lkml.org 
[lkml]   [2019]   [Oct]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [PATCH] ceph: don't allow copy_file_range when stripe_count != 1
From
Date
On Thu, 2019-10-31 at 11:49 +0000, Luis Henriques wrote:
> copy_file_range tries to use the OSD 'copy-from' operation, which simply
> performs a full object copy. Unfortunately, the implementation of this
> system call assumes that stripe_count is always set to 1 and doesn't take
> into account that the data may be striped across an object set. If the
> file layout has stripe_count different from 1, then the destination file
> data will be corrupted.
>
> For example:
>
> Consider a 8 MiB file with 4 MiB object size, stripe_count of 2 and
> stripe_size of 2 MiB; the first half of the file will be filled with 'A's
> and the second half will be filled with 'B's:
>
> 0 4M 8M Obj1 Obj2
> +------+------+ +----+ +----+
> file: | AAAA | BBBB | | AA | | AA |
> +------+------+ |----| |----|
> | BB | | BB |
> +----+ +----+
>
> If we copy_file_range this file into a new file (which needs to have the
> same file layout!), then it will start by copying the object starting at
> file offset 0 (Obj1). And then it will copy the object starting at file
> offset 4M -- which is Obj1 again.
>
> Unfortunately, the solution for this is to not allow remote object copies
> to be performed when the file layout stripe_count is not 1 and simply
> fallback to the default (VFS) copy_file_range implementation.
>
> Signed-off-by: Luis Henriques <lhenriques@suse.com>
> ---
> Hi Jeff,
>
> I hope my understanding of the whole file striping in CephFS is correct;
> I had to go re-read the whole thing to refresh my memory.
>
> Anyway, I guess that this is not really the only solution to this
> problem, but it's definitely the simplest one. copy_file_range is
> already way more complex that I had ever anticipated. I would rather
> keep this simple solution instead of adding more complexity and cover
> more corner cases. But yeah, we may want to revisit this in the
> future...
>
> [OOT: files layout is probably one of the biggest headaches to sort out
> the day we want to implement something like FIEMAP on CephFS ;-) ]
>
> Cheers,
> --
> Luis
>
> fs/ceph/file.c | 7 +++++--
> 1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index d277f71abe0b..3b0e6f9eb6a6 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -1957,9 +1957,12 @@ static ssize_t __ceph_copy_file_range(struct file *src_file, loff_t src_off,
> return -EOPNOTSUPP;
>
> if ((src_ci->i_layout.stripe_unit != dst_ci->i_layout.stripe_unit) ||
> - (src_ci->i_layout.stripe_count != dst_ci->i_layout.stripe_count) ||
> - (src_ci->i_layout.object_size != dst_ci->i_layout.object_size))
> + (src_ci->i_layout.stripe_count != 1) ||
> + (dst_ci->i_layout.stripe_count != 1) ||
> + (src_ci->i_layout.object_size != dst_ci->i_layout.object_size)) {
> + dout("Invalid src/dst files layout\n");
> return -EOPNOTSUPP;
> + }
>
> if (len < src_ci->i_layout.object_size)
> return -EOPNOTSUPP; /* no remote copy will be done */

I'm fine with restricting CFR to very simple cases, at least initially.
We can always expand it later once the need becomes clear.

That said, we should probably add a comment explaining why we're
excluding cases where the stripe count != 1 here. It doesn't need to
contain the whole commit log message you wrote, but anyone that does
want to improve this later might appreciate some breadcrumbs.

Maybe something like:

/*
* Striped file layouts require that we copy partial objects,
* but the OSD copy-from operation only supports full-object copies.
* Limit this to non-striped file layouts for now.
*/

If that sounds ok, I'll add that in and merge this later today.
Thanks,
--
Jeff Layton <jlayton@kernel.org>

\
 
 \ /
  Last update: 2019-10-31 16:29    [W:0.052 / U:0.176 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site