On Thu, 26 Sept 2024 at 18:05, Chao Liu <chao....@yeah.net> wrote: > > > The ignore_memory_transaction_failures is used for compatibility > > with legacy board models. > > > > I attempted to remove this property from the > > xilink_zynq board and replace it with unimplemented devices to > > handle devices that are not implemented on the board. > > > > Chao Liu (2): > > xilink_zynq: Add various missing unimplemented devices > > xilink-zynq-devcfg: Fix up for memory address range size not set > > correctly > > > > hw/arm/xilinx_zynq.c | 44 ++++++++++++++++++++++++++++++++++++++- > > hw/dma/xlnx-zynq-devcfg.c | 2 +- > > 2 files changed, 44 insertions(+), 2 deletions(-) > > > > -- > > 2.46.1 > > Hello, maintainers, > > Could you please provide any suggestions or feedback on the set of patches? > This is my first contribution to the QEMU community, > and it makes me feel very honored. > > Additionally, I have used creat_unimplemented_device() to > add all the unimplemented devices on the Xilinx Zynq board, > primarily referencing the Zynq DTS, located at > > roms/u-boot/arch/arm/dts/zynq-7000.dtsi.
Hi; this is on my todo list to review, but I think the big question here is how much testing you have done and on how big a range of guest software. The reason that we set ignore_memory_transaction_failures on this board (and all the others) was that it kept the behaviour the same as before we added the support for making memory transactions fail for accesses to nonexistent devices. This was the safest thing since we didn't have access to that much guest software to test on all of them. Getting rid of it for this board is a good thing, but we need to be reasonably confident that we have implemented all the devices that guests might be touching. The cross-check against the dts is helpful there, but so too is running as much real works-on-the-board code as we can, to check it still boots. thanks -- PMM
