在2024年5月21日五月 上午8:35，Max Filippov写道： [...]
+++ b/board/emulation/qemu-xtensa/MAINTAINERS @@ -0,0 +1,7 @@ +QEMU XTENSA 'VIRT' BOARD +M: Jiaxun Yang jiaxun.yang@flygoat.com +S: Maintained +F: board/emulation/qemu-xtensa/ +F: board/emulation/common/ +F: include/configs/qemu-riscv.h
This must be qemu-xtensa.h, right?
Ah yes :-)
Do you mind me to give you a M: entry? I'm pretty dumb to Xtensa stuff so it deserves an M from people know Xtensa better.
[...]
I'm not sure what CFG_SYS_IO_BASE means for the virt machine, but it looks like in the noMMU case SDRAM is going to overlap the IO range with the above base addresses and sizes. And from QEMU monitor I see the following address space distribution in the virt machine with a noMMU core:
So CFG_SYS_IO_BASE means physical address left for I/O with 1:1 mapping from virtual address. for PTP MMU I handled it in:
[PATCH 03/10] xtensa: Implement phys virt conversion for PTP_MMU
Can you confirm if noMMU core is doing the same on VA-PA mapping?
If so, that patch needs to be expanded to include noMMU as well.
Thanks - Jiaxun
address-space: memory 0000000000000000-ffffffffffffffff (prio 0, i/o): system 000000003ffe0000-000000003fffffff (prio 0, ram): xtensa.dataram0 0000000040000000-000000004001ffff (prio 0, ram): xtensa.instram0 0000000050000000-0000000053ffffff (prio 0, ram): xtensa.sysrom0 0000000060000000-00000000dfffffff (prio 0, ram): xtensa.sysram0 00000000f0000000-00000000f000ffff (prio 0, i/o): alias pcie-pio @gpex_ioport_window 0000000000000000-000000000000ffff 00000000f0100000-00000000f3ffffff (prio 0, i/o): alias pcie-ecam @pcie-mmcfg-mmio 0000000000000000-0000000003efffff 00000000f4000000-00000000fbffffff (prio 0, i/o): alias pcie-mmio @gpex_mmio_window 00000000f4000000-00000000fbffffff
i.e. there's no IO at the address 0x90000000.
-- Thanks. -- Max