Hi, On 30-Oct-24 3:19 PM, Arnd Bergmann wrote: > On Wed, Oct 30, 2024, at 14:16, Hans de Goede wrote: >> On 29-Oct-24 1:55 PM, Ilpo Järvinen wrote: >>> On Mon, 28 Oct 2024, Arnd Bergmann wrote: >>> >>> It seems that the main goal why it was put into UAPI was "to give the user >>> some reference about proper num_args and response_size for each message": >>> >>> https://lore.kernel.org/all/CAPhsuW5V0BJT+YSwv1U=hRG0k9zBWXeRd=E1n4U5hvcnwEV3mQ@xxxxxxxxxxxxxx/ >>> >>> Are we actually expecting userspace to benefit from this in C form? >>> Suma? Hans? >> >> I can see how having this available in the uapi header as documentation >> of sorts is somewhat useful. >> >> OTOH I do agree that this array should probably not be used by userspace. >> >> And there is only 1 way to find out if it is actually used (which I do not >> expect) and that is to just drop it and find out (and to be willing to >> revert the change if it breaks things). >> >> So we can either move the array in its entirety to the c-code consuming it, >> which I think would be best; or we can go with Arnd's patch + add >> >> #ifdef __KERNEL__ >> >> around the array so that it is there for people reading the header, but >> it is no longer exposed as uapi. >> > > I don't think that would work, because the 'make headers_install' > step just removes the contents of #ifdef __KERNEL__, in this case > you just end up with a uapi header that doesn't have the array. Ah I did not realize that. Ok so lets just move the definition to the .c file which actually uses the array and then add a comment like this: /* * See hsmp_msg_desc_table[] in: * https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/platform/x86/amd/hsmp.c * * for some information on number of input- and output arguments * for the various functions. * * Please find the supported list of messages and message definition * in the HSMP chapter of respective family/model PPR. */ Note please replace hsmp.c with the .c file to which the array is actually moving ... Regards, Hans
