On Sun, 17 Feb 2002, Matthew Dillon wrote:
> Whoop! I take it back. I'm still getting the errors:
>
> microuptime() went backwards (458.168990 -> 458.168882)
> microuptime() went backwards (578.609995 -> 577.929801)
> microuptime() went backwards (748.912755 -> 748.237402)
> microuptime() went backwards (775.159625 -> 775.159612)
>
> I also think this retry loop has to be done everywhere where the
> timecounter structure is accessed directly.
Yes, since the reads of all the relevant timecounter variables are
non-atomic.
> Index: kern/kern_tc.c
> ===================================================================
> RCS file: /home/ncvs/src/sys/kern/kern_tc.c,v
> retrieving revision 1.113
> diff -u -r1.113 kern_tc.c
> --- kern/kern_tc.c 7 Feb 2002 21:21:55 -0000 1.113
> +++ kern/kern_tc.c 17 Feb 2002 20:41:47 -0000
> @@ -126,8 +128,10 @@
> struct timecounter *tc;
>
> ngetmicrotime++;
> - tc = timecounter;
> - *tvp = tc->tc_microtime;
> + do {
> + tc = timecounter;
> + *tvp = tc->tc_microtime;
> + } while (tc != timecounter);
> }
>
> void
E.g., tc_mictrotime here is a timeval. It doesn't matter getting a stale
value (although getting a stale value increases the possible incoherency
of the get*() functions from 1/HZ to NTIMECOUNTER/HZ), but getting a
stale value that changed underneath the read would be bad.
Bruce
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
