On Thu 13 Feb 18:35 PST 2020, Stephen Boyd wrote: > Quoting Bjorn Andersson (2020-02-10 16:50:53) > > diff --git a/drivers/remoteproc/qcom_pil_info.c b/drivers/remoteproc/qcom_pil_info.c [..] > > + mutex_lock(&reloc_mutex); > > + _reloc = reloc; > > + mutex_unlock(&reloc_mutex); > > Ah ok, I see that mutex is protecting the pointer used for everything. > Ignore the comment above. But also, why not have every remoteproc device > point at some imem and then search through the imem for the name? Then > we don't need this driver or a lock that synchronizes these things. > Ideally we could dedicate a place in imem for each remoteproc and not > even have to search it for the string to update. Is that possible? Then > it becomes even simpler because the DT binding can point directly at the > address to write. It's not like the various images are changing that > much to the point where this location in imem is actually changing, > right? > I will check to see if these entries needs to be packed in the beginning of the array, otherwise this sounds like a good idea to simplify things. Regards, Bjorn
