Re: [Xen-devel] [PATCH] Fix performance problems with mprotect()

["Bruce Rogers" <brogers@xxxxxxxxxx>]

I believe -ENOSYS, in this case at least, correctly identifies when the new 
code is not there.  I didn't try to bump any revision, or add an explicit 
feature flag somewhere.

I would have preferred to do a one time detection by identity remapping some 
page and see if it succeeds and then just test a global flag to determine 
whether to use the new method or not, but it appeared to me at least to not be 
the Linux way (I am still somewhat new to Linux)  and was unsure how issues 
such as migrating the domain to hypervisors which might be missing this feature 
is handled, so I went with this approach of always being able to fall back to 
the previously existing method.

Ugly, I agree, but it works.

I'll look into the APIChangelog entry after more feedback rolls in.

- Bruce

>>> John Levon <levon@xxxxxxxxxxxxxxxxx> 01/05/08 6:35 AM >>>
On Fri, Jan 04, 2008 at 09:49:41PM -0700, Bruce Rogers wrote:

> While working on a database scaling problem 

The changelog entry for this patch should have a suitable entry for
http://wiki.xensource.com/xenwiki/APIChangelog, as suggested by Ian
Jackson.

(Regarding the changes, is there a sane way to detect whether the new
API is present or not? IE a sensible unique errno return or something
else.)

regards
john

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel