misc: nsm: mark device ready before registering the hwrng and misc device

nsm_device_probe() registers an hwrng provider and /dev/nsm.
hwrng_register() can make this the current RNG, and the hwrng fill kthread
finds it through the RCU-protected get_current_rng() rather than under the
rng_mutex that hwrng_register() holds, so it can be in nsm_rng_read()
offering buffers and kicking the queue before probe has returned.

The driver never calls virtio_device_ready(), so DRIVER_OK is set on its
behalf by virtio_dev_probe(), only once probe has returned. The virtio
spec is explicit about that ordering in 3.1 Device Initialization:
  |  The driver MUST NOT send any buffer available notifications to the
  |  device before setting DRIVER_OK.

A device that waits for DRIVER_OK before servicing the queue therefore
leaves that read unanswered.

Set DRIVER_OK once the virtqueue exists and before either interface is
registered, so that the queue is usable by the time anything that can reach
it is published; commit f5866db64f34 ("virtio_console: enable VQs early")
and commit 1d774589f924 ("i2c: virtio: mark device ready before registering
the adapter") do the same for the same reason. The err_hwrng and err_misc
paths then unwind a device that is live, so reset it before deleting the
virtqueue - the core only adds VIRTIO_CONFIG_S_FAILED to the status when
probe fails, it does not reset.

Fixes: b9873755a6c8 ("misc: Add Nitro Secure Module driver")
Signed-off-by: Jasper Wise <jaspwise@amazon.co.uk>
Link: https://patch.msgid.link/20260813083752.581981-1-jaspwise@amazon.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

0 files changed, 0 insertions, 0 deletions