mi: nvme_mi_mi_xfer() is overly restrictive #1027
Conversation
MI commands may be unaligned and also have payloads that are being sent and received at the same command. One case of this is AEM Sync. This wasn't seen before because that particular path does not use nvme_mi_mi_xfer(). Signed-off-by: Chuck Horkin <chorkin@microsoft.com>
chorkin
force-pushed
the
user/chorkin/mi_over_restrictive
branch
from
56e8590 to
3d60087
Compare
|
@jk-ozlabs , I realized that these checks should have been removed when we introduced the AEMs since we found out that they are not required (and actually prevent AEM flow from working right). This is just to clean up the pass through path. |
|
Looks good, but just to clarify - how does the AEM path need to use passthrough? |
It doesn't, but when I put the AEM SYNC path in (an MI config set command), we removed those (or didn't include these) restrictions. I didn't think about other paths that are also technically MI commands, which is why this wasn't removed earlier. |
|
Right, gotchya! Thanks for the explanation, all looks good. |
jk-ozlabs
left a comment
jk-ozlabs
left a comment
There was a problem hiding this comment.
LGTM, thanks.
Also confirming that the docs for nvme_mi_mi_xfer do not need updating, as they do not prescribe these restrictions.
|
@igaw can we merge this? |
|
Thanks! |
3 participants
MI commands may be unaligned and also have payloads that are being sent and recieved at the same command. One case of this is AEM Sync. This wasn't seen before because that particular path does not use nvme_mi_mi_xfer().