Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Message-ID: <96a44a4224b67b9ec36198bc0586064df5c4ad2a.camel@kernel.org>
Subject: Re: [PATCH v3] ceph: invalidate pages when doing direct/sync writes
From:   Jeff Layton <jlayton@kernel.org>
To:     =?ISO-8859-1?Q?Lu=EDs?= Henriques <lhenriques@suse.de>,
        Xiubo Li <xiubli@redhat.com>, Ilya Dryomov <idryomov@gmail.com>
Cc:     ceph-devel@vger.kernel.org, linux-kernel@vger.kernel.org
Date:   Thu, 07 Apr 2022 10:42:57 -0400
In-Reply-To: <20220407143834.7516-1-lhenriques@suse.de>
References: <20220407143834.7516-1-lhenriques@suse.de>
Content-Type: text/plain; charset="ISO-8859-15"
User-Agent: Evolution 3.42.4 (3.42.4-1.fc35) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Thu, 2022-04-07 at 15:38 +0100, Lu?s Henriques wrote:
> When doing a direct/sync write, we need to invalidate the page cache in
> the range being written to.  If we don't do this, the cache will include
> invalid data as we just did a write that avoided the page cache.
> 
> Signed-off-by: Lu?s Henriques <lhenriques@suse.de>
> ---
>  fs/ceph/file.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> Ok, here's a new attempt.  After discussion in this thread and on IRC, I
> think this is the right fix.  generic/647 now passes with and without
> encryption.  Thanks!
> 
> Changes since v2:
> - Invalidation needs to be done after a write
> 
> Changes since v1:
> - Replaced truncate_inode_pages_range() by invalidate_inode_pages2_range
> - Call fscache_invalidate with FSCACHE_INVAL_DIO_WRITE if we're doing DIO
> 
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index 5072570c2203..63e67eb60310 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -1938,6 +1938,15 @@ ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos,
>  			break;
>  		}
>  		ceph_clear_error_write(ci);
> +		ret = invalidate_inode_pages2_range(
> +				inode->i_mapping,
> +				pos >> PAGE_SHIFT,
> +				(pos + len - 1) >> PAGE_SHIFT);
> +		if (ret < 0) {
> +			dout("invalidate_inode_pages2_range returned %d\n",
> +			     ret);
> +			ret = 0;
> +		}
>  		pos += len;
>  		written += len;
>  		dout("sync_write written %d\n", written);

Looks good. I suspect we can also remove the
invalidate_indode_pages2_range call earlier in this function too. I may
roll that into this patch.

I'll give this an xfstests run with fscrypt enabled and see how it does.

Thanks!
-- 
Jeff Layton <jlayton@kernel.org>