linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Josh Poimboeuf <jpoimboe@kernel.org>
To: Jiangfeng Xiao <xiaojiangfeng@huawei.com>
Cc: Kees Cook <keescook@chromium.org>, Jann Horn <jannh@google.com>,
	gustavoars@kernel.org, akpm@linux-foundation.org,
	peterz@infradead.org, dave.hansen@linux.intel.com,
	kirill.shutemov@linux.intel.com, linux-kernel@vger.kernel.org,
	linux-hardening@vger.kernel.org, linux-mm@kvack.org,
	nixiaoming@huawei.com, kepler.chenxin@huawei.com,
	wangbing6@huawei.com, wangfangpeng1@huawei.com,
	douzhaolei@huawei.com, Russell King <linux@armlinux.org.uk>,
	linux-arm-kernel@lists.infradead.org,
	Ard Biesheuvel <ardb@kernel.org>
Subject: Re: [PATCH] usercopy: delete __noreturn from usercopy_abort
Date: Tue, 5 Mar 2024 09:58:46 -0800	[thread overview]
Message-ID: <20240305175846.qnyiru7uaa7itqba@treble> (raw)
In-Reply-To: <b274b545-9439-7ff8-e3ed-604a9ac81f65@huawei.com>

> >> For the usercopy_abort function, whether '__noreturn' is added
> >> does not affect the internal behavior of the usercopy_abort function.
> >> Therefore, it is recommended that '__noreturn' be deleted
> >> so that backtrace can work properly.
> > 
> > This isn't acceptable. Removing __noreturn this will break
> > objtool's processing of execution flow for livepatching, IBT, and
> > KCFI instrumentation. These all depend on an accurate control flow
> > descriptions, and usercopy_abort is correctly marked __noreturn.

__noreturn also has the benefit of enabling the compiler to produce more
compact code for callees.

> Thank you for providing this information.
> I'll go back to further understand how __noreturn is used
> in features such as KCFI and livepatching.

Adding ARM folks -- see
https://lkml.kernel.org/lkml/1709516385-7778-1-git-send-email-xiaojiangfeng@huawei.com
for the original bug report.

This is an off-by-one bug which is common in unwinders, due to the fact
that the address on the stack points to the return address rather than
the call address.

So, for example, when the last instruction of a function is a function
call (e.g., to a noreturn function), it can cause the unwinder to
incorrectly try to unwind from the function *after* the callee.

For ORC (x86), we fixed this by decrementing the PC for call frames (but
not exception frames).  I've seen user space unwinders do similar, for
non-signal frames.

Something like the following might fix your issue (completely untested):

diff --git a/arch/arm/include/asm/stacktrace.h b/arch/arm/include/asm/stacktrace.h
index 360f0d2406bf..4891e38cdc1f 100644
--- a/arch/arm/include/asm/stacktrace.h
+++ b/arch/arm/include/asm/stacktrace.h
@@ -21,9 +21,7 @@ struct stackframe {
 	struct llist_node *kr_cur;
 	struct task_struct *tsk;
 #endif
-#ifdef CONFIG_UNWINDER_FRAME_POINTER
 	bool ex_frame;
-#endif
 };
 
 static __always_inline
@@ -37,9 +35,8 @@ void arm_get_current_stackframe(struct pt_regs *regs, struct stackframe *frame)
 		frame->kr_cur = NULL;
 		frame->tsk = current;
 #endif
-#ifdef CONFIG_UNWINDER_FRAME_POINTER
-		frame->ex_frame = in_entry_text(frame->pc);
-#endif
+		frame->ex_frame = !!regs;
+
 }
 
 extern int unwind_frame(struct stackframe *frame);
diff --git a/arch/arm/kernel/stacktrace.c b/arch/arm/kernel/stacktrace.c
index 620aa82e3bdd..caed7436da09 100644
--- a/arch/arm/kernel/stacktrace.c
+++ b/arch/arm/kernel/stacktrace.c
@@ -154,9 +154,6 @@ static void start_stack_trace(struct stackframe *frame, struct task_struct *task
 	frame->kr_cur = NULL;
 	frame->tsk = task;
 #endif
-#ifdef CONFIG_UNWINDER_FRAME_POINTER
-	frame->ex_frame = in_entry_text(frame->pc);
-#endif
 }
 
