From mboxrd@z Thu Jan 1 00:00:00 1970 From: igor.j.konopko@intel.com (Igor Konopko) Date: Fri, 23 Nov 2018 17:08:59 +0100 Subject: Should I submit bugs against RC kernels? In-Reply-To: <20181121191913.GJ26707@localhost.localdomain> References: <20181121191913.GJ26707@localhost.localdomain> Message-ID: <7739c84c-1bc4-b4e8-eba3-48c5c5118e5d@intel.com> On 21.11.2018 20:19, Keith Busch wrote: > On Wed, Nov 21, 2018@06:21:49PM +0000, Alex_Gagniuc@Dellteam.com wrote: >> Hi, >> >> I'm not sure if submitting bugs against RC is a good idea. I am seeing >> more issues with SURPRISE!!! removal of drives under v4.20-rc3. >> >> Alex > > Hi Alex, > > Yes, you can (and should!) report sightings for issues on rc kernels to > the appropriate subsystems. > > The following should resolve the issue you're seeing. I won't be able to > test it till next Monday, though. > > --- > diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c > index 4f504e8f0669..b1ce747411de 100644 > --- a/drivers/nvme/host/pci.c > +++ b/drivers/nvme/host/pci.c > @@ -1495,6 +1495,7 @@ static void nvme_dev_remove_admin(struct nvme_dev *dev) > blk_mq_unquiesce_queue(dev->ctrl.admin_q); > blk_cleanup_queue(dev->ctrl.admin_q); > blk_mq_free_tag_set(&dev->admin_tagset); > + dev->ctrl.admin_q = NULL; > } > } > > -- Hi I also hit the same issue as Alex. I also tried the proposed fix and it helps for surprise removal scenario, but causes another issue - when during hot removal scenarios we have some ongoing calls to blk_execute_rq() - which uses admin queue, then we hit null pointer dereference on that path. I just submit other fix, which on my side helps for both hot removal scenario and also does not brake blk_execute_rq() path. Igor