wrapper: trace-clock: remove compatibility code
authorMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mon, 13 Apr 2020 20:45:47 +0000 (16:45 -0400)
committerMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mon, 13 Apr 2020 20:56:30 +0000 (16:56 -0400)
wrapper/trace-clock.h

index c51f8d3cf3652708dac979ad801b66ffc42683d0..df03c0fc46d047f54cfebf4b99a4a892c45ced9d 100644 (file)
 
 extern struct lttng_trace_clock *lttng_trace_clock;
 
-/*
- * Upstream Linux commit 27727df240c7 ("Avoid taking lock in NMI path with
- * CONFIG_DEBUG_TIMEKEEPING") introduces a buggy ktime_get_mono_fast_ns().
- * This is fixed by patch "timekeeping: Fix __ktime_get_fast_ns() regression".
- */
-#if (LTTNG_KERNEL_RANGE(4,8,0, 4,8,2) \
-       || LTTNG_KERNEL_RANGE(4,7,4, 4,7,8) \
-       || LTTNG_KERNEL_RANGE(4,4,20, 4,4,25) \
-       || LTTNG_KERNEL_RANGE(4,1,32, 4,1,35))
-#define LTTNG_CLOCK_NMI_SAFE_BROKEN
-#endif
-
 /*
  * We need clock values to be monotonically increasing per-cpu, which is
  * not strictly guaranteed by ktime_get_mono_fast_ns(). It is
@@ -51,9 +39,7 @@ extern struct lttng_trace_clock *lttng_trace_clock;
  * this feature on 64-bit architectures.
  */
 
-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0) \
-       && BITS_PER_LONG == 64 \
-       && !defined(LTTNG_CLOCK_NMI_SAFE_BROKEN))
+#if BITS_PER_LONG == 64
 #define LTTNG_USE_NMI_SAFE_CLOCK
 #endif
 
This page took 0.025267 seconds and 4 git commands to generate.