From: dai.ngo@oracle.com
To: Luis Henriques <lhenriques@suse.de>
Cc: Amir Goldstein <amir73il@gmail.com>,
Jeff Layton <jlayton@kernel.org>,
Steve French <sfrench@samba.org>,
Miklos Szeredi <miklos@szeredi.hu>,
Trond Myklebust <trond.myklebust@hammerspace.com>,
Anna Schumaker <anna.schumaker@netapp.com>,
Alexander Viro <viro@zeniv.linux.org.uk>,
"Darrick J. Wong" <darrick.wong@oracle.com>,
Dave Chinner <dchinner@redhat.com>,
Greg KH <gregkh@linuxfoundation.org>,
Nicolas Boichat <drinkcat@chromium.org>,
Ian Lance Taylor <iant@google.com>,
Luis Lozano <llozano@chromium.org>,
Andreas Dilger <adilger@dilger.ca>,
Olga Kornievskaia <aglo@umich.edu>,
Christoph Hellwig <hch@infradead.org>,
ceph-devel@vger.kernel.org, linux-kernel@vger.kernel.org,
linux-cifs@vger.kernel.org, samba-technical@lists.samba.org,
linux-fsdevel@vger.kernel.org, linux-nfs@vger.kernel.org
Subject: Re: [PATCH v8] vfs: fix copy_file_range regression in cross-fs copies
Date: Tue, 23 Feb 2021 07:29:09 -0800 [thread overview]
Message-ID: <7cc69c24-80dd-0053-24b9-3a28b0153f7e@oracle.com> (raw)
In-Reply-To: <YDTZwH7xv41Wimax@suse.de>
On 2/23/21 2:32 AM, Luis Henriques wrote:
> On Mon, Feb 22, 2021 at 08:25:27AM -0800, dai.ngo@oracle.com wrote:
>> On 2/22/21 2:24 AM, Luis Henriques wrote:
>>> A regression has been reported by Nicolas Boichat, found while using the
>>> copy_file_range syscall to copy a tracefs file. Before commit
>>> 5dae222a5ff0 ("vfs: allow copy_file_range to copy across devices") the
>>> kernel would return -EXDEV to userspace when trying to copy a file across
>>> different filesystems. After this commit, the syscall doesn't fail anymore
>>> and instead returns zero (zero bytes copied), as this file's content is
>>> generated on-the-fly and thus reports a size of zero.
>>>
>>> This patch restores some cross-filesystem copy restrictions that existed
>>> prior to commit 5dae222a5ff0 ("vfs: allow copy_file_range to copy across
>>> devices"). Filesystems are still allowed to fall-back to the VFS
>>> generic_copy_file_range() implementation, but that has now to be done
>>> explicitly.
>>>
>>> nfsd is also modified to fall-back into generic_copy_file_range() in case
>>> vfs_copy_file_range() fails with -EOPNOTSUPP or -EXDEV.
>>>
>>> Fixes: 5dae222a5ff0 ("vfs: allow copy_file_range to copy across devices")
>>> Link: https://urldefense.com/v3/__https://lore.kernel.org/linux-fsdevel/20210212044405.4120619-1-drinkcat@chromium.org/__;!!GqivPVa7Brio!P1UWThiSkxbjfjFQWNYJmCxGEkiLFyvHjH6cS-G1ZTt1z-TeqwGQgQmi49dC6w$
>>> Link: https://urldefense.com/v3/__https://lore.kernel.org/linux-fsdevel/CANMq1KDZuxir2LM5jOTm0xx*BnvW=ZmpsG47CyHFJwnw7zSX6Q@mail.gmail.com/__;Kw!!GqivPVa7Brio!P1UWThiSkxbjfjFQWNYJmCxGEkiLFyvHjH6cS-G1ZTt1z-TeqwGQgQmgCmMHzA$
>>> Link: https://urldefense.com/v3/__https://lore.kernel.org/linux-fsdevel/20210126135012.1.If45b7cdc3ff707bc1efa17f5366057d60603c45f@changeid/__;!!GqivPVa7Brio!P1UWThiSkxbjfjFQWNYJmCxGEkiLFyvHjH6cS-G1ZTt1z-TeqwGQgQmzqItkrQ$
>>> Reported-by: Nicolas Boichat <drinkcat@chromium.org>
>>> Signed-off-by: Luis Henriques <lhenriques@suse.de>
>>> ---
>>> Changes since v7
>>> - set 'ret' to '-EOPNOTSUPP' before the clone 'if' statement so that the
>>> error returned is always related to the 'copy' operation
>>> Changes since v6
>>> - restored i_sb checks for the clone operation
>>> Changes since v5
>>> - check if ->copy_file_range is NULL before calling it
>>> Changes since v4
>>> - nfsd falls-back to generic_copy_file_range() only *if* it gets -EOPNOTSUPP
>>> or -EXDEV.
>>> Changes since v3
>>> - dropped the COPY_FILE_SPLICE flag
>>> - kept the f_op's checks early in generic_copy_file_checks, implementing
>>> Amir's suggestions
>>> - modified nfsd to use generic_copy_file_range()
>>> Changes since v2
>>> - do all the required checks earlier, in generic_copy_file_checks(),
>>> adding new checks for ->remap_file_range
>>> - new COPY_FILE_SPLICE flag
>>> - don't remove filesystem's fallback to generic_copy_file_range()
>>> - updated commit changelog (and subject)
>>> Changes since v1 (after Amir review)
>>> - restored do_copy_file_range() helper
>>> - return -EOPNOTSUPP if fs doesn't implement CFR
>>> - updated commit description
>>>
>>> fs/nfsd/vfs.c | 8 +++++++-
>>> fs/read_write.c | 49 ++++++++++++++++++++++++-------------------------
>>> 2 files changed, 31 insertions(+), 26 deletions(-)
>>>
>>> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
>>> index 04937e51de56..23dab0fa9087 100644
>>> --- a/fs/nfsd/vfs.c
>>> +++ b/fs/nfsd/vfs.c
>>> @@ -568,6 +568,7 @@ __be32 nfsd4_clone_file_range(struct nfsd_file *nf_src, u64 src_pos,
>>> ssize_t nfsd_copy_file_range(struct file *src, u64 src_pos, struct file *dst,
>>> u64 dst_pos, u64 count)
>>> {
>>> + ssize_t ret;
>>> /*
>>> * Limit copy to 4MB to prevent indefinitely blocking an nfsd
>>> @@ -578,7 +579,12 @@ ssize_t nfsd_copy_file_range(struct file *src, u64 src_pos, struct file *dst,
>>> * limit like this and pipeline multiple COPY requests.
>>> */
>>> count = min_t(u64, count, 1 << 22);
>>> - return vfs_copy_file_range(src, src_pos, dst, dst_pos, count, 0);
>>> + ret = vfs_copy_file_range(src, src_pos, dst, dst_pos, count, 0);
>>> +
>>> + if (ret == -EOPNOTSUPP || ret == -EXDEV)
>>> + ret = generic_copy_file_range(src, src_pos, dst, dst_pos,
>>> + count, 0);
>>> + return ret;
>>> }
>>> __be32 nfsd4_vfs_fallocate(struct svc_rqst *rqstp, struct svc_fh *fhp,
>>> diff --git a/fs/read_write.c b/fs/read_write.c
>>> index 75f764b43418..5a26297fd410 100644
>>> --- a/fs/read_write.c
>>> +++ b/fs/read_write.c
>>> @@ -1388,28 +1388,6 @@ ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
>>> }
>>> EXPORT_SYMBOL(generic_copy_file_range);
>>> -static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>>> - struct file *file_out, loff_t pos_out,
>>> - size_t len, unsigned int flags)
>>> -{
>>> - /*
>>> - * Although we now allow filesystems to handle cross sb copy, passing
>>> - * a file of the wrong filesystem type to filesystem driver can result
>>> - * in an attempt to dereference the wrong type of ->private_data, so
>>> - * avoid doing that until we really have a good reason. NFS defines
>>> - * several different file_system_type structures, but they all end up
>>> - * using the same ->copy_file_range() function pointer.
