From mboxrd@z Thu Jan 1 00:00:00 1970 From: james.smart@broadcom.com (James Smart) Date: Mon, 10 Jul 2017 16:53:08 -0700 Subject: [PATCH 5/7] nvme: Only one AEN request In-Reply-To: <1499444581-28268-6-git-send-email-keith.busch@intel.com> References: <1499444581-28268-1-git-send-email-keith.busch@intel.com> <1499444581-28268-6-git-send-email-keith.busch@intel.com> Message-ID: On 7/7/2017 9:22 AM, Keith Busch wrote: > This driver issues only one AEN request, so all the logic to handle > multiple ones is unnecessary. Even if we did allow multiple ones, the > implementation is broken as it will create command id collisions. For > those two reasons, this patch deletes all the aer index usage, assuming > there's only one. > > As a consequence of removing the event limit, the core's callback for > these events is simplified to only handle the successful case. The > "abort" status handling for AER has been broken for some time anyway: > it was meant to reclaim the aer count from a driver cancelled command > after a reset, but that was relying on blk-mq to manage the reserved > tags. We don't use blk-mq tags for AER, though, so we will never see a > synthesized abort status. > Looks good Reviewed-by: James Smart