Hi, Robert Baldyga <r.baldyga@xxxxxxxxxxx> writes: >> Olivier Schonken <olivier.schonken@xxxxxxxxx> writes: >>> Hope I'm not raising an issue which actually is my own stupidity. >>> >>> I recently upstreamed my embedded Linux kernel to version 4.4, after which >>> I started experiencing issues with my serial gadget - acm and obex. I used >>> to be able to run a getty console on it through a systemd service, which >>> would restart everytime it was unplugged, and plugged in again. With >>> kernel 4.4, this stopped working as usual(On startup console would word, >>> but if plugged out and in again, console is frozen), and I traced it back >>> to the series of patches "usb: gadget: introduce 'enabled' flag in struct >>> usb_ep". If I revert these patches, everything works as normal again. As >>> you are the owner of these patches, I thought that you might help me >>> understand if I'm doing something wrong w.r.t. re-initializing the console >>> getty, or if something really got broken. >> >> This is short of email where the mailing list should be in Cc :-) >> >> The author of that series was Robert (now in Cc). Robert, does this ring >> any bells to you ? >> > > I have tested it with getty (without systemd) and it looks like getty > hangs up after replugging no matter if patch is applied or not, so > without systemd it won't work anyway. > > Hence how systemd handles this is the key here, however I have no idea > how systemd knows that USB device is unplugged. On device side > /dev/ttyGS0 exists regardless to ACM function state, so process which > opened this device file has no way to figure out whether connection to > host is active or not. Can you explain how systemd knows that device is > unplugged? This knowledge can be helpful. there should be a SIGHUP when cable's removed, no ? -- balbi
Attachment:
signature.asc
Description: PGP signature
