On Wed, Oct 31, 2018 at 12:54 PM J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote: > > On Fri, Oct 19, 2018 at 11:28:53AM -0400, Olga Kornievskaia wrote: > > From: Anna Schumaker <Anna.Schumaker@xxxxxxxxxx> > > I have some idea we've had some discussion about this before, but if so > I've forgotten the conclusion. Could we have more of a changelog?: > > - isn't there a race condition, or is there something preventing > the file size from changing here? No there is nothing preventing the size from changing. Just like there is nothing that prevents the file from changing if you are doing a traditional copy either. > - why are we doing this? Does this change the behavior of > copy_file_range()? We are doing this because 1. NFS spec and 2. copy_file_range semantics mandate that too. There is a whole different discussion under the client-side patch for this where the plan now is that VFS themselves are interested in making sure they are indeed enforcing the check stated by the documentation of copy_file_range call which states "copying a range beyond the end of the file" is EINVAL. I recall you argued for a "short" read instead of a EINVAL but unless VFS community is convinced to change it it'll be enforced (soon). > > --b. > > > > > Signed-off-by: Anna Schumaker <Anna.Schumaker@xxxxxxxxxx> > > --- > > fs/read_write.c | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/fs/read_write.c b/fs/read_write.c > > index 39b4a21..c60790f 100644 > > --- a/fs/read_write.c > > +++ b/fs/read_write.c > > @@ -1570,6 +1570,9 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in, > > if (unlikely(ret)) > > return ret; > > > > + if (pos_in >= i_size_read(inode_in)) > > + return -EINVAL; > > + > > if (!(file_in->f_mode & FMODE_READ) || > > !(file_out->f_mode & FMODE_WRITE) || > > (file_out->f_flags & O_APPEND)) > > -- > > 1.8.3.1
