Messages in this thread | | | Subject | Re: 64-bit shm: abi problems | From | Christoph Rohland <> | Date | 15 Dec 1999 09:18:51 +0100 |
| |
Linus Torvalds <torvalds@transmeta.com> writes:
> On Tue, 14 Dec 1999, David S. Miller wrote: > > > > Yes I see someone thought it would be a keen idea to > > change shm_segsz to size_t from int. Why? I'm quite > > happy with a 2G limit; you should have been using mmap > > well before that anyway. > > > > mmap() still doesn't support anonymous SHARED+WRITE, which is why > > people want sysv/ipv SHM. > > We really should try to fix the anon shared mmap issue, it should be > possible these days reasonably easily (it's really mostly an issue of > making sure that we can allocate "swapper_inode" inodes on the fly, with > the one special swapper_inode remaining for regular paging).
But keep in mind that mmap SHARED+WRITE does not give you everything (In fact it does not give you much at all, since you cannot unmap und mmap the same region during a run).
I am just now working on shm code using struct files for shmat. This would make shmat a special case of mmap. Also shm_open would be trivial (and is in fact what I am heading for). shm_open gives you a file descriptor to a shared memory object which you then can mmap.
My code is progressing now. I have most of the dentry/inode handling ready and have now to debug the mmap case.
Greetings Christoph
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |