[PATCH net-next v3 0/9] netconsole: stop charging netpoll users for netconsole-only data

[PATCH net-next v3 0/9] netconsole: stop charging netpoll users for netconsole-only data

[Breno Leitao]

This work continue to untangle netconsole and netpoll, improving
memory usage for netpoll users that are not netconsole.

struct netpoll is shared by every netpoll consumer in the tree:
bonding, bridge, team, vlan, macvlan, dsa and netconsole. The non-
netconsole users only need a per-port handle that lets them transmit
SKBs.

Yet struct netpoll has accumulated a pile of fields that exist only
to serve netconsole's printk path:

  - skb_pool / refill_wq: the fallback skb pool find_skb() falls back
    on when alloc_skb() returns NULL, plus the workqueue that refills
    it.
  - local_port / remote_port / remote_mac: the source and destination
    UDP ports and the destination ethernet address used to build
    outgoing log packets.

Every netpoll user that allocates a struct netpoll pays for all of
these fields, even though it never queues a single skb on the pool,
never sends a UDP frame and never looks at any of the addressing.

Move this netconsole-only state out of the shared struct into struct
netconsole_target, where its only consumer already lives.

After the series every passive netpoll consumer drops
sizeof(sk_buff_head) + sizeof(work_struct) + 2 * sizeof(u16) +
ETH_ALEN bytes per port (plus padding), and struct netpoll keeps only
what belongs to the generic poll/transmit abstraction or what netpoll
core still reads itself.

Follow-ups
----------
local_ip, remote_ip and ipv6 are still in struct netpoll. Moving the
addressing additionally requires relocating netpoll_setup() -- whose
only caller is netconsole -- and its IPv4/IPv6 helpers, so it is left
for a follow-up to keep this series focused on a single responsibility
transfer.

This is pure code motion with no functional change. The pre-existing
target/device teardown races reported against v1 have since been fixed
independently (as suggested by Jakub) and are already in the tree, so,
re-posting this one.

--

Changes in v3:
- Added a fix for a pre-existing cleanup-skip leak (released/joined targets).
- Added an skb-pool init-once patch closing an init/teardown race; rebased.
- Link to v2: https://lore.kernel.org/r/20260702-netconsole_move_more-v2-0-1ebedd921dcb@debian.org

Changes in v2:
- Rebase on net-next, now that the prerequisite netconsole
  teardown-race fixes are merged.
- Squash the local_port and remote_port moves into a single patch.
- v1: https://lore.kernel.org/r/20260524-netconsole_move_more-v1-0-909d1ab398b4@debian.org

To: David S. Miller <davem@davemloft.net>
To: Eric Dumazet <edumazet@google.com>
To: Jakub Kicinski <kuba@kernel.org>
To: Paolo Abeni <pabeni@redhat.com>
To: Simon Horman <horms@kernel.org>
To: Andrew Lunn <andrew+netdev@lunn.ch>
Cc: netdev@vger.kernel.org
Cc: asantostc@gmail.com
Cc: gustavold@gmail.com
Cc: linux-kernel@vger.kernel.org

Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
---
Breno Leitao (9):
      netconsole: clean up released targets dropped before the cleanup worker
      netpoll: export refill_skbs(), refill_skbs_work_handler(), skb_pool_flush()
      netconsole: take over skb pool lifecycle from netpoll
      netconsole: move refill_skbs_work_handler() from netpoll
      netconsole: move refill_skbs() and skb-pool sizing macros from netpoll
      netconsole: move skb_pool_flush() from netpoll
      netconsole: move skb_pool / refill_wq from struct netpoll to netconsole_target
      netconsole: move local_port / remote_port from struct netpoll to netconsole_target
      netconsole: move remote_mac from struct netpoll to netconsole_target

 drivers/net/netconsole.c | 199 +++++++++++++++++++++++++++++++++++------------
 include/linux/netpoll.h  |  18 -----
 net/core/netpoll.c       |  51 +-----------
 3 files changed, 152 insertions(+), 116 deletions(-)
---
base-commit: 34cf6dafc47441dfb6b356a095b89c3585a93714
change-id: 20260514-netconsole_move_more-888fde03dfc7

Best regards,
--  
Breno Leitao <leitao@debian.org>

[PATCH net-next v3 1/9] netconsole: clean up released targets dropped before the cleanup worker

[Breno Leitao]

drop_netconsole_target() might eventually tear down a target that
netconsole_netdev_event() had moved to target_cleanup_list but that
netconsole_process_cleanups_core() had not processed yet.

Always cleanup devices that eventually have a device attached to the
target, independent of the state.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index c1812a98365b7..a939daa07cf9d 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -1481,15 +1481,15 @@ static void drop_netconsole_target(struct config_group *group,
 
 	mutex_lock(&target_cleanup_list_lock);
 	spin_lock_irqsave(&target_list_lock, flags);
-	/* A STATE_DEACTIVATED target may have been moved to
-	 * target_cleanup_list by netconsole_netdev_event() but not yet
-	 * processed by netconsole_process_cleanups_core(). Unlinking it below
-	 * hides it from the cleanup worker, so this path has to clean it up
-	 * itself. Record that the target still owns a netpoll before the
-	 * state is downgraded.
+	/* A target moved to target_cleanup_list by netconsole_netdev_event()
+	 * but not yet processed still owns a netpoll; unlinking it below hides
+	 * it from the cleanup worker, so this path must tear it down itself.
+	 * This covers NETDEV_UNREGISTER (STATE_DEACTIVATED) and
+	 * NETDEV_RELEASE / NETDEV_JOIN (STATE_DISABLED); key off nt->np.dev,
+	 * which stays set until the netpoll is cleaned up.
 	 */
 	needs_cleanup = nt->state == STATE_ENABLED ||
-			nt->state == STATE_DEACTIVATED;
+			nt->state == STATE_DEACTIVATED || nt->np.dev;
 	/* Disable deactivated target to prevent races between resume attempt
 	 * and target removal.
 	 */

-- 
2.53.0-Meta

[PATCH net-next v3 2/9] netpoll: export refill_skbs(), refill_skbs_work_handler(), skb_pool_flush()

[Breno Leitao]

These three helpers manage the per-netpoll fallback skb pool. They
are file-static today because all of their callers live in
net/core/netpoll.c. Subsequent patches relocate the pool's owner
from struct netpoll to the only consumer that actually uses it
(netconsole), and that work needs netconsole to drive the helpers
directly while the function bodies still live here.

Drop static, add prototypes in <linux/netpoll.h>, and
EXPORT_SYMBOL_GPL() each. No behaviour change.

The exports are transitional. Each helper is moved into
drivers/net/netconsole.c later in this series, and at that point
its EXPORT_SYMBOL_GPL() and prototype are dropped. By the end of
the series no symbol introduced here remains exported.

