io_uring/rsrc: require cloned buffers to share accounting contexts

When IORING_REGISTER_CLONE_BUFFERS is used to clone buffers from uring
instance A to uring instance B, where A and B use different MMs for
accounting, the accounting can go wrong:
If uring instance A is closed before uring instance B, the pinned memory
counters for uring instance B will be decremented, even though the pinned
memory was originally accounted through uring instance A; so the MM of
uring instance B can end up with negative locked memory.

Cc: stable@vger.kernel.org
Closes: https://lore.kernel.org/r/CAG48ez1zez4bdhmeGLEFxtbFADY4Czn3CV0u9d_TMcbvRA01bg@mail.gmail.com
Fixes: 7cc2a6eadcd7 ("io_uring: add IORING_REGISTER_COPY_BUFFERS method")
Signed-off-by: Jann Horn <jannh@google.com>
Link: https://lore.kernel.org/r/20250114-uring-check-accounting-v1-1-42e4145aa743@google.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
index 964a47c8d85ec671401b41f5ccdf99ca42b030c5..688e277f033576ea6d36100b3ea52dccf093495b 100644 (file)
--- a/io_uring/rsrc.c
+++ b/io_uring/rsrc.c
@@ -928,6 +928,13 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
        int i, ret, off, nr;
        unsigned int nbufs;
 
+       /*
+        * Accounting state is shared between the two rings; that only works if
+        * both rings are accounted towards the same counters.
+        */
+       if (ctx->user != src_ctx->user || ctx->mm_account != src_ctx->mm_account)
+               return -EINVAL;
+
        /* if offsets are given, must have nr specified too */
        if (!arg->nr && (arg->dst_off || arg->src_off))
                return -EINVAL;