Hi Jiri, While playing a little bit with the CP2112 (for hid-rmi I must confess), I realized that kernel v4.9 now enforces DMA capable buffers when calling hid_hw_raw_request(). Kernel v4.8 works fine (but gives a stacktrace the first time), but v4.9 doesn't.
So I gave a check of all the other drivers in the HID tree, and it looks like only 4 drivers are not properly allocating their buffers. I'd say this is v4.9 material but I was not able to test magicmouse and lg. If this doesn't qualifies for 4.9-rc7, I think we should add the stable@ stamp, given that the chance this will break hid-rmi is huge (cp2112 is not so much an issue, given it's a devel board). Cheers, Benjamin Benjamin Tissoires (4): HID: cp2112: make transfer buffers DMA capable HID: lg: make transfer buffers DMA capable HID: magicmouse: make transfer buffers DMA capable HID: rmi: make transfer buffers DMA capable drivers/hid/hid-cp2112.c | 115 +++++++++++++++++++++++++++++-------------- drivers/hid/hid-lg.c | 12 +++-- drivers/hid/hid-magicmouse.c | 12 ++++- drivers/hid/hid-rmi.c | 10 +++- 4 files changed, 106 insertions(+), 43 deletions(-) -- 2.7.4
