Grégoire Sutre wrote: > Hi, > > Regarding the nested partition code, there is an implicit assumption > that each partition should be contained in its parent, i.e. its sectors > should also be sectors of its parent. > > This ``physical nesting'' is checked in grub_disk_read, but it would > be better to check it before that. > > The attached patch discards partitions that are invalid w.r.t. physical > nesting. This solves, in particular, a problem related to NetBSD (and > OpenBSD) disklabels. There are few ramifications of this patch. First of all some partitions which are just barely outside of the host partition will lead to something like "partition not found" errors in grub-probe. This message should be more informative (the easiest way is to issue a warning in grub-probe if partitions are discarded except some cases where it's known not to affect the functionality like 'd' "subpartitions", probably such a warning in grub proper would be too annoying though). Then if you check partitions when iterating no need to recheck in adjust_range. > > With this patch, ``external'' partitions in a disklabel simply do not > show up as BSD partitions. For instance (see bug #29956 for an image): > > MBR Partition table: > 0: NetBSD start 32, size 1000 > 1: DOS start 1040, size 1000 > > NetBSD Disklabel (stored in MBR partition 0) > 5 partitions: > # size offset fstype [fsize bsize cpg/sgs] > a: 1000 32 4.2BSD > c: 1000 32 unused > d: 2048 0 unused > e: 1000 1040 MSDOS > > > The e: partition is external: it is not contained in MBR NetBSD > partition. > > Without the patch, we get: > > $ grub-probe -m /dev/null -t drive -d /dev/rvnd0e > (/dev/rvnd0d,1,5) # this is (/dev/rvnd0d,msdos1,bsd5) > $ grub-probe -m /dev/null -t fs -d /dev/rvnd0e > grub-probe: error: unknown filesystem. > > With the patch, we get: > > niagara# grub-probe -m /dev/null -t drive -d /dev/rvnd0e > (/dev/rvnd0d,2) # this is (/dev/rvnd0d,msdos2) > niagara# grub-probe -m /dev/null -t fs -d /dev/rvnd0e > fat > > > The patch still accepts sub-partitions that start at the same > (absolute) offset as the parent. For instance, in the above example, > ls -l in grub gives both (hd1,msdos1) and (hd1,msdos1,bsd1). Should > we discard (hd0,msdos1,bsd1), i.e. require that sub-partitions start > at a strictly positive relative offset? No. SUN partitions comonly start at offset 0. > > > Grégoire > > ------------------------------------------------------------------------ > > _______________________________________________ > Grub-devel mailing list > Grub-devel@gnu.org > http://lists.gnu.org/mailman/listinfo/grub-devel
-- Regards Vladimir 'φ-coder/phcoder' Serbinenko
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Grub-devel mailing list Grub-devel@gnu.org http://lists.gnu.org/mailman/listinfo/grub-devel