Hi Bryan, On 10/25/21 08:24, Bryan O'Donoghue wrote:
+ struct mutex sync_lock;Why have an additional mutex, will the existing core::lock not do ?
I wanted to reuse it, but the core::lock in used internally in hfi and it will deadlock there if we use that one here.
shouldn't this be an -EPROBE_DEFER i.e. CORE_INIT hasn't completed/run yet so defer until it does.This fragment here looks racy to me without a DEFER above ? drivers/media/platform/qcom/venus/core.c::venus_probe()
No, we want a hard stop here. At this point the venus_core probe() has finished and it failed. Returning -EPROBE_DEFER here will just cause it to loop infinitely. -- Thanks, Tadeusz
Attachment:
OpenPGP_signature
Description: OpenPGP digital signature
