Oliver Neukum <oneukum@xxxxxxxx> writes: > Am Freitag, den 21.04.2017, 10:01 +0200 schrieb Bjørn Mork: >> There have been several reports of wdm_read returning unexpected EIO >> errors with QMI devices using the qmi_wwan driver. The reporters >> confirm that reverting prevents these errors. I have been unable to >> reproduce the bug myself, and have no explanation to offer either. But >> reverting is the safe choice here, given that the commit was an >> attempt to work around a firmware problem. Living with a firmware >> problem is still better than adding driver bugs. > > Hi, > > definitely better. Yet, ideally we'd fix both. Have you tested with > a partial reversal? Note that I have not been able to reproduce the bug myself, using similar hardware and firmware as the reporters. This is why my initial testing did not reveal it, and why I have been waiting so long before finally proposing the revert. I have been hoping to reproduce the bug so I could modify the code to solve both issues. A partial reversal which still would have made sense, would be to disable the workaround for QMI but leave it in place for MBIM. The original problem mostly affects Intel XMM7160 based MBIM devices. But there is no simple way to make such a change using the current API. And the fix is needed for v4.9 LTS. So starting with a full revert is the best option IMHO. We can consider changing the API later, and then reapply the workaround with a scope limited to MBIM only. Or some variant of it. It would really be best if we could apply it on MBIM OPEN and only then. But the protocol agnostic philosophy bites here... Bjørn -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
