Testing for regressions in Media Controller API register, ioctl, syscall,
and unregister paths. There have a few problems that result in user-after
free on media_device, media_devnode, and cdev pointers when the driver is
unbound while ioctl is in progress.
Test Procedure:
Run bin/unbind loop while ioctls are in progress.
Run rmmod and modprobe.
Disconnect the device.
Setup:
Build media_device_test
cd tools/testing/selftests/media_tests
make
Regressions test for cdev user-after free error on /dev/mediaX when driver
is unbound:
Start media_device_test to regression test media devnode dynamic alloc
and cdev user-after-free fixes. This opens media dev files and sits in
a loop running media ioctl MEDIA_IOC_DEVICE_INFO command once every 10
seconds. The idea is when device file goes away, media devnode and cdev
should stick around until this test exits.
The test for a random number of iterations or until user kills it with a
sleep 10 in between the ioctl calls.
sudo ./media_device_test -d /dev/mediaX
Regression test for media_devnode unregister race with ioctl_syscall:
Start 6 open_loop_test.sh tests with different /dev/mediaX files. When
device file goes away after unbind, device file name changes. Start the
test with possible device names. If we start with /dev/media0 for example,
after unbind, /dev/media1 or /dev/media2 could get created. The idea is
keep ioctls going while bind/unbind runs.
Copy bind_unbind_sample.txt and make changes to specify the driver name
and number to run bind and unbind. Start the bind_unbind.sh
Run dmesg looking for any user-after free errors or mutex lock errors.
>refslogtreecommitdiff|
|
rt2x00: fix clk_get call
clk_get() takes two arguments and might return ERR_PTR(), so we
have to nullify pointer on that case, to do not break further call
to clk_get_rate().
Reported-by: Felix Fietkau <nbd@nbd.name>
Fixes: 34db70b92fae ("rt2x00: add copy of clk for soc devices")
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>