diff mbox series

[net-next] selftests: net/psock_lib: Handle EINTR and EAGAIN

Message ID 20241117225950.138968-1-leocstone@gmail.com
State New
Headers show
Series [net-next] selftests: net/psock_lib: Handle EINTR and EAGAIN | expand

Commit Message

Leo Stone Nov. 17, 2024, 10:59 p.m. UTC
Make pair_udp_send_char handle EAGAIN, EINTR, and partial reads or
writes.

Signed-off-by: Leo Stone <leocstone@gmail.com>
---
 tools/testing/selftests/net/psock_lib.h | 39 +++++++++++++++++++------
 1 file changed, 30 insertions(+), 9 deletions(-)

Comments

Willem de Bruijn Nov. 18, 2024, 3:02 p.m. UTC | #1
Leo Stone wrote:
> Make pair_udp_send_char handle EAGAIN, EINTR, and partial reads or
> writes.
> 
> Signed-off-by: Leo Stone <leocstone@gmail.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
>
Leo Stone Nov. 21, 2024, 1:25 a.m. UTC | #2
Willem de Bruijn wrote:
> Did you observe actual issues or is this based on the comment in the
> existing code ("Should really handle EINTR and EAGAIN").

No, this patch was based purely off of the comment, and the examples of
similar code in other selftests, e.g. tools/testing/selftests/bpf/xsk.c:

> static int netlink_recvmsg(int sock, struct msghdr *mhdr, int flags)
> {
>       int len;
>
>       do {
>               len = recvmsg(sock, mhdr, flags);
>       } while (len < 0 && (errno == EINTR || errno == EAGAIN));
>
>       if (len < 0)
>               return -errno;
>       return len;
> }

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

If we want to handle it that way, this patch is probably not necessary,
since there is no specific signal I would want to install a handler for.

Thanks for your feedback,
Leo
diff mbox series

Patch

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;
+			} 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);