>>> - */
>>> - if (file_out->f_op->copy_file_range &&
>>> - file_out->f_op->copy_file_range == file_in->f_op->copy_file_range)
>>> - return file_out->f_op->copy_file_range(file_in, pos_in,
>>> - file_out, pos_out,
>>> - len, flags);
>>> -
>>> - return generic_copy_file_range(file_in, pos_in, file_out, pos_out, len,
>>> - flags);
>>> -}
>>> -
>>> /*
>>> * Performs necessary checks before doing a file copy
>>> *
>>> @@ -1427,6 +1405,25 @@ static int generic_copy_file_checks(struct file *file_in, loff_t pos_in,
>>> loff_t size_in;
>>> int ret;
>>> + /*
>>> + * Although we now allow filesystems to handle cross sb copy, passing
>>> + * a file of the wrong filesystem type to filesystem driver can result
>>> + * in an attempt to dereference the wrong type of ->private_data, so
>>> + * avoid doing that until we really have a good reason. NFS defines
>>> + * several different file_system_type structures, but they all end up
>>> + * using the same ->copy_file_range() function pointer.
>>> + */
>>> + if (file_out->f_op->copy_file_range) {
>>> + if (file_in->f_op->copy_file_range !=
>>> + file_out->f_op->copy_file_range)
>>> + return -EXDEV;
>>> + } else if (file_in->f_op->remap_file_range) {
>>> + if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb)
>>> + return -EXDEV;
>> I think this check is redundant, it's done in vfs_copy_file_range.
>> If this check is removed then the else clause below should be removed
>> also. Once this check and the else clause are removed then might as
>> well move the the check of copy_file_range from here to vfs_copy_file_range.
>>
> I don't think it's really redundant, although I agree is messy due to the
> fact we try to clone first instead of copying them.
>
> So, in the clone path, this is the only place where we return -EXDEV if:
>
> 1) we don't have ->copy_file_range *and*
> 2) we have ->remap_file_range but the i_sb are different.
>
> The check in vfs_copy_file_range() is only executed if:
>
> 1) we have *valid* ->copy_file_range ops and/or
> 2) we have *valid* ->remap_file_range
>
> So... if we remove the check in generic_copy_file_checks() as you suggest
> and:
> - we don't have ->copy_file_range,
> - we have ->remap_file_range but
> - the i_sb are different
>
> we'll return the -EOPNOTSUPP (the one set in line "ret = -EOPNOTSUPP;" in
> function vfs_copy_file_range() ) instead of -EXDEV.
Yes, this is the different.The NFS code handles both -EOPNOTSUPP and
-EXDEVV by doing generic_copy_file_range. Do any other consumers of
vfs_copy_file_range rely on -EXDEV and not -EOPNOTSUPP and which is
the correct error code for this case? It seems to me that -EOPNOTSUPP
is more appropriate than EXDEV when (sb1 != sb2).
>
> But I may have got it all wrong. I've looked so many times at this code
> that I'm probably useless at finding problems in it :-)
You're not alone, we all try to do the right thing :-)
-Dai
>
> Cheers,
> --
> Luís
>
>> -Dai
>>
>>> + } else {
>>> + return -EOPNOTSUPP;
>>> + }
>>> +
>>> ret = generic_file_rw_checks(file_in, file_out);
>>> if (ret)
>>> return ret;
>>> @@ -1495,6 +1492,7 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
>>> file_start_write(file_out);
>>> + ret = -EOPNOTSUPP;
>>> /*
>>> * Try cloning first, this is supported by more file systems, and
>>> * more efficient if both clone and copy are supported (e.g. NFS).
