Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <161527286194.446794.5215036039655765042.stgit@dwillia2-desk3.amr.corp.intel.com>
 <20210309073110.GA3140@lst.de>
In-Reply-To: <20210309073110.GA3140@lst.de>
From:   Dan Williams <dan.j.williams@intel.com>
Date:   Tue, 9 Mar 2021 08:36:53 -0800
Message-ID: <CAPcyv4iQShzBxyZcn+H=2DGnnpifaV_e=5yBaGA1sF+ESS0jAQ@mail.gmail.com>
Subject: Re: [PATCH] libnvdimm: Let revalidate_disk() revalidate region read-only
To:     Christoph Hellwig <hch@lst.de>
Cc:     linux-nvdimm <linux-nvdimm@lists.01.org>,
        Ming Lei <ming.lei@redhat.com>,
        "Martin K. Petersen" <martin.petersen@oracle.com>,
        Hannes Reinecke <hare@suse.de>, Jens Axboe <axboe@kernel.dk>,
        kernel test robot <lkp@intel.com>,
        Vishal Verma <vishal.l.verma@intel.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        linux-block@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Mar 8, 2021 at 11:31 PM Christoph Hellwig <hch@lst.de> wrote:
>
> On Mon, Mar 08, 2021 at 10:54:22PM -0800, Dan Williams wrote:
> > Previous kernels allowed the BLKROSET to override the disk's read-only
> > status. With that situation fixed the pmem driver needs to rely on
> > revalidate_disk() to clear the disk read-only status after the host
> > region has been marked read-write.
> >
> > Recall that when libnvdimm determines that the persistent memory has
> > lost persistence (for example lack of energy to flush from DRAM to FLASH
> > on an NVDIMM-N device) it marks the region read-only, but that state can
> > be overridden by the user via:
> >
> >    echo 0 > /sys/bus/nd/devices/regionX/read_only
> >
> > ...to date there is no notification that the region has restored
> > persistence, so the user override is the only recovery.
>
> I've just resent my series to kill of ->revalidate_disk for good, so this
> obvious makes me a little unhappy.  Given that ->revalidate_disk
> only ends up beeing called from the same path that ->open is called,
> why can't you just hook this up from the open method?
>
> Also any reason the sysfs attribute can't just directly propagate the
> information to the disk?

I should have assumed that revalidate_disk() was on the chopping
block. Let me take a look at just propagating from the region update
down to all affected disks. There's already a notification path for
regions to communicate badblocks, should be straightforward to reuse
for read-only updates.