The goal of this patch is to make the subsequente patches easy to
review.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 include/linux/netpoll.h | 3 +++
 net/core/netpoll.c      | 9 ++++++---
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index 88f7daa8560e5..a7b96e1792207 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -90,6 +90,9 @@ void netpoll_cleanup(struct netpoll *np);
 void do_netpoll_cleanup(struct netpoll *np);
 netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb);
 void netpoll_zap_completion_queue(void);
+void refill_skbs(struct netpoll *np);
+void refill_skbs_work_handler(struct work_struct *work);
+void skb_pool_flush(struct netpoll *np);
 
 #ifdef CONFIG_NETPOLL
 static inline void *netpoll_poll_lock(struct napi_struct *napi)
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index aed415d3cd74f..e062d88d10a37 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -213,7 +213,7 @@ void netpoll_poll_enable(struct net_device *dev)
 		up(&ni->dev_lock);
 }
 
-static void refill_skbs(struct netpoll *np)
+void refill_skbs(struct netpoll *np)
 {
 	struct sk_buff_head *skb_pool;
 	struct sk_buff *skb;
@@ -228,6 +228,7 @@ static void refill_skbs(struct netpoll *np)
 		skb_queue_tail(skb_pool, skb);
 	}
 }
+EXPORT_SYMBOL_GPL(refill_skbs);
 
 void netpoll_zap_completion_queue(void)
 {
@@ -351,7 +352,7 @@ netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
 }
 EXPORT_SYMBOL(netpoll_send_skb);
 
-static void skb_pool_flush(struct netpoll *np)
+void skb_pool_flush(struct netpoll *np)
 {
 	struct sk_buff_head *skb_pool;
 
@@ -359,14 +360,16 @@ static void skb_pool_flush(struct netpoll *np)
 	skb_pool = &np->skb_pool;
 	skb_queue_purge_reason(skb_pool, SKB_CONSUMED);
 }
+EXPORT_SYMBOL_GPL(skb_pool_flush);
 
-static void refill_skbs_work_handler(struct work_struct *work)
+void refill_skbs_work_handler(struct work_struct *work)
 {
 	struct netpoll *np =
 		container_of(work, struct netpoll, refill_wq);
 
 	refill_skbs(np);
 }
+EXPORT_SYMBOL_GPL(refill_skbs_work_handler);
 
 int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
 {

-- 
2.53.0-Meta

[PATCH net-next v3 3/9] netconsole: take over skb pool lifecycle from netpoll

[Breno Leitao]

The fallback skb pool fronted by find_skb() is netconsole's only client:
every other netpoll goes through __netpoll_setup() / netpoll_send_skb()
without ever touching np->skb_pool.

Today __netpoll_setup() and __netpoll_cleanup() create and destroy the
pool for everyone, paying ~48 KB of pre-allocated skbs per netpoll
instance that only netconsole uses, what a waste!

Move the responsibility to netconsole. __netpoll_setup() did this under
the RTNL, but netconsole enables targets from enabled_store() /
alloc_param_target() without it, while the teardown path flushes the pool
(cancel_work_sync() + skb_queue_purge()) under the RTNL from
netconsole_process_cleanups_core(). Initialising the queue head and the
refill work on every enable would therefore race that flush.

They only need initialising once: after a flush the queue head is left
valid and empty and cancel_work_sync() leaves the work re-armable. Set
them up in alloc_and_init(), while the target is not yet reachable, and
let the enable paths only refill the pool via refill_skbs(), which
serialises with the flush through skb_pool.lock. See discussions in [1]

Link: https://lore.kernel.org/all/alDMvD5S7TZnoD_V@gmail.com/ [1]
Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++--
 net/core/netpoll.c       | 12 +----------
 2 files changed, 52 insertions(+), 13 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index a939daa07cf9d..1f75c4bbea8b6 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -292,11 +292,33 @@ static void netcons_release_dev(struct netconsole_target *nt)
 		memset(&nt->np.dev_name, 0, IFNAMSIZ);
 }
 
+/* Seed the per-target skb pool that find_skb() falls back to. The queue
+ * head and refill work are set up once in alloc_and_init(); this only
+ * (re)fills the pool. Pair with netconsole_skb_pool_flush().
+ */
+static void netconsole_skb_pool_init(struct netconsole_target *nt)
+{
+	refill_skbs(&nt->np);
+}
+
+static void netconsole_skb_pool_flush(struct netconsole_target *nt)
+{
+	skb_pool_flush(&nt->np);
+}
+
 /* Attempts to resume logging to a deactivated target. */
 static void resume_target(struct netconsole_target *nt)
 {
+	/* Initialise the skb pool before netpoll_setup() makes nt->np.dev
+	 * visible to target_list walkers (e.g. netconsole_netdev_event),
+	 * which otherwise may move the target to the cleanup list and
+	 * call netconsole_skb_pool_flush() on uninitialised state.
+	 */
+	netconsole_skb_pool_init(nt);
+
 	if (netpoll_setup(&nt->np)) {
 		/* netpoll fails setup once, do not try again. */
+		netconsole_skb_pool_flush(nt);
 		nt->state = STATE_DISABLED;
 		return;
 	}
@@ -358,6 +380,7 @@ static void process_resume_target(struct work_struct *work)
 	rtnl_lock();
 	if (nt->state == STATE_ENABLED && nt->np.dev &&
 	    nt->np.dev->reg_state != NETREG_REGISTERED) {
+		netconsole_skb_pool_flush(nt);
 		netcons_release_dev(nt);
 		nt->state = STATE_DISABLED;
 	}
@@ -398,6 +421,9 @@ static struct netconsole_target *alloc_and_init(void)
 	eth_broadcast_addr(nt->np.remote_mac);
 	nt->state = STATE_DISABLED;
 	INIT_WORK(&nt->resume_wq, process_resume_target);
+	/* Set up the skb pool primitives once; enabling only refills it. */
+	skb_queue_head_init(&nt->np.skb_pool);
+	INIT_WORK(&nt->np.refill_wq, refill_skbs_work_handler);
 
 	return nt;
 }
@@ -417,6 +443,7 @@ static void netconsole_process_cleanups_core(void)
 	list_for_each_entry_safe(nt, tmp, &target_cleanup_list, list) {
 		/* all entries in the cleanup_list needs to be disabled */
 		WARN_ON_ONCE(nt->state == STATE_ENABLED);
+		netconsole_skb_pool_flush(nt);
 		netcons_release_dev(nt);
 		/* moved the cleaned target to target_list. Need to hold both
 		 * locks
@@ -758,9 +785,19 @@ static ssize_t enabled_store(struct config_item *item,
 		 */
 		netconsole_print_banner(&nt->np);
 
