From: Alan Cox <alan@linux.intel.com>
Date: Thu, 20 Sep 2018 00:54:55 +0000 (+0100)
Subject: timer: handle fast clocks
X-Git-Url: https://git.ndcode.org/public/gitweb.cgi?a=commitdiff_plain;h=d4a8f0b41b5d7723446afa524a35adb93d8cad6f;p=FUZIX.git

timer: handle fast clocks

On a few systems the timer tick can run too fast (eg it's 0.512 of a second
not 0.5) and we can tick ahead of the RTC time. For robustness treat being
a second ahead of locked time as we do being on time - don't tick again so
we'll slide into time.

We shouldn't end up any further out because the big jumps are only when we
have interrupts off for stuff like floppy disk I/O. In that situation we
will always lose time.
---

diff --git a/Kernel/timer.c b/Kernel/timer.c
index b0ad2ecf..78110943 100644
--- a/Kernel/timer.c
+++ b/Kernel/timer.c
@@ -112,10 +112,11 @@ void updatetod(void)
 	slide = rtcnew - rtcsec;	/* Seconds elapsed */
 	rtcsec = rtcnew;
 	/*
-	 *	FIXME: it would be nice if we see a backwards slide of 1 or 2
-	 *	to stall the OS clock so we can use an rtc to track a not so
-	 *	stable system clock.
+	 *	We assume a small negative warp in seconds is telling us
+	 *	that the clock is running too fast and we should stall.
 	 */
+	if (slide == -1)
+		return;
 addtod:
 	if (slide < 0)
 		slide += 60;		/* Seconds wrapped */