On pe, 2017-02-10 at 14:36 +, Chris Wilson wrote:
> On Fri, Feb 10, 2017 at 03:30:10PM +0200, Joonas Lahtinen wrote:
> >
> > +static unsigned long __reserve_cacheline(struct intel_guc* guc)
> > {
> > - const uint32_t cacheline_size = cache_line_size();
> > - uint32_t offset;
> > + unsig
On pe, 2017-02-10 at 16:11 +0100, Michal Wajdeczko wrote:
> On Fri, Feb 10, 2017 at 03:30:10PM +0200, Joonas Lahtinen wrote:
> >
> > Started adding proper teardown to guc_client_alloc, ended up removing
> > quite a few dead ends where errors communicating with the GuC were
> > silently ignored. Th
On pe, 2017-02-10 at 12:03 -0800, Daniele Ceraolo Spurio wrote:
>
> >
> > >
> > > +
> > > /*
> > > * Tell the GuC to allocate or deallocate a specific doorbell
> > > */
> > >
> > > -static int guc_allocate_doorbell(struct intel_guc *guc,
> > > - struct i915_guc_cli
On pe, 2017-02-10 at 11:55 -0800, Daniele Ceraolo Spurio wrote:
>
> On 10/02/17 05:30, Joonas Lahtinen wrote:
> > +static bool __test_doorbell(struct i915_guc_client *client)
> > +{
> > + return test_bit(client->doorbell_id, client->guc->doorbell_bitmap);
> > +}
>
> bikeshed: this helper is
+
/*
* Tell the GuC to allocate or deallocate a specific doorbell
*/
-static int guc_allocate_doorbell(struct intel_guc *guc,
-struct i915_guc_client *client)
+static int __create_doorbell_hw(struct i915_guc_client *client)
I would rather prefer to only ch
On 10/02/17 05:30, Joonas Lahtinen wrote:
Started adding proper teardown to guc_client_alloc, ended up removing
quite a few dead ends where errors communicating with the GuC were
silently ignored. There also seemed to be quite a few erronous
teardown actions performed in case of an error (order
On Fri, Feb 10, 2017 at 03:30:10PM +0200, Joonas Lahtinen wrote:
> Started adding proper teardown to guc_client_alloc, ended up removing
> quite a few dead ends where errors communicating with the GuC were
> silently ignored. There also seemed to be quite a few erronous
> teardown actions performed
On Fri, Feb 10, 2017 at 03:30:10PM +0200, Joonas Lahtinen wrote:
> +static unsigned long __reserve_cacheline(struct intel_guc* guc)
> {
> - const uint32_t cacheline_size = cache_line_size();
> - uint32_t offset;
> + unsigned long offset;
>
> /* Doorbell uses a single cache line
Started adding proper teardown to guc_client_alloc, ended up removing
quite a few dead ends where errors communicating with the GuC were
silently ignored. There also seemed to be quite a few erronous
teardown actions performed in case of an error (ordering wrong).
Signed-off-by: Joonas Lahtinen
C
