On Wed, Oct 14, 2026 at 09:53:47AM +0300, Leon Romanovsky wrote:
On Tue, Oct 13, 2026 at 05:25:08PM +0300, Leon Romanovsky wrote:
PCI P2PDMA applies Request and Completion Redirect throughout both paths. This misclassifies asymmetric and nested switches, and reports one answer for every kind of TLP.
Three ACS controls act on TLP attributes the Requester chooses rather than on the topology: Translation Blocking and Direct Translated P2P act on a Request's Address Type, and Completion Redirect skips Completions carrying Relaxed Ordering.
Evaluate each direction at the path divergence, decide every class from the one walk, and treat a client with ATS enabled as translating unless its driver declares per-mapping ATS.
This completes the P2PDMA side; dma-buf and mlx5 follow separately.
Signed-off-by: Leon Romanovsky leonro@nvidia.com
Changes in v11:
Bjorn,
Could you please save me from the AI-bot attack? Every rebase seems to uncover another bizarre pre-existing issue. It never ends.
Can you elaborate? I'm not sure what you'd like me to do here. I don't control the Sashiko reviews, and I thought Sashiko was recently changed to put pre-existing issues in a database but not in the email responses.
Bjorn