Messages in this thread | | | Subject | Re: splice vs O_APPEND | From | Miklos Szeredi <> | Date | Thu, 09 Oct 2008 18:04:10 +0200 |
| |
On Thu, 9 Oct 2008, Linus Torvalds wrote: > I do wonder if we shouldn't just do this in rw_verify_area(). The whole > reason for that function is that we used to have all those flock checks > etc spread out all over, and some path would inevitably just miss one > check or another. It's kind of stupid to expect low-level filesystems to > do the IS_APPEND/IS_IMMUTABLE checks.
Do we expect them? I thought we don't care if it's marked immutable or append-only after the file has been opened, same as with normal permissions.
> Comments?
Your patch still ignores O_APPEND, is that what we want? It sounds sort of strange. pwrite() for example honors O_APPEND and ignores the position, AFAICS.
Miklos
| |