Hi, On 3/11/23 21:35, Nikita Kravets wrote: > Hi, > >> The above two functions are inconsistent with the rest of the file because >> they have the return type in a separate line. > > I believe it was originally made to not exceed the limit of 80 columns > >> It's a small thing, but usually /* ... */ comments are preferred. >> Hans will tell you if you need to change it. > > Okay Yes I noticed these too during my review, but I'm fine with them as long as a single driver uses a single comment style consistently. Note most drivers/platform/x86 drivers do indeed use /* ... */, so I would not mind if you convert the comments, but this is not a blocker for merging the driver. >> Previously you said `match_string()` works here. Has something changed? > > No, I implemented it in the main repo but forgot to do it in the kernel patch.. > I'll do it in the v3 after we're finished reviewing v2 Thanks, sounds good. Only remark which I have is that the driver still has no modaliases so if this gets enabled as a module by distros, which is what most distros will do it will never load. I think you should add a dmi_system_id table together with MODULE_DEVICE_TABLE() e.g. something like this: static const struct dmi_system_id msi_dmi_table[] __initconst = { { matches = { DMI_MATCH(DMI_SYS_VENDOR, "MICRO-STAR INT"), } }, { matches = { DMI_MATCH(DMI_SYS_VENDOR, "Micro-Star International"), } }, {} }; MODULE_DEVICE_TABLE(dmi, msi_dmi_table); To make the module auto-load. But I guess it might be better to put that on the TODO list somewhere and do it once the module has had some more testing. E.g. I should really test this on my desktop's MSI B550M PRO-VDH board and see what it does there (hopefully nothing). Regards, Hans