Hi Jack,
On Thu, Sep 11, 2025 at 02:46:43PM +0100, Jack Thomson wrote:
From: Jack Thomson jackabt@amazon.com
Adding __gmem_abort and __user_mem_abort that preserve -EAGAIN results. These will be used by the pre-fault implementation which needs to retry on -EAGAIN.
-EAGAIN is a pretty clear signal that another vCPU has faulted on this memory and is in the middle of installing a mapping. Why bother with retrying?
If we conceptually treat this thing as a synthetic stage-2 abort then it should use the same EAGAIN handling as a literal stage-2 abort.
Thanks, Oliver