@@ -1063,13 +1063,17 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
*/
bytes = 0;
len = sd->total_len;
+
+ /* Don't block on output, we have to drain the direct pipe. */
flags = sd->flags;
+ sd->flags &= ~SPLICE_F_NONBLOCK;
/*
- * Don't block on output, we have to drain the direct pipe.
+ * We signal MORE until we've read sufficient data to fulfill the
+ * request and we keep signalling it if the caller set it.
*/
- sd->flags &= ~SPLICE_F_NONBLOCK;
more = sd->flags & SPLICE_F_MORE;
+ sd->flags |= SPLICE_F_MORE;
WARN_ON_ONCE(!pipe_empty(pipe->head, pipe->tail));
@@ -1085,14 +1089,12 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
sd->total_len = read_len;
/*
- * If more data is pending, set SPLICE_F_MORE
- * If this is the last data and SPLICE_F_MORE was not set
- * initially, clears it.
+ * If we now have sufficient data to fulfill the request then
+ * we clear SPLICE_F_MORE if it was not set initially.
*/
- if (read_len < len)
- sd->flags |= SPLICE_F_MORE;
- else if (!more)
+ if (read_len >= len && !more)
sd->flags &= ~SPLICE_F_MORE;
+
/*
* NOTE: nonblocking mode only applies to the input. We
* must not do the output in nonblocking mode as then we
splice_direct_to_actor() doesn't manage SPLICE_F_MORE correctly[1] - and, as a result, it incorrectly signals/fails to signal MSG_MORE when splicing to a socket. The problem I'm seeing happens when a short splice occurs because we got a short read due to hitting the EOF on a file: as the length read (read_len) is less than the remaining size to be spliced (len), SPLICE_F_MORE (and thus MSG_MORE) is set. The issue is that, for the moment, we have no way to know *why* the short read occurred and so can't make a good decision on whether we *should* keep MSG_MORE set. MSG_SENDPAGE_NOTLAST was added to work around this, but that is also set incorrectly under some circumstances - for example if a short read fills a single pipe_buffer, but the next read would return more (seqfile can do this). This was observed with the multi_chunk_sendfile tests in the tls kselftest program. Some of those tests would hang and time out when the last chunk of file was less than the sendfile request size: build/kselftest/net/tls -r tls.12_aes_gcm.multi_chunk_sendfile This has been observed before[2] and worked around in AF_TLS[3]. Fix this by making splice_direct_to_actor() always signal SPLICE_F_MORE if we haven't yet hit the requested operation size. SPLICE_F_MORE remains signalled if the user passed it in to splice() but otherwise gets cleared when we've read sufficient data to fulfill the request. If, however, we get a premature EOF from ->splice_read(), have sent at least one byte and SPLICE_F_MORE was not set by the caller, ->splice_eof() will be invoked. Signed-off-by: David Howells <dhowells@redhat.com> cc: Linus Torvalds <torvalds@linux-foundation.org> cc: Jakub Kicinski <kuba@kernel.org> cc: Jens Axboe <axboe@kernel.dk> cc: Christoph Hellwig <hch@lst.de> cc: Al Viro <viro@zeniv.linux.org.uk> cc: Matthew Wilcox <willy@infradead.org> cc: Jan Kara <jack@suse.cz> cc: Jeff Layton <jlayton@kernel.org> cc: David Hildenbrand <david@redhat.com> cc: Christian Brauner <brauner@kernel.org> cc: Chuck Lever <chuck.lever@oracle.com> cc: Boris Pismenny <borisp@nvidia.com> cc: John Fastabend <john.fastabend@gmail.com> cc: Eric Dumazet <edumazet@google.com> cc: "David S. Miller" <davem@davemloft.net> cc: Paolo Abeni <pabeni@redhat.com> cc: linux-fsdevel@vger.kernel.org cc: linux-block@vger.kernel.org cc: linux-mm@kvack.org cc: netdev@vger.kernel.org Link: https://lore.kernel.org/r/499791.1685485603@warthog.procyon.org.uk/ [1] Link: https://lore.kernel.org/r/1591392508-14592-1-git-send-email-pooja.trivedi@stackpath.com/ [2] Link: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/?id=d452d48b9f8b1a7f8152d33ef52cfd7fe1735b0a [3] --- Notes: ver #4) - Use ->splice_eof() to signal a premature EOF to the splice output. fs/splice.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-)