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, 457 insertions, 0 deletions

diff --git a/drivers/misc/lkdtm/usercopy.c b/drivers/misc/lkdtm/usercopy.c
new file mode 100644
index 000000000..67db57249
--- /dev/null
+++ b/drivers/misc/lkdtm/usercopy.c
@@ -0,0 +1,457 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * This is for all the tests related to copy_to_user() and copy_from_user()
+ * hardening.
+ */
+#include "lkdtm.h"
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/vmalloc.h>
+#include <linux/sched/task_stack.h>
+#include <linux/mman.h>
+#include <linux/uaccess.h>
+#include <asm/cacheflush.h>
+
+/*
+ * Many of the tests here end up using const sizes, but those would
+ * normally be ignored by hardened usercopy, so force the compiler
+ * into choosing the non-const path to make sure we trigger the
+ * hardened usercopy checks by added "unconst" to all the const copies,
+ * and making sure "cache_size" isn't optimized into a const.
+ */
+static volatile size_t unconst;
+static volatile size_t cache_size = 1024;
+static struct kmem_cache *whitelist_cache;
+
+static const unsigned char test_text[] = "This is a test.\n";
+
+/*
+ * Instead of adding -Wno-return-local-addr, just pass the stack address
+ * through a function to obfuscate it from the compiler.
+ */
+static noinline unsigned char *trick_compiler(unsigned char *stack)
+{
+	return stack + unconst;
+}
+
+static noinline unsigned char *do_usercopy_stack_callee(int value)
+{
+	unsigned char buf[128];
+	int i;
+
+	/* Exercise stack to avoid everything living in registers. */
+	for (i = 0; i < sizeof(buf); i++) {
+		buf[i] = value & 0xff;
+	}
+
+	/*
+	 * Put the target buffer in the middle of stack allocation
+	 * so that we don't step on future stack users regardless
+	 * of stack growth direction.
+	 */
+	return trick_compiler(&buf[(128/2)-32]);
+}
+
+static noinline void do_usercopy_stack(bool to_user, bool bad_frame)
+{
+	unsigned long user_addr;
+	unsigned char good_stack[32];
+	unsigned char *bad_stack;
+	int i;
+
+	/* Exercise stack to avoid everything living in registers. */
+	for (i = 0; i < sizeof(good_stack); i++)
+		good_stack[i] = test_text[i % sizeof(test_text)];
+
+	/* This is a pointer to outside our current stack frame. */
+	if (bad_frame) {
+		bad_stack = do_usercopy_stack_callee((uintptr_t)&bad_stack);
+	} else {
+		/* Put start address just inside stack. */
+		bad_stack = task_stack_page(current) + THREAD_SIZE;
+		bad_stack -= sizeof(unsigned long);
+	}
+
+#ifdef ARCH_HAS_CURRENT_STACK_POINTER
+	pr_info("stack     : %px\n", (void *)current_stack_pointer);
+#endif
+	pr_info("good_stack: %px-%px\n", good_stack, good_stack + sizeof(good_stack));
+	pr_info("bad_stack : %px-%px\n", bad_stack, bad_stack + sizeof(good_stack));
+
+	user_addr = vm_mmap(NULL, 0, PAGE_SIZE,
+			    PROT_READ | PROT_WRITE | PROT_EXEC,
+			    MAP_ANONYMOUS | MAP_PRIVATE, 0);
+	if (user_addr >= TASK_SIZE) {
+		pr_warn("Failed to allocate user memory\n");
+		return;
+	}
+
+	if (to_user) {
+		pr_info("attempting good copy_to_user of local stack\n");
+		if (copy_to_user((void __user *)user_addr, good_stack,
+				 unconst + sizeof(good_stack))) {
+			pr_warn("copy_to_user failed unexpectedly?!\n");
+			goto free_user;
+		}
+
+		pr_info("attempting bad copy_to_user of distant stack\n");
+		if (copy_to_user((void __user *)user_addr, bad_stack,
+				 unconst + sizeof(good_stack))) {
+			pr_warn("copy_to_user failed, but lacked Oops\n");
+			goto free_user;
+		}
+	} else {
+		/*
+		 * There isn't a safe way to not be protected by usercopy
+		 * if we're going to write to another thread's stack.
