On Wed, 2019-05-29 at 07:21 +0000, Yang, Hyungwoo wrote: > > On Tue, 2019-05-28 at 14:36 -0700, Hyungwoo Yang wrote: > > > "struct device" is embedded in "struct ishtp_cl_device". > > > so the conversion should be done by container_of() method. > > > > Which tree this patch is going to? You can't even apply on the > > mainline tree (5.2-rc2). Also you will not be able to compile even > > if you address the conflict (The patch ordering is wrong). > > Oh.. I wasn't careful to split the patch after testing. I've > corrected. > > > > > What was symptom or problem you try to address? Is there any crash > > or bug occurred? Does it happen with the mainline kernel? > > I've added the detail in commit message in v2. Basically due to wrong > usage of driver_data of ishtp client device, we see kernel crash. > Currently driver_data is set by bus driver which is wrong since > driver_data should be owned by corresponding device driver. Right > now, we see kernel crash during suspend() of cros_ec_ishtp. Yes, it > happens with the mainline kernel since cros_ec_ishtp is already > upstreamed. Technically this driver is not mainline. It will go in 5.3. The problem is cros_ec ish driver is overriding driver_data " client_data->ec_dev = ec_dev; dev->driver_data = ec_dev; " The client drivers own the driver data in its "struct ishtp_cl_device *" not the struct device *. As far as I can see the purpose of this is to get device pointer for debug purpose only. I think you can remove the dev->driver_data assignment and simply replace dev_*(dev, to dev_*(ec_dev->dev, Thanks, Srinivas > > > > > > > > > > > Change-Id: Idcbafe724e216ee6275f9d1e35a3b79cee5ce88c > > > > This tells me that you are trying to fix some Chrome issue. Don't > > include these tags for mainline kernel. > > done in v2. Thank you. > > > > > Thanks, > > Srinivas > > Thanks, > Hyungwoo
