platform/kernel/linux-rpi.git
2 years agoMerge remote-tracking branch 'stable/linux-5.15.y' into rpi-5.15.y
Dom Cobley [Tue, 26 Apr 2022 10:42:26 +0000 (11:42 +0100)]
Merge remote-tracking branch 'stable/linux-5.15.y' into rpi-5.15.y

2 years agoARM: dts: bcm2835: Switch HSM clock to firmware
Maxime Ripard [Mon, 25 Apr 2022 15:45:16 +0000 (17:45 +0200)]
ARM: dts: bcm2835: Switch HSM clock to firmware

When the bootloader boots without a display connected to the
HDMI connector, it won't initialize the HSM clock.

Then, when Linux boots, the default parent fetched by the clock
framework will be the first parent for the HSM clock, which is "gnd".

That clock is never registered, leading to the HSM clock being orphaned,
which leads to other issues with the recent clock range handling.

It's not clear what the proper solution would be, but the firmware
clocks driver doesn't have that parenting issue and the clock rate is
properly initialized for this clock whenever it hasn't been set by the
bootloader.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agooverlays: Fix pitft28/35-resistive rotate params
Phil Elwell [Sun, 24 Apr 2022 19:25:34 +0000 (20:25 +0100)]
overlays: Fix pitft28/35-resistive rotate params

The drm drivers uses the property name "rotation", not "rotate" as used
by the fbtft drivers. Fix suggested by notro.

See: https://forums.raspberrypi.com/viewtopic.php?p=1996785#p1996785

Signed-off-by: Phil Elwell <phil@raspberrypi.com>
2 years agoARM: dts: Add i2c0mux node to Model B rev 1
Phil Elwell [Tue, 19 Apr 2022 18:57:29 +0000 (19:57 +0100)]
ARM: dts: Add i2c0mux node to Model B rev 1

The lack of an i2c0mux label prevents the i2c-rtc overlay from
being applied.

See: https://github.com/raspberrypi/linux/issues/4999

Signed-off-by: Phil Elwell <phil@raspberrypi.com>
2 years agooverlays: Add "drm" parameter to pitft28-resistive
andrum99 [Sun, 24 Apr 2022 16:23:23 +0000 (17:23 +0100)]
overlays: Add "drm" parameter to pitft28-resistive

Add "drm" parameter to the pitft28-resistive, forcing use of the
mi0283qt DRM driver.

2 years agotpm_tis_spi_main: Force probe routine to run synchronously with driver and device...
Scovotto Davide [Wed, 20 Apr 2022 15:22:17 +0000 (17:22 +0200)]
tpm_tis_spi_main: Force probe routine to run synchronously with driver and device registration when IMA is enabled

Co-authored-by: Alberto Solavagione <albertosolavagione30@gmail.com>
Co-developed-by: Alberto Solavagione <albertosolavagione30@gmail.com>
Signed-off-by: Alberto Solavagione <albertosolavagione30@gmail.com>
Signed-off-by: Davide Scovotto <scovottodavide@gmail.com>
2 years agoclk-bcm2835: use subsys_initcall for the clock driver when IMA is enabled
Alberto Solavagione [Wed, 20 Apr 2022 15:15:42 +0000 (17:15 +0200)]
clk-bcm2835: use subsys_initcall for the clock driver when IMA is enabled

Co-authored-by: Davide Scovotto <scovottodavide@gmail.com>
Co-developed-by: Davide Scovotto <scovottodavide@gmail.com>
Signed-off-by: Davide Scovotto <scovottodavide@gmail.com>
Signed-off-by: Alberto Solavagione <albertosolavagione30@gmail.com>
2 years agorpivid: Use clk_get_max_rate()
Maxime Ripard [Fri, 15 Apr 2022 13:26:07 +0000 (15:26 +0200)]
rpivid: Use clk_get_max_rate()

The driver was using clk_round_rate() to figure out the maximum clock
rate that was allowed for the HEVC clock.

Since we have a function to return it directly now, let's use it.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agodrm/vc4: kms: Use maximum FIFO load for the HVS clock rate
Maxime Ripard [Wed, 13 Apr 2022 14:22:49 +0000 (16:22 +0200)]
drm/vc4: kms: Use maximum FIFO load for the HVS clock rate

The core clock computation takes into account both the load due to the
input (ie, planes) and its output (ie, encoders).

However, while the input load needs to consider all the planes, and thus
sum all of their associated loads, the output happens mostly in
parallel.

Therefore, we need to consider only the maximum of all the output loads,
and not the sum like we were doing. This resulted in a clock rate way
too high which could be discarded for being too high by the clock
framework.

Since recent changes, the clock framework will even downright reject it,
leading to a core clock being too low for its current needs.

Fixes: 16e101051f32 ("drm/vc4: Increase the core clock based on HVS load")
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agodrm/vc4: Make sure we don't end up with a core clock too high
Maxime Ripard [Fri, 25 Mar 2022 16:09:41 +0000 (17:09 +0100)]
drm/vc4: Make sure we don't end up with a core clock too high

Following the clock rate range improvements to the clock framework,
trying to set a disjoint range on a clock will now result in an error.

Thus, we can't set a minimum rate higher than the maximum reported by
the firmware, or clk_set_min_rate() will fail.

Thus we need to clamp the rate we are about to ask for to the maximum
rate possible on that clock.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agodrm/vc4: kms: Warn if clk_set_min_rate fails
Maxime Ripard [Fri, 25 Mar 2022 14:48:48 +0000 (15:48 +0100)]
drm/vc4: kms: Warn if clk_set_min_rate fails

We currently ignore the clk_set_min_rate return code assuming it would
succeed. However, it can fail if we ask for a rate higher than the
current maximum for example.

Since we can't fail in atomic_commit, at least warn on failure.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agodrm/vc4: hdmi: Rework hdmi_enable_4kp60 detection
Maxime Ripard [Thu, 24 Mar 2022 10:57:57 +0000 (11:57 +0100)]
drm/vc4: hdmi: Rework hdmi_enable_4kp60 detection

In order to support higher HDMI frequencies, users have to set the
hdmi_enable_4kp60 parameter in their config.txt file.

We were detecting this so far by calling clk_round_rate on the core
clock with the frequency we're supposed to run at when one of those
modes is enabled. Whether or not the parameter was enabled could then be
inferred by the returned rate since the maximum clock rate reported by
the firmware was one of the side effect of setting that parameter.

However, the recent clock rework we did changed what clk_round_rate was
returning to always return the minimum allowed, and thus this test
wasn't reliable anymore.

Let's instead try to set a minimum on that clock for the rate we'd like
to reach. If the maximum reported by the firmware is below the minimum
we're trying to set, the clock framework will return an error which we
then can use to infer whether the parameter is set or not.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add missing test case for ranges
Maxime Ripard [Fri, 15 Apr 2022 13:00:44 +0000 (15:00 +0200)]
clk: tests: Add missing test case for ranges

Let's add a test on the rate range after a reparenting. This fails for
now, but it's worth having it to document the corner cases we don't
support yet.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add some tests for clk_get_rate_range()
Maxime Ripard [Fri, 15 Apr 2022 12:45:17 +0000 (14:45 +0200)]
clk: tests: Add some tests for clk_get_rate_range()

Let's introduce a bunch of unit tests to make sure the values returned
by clk_get_rate_range() are sane.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Add clk_get_rate_range
Maxime Ripard [Fri, 15 Apr 2022 12:17:41 +0000 (14:17 +0200)]
clk: Add clk_get_rate_range

With the recent introduction of clock drivers that will force their
clock rate to either the minimum or maximum boundaries, it becomes
harder for clock users to discover either boundary of their clock.

Indeed, the best way to do that previously was to call clk_round_rate()
on either 0 or ULONG_MAX and count on the driver to clamp the rate to
the current boundary, but that won't work anymore.

Since any other alternative (calling clk_set_rate_range() and looking at
the returned value, calling clk_round_rate() still, or just doing
nothing) depends on how the driver will behaves, we actually are
punching a hole through the abstraction provided by the clock framework.

In order to avoid any abstraction violation, let's create a bunch of
accessors that will return the current minimum and maximum for a given
clock.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoRevert "clk: Introduce a clock request API"
Maxime Ripard [Wed, 16 Mar 2022 08:41:49 +0000 (09:41 +0100)]
Revert "clk: Introduce a clock request API"

This reverts commit 23fbabe348436e250deca7c5f2fd0caf620af174.

2 years agoRevert "clk: requests: Ignore if the pointer is null"
Maxime Ripard [Wed, 16 Mar 2022 08:41:48 +0000 (09:41 +0100)]
Revert "clk: requests: Ignore if the pointer is null"

This reverts commit f078b2c9b1901f6297154788ac07f56547ddcb7a.

2 years agoRevert "clk: requests: Dereference the request pointer after the check"
Maxime Ripard [Wed, 16 Mar 2022 08:41:44 +0000 (09:41 +0100)]
Revert "clk: requests: Dereference the request pointer after the check"

This reverts commit 12917adc036f82c4fda3b80a068d0d51c947d6a6.

2 years agoRevert "rpivid: Switch to new clock api"
Maxime Ripard [Wed, 16 Mar 2022 08:58:31 +0000 (09:58 +0100)]
Revert "rpivid: Switch to new clock api"

This reverts commit ec7556e20c2c29c3df9493248a1a4d60ed20ae38.

2 years agoRevert "bcm2835-unicam: Switch to new clock api"
Maxime Ripard [Wed, 16 Mar 2022 08:53:03 +0000 (09:53 +0100)]
Revert "bcm2835-unicam: Switch to new clock api"

This reverts commit 702228eb413876739f4fee8a9ec66b3e4e54efac.

2 years agoRevert "drm/vc4: hdmi: Convert to the new clock request API"
Maxime Ripard [Wed, 16 Mar 2022 08:52:34 +0000 (09:52 +0100)]
Revert "drm/vc4: hdmi: Convert to the new clock request API"

This reverts commit 5dbb9357519a9479eacdda130713bf0dc44e069c.

2 years agoRevert "drm/vc4: Increase the core clock based on HVS load"
Maxime Ripard [Wed, 16 Mar 2022 08:51:44 +0000 (09:51 +0100)]
Revert "drm/vc4: Increase the core clock based on HVS load"

This reverts commit 02c8543cc6940f8201e9beb601bd56421d911e83.

2 years agoRevert "drm/vc4: kms: Move clock request to our HVS state"
Maxime Ripard [Wed, 16 Mar 2022 08:51:32 +0000 (09:51 +0100)]
Revert "drm/vc4: kms: Move clock request to our HVS state"

This reverts commit c65633a429b15f9d182a5bc7d6387fecbd5b7bb0.

2 years agoclk: bcm: rpi: Run some clocks at the minimum rate allowed
Maxime Ripard [Mon, 17 Jan 2022 16:36:23 +0000 (17:36 +0100)]
clk: bcm: rpi: Run some clocks at the minimum rate allowed

The core clock and M2MC clocks are shared between some devices (Unicam
controllers and the HVS, and the HDMI controllers, respectively) that
will have various, varying, requirements depending on their current work
load.

Since those loads can require a fairly high clock rate in extreme
conditions (up to ~600MHz), we can end up running those clocks at their
maximum frequency even though we no longer require such a high rate.

