On Fri, Dec 30, 2011 at 3:54 PM, Dave Jones<davej@xxxxxxxxxx> wrote:We're getting a bunch of reports against Fedora 16
(still using 3.1) which look like some drivers are trying to
load firmware on resume from suspend, while usermodehelper
Ok, buggy drivers. You *must*not* load firmware in your resume path,
since there is no actual guarantee that any particular device will be
resumed after the disk that contains the firmware images.
So it's very simple: drivers that load firmware at resume time are
buggy. No ifs, buts, or maybes about it.
Here are some example traces:
It's isight_firmware_load(), in the isight_firmware driver. The driver
doesn't actually do anything but load the firmware, and is apparently
not very good at that either.
It should either fake a disconnect and reconnect of the device (and
let the reconnect then load the firmware through udev or something) or
it should just save the firmware image in memory from the original
load, and make the resume just re-initialize it - not load it.
It's also possible that it could be considered a USB layer bug, and
the USB layer should just not rebind the devices directly in the
resume function, but do it somehow later. HOWEVER, that would only
work for "random" USB devices that aren't in use by user space (like
disks etc might be). So I think that in general the real solution is
always just "make sure that the firmware is in memory before the
suspend even happens".
Greg - has the USB resume logic been changed lately?
Matthew? Any comments about that particular driver?
Same issue, different driver. Again, it's USB, and it's possible that
USB just makes it really hard to do this correctly (ie the "save
firmware image across suspend so that you don't have to load it at
It's also possible that we should blame the firmware code, which is
expressly written to encourage these kinds of bugs. It may be that i
tshould be the firmware code that has a "get_firmware()" +
"put_firmware()" model, and it should cache the firmware explicitly if
the config supports suspend, so that a firmware read at resume time
would actually work. The whole "request_firmware()" interface really
is very prone to these kinds of bugs.
But it's possible this could be fixed at the driver level by doing the
In this case it's the rtl8192cu driver, so Larry, Chaoming, John etc
added to the cc for that one.