diff mbox series

[RFC,-next,2/2] net: add dev_hold_track() and dev_put_track() helpers

Message ID 20211117192031.3906502-3-eric.dumazet@gmail.com (mailing list archive)
State RFC
Delegated to: Netdev Maintainers
Headers show
Series refcount: add tracking infrastructure | expand

Checks

Context Check Description
netdev/fixes_present success Fixes tag not required for -next series
netdev/subject_prefix success Link
netdev/cover_letter success Series has a cover letter
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 4760 this patch: 4760
netdev/cc_maintainers success CCed 4 of 4 maintainers
netdev/build_clang success Errors and warnings before: 862 this patch: 862
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 4917 this patch: 4917
netdev/checkpatch warning WARNING: please write a paragraph that describes the config symbol fully
netdev/kdoc fail Errors and warnings before: 0 this patch: 1
netdev/source_inline success Was 0 now: 0
netdev/tree_selection success Guessing tree name failed - patch did not apply, async

Commit Message

Eric Dumazet Nov. 17, 2021, 7:20 p.m. UTC 
From: Eric Dumazet <edumazet@google.com>

They should replace dev_hold() and dev_put().

To use these helpers, each data structure owning a refcount
should also use a "struct ref_tracker" to pair the hold and put.

Whenever a leak happens, we will get precise stack traces
of the point dev_hold_track() happened, at device dismantle phase.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/linux/netdevice.h | 23 +++++++++++++++++++++++
 net/Kconfig               |  8 ++++++++
 net/core/dev.c            |  3 +++
 3 files changed, 34 insertions(+)
diff mbox series

Patch

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 4f4a299e92de7ba9f61507ad4df7e334775c07a6..91957aa0779195a962ec95f491d826bdd536808e 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -48,6 +48,7 @@ 
 #include <uapi/linux/pkt_cls.h>
 #include <linux/hashtable.h>
 #include <linux/rbtree.h>
+#include <linux/ref_tracker.h>
 
 struct netpoll_info;
 struct device;
@@ -2181,6 +2182,7 @@  struct net_device {
 #else
 	refcount_t		dev_refcnt;
 #endif
+	struct ref_tracker_dir	refcnt_tracker;
 
 	struct list_head	link_watch_list;
 
@@ -3807,6 +3809,7 @@  void netdev_run_todo(void);
  *	@dev: network device
  *
  * Release reference to device to allow it to be freed.
+ * Try using dev_put_track() instead.
  */
 static inline void dev_put(struct net_device *dev)
 {
@@ -3824,6 +3827,7 @@  static inline void dev_put(struct net_device *dev)
  *	@dev: network device
  *
  * Hold reference to device to keep it from being freed.
+ * Try using dev_hold_track() instead.
  */
 static inline void dev_hold(struct net_device *dev)
 {
@@ -3836,6 +3840,25 @@  static inline void dev_hold(struct net_device *dev)
 	}
 }
 
+static inline void dev_hold_track(struct net_device *dev,
+				  struct ref_tracker **tracker,
+				  gfp_t gfp)
+{
+	if (dev) {
+		dev_hold(dev);
+		ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
+	}
+}
+
+static inline void dev_put_track(struct net_device *dev,
+				 struct ref_tracker **tracker)
+{
+	if (dev) {
+		ref_tracker_free(&dev->refcnt_tracker, tracker);
+		dev_put(dev);
+	}
+}
+
 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
  * and _off may be called from IRQ context, but it is caller
  * who is responsible for serialization of these calls.
diff --git a/net/Kconfig b/net/Kconfig
index 074472dfa94ae78081b7391b8ca4a73b9d0be7b7..a4743e59a35c2978ecc6d704b388ca07efe3e95c 100644
--- a/net/Kconfig
+++ b/net/Kconfig
@@ -253,6 +253,14 @@  config PCPU_DEV_REFCNT
 	  network device refcount are using per cpu variables if this option is set.
 	  This can be forced to N to detect underflows (with a performance drop).
 
+config NET_DEV_REFCNT_TRACKER
+	bool "Enable tracking in dev_put_track() and dev_hold_track()"
+	select REF_TRACKER
+	default n
+	help
+	  Enable debugging feature to track leaked device references.
+	  This adds memory and cpu costs.
+
 config RPS
 	bool
 	depends on SMP && SYSFS
diff --git a/net/core/dev.c b/net/core/dev.c
index 92c9258cbf28556e68f9112343f5ebc98b2c163b..90ee2ba8717bf74bd3a1f72f7034744773ef69c4 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -9878,6 +9878,7 @@  static void netdev_wait_allrefs(struct net_device *dev)
 			       netdev_unregister_timeout_secs * HZ)) {
 			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
 				 dev->name, refcnt);
+			ref_tracker_dir_print(&dev->refcnt_tracker, 10);
 			warning_time = jiffies;
 		}
 	}
@@ -10168,6 +10169,7 @@  struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	dev = PTR_ALIGN(p, NETDEV_ALIGN);
 	dev->padded = (char *)dev - (char *)p;
 
+	ref_tracker_dir_init(&dev->refcnt_tracker, 128);
 #ifdef CONFIG_PCPU_DEV_REFCNT
 	dev->pcpu_refcnt = alloc_percpu(int);
 	if (!dev->pcpu_refcnt)
@@ -10284,6 +10286,7 @@  void free_netdev(struct net_device *dev)
 	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
 		netif_napi_del(p);
 
+	ref_tracker_dir_exit(&dev->refcnt_tracker);
 #ifdef CONFIG_PCPU_DEV_REFCNT
 	free_percpu(dev->pcpu_refcnt);
 	dev->pcpu_refcnt = NULL;