From mboxrd@z Thu Jan 1 00:00:00 1970 From: krisman@linux.vnet.ibm.com (Gabriel Krisman Bertazi) Date: Wed, 29 Jun 2016 01:17:35 -0300 Subject: [PATCH 2/2] [RFC] nvme: enable asynchronous events notification by default In-Reply-To: <20160628151137.GA8607@localhost.localdomain> (Keith Busch's message of "Tue, 28 Jun 2016 11:11:39 -0400") References: <1466184075-10471-1-git-send-email-gpiccoli@linux.vnet.ibm.com> <1466184075-10471-3-git-send-email-gpiccoli@linux.vnet.ibm.com> <57679355.6000303@gmail.com> <576821E9.2060009@linux.vnet.ibm.com> <20160620202135.GG12936@localhost.localdomain> <20160624081726.GA13596@infradead.org> <20160628151137.GA8607@localhost.localdomain> Message-ID: <87r3bgsk68.fsf@linux.vnet.ibm.com> Keith Busch writes: > On Fri, Jun 24, 2016@01:17:26AM -0700, Christoph Hellwig wrote: >> Yes, I'm not too excited about handling the AERs in kernel space, >> having a userspace daemon to listen for them would generally be more >> useful. > > How do daemons typically want to be notified of an event like this? I'm > thinking sysfs_notify or maybe a KOBJ_CHANGE uevent. I don't know, this > is unfamiliar territory to me. For IPR (IBM Hardware RAID SCSI controller) we use KOBJ_CHANGE events for a very similar situation, in which the driver signals a userspace daemon, which goes over a sysfs attribute to collect adapter data dumps. I'm also working on another async notifications mechanism for IPR, similar to what is supported by nvme, in which I plan to use the same uevent interface for userspace notifications. -- Gabriel Krisman Bertazi