Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 5 Apr 2023 18:02:47 +0000
From:   Eric Biggers <ebiggers@kernel.org>
To:     Christoph Hellwig <hch@infradead.org>
Cc:     "Darrick J. Wong" <djwong@kernel.org>,
        Andrey Albershteyn <aalbersh@redhat.com>, dchinner@redhat.com,
        linux-xfs@vger.kernel.org, fsverity@lists.linux.dev,
        rpeterso@redhat.com, agruenba@redhat.com, xiang@kernel.org,
        chao@kernel.org, damien.lemoal@opensource.wdc.com, jth@kernel.org,
        linux-erofs@lists.ozlabs.org, linux-btrfs@vger.kernel.org,
        linux-ext4@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net,
        cluster-devel@redhat.com
Subject: Re: [PATCH v2 19/23] xfs: disable direct read path for fs-verity
 sealed files
Message-ID: <ZC23x22bxItnsANI@gmail.com>
References: <20230404145319.2057051-1-aalbersh@redhat.com>
 <20230404145319.2057051-20-aalbersh@redhat.com>
 <20230404161047.GA109974@frogsfrogsfrogs>
 <20230405150142.3jmxzo5i27bbc4c4@aalbersh.remote.csb>
 <20230405150927.GD303486@frogsfrogsfrogs>
 <ZC2YsgYRsvBejGYY@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZC2YsgYRsvBejGYY@infradead.org>
Precedence: bulk

On Wed, Apr 05, 2023 at 08:50:10AM -0700, Christoph Hellwig wrote:
> On Wed, Apr 05, 2023 at 08:09:27AM -0700, Darrick J. Wong wrote:
> > Thinking about this a little more -- I suppose we shouldn't just go
> > breaking directio reads from a verity file if we can help it.  Is there
> > a way to ask fsverity to perform its validation against some arbitrary
> > memory buffer that happens to be fs-block aligned?

You could certainly add such a function that wraps around verify_data_block().
The minimal function prototype needed (without supporting readahead or reusing
the ahash_request) would be something like the following, I think:

    bool fsverity_verify_blocks_dio(struct inode *inode, u64 pos,
                                    struct folio *folio,
                                    size_t len, size_t offset);

And I really hope that you don't want to do DIO to the *Merkle tree*, as that
would make the problem significantly harder.  I think DIO for the data, but
handling the Merkle tree in the usual way, would be okay?

> 
> That would be my preference as well.  But maybe Eric know a good reason
> why this hasn't been done yet.
> 

I believe it would be possible, especially if DIO to the Merkle tree is not in
scope.  There just hasn't been a reason to the work yet.  And ext4 and f2fs
already fall back to buffer I/O for other filesystem features, so there was
precedent for not bothering with DIO, at least in the initial version.

- Eric