[PATCH net v2] selftests/net: toeplitz: fix race on tpacket_v3 block close

[PATCH net v2] selftests/net: toeplitz: fix race on tpacket_v3 block close

[Willem de Bruijn]

From: Willem de Bruijn <willemb@google.com>

Avoid race between process wakeup and tpacket_v3 block timeout.

The test waits for cfg_timeout_msec for packets to arrive. Packets
arrive in tpacket_v3 rings, which pass packets ("frames") to the
process in batches ("blocks"). The sk waits for req3.tp_retire_blk_tov
msec to release a block.

Set the block timeout lower than the process waiting time, else
the process may find that no block has been released by the time it
scans the socket list. Convert to a ring of more than one, smaller,
blocks with shorter timeouts. Blocks must be page aligned, so >= 64KB.

Fixes: 5ebfb4cc3048 ("selftests/net: toeplitz test")
Signed-off-by: Willem de Bruijn <willemb@google.com>

---

v1->v2: replace while with ; on next line with single-line do while
---
 tools/testing/selftests/net/toeplitz.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/net/toeplitz.c b/tools/testing/selftests/net/toeplitz.c
index 90026a27eac0c..9ba03164d73a6 100644
--- a/tools/testing/selftests/net/toeplitz.c
+++ b/tools/testing/selftests/net/toeplitz.c
@@ -215,7 +215,7 @@ static char *recv_frame(const struct ring_state *ring, char *frame)
 }
 
 /* A single TPACKET_V3 block can hold multiple frames */
-static void recv_block(struct ring_state *ring)
+static bool recv_block(struct ring_state *ring)
 {
 	struct tpacket_block_desc *block;
 	char *frame;
@@ -223,7 +223,7 @@ static void recv_block(struct ring_state *ring)
 
 	block = (void *)(ring->mmap + ring->idx * ring_block_sz);
 	if (!(block->hdr.bh1.block_status & TP_STATUS_USER))
-		return;
+		return false;
 
 	frame = (char *)block;
 	frame += block->hdr.bh1.offset_to_first_pkt;
@@ -235,6 +235,8 @@ static void recv_block(struct ring_state *ring)
 
 	block->hdr.bh1.block_status = TP_STATUS_KERNEL;
 	ring->idx = (ring->idx + 1) % ring_block_nr;
+
+	return true;
 }
 
 /* simple test: sleep once unconditionally and then process all rings */
@@ -245,7 +247,7 @@ static void process_rings(void)
 	usleep(1000 * cfg_timeout_msec);
 
 	for (i = 0; i < num_cpus; i++)
-		recv_block(&rings[i]);
+		do {} while (recv_block(&rings[i]));
 
 	fprintf(stderr, "count: pass=%u nohash=%u fail=%u\n",
 		frames_received - frames_nohash - frames_error,
@@ -257,12 +259,12 @@ static char *setup_ring(int fd)
 	struct tpacket_req3 req3 = {0};
 	void *ring;
 
-	req3.tp_retire_blk_tov = cfg_timeout_msec;
+	req3.tp_retire_blk_tov = cfg_timeout_msec / 8;
 	req3.tp_feature_req_word = TP_FT_REQ_FILL_RXHASH;
 
 	req3.tp_frame_size = 2048;
 	req3.tp_frame_nr = 1 << 10;
-	req3.tp_block_nr = 2;
+	req3.tp_block_nr = 16;
 
 	req3.tp_block_size = req3.tp_frame_size * req3.tp_frame_nr;
 	req3.tp_block_size /= req3.tp_block_nr;
-- 
2.39.0.314.g84b9a713c41-goog

Re: [PATCH net v2] selftests/net: toeplitz: fix race on tpacket_v3 block close

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (master)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 18 Jan 2023 10:18:47 -0500 you wrote:
> From: Willem de Bruijn <willemb@google.com>
> 
> Avoid race between process wakeup and tpacket_v3 block timeout.
> 
> The test waits for cfg_timeout_msec for packets to arrive. Packets
> arrive in tpacket_v3 rings, which pass packets ("frames") to the
> process in batches ("blocks"). The sk waits for req3.tp_retire_blk_tov
> msec to release a block.
> 
> [...]

Here is the summary with links:
  - [net,v2] selftests/net: toeplitz: fix race on tpacket_v3 block close
    https://git.kernel.org/netdev/net/c/903848249a78

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html