Fortunately, those devices don't require an exact rate but a minimum
rate, and all the drivers are using clk_set_min_rate. Thus, we can just
rely on the fact that the clk_request minimum (which is the aggregated
minimum of all the clock users) is what we want at all times.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: bcm: rpi: Set a default minimum rate
Maxime Ripard [Mon, 17 Jan 2022 16:31:06 +0000 (17:31 +0100)]
clk: bcm: rpi: Set a default minimum rate

The M2MC clock provides the state machine clock for both HDMI
controllers.

However, if no HDMI monitor is plugged in at boot, its clock rate will
be left at 0 by the firmware and will make any register access end up in
a CPU stall, even though the clock was enabled.

We had some code in the HDMI controller to deal with this before, but it
makes more sense to have it in the clock driver. Move it there.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: bcm: rpi: Add variant structure
Maxime Ripard [Mon, 17 Jan 2022 16:19:30 +0000 (17:19 +0100)]
clk: bcm: rpi: Add variant structure

We only export a bunch of firmware clocks, and some of them require
special treatment.

This has been do so far using some tests on the clock id in various
places, but this is fairly hard to extend and doesn't scale very well.

Since we'll need some more cases in the next patches, let's switch to a
variant structure that defines the behaviour we need to have for a given
clock.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Test the clock pointer in clk_hw_get_name()
Maxime Ripard [Tue, 5 Apr 2022 13:20:27 +0000 (15:20 +0200)]
clk: Test the clock pointer in clk_hw_get_name()

Unlike __clk_get_name(), clk_hw_get_name() doesn't test wether passed
clk_hw pointer is NULL or not and dereferences it directly. This can
then lead to NULL pointer dereference.

Let's make sure the pointer isn't NULL before dereferencing it.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Zero the clk_rate_request structure
Maxime Ripard [Sat, 2 Apr 2022 12:37:39 +0000 (14:37 +0200)]
clk: Zero the clk_rate_request structure

In order to make sure we don't carry anything over from an already
existing clk_rate_request pointer we would pass to
clk_core_init_rate_req(), let's zero the entire structure before
initializing it.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Stop forwarding clk_rate_requests to the parent
Maxime Ripard [Sat, 2 Apr 2022 12:36:59 +0000 (14:36 +0200)]
clk: Stop forwarding clk_rate_requests to the parent

If the clock cannot modify its rate and has CLK_SET_RATE_PARENT,
clk_mux_determine_rate_flags() and clk_core_round_rate_nolock() will
call clk_core_round_rate_nolock() with its parent clock but use the
request of the child node either directly (clk_core_round_rate_nolock())
or by copying it (clk_mux_determine_rate_flags()).

Both cases are problematic since the parent will now have a request with
the best parent fields of the child (so pointing to itself) and the
boundaries of the child as well.

clk_core_round_rate_nolock() is even worse since we would directly
modify the caller structure if the parent was ever to modify its own
parent or its parent rate, then returning to the caller a best parent
that isn't a parent of the clock we just called clk_determine_rate()
onto.

Let's create a new function that will create a new request to forward to
the parent, clk_core_forward_rate_req() and update the relevant call
sites to that new function.

Let's also add a test to make sure we avoid regressions there.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Introduce clk_core_has_parent()
Maxime Ripard [Thu, 7 Apr 2022 12:11:37 +0000 (14:11 +0200)]
clk: Introduce clk_core_has_parent()

We will need to know if a clk_core pointer has a given parent in other
functions, so let's create a clk_core_has_parent() function that
clk_has_parent() will call into.

For good measure, let's add some unit tests as well to make sure it
works properly.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Switch from __clk_determine_rate to clk_core_round_rate_nolock
Maxime Ripard [Sat, 2 Apr 2022 11:52:55 +0000 (13:52 +0200)]
clk: Switch from __clk_determine_rate to clk_core_round_rate_nolock

clk_mux_determine_rate_flags() will call into __clk_determine_rate()
with a clk_hw pointer, while it has access to the clk_core pointer
already.

This leads to back and forth between clk_hw and clk_core, while
__clk_determine_rate will only call clk_core_round_rate_nolock() with
the clk_core pointer it retrieved from the clk_hw.

Let's simplify things a bit by calling into clk_core_round_rate_nolock
directly.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Remove redundant clk_core_init_rate_req() call
Maxime Ripard [Sat, 2 Apr 2022 12:06:53 +0000 (14:06 +0200)]
clk: Remove redundant clk_core_init_rate_req() call

Since all the users of clk_core_round_rate_nolock() will now properly
initialize, there's no need for it to initialize the request itself.

This is even dangerous, as if the clock cannot change its rate by itself
and has CLK_SET_RATE_PARENT, clk_core_round_rate_nolock() will call
itself with the parent clock but the client clk_rate_request structure.

We will then reinitialize the child request with the parent context
(parent, boundaries, etc.), which is an issue if the parent ever changes
its own parent or parent rate.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Add missing clk_core_init_rate_req calls
Maxime Ripard [Sat, 2 Apr 2022 12:01:37 +0000 (14:01 +0200)]
clk: Add missing clk_core_init_rate_req calls

Some callers of clk_core_round_rate_nolock() will initialize the
clk_rate_request structure by hand, missing a few parameters that leads
to inconsistencies in what drivers can expect from that structure.

Let's use clk_core_init_rate_req() everywhere to make sure it's built in
a consistent manner.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Introduce clk_hw_init_rate_request()
Maxime Ripard [Sat, 2 Apr 2022 12:49:56 +0000 (14:49 +0200)]
clk: Introduce clk_hw_init_rate_request()

Some drivers (at91, imx, qcom) use __clk_determine_rate directly, and
thus will need to initialise a clk_rate_request structure.

However, drivers don't have access to the function that the core uses to
initialize that structure which could prove to be useful.

Let's create a function for clock providers that will initialize a
clk_rate_request for a given clk_hw pointer and a rate.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Change clk_core_init_rate_req prototype
Maxime Ripard [Sat, 2 Apr 2022 11:48:03 +0000 (13:48 +0200)]
clk: Change clk_core_init_rate_req prototype

The expectation is that a clk_rate_request structure is supposed to be
initialized using clk_core_init_rate_req(), yet the rate we want to
request still needs to be set by hand.

Let's just pass the rate as a function argument so that callers don't
have any extra work to do.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Add our request boundaries in clk_core_init_rate_req
Maxime Ripard [Sat, 2 Apr 2022 11:45:05 +0000 (13:45 +0200)]
clk: Add our request boundaries in clk_core_init_rate_req

The expectation is that a new clk_rate_request is initialized through a
call to clk_core_init_rate_req().

However, at the moment it only fills the parent rate and clk_hw pointer,
but omits the other fields such as the clock rate boundaries.

Some users of that function will update them after calling it, but most
don't.

As we are passed the clk_core pointer, we have access to those
boundaries in clk_core_init_rate_req() however, so let's just fill it
there and remove it from the few callers that do it right.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Skip set_rate_range if our clock is orphan
Maxime Ripard [Fri, 1 Apr 2022 13:27:09 +0000 (15:27 +0200)]
clk: Skip set_rate_range if our clock is orphan

clk_set_rate_range will now force a clk_set_rate() call to
core->req_rate. However, if our clock is orphan, req_rate will be 0 and
we will thus end up calling a set_rate to 0 on potentially all that
clock subtree.

This can be fairly invasive and result in poor decisions. In such a
case, let's just store the new range but bail out and skip the set_rate.

When that clock is no longer orphan though, we should be enforcing the
new range but we don't. Let's add a skipped test to make sure we don't
forget about that corner case.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Set req_rate on reparenting
Maxime Ripard [Fri, 1 Apr 2022 13:26:46 +0000 (15:26 +0200)]
clk: Set req_rate on reparenting

If a non-rate clock started by default with a parent that never
registered, core->req_rate will be 0. The expectation is that whenever
the parent will be registered, req_rate will be updated with the new
value that has just been computed.

However, if that clock is a mux, clk_set_parent() can also make that
clock no longer orphan. In this case however, we never update req_rate.
Let's make sure it's the case for the newly unorphan clock and all its
children.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Fix clk_get_parent() documentation
Maxime Ripard [Fri, 1 Apr 2022 09:09:29 +0000 (11:09 +0200)]
clk: Fix clk_get_parent() documentation

The clk_get_parent() documentation in the header states that it will
return a valid pointer, or an error pointer on failure.

However, the documentation in the source file, and the code itself, will
return also return NULL if there isn't any parent for that clock. Let's
mention it.

An orphan clock should return NULL too, so let's add a test for it.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Take into account uncached clocks in clk_set_rate_range()
Maxime Ripard [Fri, 8 Apr 2022 08:06:11 +0000 (10:06 +0200)]
clk: Take into account uncached clocks in clk_set_rate_range()

clk_set_rate_range() will use the last requested rate for the clock when
it calls into the driver set_rate hook.

However, if CLK_GET_RATE_NOCACHE is set on that clock, the last
requested rate might not be matching the current rate of the clock. In
such a case, let's read out the rate from the hardware and use that in
our set_rate instead.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add some tests for orphan with multiple parents
Maxime Ripard [Sat, 2 Apr 2022 15:27:40 +0000 (17:27 +0200)]
clk: tests: Add some tests for orphan with multiple parents

Let's leverage the dummy mux with multiple parents we have to create a
mux whose default parent will never be registered, and thus will always
be orphan by default.

We can then create some tests to make sure that the clock API behaves
properly in such a case, and that the transition to a non-orphan clock
when we change the parent is done properly.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add tests for mux with multiple parents
Maxime Ripard [Thu, 7 Apr 2022 12:58:34 +0000 (14:58 +0200)]
clk: tests: Add tests for mux with multiple parents

We'll need to test a few corner cases that occur when we have a mux
clock whose default parent is missing.

For now, let's create the context structure and the trivial ops, along
with a test suite that just tests trivial things for now, without
considering the orphan case.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add tests for single parent mux
Maxime Ripard [Thu, 7 Apr 2022 12:21:55 +0000 (14:21 +0200)]
clk: tests: Add tests for single parent mux

We have a few tests for a mux with a single parent, testing the case
where it used to be orphan.

Let's leverage most of the code but register the clock properly to test
a few trivial things.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add tests for uncached clock
Maxime Ripard [Fri, 8 Apr 2022 07:54:39 +0000 (09:54 +0200)]
clk: tests: Add tests for uncached clock

The clock framework supports clocks that can have their rate changed
without the kernel knowing about it using the CLK_GET_RATE_NOCACHE flag.

As its name suggests, this flag turns off the rate caching in the clock
framework, reading out the rate from the hardware any time we need to
read it.

Let's add a couple of tests to make sure it works as intended.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add reference to the orphan mux bug report
Maxime Ripard [Fri, 1 Apr 2022 12:53:49 +0000 (14:53 +0200)]
clk: tests: Add reference to the orphan mux bug report

Some more context might be useful for unit-tests covering a previously
reported bug, so let's add a link to the discussion for that bug.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: tests: Add test suites description
Maxime Ripard [Fri, 1 Apr 2022 15:08:26 +0000 (17:08 +0200)]
clk: tests: Add test suites description

We start to have a few test suites, and we'll add more, so it will get
pretty confusing to figure out what is supposed to be tested in what
suite.

Let's add some comments to explain what setup they create, and what we
should be testing in every suite.

Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Drop the rate range on clk_put()
Maxime Ripard [Fri, 25 Mar 2022 16:11:44 +0000 (17:11 +0100)]
clk: Drop the rate range on clk_put()

When clk_put() is called we don't make another clk_set_rate() call to
re-evaluate the rate boundaries. This is unlike clk_set_rate_range()
that evaluates the rate again each time it is called.

However, clk_put() is essentially equivalent to clk_set_rate_range()
since after clk_put() completes the consumer's boundaries shouldn't be
enforced anymore.

Let's add a call to clk_set_rate_range() in clk_put() to make sure those
rate boundaries are dropped and the clock provider drivers can react. In
order to be as non-intrusive as possible, we'll just make that call if
the clock had non-default boundaries.

Also add a few tests to make sure this case is covered.

Fixes: c80ac50cbb37 ("clk: Always set the rate on clk_set_range_rate")
Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> # imx8mp
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> # exynos4210, meson g12b
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: test: Test clk_set_rate_range on orphan mux
Maxime Ripard [Fri, 25 Mar 2022 10:27:28 +0000 (11:27 +0100)]
clk: test: Test clk_set_rate_range on orphan mux

A bug recently affected the Tegra30 where calling clk_set_rate_range()
on a clock would make it change its rate to the minimum.

This was due to the clock in question being a mux that was orphan at
registration, which lead to the clk_core req_rate being 0, and the
clk_set_rate_range() function then calling clk_set_rate() with req_rate,
effectively making that clock running at the minimum rate allowed, even
though the initial rate was within that range.

Make a test suite to create a mux initially orphan, and then make sure
that if our clock rate was initially within a given range, then
enforcing that range won't affect it.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Add clk_drop_range
Maxime Ripard [Mon, 17 Jan 2022 15:43:52 +0000 (16:43 +0100)]
clk: Add clk_drop_range

In order to reset the range on a clock, we need to call
clk_set_rate_range with a minimum of 0 and a maximum of ULONG_MAX. Since
it's fairly inconvenient, let's introduce a clk_drop_range() function
that will do just this.

Suggested-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Always set the rate on clk_set_range_rate
Maxime Ripard [Mon, 17 Jan 2022 15:43:13 +0000 (16:43 +0100)]
clk: Always set the rate on clk_set_range_rate

When we change a clock minimum or maximum using clk_set_rate_range(),
clk_set_min_rate() or clk_set_max_rate(), the current code will only
trigger a new rate change if the rate is outside of the new boundaries.

However, a clock driver might want to always keep the clock rate to
one of its boundary, for example the minimum to keep the power
consumption as low as possible.

Since they don't always get called though, clock providers don't have the
opportunity to implement this behaviour.

Let's trigger a clk_set_rate() on the previous requested rate every time
clk_set_rate_range() is called. That way, providers that care about the
new boundaries have a chance to adjust the rate, while providers that
don't care about those new boundaries will return the same rate than
before, which will be ignored by clk_set_rate() and won't result in a
new rate change.

Suggested-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Use clamp instead of open-coding our own
Maxime Ripard [Mon, 17 Jan 2022 15:38:10 +0000 (16:38 +0100)]
clk: Use clamp instead of open-coding our own

The code in clk_set_rate_range() will, if the current rate is outside of
the new range, force it to the minimum or maximum.

Since it's running under the condition that the rate is either lower
than the minimum, or higher than the maximum, this is equivalent to
using clamp, while being less readable. Let's switch to using clamp
instead.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Always clamp the rounded rate
Maxime Ripard [Wed, 5 May 2021 13:35:34 +0000 (15:35 +0200)]
clk: Always clamp the rounded rate

The current core while setting the min and max rate properly in the
clk_request structure will not make sure that the requested rate is
within these boundaries, leaving it to each and every driver to make
sure it is.

It's not clear if this was on purpose or not, but this introduces some
inconsistencies within the API.

For example, a user setting a range and then calling clk_round_rate()
with a value outside of that range will get the same value back
(ignoring any driver adjustements), effectively ignoring the range that
was just set.

Another one, arguably worse, is that it also makes clk_round_rate() and
clk_set_rate() behave differently if there's a range and the rate being
used for both is outside that range. As we have seen, the rate will be
returned unchanged by clk_round_rate(), but clk_set_rate() will error
out returning -EINVAL.

Let's make sure the framework will always clamp the rate to the current
range found on the clock, which will fix both these inconsistencies.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoclk: Introduce Kunit Tests for the framework
Maxime Ripard [Wed, 19 Jan 2022 15:22:19 +0000 (16:22 +0100)]
clk: Introduce Kunit Tests for the framework

Let's test various parts of the rate-related clock API with the kunit
testing framework.

Cc: kunit-dev@googlegroups.com
Tested-by: Daniel Latypov <dlatypov@google.com>
Suggested-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
2 years agoLinux 5.15.35
Greg Kroah-Hartman [Wed, 20 Apr 2022 07:34:22 +0000 (09:34 +0200)]
Linux 5.15.35

Link: https://lore.kernel.org/r/20220418121200.312988959@linuxfoundation.org
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Link: https://lore.kernel.org/r/20220419073048.315594917@linuxfoundation.org
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: Fix UAF bugs in ax25 timers
Duoming Zhou [Fri, 15 Apr 2022 16:14:22 +0000 (19:14 +0300)]
ax25: Fix UAF bugs in ax25 timers

commit 82e31755e55fbcea6a9dfaae5fe4860ade17cbc0 upstream.

There are race conditions that may lead to UAF bugs in
ax25_heartbeat_expiry(), ax25_t1timer_expiry(), ax25_t2timer_expiry(),
ax25_t3timer_expiry() and ax25_idletimer_expiry(), when we call
ax25_release() to deallocate ax25_dev.

One of the UAF bugs caused by ax25_release() is shown below:

      (Thread 1)                    |      (Thread 2)
ax25_dev_device_up() //(1)          |
...                                 | ax25_kill_by_device()
ax25_bind()          //(2)          |
ax25_connect()                      | ...
 ax25_std_establish_data_link()     |
  ax25_start_t1timer()              | ax25_dev_device_down() //(3)
   mod_timer(&ax25->t1timer,..)     |
                                    | ax25_release()
   (wait a time)                    |  ...
                                    |  ax25_dev_put(ax25_dev) //(4)FREE
   ax25_t1timer_expiry()            |
    ax25->ax25_dev->values[..] //USE|  ...
     ...                            |

We increase the refcount of ax25_dev in position (1) and (2), and
decrease the refcount of ax25_dev in position (3) and (4).
The ax25_dev will be freed in position (4) and be used in
ax25_t1timer_expiry().

The fail log is shown below:
==============================================================

[  106.116942] BUG: KASAN: use-after-free in ax25_t1timer_expiry+0x1c/0x60
[  106.116942] Read of size 8 at addr ffff88800bda9028 by task swapper/0/0
[  106.116942] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.17.0-06123-g0905eec574
[  106.116942] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-14
[  106.116942] Call Trace:
...
[  106.116942]  ax25_t1timer_expiry+0x1c/0x60
[  106.116942]  call_timer_fn+0x122/0x3d0
[  106.116942]  __run_timers.part.0+0x3f6/0x520
[  106.116942]  run_timer_softirq+0x4f/0xb0
[  106.116942]  __do_softirq+0x1c2/0x651
...

This patch adds del_timer_sync() in ax25_release(), which could ensure
that all timers stop before we deallocate ax25_dev.

Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
[OP: backport to 5.15: adjust context]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: Fix NULL pointer dereferences in ax25 timers
Duoming Zhou [Fri, 15 Apr 2022 16:14:21 +0000 (19:14 +0300)]
ax25: Fix NULL pointer dereferences in ax25 timers

commit fc6d01ff9ef03b66d4a3a23b46fc3c3d8cf92009 upstream.

The previous commit 7ec02f5ac8a5 ("ax25: fix NPD bug in ax25_disconnect")
move ax25_disconnect into lock_sock() in order to prevent NPD bugs. But
there are race conditions that may lead to null pointer dereferences in
ax25_heartbeat_expiry(), ax25_t1timer_expiry(), ax25_t2timer_expiry(),
ax25_t3timer_expiry() and ax25_idletimer_expiry(), when we use
ax25_kill_by_device() to detach the ax25 device.

One of the race conditions that cause null pointer dereferences can be
shown as below:

      (Thread 1)                    |      (Thread 2)
ax25_connect()                      |
 ax25_std_establish_data_link()     |
  ax25_start_t1timer()              |
   mod_timer(&ax25->t1timer,..)     |
                                    | ax25_kill_by_device()
   (wait a time)                    |  ...
                                    |  s->ax25_dev = NULL; //(1)
   ax25_t1timer_expiry()            |
    ax25->ax25_dev->values[..] //(2)|  ...
     ...                            |

We set null to ax25_cb->ax25_dev in position (1) and dereference
the null pointer in position (2).

The corresponding fail log is shown below:
===============================================================
BUG: kernel NULL pointer dereference, address: 0000000000000050
CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.17.0-rc6-00794-g45690b7d0
RIP: 0010:ax25_t1timer_expiry+0x12/0x40
...
Call Trace:
 call_timer_fn+0x21/0x120
 __run_timers.part.0+0x1ca/0x250
 run_timer_softirq+0x2c/0x60
 __do_softirq+0xef/0x2f3
 irq_exit_rcu+0xb6/0x100
 sysvec_apic_timer_interrupt+0xa2/0xd0
...

This patch moves ax25_disconnect() before s->ax25_dev = NULL
and uses del_timer_sync() to delete timers in ax25_disconnect().
If ax25_disconnect() is called by ax25_kill_by_device() or
ax25->ax25_dev is NULL, the reason in ax25_disconnect() will be
equal to ENETUNREACH, it will wait all timers to stop before we
set null to s->ax25_dev in ax25_kill_by_device().

Fixes: 7ec02f5ac8a5 ("ax25: fix NPD bug in ax25_disconnect")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
[OP: backport to 5.15: adjust context]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: fix NPD bug in ax25_disconnect
Duoming Zhou [Fri, 15 Apr 2022 16:14:20 +0000 (19:14 +0300)]
ax25: fix NPD bug in ax25_disconnect

commit 7ec02f5ac8a5be5a3f20611731243dc5e1d9ba10 upstream.

The ax25_disconnect() in ax25_kill_by_device() is not
protected by any locks, thus there is a race condition
between ax25_disconnect() and ax25_destroy_socket().
when ax25->sk is assigned as NULL by ax25_destroy_socket(),
a NULL pointer dereference bug will occur if site (1) or (2)
dereferences ax25->sk.

ax25_kill_by_device()                | ax25_release()
  ax25_disconnect()                  |   ax25_destroy_socket()
    ...                              |
    if(ax25->sk != NULL)             |     ...
      ...                            |     ax25->sk = NULL;
      bh_lock_sock(ax25->sk); //(1)  |     ...
      ...                            |
      bh_unlock_sock(ax25->sk); //(2)|

This patch moves ax25_disconnect() into lock_sock(), which can
synchronize with ax25_destroy_socket() in ax25_release().

