On Fri, Apr 24, 2020 at 11:07:43AM -0500, Nathan Lynch wrote: > Sam Bobroff <sbobr...@linux.ibm.com> writes: > > Export rtas_error_rc() so that it can be used by other users of > > rtas_call() (which is already exported). > > This will do the right thing for your ibm,configure-pe use case in patch > 2, but the -900x => errno translations in rtas_error_rc() appear > tailored for the indicator- and sensor-related calls that currently use > it. From my reading of PAPR+, the meaning of a -900x RTAS status word > depends on the call. For example, -9002 commonly means "not authorized", > which we would typically translate to -EPERM, but rtas_error_rc() would > translate it to -ENODEV. > > Also the semantics of -9001 as a return value seem to vary a bit. > > So I don't think rtas_error_rc() should be advertised as a generically > useful facility in its current form. > > (I have had some thoughts about how firmware/hypervisor call status can > be translated to meaningful Linux error values without tedious switch > statements, which I'm happy to expand on if anyone is interested, but I > don't want to hijack your submission for that discussion.)
Ah, interesting. I'll do another version as you suggest. Cheers, Sam.
signature.asc
Description: PGP signature
