[PATCCH] add converion of H_LongBusyOrder* to -EBUSY in arch/ppc64/kernel/hvconsole.c

Ryan Arnold rsa at us.ibm.com
Tue Apr 27 23:15:19 EST 2004

On Tue, 2004-04-27 at 00:45, Paul Mackerras wrote:

> Just looking at the surrounding code, I wonder why we return -EPERM
> for the default case?  EINVAL would possibly be a better choice for a
> generic "something went wrong".

Paul, thanks for the check-in.  We return -EPERM on the default _and_ on
H_Function because H_Function indicates that a function or operation is
not supported by the hypervisor.  The default RPA return value for
invalid parameters is H_Parameter, which does get converted to -EINVAL.

Testing tended to indicate that as long as the function was supported
H_Parameter was the return value when something known was wrong but
H_Function is returned in a much more narrow case.

My reasoning for wanting -EPERM returned as the default is that the
hypervisor already returns H_Parameter for way too many ambiguous cases
(not enough unique return codes for all the operations) and I'd like to
know when something unique happens (outside the standard H_Parameter
return case).

> How are we progressing towards having something that could be sent
> upstream (as far as the hvc stuff is concerned)?

Well, I've got the code ready.  I'm still going to resubmit the compile
warning patch to Ameslab first.  Mostly I'm waiting for Hollis's HVSI
patch to get submitted/accepted on the mainline since we share a large
amount of arch functionality that would be out of synch with the current
mainline code if I were to submit my driver first.

Ryan S. Arnold

** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/

More information about the Linuxppc64-dev mailing list