Fail log:
===============================================================
BUG: kernel NULL pointer dereference, address: 0000000000000088
...
RIP: 0010:_raw_spin_lock+0x7e/0xd0
...
Call Trace:
ax25_disconnect+0xf6/0x220
ax25_device_event+0x187/0x250
raw_notifier_call_chain+0x5e/0x70
dev_close_many+0x17d/0x230
rollback_registered_many+0x1f1/0x950
unregister_netdevice_queue+0x133/0x200
unregister_netdev+0x13/0x20
...

Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
[OP: backport to 5.15: adjust context]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: fix UAF bug in ax25_send_control()
Duoming Zhou [Fri, 15 Apr 2022 16:14:19 +0000 (19:14 +0300)]
ax25: fix UAF bug in ax25_send_control()

commit 5352a761308397a0e6250fdc629bb3f615b94747 upstream.

There are UAF bugs in ax25_send_control(), when we call ax25_release()
to deallocate ax25_dev. The possible race condition is shown below:

      (Thread 1)              |     (Thread 2)
ax25_dev_device_up() //(1)    |
                              | ax25_kill_by_device()
ax25_bind()          //(2)    |
ax25_connect()                | ...
 ax25->state = AX25_STATE_1   |
 ...                          | ax25_dev_device_down() //(3)

      (Thread 3)
ax25_release()                |
 ax25_dev_put()  //(4) FREE   |
 case AX25_STATE_1:           |
  ax25_send_control()         |
   alloc_skb()       //USE    |

The refcount of ax25_dev increases in position (1) and (2), and
decreases in position (3) and (4). The ax25_dev will be freed
before dereference sites in ax25_send_control().

The following is part of the report:

[  102.297448] BUG: KASAN: use-after-free in ax25_send_control+0x33/0x210
[  102.297448] Read of size 8 at addr ffff888009e6e408 by task ax25_close/602
[  102.297448] Call Trace:
[  102.303751]  ax25_send_control+0x33/0x210
[  102.303751]  ax25_release+0x356/0x450
[  102.305431]  __sock_release+0x6d/0x120
[  102.305431]  sock_close+0xf/0x20
[  102.305431]  __fput+0x11f/0x420
[  102.305431]  task_work_run+0x86/0xd0
[  102.307130]  get_signal+0x1075/0x1220
[  102.308253]  arch_do_signal_or_restart+0x1df/0xc00
[  102.308253]  exit_to_user_mode_prepare+0x150/0x1e0
[  102.308253]  syscall_exit_to_user_mode+0x19/0x50
[  102.308253]  do_syscall_64+0x48/0x90
[  102.308253]  entry_SYSCALL_64_after_hwframe+0x44/0xae
[  102.308253] RIP: 0033:0x405ae7

This patch defers the free operation of ax25_dev and net_device after
all corresponding dereference sites in ax25_release() to avoid UAF.

Fixes: 9fd75b66b8f6 ("ax25: Fix refcount leaks caused by ax25_cb_del()")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
[OP: backport to 5.15: adjust dev_put_track()->dev_put()]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: Fix refcount leaks caused by ax25_cb_del()
Duoming Zhou [Fri, 15 Apr 2022 16:14:18 +0000 (19:14 +0300)]
ax25: Fix refcount leaks caused by ax25_cb_del()

commit 9fd75b66b8f68498454d685dc4ba13192ae069b0 upstream.

The previous commit d01ffb9eee4a ("ax25: add refcount in ax25_dev to
avoid UAF bugs") and commit feef318c855a ("ax25: fix UAF bugs of
net_device caused by rebinding operation") increase the refcounts of
ax25_dev and net_device in ax25_bind() and decrease the matching refcounts
in ax25_kill_by_device() in order to prevent UAF bugs, but there are
reference count leaks.

The root cause of refcount leaks is shown below:

     (Thread 1)                      |      (Thread 2)
ax25_bind()                          |
 ...                                 |
 ax25_addr_ax25dev()                 |
  ax25_dev_hold()   //(1)            |
  ...                                |
 dev_hold_track()   //(2)            |
 ...                                 | ax25_destroy_socket()
                                     |  ax25_cb_del()
                                     |   ...
                                     |   hlist_del_init() //(3)
                                     |
                                     |
     (Thread 3)                      |
ax25_kill_by_device()                |
 ...                                 |
 ax25_for_each(s, &ax25_list) {      |
  if (s->ax25_dev == ax25_dev) //(4) |
   ...                               |

Firstly, we use ax25_bind() to increase the refcount of ax25_dev in
position (1) and increase the refcount of net_device in position (2).
Then, we use ax25_cb_del() invoked by ax25_destroy_socket() to delete
ax25_cb in hlist in position (3) before calling ax25_kill_by_device().
Finally, the decrements of refcounts in ax25_kill_by_device() will not
be executed, because no s->ax25_dev equals to ax25_dev in position (4).

This patch adds decrements of refcounts in ax25_release() and use
lock_sock() to do synchronization. If refcounts decrease in ax25_release(),
the decrements of refcounts in ax25_kill_by_device() will not be
executed and vice versa.

Fixes: d01ffb9eee4a ("ax25: add refcount in ax25_dev to avoid UAF bugs")
Fixes: 87563a043cef ("ax25: fix reference count leaks of ax25_dev")
Fixes: feef318c855a ("ax25: fix UAF bugs of net_device caused by rebinding operation")
Reported-by: Thomas Osterried <thomas@osterried.de>
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
[OP: backport to 5.15: adjust dev_put_track()->dev_put()]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: fix UAF bugs of net_device caused by rebinding operation
Duoming Zhou [Fri, 15 Apr 2022 16:14:17 +0000 (19:14 +0300)]
ax25: fix UAF bugs of net_device caused by rebinding operation

commit feef318c855a361a1eccd880f33e88c460eb63b4 upstream.

The ax25_kill_by_device() will set s->ax25_dev = NULL and
call ax25_disconnect() to change states of ax25_cb and
sock, if we call ax25_bind() before ax25_kill_by_device().

However, if we call ax25_bind() again between the window of
ax25_kill_by_device() and ax25_dev_device_down(), the values
and states changed by ax25_kill_by_device() will be reassigned.

Finally, ax25_dev_device_down() will deallocate net_device.
If we dereference net_device in syscall functions such as
ax25_release(), ax25_sendmsg(), ax25_getsockopt(), ax25_getname()
and ax25_info_show(), a UAF bug will occur.

One of the possible race conditions is shown below:

      (USE)                   |      (FREE)
ax25_bind()                   |
                              |  ax25_kill_by_device()
ax25_bind()                   |
ax25_connect()                |    ...
                              |  ax25_dev_device_down()
                              |    ...
                              |    dev_put_track(dev, ...) //FREE
ax25_release()                |    ...
  ax25_send_control()         |
    alloc_skb()      //USE    |

the corresponding fail log is shown below:
===============================================================
BUG: KASAN: use-after-free in ax25_send_control+0x43/0x210
...
Call Trace:
  ...
  ax25_send_control+0x43/0x210
  ax25_release+0x2db/0x3b0
  __sock_release+0x6d/0x120
  sock_close+0xf/0x20
  __fput+0x11f/0x420
  ...
Allocated by task 1283:
  ...
  __kasan_kmalloc+0x81/0xa0
  alloc_netdev_mqs+0x5a/0x680
  mkiss_open+0x6c/0x380
  tty_ldisc_open+0x55/0x90
  ...
Freed by task 1969:
  ...
  kfree+0xa3/0x2c0
  device_release+0x54/0xe0
  kobject_put+0xa5/0x120
  tty_ldisc_kill+0x3e/0x80
  ...

In order to fix these UAF bugs caused by rebinding operation,
this patch adds dev_hold_track() into ax25_bind() and
corresponding dev_put_track() into ax25_kill_by_device().

Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
[OP: backport to 5.15: adjust dev_put_track()->dev_put() and
dev_hold_track()->dev_hold()]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: fix reference count leaks of ax25_dev
Duoming Zhou [Fri, 15 Apr 2022 16:14:16 +0000 (19:14 +0300)]
ax25: fix reference count leaks of ax25_dev

commit 87563a043cef044fed5db7967a75741cc16ad2b1 upstream.

The previous commit d01ffb9eee4a ("ax25: add refcount in ax25_dev
to avoid UAF bugs") introduces refcount into ax25_dev, but there
are reference leak paths in ax25_ctl_ioctl(), ax25_fwd_ioctl(),
ax25_rt_add(), ax25_rt_del() and ax25_rt_opt().

This patch uses ax25_dev_put() and adjusts the position of
ax25_addr_ax25dev() to fix reference cout leaks of ax25_dev.

Fixes: d01ffb9eee4a ("ax25: add refcount in ax25_dev to avoid UAF bugs")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/20220203150811.42256-1-duoming@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[OP: backport to 5.15: adjust context]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoax25: add refcount in ax25_dev to avoid UAF bugs
Duoming Zhou [Fri, 15 Apr 2022 16:14:15 +0000 (19:14 +0300)]
ax25: add refcount in ax25_dev to avoid UAF bugs

commit d01ffb9eee4af165d83b08dd73ebdf9fe94a519b upstream.

If we dereference ax25_dev after we call kfree(ax25_dev) in
ax25_dev_device_down(), it will lead to concurrency UAF bugs.
There are eight syscall functions suffer from UAF bugs, include
ax25_bind(), ax25_release(), ax25_connect(), ax25_ioctl(),
ax25_getname(), ax25_sendmsg(), ax25_getsockopt() and
ax25_info_show().

One of the concurrency UAF can be shown as below:

  (USE)                       |    (FREE)
                              |  ax25_device_event
                              |    ax25_dev_device_down
ax25_bind                     |    ...
  ...                         |      kfree(ax25_dev)
  ax25_fillin_cb()            |    ...
    ax25_fillin_cb_from_dev() |
  ...                         |

The root cause of UAF bugs is that kfree(ax25_dev) in
ax25_dev_device_down() is not protected by any locks.
When ax25_dev, which there are still pointers point to,
is released, the concurrency UAF bug will happen.

This patch introduces refcount into ax25_dev in order to
guarantee that there are no pointers point to it when ax25_dev
is released.

Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
[OP: backport to 5.15: adjusted context]
Signed-off-by: Ovidiu Panait <ovidiu.panait@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agocpufreq: intel_pstate: ITMT support for overclocked system
Srinivas Pandruvada [Fri, 19 Nov 2021 05:18:01 +0000 (21:18 -0800)]
cpufreq: intel_pstate: ITMT support for overclocked system

commit 03c83982a0278207709143ba78c5a470179febee upstream.

On systems with overclocking enabled, CPPC Highest Performance can be
hard coded to 0xff. In this case even if we have cores with different
highest performance, ITMT can't be enabled as the current implementation
depends on CPPC Highest Performance.

On such systems we can use MSR_HWP_CAPABILITIES maximum performance field
when CPPC.Highest Performance is 0xff.

Due to legacy reasons, we can't solely depend on MSR_HWP_CAPABILITIES as
in some older systems CPPC Highest Performance is the only way to identify
different performing cores.

Reported-by: Michael Larabel <Michael@MichaelLarabel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Tested-by: Michael Larabel <Michael@MichaelLarabel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Dimitri John Ledkov <dimitri.ledkov@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agonet: ipa: fix a build dependency
Alex Elder [Fri, 25 Feb 2022 20:15:30 +0000 (14:15 -0600)]
net: ipa: fix a build dependency

commit caef14b7530c065fb85d54492768fa48fdb5093e upstream.

