Le 03/04/2020 à 20:01, Linus Torvalds a écrit :
On Fri, Apr 3, 2020 at 12:21 AM Christophe Leroy
<christophe.leroy@xxxxxx> wrote:
Now we have user_read_access_begin() and user_write_access_begin()
in addition to user_access_begin().
I realize Al asked for this, but I don't think it really adds anything
to the series.
The "full" makes the names longer, but not really any more legible.
So I like 1-4, but am unconvinced about 5 and would prefer that to be
dropped. Sorry for the bikeshedding.
Yes I was not sure about it, that's the reason why I added it as the
last patch of the series.
And in the meantime, we see Robots reporting build failures due to
additional use of user_access_begin() in parallele to this change, so I
guess it would anyway be a challenge to perform such a change without
coordination.
And I like this series much better without the cookie that was
discussed, and just making the hard rule be that they can't nest.
Some architecture may obviously use a cookie internally if they have
some nesting behavior of their own, but it doesn't look like we have
any major reason to expose that as the actual interface.
The only other question is how to synchronize this? I'm ok with it
going through the ppc tree, for example, and just let others build on
that. Maybe using a shared immutable branch with 5.6 as a base?
Michael, can you take patches 1 to 4 ?
Otherwise, can you ack patch 4 to enable merging through another tree ?
Christophe
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
