@@ -25,7 +25,6 @@ DECLARE_DO_ERROR_INFO(do_trap_ecall_s);
DECLARE_DO_ERROR_INFO(do_trap_ecall_m);
DECLARE_DO_ERROR_INFO(do_trap_break);
-asmlinkage unsigned long get_overflow_stack(void);
asmlinkage void handle_bad_stack(struct pt_regs *regs);
#endif /* _ASM_RISCV_PROTOTYPES_H */
@@ -28,14 +28,11 @@
#define THREAD_SHIFT (PAGE_SHIFT + THREAD_SIZE_ORDER)
#define OVERFLOW_STACK_SIZE SZ_4K
-#define SHADOW_OVERFLOW_STACK_SIZE (1024)
#define IRQ_STACK_SIZE THREAD_SIZE
#ifndef __ASSEMBLY__
-extern long shadow_stack[SHADOW_OVERFLOW_STACK_SIZE / sizeof(long)];
-
#include <asm/processor.h>
#include <asm/csr.h>
@@ -37,6 +37,7 @@ void asm_offsets(void)
OFFSET(TASK_TI_PREEMPT_COUNT, task_struct, thread_info.preempt_count);
OFFSET(TASK_TI_KERNEL_SP, task_struct, thread_info.kernel_sp);
OFFSET(TASK_TI_USER_SP, task_struct, thread_info.user_sp);
+ OFFSET(TASK_TI_CPU, task_struct, thread_info.cpu);
OFFSET(TASK_THREAD_F0, task_struct, thread.fstate.f[0]);
OFFSET(TASK_THREAD_F1, task_struct, thread.fstate.f[1]);
@@ -223,54 +223,16 @@ END(ret_from_exception)
#ifdef CONFIG_VMAP_STACK
ENTRY(handle_kernel_stack_overflow)
- la sp, shadow_stack
- addi sp, sp, SHADOW_OVERFLOW_STACK_SIZE
-
- //save caller register to shadow stack
- addi sp, sp, -(PT_SIZE_ON_STACK)
- REG_S x1, PT_RA(sp)
- REG_S x5, PT_T0(sp)
- REG_S x6, PT_T1(sp)
- REG_S x7, PT_T2(sp)
- REG_S x10, PT_A0(sp)
- REG_S x11, PT_A1(sp)
- REG_S x12, PT_A2(sp)
- REG_S x13, PT_A3(sp)
- REG_S x14, PT_A4(sp)
- REG_S x15, PT_A5(sp)
- REG_S x16, PT_A6(sp)
- REG_S x17, PT_A7(sp)
- REG_S x28, PT_T3(sp)
- REG_S x29, PT_T4(sp)
- REG_S x30, PT_T5(sp)
- REG_S x31, PT_T6(sp)
-
- la ra, restore_caller_reg
- tail get_overflow_stack
-
-restore_caller_reg:
- //save per-cpu overflow stack
- REG_S a0, -8(sp)
- //restore caller register from shadow_stack
- REG_L x1, PT_RA(sp)
- REG_L x5, PT_T0(sp)
- REG_L x6, PT_T1(sp)
- REG_L x7, PT_T2(sp)
- REG_L x10, PT_A0(sp)
- REG_L x11, PT_A1(sp)
- REG_L x12, PT_A2(sp)
- REG_L x13, PT_A3(sp)
- REG_L x14, PT_A4(sp)
- REG_L x15, PT_A5(sp)
- REG_L x16, PT_A6(sp)
- REG_L x17, PT_A7(sp)
- REG_L x28, PT_T3(sp)
- REG_L x29, PT_T4(sp)
- REG_L x30, PT_T5(sp)
- REG_L x31, PT_T6(sp)
+ la sp, overflow_stack
+ /* use tp as tmp register since we can restore it from CSR_SCRATCH */
+ REG_L tp, TASK_TI_CPU(tp)
+ slli tp, tp, RISCV_LGPTR
+ add tp, sp, tp
+ REG_L sp, 0(tp)
+
+ /* restore tp */
+ csrr tp, CSR_SCRATCH
- //load per-cpu overflow stack
- REG_L sp, -8(sp)
addi sp, sp, -(PT_SIZE_ON_STACK)
//save context to overflow stack
@@ -245,23 +245,12 @@ int is_valid_bugaddr(unsigned long pc)
#endif /* CONFIG_GENERIC_BUG */
#ifdef CONFIG_VMAP_STACK
-static DEFINE_PER_CPU(unsigned long [OVERFLOW_STACK_SIZE/sizeof(long)],
- overflow_stack)__aligned(16);
-/*
- * shadow stack, handled_ kernel_ stack_ overflow(in kernel/entry.S) is used
- * to get per-cpu overflow stack(get_overflow_stack).