An IPA build problem arose in the linux-next tree the other day.
The problem is that a recent commit adds a new dependency on some
code, and the Kconfig file for IPA doesn't reflect that dependency.
As a result, some configurations can fail to build (particularly
when COMPILE_TEST is enabled).

The recent patch adds calls to qmp_get(), qmp_put(), and qmp_send(),
and those are built based on the QCOM_AOSS_QMP config option.  If
that symbol is not defined, stubs are defined, so we just need to
ensure QCOM_AOSS_QMP is compatible with QCOM_IPA, or it's not
defined.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Fixes: 34a081761e4e3 ("net: ipa: request IPA register values be retained")
Signed-off-by: Alex Elder <elder@linaro.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agosoc: qcom: aoss: Fix missing put_device call in qmp_get
Miaoqian Lin [Sat, 8 Jan 2022 09:59:31 +0000 (09:59 +0000)]
soc: qcom: aoss: Fix missing put_device call in qmp_get

commit 4b41a9d0fe3db5f91078a380f62f0572c3ecf2dd upstream.

The reference taken by 'of_find_device_by_node()' must be released when
not needed anymore.
Add the corresponding 'put_device()' in the error handling paths.

Fixes: 8c75d585b931 ("soc: qcom: aoss: Expose send for generic usecase")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20220108095931.21527-1-linmq006@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agocpu/hotplug: Remove the 'cpu' member of cpuhp_cpu_state
Steven Price [Mon, 11 Apr 2022 15:22:32 +0000 (16:22 +0100)]
cpu/hotplug: Remove the 'cpu' member of cpuhp_cpu_state

commit b7ba6d8dc3569e49800ef0136799f26f43e237e8 upstream.

Currently the setting of the 'cpu' member of struct cpuhp_cpu_state in
cpuhp_create() is too late as it is used earlier in _cpu_up().

If kzalloc_node() in __smpboot_create_thread() fails then the rollback will
be done with st->cpu==0 causing CPU0 to be erroneously set to be dying,
causing the scheduler to get mightily confused and throw its toys out of
the pram.

However the cpu number is actually available directly, so simply remove
the 'cpu' member and avoid the problem in the first place.

Fixes: 2ea46c6fc945 ("cpumask/hotplug: Fix cpu_dying() state tracking")
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20220411152233.474129-2-steven.price@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agodrm/i915: Sunset igpu legacy mmap support based on GRAPHICS_VER_FULL
Matt Roper [Thu, 7 Apr 2022 16:18:39 +0000 (09:18 -0700)]
drm/i915: Sunset igpu legacy mmap support based on GRAPHICS_VER_FULL

commit 1acb34e7dd7720a1fff00cbd4d000ec3219dc9d6 upstream.

The intent of the version check in the mmap ioctl was to maintain
support for existing platforms (i.e., ADL/RPL and earlier), but drop
support on all future igpu platforms.  As we've seen on the dgpu side,
the hardware teams are using a more fine-grained numbering system for IP
version numbers these days, so it's possible the version number
associated with our next igpu could be some form of "12.xx" rather than
13 or higher.  Comparing against the full ver.release number will ensure
the intent of the check is maintained no matter what numbering the
hardware teams settle on.

Fixes: d3f3baa3562a ("drm/i915: Reinstate the mmap ioctl for some platforms")
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220407161839.1073443-1-matthew.d.roper@intel.com
(cherry picked from commit 8e7e5c077cd57ee9a36d58c65f07257dc49a88d5)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agodma-direct: avoid redundant memory sync for swiotlb
Chao Gao [Wed, 13 Apr 2022 06:32:22 +0000 (08:32 +0200)]
dma-direct: avoid redundant memory sync for swiotlb

commit 9e02977bfad006af328add9434c8bffa40e053bb upstream.

When we looked into FIO performance with swiotlb enabled in VM, we found
swiotlb_bounce() is always called one more time than expected for each DMA
read request.

It turns out that the bounce buffer is copied to original DMA buffer twice
after the completion of a DMA request (one is done by in
dma_direct_sync_single_for_cpu(), the other by swiotlb_tbl_unmap_single()).
But the content in bounce buffer actually doesn't change between the two
rounds of copy. So, one round of copy is redundant.

Pass DMA_ATTR_SKIP_CPU_SYNC flag to swiotlb_tbl_unmap_single() to
skip the memory copy in it.

This fix increases FIO 64KB sequential read throughput in a guest with
swiotlb=force by 5.6%.

Fixes: 55897af63091 ("dma-direct: merge swiotlb_dma_ops into the dma_direct code")
Reported-by: Wang Zhaoyang1 <zhaoyang1.wang@intel.com>
Reported-by: Gao Liang <liang.gao@intel.com>
Signed-off-by: Chao Gao <chao.gao@intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agotimers: Fix warning condition in __run_timers()
Anna-Maria Behnsen [Tue, 5 Apr 2022 19:17:32 +0000 (21:17 +0200)]
timers: Fix warning condition in __run_timers()

commit c54bc0fc84214b203f7a0ebfd1bd308ce2abe920 upstream.

When the timer base is empty, base::next_expiry is set to base::clk +
NEXT_TIMER_MAX_DELTA and base::next_expiry_recalc is false. When no timer
is queued until jiffies reaches base::next_expiry value, the warning for
not finding any expired timer and base::next_expiry_recalc is false in
__run_timers() triggers.

To prevent triggering the warning in this valid scenario
base::timers_pending needs to be added to the warning condition.

Fixes: 31cd0e119d50 ("timers: Recalculate next timer interrupt only when necessary")
Reported-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/20220405191732.7438-3-anna-maria@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agodt-bindings: net: snps: remove duplicate name
Dongjin Yang [Mon, 4 Apr 2022 02:28:57 +0000 (11:28 +0900)]
dt-bindings: net: snps: remove duplicate name

commit ce8b3ad1071b764e963d9b08ac34ffddddf12da6 upstream.

snps,dwmac has duplicated name for loongson,ls2k-dwmac and
loongson,ls7a-dwmac.

Signed-off-by: Dongjin Yang <dj76.yang@samsung.com>
Fixes: 68277749a013 ("dt-bindings: dwmac: Add bindings for new Loongson SoC and bridge chip")
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220404022857epcms1p6e6af1a6a86569f339e50c318abde7d3c@epcms1p6
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoi2c: pasemi: Wait for write xfers to finish
Martin Povišer [Tue, 29 Mar 2022 18:38:17 +0000 (20:38 +0200)]
i2c: pasemi: Wait for write xfers to finish

commit bd8963e602c77adc76dbbbfc3417c3cf14fed76b upstream.

Wait for completion of write transfers before returning from the driver.
At first sight it may seem advantageous to leave write transfers queued
for the controller to carry out on its own time, but there's a couple of
issues with it:

 * Driver doesn't check for FIFO space.

 * The queued writes can complete while the driver is in its I2C read
   transfer path which means it will get confused by the raising of
   XEN (the 'transaction ended' signal). This can cause a spurious
   ENODATA error due to premature reading of the MRXFIFO register.

Adding the wait fixes some unreliability issues with the driver. There's
some efficiency cost to it (especially with pasemi_smb_waitready doing
its polling), but that will be alleviated once the driver receives
interrupt support.

Fixes: beb58aa39e6e ("i2c: PA Semi SMBus driver")
Signed-off-by: Martin Povišer <povik+lin@cutebit.org>
Reviewed-by: Sven Peter <sven@svenpeter.dev>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agosmp: Fix offline cpu check in flush_smp_call_function_queue()
Nadav Amit [Sat, 19 Mar 2022 07:20:15 +0000 (00:20 -0700)]
smp: Fix offline cpu check in flush_smp_call_function_queue()

commit 9e949a3886356fe9112c6f6f34a6e23d1d35407f upstream.

The check in flush_smp_call_function_queue() for callbacks that are sent
to offline CPUs currently checks whether the queue is empty.

However, flush_smp_call_function_queue() has just deleted all the
callbacks from the queue and moved all the entries into a local list.
This checks would only be positive if some callbacks were added in the
short time after llist_del_all() was called. This does not seem to be
the intention of this check.

Change the check to look at the local list to which the entries were
moved instead of the queue from which all the callbacks were just
removed.

Fixes: 8d056c48e4862 ("CPU hotplug, smp: flush any pending IPI callbacks before CPU offline")
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20220319072015.1495036-1-namit@vmware.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoi2c: dev: check return value when calling dev_set_name()
Andy Shevchenko [Mon, 11 Apr 2022 18:07:51 +0000 (21:07 +0300)]
i2c: dev: check return value when calling dev_set_name()

commit 993eb48fa199b5f476df8204e652eff63dd19361 upstream.

If dev_set_name() fails, the dev_name() is null, check the return
value of dev_set_name() to avoid the null-ptr-deref.

Fixes: 1413ef638aba ("i2c: dev: Fix the race between the release of i2c_dev and cdev")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agodm integrity: fix memory corruption when tag_size is less than digest size
Mikulas Patocka [Sun, 3 Apr 2022 18:38:22 +0000 (14:38 -0400)]
dm integrity: fix memory corruption when tag_size is less than digest size

commit 08c1af8f1c13bbf210f1760132f4df24d0ed46d6 upstream.

It is possible to set up dm-integrity in such a way that the
"tag_size" parameter is less than the actual digest size. In this
situation, a part of the digest beyond tag_size is ignored.

In this case, dm-integrity would write beyond the end of the
ic->recalc_tags array and corrupt memory. The corruption happened in
integrity_recalc->integrity_sector_checksum->crypto_shash_final.

Fix this corruption by increasing the tags array so that it has enough
padding at the end to accomodate the loop in integrity_recalc() being
able to write a full digest size for the last member of the tags
array.

Cc: stable@vger.kernel.org # v4.19+
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoARM: davinci: da850-evm: Avoid NULL pointer dereference
Nathan Chancellor [Thu, 23 Dec 2021 22:21:41 +0000 (15:21 -0700)]
ARM: davinci: da850-evm: Avoid NULL pointer dereference

commit 83a1cde5c74bfb44b49cb2a940d044bb2380f4ea upstream.

With newer versions of GCC, there is a panic in da850_evm_config_emac()
when booting multi_v5_defconfig in QEMU under the palmetto-bmc machine:

Unable to handle kernel NULL pointer dereference at virtual address 00000020
pgd = (ptrval)
[00000020] *pgd=00000000
Internal error: Oops: 5 [#1] PREEMPT ARM
Modules linked in:
CPU: 0 PID: 1 Comm: swapper Not tainted 5.15.0 #1
Hardware name: Generic DT based system
PC is at da850_evm_config_emac+0x1c/0x120
LR is at do_one_initcall+0x50/0x1e0

The emac_pdata pointer in soc_info is NULL because davinci_soc_info only
gets populated on davinci machines but da850_evm_config_emac() is called
on all machines via device_initcall().

Move the rmii_en assignment below the machine check so that it is only
dereferenced when running on a supported SoC.

Fixes: bae105879f2f ("davinci: DA850/OMAP-L138 EVM: implement autodetect of RMII PHY")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Bartosz Golaszewski <brgl@bgdev.pl>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/YcS4xVWs6bQlQSPC@archlinux-ax161/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agotick/nohz: Use WARN_ON_ONCE() to prevent console saturation
Paul Gortmaker [Mon, 6 Dec 2021 14:59:50 +0000 (09:59 -0500)]
tick/nohz: Use WARN_ON_ONCE() to prevent console saturation

commit 40e97e42961f8c6cc7bd5fe67cc18417e02d78f1 upstream.

