target/hppa: fix PSW Q bit behaviour to match hardware

PA-RISC specification says: "Setting the PSW Q-bit, PSW{28}, to 1
with this instruction, if it was not already 1, is an undefined
operation." However, at least HP-UX 10.20 sets the Q bit from 0 to 1
with the SSM instruction. Tested this both on HP9000/712 and
HP9000/785/C3750, both machines set the Q bit from 0 to 1 without
exception. This makes HP-UX 10.20 progress a little bit further.

Signed-off-by: Sven Schnelle <svens@stackframe.org>
Message-Id: <20190129191402.29539-1-svens@stackframe.org>
[rth: Add a comment to the code as well.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

diff --git a/target/hppa/op_helper.c b/target/hppa/op_helper.c
index 912e8d5be42441b4d92c7d29e524d0d9fba83e3b..6bf478e7b02e80ed6050e991c4071893199a9bb0 100644 (file)
--- a/target/hppa/op_helper.c
+++ b/target/hppa/op_helper.c
@@ -665,11 +665,15 @@ void HELPER(reset)(CPUHPPAState *env)
 target_ureg HELPER(swap_system_mask)(CPUHPPAState *env, target_ureg nsm)
 {
     target_ulong psw = env->psw;
-    /* ??? On second reading this condition simply seems
-       to be undefined rather than a diagnosed trap.  */
-    if (nsm & ~psw & PSW_Q) {
-        hppa_dynamic_excp(env, EXCP_ILL, GETPC());
-    }
+    /*
+     * Setting the PSW Q bit to 1, if it was not already 1, is an
+     * undefined operation.
+     *
+     * However, HP-UX 10.20 does this with the SSM instruction.
+     * Tested this on HP9000/712 and HP9000/785/C3750 and both
+     * machines set the Q bit from 0 to 1 without an exception,
+     * so let this go without comment.
+     */
     env->psw = (psw & ~PSW_SM) | (nsm & PSW_SM);
     return psw & PSW_SM;
 }