Re: Fwd: Re: [PATCH] PCI I/O space

  • From: Robert Millan <rmh@xxxxxxxxxxx>
  • To: pooka@xxxxxx, rumpkernel-users@xxxxxxxxxxxxx
  • Date: Fri, 12 Jun 2015 23:25:58 +0200


Hi Antti,

El 08/06/15 a les 07:53, Antti Kantee ha escrit:
>> +int rumpcomp_pci_init(int, int *);
>
> Now that rumpcomp_userfeatures_pci.h is available, that signature can be
simplified.

That'd be "int rumpcomp_pci_init(void);" ?

El 08/06/15 a les 23:20, Antti Kantee ha escrit:

I'm just worried that your print will look too much like an error when it is
not necessary so. Perhaps there's a way to make the print softer, something to
indicate that some features are missing most likely because the program is not
being run as root or with CAP_SYS_RAWIO. In other words, treat EPERM as the
common case error.

Also, the print and the call doesn't match: pci_init() vs. "iopl failed".

How about "pci: initialization returned %d, some features may be unavailable" ?

Finally, commenting on the hypercall side of the patch, you should not return
the return value of iopl(), but rather the translated errno.

Are there any facilities and/or standard way of doing this? Or should I just
use an ad-hoc switch/case with possible iopl() errnos?

Thanks

--
Robert Millan

Other related posts: