From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- lib/crypto/arm64/aes.h | 219 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 219 insertions(+) create mode 100644 lib/crypto/arm64/aes.h (limited to 'lib/crypto/arm64/aes.h') diff --git a/lib/crypto/arm64/aes.h b/lib/crypto/arm64/aes.h new file mode 100644 index 000000000..9e9e45a6f --- /dev/null +++ b/lib/crypto/arm64/aes.h @@ -0,0 +1,219 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * AES block cipher, optimized for ARM64 + * + * Copyright (C) 2013 - 2017 Linaro Ltd + * Copyright 2026 Google LLC + */ + +#include +#include +#include +#include + +static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_neon); +static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_aes); + +struct aes_block { + u8 b[AES_BLOCK_SIZE]; +}; + +asmlinkage void __aes_arm64_encrypt(const u32 rk[], u8 out[AES_BLOCK_SIZE], + const u8 in[AES_BLOCK_SIZE], int rounds); +asmlinkage void __aes_arm64_decrypt(const u32 inv_rk[], u8 out[AES_BLOCK_SIZE], + const u8 in[AES_BLOCK_SIZE], int rounds); +asmlinkage void __aes_ce_encrypt(const u32 rk[], u8 out[AES_BLOCK_SIZE], + const u8 in[AES_BLOCK_SIZE], int rounds); +asmlinkage void __aes_ce_decrypt(const u32 inv_rk[], u8 out[AES_BLOCK_SIZE], + const u8 in[AES_BLOCK_SIZE], int rounds); +asmlinkage u32 __aes_ce_sub(u32 l); +asmlinkage void __aes_ce_invert(struct aes_block *out, + const struct aes_block *in); +asmlinkage void neon_aes_mac_update(u8 const in[], u32 const rk[], int rounds, + size_t blocks, u8 dg[], int enc_before, + int enc_after); + +/* + * Expand an AES key using the crypto extensions if supported and usable or + * generic code otherwise. The expanded key format is compatible between the + * two cases. The outputs are @rndkeys (required) and @inv_rndkeys (optional). + */ +static void aes_expandkey_arm64(u32 rndkeys[], u32 *inv_rndkeys, + const u8 *in_key, int key_len, int nrounds) +{ + /* + * The AES key schedule round constants + */ + static u8 const rcon[] = { + 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1b, 0x36, + }; + + u32 kwords = key_len / sizeof(u32); + struct aes_block *key_enc, *key_dec; + int i, j; + + if (!static_branch_likely(&have_aes) || unlikely(!may_use_simd())) { + aes_expandkey_generic(rndkeys, inv_rndkeys, in_key, key_len); + return; + } + + for (i = 0; i < kwords; i++) + rndkeys[i] = get_unaligned_le32(&in_key[i * sizeof(u32)]); + + scoped_ksimd() { + for (i = 0; i < sizeof(rcon); i++) { + u32 *rki = &rndkeys[i * kwords]; + u32 *rko = rki + kwords; + + rko[0] = ror32(__aes_ce_sub(rki[kwords - 1]), 8) ^ + rcon[i] ^ rki[0]; + rko[1] = rko[0] ^ rki[1]; + rko[2] = rko[1] ^ rki[2]; + rko[3] = rko[2] ^ rki[3]; + + if (key_len == AES_KEYSIZE_192) { + if (i >= 7) + break; + rko[4] = rko[3] ^ rki[4]; + rko[5] = rko[4] ^ rki[5]; + } else if (key_len == AES_KEYSIZE_256) { + if (i >= 6) + break; + rko[4] = __aes_ce_sub(rko[3]) ^ rki[4]; + rko[5] = rko[4] ^ rki[5]; + rko[6] = rko[5] ^ rki[6]; + rko[7] = rko[6] ^ rki[7]; + } + } + + /* + * Generate the decryption keys for the Equivalent Inverse + * Cipher. This involves reversing the order of the round + * keys, and applying the Inverse Mix Columns transformation on + * all but the first and the last one. + */ + if (inv_rndkeys) { + key_enc = (struct aes_block *)rndkeys; + key_dec = (struct aes_block *)inv_rndkeys; + j = nrounds; + + key_dec[0] = key_enc[j]; + for (i = 1, j--; j > 0; i++, j--) + __aes_ce_invert(key_dec + i, key_enc + j); + key_dec[i] = key_enc[0]; + } + } +} + +static void aes_preparekey_arch(union aes_enckey_arch *k, + union aes_invkey_arch *inv_k, + const u8 *in_key, int key_len, int nrounds) +{ + aes_expandkey_arm64(k->rndkeys, inv_k ? inv_k->inv_rndkeys : NULL, + in_key, key_len, nrounds); +} + +/* + * This is here temporarily until the remaining AES mode implementations are + * migrated from arch/arm64/crypto/ to lib/crypto/arm64/. + */ +int ce_aes_expandkey(struct crypto_aes_ctx *ctx, const u8 *in_key, + unsigned int key_len) +{ + if (aes_check_keylen(key_len) != 0) + return -EINVAL; + ctx->key_length = key_len; + aes_expandkey_arm64(ctx->key_enc, ctx->key_dec, in_key, key_len, + 6 + key_len / 4); + return 0; +} +EXPORT_SYMBOL(ce_aes_expandkey); + +EXPORT_SYMBOL_NS_GPL(neon_aes_ecb_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_ecb_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_cbc_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_cbc_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_cbc_cts_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_cbc_cts_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_ctr_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_xctr_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_xts_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_xts_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_essiv_cbc_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(neon_aes_essiv_cbc_decrypt, "CRYPTO_INTERNAL"); + +EXPORT_SYMBOL_NS_GPL(ce_aes_ecb_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_ecb_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_cbc_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_cbc_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_cbc_cts_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_cbc_cts_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_ctr_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_xctr_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_xts_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_xts_decrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_essiv_cbc_encrypt, "CRYPTO_INTERNAL"); +EXPORT_SYMBOL_NS_GPL(ce_aes_essiv_cbc_decrypt, "CRYPTO_INTERNAL"); +#if IS_MODULE(CONFIG_CRYPTO_AES_ARM64_CE_CCM) +EXPORT_SYMBOL_NS_GPL(ce_aes_mac_update, "CRYPTO_INTERNAL"); +#endif + +static void aes_encrypt_arch(const struct aes_enckey *key, + u8 out[AES_BLOCK_SIZE], + const u8 in[AES_BLOCK_SIZE]) +{ + if (static_branch_likely(&have_aes) && likely(may_use_simd())) { + scoped_ksimd() + __aes_ce_encrypt(key->k.rndkeys, out, in, key->nrounds); + } else { + __aes_arm64_encrypt(key->k.rndkeys, out, in, key->nrounds); + } +} + +static void aes_decrypt_arch(const struct aes_key *key, + u8 out[AES_BLOCK_SIZE], + const u8 in[AES_BLOCK_SIZE]) +{ + if (static_branch_likely(&have_aes) && likely(may_use_simd())) { + scoped_ksimd() + __aes_ce_decrypt(key->inv_k.inv_rndkeys, out, in, + key->nrounds); + } else { + __aes_arm64_decrypt(key->inv_k.inv_rndkeys, out, in, + key->nrounds); + } +} + +#if IS_ENABLED(CONFIG_CRYPTO_LIB_AES_CBC_MACS) +#define aes_cbcmac_blocks_arch aes_cbcmac_blocks_arch +static bool aes_cbcmac_blocks_arch(u8 h[AES_BLOCK_SIZE], + const struct aes_enckey *key, const u8 *data, + size_t nblocks, bool enc_before, + bool enc_after) +{ + if (static_branch_likely(&have_neon) && likely(may_use_simd())) { + scoped_ksimd() { + if (static_branch_likely(&have_aes)) + ce_aes_mac_update(data, key->k.rndkeys, + key->nrounds, nblocks, h, + enc_before, enc_after); + else + neon_aes_mac_update(data, key->k.rndkeys, + key->nrounds, nblocks, h, + enc_before, enc_after); + } + return true; + } + return false; +} +#endif /* CONFIG_CRYPTO_LIB_AES_CBC_MACS */ + +#define aes_mod_init_arch aes_mod_init_arch +static void aes_mod_init_arch(void) +{ + if (cpu_have_named_feature(ASIMD)) { + static_branch_enable(&have_neon); + if (cpu_have_named_feature(AES)) + static_branch_enable(&have_aes); + } +} -- cgit v1.3.1