On Tue, Nov 12, 2024 at 04:38:42PM -0800, Linus Torvalds wrote: > Looking at that locking code in fadvise() just for the f_mode use does > make me think this would be a really good cleanup. > > I note that our fcntl code seems buggy as-is, because while it does > use f_lock for assignments (good), it clearly does *not* use them for > reading. > > So it looks like you can actually read inconsistent values. > > I get the feeling that f_flags would want WRITE_ONCE/READ_ONCE in > _addition_ to the f_lock use it has. AFAICS, fasync logics is the fishy part - the rest should be sane. > The f_mode thing with fadvise() smells like the same bug. Just because > the modifications are serialized wrt each other doesn't mean that > readers are then automatically ok. Reads are also under ->f_lock in there, AFAICS... Another thing in the vicinity is ->f_mode modifications after the calls of anon_inode_getfile() in several callers - probably ought to switch those to anon_inode_getfile_fmode(). That had been discussed back in April when the function got merged, but "convert to using it" followup series hadn't materialized...
