We have refactored and extended this - document it. Signed-off-by: Daniel Axtens <dja@xxxxxxxxxx> --- Documentation/gpu/default_display.rst | 93 +++++++++++++++++++++++++++++++++++ Documentation/gpu/index.rst | 1 + 2 files changed, 94 insertions(+) create mode 100644 Documentation/gpu/default_display.rst diff --git a/Documentation/gpu/default_display.rst b/Documentation/gpu/default_display.rst new file mode 100644 index 000000000000..3c190611564e --- /dev/null +++ b/Documentation/gpu/default_display.rst @@ -0,0 +1,93 @@ +======================= +Default Display Devices +======================= + +Overview +======== + +.. kernel-doc:: drivers/gpu/vga/default_display.c + :doc: overview + + +Why do we need this? +==================== + +The default device is used to set the ``boot_vga`` per-device sysfs +file, which is used by user-space. Most notably, Xorg reads this file +via libpciaccess in order to facilitate auto-configuration. + + +History +======= + +When the VGA arbiter was introduced, it would pick a default device on +boot. As the arbiter exists to arbitrate access to legacy resources, +it would only pick a card that could be accessed through legacy areas. +(See the :doc:`vgaarbiter` documentation for more.) + +The arbiter was later extended on x86 and IA64 to consider the EFI +framebuffer. + +This is all well and good if you have legacy resources or +EFI. However, some systems do not have either of those. For example, +on POWER8: [0]_ + + - There is no IO space at all + + - We configure the 32-bit MMIO window to be around 3..4G (to avoid + overlapping with DMA space below it). + +So effectively, there is no path to the legacy areas. + +This means the VGA arbiter will not pick a default device on these +platforms. So, on powerpc, a class hook was added to mark a default +device (``arch/powerpc/kernel/pci-common.c``), independent of the +arbiter. + +When this issue arose on an arm64 system, it was decided that a generic +approach would be better than more special cases. Therefore, the +default device selection was factored out, and it now operates using +the priority list described in the Overview. + +API +=== + +Public +------ + +.. kernel-doc:: drivers/gpu/vga/default_display.c + :export: + +Private +------- + +.. kernel-doc:: drivers/gpu/vga/default_display.c + :internal: + +Future Work +=========== + +There is no support for non-PCI VGA devices being marked as default. +The following comment, extracted from an earlier version of +:c:func:`pci_default_display()` might help: + + If your VGA default device is not PCI, you'll have to override this + and return NULL here. In this case, I assume it will not conflict + with any PCI card. If this is not true, I'll have to define two + archs hooks for enabling/disabling the VGA default device if that + is possible. + + This may be a problem with real _ISA_ VGA cards, in addition to a + PCI one. I don't know at this point how to deal with that card. Can + theirs IOs be disabled at all ? If not, then I suppose it's a matter + of having the proper arch hook telling us about it, so we basically + never allow anybody to succeed a ``vga_get()``... + +Currently there is also no way to support non-VGA-class PCI devices as +default display devices. + + +References +========== + +.. [0] https://www.spinics.net/lists/linux-pci/msg64142.html diff --git a/Documentation/gpu/index.rst b/Documentation/gpu/index.rst index 35d673bf9b56..8083d84f2334 100644 --- a/Documentation/gpu/index.rst +++ b/Documentation/gpu/index.rst @@ -16,6 +16,7 @@ Linux GPU Driver Developer's Guide tegra tinydrm vc4 + default_display vga-switcheroo vgaarbiter bridge/dw-hdmi -- 2.11.0
