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

On Mon, 2 Nov 2020 at 23:08, Daniel Borkmann daniel@iogearbox.net wrote:

On 10/30/20 1:13 PM, Weqaar Janjua wrote:

...

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

# 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_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

Structure of the patch set:

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

Thanks: Weqaar

Weqaar Janjua (5): selftests/xsk: xsk selftests framework selftests/xsk: xsk selftests - SKB POLL, NOPOLL selftests/xsk: xsk selftests - DRV POLL, NOPOLL selftests/xsk: xsk selftests - Socket Teardown - SKB, DRV selftests/xsk: xsk selftests - Bi-directional Sockets - SKB, DRV

Thanks a lot for adding the selftests, Weqaar! Given this needs to copy quite a bit of BPF selftest base infra e.g. from Makefiles I'd prefer if you could place these under selftests/bpf/ instead to avoid duplicating changes into two locations. I understand that these tests don't integrate well into test_progs, but for example see test_tc_redirect.sh or test_tc_edt.sh for stand-alone tests which could be done similarly with the xsk ones. Would be great if you could integrate them and spin a v2 with that.

Thanks, Daniel

Hi Daniel,

Appreciate the pointers and suggestions which I will re-evaluate against merging of selftests/xsk into selftests/bpf, let me explain a bit to get your opinion re-evaluated on this - perhaps selftests/xsk could still work (as per my clarifications below) or we somehow have selftests/bpf/Makefile trigger selftests/bpf/test_xsk/Makefile<or whatever>.

I had a look at selftests/bpf earlier, the problem was the same as you indicated - xsk tests do not integrate well into selftests/bpf as the semantics in the top level Makefile for both do no merge well - the way xsk tests are designed is systematically different.

If you look closely into the patch set patches -> 2/5 - 5/5, that is where the major difference shows up, xsk/xdpprogs/xdpxceiver.* is a self-contained binary utilized for testing and it is a major piece of code: tools/testing/selftests/xsk/xdpprogs/Makefile | 64 ++ .../selftests/xsk/xdpprogs/Makefile.target | 68 ++ .../selftests/xsk/xdpprogs/xdpxceiver.c | 1000 +++++++++++++++++ .../selftests/xsk/xdpprogs/xdpxceiver.h | 159 +++ ... 21 files changed, 1833 insertions(+)

Bits that -> copy quite a bit of BPF selftest base infra isn't the top-level selftests/xsk/Makefile|*, it is these (patches -> 2/5 - 5/5): tools/testing/selftests/xsk/xdpprogs/Makefile | 64 ++ .../selftests/xsk/xdpprogs/Makefile.target | 68 ++

This is 132++ of 1833++ or ~7% of similarity.

I had a look at these today, and some other tests earlier: - test_tc_redirect.sh -> test_xdp_redirect.sh (perhaps you meant this?) - test_tc_edt.sh

Patch 1/5 might look similar to these^^ as that sets up the base xsk test framework, but the xsk tests really start with patches -> 2/5 - 5/5 which will give you a better picture, please let me know if this clarifies my intent for selftests/xsk/, or otherwise if you insist merging into selftests/bpf, I will then go ahead and work it out accordingly.

Thanks,

/Weqaar