vsock/test: fix send_buf()/recv_buf() EINTR handling
authorStefano Garzarella <sgarzare@redhat.com>
Fri, 3 Apr 2026 09:32:51 +0000 (11:32 +0200)
committerJakub Kicinski <kuba@kernel.org>
Tue, 7 Apr 2026 01:46:03 +0000 (18:46 -0700)
When send() or recv() returns -1 with errno == EINTR, the code skips
the break but still adds the return value to nwritten/nread, making it
decrease by 1. This leads to wrong buffer offsets and wrong bytes count.

Fix it by explicitly continuing the loop on EINTR, so the return value
is only added when it is positive.

Fixes: a8ed71a27ef5 ("vsock/test: add recv_buf() utility function")
Fixes: 12329bd51fdc ("vsock/test: add send_buf() utility function")
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Luigi Leonardi <leonardi@redhat.com>
Link: https://patch.msgid.link/20260403093251.30662-1-sgarzare@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
tools/testing/vsock/util.c

index 9430ef5..1fe1338 100644 (file)
@@ -344,7 +344,9 @@ void send_buf(int fd, const void *buf, size_t len, int flags,
                ret = send(fd, buf + nwritten, len - nwritten, flags);
                timeout_check("send");
 
-               if (ret == 0 || (ret < 0 && errno != EINTR))
+               if (ret < 0 && errno == EINTR)
+                       continue;
+               if (ret <= 0)
                        break;
 
                nwritten += ret;
@@ -396,7 +398,9 @@ void recv_buf(int fd, void *buf, size_t len, int flags, ssize_t expected_ret)
                ret = recv(fd, buf + nread, len - nread, flags);
                timeout_check("recv");
 
-               if (ret == 0 || (ret < 0 && errno != EINTR))
+               if (ret < 0 && errno == EINTR)
+                       continue;
+               if (ret <= 0)
                        break;
 
                nread += ret;