On Wed, 2017-05-03 at 15:42 +0300, Amir Goldstein wrote: > On Wed, May 3, 2017 at 2:00 PM, Andy Shevchenko > <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote: > > On Wed, 2017-05-03 at 13:05 +0300, Amir Goldstein wrote: > > > On Wed, May 3, 2017 at 12:13 PM, Andy Shevchenko > I can hear the voices of > > > those saying that there should be a 'natural' helper > > > uuid_is_null(u8 > > > *) for the > > > users that represent uuid as u8[16] (i.e. filesystems and sb- > > > >s_uuid). > > > > u8 * doesn't represent UUID as a type. > > Perhaps we need to reflect this in the name of the function somehow. > > > > OK, since you have no time nor the intention to convert filesystem > code > to use uuid_le uuid_be I suppose. I noticed core developers doesn't support endianess feature of UUID: https://lkml.org/lkml/2016/5/25/724 That's also why I don't want to touch filesystem stuff. > (not should you), I see 2 options for filesystems/VFS: > > 1. Use the simple u8* libnvdimm helper proposed by this patch > 2. Hoist uuid_t + helpers from fs/xfs/uuid.* to linux/uuid.h > lib/uuid.c > > If option #2 is preferred, I think we should conform to libuuid's > helper > name uuid_is_null() (instead of freebsd's uuid_is_nil()) I think option #2 is preferred and we need actually to submit some generic uuid_{be|le}_cmp() helpers. For now it's blocked by Gcc bug. We may remove const from the prototypes for now and add them later when the bug will be fixed. > and maybe > convert libnvdimm and bluetooth to use the xfs helpers as well. > > Christoph, > > Since you got me started on this helper, do you have a preference? -- Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> Intel Finland Oy
