[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v2 2/2] kexec: remove spinlock now that all KEXEC hypercall ops are protected at the top-level



On Wed, Apr 19, 2017 at 05:20:50AM -0600, Jan Beulich wrote:
> >>> On 19.04.17 at 12:56, <daniel.kiper@xxxxxxxxxx> wrote:
> > On Tue, Apr 18, 2017 at 04:49:48AM -0600, Jan Beulich wrote:
> >> >>> On 17.04.17 at 21:09, <eric.devolder@xxxxxxxxxx> wrote:
> >> > The spinlock in kexec_swap_images() was removed as
> >> > this function is only reachable on the kexec hypercall, which is
> >> > now protected at the top-level in do_kexec_op_internal(),
> >> > thus the local spinlock is no longer necessary.
> >>
> >> But perhaps leave an ASSERT() there, making sure the in-hypercall
> >> flag is set?
> >
> > I am not sure why but if at all I think that we should also consider
> > other key kexec functions then. Or put ASSERT() into do_kexec_op_internal()
> > just before "switch ( op )".
>
> The point of my placement suggestion was that the ASSERT()
> effectively replaces the lock acquire - the places you name
> didn't previously require any synchronization.

After the first patch of this series kexec_swap_images() cannot be
started twice in parallel. So, I do not see the point of ASSERT() here.
Or let's say we wish to have it to double check that "the in-hypercall
flag is set". AIUI, it is your original idea. However, then I think that
we should have an ASSERT() at least in kexec_load_slot() because parallel
loads make issues too. And we can go higher to feel more safe. That is
why I suggested do_kexec_op_internal() as the final resting place for
an ASSERT(). Simply it looks to me the safest approach. Am I missing
something?

Daniel

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

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.