Netdev List
 help / color / mirror / Atom feed
From: Petr Oros <poros@redhat.com>
To: netdev@vger.kernel.org
Cc: David Ahern <dsahern@kernel.org>,
	Stephen Hemminger <stephen@networkplumber.org>,
	Ivan Vecera <ivecera@redhat.com>, Jiri Pirko <jiri@resnulli.us>,
	Petr Oros <poros@redhat.com>
Subject: [PATCH iproute2-next 1/2] mnl_utils: add a dump helper for NLM_F_DUMP_INTR
Date: Tue, 10 Nov 2026 12:01:00 +0100	[thread overview]
Message-ID: <20261110110101.2155953-2-poros@redhat.com> (raw)
In-Reply-To: <20261110110101.2155953-1-poros@redhat.com>

The kernel sets NLM_F_DUMP_INTR on a dump message when the dumped
objects changed between two parts of the dump, so the result can miss an
object or list it twice. mnl_cb_run2() answers the flag with EINTR on
the first such message. mnlu_gen_socket_sndrcv() then prints "kernel
answers: Interrupted system call" and returns with the rest of the dump
still queued in the socket. The caller has passed a part of the objects
to its callback by then and cannot tell an interrupted dump from a
request that failed.

Add mnlu_gen_socket_sndrcv_dump(). It clears the flag in every received
message before the callbacks run, reads the dump to its end and tells
the caller through a bool that the dump was interrupted. The caller can
then repeat the request on the same socket or warn the user, as
rtnl_dump_filter() does for rtnetlink dumps.

The receive loop is shared with mnlu_socket_recv_run() and
mnlu_gen_socket_sndrcv(), which behave as before.

Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Petr Oros <poros@redhat.com>
---
 include/mnl_utils.h |  5 ++++
 lib/mnl_utils.c     | 58 +++++++++++++++++++++++++++++++++++++++------
 2 files changed, 56 insertions(+), 7 deletions(-)

diff --git a/include/mnl_utils.h b/include/mnl_utils.h
index 4df9e398e59165..82c26f05742ce3 100644
--- a/include/mnl_utils.h
+++ b/include/mnl_utils.h
@@ -2,6 +2,8 @@
 #ifndef __MNL_UTILS_H__
 #define __MNL_UTILS_H__ 1
 
+#include <stdbool.h>
+
 struct mnlu_gen_socket {
 	struct mnl_socket *nl;
 	char *buf;
@@ -22,6 +24,9 @@ struct nlmsghdr *mnlu_gen_socket_cmd_prepare(struct mnlu_gen_socket *nlg,
 					     uint8_t cmd, uint16_t flags);
 int mnlu_gen_socket_sndrcv(struct mnlu_gen_socket *nlg, const struct nlmsghdr *nlh,
 			   mnl_cb_t data_cb, void *data);
+int mnlu_gen_socket_sndrcv_dump(struct mnlu_gen_socket *nlg,
+				const struct nlmsghdr *nlh,
+				mnl_cb_t data_cb, void *data, bool *intr);
 
 struct mnl_socket *mnlu_socket_open(int bus);
 int mnl_add_nl_group(struct mnl_socket *nl, unsigned int group);
diff --git a/lib/mnl_utils.c b/lib/mnl_utils.c
index 43b4cd1a9e40d4..4b971797777cb3 100644
--- a/lib/mnl_utils.c
+++ b/lib/mnl_utils.c
@@ -102,8 +102,29 @@ static mnl_cb_t mnlu_cb_array[NLMSG_MIN_TYPE] = {
 	[NLMSG_OVERRUN]	= mnlu_cb_noop,
 };
 
-int mnlu_socket_recv_run(struct mnl_socket *nl, unsigned int seq, void *buf, size_t buf_size,
-			 mnl_cb_t cb, void *data)
+/* mnl_cb_run2() fails with EINTR on the first message with NLM_F_DUMP_INTR
+ * and leaves the rest of the dump in the socket. Clear the flag, so that
+ * the dump is read to its end, and tell the caller that it was there.
+ */
+static bool mnlu_dump_intr_clear(void *buf, size_t len)
+{
+	struct nlmsghdr *nlh = buf;
+	bool intr = false;
+	int rem = len;
+
+	for (; mnl_nlmsg_ok(nlh, rem); nlh = mnl_nlmsg_next(nlh, &rem)) {
+		if (nlh->nlmsg_flags & NLM_F_DUMP_INTR) {
+			nlh->nlmsg_flags &= ~NLM_F_DUMP_INTR;
+			intr = true;
+		}
+	}
+
+	return intr;
+}
+
+static int __mnlu_socket_recv_run(struct mnl_socket *nl, unsigned int seq,
+				  void *buf, size_t buf_size,
+				  mnl_cb_t cb, void *data, bool *intr)
 {
 	unsigned int portid = mnl_socket_get_portid(nl);
 	int err;
@@ -112,6 +133,8 @@ int mnlu_socket_recv_run(struct mnl_socket *nl, unsigned int seq, void *buf, siz
 		err = mnl_socket_recvfrom(nl, buf, buf_size);
 		if (err <= 0)
 			break;
+		if (intr)
+			*intr |= mnlu_dump_intr_clear(buf, err);
 		err = mnl_cb_run2(buf, err, seq, portid,
 				  cb, data, mnlu_cb_array,
 				  ARRAY_SIZE(mnlu_cb_array));
@@ -120,6 +143,12 @@ int mnlu_socket_recv_run(struct mnl_socket *nl, unsigned int seq, void *buf, siz
 	return err;
 }
 
