On 04/20/2018 02:57 AM, Michael Clark wrote:
On Wed, Apr 18, 2018 at 10:16 PM, KONRAD Frederic < frederic.kon...@adacore.com> wrote:The htif device is supposed to be mapped over an other subregion. So increase its priority to one to avoid any conflict. Here is the output of info mtree: Before: (qemu) info mtree address-space: memory 0000000000000000-ffffffffffffffff (prio 0, i/o): system 0000000000000000-000000000000000f (prio 0, i/o): riscv.htif.uart 0000000000000000-0000000000011fff (prio 0, ram): riscv.spike.bootrom 0000000002000000-000000000200ffff (prio 0, i/o): riscv.sifive.clint 0000000080000000-0000000087ffffff (prio 0, ram): riscv.spike.ram address-space: I/O 0000000000000000-000000000000ffff (prio 0, i/o): io address-space: cpu-memory-0 0000000000000000-ffffffffffffffff (prio 0, i/o): system 0000000000000000-000000000000000f (prio 0, i/o): riscv.htif.uart 0000000000000000-0000000000011fff (prio 0, ram): riscv.spike.bootrom 0000000002000000-000000000200ffff (prio 0, i/o): riscv.sifive.clint 0000000080000000-0000000087ffffff (prio 0, ram): riscv.spike.ram After: (qemu) info mtree address-space: memory 0000000000000000-ffffffffffffffff (prio 0, i/o): system 0000000000000000-000000000000000f (prio 1, i/o): riscv.htif.uart 0000000000000000-0000000000011fff (prio 0, ram): riscv.spike.bootrom 0000000002000000-000000000200ffff (prio 0, i/o): riscv.sifive.clint 0000000080000000-0000000087ffffff (prio 0, ram): riscv.spike.ram address-space: I/O 0000000000000000-000000000000ffff (prio 0, i/o): io address-space: cpu-memory-0 0000000000000000-ffffffffffffffff (prio 0, i/o): system 0000000000000000-000000000000000f (prio 1, i/o): riscv.htif.uart 0000000000000000-0000000000011fff (prio 0, ram): riscv.spike.bootrom 0000000002000000-000000000200ffff (prio 0, i/o): riscv.sifive.clint 0000000080000000-0000000087ffffff (prio 0, ram): riscv.spike.ram Signed-off-by: KONRAD Frederic <frederic.kon...@adacore.com>Reviewed-by: Michael Clark <m...@sifive.com> BTW if you like I can incorporate these into the riscv tree. e.g. here, feel free to make a PR in GitHub: https://github.com/riscv/riscv-qemu/tree/qemu-2.13-for-upstream (however the riscv.org tree has a lot of changes in it so you might be better off getting these changes merged in upstream as the review backlog is large) It's just that i'd like to keep the trees in sync with upstream while at the same time incorporating as many fixes as I can in the riscv tree (before they are in upstream) so that the riscv tree is complete for users who want to pull a branch that includes the outstanding riscv fixes. I have scripts that can merge mutliple branches into 'riscv-all' in the riscv github so if you make a PR against the riscv github I can merge them in that tree, and drop them if they get accepted upstream independently. I'm pulling from master frequently so its fine either way. Just i'd like the riscv tree to accumulate all of the fixes that haven't yet made it upstream.
Hi Michael, Ok I'll respin on the list then and try to do a PR on GitHub. Thanks, Fred
---hw/riscv/riscv_htif.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/hw/riscv/riscv_htif.c b/hw/riscv/riscv_htif.c index 6e687f2..48e5452 100644 --- a/hw/riscv/riscv_htif.c +++ b/hw/riscv/riscv_htif.c @@ -253,8 +253,9 @@ HTIFState *htif_mm_init(MemoryRegion *address_space, MemoryRegion *main_mem, htif_be_change, s, NULL, true); if (address_symbol_set) { memory_region_init_io(&s->mmio, NULL, &htif_mm_ops, s, - TYPE_HTIF_UART, size); - memory_region_add_subregion(address_space, base, &s->mmio); + TYPE_HTIF_UART, size); + memory_region_add_subregion_overlap(address_space, base, + &s->mmio, 1); } return s; -- 1.8.3.1
