netdev - Re: [PATCH net-next] selftests: net/psock

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]

Message-ID: <673b56f7768fc_1ced1f294fa@willemb.c.googlers.com.notmuch>
Date: Mon, 18 Nov 2024 10:02:15 -0500
From: Willem de Bruijn <willemdebruijn.kernel@...il.com>
To: Leo Stone <leocstone@...il.com>, 
 davem@...emloft.net, 
 edumazet@...gle.com, 
 kuba@...nel.org, 
 pabeni@...hat.com, 
 horms@...nel.org, 
 shuah@...nel.org
Cc: Leo Stone <leocstone@...il.com>, 
 netdev@...r.kernel.org, 
 linux-kselftest@...r.kernel.org, 
 linux-kernel@...r.kernel.org, 
 linux-kernel-mentees@...ts.linuxfoundation.org
Subject: Re: [PATCH net-next] selftests: net/psock_lib: Handle EINTR and
 EAGAIN

Leo Stone wrote:
> Make pair_udp_send_char handle EAGAIN, EINTR, and partial reads or
> writes.
> 
> Signed-off-by: Leo Stone <leocstone@...il.com>

Did you observe actual issues or is this based on the comment in the
existing code ("Should really handle EINTR and EAGAIN").

AFAIK all the users of psock_lib use default blocking IO, so should
not see EAGAIN.

A simpler approach to dealing with EINTR is to ask glibc to restart
with sigaction or siginterrupt.

> ---
>  tools/testing/selftests/net/psock_lib.h | 39 +++++++++++++++++++------
>  1 file changed, 30 insertions(+), 9 deletions(-)
> 
> diff --git a/tools/testing/selftests/net/psock_lib.h b/tools/testing/selftests/net/psock_lib.h
> index 6e4fef560873..e28bff95c8e2 100644
> --- a/tools/testing/selftests/net/psock_lib.h
> +++ b/tools/testing/selftests/net/psock_lib.h
> @@ -13,6 +13,7 @@
>  #include <string.h>
>  #include <arpa/inet.h>
>  #include <unistd.h>
> +#include <stdbool.h>
>  
>  #include "kselftest.h"
>  
> @@ -110,21 +111,41 @@ static __maybe_unused void pair_udp_open(int fds[], uint16_t port)
>  	}
>  }
>  
> +static void read_write_checked(int fd, char *buf, size_t sz, bool is_write)
> +{
> +	int bytes_processed = 0;
> +	int ret;
> +
> +	do {
> +		if (is_write)
> +			ret = write(fd, buf + bytes_processed,
> +				    sz - bytes_processed);
> +		else
> +			ret = read(fd, buf + bytes_processed,
> +				   sz - bytes_processed);
> +		if (ret == -1) {
> +			if (errno == EAGAIN || errno == EINTR) {
> +				continue;

Instead of busy looping, on return with EAGAIN should probably block.

> +			} else {
> +				fprintf(stderr, "ERROR: %s failed, bytes left=%lu\n",
> +					is_write ? "send" : "recv",
> +					sz - bytes_processed);
> +				exit(1);
> +			}
> +		}
> +		bytes_processed += ret;
> +	} while (bytes_processed < sz);
> +}
> +
>  static __maybe_unused void pair_udp_send_char(int fds[], int num, char payload)
>  {
>  	char buf[DATA_LEN], rbuf[DATA_LEN];
>  
>  	memset(buf, payload, sizeof(buf));
>  	while (num--) {
> -		/* Should really handle EINTR and EAGAIN */
> -		if (write(fds[0], buf, sizeof(buf)) != sizeof(buf)) {
> -			fprintf(stderr, "ERROR: send failed left=%d\n", num);
> -			exit(1);
> -		}
> -		if (read(fds[1], rbuf, sizeof(rbuf)) != sizeof(rbuf)) {
> -			fprintf(stderr, "ERROR: recv failed left=%d\n", num);
> -			exit(1);
> -		}
> +		read_write_checked(fds[0], buf, sizeof(buf), true);
> +		read_write_checked(fds[1], rbuf, sizeof(rbuf), false);
> +
>  		if (memcmp(buf, rbuf, sizeof(buf))) {
>  			fprintf(stderr, "ERROR: data failed left=%d\n", num);
>  			exit(1);
> -- 
> 2.39.5
>