On Wed, Dec 04, 2019 at 03:28:30PM +0000, Ard Biesheuvel wrote: > On Wed, 4 Dec 2019 at 15:23, Arvind Sankar <nivedita@xxxxxxxxxxxx> wrote: > > > > On Wed, Dec 04, 2019 at 03:03:04PM +0000, Ard Biesheuvel wrote: > > > On Tue, 3 Dec 2019 at 21:47, Arvind Sankar <nivedita@xxxxxxxxxxxx> wrote: > > > > > > > > There are two ways the status return of setup_gop32/64 could be > > > > incorrect. > > > > > > > > 1. If we don't find a usable GOP because none of them have a framebuffer > > > > (i.e. they were all PIXEL_BLT_ONLY), but all the efi calls succeeded, we > > > > will return EFI_SUCCESS even though we didn't find a usable GOP. Return > > > > EFI_NOT_FOUND instead, allowing the caller to probe for UGA instead. > > > > > > > > 2. If we do find a usable GOP, but one of the subsequent ones fails in > > > > an EFI call while checking if any support console output, we may return > > > > an EFI error code even though we found a usable GOP. Fix this by always > > > > returning EFI_SUCCESS if we got a usable GOP. > > > > > > > > > > Please split this into 2 patches > > > > Not quite following -- what should be the two pieces? Are you saying to > > split the success return and the failure return into two patches -- that > > seems excessively fine-grained. > > > > Your commit log describes two distinct changes 1. and 2. The patch > addresses those two issues, and in addition, moves variable > declarations around. The result is much more difficult to read than > necessary. > Ok fair enough. I'll do that in v2 and just leave out the declaration move then.
