x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 180 insertions, 0 deletions

diff --git a/net/sysctl_net.c b/net/sysctl_net.c
new file mode 100644
index 000000000..e190a639e
--- /dev/null
+++ b/net/sysctl_net.c
@@ -0,0 +1,180 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* -*- linux-c -*-
+ * sysctl_net.c: sysctl interface to net subsystem.
+ *
+ * Begun April 1, 1996, Mike Shaver.
+ * Added /proc/sys/net directories for each protocol family. [MS]
+ *
+ * Revision 1.2  1996/05/08  20:24:40  shaver
+ * Added bits for NET_BRIDGE and the NET_IPV4_ARP stuff and
+ * NET_IPV4_IP_FORWARD.
+ *
+ *
+ */
+
+#include <linux/mm.h>
+#include <linux/export.h>
+#include <linux/sysctl.h>
+#include <linux/nsproxy.h>
+
+#include <net/sock.h>
+
+#ifdef CONFIG_INET
+#include <net/ip.h>
+#endif
+
+#ifdef CONFIG_NET
+#include <linux/if_ether.h>
+#endif
+
+static struct ctl_table_set *
+net_ctl_header_lookup(struct ctl_table_root *root)
+{
+	return &current->nsproxy->net_ns->sysctls;
+}
+
+static int is_seen(struct ctl_table_set *set)
+{
+	return &current->nsproxy->net_ns->sysctls == set;
+}
+
+/* Return standard mode bits for table entry. */
+static int net_ctl_permissions(struct ctl_table_header *head,
+			       const struct ctl_table *table)
+{
+	struct net *net = container_of(head->set, struct net, sysctls);
+
+	/* Allow network administrator to have same access as root. */
+	if (ns_capable_noaudit(net->user_ns, CAP_NET_ADMIN)) {
+		int mode = (table->mode >> 6) & 7;
+		return (mode << 6) | (mode << 3) | mode;
+	}
+
+	return table->mode;
+}
+
+static void net_ctl_set_ownership(struct ctl_table_header *head,
+				  kuid_t *uid, kgid_t *gid)
+{
+	struct net *net = container_of(head->set, struct net, sysctls);
+	kuid_t ns_root_uid;
+	kgid_t ns_root_gid;
+
+	ns_root_uid = make_kuid(net->user_ns, 0);
+	if (uid_valid(ns_root_uid))
+		*uid = ns_root_uid;
+
+	ns_root_gid = make_kgid(net->user_ns, 0);
+	if (gid_valid(ns_root_gid))
+		*gid = ns_root_gid;
+}
+
+static struct ctl_table_root net_sysctl_root = {
+	.lookup = net_ctl_header_lookup,
+	.permissions = net_ctl_permissions,
+	.set_ownership = net_ctl_set_ownership,
+};
+
+static int __net_init sysctl_net_init(struct net *net)
+{
+	setup_sysctl_set(&net->sysctls, &net_sysctl_root, is_seen);
+	return 0;
+}
+
+static void __net_exit sysctl_net_exit(struct net *net)
+{
+	retire_sysctl_set(&net->sysctls);
+}
+
+static struct pernet_operations sysctl_pernet_ops = {
+	.init = sysctl_net_init,
+	.exit = sysctl_net_exit,
+};
+
+static struct ctl_table_header *net_header;
+__init int net_sysctl_init(void)
+{
+	static struct ctl_table empty[1];
+	int ret = -ENOMEM;
+	/* Avoid limitations in the sysctl implementation by
+	 * registering "/proc/sys/net" as an empty directory not in a
+	 * network namespace.
+	 */
+	net_header = register_sysctl_sz("net", empty, 0);
+	if (!net_header)
+		goto out;
+	ret = register_pernet_subsys(&sysctl_pernet_ops);
+	if (ret)
+		goto out1;
+out:
+	return ret;
+out1:
+	unregister_sysctl_table(net_header);
+	net_header = NULL;
+	goto out;
+}
+
+/* Return error when sysctls for non-init netns are unsafe by verifying:
+ * 1) being read-only, or
+ * 2) having a data pointer which points outside of the global kernel/module
+ *    data segment, and rather into the heap where a per-net object was
+ *    allocated.
+ */
+static int ensure_safe_net_sysctl(struct net *net, const char *path,
+				  const struct ctl_table *table,
+				  size_t table_size)
+{
+	const struct ctl_table *ent;
+
+	pr_debug("Registering net sysctl (net %p): %s\n", net, path);
+	ent = table;
+	for (size_t i = 0; i < table_size; ent++, i++) {
+		unsigned long addr;
+		const char *where;
+
+		pr_debug("  procname=%s mode=%o proc_handler=%ps data=%p\n",
+			 ent->procname, ent->mode, ent->proc_handler, ent->data);
+
+		/* If it's not writable inside the netns, then it can't hurt. */
+		if ((ent->mode & 0222) == 0) {
+			pr_debug("    Not writable by anyone\n");
+			continue;
+		}
+
+		/* Where does data point? */
+		addr = (unsigned long)ent->data;
+		if (is_module_address(addr))
+			where = "module";
+		else if (is_kernel_core_data(addr))
+			where = "kernel";
+		else
+			continue;
+
+		/* Warn on netns leak. */
+		WARN(1, "sysctl %s/%s: data points to %s global data: %ps\n",
+		     path, ent->procname, where, ent->data);
+
+		return -EACCES;
+	}
+
+	return 0;
+}
+
+struct ctl_table_header *register_net_sysctl_sz(struct net *net,
+						const char *path,
+						const struct ctl_table *table,
+						size_t table_size)
+{
+	if (!net_eq(net, &init_net))
+		if (ensure_safe_net_sysctl(net, path, table, table_size))
+			return NULL;
+
+	return __register_sysctl_table(&net->sysctls, path, table, table_size);
+}
+EXPORT_SYMBOL_GPL(register_net_sysctl_sz);
+
+void unregister_net_sysctl_table(struct ctl_table_header *header)
+{
+	unregister_sysctl_table(header);
+}
+EXPORT_SYMBOL_GPL(unregister_net_sysctl_table);