(I think Dmitry noticed the same while I wrote this.)
On Mon, Oct 10, 2016 at 04:47:08PM -0700, Brian Norris wrote:
> [*] The other cases are in error handling cases. I guess I should make
> sure those didn't race too...
Ah, well I think I missed one case:
For the async FW request code path, the callback mwifiex_fw_dpc() can
fail to load FW, and so it unwinds with:
if (adapter->if_ops.unregister_dev)
adapter->if_ops.unregister_dev(adapter);
This all happens after probe() is finished, so we can have a race on
card->adapter being read in suspend() and written in the $subject patch.
Can I revoke my Reviewed-by? (Or make it Reviewed-and-found-wanting-by?)
Brian