 void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,
@@ -167,6 +164,7 @@ void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,
 	if (regs) {
 		start_stack_trace(&frame, NULL, regs->ARM_fp, regs->ARM_sp,
 				  regs->ARM_lr, regs->ARM_pc);
+		frame.ex_frame = true;
 	} else if (task != current) {
 #ifdef CONFIG_SMP
 		/*
@@ -180,6 +178,7 @@ void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,
 				  thread_saved_sp(task), 0,
 				  thread_saved_pc(task));
 #endif
+		frame.ex_frame = false;
 	} else {
 here:
 		start_stack_trace(&frame, task,
@@ -187,6 +186,7 @@ void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,
 				  current_stack_pointer,
 				  (unsigned long)__builtin_return_address(0),
 				  (unsigned long)&&here);
+		frame.ex_frame = false;
 		/* skip this function */
 		if (unwind_frame(&frame))
 			return;
diff --git a/arch/arm/kernel/traps.c b/arch/arm/kernel/traps.c
index 3bad79db5d6e..46a5b1eb3f0a 100644
--- a/arch/arm/kernel/traps.c
+++ b/arch/arm/kernel/traps.c
@@ -84,10 +84,10 @@ void dump_backtrace_entry(unsigned long where, unsigned long from,
 	printk("%sFunction entered at [<%08lx>] from [<%08lx>]\n",
 		loglvl, where, from);
 #elif defined CONFIG_BACKTRACE_VERBOSE
-	printk("%s[<%08lx>] (%ps) from [<%08lx>] (%pS)\n",
+	printk("%s[<%08lx>] (%ps) from [<%08lx>] (%pB)\n",
 		loglvl, where, (void *)where, from, (void *)from);
 #else
-	printk("%s %ps from %pS\n", loglvl, (void *)where, (void *)from);
+	printk("%s %ps from %pB\n", loglvl, (void *)where, (void *)from);
 #endif
 
 	if (in_entry_text(from) && end <= ALIGN(frame, THREAD_SIZE))
diff --git a/arch/arm/kernel/unwind.c b/arch/arm/kernel/unwind.c
index 9d2192156087..99ded32196af 100644
--- a/arch/arm/kernel/unwind.c
+++ b/arch/arm/kernel/unwind.c
@@ -407,7 +407,7 @@ int unwind_frame(struct stackframe *frame)
 {
 	const struct unwind_idx *idx;
 	struct unwind_ctrl_block ctrl;
-	unsigned long sp_low;
+	unsigned long sp_low, pc;
 
 	/* store the highest address on the stack to avoid crossing it*/
 	sp_low = frame->sp;
@@ -417,19 +417,22 @@ int unwind_frame(struct stackframe *frame)
 	pr_debug("%s(pc = %08lx lr = %08lx sp = %08lx)\n", __func__,
 		 frame->pc, frame->lr, frame->sp);
 
-	idx = unwind_find_idx(frame->pc);
+	pc = frame->ex_frame ? frame->pc : frame->pc - 4;
+
+	idx = unwind_find_idx(pc);
 	if (!idx) {
-		if (frame->pc && kernel_text_address(frame->pc)) {
-			if (in_module_plt(frame->pc) && frame->pc != frame->lr) {
+		if (kernel_text_address(pc)) {
+			if (in_module_plt(pc) && frame->pc != frame->lr) {
 				/*
 				 * Quoting Ard: Veneers only set PC using a
 				 * PC+immediate LDR, and so they don't affect
 				 * the state of the stack or the register file
 				 */
 				frame->pc = frame->lr;
+				frame->ex_frame = false;
 				return URC_OK;
 			}
-			pr_warn("unwind: Index not found %08lx\n", frame->pc);
+			pr_warn("unwind: Index not found %08lx\n", pc);
 		}
 		return -URC_FAILURE;
 	}
@@ -442,7 +445,7 @@ int unwind_frame(struct stackframe *frame)
 	if (idx->insn == 1)
 		/* can't unwind */
 		return -URC_FAILURE;
-	else if (frame->pc == prel31_to_addr(&idx->addr_offset)) {
+	else if (frame->ex_frame && pc == prel31_to_addr(&idx->addr_offset)) {
 		/*
 		 * Unwinding is tricky when we're halfway through the prologue,
 		 * since the stack frame that the unwinder expects may not be
@@ -451,9 +454,10 @@ int unwind_frame(struct stackframe *frame)
 		 * a function, we are still effectively in the stack frame of
 		 * the caller, and the unwind info has no relevance yet.
 		 */
-		if (frame->pc == frame->lr)
+		if (pc == frame->lr)
 			return -URC_FAILURE;
 		frame->pc = frame->lr;
+		frame->ex_frame = false;
 		return URC_OK;
 	} else if ((idx->insn & 0x80000000) == 0)
 		/* prel31 to the unwind table */
@@ -515,6 +519,7 @@ int unwind_frame(struct stackframe *frame)
 	frame->lr = ctrl.vrs[LR];
 	frame->pc = ctrl.vrs[PC];
 	frame->lr_addr = ctrl.lr_addr;
+	frame->ex_frame = false;
 