+		/* Initialise the skb pool before netpoll_setup() so the pool
+		 * is valid as soon as nt->np.dev becomes visible to
+		 * target_list walkers (netconsole_netdev_event), which would
+		 * otherwise call netconsole_skb_pool_flush() on uninitialised
+		 * state.
+		 */
+		netconsole_skb_pool_init(nt);
+
 		ret = netpoll_setup(&nt->np);
-		if (ret)
+		if (ret) {
+			netconsole_skb_pool_flush(nt);
 			goto out_unlock;
+		}
 
 		nt->state = STATE_ENABLED;
 		pr_info("network logging started\n");
@@ -1514,8 +1551,10 @@ static void drop_netconsole_target(struct config_group *group,
 	 * netpoll_cleanup() is idempotent (it skips when np->dev is NULL), so
 	 * it is safe even if the cleanup worker already tore the netpoll down.
 	 */
-	if (needs_cleanup)
+	if (needs_cleanup) {
+		netconsole_skb_pool_flush(nt);
 		netpoll_cleanup(&nt->np);
+	}
 
 	config_item_put(&nt->group.cg_item);
 }
@@ -2330,10 +2369,18 @@ static struct netconsole_target *alloc_param_target(char *target_config,
 	if (err)
 		goto fail;
 
+	/* Initialise the skb pool before netpoll_setup() so the pool is
+	 * valid as soon as nt->np.dev becomes visible. The target is not
+	 * yet on target_list, so a netdev event cannot reach it here, but
+	 * mirror the configfs path for symmetry.
+	 */
+	netconsole_skb_pool_init(nt);
+
 	err = netpoll_setup(&nt->np);
 	if (err) {
 		pr_err("Not enabling netconsole for %s%d. Netpoll setup failed\n",
 		       NETCONSOLE_PARAM_TARGET_PREFIX, cmdline_count);
+		netconsole_skb_pool_flush(nt);
 		if (!IS_ENABLED(CONFIG_NETCONSOLE_DYNAMIC))
 			/* only fail if dynamic reconfiguration is set,
 			 * otherwise, keep the target in the list, but disabled.
@@ -2355,6 +2402,8 @@ static struct netconsole_target *alloc_param_target(char *target_config,
 static void free_param_target(struct netconsole_target *nt)
 {
 	cancel_work_sync(&nt->resume_wq);
+	if (nt->state == STATE_ENABLED)
+		netconsole_skb_pool_flush(nt);
 	netpoll_cleanup(&nt->np);
 #ifdef	CONFIG_NETCONSOLE_DYNAMIC
 	kfree(nt->userdata);
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index e062d88d10a37..58f30a4d5eb0f 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -377,9 +377,6 @@ int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
 	const struct net_device_ops *ops;
 	int err;
 
-	skb_queue_head_init(&np->skb_pool);
-	INIT_WORK(&np->refill_wq, refill_skbs_work_handler);
-
 	if (ndev->priv_flags & IFF_DISABLE_NETPOLL) {
 		np_err(np, "%s doesn't support polling, aborting\n",
 		       ndev->name);
@@ -414,9 +411,6 @@ int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
 	np->dev = ndev;
 	strscpy(np->dev_name, ndev->name, IFNAMSIZ);
 
-	/* fill up the skb queue */
-	refill_skbs(np);
-
 	/* last thing to do is link it to the net device structure */
 	rcu_assign_pointer(ndev->npinfo, npinfo);
 
@@ -606,7 +600,7 @@ int netpoll_setup(struct netpoll *np)
 
 	err = __netpoll_setup(np, ndev);
 	if (err)
-		goto flush;
+		goto put;
 	rtnl_unlock();
 
 	/* Make sure all NAPI polls which started before dev->npinfo
@@ -617,8 +611,6 @@ int netpoll_setup(struct netpoll *np)
 
 	return 0;
 
-flush:
-	skb_pool_flush(np);
 put:
 	DEBUG_NET_WARN_ON_ONCE(np->dev);
 	if (ip_overwritten)
@@ -662,8 +654,6 @@ static void __netpoll_cleanup(struct netpoll *np)
 		disable_delayed_work_sync(&npinfo->tx_work);
 		call_rcu(&npinfo->rcu, rcu_cleanup_netpoll_info);
 	}
-
-	skb_pool_flush(np);
 }
 
 void __netpoll_free(struct netpoll *np)

-- 
2.53.0-Meta

[PATCH net-next v3 4/9] netconsole: move refill_skbs_work_handler() from netpoll

[Breno Leitao]

The work handler is wired via INIT_WORK() in
netconsole_skb_pool_init() and has no other callers since the
previous patch took the skb pool lifecycle out of __netpoll_setup().
Move the function body into drivers/net/netconsole.c as a
file-static helper, drop EXPORT_SYMBOL_GPL() and remove the
prototype from <linux/netpoll.h>.

Pure code motion: the body is unchanged and still calls the
exported refill_skbs() in net/core/netpoll.c.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 8 ++++++++
 include/linux/netpoll.h  | 1 -
 net/core/netpoll.c       | 9 ---------
 3 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index 1f75c4bbea8b6..96d9a47312cd7 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -292,6 +292,14 @@ static void netcons_release_dev(struct netconsole_target *nt)
 		memset(&nt->np.dev_name, 0, IFNAMSIZ);
 }
 
+static void refill_skbs_work_handler(struct work_struct *work)
+{
+	struct netpoll *np =
+		container_of(work, struct netpoll, refill_wq);
+
+	refill_skbs(np);
+}
+
 /* Seed the per-target skb pool that find_skb() falls back to. The queue
  * head and refill work are set up once in alloc_and_init(); this only
  * (re)fills the pool. Pair with netconsole_skb_pool_flush().
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index a7b96e1792207..51e5863d8e678 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -91,7 +91,6 @@ void do_netpoll_cleanup(struct netpoll *np);
 netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb);
 void netpoll_zap_completion_queue(void);
 void refill_skbs(struct netpoll *np);
-void refill_skbs_work_handler(struct work_struct *work);
 void skb_pool_flush(struct netpoll *np);
 
 #ifdef CONFIG_NETPOLL
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index 58f30a4d5eb0f..93a16faf808c8 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -362,15 +362,6 @@ void skb_pool_flush(struct netpoll *np)
 }
 EXPORT_SYMBOL_GPL(skb_pool_flush);
 
-void refill_skbs_work_handler(struct work_struct *work)
-{
-	struct netpoll *np =
-		container_of(work, struct netpoll, refill_wq);
-
-	refill_skbs(np);
-}
-EXPORT_SYMBOL_GPL(refill_skbs_work_handler);
-
 int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
 {
 	struct netpoll_info *npinfo;

-- 
2.53.0-Meta

[PATCH net-next v3 5/9] netconsole: move refill_skbs() and skb-pool sizing macros from netpoll

[Breno Leitao]

refill_skbs() is now only called from netconsole (directly via
netconsole_skb_pool_init() and indirectly via the just-moved
refill_skbs_work_handler()), and the MAX_UDP_CHUNK / MAX_SKBS /
MAX_SKB_SIZE macros are private to it. Move them all into
drivers/net/netconsole.c.

