Hello Jules, On 29.01.21 11:51, Jules Maselbas wrote: > Hi Ahmad, > > On Fri, Jan 29, 2021 at 10:51:18AM +0100, Ahmad Fatoum wrote: >> Hello Jules, >> >> On 27.01.21 17:49, Jules Maselbas wrote: >>> File system operation shouldn't be executed in a poller. Use >>> a workqueue to delay filesystem operation to command context. >>> >>> This is an RFC, extra work must be done to properly handle error >>> cases and dfu cleanup. >> >> I erroneously thought the poller is within the DFU bits. I wonder what >> side-effect moving the whole USB gadget polling into a workqueue would >> have. In that case, we wouldn't need to any changes for DFU itself. >> > If I understood correctly you're suggesting to wrap the entire dfu > gadget inside a poller. I have not tried this and it might work. > However wrapping each fs operation allow the dfu gadget to respond > to GET_STATUS queries will erase/write operation are on-going. To allow running in the "background", the UDC calls into the gadget code in a poller. I was wondering if replacing that poller with a workqueue would just make the problem go away, not for DFU. It now know it wouldn't: Work queues work in shell context and doing console stuff in shell context (via ttyACM gadget) is not a wise choice. So please dismiss my suggestion ^^. > > Jules > > -- Pengutronix e.K. | | Steuerwalder Str. 21 | http://www.pengutronix.de/ | 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox