On 12/8/24 02:48, Richard Henderson wrote:
On 8/12/24 02:54, Philippe Mathieu-Daudé wrote:
When refactoring page_table_walk_refill() in commit 4e999bf419
we replaced the execution mode and forced it to kernel mode.
Restore the previous behavior to also get supervisor / user modes.
Reported-by: Thomas Petazzoni <thomas.petazz...@bootlin.com>
Reported-by: Waldemar Brodkorb <w...@uclibc-ng.org>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2470
Fixes: 4e999bf419 ("target/mips: Pass ptw_mmu_idx down from
mips_cpu_tlb_fill")
Signed-off-by: Philippe Mathieu-Daudé <phi...@linaro.org>
---
target/mips/tcg/sysemu/tlb_helper.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/target/mips/tcg/sysemu/tlb_helper.c
b/target/mips/tcg/sysemu/tlb_helper.c
index 3ba6d369a6..e7ae4f0bef 100644
--- a/target/mips/tcg/sysemu/tlb_helper.c
+++ b/target/mips/tcg/sysemu/tlb_helper.c
@@ -940,8 +940,9 @@ bool mips_cpu_tlb_fill(CPUState *cs, vaddr
address, int size,
* Memory reads during hardware page table walking are
performed
* as if they were kernel-mode load instructions.
*/
- int ptw_mmu_idx = (env->hflags & MIPS_HFLAG_ERL ?
- MMU_ERL_IDX : MMU_KERNEL_IDX);
+ int ptw_mmu_idx = (env->hflags & MIPS_HFLAG_ERL)
+ ? MMU_ERL_IDX
+ : (env->hflags & MIPS_HFLAG_KSU);
This contradicts the comment above.
If this code change is correct, then the comment isn't.
OK.
But the comment certainly makes sense -- page tables are never
accessible to user mode.
But we are still dropping the supervisor mode, so OK if I
reword as:
"Restore the previous behavior to also get supervisor modes."
?
