx_ring[0].base is NULL in vmxnet3_rq_cleanup() and exit early if so.

The error log in our fault-injection testing is shown as follows:

[   65.220135] BUG: kernel NULL pointer dereference, address: 0000000000000008
...
[   65.222633] RIP: 0010:vmxnet3_rq_cleanup_all+0x396/0x4e0 [vmxnet3]
...
[   65.227977] Call Trace:
...
[   65.228262]  vmxnet3_quiesce_dev+0x80f/0x8a0 [vmxnet3]
[   65.228580]  vmxnet3_close+0x2c4/0x3f0 [vmxnet3]
[   65.228866]  __dev_close_many+0x288/0x350
[   65.229607]  dev_close_many+0xa4/0x480
[   65.231124]  dev_close+0x138/0x230
[   65.231933]  vmxnet3_force_close+0x1f0/0x240 [vmxnet3]
[   65.232248]  vmxnet3_change_mtu+0x75d/0x920 [vmxnet3]
...

Fixes: d1a890fa37f27 ("net: VMware virtual Ethernet NIC driver: vmxnet3")
Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Zixuan Fu <r33s3n6@gmail.com>
Link: https://lore.kernel.org/r/20220514050711.2636709-1-r33s3n6@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: vmxnet3: fix possible use-after-free bugs in vmxnet3_rq_alloc_rx_buf()

[ Upstream commit 9e7fef9521e73ca8afd7da9e58c14654b02dfad8 ]

In vmxnet3_rq_alloc_rx_buf(), when dma_map_single() fails, rbi->skb is
freed immediately. Similarly, in another branch, when dma_map_page() fails,
rbi->page is also freed. In the two cases, vmxnet3_rq_alloc_rx_buf()
returns an error to its callers vmxnet3_rq_init() -> vmxnet3_rq_init_all()
-> vmxnet3_activate_dev(). Then vmxnet3_activate_dev() calls
vmxnet3_rq_cleanup_all() in error handling code, and rbi->skb or rbi->page
are freed again in vmxnet3_rq_cleanup_all(), causing use-after-free bugs.

To fix these possible bugs, rbi->skb and rbi->page should be cleared after
they are freed.

The error log in our fault-injection testing is shown as follows:

[   14.319016] BUG: KASAN: use-after-free in consume_skb+0x2f/0x150
...
[   14.321586] Call Trace:
...
[   14.325357]  consume_skb+0x2f/0x150
[   14.325671]  vmxnet3_rq_cleanup_all+0x33a/0x4e0 [vmxnet3]
[   14.326150]  vmxnet3_activate_dev+0xb9d/0x2ca0 [vmxnet3]
[   14.326616]  vmxnet3_open+0x387/0x470 [vmxnet3]
...
[   14.361675] Allocated by task 351:
...
[   14.362688]  __netdev_alloc_skb+0x1b3/0x6f0
[   14.362960]  vmxnet3_rq_alloc_rx_buf+0x1b0/0x8d0 [vmxnet3]
[   14.363317]  vmxnet3_activate_dev+0x3e3/0x2ca0 [vmxnet3]
[   14.363661]  vmxnet3_open+0x387/0x470 [vmxnet3]
...
[   14.367309]
[   14.367412] Freed by task 351:
...
[   14.368932]  __dev_kfree_skb_any+0xd2/0xe0
[   14.369193]  vmxnet3_rq_alloc_rx_buf+0x71e/0x8d0 [vmxnet3]
[   14.369544]  vmxnet3_activate_dev+0x3e3/0x2ca0 [vmxnet3]
[   14.369883]  vmxnet3_open+0x387/0x470 [vmxnet3]
[   14.370174]  __dev_open+0x28a/0x420
[   14.370399]  __dev_change_flags+0x192/0x590
[   14.370667]  dev_change_flags+0x7a/0x180
[   14.370919]  do_setlink+0xb28/0x3570
[   14.371150]  rtnl_newlink+0x1160/0x1740
[   14.371399]  rtnetlink_rcv_msg+0x5bf/0xa50
[   14.371661]  netlink_rcv_skb+0x1cd/0x3e0
[   14.371913]  netlink_unicast+0x5dc/0x840
[   14.372169]  netlink_sendmsg+0x856/0xc40
[   14.372420]  ____sys_sendmsg+0x8a7/0x8d0
[   14.372673]  __sys_sendmsg+0x1c2/0x270
[   14.372914]  do_syscall_64+0x41/0x90
[   14.373145]  entry_SYSCALL_64_after_hwframe+0x44/0xae
...

Fixes: 5738a09d58d5a ("vmxnet3: fix checks for dma mapping errors")
Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Zixuan Fu <r33s3n6@gmail.com>
Link: https://lore.kernel.org/r/20220514050656.2636588-1-r33s3n6@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: systemport: Fix an error handling path in bcm_sysport_probe()

[ Upstream commit ef6b1cd11962aec21c58d137006ab122dbc8d6fd ]

if devm_clk_get_optional() fails, we still need to go through the error
handling path.

Add the missing goto.

Fixes: 6328a126896ea ("net: systemport: Manage Wake-on-LAN clock")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/99d70634a81c229885ae9e4ee69b2035749f7edc.1652634040.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Revert "PCI: aardvark: Rewrite IRQ code to chained IRQ handler"

[ Upstream commit a3b69dd0ad6265c29c4b6fb381cd76fb3bebdf8c ]

This reverts commit 1571d67dc190e50c6c56e8f88cdc39f7cc53166e.

This commit broke support for setting interrupt affinity. It looks like
that it is related to the chained IRQ handler. Revert this commit until
issue with setting interrupt affinity is fixed.

Fixes: 1571d67dc190 ("PCI: aardvark: Rewrite IRQ code to chained IRQ handler")
Link: https://lore.kernel.org/r/20220515125815.30157-1-pali@kernel.org
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nft_flow_offload: fix offload with pppoe + vlan

[ Upstream commit 2456074935003b66c40f78df6adfc722435d43ea ]

When running a combination of PPPoE on top of a VLAN, we need to set
info->outdev to the PPPoE device, otherwise PPPoE encap is skipped
during software offload.

Fixes: 72efd585f714 ("netfilter: flowtable: add pppoe support")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: fix dev_fill_forward_path with pppoe + bridge

[ Upstream commit cf2df74e202d81b09f09d84c2d8903e0e87e9274 ]

When calling dev_fill_forward_path on a pppoe device, the provided destination
address is invalid. In order for the bridge fdb lookup to succeed, the pppoe
code needs to update ctx->daddr to the correct value.
Fix this by storing the address inside struct net_device_path_ctx

Fixes: f6efc675c9dd ("net: ppp: resolve forwarding path for bridge pppoe devices")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nft_flow_offload: skip dst neigh lookup for ppp devices

[ Upstream commit 45ca3e61999e9a30ca2b7cfbf9da8a9f8d13be31 ]

The dst entry does not contain a valid hardware address, so skip the lookup
in order to avoid running into errors here.
The proper hardware address is filled in from nft_dev_path_info

Fixes: 72efd585f714 ("netfilter: flowtable: add pppoe support")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>