[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PATCH v2 07/19] hw/ppc: Avoid decrementer rounding errors
From: |
Nicholas Piggin |
Subject: |
[PATCH v2 07/19] hw/ppc: Avoid decrementer rounding errors |
Date: |
Tue, 8 Aug 2023 14:19:49 +1000 |
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>
---
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 13eb45f4b7..a397820d9c 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;
}
}
--
2.40.1
- [PATCH v2 for-8.2 00/19] ppc: record-replay enablement and fixes, Nicholas Piggin, 2023/08/08
- [PATCH v2 01/19] ppc/vhyp: reset exception state when handling vhyp hcall, Nicholas Piggin, 2023/08/08
- [PATCH v2 02/19] ppc/vof: Fix missed fields in VOF cleanup, Nicholas Piggin, 2023/08/08
- [PATCH v2 03/19] hw/ppc/ppc.c: Tidy over-long lines, Nicholas Piggin, 2023/08/08
- [PATCH v2 04/19] hw/ppc: Introduce functions for conversion between timebase and nanoseconds, Nicholas Piggin, 2023/08/08
- [PATCH v2 05/19] host-utils: Add muldiv64_round_up, Nicholas Piggin, 2023/08/08
- [PATCH v2 06/19] hw/ppc: Round up the decrementer interval when converting to ns, Nicholas Piggin, 2023/08/08
- [PATCH v2 07/19] hw/ppc: Avoid decrementer rounding errors,
Nicholas Piggin <=
- [PATCH v2 08/19] target/ppc: Sign-extend large decrementer to 64-bits, Nicholas Piggin, 2023/08/08
- [PATCH v2 09/19] hw/ppc: Always store the decrementer value, Nicholas Piggin, 2023/08/08
- [PATCH v2 10/19] target/ppc: Migrate DECR SPR, Nicholas Piggin, 2023/08/08
- [PATCH v2 11/19] hw/ppc: Reset timebase facilities on machine reset, Nicholas Piggin, 2023/08/08
- [PATCH v2 12/19] hw/ppc: Read time only once to perform decrementer write, Nicholas Piggin, 2023/08/08
- [PATCH v2 13/19] target/ppc: Fix CPU reservation migration for record-replay, Nicholas Piggin, 2023/08/08
- [PATCH v2 14/19] target/ppc: Fix timebase reset with record-replay, Nicholas Piggin, 2023/08/08
- [PATCH v2 15/19] spapr: Fix machine reset deadlock from replay-record, Nicholas Piggin, 2023/08/08