In article <20050216230852.GB3918@antioche.eu.org>,
Manuel Bouyer  <bouyer@antioche.eu.org> wrote:
>-=-=-=-=-=-
>
>Hi,
>with the attached test program, I stat some disk device. Sometimes I
>get the real size of the parition, sometimes I get 0:
>rochebonne# disklabel wd2
>[...]
>8 partitions:
>#        size    offset     fstype [fsize bsize cpg/sgs]
> d: 156301488         0     unused      0     0        # (Cyl.      0 - 155060)
> e:  59999184         0     4.4LFS      0     0     0  # (Cyl.      0 -  59522)
> f:  59999184  59999184     4.2BSD   2048 16384 28808  # (Cyl.  59523 - 119045)
> g:  36303120 119998368     4.2BSD      0     0     0  # (Cyl. 119046 - 155060)
>rochebonne# /tmp/tstat /dev/rwd2d
>dev 0x0 rdev 0x313 size 80026361856 blocks 65536
>rochebonne# /tmp/tstat /dev/rwd2e
>dev 0x0 rdev 0x314 size 30719582208 blocks 65536
>rochebonne# /tmp/tstat /dev/rwd2f
>dev 0x0 rdev 0x315 size 0 blocks 65536
>rochebonne# /tmp/tstat /dev/rwd2g
>dev 0x0 rdev 0x316 size 0 blocks 65536
>rochebonne# /tmp/tstat /dev/wd2g
>dev 0x0 rdev 0x16 size 0 blocks 2048
>rochebonne# /tmp/tstat /dev/wd2f
>dev 0x0 rdev 0x15 size 30719582208 blocks 2048
>rochebonne# /tmp/tstat /dev/wd2e
>dev 0x0 rdev 0x14 size 30719582208 blocks 2048
>rochebonne# /tmp/tstat /dev/wd2d
>dev 0x0 rdev 0x13 size 0 blocks 2048
>
>Any reason for such a behavior ?
>
>-- 
>Manuel Bouyer <bouyer@antioche.eu.org>
>     NetBSD: 26 ans d'experience feront toujours la difference
>--
>
>-=-=-=-=-=-
>
>#include <sys/stat.h>
>
>main(int argc, char **argv)
>{
>	struct stat sb;
>
>	if (stat(argv[1], &sb) < 0) {
>		perror("stat");
>		exit(1);
>	}
>	printf("dev 0x%x rdev 0x%x size %lld blocks %d\n",
>	    (int)sb.st_dev, (int)sb.st_rdev, (long long)sb.st_size,
>	    (int)sb.st_blksize);
>}

Looks like a bug to me. It gets the right size if the vnode is cached.

so doing:

    dd if=/dev/foo of=/dev/null count=1

fixes it.

christos