MAX_UDP_CHUNK and MAX_SKB_SIZE were promoted to <linux/netpoll.h>
by commit 6c537b845c99 ("netconsole: do not dequeue pooled skbs that
cannot satisfy len") so find_skb() could detect oversized requests
against the same value refill_skbs() used. With both functions now
local to netconsole, the shared definition no longer needs to live
in the header.

Pure code motion: bodies and pool sizing semantics are unchanged.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 29 +++++++++++++++++++++++++++++
 include/linux/netpoll.h  | 15 ---------------
 net/core/netpoll.c       | 23 -----------------------
 3 files changed, 29 insertions(+), 38 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index 96d9a47312cd7..efeada762536f 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -61,6 +61,19 @@ MODULE_IMPORT_NS("NETDEV_INTERNAL");
 #define MAX_USERDATA_ITEMS		256
 #define MAX_PRINT_CHUNK			1000
 
+/*
+ * Sizing for the per-target fallback skb pool consulted by find_skb()
+ * when its GFP_ATOMIC allocation fails so messages still get out under
+ * memory pressure.
+ */
+#define MAX_UDP_CHUNK			1460
+#define MAX_SKBS			32
+#define MAX_SKB_SIZE							\
+	(sizeof(struct ethhdr) +					\
+	 sizeof(struct iphdr) +						\
+	 sizeof(struct udphdr) +					\
+	 MAX_UDP_CHUNK)
+
 static char config[MAX_PARAM_LENGTH];
 module_param_string(netconsole, config, MAX_PARAM_LENGTH, 0);
 MODULE_PARM_DESC(netconsole, " netconsole=[src-port]@[src-ip]/[dev],[tgt-port]@<tgt-ip>/[tgt-macaddr]");
@@ -292,6 +305,22 @@ static void netcons_release_dev(struct netconsole_target *nt)
 		memset(&nt->np.dev_name, 0, IFNAMSIZ);
 }
 
+static void refill_skbs(struct netpoll *np)
+{
+	struct sk_buff_head *skb_pool;
+	struct sk_buff *skb;
+
+	skb_pool = &np->skb_pool;
+
+	while (READ_ONCE(skb_pool->qlen) < MAX_SKBS) {
+		skb = alloc_skb(MAX_SKB_SIZE, GFP_ATOMIC | __GFP_NOWARN);
+		if (!skb)
+			break;
+
+		skb_queue_tail(skb_pool, skb);
+	}
+}
+
 static void refill_skbs_work_handler(struct work_struct *work)
 {
 	struct netpoll *np =
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index 51e5863d8e678..7e2fbce863e9b 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -21,20 +21,6 @@ union inet_addr {
 	struct in6_addr	in6;
 };
 
-/*
- * Maximum payload netpoll's preallocated skb pool can carry. Keep this in
- * sync with the buffer size used by refill_skbs() in net/core/netpoll.c;
- * callers (e.g. netconsole) use it to detect requests the pool can never
- * satisfy and avoid dequeuing a pooled skb that would later trip
- * skb_over_panic() in skb_put().
- */
-#define MAX_UDP_CHUNK	1460
-#define MAX_SKB_SIZE						\
-	(sizeof(struct ethhdr) +				\
-	 sizeof(struct iphdr) +					\
-	 sizeof(struct udphdr) +				\
-	 MAX_UDP_CHUNK)
-
 struct netpoll {
 	struct net_device *dev;
 	netdevice_tracker dev_tracker;
@@ -90,7 +76,6 @@ void netpoll_cleanup(struct netpoll *np);
 void do_netpoll_cleanup(struct netpoll *np);
 netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb);
 void netpoll_zap_completion_queue(void);
-void refill_skbs(struct netpoll *np);
 void skb_pool_flush(struct netpoll *np);
 
 #ifdef CONFIG_NETPOLL
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index 93a16faf808c8..9ca695f64210c 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -36,12 +36,6 @@
 #include <trace/events/napi.h>
 #include <linux/kconfig.h>
 
-/*
- * We maintain a small pool of fully-sized skbs, to make sure the
- * message gets out even in extreme OOM situations.
- */
-
-#define MAX_SKBS 32
 #define USEC_PER_POLL	50
 
 static unsigned int carrier_timeout = 4;
@@ -213,23 +207,6 @@ void netpoll_poll_enable(struct net_device *dev)
 		up(&ni->dev_lock);
 }
 
-void refill_skbs(struct netpoll *np)
-{
-	struct sk_buff_head *skb_pool;
-	struct sk_buff *skb;
-
-	skb_pool = &np->skb_pool;
-
-	while (READ_ONCE(skb_pool->qlen) < MAX_SKBS) {
-		skb = alloc_skb(MAX_SKB_SIZE, GFP_ATOMIC | __GFP_NOWARN);
-		if (!skb)
-			break;
-
-		skb_queue_tail(skb_pool, skb);
-	}
-}
-EXPORT_SYMBOL_GPL(refill_skbs);
-
 void netpoll_zap_completion_queue(void)
 {
 	unsigned long flags;

-- 
2.53.0-Meta

[PATCH net-next v3 6/9] netconsole: move skb_pool_flush() from netpoll

[Breno Leitao]

skb_pool_flush() has no callers left in net/core/netpoll.c after
netconsole took over the pool lifecycle. Inline its body into
netconsole_skb_pool_flush() (the only caller) and drop the function
and its export from netpoll. The prototype goes from
<linux/netpoll.h>.

