Re: [PATCH bpf-next v2 0/5] selftests/bpf: xsk selftests

On 2020-11-23 13:20, Björn Töpel wrote:

On 2020-11-21 01:31, Yonghong Song wrote:

quoted

On 11/20/20 5:00 AM, Weqaar Janjua wrote:

quoted

This patch set adds AF_XDP selftests based on veth to selftests/bpf.

# Topology:
# ---------
#                 -----------
#               _ | Process | _
#              /  -----------  \
#             /        |        \
#            /         |         \
#      -----------     |     -----------
#      | Thread1 |     |     | Thread2 |
#      -----------     |     -----------
#           |          |          |
#      -----------     |     -----------
#      |  xskX   |     |     |  xskY   |
#      -----------     |     -----------
#           |          |          |
#      -----------     |     ----------
#      |  vethX  | --------- |  vethY |
#      -----------   peer    ----------
#           |          |          |
#      namespaceX      |     namespaceY

These selftests test AF_XDP SKB and Native/DRV modes using veth Virtual
Ethernet interfaces.

The test program contains two threads, each thread is single socket with
a unique UMEM. It validates in-order packet delivery and packet content
by sending packets to each other.

Prerequisites setup by script test_xsk_prerequisites.sh:

    Set up veth interfaces as per the topology shown ^^:
    * setup two veth interfaces and one namespace
    ** veth<xxxx> in root namespace
    ** veth<yyyy> in af_xdp<xxxx> namespace
    ** namespace af_xdp<xxxx>
    * create a spec file veth.spec that includes this run-time 
configuration
      that is read by test scripts - filenames prefixed with test_xsk_
    *** xxxx and yyyy are randomly generated 4 digit numbers used to 
avoid
        conflict with any existing interface

The following tests are provided:

1. AF_XDP SKB mode
    Generic mode XDP is driver independent, used when the driver does
    not have support for XDP. Works on any netdevice using sockets and
    generic XDP path. XDP hook from netif_receive_skb().
    a. nopoll - soft-irq processing
    b. poll - using poll() syscall
    c. Socket Teardown
       Create a Tx and a Rx socket, Tx from one socket, Rx on another.
       Destroy both sockets, then repeat multiple times. Only nopoll 
mode
      is used
    d. Bi-directional Sockets
       Configure sockets as bi-directional tx/rx sockets, sets up fill
      and completion rings on each socket, tx/rx in both directions.
      Only nopoll mode is used

2. AF_XDP DRV/Native mode
    Works on any netdevice with XDP_REDIRECT support, driver dependent.
    Processes packets before SKB allocation. Provides better performance
    than SKB. Driver hook available just after DMA of buffer descriptor.
    a. nopoll
    b. poll
    c. Socket Teardown
    d. Bi-directional Sockets
    * Only copy mode is supported because veth does not currently 
support
      zero-copy mode

Total tests: 8

Flow:
* Single process spawns two threads: Tx and Rx
* Each of these two threads attach to a veth interface within their
   assigned namespaces
* Each thread creates one AF_XDP socket connected to a unique umem
   for each veth interface
* Tx thread transmits 10k packets from veth<xxxx> to veth<yyyy>
* Rx thread verifies if all 10k packets were received and delivered
   in-order, and have the right content

v2 changes:
* Move selftests/xsk to selftests/bpf
* Remove Makefiles under selftests/xsk, and utilize 
selftests/bpf/Makefile

Structure of the patch set:

Patch 1: This patch adds XSK Selftests framework under selftests/bpf
Patch 2: Adds tests: SKB poll and nopoll mode, and mac-ip-udp debug
Patch 3: Adds tests: DRV poll and nopoll mode
Patch 4: Adds tests: SKB and DRV Socket Teardown
Patch 5: Adds tests: SKB and DRV Bi-directional Sockets

I just want to report that after applying the above 5 patches
on top of bpf-next commit 450d060e8f75 ("bpftool: Add {i,d}tlb_misses 
support for bpftool profile"), I hit the following error with below 
command sequences:

  $ ./test_xsk_prerequisites.sh
  $ ./test_xsk_skb_poll.sh
# Interface found: ve1480
# Interface found: ve9258
# NS switched: af_xdp9258
1..1
# Interface [ve9258] vector [Rx]
# Interface [ve1480] vector [Tx]
# Sending 10000 packets on interface ve1480
[  331.741244] ------------[ cut here ]------------
[  331.741741] kernel BUG at net/core/skbuff.c:1621!
[  331.742265] invalid opcode: 0000 [#1] PREEMPT SMP PTI
[  331.742837] CPU: 0 PID: 1883 Comm: xdpxceiver Not tainted 
5.10.0-rc3+ #1037
[  331.743468] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), 
BIOS 1.9.3
-1.el7.centos 04/01/2014
[  331.744300] RIP: 0010:pskb_expand_head+0x27b/0x310

Ugh, looks like the tests are working. :-P

This is a BUG_ON(skb_shared(skb)) trigger, related to the skbuff 
refcount changes done recently in AF_XDP.

I'll cook a patch! Thanks for the report!

Posted a fix [1].

Please not that it's for the bpf tree, so when Weqaar pushes the v3 of
the selftests to bpf-next, [1] needs to be pulled in.



Björn

[1] 
https://lore.kernel.org/bpf/20201123131215.136131-1-bjorn.topel@gmail.com/ (local)

Björn