Re: [PATCH 1/6] block: ensure we don't truncate top bits of the request command flags

From: Jens Axboe
Date: Fri Mar 25 2016 - 00:19:36 EST

On 03/24/2016 08:08 PM, Mike Christie wrote:
On 03/22/2016 02:01 PM, Jens Axboe wrote:
On 03/22/2016 12:59 PM, Christoph Hellwig wrote:
On Tue, Mar 22, 2016 at 11:55:15AM -0600, Jens Axboe wrote:
Some of the flags that we want to use from the make_request_fn path
are now larger than 32-bit, so change the functions involved to
accept an u64 instead of an unsigned int.

When did we start doing that? We really should merge Mike's split
of the operation style flags into the cmd_type before making things
even worse in the flags area.

Just now, and I ran into it last week as well, for a test patch on cfq
that passed in higher flags for get_request -> may_queue() as well. We
can do Mike's split first, I think it's a good cleanup. As a standalone
series, I needed it though.

Hey, did you want any changes on that patchset? I was going to repost it
with the kbuild fix against linux-next, but I can make any changes you
wanted first.

I don't believe I've ever been CC'ed on the posting, or it even being posted on the block list? If so, I don't see it... I did become aware of it since Christoph CC'ed me in. In general, I think it looks good, at least the end results. It's a bit murky in the middle, and the commit messages need some help. So go over everything, sanitize it, and repost it. I don't like the current pure flag based scheme we have, it's a mess of ops and modifiers. So splitting that up is definitely a good thing.

Jens Axboe