netdev - [PATCH net v2] inet: bring NLM_DONE out to a separate recv() in inet_dump

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>] [day] [month] [year] [list]
Date: Sun,  2 Jun 2024 14:55:52 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: davem@...emloft.net
Cc: netdev@...r.kernel.org,
	edumazet@...gle.com,
	pabeni@...hat.com,
	Jakub Kicinski <kuba@...nel.org>,
	Jaroslav Pulchart <jaroslav.pulchart@...ddata.com>,
	dsahern@...nel.org
Subject: [PATCH net v2] inet: bring NLM_DONE out to a separate recv() in inet_dump_ifaddr()

Jaroslav reports Dell's OMSA Systems Management Data Engine
expects NLM_DONE in a separate recvmsg(), so revert to the
old behavior.

This is the same kind of fix as we added in
commit 460b0d33cf10 ("inet: bring NLM_DONE out to a separate recv() again")
so wrap the logic into a helper, to make it easier to keep track
of which dump handles we know to require legacy handling
(and possibly one day let sockets opt into not doing this).

Tested:

  ./cli.py --dbg-small-recv 4096 --spec netlink/specs/rt_addr.yaml \
           --dump getaddr --json '{"ifa-family": 2}'

  ./cli.py --dbg-small-recv 4096 --spec netlink/specs/rt_route.yaml \
           --dump getroute --json '{"rtm-family": 2}'

  ./cli.py --dbg-small-recv 4096 --spec netlink/specs/rt_link.yaml \
           --dump getlink

Fixes: 3e41af90767d ("rtnetlink: use xarray iterator to implement rtnl_dump_ifinfo()")
Fixes: cdb2f80f1c10 ("inet: use xa_array iterator to implement inet_dump_ifaddr()")
Reported-by: Jaroslav Pulchart <jaroslav.pulchart@...ddata.com>
Link: https://lore.kernel.org/all/CAK8fFZ7MKoFSEzMBDAOjoUt+vTZRRQgLDNXEOfdCCXSoXXKE0g@mail.gmail.com
Signed-off-by: Jakub Kicinski <kuba@...nel.org>
---
v2:
 - adjust the comment in the kdoc
 - add getlink

CC: dsahern@...nel.org
---
 include/net/netlink.h   | 40 ++++++++++++++++++++++++++++++++++++++++
 net/core/rtnetlink.c    |  5 ++++-
 net/ipv4/devinet.c      |  3 +++
 net/ipv4/fib_frontend.c |  5 +----
 4 files changed, 48 insertions(+), 5 deletions(-)

diff --git a/include/net/netlink.h b/include/net/netlink.h
index e78ce008e07c..e41069edaea1 100644
--- a/include/net/netlink.h
+++ b/include/net/netlink.h
@@ -1198,6 +1198,46 @@ nl_dump_check_consistent(struct netlink_callback *cb,
 	cb->prev_seq = cb->seq;
 }
 
+/**
+ * nl_dump_legacy_retval - get legacy return code for netlink dumps
+ * @err: error encountered during dump, negative errno or zero
+ * @skb: skb with the dump results
+ *
+ * Netlink dump callbacks get called multiple times per dump, because
+ * all the objects may not fit into a single skb. Whether another iteration
+ * is necessary gets decided based on the return value of the callback
+ * (with 0 meaning "end reached").
+ *
+ * The semantics used to be more complicated, with positive return values
+ * meaning "continue" and negative meaning "end with an error". A lot of
+ * handlers simplified this to return skb->len ? : -errno. Meaning that zero
+ * would only be returned when skb was empty, requiring another recvmsg()
+ * syscall just to get the NLM_DONE message.
+ *
+ * The current semantics allow handlers to also return -EMSGSIZE to continue.
+ *
+ * Unfortunately, some user space has started to depend on the NLM_DONE
+ * message being returned individually, in a separate recvmsg(). Select
+ * netlink dumps must preserve those semantics.
+ *
+ * This helper wraps the "legacy logic" and serves as an annotation for
+ * dumps which are known to require legacy handling.
+ *
+ * When used in combination with for_each_netdev_dump() - make sure to
+ * invalidate the ifindex when iteration is done. for_each_netdev_dump()
+ * does not move the iterator index "after" the last valid entry.
+ *
+ * NOTE: Do not use this helper in new code or dumps without legacy users!
+ *
+ * Return: return code to use for a dump handler
+ */
+static inline int nl_dump_legacy_retval(int err, const struct sk_buff *skb)
+{
+	if (err < 0 && err != -EMSGSIZE)
+		return err;
+	return skb->len;
+}
+
 /**************************************************************************
  * Netlink Attributes
  **************************************************************************/
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index b86b0a87367d..522bbd70c205 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -2284,8 +2284,11 @@ static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 				       ext_filter_mask, 0, NULL, 0,
 				       netnsid, GFP_KERNEL);
 		if (err < 0)
-			break;
+			goto done;
 	}
+	ctx->ifindex = INT_MAX;
+done:
+	err = nl_dump_legacy_retval(err, skb);
 	cb->seq = tgt_net->dev_base_seq;
 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 	if (netnsid >= 0)
diff --git a/net/ipv4/devinet.c b/net/ipv4/devinet.c
index 96accde527da..4666e1ee2c14 100644
--- a/net/ipv4/devinet.c
+++ b/net/ipv4/devinet.c
@@ -1911,7 +1911,10 @@ static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
 		if (err < 0)
 			goto done;
 	}
+	ctx->ifindex = INT_MAX;
 done:
+	err = nl_dump_legacy_retval(err, skb);
+
 	if (fillargs.netnsid >= 0)
 		put_net(tgt_net);
 	rcu_read_unlock();
diff --git a/net/ipv4/fib_frontend.c b/net/ipv4/fib_frontend.c
index c484b1c0fc00..100d77eafe35 100644
--- a/net/ipv4/fib_frontend.c
+++ b/net/ipv4/fib_frontend.c
@@ -1051,10 +1051,7 @@ static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
 		}
 	}
 
-	/* Don't let NLM_DONE coalesce into a message, even if it could.
-	 * Some user space expects NLM_DONE in a separate recv().
-	 */
-	err = skb->len;
+	err = nl_dump_legacy_retval(err, skb);
 out:
 
 	cb->args[1] = e;
-- 
2.45.1