On 05/17/16 14:08, Peter Ujfalusi wrote: > On 05/17/16 14:02, Geert Uytterhoeven wrote: >> On Tue, May 17, 2016 at 12:55 PM, Geert Uytterhoeven >> <geert@xxxxxxxxxxxxxx> wrote: >>> With today's sound-asoc/for-next, I'm seeing on r8a7791/koelsch and >>> sh73a0/kzm9g: >>> >>> kernel BUG at drivers/base/regmap/regcache.c:347! >>> Internal error: Oops - BUG: 0 [#1] SMP ARM >>> Modules linked in: >>> CPU: 0 PID: 20 Comm: kworker/0:1 Not tainted >>> 4.6.0-koelsch-07135-g1481dd77fa7d01cb #2621 >>> Hardware name: Generic R8A7791 (Flattened Device Tree) >>> Workqueue: events soc_resume_deferred >>> task: ef512100 ti: ef5a6000 task.ti: ef5a6000 >>> PC is at regcache_sync+0xc/0x118 >>> LR is at ak4642_resume+0x20/0x28 >> >> This is during resume from s2ram. > > if you revert the patch does the ak4642_resume() is called at all, I think not. > But in the driver does not implement the .suspend callback, probably this > never worked correctly in ak4642? The reason is different, the bug is caused by: BUG_ON(!map->cache_ops); in regcache.c None of the regmap_config have set the .cache_type, so they do not have cache, so sync is not available for them -> regmap_sync() BUGs out. It can be fixed by either setting .cache_type for the regmap_config structures or by removing the resume callback, as it is broken. -- Péter
