On Thu, Mar 26, 2009 at 05:55:00PM +0100, Daniel Veillard wrote:
On Thu, Mar 26, 2009 at 01:19:12PM +0000, Daniel P. Berrange wrote:
On Tue, Mar 24, 2009 at 03:29:21PM -0400, Cole Robinson wrote: [...] Having read the man page again, I'm inclined to say using st_blksize is always wrong no matter what, because it is quite clear that 'st_blocks' is always in 512 byte units. So perhaps we might be better of doing
#ifndef DEV_BSIZE #define DEV_BSIZE 512 #endif
And then always using DEV_BSIZE.
In those kind of cases I go down to the spec and it states (in the informative section though):
----------------------------------------- http://www.opengroup.org/onlinepubs/000095399/basedefs/sys/stat.h.html "The unit for the st_blocks member of the stat structure is not defined within IEEE Std 1003.1-2001. In some implementations it is 512 bytes. It may differ on a file system basis. There is no correlation between values of the st_blocks and st_blksize, and the f_bsize (from <sys/statvfs.h>) structure members.
Traditionally, some implementations defined the multiplier for st_blocks in <sys/param.h> as the symbol DEV_BSIZE." -----------------------------------------
So I agree with Dan, we need to drop st_blksize in any volume size computation, and fallback to 512 if not defined, apparently only src/storage_backend.c referenced it :-)
I still find the "It may differ on a file system basis" to be a bit frightening considering the sandard doesn't seems to indicate how to extract that information from the filesystem :-( , oh well ...
I guess if we come across a filesystem where it is not 512, then someone will have created an ioctl() or equivalent to let us find out the true value. Failing that, relying on 512 seems like a good plan. Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|