On Jan 3, 2009, at 10:13 AM, M. Warner Losh wrote: > : > But my problems are with a platform that has BSD labels, and not > GPT > : > labels, with a proper BSD label on the device. > : > : You don't have GEOM_BSD to service the DIOCGDINFO ioctl(2) > : anymore. We need to change fsck(8) so that it doesn't work > : only on BSD disk labels, but instead obtains the partition > : type (if applicable) in a more generic way so that it also > : applies to other partitioning schemes. > > Agreed. Do you have a proposal for how to do that? The best way to do it is to introduce a function that returns the file system name/type. It could be used by fsck(8), newfs(8), mount(8), etc to determine which variant to invoke. This function is best defined in libgeom, where it can get all the data it needs from the XML. geom_part puts it there anyway. Doing it in userland also makes it easier to actual read the partition and see if there's already a file system. It's one thing to use the partition type as a best guess, but lacking a partitioning scheme on the disk, it may be beneficial to pursue other avenues. Not to mention that it's better to return the actual file system that's there than some guess based on partition type. Thus: I'd like to eliminate the ioctl and do all or most of the work in user space. Thoughts? -- Marcel Moolenaar xcllnt_at_mac.comReceived on Sat Jan 03 2009 - 18:46:18 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:39 UTC