On 2/7/26 18:05, Richard Henderson wrote:
On 7/1/26 07:08, Philippe Mathieu-Daudé wrote:
Sorry for not being clear enough, I meant the TLB_MMIO & TLB_WATCHPOINT
checks just after:
+ /* The first active element crosses a page boundary. */
+ flags |= info.page[1].flags;
+ if (unlikely(flags & TLB_MMIO)) {
+ /* Some page is MMIO, see below. */
+ goto do_fault;
+ }
+ if (unlikely(flags & TLB_WATCHPOINT) &&
+ (cpu_watchpoint_address_matches
+ (env_cpu(env), addr + mem_off, 1 << msz)
+ & BP_MEM_READ)) {
+ /* Watchpoint hit, see below. */
+ goto do_fault;
+ }
We still have TLB_WATCHPOINT == 0 for user-only, so that won't actually
fire.
We set TLB_MMIO to non-zero for user-only in 6d03226b422 so that we
could force slow paths for plugins. These days we should probably
rearrange things so that TLB_FORCE_SLOW is used for that instead, and
make TLB_MMIO be 0 again.
OK. I'm still stuck by this function complexity, and have the
feeling some of this should be handled at the lower mte_probe()
level (maybe extending it) up to allocation_tag_mem_probe() which
is clearer. But I have no expertise here so my comments might be
non-sense, no offense :)
r~
+ /* TODO: MTE check. */
+ /*
+ * Use the slow path for cross-page handling.
+ * This is RAM, without a watchpoint, and will not trap.
+ */
+ tlb_fn(env, vd, reg_off, addr + mem_off, retaddr);
+ goto second_page;
}
}
