Messages in this thread | | | From | "Stephen C. Tweedie" <> | Date | Fri, 12 Nov 1999 17:52:41 +0000 (GMT) | Subject | Re: kiobuf using kernel pages |
| |
Hi,
On Thu, 11 Nov 1999 08:48:16 -0800 (PST), Matt Robinson <yakker@cthulhu.engr.sgi.com> said:
> There is a mechanism in the patch to pass AS_KERNEL to map_user_kiobuf() > so that the kernel pages are tagged appropriately for raw I/O. Given > that, when sd_raw_rw() is called, if the f_reada flag of the file > pointer is DUMP_KIOBUF_NUMBER, we automatically set AS_KERNEL.
I'm munging the 2.3 kiobuf/raw IO stuff at the moment, and yes, mapping kernel pages is on the agenda. However, I really really do not want to see it done this way --- it _must_ be the caller who deals with the address space stuff, not the low-level raw_rw function. The IO function should just accept an arbitrary kiobuf for the IO, no more, no less.
I'm adding a proper set of documentation on the kiobuf IO model in the current patches to show how it is designed to be used. Getting the abstraction right is really important here: it not only results in cleaner code, it will also enable the IO functions to be reused flexibly.
--Stephen
- 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/
| |