>>> @@ -1513,9 +1511,10 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
>>> }
>>> }
>>> - ret = do_copy_file_range(file_in, pos_in, file_out, pos_out, len,
>>> - flags);
>>> - WARN_ON_ONCE(ret == -EOPNOTSUPP);
>>> + if (file_out->f_op->copy_file_range)
>>> + ret = file_out->f_op->copy_file_range(file_in, pos_in,
>>> + file_out, pos_out,
>>> + len, flags);
>>> done:
>>> if (ret > 0) {
>>> fsnotify_access(file_in);
next prev parent reply other threads:[~2021-02-23 15:31 UTC|newest]
Thread overview: 93+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <CAOQ4uxiFGjdvX2-zh5o46pn7RZhvbGHH0wpzLPuPOom91FwWeQ@mail.gmail.com>
2021-02-15 15:43 ` [PATCH v2] vfs: prevent copy_file_range to copy across devices Luis Henriques
2021-02-15 16:02 ` Trond Myklebust
2021-02-16 0:25 ` Steve French
2021-02-15 16:34 ` Amir Goldstein
2021-02-15 16:53 ` Trond Myklebust
2021-02-15 17:24 ` Amir Goldstein
2021-02-15 18:57 ` Trond Myklebust
2021-02-15 19:43 ` Amir Goldstein
2021-02-16 11:17 ` Luis Henriques
2021-02-16 11:28 ` gregkh
2021-02-16 12:01 ` Luis Henriques
2021-02-16 12:08 ` Greg KH
2021-02-16 13:51 ` Amir Goldstein
2021-02-16 16:42 ` Luis Henriques
2021-02-16 17:44 ` Amir Goldstein
2021-02-16 18:55 ` Luis Henriques
2021-02-16 19:20 ` Amir Goldstein
2021-02-16 19:27 ` Anna Schumaker
2021-02-16 19:31 ` Steve French
2021-02-16 19:40 ` Amir Goldstein
2021-02-16 21:15 ` Steve French
2021-02-17 8:08 ` Amir Goldstein
2021-02-17 17:26 ` [PATCH v3] vfs: fix copy_file_range regression in cross-fs copies Luis Henriques
2021-02-17 20:47 ` Amir Goldstein
2021-02-18 0:56 ` Nicolas Boichat
2021-02-18 5:32 ` Olga Kornievskaia
2021-02-18 6:47 ` Amir Goldstein
2021-02-18 16:28 ` Olga Kornievskaia
2021-02-18 7:43 ` Christoph Hellwig
2021-02-18 0:50 ` [PATCH v2] vfs: prevent copy_file_range to copy across devices Andreas Dilger
2021-02-18 7:34 ` gregkh
2021-02-16 18:54 ` Andreas Dilger
2021-02-17 4:45 ` Nicolas Boichat
2021-02-18 7:42 ` Christoph Hellwig
2021-02-18 9:10 ` Amir Goldstein
2021-02-18 10:29 ` Luis Henriques
2021-02-18 12:15 ` Luis Henriques
2021-02-18 12:49 ` Amir Goldstein
2021-02-18 14:36 ` [PATCH v4] vfs: fix copy_file_range regression in cross-fs copies Luis Henriques
2021-02-18 14:58 ` Amir Goldstein
2021-02-18 15:17 ` [PATCH v5] " Luis Henriques
2021-02-18 15:53 ` Amir Goldstein
2021-02-18 16:35 ` Luis Henriques
2021-02-18 17:18 ` [PATCH v6] " Luis Henriques
2021-02-19 21:18 ` Olga Kornievskaia
2021-02-19 21:52 ` Amir Goldstein
2021-02-21 19:58 ` [PATCH v7] " Luis Henriques
2021-02-22 3:00 ` Nicolas Boichat
2021-02-22 10:24 ` [PATCH v8] " Luis Henriques
2021-02-22 10:46 ` Amir Goldstein
2021-02-22 16:25 ` dai.ngo
2021-02-23 10:32 ` Luis Henriques
2021-02-23 15:28 ` Amir Goldstein
2021-02-23 15:29 ` dai.ngo [this message]
2021-02-23 16:02 ` dai.ngo
2021-02-23 16:47 ` Amir Goldstein
2021-02-23 16:57 ` dai.ngo
[not found] ` <e3eed18b-fc7e-e687-608b-7f662017329c@oracle.com>
2021-02-23 17:33 ` Amir Goldstein
2021-02-24 0:13 ` dai.ngo
2021-02-23 17:56 ` Luis Henriques
2021-02-23 17:13 ` Olga Kornievskaia
2021-02-24 1:00 ` Olga Kornievskaia
2021-02-24 10:23 ` Luis Henriques
2021-02-24 10:44 ` Nicolas Boichat
2021-04-09 5:23 ` Nicolas Boichat
2021-04-09 13:39 ` Luis Henriques
2021-04-09 13:50 ` Amir Goldstein
2021-04-23 4:40 ` Nicolas Boichat
2021-05-03 8:54 ` Luis Henriques
2021-02-24 14:23 ` [PATCH] copy_file_range.2: Kernel v5.12 updates Luis Henriques
2021-02-24 16:10 ` Amir Goldstein
2021-02-25 10:21 ` Luis Henriques
2021-02-26 10:13 ` Alejandro Colomar (man-pages)
2021-02-26 10:34 ` Amir Goldstein
2021-02-26 11:15 ` Alejandro Colomar (man-pages)
2021-02-26 13:59 ` Jeff Layton
2021-02-26 21:26 ` Alejandro Colomar (man-pages)
2021-02-26 22:18 ` Alejandro Colomar (man-pages)
2021-02-27 5:41 ` Amir Goldstein
2021-02-27 12:20 ` Alejandro Colomar (man-pages)
2021-02-27 13:49 ` [RFC v2] copy_file_range.2: Update cross-filesystem support for 5.12 Alejandro Colomar
2021-02-27 16:00 ` Amir Goldstein
2021-02-27 23:08 ` [PATCH] copy_file_range.2: Kernel v5.12 updates Steve French
2021-02-28 7:35 ` Amir Goldstein
2021-02-28 22:25 ` Steve French
2021-03-01 6:18 ` Amir Goldstein
2021-03-01 14:41 ` [RFC v3] copy_file_range.2: Update cross-filesystem support for 5.12 Alejandro Colomar
2021-03-01 14:58 ` Amir Goldstein
2021-03-04 9:38 ` [RFC v4] " Alejandro Colomar
2021-03-04 17:13 ` Darrick J. Wong
2021-03-04 18:24 ` Alejandro Colomar (man-pages)
2021-03-04 23:50 ` Darrick J. Wong
2021-02-18 20:41 ` [PATCH v2] vfs: prevent copy_file_range to copy across devices Steve French
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=7cc69c24-80dd-0053-24b9-3a28b0153f7e@oracle.com \
--to=dai.ngo@oracle.com \
--cc=adilger@dilger.ca \
--cc=aglo@umich.edu \
--cc=amir73il@gmail.com \
--cc=anna.schumaker@netapp.com \
--cc=ceph-devel@vger.kernel.org \
--cc=darrick.wong@oracle.com \
--cc=dchinner@redhat.com \
--cc=drinkcat@chromium.org \
--cc=gregkh@linuxfoundation.org \
--cc=hch@infradead.org \
--cc=iant@google.com \
--cc=jlayton@kernel.org \
--cc=lhenriques@suse.de \
--cc=linux-cifs@vger.kernel.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nfs@vger.kernel.org \
--cc=llozano@chromium.org \
--cc=miklos@szeredi.hu \
--cc=samba-technical@lists.samba.org \
--cc=sfrench@samba.org \
--cc=trond.myklebust@hammerspace.com \
--cc=viro@zeniv.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).