On Tue, Jan 16, 2018 at 03:35:20PM -0200, Eduardo Habkost wrote:
On Fri, Jan 12, 2018 at 08:31:16PM +0100, Kashyap Chamarthy wrote:
Currently, the CPU feature 'name' XML attribute, as in:
[...]
--- docs/formatdomain.html.in | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)
diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index d272cc1ba..e717fb3aa 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -1454,6 +1454,23 @@
<span class="since">Since 0.8.5</span> the <code>policy</code> attribute can be omitted and will default to <code>require</code>. + + Individual CPU feature names can be specified as part of the + <code>name</code> attribute.
Isn't this "should" instead of "can"? Does it make sense to have a 'feature' element without a 'name' attribute?
Good catch. Near as I see, it doesn't. So I'll: s/can/should.
The list of known CPU feature + names (e.g. 'vmx', 'cmt', et cetera) can be found in the same + file as CPU models -- <code>cpu_map.xml</code>. For example, + to explicitly specify the 'pcid' feature with Intel IvyBridge + CPU model:
Another paragraph above already says "The list of known feature names can be found in the same file as CPU models". If you think the existing paragraph is not enough, I suggest rewriting it so the document won't repeat exactly the same thing.
True. How about this rewrite: "Once you choose a feature (e.g. 'pcid') from the `cpu_map.xml`, to specify it explicitly with the Intel IvyBridge CPU model [...]" I'll consider whether to also add a note that before specifying extra CPU feature flags, one should check if the named CPU models provided by libvirt already include the said flags. [...] -- /kashyap