Re: [Xen-devel] -EINTR return in domain_relinquish_resources

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 23.01.15 at 18:21, <konrad.wilk@xxxxxxxxxx> wrote:
> On Fri, Jan 23, 2015 at 04:03:55PM +0000, Jan Beulich wrote:
>> >>> On 23.01.15 at 16:46, <konrad.wilk@xxxxxxxxxx> wrote:
>> > Subject: [PATCH] domain: In vcpu_destroy_pagetables we can return -ERESTART
>> >  instead of -EINTR
>> > 
>> > which has the side effect that domain_relinquish_resources will stop
>> > and return to user-space -EINTR - which it is not equipped to deal with.
>> 
>> The title read wrong, especially on its own, as it appears to
>> state the inverse thing of what you do in the patch. Perhaps
>> 
>> x86: vcpu_destroy_pagetables() must not return -EINTR
>> 
>> with the initial part of the description adjusted accordingly?
>> 
>> > +    /*
>> > +     * The put_page_and_type_preemptible is liable to return -EINTR. Other
>> > +     * callers of it filter the -EINTR to whatever they deem applicable - 
>> > in
>> > +     * this case we MUST do it as the caller of this function will return 
>> > the
>> > +     * error code to userspace. And userspace for domain destruction 
>> > expects
>> > +     * -EAGAIN (domain_relinquish_resources converts ERESTART to -EAGAIN).
>> > +     */
>> 
>> This is still misleading, as it kind of implies that the function has only
>> that one caller. Don't talk about domain_relinquish_resources() and
>> EAGAIN at all.
> 
> Right. I somehow managed to miss the other caller of 
> vcpu_destroy_pagetables.
> 
> Please see following patch:

Looks good. I don't see a reason not to apply it without you doing
a formal submission.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel