Hello, At last I get some time to get back to this
Thomas Schwinge, le Tue 17 Feb 2009 15:29:32 +0100, a écrit : > diff -u -p -r1.11 io-stat.c > --- libdiskfs/io-stat.c 12 Apr 2001 19:43:43 -0000 1.11 > +++ libdiskfs/io-stat.c 17 Feb 2009 14:27:30 -0000 > @@ -45,6 +46,9 @@ diskfs_S_io_stat (struct protid *cred, > if (cred->po->shadow_root == np || np == diskfs_root_node) > statbuf->st_mode |= S_IROOT; > > + if (diskfs_modify_stat != NULL) > + diskfs_modify_stat (statbuf); > + > mutex_unlock (&np->lock); > > return 0; Mmm, so you're rounding it just before revealing it through io_stat()? I do not like this. Some other code within libdiskfs could do the same as userland: use the fine-grain precision timestamp, and then choke when ext2fs gives back the coarse-grain precision timestamp. Why not doing it right when setting new timestamps, so that everybody gets the same value? Samuel