While running some testing on code that happened to allow the variable
tick_nohz_full_running to get set but with no "possible" NOHZ cores to
back up that setting, this warning triggered:

        if (unlikely(tick_do_timer_cpu == TICK_DO_TIMER_NONE))
                WARN_ON(tick_nohz_full_running);

The console was overwhemled with an endless stream of one WARN per tick
per core and there was no way to even see what was going on w/o using a
serial console to capture it and then trace it back to this.

Change it to WARN_ON_ONCE().

Fixes: 08ae95f4fd3b ("nohz_full: Allow the boot CPU to be nohz_full")
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20211206145950.10927-3-paul.gortmaker@windriver.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agogenirq/affinity: Consider that CPUs on nodes can be unbalanced
Rei Yamamoto [Thu, 31 Mar 2022 00:33:09 +0000 (09:33 +0900)]
genirq/affinity: Consider that CPUs on nodes can be unbalanced

commit 08d835dff916bfe8f45acc7b92c7af6c4081c8a7 upstream.

If CPUs on a node are offline at boot time, the number of nodes is
different when building affinity masks for present cpus and when building
affinity masks for possible cpus. This causes the following problem:

In the case that the number of vectors is less than the number of nodes
there are cases where bits of masks for present cpus are overwritten when
building masks for possible cpus.

Fix this by excluding CPUs, which are not part of the current build mask
(present/possible).

[ tglx: Massaged changelog and added comment ]

Fixes: b82592199032 ("genirq/affinity: Spread IRQs to all available NUMA nodes")
Signed-off-by: Rei Yamamoto <yamamoto.rei@jp.fujitsu.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220331003309.10891-1-yamamoto.rei@jp.fujitsu.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agox86/tsx: Disable TSX development mode at boot
Pawan Gupta [Thu, 10 Mar 2022 22:02:09 +0000 (14:02 -0800)]
x86/tsx: Disable TSX development mode at boot

commit 400331f8ffa3bec5c561417e5eec6848464e9160 upstream.

A microcode update on some Intel processors causes all TSX transactions
to always abort by default[*]. Microcode also added functionality to
re-enable TSX for development purposes. With this microcode loaded, if
tsx=on was passed on the cmdline, and TSX development mode was already
enabled before the kernel boot, it may make the system vulnerable to TSX
Asynchronous Abort (TAA).

To be on safer side, unconditionally disable TSX development mode during
boot. If a viable use case appears, this can be revisited later.

  [*]: Intel TSX Disable Update for Selected Processors, doc ID: 643557

  [ bp: Drop unstable web link, massage heavily. ]

Suggested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Suggested-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Tested-by: Neelima Krishnan <neelima.krishnan@intel.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/347bd844da3a333a9793c6687d4e4eb3b2419a3e.1646943780.git.pawan.kumar.gupta@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agox86/tsx: Use MSR_TSX_CTRL to clear CPUID bits
Pawan Gupta [Thu, 10 Mar 2022 22:00:59 +0000 (14:00 -0800)]
x86/tsx: Use MSR_TSX_CTRL to clear CPUID bits

commit 258f3b8c3210b03386e4ad92b4bd8652b5c1beb3 upstream.

tsx_clear_cpuid() uses MSR_TSX_FORCE_ABORT to clear CPUID.RTM and
CPUID.HLE. Not all CPUs support MSR_TSX_FORCE_ABORT, alternatively use
MSR_IA32_TSX_CTRL when supported.

  [ bp: Document how and why TSX gets disabled. ]

Fixes: 293649307ef9 ("x86/tsx: Clear CPUID bits when TSX always force aborts")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Tested-by: Neelima Krishnan <neelima.krishnan@intel.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/5b323e77e251a9c8bcdda498c5cc0095be1e1d3c.1646943780.git.pawan.kumar.gupta@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agodrm/amdgpu: Enable gfxoff quirk on MacBook Pro
Tomasz Moń [Wed, 6 Apr 2022 19:49:21 +0000 (21:49 +0200)]
drm/amdgpu: Enable gfxoff quirk on MacBook Pro

commit 4593c1b6d159f1e5c35c07a7f125e79e5a864302 upstream.

Enabling gfxoff quirk results in perfectly usable graphical user
interface on MacBook Pro (15-inch, 2019) with Radeon Pro Vega 20 4 GB.

Without the quirk, X server is completely unusable as every few seconds
there is gpu reset due to ring gfx timeout.

Signed-off-by: Tomasz Moń <desowin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agodrm/amd/display: don't ignore alpha property on pre-multiplied mode
Melissa Wen [Tue, 29 Mar 2022 20:18:35 +0000 (19:18 -0100)]
drm/amd/display: don't ignore alpha property on pre-multiplied mode

commit e4f1541caf60fcbe5a59e9d25805c0b5865e546a upstream.

"Pre-multiplied" is the default pixel blend mode for KMS/DRM, as
documented in supported_modes of drm_plane_create_blend_mode_property():
https://cgit.freedesktop.org/drm/drm-misc/tree/drivers/gpu/drm/drm_blend.c

In this mode, both 'pixel alpha' and 'plane alpha' participate in the
calculation, as described by the pixel blend mode formula in KMS/DRM
documentation:

out.rgb = plane_alpha * fg.rgb +
          (1 - (plane_alpha * fg.alpha)) * bg.rgb

Considering the blend config mechanisms we have in the driver so far,
the alpha mode that better fits this blend mode is the
_PER_PIXEL_ALPHA_COMBINED_GLOBAL_GAIN, where the value for global_gain
is the plane alpha (global_alpha).

With this change, alpha property stops to be ignored. It also addresses
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1734

v2:
 * keep the 8-bit value for global_alpha_value (Nicholas)
 * correct the logical ordering for combined global gain (Nicholas)
 * apply to dcn10 too (Nicholas)

Signed-off-by: Melissa Wen <mwen@igalia.com>
Tested-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Tested-by: Simon Ser <contact@emersion.fr>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoipv6: fix panic when forwarding a pkt with no in6 dev
Nicolas Dichtel [Fri, 8 Apr 2022 14:03:42 +0000 (16:03 +0200)]
ipv6: fix panic when forwarding a pkt with no in6 dev

commit e3fa461d8b0e185b7da8a101fe94dfe6dd500ac0 upstream.

kongweibin reported a kernel panic in ip6_forward() when input interface
has no in6 dev associated.

The following tc commands were used to reproduce this panic:
tc qdisc del dev vxlan100 root
tc qdisc add dev vxlan100 root netem corrupt 5%

CC: stable@vger.kernel.org
Fixes: ccd27f05ae7b ("ipv6: fix 'disable_policy' for fwd packets")
Reported-by: kongweibin <kongweibin2@huawei.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agonl80211: correctly check NL80211_ATTR_REG_ALPHA2 size
Johannes Berg [Mon, 11 Apr 2022 09:42:03 +0000 (11:42 +0200)]
nl80211: correctly check NL80211_ATTR_REG_ALPHA2 size

commit 6624bb34b4eb19f715db9908cca00122748765d7 upstream.

We need this to be at least two bytes, so we can access
alpha2[0] and alpha2[1]. It may be three in case some
userspace used NUL-termination since it was NLA_STRING
(and we also push it out with NUL-termination).

Cc: stable@vger.kernel.org
Reported-by: Lee Jones <lee.jones@linaro.org>
Link: https://lore.kernel.org/r/20220411114201.fd4a31f06541.Ie7ff4be2cf348d8cc28ed0d626fc54becf7ea799@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoALSA: pcm: Test for "silence" field in struct "pcm_format_data"
Fabio M. De Francesco [Sat, 9 Apr 2022 01:26:55 +0000 (03:26 +0200)]
ALSA: pcm: Test for "silence" field in struct "pcm_format_data"

commit 2f7a26abb8241a0208c68d22815aa247c5ddacab upstream.

Syzbot reports "KASAN: null-ptr-deref Write in
snd_pcm_format_set_silence".[1]

It is due to missing validation of the "silence" field of struct
"pcm_format_data" in "pcm_formats" array.

Add a test for valid "pat" and, if it is not so, return -EINVAL.

[1] https://lore.kernel.org/lkml/000000000000d188ef05dc2c7279@google.com/

Reported-and-tested-by: syzbot+205eb15961852c2c5974@syzkaller.appspotmail.com
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220409012655.9399-1-fmdefrancesco@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoALSA: hda/realtek: add quirk for Lenovo Thinkpad X12 speakers
Tao Jin [Sat, 9 Apr 2022 22:44:24 +0000 (18:44 -0400)]
ALSA: hda/realtek: add quirk for Lenovo Thinkpad X12 speakers

commit 264fb03497ec1c7841bba872571bcd11beed57a7 upstream.

For this specific device on Lenovo Thinkpad X12 tablet, the verbs were
dumped by qemu running a guest OS that init this codec properly.
After studying the dump, it turns out that
the same quirk used by the other Lenovo devices can be reused.

The patch was tested working against the mainline kernel.

Cc: <stable@vger.kernel.org>
Signed-off-by: Tao Jin <tao-j@outlook.com>
Link: https://lore.kernel.org/r/CO6PR03MB6241CD73310B37858FE64C85E1E89@CO6PR03MB6241.namprd03.prod.outlook.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoALSA: hda/realtek: Add quirk for Clevo PD50PNT
Tim Crawford [Tue, 5 Apr 2022 18:20:29 +0000 (12:20 -0600)]
ALSA: hda/realtek: Add quirk for Clevo PD50PNT

commit 9eb6f5c388060d8cef3c8b616cc31b765e022359 upstream.

Fixes speaker output and headset detection on Clevo PD50PNT.

Signed-off-by: Tim Crawford <tcrawford@system76.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220405182029.27431-1-tcrawford@system76.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agobtrfs: mark resumed async balance as writing
Naohiro Aota [Tue, 29 Mar 2022 06:55:58 +0000 (15:55 +0900)]
btrfs: mark resumed async balance as writing

commit a690e5f2db4d1dca742ce734aaff9f3112d63764 upstream.

When btrfs balance is interrupted with umount, the background balance
resumes on the next mount. There is a potential deadlock with FS freezing
here like as described in commit 26559780b953 ("btrfs: zoned: mark
relocation as writing"). Mark the process as sb_writing to avoid it.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
CC: stable@vger.kernel.org # 4.9+
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agobtrfs: fix root ref counts in error handling in btrfs_get_root_ref
Jia-Ju Bai [Thu, 24 Mar 2022 13:44:54 +0000 (06:44 -0700)]
btrfs: fix root ref counts in error handling in btrfs_get_root_ref

commit 168a2f776b9762f4021421008512dd7ab7474df1 upstream.

In btrfs_get_root_ref(), when btrfs_insert_fs_root() fails,
btrfs_put_root() can happen for two reasons:

- the root already exists in the tree, in that case it returns the
  reference obtained in btrfs_lookup_fs_root()

- another error so the cleanup is done in the fail label

Calling btrfs_put_root() unconditionally would lead to double decrement
of the root reference possibly freeing it in the second case.

Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Fixes: bc44d7c4b2b1 ("btrfs: push btrfs_grab_fs_root into btrfs_get_fs_root")
CC: stable@vger.kernel.org # 5.10+
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoath9k: Fix usage of driver-private space in tx_info
Toke Høiland-Jørgensen [Mon, 4 Apr 2022 20:48:00 +0000 (22:48 +0200)]
ath9k: Fix usage of driver-private space in tx_info

commit 5a6b06f5927c940fa44026695779c30b7536474c upstream.

The ieee80211_tx_info_clear_status() helper also clears the rate counts and
the driver-private part of struct ieee80211_tx_info, so using it breaks
quite a few other things. So back out of using it, and instead define a
ath-internal helper that only clears the area between the
status_driver_data and the rates info. Combined with moving the
ath_frame_info struct to status_driver_data, this avoids clearing anything
we shouldn't be, and so we can keep the existing code for handling the rate
information.

While fixing this I also noticed that the setting of
tx_info->status.rates[tx_rateindex].count on hardware underrun errors was
always immediately overridden by the normal setting of the same fields, so
rearrange the code so that the underrun detection actually takes effect.

The new helper could be generalised to a 'memset_between()' helper, but
leave it as a driver-internal helper for now since this needs to go to
stable.

Cc: stable@vger.kernel.org
Reported-by: Peter Seiderer <ps.report@gmx.net>
Fixes: 037250f0a45c ("ath9k: Properly clear TX status area before reporting to mac80211")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Reviewed-by: Peter Seiderer <ps.report@gmx.net>
Tested-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20220404204800.2681133-1-toke@toke.dk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoath9k: Properly clear TX status area before reporting to mac80211
Toke Høiland-Jørgensen [Wed, 30 Mar 2022 16:44:09 +0000 (18:44 +0200)]
ath9k: Properly clear TX status area before reporting to mac80211

commit 037250f0a45cf9ecf5b52d4b9ff8eadeb609c800 upstream.

The ath9k driver was not properly clearing the status area in the
ieee80211_tx_info struct before reporting TX status to mac80211. Instead,
it was manually filling in fields, which meant that fields introduced later
were left as-is.

Conveniently, mac80211 actually provides a helper to zero out the status
area, so use that to make sure we zero everything.

The last commit touching the driver function writing the status information
seems to have actually been fixing an issue that was also caused by the
area being uninitialised; but it only added clearing of a single field
instead of the whole struct. That is now redundant, though, so revert that
commit and use it as a convenient Fixes tag.

Fixes: cc591d77aba1 ("ath9k: Make sure to zero status.tx_time before reporting TX status")
Reported-by: Bagas Sanjaya <bagasdotme@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20220330164409.16645-1-toke@toke.dk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agocifs: verify that tcon is valid before dereference in cifs_kill_sb
Ronnie Sahlberg [Wed, 13 Apr 2022 00:02:17 +0000 (10:02 +1000)]
cifs: verify that tcon is valid before dereference in cifs_kill_sb

commit 8b6c58458ee3206dde345fce327a4cb83e69caf9 upstream.

On umount, cifs_sb->tlink_tree might contain entries that do not represent
a valid tcon.
Check the tcon for error before we dereference it.

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Cc: stable@vger.kernel.org
Reviewed-by: Shyam Prasad N <sprasad@microsoft.com>
Reported-by: Xiaoli Feng <xifeng@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agogcc-plugins: latent_entropy: use /dev/urandom
Jason A. Donenfeld [Tue, 5 Apr 2022 22:28:15 +0000 (00:28 +0200)]
gcc-plugins: latent_entropy: use /dev/urandom

commit c40160f2998c897231f8454bf797558d30a20375 upstream.

While the latent entropy plugin mostly doesn't derive entropy from
get_random_const() for measuring the call graph, when __latent_entropy is
applied to a constant, then it's initialized statically to output from
get_random_const(). In that case, this data is derived from a 64-bit
seed, which means a buffer of 512 bits doesn't really have that amount
of compile-time entropy.

This patch fixes that shortcoming by just buffering chunks of
/dev/urandom output and doling it out as requested.

At the same time, it's important that we don't break the use of
-frandom-seed, for people who want the runtime benefits of the latent
entropy plugin, while still having compile-time determinism. In that
case, we detect whether gcc's set_random_seed() has been called by
making a call to get_random_seed(noinit=true) in the plugin init
function, which is called after set_random_seed() is called but before
anything that calls get_random_seed(noinit=false), and seeing if it's
zero or not. If it's not zero, we're in deterministic mode, and so we
just generate numbers with a basic xorshift prng.

Note that we don't detect if -frandom-seed is being used using the
documented local_tick variable, because it's assigned via:
   local_tick = (unsigned) tv.tv_sec * 1000 + tv.tv_usec / 1000;
which may well overflow and become -1 on its own, and so isn't
reliable: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105171

[kees: The 256 byte rnd_buf size was chosen based on average (250),
 median (64), and std deviation (575) bytes of used entropy for a
 defconfig x86_64 build]

Fixes: 38addce8b600 ("gcc-plugins: Add latent_entropy plugin")
Cc: stable@vger.kernel.org
Cc: PaX Team <pageexec@freemail.hu>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20220405222815.21155-1-Jason@zx2c4.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agomemory: renesas-rpc-if: fix platform-device leak in error path
Johan Hovold [Thu, 3 Mar 2022 18:06:32 +0000 (19:06 +0100)]
memory: renesas-rpc-if: fix platform-device leak in error path

commit b452dbf24d7d9a990d70118462925f6ee287d135 upstream.

Make sure to free the flash platform device in the event that
registration fails during probe.

Fixes: ca7d8b980b67 ("memory: add Renesas RPC-IF driver")
Cc: stable@vger.kernel.org # 5.8
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Johan Hovold <johan@kernel.org>
Link: https://lore.kernel.org/r/20220303180632.3194-1-johan@kernel.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoSUNRPC: Fix NFSD's request deferral on RDMA transports
Chuck Lever [Fri, 1 Apr 2022 21:08:21 +0000 (17:08 -0400)]
SUNRPC: Fix NFSD's request deferral on RDMA transports

commit 773f91b2cf3f52df0d7508fdbf60f37567cdaee4 upstream.

Trond Myklebust reports an NFSD crash in svc_rdma_sendto(). Further
investigation shows that the crash occurred while NFSD was handling
a deferred request.

This patch addresses two inter-related issues that prevent request
deferral from working correctly for RPC/RDMA requests:

1. Prevent the crash by ensuring that the original
   svc_rqst::rq_xprt_ctxt value is available when the request is
   revisited. Otherwise svc_rdma_sendto() does not have a Receive
   context available with which to construct its reply.

2. Possibly since before commit 71641d99ce03 ("svcrdma: Properly
   compute .len and .buflen for received RPC Calls"),
   svc_rdma_recvfrom() did not include the transport header in the
   returned xdr_buf. There should have been no need for svc_defer()
   and friends to save and restore that header, as of that commit.
   This issue is addressed in a backport-friendly way by simply
   having svc_rdma_recvfrom() set rq_xprt_hlen to zero
   unconditionally, just as svc_tcp_recvfrom() does. This enables
   svc_deferred_recv() to correctly reconstruct an RPC message
   received via RPC/RDMA.

Reported-by: Trond Myklebust <trondmy@hammerspace.com>
Link: https://lore.kernel.org/linux-nfs/82662b7190f26fb304eb0ab1bb04279072439d4e.camel@hammerspace.com/
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoKVM: x86/mmu: Resolve nx_huge_pages when kvm.ko is loaded
Sean Christopherson [Thu, 31 Mar 2022 22:13:59 +0000 (22:13 +0000)]
KVM: x86/mmu: Resolve nx_huge_pages when kvm.ko is loaded

commit 1d0e84806047f38027d7572adb4702ef7c09b317 upstream.

Resolve nx_huge_pages to true/false when kvm.ko is loaded, leaving it as
-1 is technically undefined behavior when its value is read out by
param_get_bool(), as boolean values are supposed to be '0' or '1'.

Alternatively, KVM could define a custom getter for the param, but the
auto value doesn't depend on the vendor module in any way, and printing
"auto" would be unnecessarily unfriendly to the user.

In addition to fixing the undefined behavior, resolving the auto value
also fixes the scenario where the auto value resolves to N and no vendor
module is loaded.  Previously, -1 would result in Y being printed even
though KVM would ultimately disable the mitigation.

Rename the existing MMU module init/exit helpers to clarify that they're
invoked with respect to the vendor module, and add comments to document
why KVM has two separate "module init" flows.

  =========================================================================
  UBSAN: invalid-load in kernel/params.c:320:33
  load of value 255 is not a valid value for type '_Bool'
  CPU: 6 PID: 892 Comm: tail Not tainted 5.17.0-rc3+ #799
  Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 0.0.0 02/06/2015
  Call Trace:
   <TASK>
   dump_stack_lvl+0x34/0x44
   ubsan_epilogue+0x5/0x40
   __ubsan_handle_load_invalid_value.cold+0x43/0x48
   param_get_bool.cold+0xf/0x14
   param_attr_show+0x55/0x80
   module_attr_show+0x1c/0x30
   sysfs_kf_seq_show+0x93/0xc0
   seq_read_iter+0x11c/0x450
   new_sync_read+0x11b/0x1a0
   vfs_read+0xf0/0x190
   ksys_read+0x5f/0xe0
   do_syscall_64+0x3b/0xc0
   entry_SYSCALL_64_after_hwframe+0x44/0xae
   </TASK>
  =========================================================================

Fixes: b8e8c8303ff2 ("kvm: mmu: ITLB_MULTIHIT mitigation")
Cc: stable@vger.kernel.org
Reported-by: Bruno Goncalves <bgoncalv@redhat.com>
Reported-by: Jan Stancek <jstancek@redhat.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220331221359.3912754-1-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agomm: kmemleak: take a full lowmem check in kmemleak_*_phys()
Patrick Wang [Fri, 15 Apr 2022 02:14:04 +0000 (19:14 -0700)]
mm: kmemleak: take a full lowmem check in kmemleak_*_phys()

commit 23c2d497de21f25898fbea70aeb292ab8acc8c94 upstream.

The kmemleak_*_phys() apis do not check the address for lowmem's min
boundary, while the caller may pass an address below lowmem, which will
trigger an oops:

  # echo scan > /sys/kernel/debug/kmemleak
  Unable to handle kernel paging request at virtual address ff5fffffffe00000
  Oops [#1]
  Modules linked in:
  CPU: 2 PID: 134 Comm: bash Not tainted 5.18.0-rc1-next-20220407 #33
  Hardware name: riscv-virtio,qemu (DT)
  epc : scan_block+0x74/0x15c
   ra : scan_block+0x72/0x15c
  epc : ffffffff801e5806 ra : ffffffff801e5804 sp : ff200000104abc30
   gp : ffffffff815cd4e8 tp : ff60000004cfa340 t0 : 0000000000000200
   t1 : 00aaaaaac23954cc t2 : 00000000000003ff s0 : ff200000104abc90
   s1 : ffffffff81b0ff28 a0 : 0000000000000000 a1 : ff5fffffffe01000
   a2 : ffffffff81b0ff28 a3 : 0000000000000002 a4 : 0000000000000001
   a5 : 0000000000000000 a6 : ff200000104abd7c a7 : 0000000000000005
   s2 : ff5fffffffe00ff9 s3 : ffffffff815cd998 s4 : ffffffff815d0e90
   s5 : ffffffff81b0ff28 s6 : 0000000000000020 s7 : ffffffff815d0eb0
   s8 : ffffffffffffffff s9 : ff5fffffffe00000 s10: ff5fffffffe01000
   s11: 0000000000000022 t3 : 00ffffffaa17db4c t4 : 000000000000000f
   t5 : 0000000000000001 t6 : 0000000000000000
  status: 0000000000000100 badaddr: ff5fffffffe00000 cause: 000000000000000d
    scan_gray_list+0x12e/0x1a6
    kmemleak_scan+0x2aa/0x57e
    kmemleak_write+0x32a/0x40c
    full_proxy_write+0x56/0x82
    vfs_write+0xa6/0x2a6
    ksys_write+0x6c/0xe2
    sys_write+0x22/0x2a
    ret_from_syscall+0x0/0x2

The callers may not quite know the actual address they pass(e.g. from
devicetree).  So the kmemleak_*_phys() apis should guarantee the address
they finally use is in lowmem range, so check the address for lowmem's
min boundary.