Pure code motion: cancel_work_sync() + skb_queue_purge_reason()
semantics are unchanged.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c |  5 ++++-
 include/linux/netpoll.h  |  1 -
 net/core/netpoll.c       | 10 ----------
 3 files changed, 4 insertions(+), 12 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index efeada762536f..742723c5a7d46 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -340,7 +340,10 @@ static void netconsole_skb_pool_init(struct netconsole_target *nt)
 
 static void netconsole_skb_pool_flush(struct netconsole_target *nt)
 {
-	skb_pool_flush(&nt->np);
+	struct netpoll *np = &nt->np;
+
+	cancel_work_sync(&np->refill_wq);
+	skb_queue_purge_reason(&np->skb_pool, SKB_CONSUMED);
 }
 
 /* Attempts to resume logging to a deactivated target. */
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index 7e2fbce863e9b..1216b5c237ce4 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -76,7 +76,6 @@ void netpoll_cleanup(struct netpoll *np);
 void do_netpoll_cleanup(struct netpoll *np);
 netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb);
 void netpoll_zap_completion_queue(void);
-void skb_pool_flush(struct netpoll *np);
 
 #ifdef CONFIG_NETPOLL
 static inline void *netpoll_poll_lock(struct napi_struct *napi)
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index 9ca695f64210c..f8da1048ea3ab 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -329,16 +329,6 @@ netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
 }
 EXPORT_SYMBOL(netpoll_send_skb);
 
-void skb_pool_flush(struct netpoll *np)
-{
-	struct sk_buff_head *skb_pool;
-
-	cancel_work_sync(&np->refill_wq);
-	skb_pool = &np->skb_pool;
-	skb_queue_purge_reason(skb_pool, SKB_CONSUMED);
-}
-EXPORT_SYMBOL_GPL(skb_pool_flush);
-
 int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
 {
 	struct netpoll_info *npinfo;

-- 
2.53.0-Meta

[PATCH net-next v3 7/9] netconsole: move skb_pool / refill_wq from struct netpoll to netconsole_target

[Breno Leitao]

These two fields back the fallback skb pool that find_skb() uses.
Every helper that touches them lives in netconsole now (refill_skbs,
refill_skbs_work_handler, netconsole_skb_pool_init,
netconsole_skb_pool_flush, find_skb, netcons_skb_pop), so the data
can move alongside its only consumer.

Add skb_pool and refill_wq to struct netconsole_target, drop them
from struct netpoll.

This will save 48-bytes for every netpoll user instance (except
netconsole that will have it in netconsole target struct).

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 53 +++++++++++++++++++++++++++---------------------
 include/linux/netpoll.h  |  2 --
 2 files changed, 30 insertions(+), 25 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index 742723c5a7d46..cf591ae667362 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -183,6 +183,11 @@ enum target_state {
  *		remote_mac	(read-write)
  * @buf:	The buffer used to send the full msg to the network stack
  * @resume_wq:	Workqueue to resume deactivated target
+ * @skb_pool:	Per-target fallback skb pool consulted by find_skb() when
+ *		its GFP_ATOMIC allocation fails. Lifetime brackets a
+ *		successful netpoll_setup() / netpoll_cleanup() pair on @np.
+ * @refill_wq:	Work item that asynchronously tops @skb_pool back up to
+ *		MAX_SKBS after find_skb() drains an entry.
  */
 struct netconsole_target {
 	struct list_head	list;
@@ -208,6 +213,8 @@ struct netconsole_target {
 	 */
 	char			buf[MAX_PRINT_CHUNK + 1];
 	struct work_struct	resume_wq;
+	struct sk_buff_head	skb_pool;
+	struct work_struct	refill_wq;
 };
 
 #ifdef	CONFIG_NETCONSOLE_DYNAMIC
@@ -305,13 +312,11 @@ static void netcons_release_dev(struct netconsole_target *nt)
 		memset(&nt->np.dev_name, 0, IFNAMSIZ);
 }
 
-static void refill_skbs(struct netpoll *np)
+static void refill_skbs(struct netconsole_target *nt)
 {
-	struct sk_buff_head *skb_pool;
+	struct sk_buff_head *skb_pool = &nt->skb_pool;
 	struct sk_buff *skb;
 
-	skb_pool = &np->skb_pool;
-
 	while (READ_ONCE(skb_pool->qlen) < MAX_SKBS) {
 		skb = alloc_skb(MAX_SKB_SIZE, GFP_ATOMIC | __GFP_NOWARN);
 		if (!skb)
@@ -323,10 +328,10 @@ static void refill_skbs(struct netpoll *np)
 
 static void refill_skbs_work_handler(struct work_struct *work)
 {
-	struct netpoll *np =
-		container_of(work, struct netpoll, refill_wq);
+	struct netconsole_target *nt =
+		container_of(work, struct netconsole_target, refill_wq);
 
-	refill_skbs(np);
+	refill_skbs(nt);
 }
 
 /* Seed the per-target skb pool that find_skb() falls back to. The queue
@@ -335,15 +340,13 @@ static void refill_skbs_work_handler(struct work_struct *work)
  */
 static void netconsole_skb_pool_init(struct netconsole_target *nt)
 {
-	refill_skbs(&nt->np);
+	refill_skbs(nt);
 }
 
 static void netconsole_skb_pool_flush(struct netconsole_target *nt)
 {
-	struct netpoll *np = &nt->np;
-
-	cancel_work_sync(&np->refill_wq);
-	skb_queue_purge_reason(&np->skb_pool, SKB_CONSUMED);
+	cancel_work_sync(&nt->refill_wq);
+	skb_queue_purge_reason(&nt->skb_pool, SKB_CONSUMED);
 }
 
 /* Attempts to resume logging to a deactivated target. */
@@ -462,8 +465,8 @@ static struct netconsole_target *alloc_and_init(void)
 	nt->state = STATE_DISABLED;
 	INIT_WORK(&nt->resume_wq, process_resume_target);
 	/* Set up the skb pool primitives once; enabling only refills it. */
-	skb_queue_head_init(&nt->np.skb_pool);
-	INIT_WORK(&nt->np.refill_wq, refill_skbs_work_handler);
+	skb_queue_head_init(&nt->skb_pool);
+	INIT_WORK(&nt->refill_wq, refill_skbs_work_handler);
 
 	return nt;
 }
@@ -1785,7 +1788,7 @@ static struct notifier_block netconsole_netdev_notifier = {
  * pool locks and is therefore not NMI-safe. Skip the refill when called
  * from NMI context; the next non-NMI caller will top the pool back up.
  */
-static struct sk_buff *netcons_skb_pop(struct netpoll *np, int len)
+static struct sk_buff *netcons_skb_pop(struct netconsole_target *nt, int len)
 {
 	struct sk_buff *skb;
 
@@ -1797,19 +1800,21 @@ static struct sk_buff *netcons_skb_pop(struct netpoll *np, int len)
 		if (!in_nmi())
 			net_warn_ratelimited("netconsole: dropping message, requested skb len %d exceeds pool buffer size %zu on %s\n",
 					     len, (size_t)MAX_SKB_SIZE,
-					     np->dev->name);
+					     nt->np.dev->name);
 		return NULL;
 	}
 
-	skb = skb_dequeue(&np->skb_pool);
+	skb = skb_dequeue(&nt->skb_pool);
 	if (!in_nmi())
-		schedule_work(&np->refill_wq);
+		schedule_work(&nt->refill_wq);
 
 	return skb;
 }
 
-static struct sk_buff *find_skb(struct netpoll *np, int len, int reserve)
+static struct sk_buff *find_skb(struct netconsole_target *nt, int len,
+				int reserve)
 {
+	struct netpoll *np = &nt->np;
 	int count = 0;
 	struct sk_buff *skb;
 
