Re: [PATCH 1/5] common: don't (kind of) open-code rcu_lock_domain_by_any_id()

[Jan Beulich <jbeulich@xxxxxxxx>]

On 05.01.2021 14:38, Andrew Cooper wrote:
> On 05/01/2021 13:24, Jan Beulich wrote:
>> Even more so when using rcu_lock_domain_by_id() in place of the more
>> efficient rcu_lock_current_domain().
>>
>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> 
> Acked-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>

Thanks.

>> ---
>> Besides get_pg_owner(), gnttab_copy_lock_domain() has similar strange(?)
>> behavior: They accept DOMID_SELF, but not the resolved ID of the caller.
> 
> I queried that behaviour years and years ago.
> 
> According to Keir, it is part of the attempt to force guests to not know
> their own domid.  Xen itself does (or did) its very hardest to prevent a
> domU knowing its own domid, while at the same time the "domid" key in
> xenstore is mandatory for setting up PV rings.
> 
> I'd personally think was a short sighted move.

Let me make another patch then, unless you can see reasons we should
stick to the current behavior. Figuring out its own domid is possible
for a domain, after all, and even shouldn't be overly difficult.

Jan