On 23.10.18 11:16, Jouke Witteveen wrote: > On Tue, Oct 23, 2018 at 6:06 AM Henrique de Moraes Holschuh > <hmh@xxxxxxxxxx> wrote: >> >> On Mon, 15 Oct 2018, Dmitry Torokhov wrote: >>> On Sat, Oct 13, 2018 at 6:32 AM Jaak Ristioja <jaak@xxxxxxxxxxx> wrote: >>>> On 27.08.2018 19:22, Jaak Ristioja wrote: >>>>> Upgrading Linux from 4.16 to 4.17, a ThinkPad L570 started receiving >>>>> spurious input events, mostly right mouse button click events, but also >>>>> cursor jumps. >>>>> >>>>> I have not attempted to understand whether these events come from the >>>>> trackpoint or touchpad or some other driver, but I managed to bisect >>>>> this issue to commit a986c75a7df0 titled "platform/x86: thinkpad_acpi: >>>>> Add 2nd Fan Support for Thinkpad P50" by Alexander Kappner. >>>>> >>>>> Apparently the quirk mitigation is applied when the BIOS version begins >>>>> with N1. The BIOS version on the L570 in question is N1XET57W (1.35 ) >>>>> which is probably why this commit causes the described problems on the >>>>> ThinkPad L570. How exactly? - I don't know. >>>>> >>>>> The issue did not reproduce when running some stable 4.17 and 4.18 >>>>> kernels with commit a986c75a7df0 reverted. >>>>> >>>>> Please fix this for future kernels. Thanks! :) >>>> >>>> Ping. Do you need any additional information? >>> >>> Sounds like we need tighter check for the quirk, maybe based on >>> DMI/Board name? Can we revert the offending commit for now? >> >> Apparently it has to be fixed properly, because if I understood the >> issue correctly, every two-character quirk will trigger on *any* >> three-character model that starts with those two characters. > > This should be (and hopefully is) incorrect. > In the tpacpi_check_quirks function, equality between the quirk bios > and the actual model bios is tested. For two-character models there > are three nonzero bytes, for two character models there are only two. > Therefore, a two-character quirk should not trigger on any > three-character model. > >> We could revert the three-character quirk support, but this would cause >> regressions as well. > > I doubt this would fix anything. Maybe the bios id for the L570 that > is used by the driver is not the one Jaak Ristioja reported? The > detected string is reported in the dmesg output. I don't see the BIOS version in the dmesg output. The BIOS IDs do start with N1 thou as this is apparent from dmidecode output on the L570 and from the BIOS changelog as well: https://download.lenovo.com/pccbbs/mobiles/n1xuj24w.txt Best regards, J
