On Sun, Mar 24, 2013 at 05:55:22PM -0400, Kevin O'Connor wrote:
On Sun, Mar 24, 2013 at 09:45:54PM +0200, Michael S. Tsirkin wrote:
On Sun, Mar 24, 2013 at 03:21:08PM -0400, Kevin O'Connor wrote:
On Sun, Mar 24, 2013 at 09:14:47PM +0200, Michael S. Tsirkin wrote:
I don't exactly understand what do you mean by "file".
A fw_cfg entry added using QEMU's fw_cfg_add_file() instead of fw_cfg_add_bytes().
Looks good overall. Some further proposals:
- I think we need runtime patching so we'll need another one that adds stuff from memory but yes it can look like a file for seabios.
- paths exposed to seabios are relative at the moment. So "acpi/*" ?
fw_cfg_add_file() does read from memory. It's not tied to the host filesystem or guest filesystem.
- add option to disable builtin tables at compile time
Agreed.
to keep the ability to develop seabios code in-tree, make it depend on compile option disabled by default.
similarly in QEMU, put code there disabled first, so developers can play with it.
when all is ready, flip the switch to enable in both places.
Sure - we can do that if needed.
-Kevin
Okay I'm pretty close to posting some patches that advance this project further, but wanted to check something beforehand: there are several tables that point to other tables (for example: FADT points to DSDT). What I did is provide a list of fixups such that bios can patch in pointers without any need to understand what's what. Thoughts?