+		 */
+		if (!bad_frame)
+			goto free_user;
+
+		pr_info("attempting good copy_from_user of local stack\n");
+		if (copy_from_user(good_stack, (void __user *)user_addr,
+				   unconst + sizeof(good_stack))) {
+			pr_warn("copy_from_user failed unexpectedly?!\n");
+			goto free_user;
+		}
+
+		pr_info("attempting bad copy_from_user of distant stack\n");
+		if (copy_from_user(bad_stack, (void __user *)user_addr,
+				   unconst + sizeof(good_stack))) {
+			pr_warn("copy_from_user failed, but lacked Oops\n");
+			goto free_user;
+		}
+	}
+
+free_user:
+	vm_munmap(user_addr, PAGE_SIZE);
+}
+
+/*
+ * This checks for whole-object size validation with hardened usercopy,
+ * with or without usercopy whitelisting.
+ */
+static void do_usercopy_slab_size(bool to_user)
+{
+	unsigned long user_addr;
+	unsigned char *one, *two;
+	void __user *test_user_addr;
+	void *test_kern_addr;
+	size_t size = unconst + 1024;
+
+	one = kmalloc(size, GFP_KERNEL);
+	two = kmalloc(size, GFP_KERNEL);
+	if (!one || !two) {
+		pr_warn("Failed to allocate kernel memory\n");
+		goto free_kernel;
+	}
+
+	user_addr = vm_mmap(NULL, 0, PAGE_SIZE,
+			    PROT_READ | PROT_WRITE | PROT_EXEC,
+			    MAP_ANONYMOUS | MAP_PRIVATE, 0);
+	if (user_addr >= TASK_SIZE) {
+		pr_warn("Failed to allocate user memory\n");
+		goto free_kernel;
+	}
+
+	memset(one, 'A', size);
+	memset(two, 'B', size);
+
+	test_user_addr = (void __user *)(user_addr + 16);
+	test_kern_addr = one + 16;
+
+	if (to_user) {
+		pr_info("attempting good copy_to_user of correct size\n");
+		if (copy_to_user(test_user_addr, test_kern_addr, size / 2)) {
+			pr_warn("copy_to_user failed unexpectedly?!\n");
+			goto free_user;
+		}
+
+		pr_info("attempting bad copy_to_user of too large size\n");
+		if (copy_to_user(test_user_addr, test_kern_addr, size)) {
+			pr_warn("copy_to_user failed, but lacked Oops\n");
+			goto free_user;
+		}
+	} else {
+		pr_info("attempting good copy_from_user of correct size\n");
+		if (copy_from_user(test_kern_addr, test_user_addr, size / 2)) {
+			pr_warn("copy_from_user failed unexpectedly?!\n");
+			goto free_user;
+		}
+
+		pr_info("attempting bad copy_from_user of too large size\n");
+		if (copy_from_user(test_kern_addr, test_user_addr, size)) {
+			pr_warn("copy_from_user failed, but lacked Oops\n");
+			goto free_user;
+		}
+	}
+	pr_err("FAIL: bad usercopy not detected!\n");
+	pr_expected_config_param(CONFIG_HARDENED_USERCOPY, "hardened_usercopy");
+
+free_user:
+	vm_munmap(user_addr, PAGE_SIZE);
+free_kernel:
+	kfree(one);
+	kfree(two);
+}
+
+/*
+ * This checks for the specific whitelist window within an object. If this
+ * test passes, then do_usercopy_slab_size() tests will pass too.
+ */
+static void do_usercopy_slab_whitelist(bool to_user)
+{
+	unsigned long user_alloc;
+	unsigned char *buf = NULL;
+	unsigned char __user *user_addr;
+	size_t offset, size;
+
+	/* Make sure cache was prepared. */
+	if (!whitelist_cache) {
+		pr_warn("Failed to allocate kernel cache\n");
+		return;
+	}
+
+	/*
+	 * Allocate a buffer with a whitelisted window in the buffer.
