On 27.04.19 10:58, Esben Haabendal wrote: Hi folks, > That said, the purpose of UPF_BOOT_AUTOCONF (for 8250 driver) is to> request and map the register memory. So when that is already done by> the parent MFD driver, I think it is silly to workaround problems caused> by UPF_BOOT_AUTOCONF being force setted, when it really shouldn't. I tend to agree. Maybe we should give serial8250_register_8250_port() some flags for controlling this, or add another function for those cases. A minimal-invasive approach could be introducing an serial8250_register_8250_port_ext() with extra parameters, and let serial8250_register_8250_port() just call it. By the way: the mem-mapping code pathes in serial8250 are a bit strange anyways, eg. serial8250_port_size() is always called when map/unmap is done, and this function does special logic for certain devices. IMHO, it this should be called only once on device init and later just use the already computed value: https://github.com/metux/linux/commit/7ec63b699c910228b92cfb27eb8edfda90fdfd63 (haven't sent this queue yet, but feel free to cherry-pick if you like) --mtx -- Enrico Weigelt, metux IT consult Free software and Linux embedded engineering info@xxxxxxxxx -- +49-151-27565287
