Message-ID: <34d65539-015e-23c8-cf5e-f34bd5795e52@sberdevices.ru>
Date:   Tue, 14 Mar 2023 13:41:03 +0300
From:   Arseniy Krasnov <avkrasnov@...rdevices.ru>
To:     Stefan Hajnoczi <stefanha@...hat.com>,
        Stefano Garzarella <sgarzare@...hat.com>,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        Bobby Eshleman <bobby.eshleman@...edance.com>
CC:     <kvm@...r.kernel.org>, <virtualization@...ts.linux-foundation.org>,
        <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <kernel@...rdevices.ru>, <oxffffaa@...il.com>,
        <avkrasnov@...rdevices.ru>
Subject: [PATCH net v4 0/4] several updates to virtio/vsock

Hello,

this patchset evolved from previous v2 version (see link below). It does
several updates to virtio/vsock:
1) Changes 'virtio_transport_inc/dec_rx_pkt()' interface. Now instead of
   using skbuff state ('head' and 'data' pointers) to update 'fwd_cnt'
   and 'rx_bytes', integer value is passed as an input argument. This
   makes code more simple, because in this case we don't need to update
   skbuff state before calling 'virtio_transport_inc/dec_rx_pkt()'. In
   more common words - we don't need to change skbuff state to update
   'rx_bytes' and 'fwd_cnt' correctly.
2) For SOCK_STREAM, when copying data to user fails, current skbuff is
   not dropped. Next read attempt will use same skbuff and last offset.
   Instead of 'skb_dequeue()', 'skb_peek()' + '__skb_unlink()' are used.
   This behaviour was implemented before skbuff support.
3) For SOCK_SEQPACKET it removes unneeded 'skb_pull()' call, because for
   this type of socket each skbuff is used only once: after removing it
   from socket's queue, it will be freed anyway.

Test for 2) also added:
Test tries to 'recv()' data to NULL buffer, then does 'recv()' with valid
buffer. For SOCK_STREAM second 'recv()' must return data, because skbuff
must not be dropped, but for SOCK_SEQPACKET skbuff will be dropped by
kernel, and 'recv()' will return EAGAIN.

Link to v1 on lore:
https://lore.kernel.org/netdev/c2d3e204-89d9-88e9-8a15-3fe027e56b4b@sberdevices.ru/

Link to v2 on lore:
https://lore.kernel.org/netdev/a7ab414b-5e41-c7b6-250b-e8401f335859@sberdevices.ru/

Link to v3 on lore:
https://lore.kernel.org/netdev/0abeec42-a11d-3a51-453b-6acf76604f2e@sberdevices.ru/

Change log:

v1 -> v2:
 - For SOCK_SEQPACKET call 'skb_pull()' also in case of copy failure or
   dropping skbuff (when we just waiting message end).
 - Handle copy failure for SOCK_STREAM in the same manner (plus free
   current skbuff).
 - Replace bug repdroducer with new test in vsock_test.c

v2 -> v3:
 - Replace patch which removes 'skb->len' subtraction from function
   'virtio_transport_dec_rx_pkt()' with patch which updates functions
   'virtio_transport_inc/dec_rx_pkt()' by passing integer argument
   instead of skbuff pointer.
 - Replace patch which drops skbuff when copying to user fails with
   patch which changes this behaviour by keeping skbuff in queue until
   it has no data.
 - Add patch for SOCK_SEQPACKET which removes redundant 'skb_pull()'
   call on read.
 - I remove "Fixes" tag from all patches, because all of them now change
   code logic, not only fix something.

v3 -> v4:
 - Update commit messages in all patches except test.
 - Add "Fixes" tag to all patches except test.

Arseniy Krasnov (4):
  virtio/vsock: don't use skbuff state to account credit
  virtio/vsock: remove redundant 'skb_pull()' call
  virtio/vsock: don't drop skbuff on copy failure
  test/vsock: copy to user failure test

 net/vmw_vsock/virtio_transport_common.c |  29 +++---
 tools/testing/vsock/vsock_test.c        | 118 ++++++++++++++++++++++++
 2 files changed, 131 insertions(+), 16 deletions(-)

-- 
2.25.1

Powered by blists - more mailing lists