+	 */
+	buf = kmem_cache_alloc(whitelist_cache, GFP_KERNEL);
+	if (!buf) {
+		pr_warn("Failed to allocate buffer from whitelist cache\n");
+		goto free_alloc;
+	}
+
+	/* Allocate user memory we'll poke at. */
+	user_alloc = vm_mmap(NULL, 0, PAGE_SIZE,
+			    PROT_READ | PROT_WRITE | PROT_EXEC,
+			    MAP_ANONYMOUS | MAP_PRIVATE, 0);
+	if (user_alloc >= TASK_SIZE) {
+		pr_warn("Failed to allocate user memory\n");
+		goto free_alloc;
+	}
+	user_addr = (void __user *)user_alloc;
+
+	memset(buf, 'B', cache_size);
+
+	/* Whitelisted window in buffer, from kmem_cache_create_usercopy. */
+	offset = (cache_size / 4) + unconst;
+	size = (cache_size / 16) + unconst;
+
+	if (to_user) {
+		pr_info("attempting good copy_to_user inside whitelist\n");
+		if (copy_to_user(user_addr, buf + offset, size)) {
+			pr_warn("copy_to_user failed unexpectedly?!\n");
+			goto free_user;
+		}
+
+		pr_info("attempting bad copy_to_user outside whitelist\n");
+		if (copy_to_user(user_addr, buf + offset - 1, size)) {
+			pr_warn("copy_to_user failed, but lacked Oops\n");
+			goto free_user;
+		}
+	} else {
+		pr_info("attempting good copy_from_user inside whitelist\n");
+		if (copy_from_user(buf + offset, user_addr, size)) {
+			pr_warn("copy_from_user failed unexpectedly?!\n");
+			goto free_user;
+		}
+
+		pr_info("attempting bad copy_from_user outside whitelist\n");
+		if (copy_from_user(buf + offset - 1, user_addr, size)) {
+			pr_warn("copy_from_user failed, but lacked Oops\n");
+			goto free_user;
+		}
+	}
+	pr_err("FAIL: bad usercopy not detected!\n");
+	pr_expected_config_param(CONFIG_HARDENED_USERCOPY, "hardened_usercopy");
+
+free_user:
+	vm_munmap(user_alloc, PAGE_SIZE);
+free_alloc:
+	if (buf)
+		kmem_cache_free(whitelist_cache, buf);
+}
+
+/* Callable tests. */
+static void lkdtm_USERCOPY_SLAB_SIZE_TO(void)
+{
+	do_usercopy_slab_size(true);
+}
+
+static void lkdtm_USERCOPY_SLAB_SIZE_FROM(void)
+{
+	do_usercopy_slab_size(false);
+}
+
+static void lkdtm_USERCOPY_SLAB_WHITELIST_TO(void)
+{
+	do_usercopy_slab_whitelist(true);
+}
+
+static void lkdtm_USERCOPY_SLAB_WHITELIST_FROM(void)
+{
+	do_usercopy_slab_whitelist(false);
+}
+
+static void lkdtm_USERCOPY_STACK_FRAME_TO(void)
+{
+	do_usercopy_stack(true, true);
+}
+
+static void lkdtm_USERCOPY_STACK_FRAME_FROM(void)
+{
+	do_usercopy_stack(false, true);
+}
+
+static void lkdtm_USERCOPY_STACK_BEYOND(void)
+{
+	do_usercopy_stack(true, false);
+}
+
+static void lkdtm_USERCOPY_KERNEL(void)
+{
+	unsigned long user_addr;
+
+	user_addr = vm_mmap(NULL, 0, PAGE_SIZE,
+			    PROT_READ | PROT_WRITE | PROT_EXEC,
+			    MAP_ANONYMOUS | MAP_PRIVATE, 0);
+	if (user_addr >= TASK_SIZE) {
+		pr_warn("Failed to allocate user memory\n");
+		return;
+	}
+
+	pr_info("attempting good copy_to_user from kernel rodata: %px\n",
+		test_text);
+	if (copy_to_user((void __user *)user_addr, test_text,
+			 unconst + sizeof(test_text))) {
+		pr_warn("copy_to_user failed unexpectedly?!\n");
+		goto free_user;
+	}
+
+	pr_info("attempting bad copy_to_user from kernel text: %px\n",
+		vm_mmap);
+	if (copy_to_user((void __user *)user_addr, vm_mmap,
+			 unconst + PAGE_SIZE)) {
+		pr_warn("copy_to_user failed, but lacked Oops\n");
+		goto free_user;
+	}
+	pr_err("FAIL: bad copy_to_user() not detected!\n");
+	pr_expected_config_param(CONFIG_HARDENED_USERCOPY, "hardened_usercopy");
+
+free_user:
+	vm_munmap(user_addr, PAGE_SIZE);
+}
+
+/*
+ * This expects "kaddr" to point to a PAGE_SIZE allocation, which means
+ * a more complete test that would include copy_from_user() would risk
+ * memory corruption. Just test copy_to_user() here, as that exercises
+ * almost exactly the same code paths.
