arm64: vdso: defer shifting of nanosecond component of timespec

Shifting the nanosecond component of the computed timespec early can
lead to sub-ns inaccuracies when using the truncated value as input to
further arithmetic for things like conversions to monotonic time.

This patch defers the timespec shifting until after the final value has
been computed.
Reported-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

arch/arm64/kernel/vdso.c

@@ -239,7 +239,7 @@ void update_vsyscall(struct timekeeper *tk)
 	if (!use_syscall) {
 		vdso_data->cs_cycle_last	= tk->clock->cycle_last;
 		vdso_data->xtime_clock_sec	= tk->xtime_sec;
-		vdso_data->xtime_clock_nsec	= tk->xtime_nsec >> tk->shift;
+		vdso_data->xtime_clock_nsec	= tk->xtime_nsec;
 		vdso_data->cs_mult		= tk->mult;
 		vdso_data->cs_shift		= tk->shift;
 		vdso_data->wtm_clock_sec	= tk->wall_to_monotonic.tv_sec;

arch/arm64/kernel/vdso/gettimeofday.S

@@ -66,6 +66,7 @@ ENTRY(__kernel_gettimeofday)
 
 	/* Convert ns to us. */
 	mov	x13, #1000
+	lsl	x13, x13, x12
 	udiv	x11, x11, x13
 	stp	x10, x11, [x0, #TVAL_TV_SEC]
 2:
@@ -136,11 +137,13 @@ ENTRY(__kernel_clock_gettime)
 4:
 	/* Add on wtm timespec. */
 	add	x10, x10, x13
+	lsl	x14, x14, x12
 	add	x11, x11, x14
 
 	/* Normalise the new timespec. */
 	mov	x15, #NSEC_PER_SEC_LO16
 	movk	x15, #NSEC_PER_SEC_HI16, lsl #16
+	lsl	x15, x15, x12
 	cmp	x11, x15
 	b.lt	5f
 	sub	x11, x11, x15
@@ -152,6 +155,7 @@ ENTRY(__kernel_clock_gettime)
 	sub	x10, x10, #1
 
 6:	/* Store to the user timespec. */
+	lsr	x11, x11, x12
 	stp	x10, x11, [x1, #TSPEC_TV_SEC]
 	mov	x0, xzr
 	ret	x2
@@ -204,7 +208,7 @@ ENDPROC(__kernel_clock_getres)
  * Clobbers the temporary registers (x9 - x15).
  * Returns:
  *  - w9		= vDSO sequence counter
- *  - (x10, x11)	= (ts->tv_sec, ts->tv_nsec)
+ *  - (x10, x11)	= (ts->tv_sec, shifted ts->tv_nsec)
  *  - w12		= cs_shift
  */
 ENTRY(__do_get_tspec)
@@ -226,11 +230,11 @@ ENTRY(__do_get_tspec)
 	movn	x15, #0xff00, lsl #48
 	and	x10, x15, x10
 	mul	x10, x10, x11
-	lsr	x10, x10, x12
 
 	/* Use the kernel time to calculate the new timespec. */
 	mov	x11, #NSEC_PER_SEC_LO16
 	movk	x11, #NSEC_PER_SEC_HI16, lsl #16
+	lsl	x11, x11, x12
 	add	x15, x10, x14
 	udiv	x14, x15, x11
 	add	x10, x13, x14