I have encountered two issues with this module after switching to 2.6.16: - the first one is that inter_module* functions have finaly been removed from Linux, but thinkpad module was still using them. I will attach a patch to convert it to the EXPORT_SYMBOL scheme. I don't know whether this patch should be applied unconditionnaly, or only for >=2.6.16 (should work with older versions, sure, but why make the change if it is not required?), so i will leave you integration to the src_unpack(). - the second one is that the thinkpadpm code is no more used in thinkpad-5.9, but the thinkpad module was still declaring an obsolete "enable_thinkpadpm" parameter, leading to this error: # modprobe thinkpad FATAL: Error inserting thinkpad (/lib/modules/2.6.16-archck1/thinkpad/thinkpad.ko): Invalid argument # dmesg | tail -n 1 thinkpad: falsely claims to have parameter enable_thinkpadpm Well, actually, this one i'm not sure whether it is specific to 2.6.16 kernel. I've also switched to gcc4 recently, so maybe that's him who optimize-out an unused variable, or something like that. I will report back later after trying a 2.6.15 again. Anyway, the patch for this one is simple too: it drops any mention of the obsolete parameter in the thinkpad module.
Created attachment 83027 [details, diff] thinkpad-5.9-remove-inter_module.patch patch for issue #1
Created attachment 83028 [details, diff] thinkpad-5.9-remove-thinkpadpm-argument.patch patch for issue #2
(In reply to comment #0) > > Well, actually, this one i'm not sure whether it is specific to 2.6.16 kernel. I have just tested on 2.6.15, and the unpatched module doesn't make kernel complain about this parameter. So it seems to be specific to 2.6.16, although i don't know where the difference comes from. I have also tested both patches on 2.6.15, and they seems to work fine there too. So i still don't have a clue whether they should be applied only for `kernel_is gt 2 6 16` or not.
Tested here on my Thinkpad 600E, thanks for the patches, committed to CVS, please let me know if there are any problems. Will try to 5.9-r1 stable in 2-3 weeks.
(In reply to comment #4) > Tested here on my Thinkpad 600E, thanks for the patches, committed to CVS, > please let me know if there are any problems. Will try to 5.9-r1 stable in 2-3 > weeks. > Have you forgotten to stable the new version or are problems?
Yes I had forgotten to request stabelization. Will open a bug shortly, unless you get to it before I get the chance
TGL: I'm not kernel hacker, thus may be I'm wrong but it seems to me that it's better to use pxint_doSmapi = (pxint_do_t)symbol_request( smapi_do ); instead of, suggested in your patch pxint_doSuperio = (pxint_do_t)try_then_request_module( __symbol_get("superio_do") , _szSuperioName ); symbol_request is defined include/linux/module.h and it's definition is very similar to the code you wrote, but there are some differences. FEX, it uses symbol_get instead of __symbol_get and thus takes into account MODULE_SYMBOL_PREFIX. What do you think about this?