+ */
+static void do_usercopy_page_span(const char *name, void *kaddr)
+{
+	unsigned long uaddr;
+
+	uaddr = vm_mmap(NULL, 0, PAGE_SIZE, PROT_READ | PROT_WRITE,
+			MAP_ANONYMOUS | MAP_PRIVATE, 0);
+	if (uaddr >= TASK_SIZE) {
+		pr_warn("Failed to allocate user memory\n");
+		return;
+	}
+
+	/* Initialize contents. */
+	memset(kaddr, 0xAA, PAGE_SIZE);
+
+	/* Bump the kaddr forward to detect a page-spanning overflow. */
+	kaddr += PAGE_SIZE / 2;
+
+	pr_info("attempting good copy_to_user() from kernel %s: %px\n",
+		name, kaddr);
+	if (copy_to_user((void __user *)uaddr, kaddr,
+			 unconst + (PAGE_SIZE / 2))) {
+		pr_err("copy_to_user() failed unexpectedly?!\n");
+		goto free_user;
+	}
+
+	pr_info("attempting bad copy_to_user() from kernel %s: %px\n",
+		name, kaddr);
+	if (copy_to_user((void __user *)uaddr, kaddr, unconst + PAGE_SIZE)) {
+		pr_warn("Good, copy_to_user() failed, but lacked Oops(?!)\n");
+		goto free_user;
+	}
+
+	pr_err("FAIL: bad copy_to_user() not detected!\n");
+	pr_expected_config_param(CONFIG_HARDENED_USERCOPY, "hardened_usercopy");
+
+free_user:
+	vm_munmap(uaddr, PAGE_SIZE);
+}
+
+static void lkdtm_USERCOPY_VMALLOC(void)
+{
+	void *addr;
+
+	addr = vmalloc(PAGE_SIZE);
+	if (!addr) {
+		pr_err("vmalloc() failed!?\n");
+		return;
+	}
+	do_usercopy_page_span("vmalloc", addr);
+	vfree(addr);
+}
+
+static void lkdtm_USERCOPY_FOLIO(void)
+{
+	struct folio *folio;
+	void *addr;
+
+	/*
+	 * FIXME: Folio checking currently misses 0-order allocations, so
+	 * allocate and bump forward to the last page.
+	 */
+	folio = folio_alloc(GFP_KERNEL | __GFP_ZERO, 1);
+	if (!folio) {
+		pr_err("folio_alloc() failed!?\n");
+		return;
+	}
+	addr = folio_address(folio);
+	if (addr)
+		do_usercopy_page_span("folio", addr + PAGE_SIZE);
+	else
+		pr_err("folio_address() failed?!\n");
+	folio_put(folio);
+}
+
+void __init lkdtm_usercopy_init(void)
+{
+	/* Prepare cache that lacks SLAB_USERCOPY flag. */
+	whitelist_cache =
+		kmem_cache_create_usercopy("lkdtm-usercopy", cache_size,
+					   0, 0,
+					   cache_size / 4,
+					   cache_size / 16,
+					   NULL);
+}
+
+void __exit lkdtm_usercopy_exit(void)
+{
+	kmem_cache_destroy(whitelist_cache);
+}
+
+static struct crashtype crashtypes[] = {
+	CRASHTYPE(USERCOPY_SLAB_SIZE_TO),
+	CRASHTYPE(USERCOPY_SLAB_SIZE_FROM),
+	CRASHTYPE(USERCOPY_SLAB_WHITELIST_TO),
+	CRASHTYPE(USERCOPY_SLAB_WHITELIST_FROM),
+	CRASHTYPE(USERCOPY_STACK_FRAME_TO),
+	CRASHTYPE(USERCOPY_STACK_FRAME_FROM),
+	CRASHTYPE(USERCOPY_STACK_BEYOND),
+	CRASHTYPE(USERCOPY_VMALLOC),
+	CRASHTYPE(USERCOPY_FOLIO),
+	CRASHTYPE(USERCOPY_KERNEL),
+};
+
+struct crashtype_category usercopy_crashtypes = {
+	.crashtypes = crashtypes,
+	.len	    = ARRAY_SIZE(crashtypes),
+};