These patches are supposed to fix a problem Daniel Axtens found on the HiSilicon D05 board. The VGA device there is behind a bridge that doesn't support PCI_BRIDGE_CTL_VGA, so the arbiter never selects the device as the default.
The first patch extends the arbiter so that if it can't find an enabled VGA device with legacy resources, it selects the first enabled device *without* legacy resources (this is what fixes the D05). If that fails, it selects the first device that isn't enabled. The combination of both changes should make the current powerpc fixup_vga() quirk unnecessary.
N.B. It changes the powerpc behavior: if there are several enabled VGA devices, the current quirk selects the last one, while this patch selects the first one. If this is a problem, I can drop that part of the patch and keep the quirk.
The second patch pulls out this fallback device detection (and the EFI override) from vga_arb_device_init() to make it easier to read.
Changes in v2: - set default device to vgadev->pdev instead of the pdev left over from a previous iteration through the list (thanks to Julien Thierry)
- added "Tested-by: Zhou Wang" (I'm pretty sure the change above doesn't invalidate this testing)
---
Bjorn Helgaas (2): vgaarb: Select a default VGA device even if there's no legacy VGA vgaarb: Factor out EFI and fallback default device selection
arch/powerpc/kernel/pci-common.c | 12 ------ drivers/gpu/vga/vgaarb.c | 72 +++++++++++++++++++++++++++++--------- 2 files changed, 55 insertions(+), 29 deletions(-)