Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <c1fd63b9-42ea-fa83-ecb1-9af715e37ffa@redhat.com>
Date:   Fri, 19 May 2023 16:40:26 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.10.0
Subject: Re: [PATCH v20 13/32] ceph: Provide a splice-read stub
To:     David Howells <dhowells@redhat.com>, Jens Axboe <axboe@kernel.dk>,
        Al Viro <viro@zeniv.linux.org.uk>,
        Christoph Hellwig <hch@infradead.org>
Cc:     Matthew Wilcox <willy@infradead.org>, Jan Kara <jack@suse.cz>,
        Jeff Layton <jlayton@kernel.org>,
        David Hildenbrand <david@redhat.com>,
        Jason Gunthorpe <jgg@nvidia.com>,
        Logan Gunthorpe <logang@deltatee.com>,
        Hillf Danton <hdanton@sina.com>,
        Christian Brauner <brauner@kernel.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        linux-fsdevel@vger.kernel.org, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        Christoph Hellwig <hch@lst.de>,
        Ilya Dryomov <idryomov@gmail.com>, ceph-devel@vger.kernel.org
References: <20230519074047.1739879-1-dhowells@redhat.com>
 <20230519074047.1739879-14-dhowells@redhat.com>
Content-Language: en-US
From:   Xiubo Li <xiubli@redhat.com>
In-Reply-To: <20230519074047.1739879-14-dhowells@redhat.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 5/19/23 15:40, David Howells wrote:
> Provide a splice_read stub for Ceph.  This does the inode shutdown check
> before proceeding and jumps to direct_splice_read() if O_DIRECT is set, the
> file has inline data or is a synchronous file.
>
> We try and get FILE_RD and either FILE_CACHE and/or FILE_LAZYIO caps and
> hold them across filemap_splice_read().  If we fail to get FILE_CACHE or
> FILE_LAZYIO capabilities, we use direct_splice_read() instead.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Christoph Hellwig <hch@lst.de>
> cc: Al Viro <viro@zeniv.linux.org.uk>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Xiubo Li <xiubli@redhat.com>
> cc: Ilya Dryomov <idryomov@gmail.com>
> cc: Jeff Layton <jlayton@kernel.org>
> cc: ceph-devel@vger.kernel.org
> cc: linux-fsdevel@vger.kernel.org
> cc: linux-block@vger.kernel.org
> cc: linux-mm@kvack.org
> ---
>   fs/ceph/file.c | 66 +++++++++++++++++++++++++++++++++++++++++++++++++-
>   1 file changed, 65 insertions(+), 1 deletion(-)
>
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index f4d8bf7dec88..382dd5901748 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -1745,6 +1745,70 @@ static ssize_t ceph_read_iter(struct kiocb *iocb, struct iov_iter *to)
>   	return ret;
>   }
>   
> +/*
> + * Wrap filemap_splice_read with checks for cap bits on the inode.
> + * Atomically grab references, so that those bits are not released
> + * back to the MDS mid-read.
> + */
> +static ssize_t ceph_splice_read(struct file *in, loff_t *ppos,
> +				struct pipe_inode_info *pipe,
> +				size_t len, unsigned int flags)
> +{
> +	struct ceph_file_info *fi = in->private_data;
> +	struct inode *inode = file_inode(in);
> +	struct ceph_inode_info *ci = ceph_inode(inode);
> +	ssize_t ret;
> +	int want = 0, got = 0;
> +	CEPH_DEFINE_RW_CONTEXT(rw_ctx, 0);
> +
> +	dout("splice_read %p %llx.%llx %llu~%zu trying to get caps on %p\n",
> +	     inode, ceph_vinop(inode), *ppos, len, inode);
> +
> +	if (ceph_inode_is_shutdown(inode))
> +		return -ESTALE;
> +
> +	if ((in->f_flags & O_DIRECT) ||
> +	    ceph_has_inline_data(ci) ||
> +	    (fi->flags & CEPH_F_SYNC))
> +		return direct_splice_read(in, ppos, pipe, len, flags);
> +
> +	ceph_start_io_read(inode);
> +
> +	want = CEPH_CAP_FILE_CACHE;
> +	if (fi->fmode & CEPH_FILE_MODE_LAZY)
> +		want |= CEPH_CAP_FILE_LAZYIO;
> +
> +	ret = ceph_get_caps(in, CEPH_CAP_FILE_RD, want, -1, &got);
> +	if (ret < 0) {
> +		ceph_end_io_read(inode);
> +		return ret;
> +	}
> +
> +	if ((got & (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO)) == 0) {
> +		dout("splice_read/sync %p %llx.%llx %llu~%zu got cap refs on %s\n",
> +		     inode, ceph_vinop(inode), *ppos, len,
> +		     ceph_cap_string(got));
> +
> +		ceph_end_io_read(inode);
> +		return direct_splice_read(in, ppos, pipe, len, flags);

Shouldn't we release cap ref before returning here ?

Thanks

- Xiubo


> +	}
> +
> +	dout("splice_read %p %llx.%llx %llu~%zu got cap refs on %s\n",
> +	     inode, ceph_vinop(inode), *ppos, len, ceph_cap_string(got));
> +
> +	rw_ctx.caps = got;
> +	ceph_add_rw_context(fi, &rw_ctx);
> +	ret = filemap_splice_read(in, ppos, pipe, len, flags);
> +	ceph_del_rw_context(fi, &rw_ctx);
> +
> +	dout("splice_read %p %llx.%llx dropping cap refs on %s = %zd\n",
> +	     inode, ceph_vinop(inode), ceph_cap_string(got), ret);
> +	ceph_put_cap_refs(ci, got);
> +
> +	ceph_end_io_read(inode);
> +	return ret;
> +}
> +
>   /*
>    * Take cap references to avoid releasing caps to MDS mid-write.
>    *
> @@ -2593,7 +2657,7 @@ const struct file_operations ceph_file_fops = {
>   	.lock = ceph_lock,
>   	.setlease = simple_nosetlease,
>   	.flock = ceph_flock,
> -	.splice_read = generic_file_splice_read,
> +	.splice_read = ceph_splice_read,
>   	.splice_write = iter_file_splice_write,
>   	.unlocked_ioctl = ceph_ioctl,
>   	.compat_ioctl = compat_ptr_ioctl,
>