[T10] Should this WRITE STREAM error response be changed?

Gerry Houlder gerry.houlder at seagate.com
Mon Nov 14 08:51:32 PST 2016


Hi,

i received a comment that NVMe Write Stream error handling has a difference
with respect to SAS error handling and there is a request to make them the
same in this regard.

With SAS, if the target device receives a Write Stream command with a
Stream_ID that is not valid then the target is required to reject the
command with CHECK CONDITION and the host has to figure out how it wants to
retry the command.

With NVMe, if the device receives a Write Stream command with a Stream_ID
that is not valid then the device handles the Write Stream command like it
is a regular non-stream write command. This means the Stream_ID is ignored
and the write proceeds normally without the hint of the Stream_ID. This is
recommended as friendlier to the host in that the overhead of handling a
rejected command and retrying the command is avoided.

Should SBC-4 be changed to specify that an invalid Stream_ID is ignored the
the write command proceeds as a non-streaming write command? I'd like to
hears some other opinions (either for or against) before I write a proposal
along these lines.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.t10.org/pipermail/t10/attachments/20161114/1097fd71/attachment.html>


More information about the T10 mailing list