netdev - Re: [bpf-next] selftests/bpf: fix race in test_tcp

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 16 Aug 2019 09:13:02 -0700
From:   Stanislav Fomichev <sdf@...ichev.me>
To:     Petar Penkov <ppenkov.kernel@...il.com>
Cc:     netdev@...r.kernel.org, bpf@...r.kernel.org, davem@...emloft.net,
        ast@...nel.org, daniel@...earbox.net, sdf@...gle.com,
        Petar Penkov <ppenkov@...gle.com>
Subject: Re: [bpf-next] selftests/bpf: fix race in test_tcp_rtt test

On 08/16, Petar Penkov wrote:
> From: Petar Penkov <ppenkov@...gle.com>
> 
> There is a race in this test between receiving the ACK for the
> single-byte packet sent in the test, and reading the values from the
> map.
> 
> This patch fixes this by having the client wait until there are no more
> unacknowledged packets.
> 
> Before:
> for i in {1..1000}; do ../net/in_netns.sh ./test_tcp_rtt; \
> done | grep -c PASSED
> < trimmed error messages >
> 993
> 
> After:
> for i in {1..10000}; do ../net/in_netns.sh ./test_tcp_rtt; \
> done | grep -c PASSED
> 10000
> 
> Fixes: b55873984dab ("selftests/bpf: test BPF_SOCK_OPS_RTT_CB")
> Signed-off-by: Petar Penkov <ppenkov@...gle.com>
> ---
>  tools/testing/selftests/bpf/test_tcp_rtt.c | 31 ++++++++++++++++++++++
>  1 file changed, 31 insertions(+)
> 
> diff --git a/tools/testing/selftests/bpf/test_tcp_rtt.c b/tools/testing/selftests/bpf/test_tcp_rtt.c
> index 90c3862f74a8..2b4754473956 100644
> --- a/tools/testing/selftests/bpf/test_tcp_rtt.c
> +++ b/tools/testing/selftests/bpf/test_tcp_rtt.c
> @@ -6,6 +6,7 @@
>  #include <sys/types.h>
>  #include <sys/socket.h>
>  #include <netinet/in.h>
> +#include <netinet/tcp.h>
>  #include <pthread.h>
>  
>  #include <linux/filter.h>
> @@ -34,6 +35,30 @@ static void send_byte(int fd)
>  		error(1, errno, "Failed to send single byte");
>  }
>  
> +static int wait_for_ack(int fd, int retries)
> +{
> +	struct tcp_info info;
> +	socklen_t optlen;
> +	int i, err;
> +
> +	for (i = 0; i < retries; i++) {
> +		optlen = sizeof(info);
> +		err = getsockopt(fd, SOL_TCP, TCP_INFO, &info, &optlen);
> +		if (err < 0) {
> +			log_err("Failed to lookup TCP stats");
> +			return err;
> +		}
> +
> +		if (info.tcpi_unacked == 0)
> +			return 0;
> +
> +		sleep(1);
Isn't it too big of a hammer? Maybe usleep(10) here and do x100 retries
instead?

> +	}
> +
> +	log_err("Did not receive ACK");
> +	return -1;
> +}
> +
>  static int verify_sk(int map_fd, int client_fd, const char *msg, __u32 invoked,
>  		     __u32 dsack_dups, __u32 delivered, __u32 delivered_ce,
>  		     __u32 icsk_retransmits)
> @@ -149,6 +174,11 @@ static int run_test(int cgroup_fd, int server_fd)
>  			 /*icsk_retransmits=*/0);
>  
>  	send_byte(client_fd);
> +	if (wait_for_ack(client_fd, 5) < 0) {
> +		err = -1;
> +		goto close_client_fd;
> +	}
> +
>  
>  	err += verify_sk(map_fd, client_fd, "first payload byte",
>  			 /*invoked=*/2,
> @@ -157,6 +187,7 @@ static int run_test(int cgroup_fd, int server_fd)
>  			 /*delivered_ce=*/0,
>  			 /*icsk_retransmits=*/0);
>  
> +close_client_fd:
>  	close(client_fd);
>  
>  close_bpf_object:
> -- 
> 2.23.0.rc1.153.gdeed80330f-goog
>