Re: [RFC PATCH v3 2/4] arm64: Check the return PC against unreliable code sections
From: Madhavan T. Venkataraman <hidden>
Date: 2021-05-05 20:00:49
Also in:
live-patching, lkml
OK. I will make all the changes you suggested. Thanks! Madhavan On 5/5/21 2:30 PM, Ard Biesheuvel wrote:
On Mon, 3 May 2021 at 19:38, [off-list ref] wrote:quoted
From: "Madhavan T. Venkataraman" <redacted> Create a sym_code_ranges[] array to cover the following text sections that contain functions defined as SYM_CODE_*(). These functions are low-level functions (and do not have a proper frame pointer prolog and epilog). So, they are inherently unreliable from a stack unwinding perspective. .entry.text .idmap.text .hyp.idmap.text .hyp.text .hibernate_exit.text .entry.tramp.text If a return PC falls in any of these, mark the stack trace unreliable. The only exception to this is - if the unwinder has reached the last frame already, it will not mark the stack trace unreliable since there is no more unwinding to do. E.g., - ret_from_fork() occurs at the end of the stack trace of kernel tasks. - el0_*() functions occur at the end of EL0 exception stack traces. This covers all user task entries into the kernel. NOTE: - EL1 exception handlers are in .entry.text. So, stack traces that contain those functions will be marked not reliable. This covers interrupts, exceptions and breakpoints encountered while executing in the kernel. - At the end of an interrupt, the kernel can preempt the current task if required. So, the stack traces of all preempted tasks will show the interrupt frame and will be considered unreliable. Signed-off-by: Madhavan T. Venkataraman <redacted> --- arch/arm64/kernel/stacktrace.c | 54 ++++++++++++++++++++++++++++++++++ 1 file changed, 54 insertions(+)diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c index c21a1bca28f3..1ff14615a55a 100644 --- a/arch/arm64/kernel/stacktrace.c +++ b/arch/arm64/kernel/stacktrace.c@@ -15,9 +15,48 @@ #include <asm/irq.h> #include <asm/pointer_auth.h> +#include <asm/sections.h> #include <asm/stack_pointer.h> #include <asm/stacktrace.h> +struct code_range { + unsigned long start; + unsigned long end; +}; + +struct code_range sym_code_ranges[] =This should be static and constquoted
+{ + /* non-unwindable ranges */ + { (unsigned long)__entry_text_start, + (unsigned long)__entry_text_end }, + { (unsigned long)__idmap_text_start, + (unsigned long)__idmap_text_end }, + { (unsigned long)__hyp_idmap_text_start, + (unsigned long)__hyp_idmap_text_end }, + { (unsigned long)__hyp_text_start, + (unsigned long)__hyp_text_end }, +#ifdef CONFIG_HIBERNATION + { (unsigned long)__hibernate_exit_text_start, + (unsigned long)__hibernate_exit_text_end }, +#endif +#ifdef CONFIG_UNMAP_KERNEL_AT_EL0 + { (unsigned long)__entry_tramp_text_start, + (unsigned long)__entry_tramp_text_end }, +#endif + { /* sentinel */ } +}; + +static struct code_range *lookup_range(unsigned long pc)const struct code_range *quoted
+{ + struct code_range *range;const struct code_range *quoted
+ + for (range = sym_code_ranges; range->start; range++) { + if (pc >= range->start && pc < range->end) + return range; + } + return range; +} + /* * AArch64 PCS assigns the frame pointer to x29. *@@ -43,6 +82,7 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame) { unsigned long fp = frame->fp; struct stack_info info; + struct code_range *range;const struct code_range *quoted
frame->reliable = true;@@ -103,6 +143,8 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame) return 0; } + range = lookup_range(frame->pc); + #ifdef CONFIG_FUNCTION_GRAPH_TRACER if (tsk->ret_stack && frame->pc == (unsigned long)return_to_handler) {@@ -118,9 +160,21 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame) return -EINVAL; frame->pc = ret_stack->ret; frame->pc = ptrauth_strip_insn_pac(frame->pc); + return 0; } #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ + if (!range->start) + return 0; + + /* + * The return PC falls in an unreliable function. If the final frame + * has been reached, no more unwinding is needed. Otherwise, mark the + * stack trace not reliable. + */ + if (frame->fp) + frame->reliable = false; + return 0; } NOKPROBE_SYMBOL(unwind_frame); --2.25.1 _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
_______________________________________________ linux-arm-kernel mailing list linux-arm-kernel@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel