Re:Re: [PATCH] xfsprogs: io/copy_range: cover corner case (fd_in == fd_out)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




At 2019-09-04 10:03:57, "Zorro Lang" <zlang@xxxxxxxxxx> wrote:
>On Wed, Sep 04, 2019 at 12:31:16AM +0800, 尹剑虹 wrote:
>> We need cover the scenario that fd_in == fd_out
>> not just same path.
>
>Please reply to mail list, not 'me' only, to get more review:)
>
>The patch which you're trying to cover is commit 9ab70ca653 as below[1].
>From the code, I really doubt if you need same `struct file`, looks like
>you need same `struct inode`.
>
>Have you tried to test on same inode but not same 'fd'? I'm not a CIFS
>expert, can CIFS have same file with different inode?

Thanks Zorro, you are right. same fd is not necessary;

drop this patch.

>
>Thanks,
>Zorro
>
>[1]
>commit 9ab70ca653307771589e1414102c552d8dbdbbef
>Author: Kovtunenko Oleksandr <alexander198961@xxxxxxxxx>
>Date:   Tue May 14 05:52:34 2019 +0000
>
>    Fixed https://bugzilla.kernel.org/show_bug.cgi?id=202935 allow write on the same file
>    
>    Copychunk allows source and target to be on the same file.
>    For details on restrictions see MS-SMB2 3.3.5.15.6
>    
>    Signed-off-by: Kovtunenko Oleksandr <alexander198961@xxxxxxxxx>
>    Signed-off-by: Steve French <stfrench@xxxxxxxxxxxxx>
>
>diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
>index b1a5fcfa3ce1..d0cb042732cb 100644
>--- a/fs/cifs/cifsfs.c
>+++ b/fs/cifs/cifsfs.c
>@@ -1070,11 +1070,6 @@ ssize_t cifs_file_copychunk_range(unsigned int xid,
> 
>        cifs_dbg(FYI, "copychunk range\n");
> 
>-       if (src_inode == target_inode) {
>-               rc = -EINVAL;
>-               goto out;
>-       }
>-
>        if (!src_file->private_data || !dst_file->private_data) {
>                rc = -EBADF;
>                cifs_dbg(VFS, "missing cifsFileInfo on copy range src file\n");
>
>> 
>> #Did you read the summary and commit log?
>> 
>> 
>> | |
>> 尹剑虹
>> |
>> |
>> 邮箱:yin-jianhong@xxxxxxx
>> |
>> 
>> 签名由 网易邮箱大师 定制
>> 
>> On 09/03/2019 21:19, Zorro Lang wrote:
>> On Tue, Sep 03, 2019 at 07:59:43PM +0800, Zorro Lang wrote:
>> > On Tue, Sep 03, 2019 at 06:56:32PM +0800, Jianhong.Yin wrote:
>> > > Related bug:
>> > >   copy_file_range return "Invalid argument" when copy in the same file
>> > >   https://bugzilla.kernel.org/show_bug.cgi?id=202935
>> > >
>> > > if argument of option -f is "-", use current file->fd as fd_in
>> > >
>> > > Usage:
>> > >   xfs_io -c 'copy_range -f -' some_file
>> > >
>> > > Signed-off-by: Jianhong Yin <yin-jianhong@xxxxxxx>
>> > > ---
>> >
>> > Hi,
>> >
>> > Actually, I'm thinking about if you need same 'fd' or same file path?
>> > If you just need same file path, I think
>> >
>> >   # xfs_io -c "copy_range testfile" testfile
>> >
>> > already can help that. The only one problem stop you doing that is
>> > "copy_dst_truncate()".
>> >
>> > If all above I suppose is right, we can turn to talk about if that
>> > copy_dst_truncate() is necessary, or how can we skip it.
>> 
>> I just checked, the copy_dst_truncate() is only called when:
>> 
>>  if (src == 0 && dst == 0 && len == 0) {
>> 
>> So if you can give your reproducer a "length"(or offset), likes:
>> 
>>  # xfs_io -c "copy_range -l 64k testfile" testfile
>> 
>> You can avoid the copy_dst_truncate() too.
>> 
>> Is that helpful?
>> 
>> Thanks,
>> Zorro
>> 
>> >
>> > Thanks,
>> > Zorro
>> >
>> > >  io/copy_file_range.c | 27 ++++++++++++++++++---------
>> > >  1 file changed, 18 insertions(+), 9 deletions(-)
>> > >
>> > > diff --git a/io/copy_file_range.c b/io/copy_file_range.c
>> > > index b7b9fd88..2dde8a31 100644
>> > > --- a/io/copy_file_range.c
>> > > +++ b/io/copy_file_range.c
>> > > @@ -28,6 +28,7 @@ copy_range_help(void)
>> > >                            at position 0\n\
>> > >   'copy_range -f 2' - copies all bytes from open file 2 into the current open file\n\
>> > >                            at position 0\n\
>> > > + 'copy_range -f -' - copies all bytes from current open file append the current open file\n\
>> > >  "));
>> > >  }
>> > >  
>> > > @@ -114,11 +115,15 @@ copy_range_f(int argc, char **argv)
>> > >                 }
>> > >                 break;
>> > >            case 'f':
>> > > -               src_file_nr = atoi(argv[1]);
>> > > -               if (src_file_nr < 0 || src_file_nr >= filecount) {
>> > > -                    printf(_("file value %d is out of range (0-%d)\n"),
>> > > -                         src_file_nr, filecount - 1);
>> > > -                    return 0;
>> > > +               if (strcmp(argv[1], "-"))
>> > > +                    src_file_nr = (file - &filetable[0]) / sizeof(fileio_t);
>> > > +               else {
>> > > +                    src_file_nr = atoi(argv[1]);
>> > > +                    if (src_file_nr < 0 || src_file_nr >= filecount) {
>> > > +                         printf(_("file value %d is out of range (0-%d)\n"),
>> > > +                              src_file_nr, filecount - 1);
>> > > +                         return 0;
>> > > +                    }
>> > >                 }
>> > >                 /* Expect no src_path arg */
>> > >                 src_path_arg = 0;
>> > > @@ -147,10 +152,14 @@ copy_range_f(int argc, char **argv)
>> > >            }
>> > >            len = sz;
>> > >  
>> > > -          ret = copy_dst_truncate();
>> > > -          if (ret < 0) {
>> > > -               ret = 1;
>> > > -               goto out;
>> > > +          if (fd != file->fd) {
>> > > +               ret = copy_dst_truncate();
>> > > +               if (ret < 0) {
>> > > +                    ret = 1;
>> > > +                    goto out;
>> > > +               }
>> > > +          } else {
>> > > +               dst = sz;
>> > >            }
>> > >       }
>> > >  
>> > > --
>> > > 2.17.2
>> > >




[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux