On 02/07/17 22:02, Phil Dennis-Jordan wrote: > On 7 February 2017 at 20:54, Laszlo Ersek <ler...@redhat.com> wrote:
>> I filed <https://bugzilla.tianocore.org/show_bug.cgi?id=368>. > > That looks nice and thorough. > >>> Your EDK2 patch >> >> For the record: >> [1] https://lists.01.org/pipermail/edk2-devel/2017-February/007072.html >> >>> fixes the problem with values OVMF writes to >>> DSDT/X_DSDT, but the issue of it refusing Qemu's linker commands for >>> those two still needs to be solved so that SeaBIOS can boot a wide >>> range of OSes. >>> >>> I'd be happy to give writing the memoising patch a go myself if that >>> helps. Looks like setting up an ORDERED_COLLECTION during phase 2 >>> might be the simplest solution? >> >> Thanks for the offer. :) >> >> * If you'd like to help me with my load and reach a good "development >> throughput", then I prefer to write the patch myself. On this >> *specific* occasion, I think it will be faster. >> >> I intend to send an OVMF series this week that addresses both >> TianoCore#368 (see above) and TianoCore#359 too (mentioned earlier). >> Both are for OvmfPkg/AcpiPlatformDxe, and it makes sense to round >> them up. >> >> I plan to CC the QEMU stakeholders as appropriate. Your feedback >> would be greatly appreciated; the same way as [1] is very helpful >> (thanks again for it!) >> >> * If your main interest is rather to get into OVMF development, then I >> positively welcome that, and encourage you to send the patch. >> Completing the patch will likely take longer (the edk2 coding style >> is... arcane... and your reviewer is somewhat pedantic :)), but if >> you plan to get into OVMF development, then it's worth both our times. >> >> (The above should not be misunderstood as "Laszlo doesn't value one-off >> contributions" -- it really depends on feature size and area. Hence the >> emphasis on "specific" above.) >> >> Your call :) If possible, please let me know it tomorrow. > > I actually ended up writing said pointer memoisation patch yesterday, > and it appears to work fine in initial tests. I still need to fully > work my way through the edk2 coding and contribution guidelines, so > I'll need to re-visit that patch with a fine tooth comb before > submitting it. In any case, here it is so far: > https://github.com/pmj/edk2/commit/58e0510c6da62d5a985b97e9bff84bc53442d3fe > > I am intending to submit more patches to edk2 over time - like this > one, they'll mostly be based on Reza Jelveh's GSoC project from a few > years ago. Some of his work on getting OS X/macOS guests working in > Qemu/OVMF have got upstreamed, most of it has not. I'm hoping I can > improve the ratio a little. > > I've also written an EFI framebuffer driver for the VMware virtual > SVGA adapter in Qemu, which again I need to tidy up to conform with > the coding conventions before submitting. (The only advantage of this > vs. QXL or virtio-gpu being that there are OSes, including OSX/macOS > for which there exist drivers for neither QXL nor virtio-gpu.) > > So I guess I may as well get some practice. :-) > > > I anticipate submitting the memoisation patch for review on Thursday > or Friday as I'll be out tomorrow. Sounds good to me! Thanks, Laszlo
