net: Implement fault injection forcing skb reallocation

Introduce a fault injection mechanism to force skb reallocation. The
primary goal is to catch bugs related to pointer invalidation after
potential skb reallocation.

The fault injection mechanism aims to identify scenarios where callers
retain pointers to various headers in the skb but fail to reload these
pointers after calling a function that may reallocate the data. This
type of bug can lead to memory corruption or crashes if the old,
now-invalid pointers are used.

By forcing reallocation through fault injection, we can stress-test code
paths and ensure proper pointer management after potential skb
reallocations.

Add a hook for fault injection in the following functions:

* pskb_trim_rcsum()
* pskb_may_pull_reason()
* pskb_trim()

As the other fault injection mechanism, protect it under a debug Kconfig
called CONFIG_FAIL_SKB_REALLOC.

This patch was *heavily* inspired by Jakub's proposal from:
https://lore.kernel.org/all/[e-mail]/

CC: Akinobu Mita <[e-mail]>
Suggested-by: Jakub Kicinski <[e-mail]>
Signed-off-by: Breno Leitao <[e-mail]>
Reviewed-by: Akinobu Mita <[e-mail]>
Acked-by: Paolo Abeni <[e-mail]>
Acked-by: Guillaume Nault <[e-mail]>
Link: https://patch.msgid.link/[e-mail]
Signed-off-by: Paolo Abeni <[e-mail]>

Documentation/admin-guide/kernel-parameters.txt

···
 	failslab=
 	fail_usercopy=
 	fail_page_alloc=
+	fail_skb_realloc=
 	fail_make_request=[KNL]
 			General fault injection mechanism.
 			Format: <interval>,<probability>,<space>,<times>

Documentation/fault-injection/fault-injection.rst

···
   ALLOW_ERROR_INJECTION() macro, by setting debugfs entries
   under /sys/kernel/debug/fail_function. No boot option supported.
 
+- fail_skb_realloc
+
+  inject skb (socket buffer) reallocation events into the network path. The
+  primary goal is to identify and prevent issues related to pointer
+  mismanagement in the network subsystem.  By forcing skb reallocation at
+  strategic points, this feature creates scenarios where existing pointers to
+  skb headers become invalid.
+
+  When the fault is injected and the reallocation is triggered, cached pointers
+  to skb headers and data no longer reference valid memory locations. This
+  deliberate invalidation helps expose code paths where proper pointer updating
+  is neglected after a reallocation event.
+
+  By creating these controlled fault scenarios, the system can catch instances
+  where stale pointers are used, potentially leading to memory corruption or
+  system instability.
+
+  To select the interface to act on, write the network name to
+  /sys/kernel/debug/fail_skb_realloc/devname.
+  If this field is left empty (which is the default value), skb reallocation
+  will be forced on all network interfaces.
+
+  The effectiveness of this fault detection is enhanced when KASAN is
+  enabled, as it helps identify invalid memory references and use-after-free
+  (UAF) issues.
+
 - NVMe fault injection
 
   inject NVMe status code and retry flag on devices permitted by setting
···
 	use a negative errno, you better use 'printf' instead of 'echo', e.g.:
 	$ printf %#x -12 > retval
 
+- /sys/kernel/debug/fail_skb_realloc/devname:
+
+        Specifies the network interface on which to force SKB reallocation.  If
+        left empty, SKB reallocation will be applied to all network interfaces.
+
+        Example usage::
+
+          # Force skb reallocation on eth0
+          echo "eth0" > /sys/kernel/debug/fail_skb_realloc/devname
+
+          # Clear the selection and force skb reallocation on all interfaces
+          echo "" > /sys/kernel/debug/fail_skb_realloc/devname
+
 Boot option
 ^^^^^^^^^^^
 
···
 	fail_usercopy=
 	fail_make_request=
 	fail_futex=
+	fail_skb_realloc=
 	mmc_core.fail_request=<interval>,<probability>,<space>,<times>
 
 proc entries

include/linux/skbuff.h

···
 #endif /* CONFIG_DEBUG_NET */
 }
 
+#if defined(CONFIG_FAIL_SKB_REALLOC)
+void skb_might_realloc(struct sk_buff *skb);
+#else
+static inline void skb_might_realloc(struct sk_buff *skb) {}
+#endif
+
 /*
  *	Add data to an sk_buff
  */
···
 pskb_may_pull_reason(struct sk_buff *skb, unsigned int len)
 {
 	DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
+	skb_might_realloc(skb);
 
 	if (likely(len <= skb_headlen(skb)))
 		return SKB_NOT_DROPPED_YET;
···
 
 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
 {
+	skb_might_realloc(skb);
 	return (len < skb->len) ? __pskb_trim(skb, len) : 0;
 }
 
···
 
 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
 {
+	skb_might_realloc(skb);
 	if (likely(len >= skb->len))
 		return 0;
 	return pskb_trim_rcsum_slow(skb, len);

lib/Kconfig.debug

···
 	  Provide fault-injection capability for SunRPC and
 	  its consumers.
 
+config FAIL_SKB_REALLOC
+	bool "Fault-injection capability forcing skb to reallocate"
+	depends on FAULT_INJECTION_DEBUG_FS
+	help
+	  Provide fault-injection capability that forces the skb to be
+	  reallocated, catching possible invalid pointers to the skb.
+
+	  For more information, check
+	  Documentation/dev-tools/fault-injection/fault-injection.rst
+
 config FAULT_INJECTION_CONFIGFS
 	bool "Configfs interface for fault-injection capabilities"
 	depends on FAULT_INJECTION

net/core/Makefile

···
 obj-$(CONFIG_NET_TEST) += net_test.o
 obj-$(CONFIG_NET_DEVMEM) += devmem.o
 obj-$(CONFIG_DEBUG_NET_SMALL_RTNL) += rtnl_net_debug.o
+obj-$(CONFIG_FAIL_SKB_REALLOC) += skb_fault_injection.o

net/core/skb_fault_injection.c

···
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <linux/debugfs.h>
+#include <linux/fault-inject.h>
+#include <linux/netdevice.h>
+#include <linux/skbuff.h>
+
+static struct {
+	struct fault_attr attr;
+	char devname[IFNAMSIZ];
+	bool filtered;
+} skb_realloc = {
+	.attr = FAULT_ATTR_INITIALIZER,
+	.filtered = false,
+};
+
+static bool should_fail_net_realloc_skb(struct sk_buff *skb)
+{
+	struct net_device *net = skb->dev;
+
+	if (skb_realloc.filtered &&
+	    strncmp(net->name, skb_realloc.devname, IFNAMSIZ))
+		/* device name filter set, but names do not match */
+		return false;
+
+	if (!should_fail(&skb_realloc.attr, 1))
+		return false;
+
+	return true;
+}
+ALLOW_ERROR_INJECTION(should_fail_net_realloc_skb, TRUE);
+
+void skb_might_realloc(struct sk_buff *skb)
+{
+	if (!should_fail_net_realloc_skb(skb))
+		return;
+
+	pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
+}
+EXPORT_SYMBOL(skb_might_realloc);
+
+static int __init fail_skb_realloc_setup(char *str)
+{
+	return setup_fault_attr(&skb_realloc.attr, str);
+}
+__setup("fail_skb_realloc=", fail_skb_realloc_setup);
+
+static void reset_settings(void)
+{
+	skb_realloc.filtered = false;
+	memset(&skb_realloc.devname, 0, IFNAMSIZ);
+}
+
+static ssize_t devname_write(struct file *file, const char __user *buffer,
+			     size_t count, loff_t *ppos)
+{
+	ssize_t ret;
+
+	reset_settings();
+	ret = simple_write_to_buffer(&skb_realloc.devname, IFNAMSIZ,
+				     ppos, buffer, count);
+	if (ret < 0)
+		return ret;
+
+	skb_realloc.devname[IFNAMSIZ - 1] = '\0';
+	/* Remove a possible \n at the end of devname */
+	strim(skb_realloc.devname);
+
+	if (strnlen(skb_realloc.devname, IFNAMSIZ))
+		skb_realloc.filtered = true;
+
+	return count;
+}
+
+static ssize_t devname_read(struct file *file,
+			    char __user *buffer,
+			    size_t size, loff_t *ppos)
+{
+	if (!skb_realloc.filtered)
+		return 0;
+
+	return simple_read_from_buffer(buffer, size, ppos, &skb_realloc.devname,
+				       strlen(skb_realloc.devname));
+}
+
+static const struct file_operations devname_ops = {
+	.write = devname_write,
+	.read = devname_read,
+};
+
+static int __init fail_skb_realloc_debugfs(void)
+{
+	umode_t mode = S_IFREG | 0600;
+	struct dentry *dir;
+
+	dir = fault_create_debugfs_attr("fail_skb_realloc", NULL,
+					&skb_realloc.attr);
+	if (IS_ERR(dir))
+		return PTR_ERR(dir);
+
+	debugfs_create_file("devname", mode, dir, NULL, &devname_ops);
+
+	return 0;
+}
+
+late_initcall(fail_skb_realloc_debugfs);