Re: [PATCH v5 2/7] dax: introduce dax device flag DAXDEV_RECOVERY

From: Dan Williams <dan.j.williams@intel.com>
To: Christoph Hellwig <hch@infradead.org>
Cc: Jane Chu <jane.chu@oracle.com>,
	"david@fromorbit.com" <david@fromorbit.com>,
	 "djwong@kernel.org" <djwong@kernel.org>,
	"vishal.l.verma@intel.com" <vishal.l.verma@intel.com>,
	 "dave.jiang@intel.com" <dave.jiang@intel.com>,
	"agk@redhat.com" <agk@redhat.com>,
	 "snitzer@redhat.com" <snitzer@redhat.com>,
	"dm-devel@redhat.com" <dm-devel@redhat.com>,
	 "ira.weiny@intel.com" <ira.weiny@intel.com>,
	"willy@infradead.org" <willy@infradead.org>,
	 "vgoyal@redhat.com" <vgoyal@redhat.com>,
	 "linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
	 "nvdimm@lists.linux.dev" <nvdimm@lists.linux.dev>,
	 "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	 "linux-xfs@vger.kernel.org" <linux-xfs@vger.kernel.org>
Subject: Re: [PATCH v5 2/7] dax: introduce dax device flag DAXDEV_RECOVERY
Date: Thu, 3 Feb 2022 21:32:27 -0800	[thread overview]
Message-ID: <CAPcyv4hCf0WpRyNx4vo0_+-w1ABX0cJTyLozPYEqiqR0i_H1_Q@mail.gmail.com> (raw)
In-Reply-To: <CAPcyv4i99BhF+JndtanBuOWRc3eh1C=-CyswhvLDeDSeTHSUZw@mail.gmail.com>

On Thu, Feb 3, 2022 at 9:17 PM Dan Williams <dan.j.williams@intel.com> wrote:
>
> On Thu, Feb 3, 2022 at 5:43 AM Christoph Hellwig <hch@infradead.org> wrote:
> >
> > On Wed, Feb 02, 2022 at 09:27:42PM +0000, Jane Chu wrote:
> > > Yeah, I see.  Would you suggest a way to pass the indication from
> > > dax_iomap_iter to dax_direct_access that the caller intends the
> > > callee to ignore poison in the range because the caller intends
> > > to do recovery_write? We tried adding a flag to dax_direct_access, and
> > > that wasn't liked if I recall.
> >
> > To me a flag seems cleaner than this magic, but let's wait for Dan to
> > chime in.
>
> So back in November I suggested modifying the kaddr, mainly to avoid
> touching all the dax_direct_access() call sites [1]. However, now
> seeing the code and Chrisoph's comment I think this either wants type
> safety (e.g. 'dax_addr_t *'), or just add a new flag. Given both of
> those options involve touching all dax_direct_access() call sites and
> a @flags operation is more extensible if any other scenarios arrive
> lets go ahead and plumb a flag and skip the magic.

Just to be clear we are talking about a flow like:

        flags = 0;
        rc = dax_direct_access(..., &kaddr, flags, ...);
        if (unlikely(rc)) {
                flags |= DAX_RECOVERY;
                dax_direct_access(..., &kaddr, flags, ...);
                return dax_recovery_{read,write}(..., kaddr, ...);
        }
        return copy_{mc_to_iter,from_iter_flushcache}(...);