[PATCH 3/4] um: chan_user: retry partial writes

[benjamin at sipsolutions.net]

From: Benjamin Berg <benjamin.berg at intel.com>

In the next commit, we are going to set the output FD to be blocking.
Once that is done, the write() may be short if an interrupt happens
while it is writing out data. As such, to properly catch an EINTR error,
we need to retry the write.

Signed-off-by: Benjamin Berg <benjamin.berg at intel.com>
---
 arch/um/drivers/chan_user.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/arch/um/drivers/chan_user.c b/arch/um/drivers/chan_user.c
index 29c68820522a..339aae56b6f1 100644
--- a/arch/um/drivers/chan_user.c
+++ b/arch/um/drivers/chan_user.c
@@ -37,11 +37,23 @@ int generic_read(int fd, char *c_out, void *unused)
 
 int generic_write(int fd, const char *buf, int n, void *unused)
 {
+	int written = 0;
 	int err;
 
-	CATCH_EINTR(err = write(fd, buf, n));
-	if (err > 0)
-		return err;
+	/* The FD may be in blocking mode, as such, need to retry short writes,
+	 * they may have been interrupted by a signal.
+	 */
+	do {
+		errno = 0;
+		err = write(fd, buf + written, n - written);
+		if (err > 0) {
+			written += err;
+			continue;
+		}
+	} while (err < 0 && errno == EINTR);
+
+	if (written > 0)
+		return written;
 	else if (errno == EAGAIN)
 		return 0;
 	else if (err == 0)
-- 
2.41.0