Bug 12808
| Summary: | [PATCH] vmstat -i output wraps to negative values | ||
|---|---|---|---|
| Product: | Base System | Reporter: | toasty <toasty> |
| Component: | bin | Assignee: | Nick Hibma <n_hibma> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | ||
| Priority: | Normal | ||
| Version: | 4.0-CURRENT | ||
| Hardware: | Any | ||
| OS: | Any | ||
> # vmstat -i
> interrupt total rate
> fxp0 irq11 1986997576 618
> ata-pci0 irq14 43093267 13
> ata-pci0 irq15 85069735 26
> pn0 irq10 1083789644 337
> ahc0 irq9 433666 0
> clk irq0 321001949 99
> rtc irq8 410878329 127
> Total -363703130 -113
What's the uptime for this machine? :-)
> The 'total' variable is only a signed long. Could this possibly be changed
> to an unsigned long or even a long long?
It definitely should be unsigned, so overflows will make it count from 0
and up.
> >Fix:
Reboot more often, for example through a crontab?
2 2 1 1 * /sbin/reboot
Or apply the following patch. This converts the values into unsigned
long. A better solution would be to convert the values in the kernel and
in vmstat.c to quads I guess.
--- vmstat.c.old Wed May 19 15:20:54 1999
+++ vmstat.c Mon Jul 26 09:27:26 1999
@@ -765,7 +765,7 @@
void
dointr()
{
- register long *intrcnt, inttotal, uptime;
+ register unsigned long *intrcnt, inttotal, uptime;
register int nintr, inamlen;
register char *intrname;
@@ -784,12 +784,12 @@
nintr /= sizeof(long);
while (--nintr >= 0) {
if (*intrcnt)
- (void)printf("%-12s %8ld %8ld\n", intrname,
+ (void)printf("%-12s %8lu %8lu\n", intrname,
*intrcnt, *intrcnt / uptime);
intrname += strlen(intrname) + 1;
inttotal += *intrcnt++;
}
- (void)printf("Total %8ld %8ld\n", inttotal, inttotal /
uptime);
+ (void)printf("Total %8lu %8lu\n", inttotal, inttotal /
uptime);
}
#define MAX_KMSTATS 200
(And the same patch uuencoded)
begin 644 vmstat.c.patch
M+2TM('9M<W1A="YC+F]L9`E7960@36%Y(#$Y(#$U.C(P.C4T(#$Y.3D**RLK
M('9M<W1A="YC"4UO;B!*=6P@,C8@,#DZ,C<Z,C8@,3DY.0I`0"`M-S8U+#<@
M*S<V-2PW($!`"B!V;VED"B!D;VEN='(H*0H@>PHM"7)E9VES=&5R(&QO;F<@
M*FEN=')C;G0L(&EN='1O=&%L+"!U<'1I;64["BL)<F5G:7-T97(@=6YS:6=N
M960@;&]N9R`J:6YT<F-N="P@:6YT=&]T86PL('5P=&EM93L*(`ER96=I<W1E
M<B!I;G0@;FEN='(L(&EN86UL96X["B`)<F5G:7-T97(@8VAA<B`J:6YT<FYA
M;64["B`*0$`@+3<X-"PQ,B`K-S@T+#$R($!`"B`);FEN='(@+ST@<VEZ96]F
M*&QO;F<I.PH@"7=H:6QE("@M+6YI;G1R(#X](#`I('L*(`D):68@*"II;G1R
M8VYT*0HM"0D)*'9O:60I<')I;G1F*"(E+3$R<R`E.&QD("4X;&1<;B(L(&EN
M=')N86UE+`HK"0D)*'9O:60I<')I;G1F*"(E+3$R<R`E.&QU("4X;'5<;B(L
M(&EN=')N86UE+`H@"0D)("`@("II;G1R8VYT+"`J:6YT<F-N="`O('5P=&EM
M92D["B`)"6EN=')N86UE("L]('-T<FQE;BAI;G1R;F%M92D@*R`Q.PH@"0EI
M;G1T;W1A;"`K/2`J:6YT<F-N="LK.PH@"7T*+0DH=F]I9"EP<FEN=&8H(E1O
M=&%L("`@("`@("`E.&QD("4X;&1<;B(L(&EN='1O=&%L+"!I;G1T;W1A;"`O
M('5P=&EM92D["BL)*'9O:60I<')I;G1F*")4;W1A;"`@("`@("`@)3AL=2`E
M.&QU7&XB+"!I;G1T;W1A;"P@:6YT=&]T86P@+R!U<'1I;64I.PH@?0H@"B`C
79&5F:6YE"4U!6%]+35-40513"3(P,`IT
`
end
Nick
--
ISIS/STA, T.P.270, Joint Research Centre, 21020 Ispra, Italy
Responsible Changed From-To: freebsd-bugs->n_hibma remind me to commit the patch after review State Changed From-To: open->closed Committed. Thanks!. > > # vmstat -i > > interrupt total rate > > fxp0 irq11 1986997576 618 > > ata-pci0 irq14 43093267 13 > > ata-pci0 irq15 85069735 26 > > pn0 irq10 1083789644 337 > > ahc0 irq9 433666 0 > > clk irq0 321001949 99 > > rtc irq8 410878329 127 > > Total -363703130 -113 > > What's the uptime for this machine? :-) > su-2.03# uname -a FreeBSD nfs.dragondata.com 4.0-CURRENT FreeBSD 4.0-CURRENT #2: Thu May 6 00:18:24 CDT 1999 toasty@nfs.dragondata.com:/usr/src/sys/compile/NFS i386 su-2.03# w 2:11PM up 38 days, 2:02, 2 users, load averages: 1.60, 1.64, 1.65 38 days isn't an reasonable time to wraparound. :) Even worse, is another machine of mine: # vmstat -i interrupt total rate clk0 irq0 -1099210494 -34 rtc0 irq8 -204812585 -6 fdc0 irq6 1 0 wdc0 irq14 184631379 5 sc0 irq1 254 0 ep0 irq10 -758217127 -23 Total -1877608572 -58 # w 2:12PM up 369 days, 20:15, 1 user, load averages: 0.38, 0.37, 0.33 Everything else appears ok in vmstat though. :) > > >Fix: > > Reboot more often, for example through a crontab? > > 2 2 1 1 * /sbin/reboot > Nah, 38 day uptimes aren't something to reboot out of. :) > > Or apply the following patch. This converts the values into unsigned > long. A better solution would be to convert the values in the kernel and > in vmstat.c to quads I guess. > Quads would be nice, especially considering my 369 day uptime box. :) Thanks, Kevin |
The output of vmstat -i will wrap around to negative numbers. # vmstat -i interrupt total rate fxp0 irq11 1986997576 618 ata-pci0 irq14 43093267 13 ata-pci0 irq15 85069735 26 pn0 irq10 1083789644 337 ahc0 irq9 433666 0 clk irq0 321001949 99 rtc irq8 410878329 127 Total -363703130 -113 The 'total' variable is only a signed long. Could this possibly be changed to an unsigned long or even a long long? Fix: In vmstat.c: void dointr() { register long *intrcnt, inttotal, uptime; inttotal should be changed to an unsigned long, if not a larger datatype.