8 Sep
2021
8 Sep
'21
3:29 p.m.
On 08/09/2021 13:26, Mark Brown wrote:
On Wed, Sep 08, 2021 at 12:28:05PM +0100, Srinivas Kandagatla wrote:
On 07/09/2021 16:04, Pierre-Louis Bossart wrote:
+ spin_lock(&apm->lock); + idr_alloc(&apm->graph_idr, graph, graph_id, + graph_id + 1, GFP_ATOMIC);
does this need to be ATOMIC?
We are inside spinlock.
but this is not used in an interrupt handler or anything that isn't in a process context, is it?
No, it is in process context only.
So why is it a spinlock then?
Only reason I ended-up with spin locks is because am guessing that busy loops in this case are not going to be for very long as its mostly dealing with idr resource.
On the other hand, I don't have any issues in moving this to mutex.
--srini