diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2026-05-12-16-47-23.gh-issue-139808.iIs7_E.rst b/Misc/NEWS.d/next/Core_and_Builtins/2026-05-12-16-47-23.gh-issue-139808.iIs7_E.rst new file mode 100644 index 00000000000000..3e9d930bf1de89 --- /dev/null +++ b/Misc/NEWS.d/next/Core_and_Builtins/2026-05-12-16-47-23.gh-issue-139808.iIs7_E.rst @@ -0,0 +1,2 @@ +Add branch protections for AArch64 (BTI/PAC) in assembly code used by +:option:`-X perf_jit <-X>` (Linux perf profiler integration). diff --git a/Python/asm_trampoline.S b/Python/asm_trampoline.S index 93adae3d99038f..9f3ca909ab7d85 100644 --- a/Python/asm_trampoline.S +++ b/Python/asm_trampoline.S @@ -1,3 +1,5 @@ +#include "asm_trampoline_aarch64.h" + .text #if defined(__APPLE__) .globl __Py_trampoline_func_start @@ -29,10 +31,12 @@ _Py_trampoline_func_start: #if defined(__aarch64__) && defined(__AARCH64EL__) && !defined(__ILP32__) // ARM64 little endian, 64bit ABI // generate with aarch64-linux-gnu-gcc 12.1 + SIGN_LR stp x29, x30, [sp, -16]! mov x29, sp blr x3 ldp x29, x30, [sp], 16 + VERIFY_LR ret #endif #ifdef __riscv diff --git a/Python/asm_trampoline_aarch64.h b/Python/asm_trampoline_aarch64.h new file mode 100644 index 00000000000000..bc83aa460b6860 --- /dev/null +++ b/Python/asm_trampoline_aarch64.h @@ -0,0 +1,56 @@ +#ifndef ASM_TRAMPOLINE_AARCH_64_H_ +#define ASM_TRAMPOLINE_AARCH_64_H_ + +/* + * References: + * - https://developer.arm.com/documentation/101028/0012/5--Feature-test-macros + * - https://github.com/ARM-software/abi-aa/blob/main/aaelf64/aaelf64.rst + */ + +#if defined(__ARM_FEATURE_BTI_DEFAULT) && __ARM_FEATURE_BTI_DEFAULT == 1 + #define BTI_J hint 36 /* bti j: for jumps, IE br instructions */ + #define BTI_C hint 34 /* bti c: for calls, IE bl instructions */ + #define GNU_PROPERTY_AARCH64_BTI 1 /* bit 0 GNU Notes is for BTI support */ +#else + #define BTI_J + #define BTI_C + #define GNU_PROPERTY_AARCH64_BTI 0 +#endif + +#if defined(__ARM_FEATURE_PAC_DEFAULT) + #if __ARM_FEATURE_PAC_DEFAULT & 1 + #define SIGN_LR hint 25 /* paciasp: sign with the A key */ + #define VERIFY_LR hint 29 /* autiasp: verify with the A key */ + #elif __ARM_FEATURE_PAC_DEFAULT & 2 + #define SIGN_LR hint 27 /* pacibsp: sign with the b key */ + #define VERIFY_LR hint 31 /* autibsp: verify with the b key */ + #endif + #define GNU_PROPERTY_AARCH64_POINTER_AUTH 2 /* bit 1 GNU Notes is for PAC support */ +#else + #define SIGN_LR BTI_C + #define VERIFY_LR + #define GNU_PROPERTY_AARCH64_POINTER_AUTH 0 +#endif + +#if defined(__ARM_FEATURE_GCS_DEFAULT) && __ARM_FEATURE_GCS_DEFAULT == 1 + #define GNU_PROPERTY_AARCH64_GCS 4 /* bit 2 GNU Notes is for GCS support */ +#else + #define GNU_PROPERTY_AARCH64_GCS 0 +#endif + +/* Add the BTI, PAC and GCS support to GNU Notes section */ +#if GNU_PROPERTY_AARCH64_BTI != 0 || GNU_PROPERTY_AARCH64_POINTER_AUTH != 0 || GNU_PROPERTY_AARCH64_GCS != 0 + .pushsection .note.gnu.property, "a"; /* Start a new allocatable section */ + .balign 8; /* align it on a byte boundry */ + .long 4; /* size of "GNU\0" */ + .long 0x10; /* size of descriptor */ + .long 0x5; /* NT_GNU_PROPERTY_TYPE_0 */ + .asciz "GNU"; + .long 0xc0000000; /* GNU_PROPERTY_AARCH64_FEATURE_1_AND */ + .long 4; /* Four bytes of data */ + .long (GNU_PROPERTY_AARCH64_BTI|GNU_PROPERTY_AARCH64_POINTER_AUTH|GNU_PROPERTY_AARCH64_GCS); /* BTI, PAC or GCS is enabled */ + .long 0; /* padding for 8 byte alignment */ + .popsection; /* end the section */ +#endif + +#endif diff --git a/Python/jit_unwind.c b/Python/jit_unwind.c index 646106f0a9655c..0941ed593ff7d1 100644 --- a/Python/jit_unwind.c +++ b/Python/jit_unwind.c @@ -60,6 +60,9 @@ enum { DWRF_CFA_offset_extended_sf = 0x11, // Extended signed offset DWRF_CFA_advance_loc = 0x40, // Advance location counter DWRF_CFA_offset = 0x80, // Simple offset instruction +#if defined(__aarch64__) + DWRF_CFA_AARCH64_negate_ra_state = 0x2d, // Toggle return address signing state +#endif DWRF_CFA_restore = 0xc0 // Restore register }; @@ -562,6 +565,13 @@ static void elf_init_ehframe_perf(ELFObjectContext* ctx) { DWRF_UV(8); // New offset: SP + 8 #elif defined(__aarch64__) && defined(__AARCH64EL__) && !defined(__ILP32__) /* AArch64 calling convention unwinding rules */ +#if defined(__ARM_FEATURE_PAC_DEFAULT) || \ + (defined(__ARM_FEATURE_BTI_DEFAULT) && __ARM_FEATURE_BTI_DEFAULT == 1) + DWRF_U8(DWRF_CFA_advance_loc | 1); // Advance past SIGN_LR (4 bytes) +#endif +#if defined(__ARM_FEATURE_PAC_DEFAULT) + DWRF_U8(DWRF_CFA_AARCH64_negate_ra_state); // Saved LR is PAC-signed from here +#endif DWRF_U8(DWRF_CFA_advance_loc | 1); // Advance by 1 instruction (4 bytes) DWRF_U8(DWRF_CFA_def_cfa_offset); // CFA = SP + 16 DWRF_UV(16); // Stack pointer moved by 16 bytes @@ -570,6 +580,9 @@ static void elf_init_ehframe_perf(ELFObjectContext* ctx) { DWRF_U8(DWRF_CFA_offset | DWRF_REG_RA); // x30 (link register) saved DWRF_UV(1); // At CFA-8 (1 * 8 = 8 bytes from CFA) DWRF_U8(DWRF_CFA_advance_loc | 3); // Advance by 3 instructions (12 bytes) +#if defined(__ARM_FEATURE_PAC_DEFAULT) + DWRF_U8(DWRF_CFA_AARCH64_negate_ra_state); // LR is authenticated, no longer PAC-signed +#endif DWRF_U8(DWRF_CFA_def_cfa_register); // CFA = FP (x29) + 16 DWRF_UV(DWRF_REG_FP); DWRF_U8(DWRF_CFA_restore | DWRF_REG_RA); // Restore x30 - NO DWRF_UV() after this!