For single-binary, we would really like to have a common TARGET_PAGE_BITS_MIN.
AVR's use of TARGET_PAGE_BITS == 8 is unfortunately too small. This was due to having mmio in the first 256 or 512 bytes and RAM starting immediately afterward -- the softmmu page table mapping really doesn't like mixed i/o and sram on the same page. For v2, my solution is to map the balance of the first page with a ram device instead of normal ram. This way, the entire first page is always i/o. In addition, add some symbolic names for the cpu registers in i/o 0x38-0x3f. r~ Richard Henderson (11): target/avr: Fix buffer read in avr_print_insn target/avr: Improve decode of LDS, STS hw/core/cpu: Use size_t for memory_rw_debug len argument target/avr: Remove OFFSET_CPU_REGISTERS target/avr: Remove NUMBER_OF_IO_REGISTERS target/avr: Add defines for i/o port registers target/avr: Move cpu register accesses into system memory target/avr: Use cpu_stb_mmuidx_ra in helper_fullwr target/avr: Use do_stb in avr_cpu_do_interrupt hw/avr: Prepare for TARGET_PAGE_SIZE > 256 target/avr: Increase TARGET_PAGE_BITS to 10 hw/avr/atmega.h | 1 + include/hw/core/cpu.h | 2 +- target/avr/cpu-param.h | 8 +- target/avr/cpu.h | 21 ++- target/avr/helper.h | 3 - target/sparc/cpu.h | 2 +- hw/avr/atmega.c | 39 +++++- target/avr/cpu.c | 16 +++ target/avr/disas.c | 21 ++- target/avr/helper.c | 262 ++++++++++++++++---------------------- target/avr/translate.c | 44 ++++--- target/sparc/mmu_helper.c | 2 +- target/avr/insn.decode | 7 +- 13 files changed, 222 insertions(+), 206 deletions(-) -- 2.43.0
