On Fri, Mar 06, 2026 at 04:22:16PM -0400, Jason Gunthorpe wrote:

On Fri, Mar 06, 2026 at 12:18:24PM -0800, Nicolin Chen wrote:

quoted

IIUIC, we would have two timeouts to identify the device(s), so we
wouldn't need to give away the optimization of batching ATCI cmds?

Yes we should not harm success path performance to compensate for a
RAS error case.

quoted

Will letting a faulty device time out once again give it a window
to corrupt the memory?

No, you are doing this within the flush callback so everything relying
on that flush should remain held.

Oh, right! I didn't see that coming..

I will rebase the next version on the invalidation series as we
have multiple things relying on that now.

Thanks
Nicolin