From mboxrd@z Thu Jan 1 00:00:00 1970 From: hch@lst.de (Christoph Hellwig) Date: Mon, 5 Jun 2017 10:18:17 +0200 Subject: [PATCH V2] nvme: fix nvme_remove going to uninterruptible sleep for ever In-Reply-To: References: <20170530071610.GA2679@hercules.tuxera.com> <4da7c939-1f54-80e5-48fc-06e58e14f018@grimberg.me> <20170530142346.GA39428@dhcp-216.srv.tuxera.com> <20170601114338.GA24855@lst.de> Message-ID: <20170605081817.GA22122@lst.de> On Sun, Jun 04, 2017@06:28:15PM +0300, Sagi Grimberg wrote: > >> I think we need the NVME_CTRL_SCHED_RESET state. In fact I'm pretty >> sure some time in the past I already had it in a local tree as a >> generalization of rdma and loop already use NVME_CTRL_RESETTING >> (they set it before queueing the reset work). > > I don't remember having it, but I might be wrong... > > Can you explain again why you think we need it? Sorry for being > difficult, but I'm not exactly sure why it makes things better > or simpler. Motly that we can treat a controller as under reset before scheduling the reset work, both to prevent multiple schedules, and to make checks like the one in nvme_should_reset robus. But I think something along the lines of the earlier patch from Rakesh that just sets the RESETTING state earlier + the cancel_work_sync suggested by you should also work for that purpose. So maybe that's the way to go after all.