Al Viro <viro@xxxxxxxxxxxxxxxxxx> writes: > On Thu, Oct 23, 2014 at 05:01:06PM +0100, Al Viro wrote: > >> Hmm... Why hadn't d_splice_alias() caught that, though? > > Aha. It's not namei_msdos.c part, it's namei_vfat.c one. And there > we don't call d_splice_alias() on the affected path... > > OK, so your check isn't enough. What we need there is this: > if (alias && alias->d_parent == dentry->d_parent && ...) > Otherwise that d_move() isn't safe at all. Hm, sounds like I'm missing something. what case has different ->d_parent on alias if prevented by my check? > Moreover, for directories we don't want to bother with that codepath > at all - d_splice_alias() will do that d_move() just fine there. d_splice_alias() calls __d_find_alias() with want_discon==1, so __d_find_alias() doesn't return dentry, and d_splice_alias() doesn't use d_move() path, right? -- OGAWA Hirofumi <hirofumi@xxxxxxxxxxxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