- */
-long shadow_stack[SHADOW_OVERFLOW_STACK_SIZE/sizeof(long)];
-asmlinkage unsigned long get_overflow_stack(void)
-{
- return (unsigned long)this_cpu_ptr(overflow_stack) +
- OVERFLOW_STACK_SIZE;
-}
+u8 *overflow_stack[NR_CPUS] __ro_after_init __aligned(16);
asmlinkage void handle_bad_stack(struct pt_regs *regs)
{
unsigned long tsk_stk = (unsigned long)current->stack;
- unsigned long ovf_stk = (unsigned long)this_cpu_ptr(overflow_stack);
+ unsigned long ovf_stk = (unsigned long)overflow_stack[raw_smp_processor_id()];
console_verbose();
@@ -269,7 +258,7 @@ asmlinkage void handle_bad_stack(struct pt_regs *regs)
pr_emerg("Task stack: [0x%016lx..0x%016lx]\n",
tsk_stk, tsk_stk + THREAD_SIZE);
pr_emerg("Overflow stack: [0x%016lx..0x%016lx]\n",
- ovf_stk, ovf_stk + OVERFLOW_STACK_SIZE);
+ ovf_stk - OVERFLOW_STACK_SIZE, ovf_stk);
__show_regs(regs);
panic("Kernel stack overflow");
@@ -277,4 +266,19 @@ asmlinkage void handle_bad_stack(struct pt_regs *regs)
for (;;)
wait_for_interrupt();
}
+
+static int __init alloc_overflow_stacks(void)
+{
+ u8 *s;
+ int cpu;
+
+ for_each_possible_cpu(cpu) {
+ s = (u8 *)__get_free_pages(GFP_KERNEL, get_order(OVERFLOW_STACK_SIZE));
+ if (WARN_ON(!s))
+ return -ENOMEM;
+ overflow_stack[cpu] = &s[OVERFLOW_STACK_SIZE];
+ }
+ return 0;
+}
+early_initcall(alloc_overflow_stacks);
#endif
Currently, when detecting vmap stack overflow, riscv firstly switches to the so called shadow stack, then use this shadow stack to call the get_overflow_stack() to get the overflow stack. However, there's a race here if two or more harts use the same shadow stack at the same time. To solve this race, we rely on two facts: 1. the content of kernel thread pointer I.E "tp" register can still be gotten from the CSR_SCRATCH register, thus we can clobber tp under the condition that we restore tp from CSR_SCRATCH later. 2. Once vmap stack overflow happen, panic is coming soon, no performance concern at all, so we don't need to define the overflow stack as percpu var, we can simplify it into a pointer array which points to allocated pages. Thus we can use tp as a tmp register to get the cpu id to calculate the offset of overflow stack pointer array for each cpu w/o shadow stack any more. Thus the race condition is removed as a side effect. NOTE: we can use similar mechanism to let each cpu use different shadow stack to fix the race codition, but if we can remove shadow stack usage totally, why not. Signed-off-by: Jisheng Zhang <jszhang@kernel.org> Fixes: 31da94c25aea ("riscv: add VMAP_STACK overflow detection") --- arch/riscv/include/asm/asm-prototypes.h | 1 - arch/riscv/include/asm/thread_info.h | 3 -- arch/riscv/kernel/asm-offsets.c | 1 + arch/riscv/kernel/entry.S | 56 ++++--------------------- arch/riscv/kernel/traps.c | 32 +++++++------- 5 files changed, 28 insertions(+), 65 deletions(-)