@@ -1818,7 +1823,7 @@ static struct sk_buff *find_skb(struct netpoll *np, int len, int reserve)
 
 	skb = alloc_skb(len, GFP_ATOMIC | __GFP_NOWARN);
 	if (!skb)
-		skb = netcons_skb_pop(np, len);
+		skb = netcons_skb_pop(nt, len);
 
 	if (!skb) {
 		if (++count < 10) {
@@ -1940,8 +1945,10 @@ static void push_ipv6(struct netpoll *np, struct sk_buff *skb, int len)
 	skb->protocol = htons(ETH_P_IPV6);
 }
 
-static int netpoll_send_udp(struct netpoll *np, const char *msg, int len)
+static int netpoll_send_udp(struct netconsole_target *nt, const char *msg,
+			    int len)
 {
+	struct netpoll *np = &nt->np;
 	int total_len, ip_len, udp_len;
 	struct sk_buff *skb;
 
@@ -1956,7 +1963,7 @@ static int netpoll_send_udp(struct netpoll *np, const char *msg, int len)
 
 	total_len = ip_len + LL_RESERVED_SPACE(np->dev);
 
-	skb = find_skb(np, total_len + np->dev->needed_tailroom,
+	skb = find_skb(nt, total_len + np->dev->needed_tailroom,
 		       total_len - len);
 	if (!skb)
 		return -ENOMEM;
@@ -1987,7 +1994,7 @@ static int netpoll_send_udp(struct netpoll *np, const char *msg, int len)
  */
 static void send_udp(struct netconsole_target *nt, const char *msg, int len)
 {
-	int result = netpoll_send_udp(&nt->np, msg, len);
+	int result = netpoll_send_udp(nt, msg, len);
 
 	if (IS_ENABLED(CONFIG_NETCONSOLE_DYNAMIC)) {
 		if (result == NET_XMIT_DROP) {
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index 1216b5c237ce4..f377fdf7839ca 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -37,8 +37,6 @@ struct netpoll {
 	bool ipv6;
 	u16 local_port, remote_port;
 	u8 remote_mac[ETH_ALEN];
-	struct sk_buff_head skb_pool;
-	struct work_struct refill_wq;
 };
 
 #define np_info(np, fmt, ...)				\

-- 
2.53.0-Meta

[PATCH net-next v3 8/9] netconsole: move local_port / remote_port from struct netpoll to netconsole_target

[Breno Leitao]

The source and destination UDP ports live in struct netpoll but are
netconsole configuration. No other netpoll user (bonding, team, vlan,
bridge, macvlan, dsa) touches np->local_port or np->remote_port; they
only use the netpoll TX/forwarding path. Only netconsole's UDP framing
and its configfs/cmdline interface read these fields.

Move both into struct netconsole_target and convert the three helpers
that read them - push_udp(), netconsole_print_banner() and
netconsole_parser_cmdline() - to take the netconsole_target. The
configfs show/store handlers already have the target in hand.

No functional change; the local_port / remote_port sysfs attributes are
unchanged.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 47 ++++++++++++++++++++++++++---------------------
 include/linux/netpoll.h  |  1 -
 2 files changed, 26 insertions(+), 22 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index cf591ae667362..d0739b45f66e1 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -175,12 +175,12 @@ enum target_state {
  * @np:		The netpoll structure for this target.
  *		Contains the other userspace visible parameters:
  *		dev_name	(read-write)
- *		local_port	(read-write)
- *		remote_port	(read-write)
  *		local_ip	(read-write)
  *		remote_ip	(read-write)
  *		local_mac	(read-only)
  *		remote_mac	(read-write)
+ * @local_port:	Source UDP port of the target (read-write).
+ * @remote_port: Destination UDP port of the target (read-write).
  * @buf:	The buffer used to send the full msg to the network stack
  * @resume_wq:	Workqueue to resume deactivated target
  * @skb_pool:	Per-target fallback skb pool consulted by find_skb() when
@@ -208,6 +208,7 @@ struct netconsole_target {
 	bool			extended;
 	bool			release;
 	struct netpoll		np;
+	u16			local_port, remote_port;
 	/* protected by target_list_lock; +1 gives scnprintf() room for its
 	 * NUL terminator so a full MAX_PRINT_CHUNK payload is not truncated
 	 */
@@ -459,8 +460,8 @@ static struct netconsole_target *alloc_and_init(void)
 
 	nt->np.name = "netconsole";
 	strscpy(nt->np.dev_name, "eth0", IFNAMSIZ);
-	nt->np.local_port = 6665;
-	nt->np.remote_port = 6666;
+	nt->local_port = 6665;
+	nt->remote_port = 6666;
 	eth_broadcast_addr(nt->np.remote_mac);
 	nt->state = STATE_DISABLED;
 	INIT_WORK(&nt->resume_wq, process_resume_target);
@@ -499,16 +500,18 @@ static void netconsole_process_cleanups_core(void)
 	mutex_unlock(&target_cleanup_list_lock);
 }
 
-static void netconsole_print_banner(struct netpoll *np)
+static void netconsole_print_banner(struct netconsole_target *nt)
 {
-	np_info(np, "local port %d\n", np->local_port);
+	struct netpoll *np = &nt->np;
+
+	np_info(np, "local port %d\n", nt->local_port);
 	if (np->ipv6)
 		np_info(np, "local IPv6 address %pI6c\n", &np->local_ip.in6);
 	else
 		np_info(np, "local IPv4 address %pI4\n", &np->local_ip.ip);
 	np_info(np, "interface name '%s'\n", np->dev_name);
 	np_info(np, "local ethernet address '%pM'\n", np->dev_mac);
-	np_info(np, "remote port %d\n", np->remote_port);
+	np_info(np, "remote port %d\n", nt->remote_port);
 	if (np->ipv6)
 		np_info(np, "remote IPv6 address %pI6c\n", &np->remote_ip.in6);
 	else
@@ -631,12 +634,12 @@ static ssize_t dev_name_show(struct config_item *item, char *buf)
 
 static ssize_t local_port_show(struct config_item *item, char *buf)
 {
-	return sysfs_emit(buf, "%d\n", to_target(item)->np.local_port);
+	return sysfs_emit(buf, "%d\n", to_target(item)->local_port);
 }
 
 static ssize_t remote_port_show(struct config_item *item, char *buf)
 {
-	return sysfs_emit(buf, "%d\n", to_target(item)->np.remote_port);
+	return sysfs_emit(buf, "%d\n", to_target(item)->remote_port);
 }
 
 static ssize_t local_ip_show(struct config_item *item, char *buf)
@@ -826,7 +829,7 @@ static ssize_t enabled_store(struct config_item *item,
 		 * Skip netconsole_parser_cmdline() -- all the attributes are
 		 * already configured via configfs. Just print them out.
 		 */
-		netconsole_print_banner(&nt->np);
+		netconsole_print_banner(nt);
 
 		/* Initialise the skb pool before netpoll_setup() so the pool
 		 * is valid as soon as nt->np.dev becomes visible to
@@ -965,7 +968,7 @@ static ssize_t local_port_store(struct config_item *item, const char *buf,
 		goto out_unlock;
 	}
 
-	ret = kstrtou16(buf, 10, &nt->np.local_port);
+	ret = kstrtou16(buf, 10, &nt->local_port);
 	if (ret < 0)
 		goto out_unlock;
 	ret = count;
@@ -987,7 +990,7 @@ static ssize_t remote_port_store(struct config_item *item,
 		goto out_unlock;
 	}
 
-	ret = kstrtou16(buf, 10, &nt->np.remote_port);
+	ret = kstrtou16(buf, 10, &nt->remote_port);
 	if (ret < 0)
 		goto out_unlock;
 	ret = count;
@@ -1863,8 +1866,9 @@ static void netpoll_udp_checksum(struct netpoll *np, struct sk_buff *skb,
 		udph->check = CSUM_MANGLED_0;
 }
 
-static void push_udp(struct netpoll *np, struct sk_buff *skb, int len)
+static void push_udp(struct netconsole_target *nt, struct sk_buff *skb, int len)
 {
+	struct netpoll *np = &nt->np;
 	struct udphdr *udph;
 	int udp_len;
 
@@ -1874,8 +1878,8 @@ static void push_udp(struct netpoll *np, struct sk_buff *skb, int len)
 	skb_reset_transport_header(skb);
 
 	udph = udp_hdr(skb);
-	udph->source = htons(np->local_port);
-	udph->dest = htons(np->remote_port);
+	udph->source = htons(nt->local_port);
+	udph->dest = htons(nt->remote_port);
 	udph->len = htons(udp_len);
 
 	netpoll_udp_checksum(np, skb, len);
@@ -1971,7 +1975,7 @@ static int netpoll_send_udp(struct netconsole_target *nt, const char *msg,
 	skb_copy_to_linear_data(skb, msg, len);
 	skb_put(skb, len);
 
-	push_udp(np, skb, len);
+	push_udp(nt, skb, len);
 	if (np->ipv6)
 		push_ipv6(np, skb, len);
 	else
@@ -2289,8 +2293,9 @@ __releases(&target_list_lock)
 	spin_unlock_irqrestore(&target_list_lock, flags);
 }
 
-static int netconsole_parser_cmdline(struct netpoll *np, char *opt)
+static int netconsole_parser_cmdline(struct netconsole_target *nt, char *opt)
 {
+	struct netpoll *np = &nt->np;
 	bool ipversion_set = false;
 	char *cur = opt;
 	char *delim;
@@ -2301,7 +2306,7 @@ static int netconsole_parser_cmdline(struct netpoll *np, char *opt)
 		if (!delim)
 			goto parse_failed;
 		*delim = 0;
-		if (kstrtou16(cur, 10, &np->local_port))
+		if (kstrtou16(cur, 10, &nt->local_port))
 			goto parse_failed;
 		cur = delim;
 	}
@@ -2348,7 +2353,7 @@ static int netconsole_parser_cmdline(struct netpoll *np, char *opt)
 		*delim = 0;
 		if (*cur == ' ' || *cur == '\t')
 			np_info(np, "warning: whitespace is not allowed\n");
-		if (kstrtou16(cur, 10, &np->remote_port))
+		if (kstrtou16(cur, 10, &nt->remote_port))
 			goto parse_failed;
 		cur = delim;
 	}
@@ -2374,7 +2379,7 @@ static int netconsole_parser_cmdline(struct netpoll *np, char *opt)
 			goto parse_failed;
 	}
 
-	netconsole_print_banner(np);
+	netconsole_print_banner(nt);
 
 	return 0;
 
@@ -2412,7 +2417,7 @@ static struct netconsole_target *alloc_param_target(char *target_config,
 	}
 
 	/* Parse parameters and setup netpoll */
-	err = netconsole_parser_cmdline(&nt->np, target_config);
+	err = netconsole_parser_cmdline(nt, target_config);
 	if (err)
 		goto fail;
 
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index f377fdf7839ca..5ca79fa7d9431 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -35,7 +35,6 @@ struct netpoll {
 
 	union inet_addr local_ip, remote_ip;
 	bool ipv6;
-	u16 local_port, remote_port;
 	u8 remote_mac[ETH_ALEN];
 };
 

-- 
2.53.0-Meta

[PATCH net-next v3 9/9] netconsole: move remote_mac from struct netpoll to netconsole_target

[Breno Leitao]

The destination ethernet address is netconsole configuration: no other
netpoll user (bonding, team, vlan, bridge, macvlan, dsa) references
np->remote_mac, only netconsole's ethernet framing and its
configfs/cmdline interface do.

Move it into struct netconsole_target and convert push_eth() to take the
netconsole_target; netconsole_print_banner() and
netconsole_parser_cmdline() already take it. The configfs show/store
handlers and alloc_and_init() reach the field directly.

No functional change; the remote_mac sysfs attribute is unchanged.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/netconsole.c | 20 +++++++++++---------
 include/linux/netpoll.h  |  1 -
 2 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
index d0739b45f66e1..49b2243a20b2d 100644
--- a/drivers/net/netconsole.c
+++ b/drivers/net/netconsole.c
@@ -178,9 +178,9 @@ enum target_state {
  *		local_ip	(read-write)
  *		remote_ip	(read-write)
  *		local_mac	(read-only)
- *		remote_mac	(read-write)
  * @local_port:	Source UDP port of the target (read-write).
  * @remote_port: Destination UDP port of the target (read-write).
+ * @remote_mac:	Destination ethernet address of the target (read-write).
  * @buf:	The buffer used to send the full msg to the network stack
  * @resume_wq:	Workqueue to resume deactivated target
  * @skb_pool:	Per-target fallback skb pool consulted by find_skb() when
@@ -209,6 +209,7 @@ struct netconsole_target {
 	bool			release;
 	struct netpoll		np;
 	u16			local_port, remote_port;
+	u8			remote_mac[ETH_ALEN];
 	/* protected by target_list_lock; +1 gives scnprintf() room for its
 	 * NUL terminator so a full MAX_PRINT_CHUNK payload is not truncated
 	 */
@@ -462,7 +463,7 @@ static struct netconsole_target *alloc_and_init(void)
 	strscpy(nt->np.dev_name, "eth0", IFNAMSIZ);
 	nt->local_port = 6665;
 	nt->remote_port = 6666;
-	eth_broadcast_addr(nt->np.remote_mac);
+	eth_broadcast_addr(nt->remote_mac);
 	nt->state = STATE_DISABLED;
 	INIT_WORK(&nt->resume_wq, process_resume_target);
 	/* Set up the skb pool primitives once; enabling only refills it. */
@@ -516,7 +517,7 @@ static void netconsole_print_banner(struct netconsole_target *nt)
 		np_info(np, "remote IPv6 address %pI6c\n", &np->remote_ip.in6);
 	else
 		np_info(np, "remote IPv4 address %pI4\n", &np->remote_ip.ip);
-	np_info(np, "remote ethernet address %pM\n", np->remote_mac);
+	np_info(np, "remote ethernet address %pM\n", nt->remote_mac);
 }
 
 /* Parse the string and populate the `inet_addr` union. Return 0 if IPv4 is
@@ -672,7 +673,7 @@ static ssize_t local_mac_show(struct config_item *item, char *buf)
 
 static ssize_t remote_mac_show(struct config_item *item, char *buf)
 {
-	return sysfs_emit(buf, "%pM\n", to_target(item)->np.remote_mac);
+	return sysfs_emit(buf, "%pM\n", to_target(item)->remote_mac);
 }
 
 static ssize_t transmit_errors_show(struct config_item *item, char *buf)
@@ -1077,7 +1078,7 @@ static ssize_t remote_mac_store(struct config_item *item, const char *buf,
 		goto out_unlock;
 	if (buf[MAC_ADDR_STR_LEN] && buf[MAC_ADDR_STR_LEN] != '\n')
 		goto out_unlock;
-	memcpy(nt->np.remote_mac, remote_mac, ETH_ALEN);
+	memcpy(nt->remote_mac, remote_mac, ETH_ALEN);
 
 	ret = count;
 out_unlock:
@@ -1885,14 +1886,15 @@ static void push_udp(struct netconsole_target *nt, struct sk_buff *skb, int len)
 	netpoll_udp_checksum(np, skb, len);
 }
 
-static void push_eth(struct netpoll *np, struct sk_buff *skb)
+static void push_eth(struct netconsole_target *nt, struct sk_buff *skb)
 {
+	struct netpoll *np = &nt->np;
 	struct ethhdr *eth;
 
 	eth = skb_push(skb, ETH_HLEN);
 	skb_reset_mac_header(skb);
 	ether_addr_copy(eth->h_source, np->dev->dev_addr);
-	ether_addr_copy(eth->h_dest, np->remote_mac);
+	ether_addr_copy(eth->h_dest, nt->remote_mac);
 	if (np->ipv6)
 		eth->h_proto = htons(ETH_P_IPV6);
 	else
@@ -1980,7 +1982,7 @@ static int netpoll_send_udp(struct netconsole_target *nt, const char *msg,
 		push_ipv6(np, skb, len);
 	else
 		push_ipv4(np, skb, len);
-	push_eth(np, skb);
+	push_eth(nt, skb);
 	skb->dev = np->dev;
 
 	return (int)netpoll_send_skb(np, skb);
@@ -2375,7 +2377,7 @@ static int netconsole_parser_cmdline(struct netconsole_target *nt, char *opt)
 
 	if (*cur != 0) {
 		/* MAC address */
-		if (!mac_pton(cur, np->remote_mac))
+		if (!mac_pton(cur, nt->remote_mac))
 			goto parse_failed;
 	}
 
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index 5ca79fa7d9431..79315461a7b1e 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -35,7 +35,6 @@ struct netpoll {
 
 	union inet_addr local_ip, remote_ip;
 	bool ipv6;
-	u8 remote_mac[ETH_ALEN];
 };
 
 #define np_info(np, fmt, ...)				\

-- 
2.53.0-Meta

Re: [PATCH net-next v3 0/9] netconsole: stop charging netpoll users for netconsole-only data

[Simon Horman]

On Fri, Jul 10, 2026 at 04:38:52AM -0700, Breno Leitao wrote:
> This work continue to untangle netconsole and netpoll, improving
> memory usage for netpoll users that are not netconsole.
> 
> struct netpoll is shared by every netpoll consumer in the tree:
> bonding, bridge, team, vlan, macvlan, dsa and netconsole. The non-
> netconsole users only need a per-port handle that lets them transmit
> SKBs.
> 
> Yet struct netpoll has accumulated a pile of fields that exist only
> to serve netconsole's printk path:
> 
>   - skb_pool / refill_wq: the fallback skb pool find_skb() falls back
>     on when alloc_skb() returns NULL, plus the workqueue that refills
>     it.
>   - local_port / remote_port / remote_mac: the source and destination
>     UDP ports and the destination ethernet address used to build
>     outgoing log packets.
> 
> Every netpoll user that allocates a struct netpoll pays for all of
> these fields, even though it never queues a single skb on the pool,
> never sends a UDP frame and never looks at any of the addressing.
> 
> Move this netconsole-only state out of the shared struct into struct
> netconsole_target, where its only consumer already lives.
> 
> After the series every passive netpoll consumer drops
> sizeof(sk_buff_head) + sizeof(work_struct) + 2 * sizeof(u16) +
> ETH_ALEN bytes per port (plus padding), and struct netpoll keeps only
> what belongs to the generic poll/transmit abstraction or what netpoll
> core still reads itself.
> 
> Follow-ups
> ----------
> local_ip, remote_ip and ipv6 are still in struct netpoll. Moving the
> addressing additionally requires relocating netpoll_setup() -- whose
> only caller is netconsole -- and its IPv4/IPv6 helpers, so it is left
> for a follow-up to keep this series focused on a single responsibility
> transfer.
> 
> This is pure code motion with no functional change. The pre-existing
> target/device teardown races reported against v1 have since been fixed
> independently (as suggested by Jakub) and are already in the tree, so,
> re-posting this one.
> 
> --
> 
> Changes in v3:
> - Added a fix for a pre-existing cleanup-skip leak (released/joined targets).
> - Added an skb-pool init-once patch closing an init/teardown race; rebased.
> - Link to v2: https://lore.kernel.org/r/20260702-netconsole_move_more-v2-0-1ebedd921dcb@debian.org
> 
> Changes in v2:
> - Rebase on net-next, now that the prerequisite netconsole
>   teardown-race fixes are merged.
> - Squash the local_port and remote_port moves into a single patch.
> - v1: https://lore.kernel.org/r/20260524-netconsole_move_more-v1-0-909d1ab398b4@debian.org

For the series:

Reviewed-by: Simon Horman <horms@kernel.org>