On 12.03.19 09:49, Baolin Wang wrote: > I still do not think the new API is suitable for this case. Since we > can have optional multiple IO resources, so the original code will not > return errors if we did not get the IO resources, but we must cast > errors if we failed to do ioremap. But you ignore the errors of > ioremap, which is not good. hmm, maybe we can differenciate on the error code ? patch up devm_platform_ioremap_resource() so it returns -ENOENT if there's no such resource defined. I suppose, ioremap() doesn't have a valid case for -ENOENT (but haven't checked yet). --mtx -- Enrico Weigelt, metux IT consult Free software and Linux embedded engineering info@xxxxxxxxx -- +49-151-27565287