Link: https://lkml.kernel.org/r/20220413122925.33856-1-patrick.wang.shcn@gmail.com
Signed-off-by: Patrick Wang <patrick.wang.shcn@gmail.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agomm: fix unexpected zeroed page mapping with zram swap
Minchan Kim [Fri, 15 Apr 2022 02:13:46 +0000 (19:13 -0700)]
mm: fix unexpected zeroed page mapping with zram swap

commit e914d8f00391520ecc4495dd0ca0124538ab7119 upstream.

Two processes under CLONE_VM cloning, user process can be corrupted by
seeing zeroed page unexpectedly.

      CPU A                        CPU B

  do_swap_page                do_swap_page
  SWP_SYNCHRONOUS_IO path     SWP_SYNCHRONOUS_IO path
  swap_readpage valid data
    swap_slot_free_notify
      delete zram entry
                              swap_readpage zeroed(invalid) data
                              pte_lock
                              map the *zero data* to userspace
                              pte_unlock
  pte_lock
  if (!pte_same)
    goto out_nomap;
  pte_unlock
  return and next refault will
  read zeroed data

The swap_slot_free_notify is bogus for CLONE_VM case since it doesn't
increase the refcount of swap slot at copy_mm so it couldn't catch up
whether it's safe or not to discard data from backing device.  In the
case, only the lock it could rely on to synchronize swap slot freeing is
page table lock.  Thus, this patch gets rid of the swap_slot_free_notify
function.  With this patch, CPU A will see correct data.

      CPU A                        CPU B

  do_swap_page                do_swap_page
  SWP_SYNCHRONOUS_IO path     SWP_SYNCHRONOUS_IO path
                              swap_readpage original data
                              pte_lock
                              map the original data
                              swap_free
                                swap_range_free
                                  bd_disk->fops->swap_slot_free_notify
  swap_readpage read zeroed data
                              pte_unlock
  pte_lock
  if (!pte_same)
    goto out_nomap;
  pte_unlock
  return
  on next refault will see mapped data by CPU B

The concern of the patch would increase memory consumption since it
could keep wasted memory with compressed form in zram as well as
uncompressed form in address space.  However, most of cases of zram uses
no readahead and do_swap_page is followed by swap_free so it will free
the compressed form from in zram quickly.

Link: https://lkml.kernel.org/r/YjTVVxIAsnKAXjTd@google.com
Fixes: 0bcac06f27d7 ("mm, swap: skip swapcache for swapin of synchronous device")
Reported-by: Ivan Babrou <ivan@cloudflare.com>
Tested-by: Ivan Babrou <ivan@cloudflare.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: David Hildenbrand <david@redhat.com>
Cc: <stable@vger.kernel.org> [4.14+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agomm, page_alloc: fix build_zonerefs_node()
Juergen Gross [Fri, 15 Apr 2022 02:13:43 +0000 (19:13 -0700)]
mm, page_alloc: fix build_zonerefs_node()

commit e553f62f10d93551eb883eca227ac54d1a4fad84 upstream.

Since commit 6aa303defb74 ("mm, vmscan: only allocate and reclaim from
zones with pages managed by the buddy allocator") only zones with free
memory are included in a built zonelist.  This is problematic when e.g.
all memory of a zone has been ballooned out when zonelists are being
rebuilt.

The decision whether to rebuild the zonelists when onlining new memory
is done based on populated_zone() returning 0 for the zone the memory
will be added to.  The new zone is added to the zonelists only, if it
has free memory pages (managed_zone() returns a non-zero value) after
the memory has been onlined.  This implies, that onlining memory will
always free the added pages to the allocator immediately, but this is
not true in all cases: when e.g. running as a Xen guest the onlined new
memory will be added only to the ballooned memory list, it will be freed
only when the guest is being ballooned up afterwards.

Another problem with using managed_zone() for the decision whether a
zone is being added to the zonelists is, that a zone with all memory
used will in fact be removed from all zonelists in case the zonelists
happen to be rebuilt.

Use populated_zone() when building a zonelist as it has been done before
that commit.

There was a report that QubesOS (based on Xen) is hitting this problem.
Xen has switched to use the zone device functionality in kernel 5.9 and
QubesOS wants to use memory hotplugging for guests in order to be able
to start a guest with minimal memory and expand it as needed.  This was
the report leading to the patch.

Link: https://lkml.kernel.org/r/20220407120637.9035-1-jgross@suse.com
Fixes: 6aa303defb74 ("mm, vmscan: only allocate and reclaim from zones with pages managed by the buddy allocator")
Signed-off-by: Juergen Gross <jgross@suse.com>
Reported-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agomm/secretmem: fix panic when growing a memfd_secret
Axel Rasmussen [Fri, 15 Apr 2022 02:13:31 +0000 (19:13 -0700)]
mm/secretmem: fix panic when growing a memfd_secret

commit f9b141f93659e09a52e28791ccbaf69c273b8e92 upstream.

When one tries to grow an existing memfd_secret with ftruncate, one gets
a panic [1].  For example, doing the following reliably induces the
panic:

    fd = memfd_secret();

    ftruncate(fd, 10);
    ptr = mmap(NULL, 10, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
    strcpy(ptr, "123456789");

    munmap(ptr, 10);
    ftruncate(fd, 20);

The basic reason for this is, when we grow with ftruncate, we call down
into simple_setattr, and then truncate_inode_pages_range, and eventually
we try to zero part of the memory.  The normal truncation code does this
via the direct map (i.e., it calls page_address() and hands that to
memset()).

For memfd_secret though, we specifically don't map our pages via the
direct map (i.e.  we call set_direct_map_invalid_noflush() on every
fault).  So the address returned by page_address() isn't useful, and
when we try to memset() with it we panic.

This patch avoids the panic by implementing a custom setattr for
memfd_secret, which detects resizes specifically (setting the size for
the first time works just fine, since there are no existing pages to try
to zero), and rejects them with EINVAL.

One could argue growing should be supported, but I think that will
require a significantly more lengthy change.  So, I propose a minimal
fix for the benefit of stable kernels, and then perhaps to extend
memfd_secret to support growing in a separate patch.

[1]:

  BUG: unable to handle page fault for address: ffffa0a889277028
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD afa01067 P4D afa01067 PUD 83f909067 PMD 83f8bf067 PTE 800ffffef6d88060
  Oops: 0002 [#1] PREEMPT SMP DEBUG_PAGEALLOC PTI
  CPU: 0 PID: 281 Comm: repro Not tainted 5.17.0-dbg-DEV #1
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1 04/01/2014
  RIP: 0010:memset_erms+0x9/0x10
  Code: c1 e9 03 40 0f b6 f6 48 b8 01 01 01 01 01 01 01 01 48 0f af c6 f3 48 ab 89 d1 f3 aa 4c 89 c8 c3 90 49 89 f9 40 88 f0 48 89 d1 <f3> aa 4c 89 c8 c3 90 49 89 fa 40 0f b6 ce 48 b8 01 01 01 01 01 01
  RSP: 0018:ffffb932c09afbf0 EFLAGS: 00010246
  RAX: 0000000000000000 RBX: ffffda63c4249dc0 RCX: 0000000000000fd8
  RDX: 0000000000000fd8 RSI: 0000000000000000 RDI: ffffa0a889277028
  RBP: ffffb932c09afc00 R08: 0000000000001000 R09: ffffa0a889277028
  R10: 0000000000020023 R11: 0000000000000000 R12: ffffda63c4249dc0
  R13: ffffa0a890d70d98 R14: 0000000000000028 R15: 0000000000000fd8
  FS:  00007f7294899580(0000) GS:ffffa0af9bc00000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: ffffa0a889277028 CR3: 0000000107ef6006 CR4: 0000000000370ef0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   ? zero_user_segments+0x82/0x190
   truncate_inode_partial_folio+0xd4/0x2a0
   truncate_inode_pages_range+0x380/0x830
   truncate_setsize+0x63/0x80
   simple_setattr+0x37/0x60
   notify_change+0x3d8/0x4d0
   do_sys_ftruncate+0x162/0x1d0
   __x64_sys_ftruncate+0x1c/0x20
   do_syscall_64+0x44/0xa0
   entry_SYSCALL_64_after_hwframe+0x44/0xae
  Modules linked in: xhci_pci xhci_hcd virtio_net net_failover failover virtio_blk virtio_balloon uhci_hcd ohci_pci ohci_hcd evdev ehci_pci ehci_hcd 9pnet_virtio 9p netfs 9pnet
  CR2: ffffa0a889277028

[lkp@intel.com: secretmem_iops can be static]
Signed-off-by: kernel test robot <lkp@intel.com>
[axelrasmussen@google.com: return EINVAL]

Link: https://lkml.kernel.org/r/20220324210909.1843814-1-axelrasmussen@google.com
Link: https://lkml.kernel.org/r/20220412193023.279320-1-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Cc: kernel test robot <lkp@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agoperf/imx_ddr: Fix undefined behavior due to shift overflowing the constant
Borislav Petkov [Tue, 5 Apr 2022 15:15:15 +0000 (17:15 +0200)]
perf/imx_ddr: Fix undefined behavior due to shift overflowing the constant

[ Upstream commit d02b4dd84e1a90f7f1444d027c0289bf355b0d5a ]

Fix:

  In file included from <command-line>:0:0:
  In function ‘ddr_perf_counter_enable’,
      inlined from ‘ddr_perf_irq_handler’ at drivers/perf/fsl_imx8_ddr_perf.c:651:2:
  ././include/linux/compiler_types.h:352:38: error: call to ‘__compiletime_assert_729’ \
declared with attribute error: FIELD_PREP: mask is not constant
    _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
...

See https://lore.kernel.org/r/YkwQ6%2BtIH8GQpuct@zn.tnic for the gory
details as to why it triggers with older gccs only.

Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Frank Li <Frank.li@nxp.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Shawn Guo <shawnguo@kernel.org>
Cc: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Pengutronix Kernel Team <kernel@pengutronix.de>
Cc: Fabio Estevam <festevam@gmail.com>
Cc: NXP Linux Team <linux-imx@nxp.com>
Cc: linux-arm-kernel@lists.infradead.org
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20220405151517.29753-10-bp@alien8.de
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>