Fix coroutine implementaion on Windows-Arm64

When setjmp/longjmp/exceptions are used on Windows it's necessary to store+restore additional information from the TEB.
I didn't find any official documentation about the values to be saved, but found the corresponding boost/context implemenataion:
    abf8e04e23

This is similar to the special TIB handling on x86/x86_64 on Windows.

Without this fix an exception in a fiber segfaults without any output:
  ruby -e "Fiber.new{ raise 'test' }.resume"
This commit is contained in:
Lars Kanis 2024-12-16 21:02:50 +01:00 committed by Nobuyoshi Nakada
parent 11e120dfce
commit 187b8fdb69
Notes: git 2024-12-17 00:46:44 +00:00
2 changed files with 72 additions and 24 deletions

View File

@ -21,6 +21,13 @@
# error "-mbranch-protection flag specified b-key but Context.S does not support this"
#endif
#if defined(_WIN32)
## Add more space for certain TEB values on each stack
#define TEB_OFFSET 0x20
#else
#define TEB_OFFSET 0x00
#endif
## NOTE(PAC): Use we HINT mnemonics instead of PAC mnemonics to
## keep compatibility with those assemblers that don't support PAC.
##
@ -39,19 +46,34 @@ PREFIXED_SYMBOL(coroutine_transfer):
hint #34
#endif
# Make space on the stack for caller registers
sub sp, sp, 0xa0
sub sp, sp, 0xa0 + TEB_OFFSET
# Save caller registers
stp d8, d9, [sp, 0x00]
stp d10, d11, [sp, 0x10]
stp d12, d13, [sp, 0x20]
stp d14, d15, [sp, 0x30]
stp x19, x20, [sp, 0x40]
stp x21, x22, [sp, 0x50]
stp x23, x24, [sp, 0x60]
stp x25, x26, [sp, 0x70]
stp x27, x28, [sp, 0x80]
stp x29, x30, [sp, 0x90]
stp d8, d9, [sp, 0x00 + TEB_OFFSET]
stp d10, d11, [sp, 0x10 + TEB_OFFSET]
stp d12, d13, [sp, 0x20 + TEB_OFFSET]
stp d14, d15, [sp, 0x30 + TEB_OFFSET]
stp x19, x20, [sp, 0x40 + TEB_OFFSET]
stp x21, x22, [sp, 0x50 + TEB_OFFSET]
stp x23, x24, [sp, 0x60 + TEB_OFFSET]
stp x25, x26, [sp, 0x70 + TEB_OFFSET]
stp x27, x28, [sp, 0x80 + TEB_OFFSET]
stp x29, x30, [sp, 0x90 + TEB_OFFSET]
#if defined(_WIN32)
# Save certain values from Thread Environment Block (TEB)
# x18 points to the TEB on Windows
# Read TeStackBase and TeStackLimit at ksarm64.h from TEB
ldp x5, x6, [x18, #0x08]
# Save them
stp x5, x6, [sp, #0x00]
# Read TeDeallocationStack at ksarm64.h from TEB
ldr x5, [x18, #0x1478]
# Read TeFiberData at ksarm64.h from TEB
ldr x6, [x18, #0x20]
# Save current fiber data and deallocation stack
stp x5, x6, [sp, #0x10]
#endif
# Save stack pointer to x0 (first argument)
mov x2, sp
@ -61,20 +83,33 @@ PREFIXED_SYMBOL(coroutine_transfer):
ldr x3, [x1, 0]
mov sp, x3
#if defined(_WIN32)
# Restore stack base and limit
ldp x5, x6, [sp, #0x00]
# Write TeStackBase and TeStackLimit at ksarm64.h to TEB
stp x5, x6, [x18, #0x08]
# Restore fiber data and deallocation stack
ldp x5, x6, [sp, #0x10]
# Write TeDeallocationStack at ksarm64.h to TEB
str x5, [x18, #0x1478]
# Write TeFiberData at ksarm64.h to TEB
str x6, [x18, #0x20]
#endif
# Restore caller registers
ldp d8, d9, [sp, 0x00]
ldp d10, d11, [sp, 0x10]
ldp d12, d13, [sp, 0x20]
ldp d14, d15, [sp, 0x30]
ldp x19, x20, [sp, 0x40]
ldp x21, x22, [sp, 0x50]
ldp x23, x24, [sp, 0x60]
ldp x25, x26, [sp, 0x70]
ldp x27, x28, [sp, 0x80]
ldp x29, x30, [sp, 0x90]
ldp d8, d9, [sp, 0x00 + TEB_OFFSET]
ldp d10, d11, [sp, 0x10 + TEB_OFFSET]
ldp d12, d13, [sp, 0x20 + TEB_OFFSET]
ldp d14, d15, [sp, 0x30 + TEB_OFFSET]
ldp x19, x20, [sp, 0x40 + TEB_OFFSET]
ldp x21, x22, [sp, 0x50 + TEB_OFFSET]
ldp x23, x24, [sp, 0x60 + TEB_OFFSET]
ldp x25, x26, [sp, 0x70 + TEB_OFFSET]
ldp x27, x28, [sp, 0x80 + TEB_OFFSET]
ldp x29, x30, [sp, 0x90 + TEB_OFFSET]
# Pop stack frame
add sp, sp, 0xa0
add sp, sp, 0xa0 + TEB_OFFSET
#if defined(__ARM_FEATURE_PAC_DEFAULT) && (__ARM_FEATURE_PAC_DEFAULT != 0)
# autiasp: Authenticate x30 (LR) with SP and key A

View File

@ -17,7 +17,13 @@
#define COROUTINE __attribute__((noreturn)) void
enum {COROUTINE_REGISTERS = 0xa0 / 8};
#if defined(_WIN32)
#define TEB_OFFSET 0x20
#else
#define TEB_OFFSET 0x00
#endif
enum {COROUTINE_REGISTERS = (0xa0 + TEB_OFFSET) / 8};
#if defined(__SANITIZE_ADDRESS__)
#define COROUTINE_SANITIZE_ADDRESS
@ -87,7 +93,14 @@ static inline void coroutine_initialize(
memset(context->stack_pointer, 0, sizeof(void*) * COROUTINE_REGISTERS);
void *addr = (void*)(uintptr_t)start;
context->stack_pointer[0x98 / 8] = ptrauth_sign_instruction_addr(addr, (void*)top);
context->stack_pointer[(0x98 + TEB_OFFSET) / 8] = ptrauth_sign_instruction_addr(addr, (void*)top);
#if defined(_WIN32)
// save top address of stack as base in TEB
context->stack_pointer[0x00 / 8] = (char*)stack + size;
// save botton address of stack as limit and deallocation stack in TEB
context->stack_pointer[0x08 / 8] = stack;
context->stack_pointer[0x10 / 8] = stack;
#endif
}
struct coroutine_context * coroutine_transfer(struct coroutine_context * current, struct coroutine_context * target);