 	return URC_OK;
 }
@@ -544,6 +549,7 @@ void unwind_backtrace(struct pt_regs *regs, struct task_struct *tsk,
 		 */
 here:
 		frame.pc = (unsigned long)&&here;
+		frame.ex_frame = false;
 	} else {
 		/* task blocked in __switch_to */
 		frame.fp = thread_saved_fp(tsk);
@@ -554,11 +560,12 @@ void unwind_backtrace(struct pt_regs *regs, struct task_struct *tsk,
 		 */
 		frame.lr = 0;
 		frame.pc = thread_saved_pc(tsk);
+		frame.ex_frame = false;
 	}
 
 	while (1) {
 		int urc;
-		unsigned long where = frame.pc;
+		unsigned long where = frame.ex_frame ? frame.pc : frame.pc - 4;
 
 		urc = unwind_frame(&frame);
 		if (urc < 0)


  reply	other threads:[~2024-03-05 17:58 UTC|newest]

Thread overview: 42+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-04  1:39 Jiangfeng Xiao
2024-03-04 15:15 ` Jann Horn
2024-03-04 17:40   ` Kees Cook
2024-03-05  3:31     ` Jiangfeng Xiao
2024-03-05  9:32       ` Kees Cook
2024-03-05 11:38         ` Jiangfeng Xiao
2024-03-05 17:58           ` Josh Poimboeuf [this message]
2024-03-06  4:00             ` Jiangfeng Xiao
2024-03-06  9:52             ` Russell King (Oracle)
2024-03-06 16:02               ` Josh Poimboeuf
2024-03-09 14:58               ` David Laight
2024-03-18  4:01             ` Jiangfeng Xiao
2024-03-05  2:54   ` Jiangfeng Xiao
2024-03-05  3:12     ` Jiangfeng Xiao
2024-03-20  2:19 ` [PATCH] ARM: unwind: improve unwinders for noreturn case Jiangfeng Xiao
2024-03-20  2:46   ` Kees Cook
2024-03-20  3:30     ` Jiangfeng Xiao
2024-03-20  3:34       ` Matthew Wilcox
2024-03-20  3:46         ` Jiangfeng Xiao
2024-03-20  3:44 ` [PATCH v2] " Jiangfeng Xiao
2024-03-20  8:45   ` Russell King (Oracle)
2024-03-20 15:30     ` Jiangfeng Xiao
2024-03-20 19:40       ` Russell King (Oracle)
2024-03-21  9:44         ` Jiangfeng Xiao
2024-03-21 10:22           ` David Laight
2024-03-21 11:23             ` Russell King (Oracle)
2024-03-21 12:07               ` David Laight
2024-03-21 12:22                 ` Russell King (Oracle)
2024-03-21 12:57                   ` David Laight
2024-03-21 13:08                     ` Russell King (Oracle)
2024-03-21 14:37                       ` David Laight
2024-03-21 14:56                         ` Russell King (Oracle)
2024-03-21 15:20                           ` David Laight
2024-03-21 15:33                             ` Russell King (Oracle)
2024-03-21 22:43               ` Ard Biesheuvel
2024-03-22  0:08                 ` Russell King (Oracle)
2024-03-22  9:24                   ` David Laight
2024-03-22  9:52                     ` Russell King (Oracle)
2024-03-22 12:54                       ` Jiangfeng Xiao
2024-03-22 14:16                       ` David Laight
2024-03-20 15:41 ` [PATCH v3] " Jiangfeng Xiao
2024-03-20 19:42   ` Russell King (Oracle)

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240305175846.qnyiru7uaa7itqba@treble \
    --to=jpoimboe@kernel.org \
    --cc=akpm@linux-foundation.org \
    --cc=ardb@kernel.org \
    --cc=dave.hansen@linux.intel.com \
    --cc=douzhaolei@huawei.com \
    --cc=gustavoars@kernel.org \
    --cc=jannh@google.com \
    --cc=keescook@chromium.org \
    --cc=kepler.chenxin@huawei.com \
    --cc=kirill.shutemov@linux.intel.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-hardening@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux@armlinux.org.uk \
    --cc=nixiaoming@huawei.com \
    --cc=peterz@infradead.org \
    --cc=wangbing6@huawei.com \
    --cc=wangfangpeng1@huawei.com \
    --cc=xiaojiangfeng@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox