On 10/25/22 19:23, Claudio Fontana wrote:
+/*
+ * The cpu state corresponding to 'host_pc' is restored.
+ * When reset_icount is true, current TB will be interrupted and
+ * icount should be recalculated.
+ */
+static void cpu_restore_state_from_tb(CPUState *cpu, TranslationBlock *tb,
+ uintptr_t host_pc, bool reset_icount)
+{
+ uint64_t data[TARGET_INSN_START_WORDS];
+#ifdef CONFIG_PROFILER
+ TCGProfile *prof = &tcg_ctx->prof;
+ int64_t ti = profile_getclock();
+#endif
+ int insns_left = cpu_unwind_data_from_tb(tb, host_pc, data);
+
+ if (insns_left < 0) {
+ return;
+ }
Is the -1 return value some error condition to do anything about, log, tcg
assert, or ...,
under some DEBUG_* condition, or ignored as done here?
Interesting question.
By presenting this tb, have we asserted that host_pc is within (otherwise, why select this
tb). But if we didn't find host_pc within the unwind data... that suggests that the tcg
backend code generation may be wrong, generating an exception at an unexpected point.
But for the purposes of this patch, it is no change in behaviour. Previously we returned
from the function without goto found.
r~
