From: Nicholas Piggin <npiggin@gmail.com>

The decrementer register contains a relative time in timebase units.
When writing to DECR this is converted and stored as an absolute value
in nanosecond units, reading DECR converts back to relative timebase.

The tb<->ns conversion of the relative part can cause rounding such that
a value writen to the decrementer can read back a different, with time
held constant. This is a particular problem for a deterministic icount
and record-replay trace.

Fix this by storing the absolute value in timebase units rather than
nanoseconds. The math before:
  store:  decr_next = now_ns + decr * ns_per_sec / tb_per_sec
  load:        decr = (decr_next - now_ns) * tb_per_sec / ns_per_sec
  load(store): decr = decr * ns_per_sec / tb_per_sec * tb_per_sec /
                      ns_per_sec

After:
  store:  decr_next = now_ns * tb_per_sec / ns_per_sec + decr
  load:        decr = decr_next - now_ns * tb_per_sec / ns_per_sec
  load(store): decr = decr

Fixes: 9fddaa0c0cab ("PowerPC merge: real time TB and decrementer - faster and simpler exception handling (Jocelyn Mayer)")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
---
 hw/ppc/ppc.c | 39 ++++++++++++++++++++++++---------------
 1 file changed, 24 insertions(+), 15 deletions(-)

diff --git a/hw/ppc/ppc.c b/hw/ppc/ppc.c
index 13eb45f4b77e..a397820d9c34 100644
--- a/hw/ppc/ppc.c
+++ b/hw/ppc/ppc.c
@@ -711,16 +711,17 @@ bool ppc_decr_clear_on_delivery(CPUPPCState *env)
 static inline int64_t _cpu_ppc_load_decr(CPUPPCState *env, uint64_t next)
 {
     ppc_tb_t *tb_env = env->tb_env;
-    int64_t decr, diff;
+    uint64_t now, n;
+    int64_t decr;
 
-    diff = next - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
-    if (diff >= 0) {
-        decr = ns_to_tb(tb_env->decr_freq, diff);
-    } else if (tb_env->flags & PPC_TIMER_BOOKE) {
+    now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+    n = ns_to_tb(tb_env->decr_freq, now);
+    if (next > n && tb_env->flags & PPC_TIMER_BOOKE) {
         decr = 0;
-    }  else {
-        decr = -ns_to_tb(tb_env->decr_freq, -diff);
+    } else {
+        decr = next - n;
     }
+
     trace_ppc_decr_load(decr);
 
     return decr;
@@ -857,13 +858,18 @@ static void __cpu_ppc_store_decr(PowerPCCPU *cpu, uint64_t *nextp,
         (*lower_excp)(cpu);
     }
 
-    /* Calculate the next timer event */
+    /*
+     * Calculate the next decrementer event and set a timer.
+     * decr_next is in timebase units to keep rounding simple. Note it is
+     * not adjusted by tb_offset because if TB changes via tb_offset changing,
+     * decrementer does not change, so not directly comparable with TB.
+     */
     now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
-    next = now + tb_to_ns_round_up(tb_env->decr_freq, value);
+    next = ns_to_tb(tb_env->decr_freq, now) + value;
     *nextp = next;
 
     /* Adjust timer */
-    timer_mod(timer, next);
+    timer_mod(timer, tb_to_ns_round_up(tb_env->decr_freq, next));
 }
 
 static inline void _cpu_ppc_store_decr(PowerPCCPU *cpu, target_ulong decr,
@@ -1177,12 +1183,15 @@ static void start_stop_pit (CPUPPCState *env, ppc_tb_t *tb_env, int is_excp)
     } else {
         trace_ppc4xx_pit_start(ppc40x_timer->pit_reload);
         now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
-        next = now + tb_to_ns_round_up(tb_env->decr_freq,
-                                       ppc40x_timer->pit_reload);
-        if (is_excp)
-            next += tb_env->decr_next - now;
+
+        if (is_excp) {
+            tb_env->decr_next += ppc40x_timer->pit_reload;
+        } else {
+            tb_env->decr_next = ns_to_tb(tb_env->decr_freq, now)
+                                + ppc40x_timer->pit_reload;
+        }
+        next = tb_to_ns_round_up(tb_env->decr_freq, tb_env->decr_next);
         timer_mod(tb_env->decr_timer, next);
-        tb_env->decr_next = next;
     }
 }
 
-- 
2.41.0