+int mnlu_socket_recv_run(struct mnl_socket *nl, unsigned int seq, void *buf, size_t buf_size,
+			 mnl_cb_t cb, void *data)
+{
+	return __mnlu_socket_recv_run(nl, seq, buf, buf_size, cb, data, NULL);
+}
+
 static int ctrl_attrs_cb(const struct nlattr *attr, void *data)
 {
 	int type = mnl_attr_get_type(attr);
@@ -242,20 +271,29 @@ struct nlmsghdr *mnlu_gen_socket_cmd_prepare(struct mnlu_gen_socket *nlg,
 					    nlg->version);
 }
 
-int mnlu_gen_socket_sndrcv(struct mnlu_gen_socket *nlg, const struct nlmsghdr *nlh,
-			   mnl_cb_t data_cb, void *data)
+/* Like mnlu_gen_socket_sndrcv(), for a dump request. A dump the kernel
+ * marked as interrupted is read to its end and reported in @intr. The
+ * objects passed to @data_cb can then be incomplete or duplicated and it
+ * is up to the caller to repeat the dump or to warn the user.
+ */
+int mnlu_gen_socket_sndrcv_dump(struct mnlu_gen_socket *nlg,
+				const struct nlmsghdr *nlh,
+				mnl_cb_t data_cb, void *data, bool *intr)
 {
 	int err;
 
+	if (intr)
+		*intr = false;
+
 	err = mnl_socket_sendto(nlg->nl, nlh, nlh->nlmsg_len);
 	if (err < 0) {
 		perror("Failed to send data");
 		return -errno;
 	}
 
-	err = mnlu_socket_recv_run(nlg->nl, nlh->nlmsg_seq, nlg->buf,
-				   MNL_SOCKET_BUFFER_SIZE,
-				   data_cb, data);
+	err = __mnlu_socket_recv_run(nlg->nl, nlh->nlmsg_seq, nlg->buf,
+				     MNL_SOCKET_BUFFER_SIZE,
+				     data_cb, data, intr);
 	if (err < 0) {
 		fprintf(stderr, "kernel answers: %s\n", strerror(errno));
 		return -errno;
@@ -263,6 +301,12 @@ int mnlu_gen_socket_sndrcv(struct mnlu_gen_socket *nlg, const struct nlmsghdr *n
 	return 0;
 }
 
+int mnlu_gen_socket_sndrcv(struct mnlu_gen_socket *nlg, const struct nlmsghdr *nlh,
+			   mnl_cb_t data_cb, void *data)
+{
+	return mnlu_gen_socket_sndrcv_dump(nlg, nlh, data_cb, data, NULL);
+}
+
 int mnlu_gen_socket_recv_run(struct mnlu_gen_socket *nlg, mnl_cb_t cb,
 			     void *data)
 {
-- 
2.56.0


  reply	other threads:[~2026-11-10 11:01 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-10 11:00 [PATCH iproute2-next 0/2] dpll: do not fail on interrupted dumps Petr Oros
2026-11-10 11:01 ` Petr Oros [this message]
2026-11-18 22:58   ` [PATCH iproute2-next 1/2] mnl_utils: add a dump helper for NLM_F_DUMP_INTR David Ahern
2026-11-10 11:01 ` [PATCH iproute2-next 2/2] dpll: warn about interrupted dumps Petr Oros

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261110110101.2155953-2-poros@redhat.com \
    --to=poros@redhat.com \
    --cc=dsahern@kernel.org \
    --cc=ivecera@redhat.com \
    --cc=jiri@resnulli.us \
    --cc=netdev@vger.kernel.org \
    --cc=stephen@networkplumber.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox