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/mpi/mpicoder.c | 417 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 417 insertions(+) create mode 100644 lib/crypto/mpi/mpicoder.c (limited to 'lib/crypto/mpi/mpicoder.c') diff --git a/lib/crypto/mpi/mpicoder.c b/lib/crypto/mpi/mpicoder.c new file mode 100644 index 000000000..9359a58c2 --- /dev/null +++ b/lib/crypto/mpi/mpicoder.c @@ -0,0 +1,417 @@ +/* mpicoder.c - Coder for the external representation of MPIs + * Copyright (C) 1998, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA + */ + +#include +#include +#include +#include +#include +#include +#include "mpi-internal.h" + +#define MAX_EXTERN_MPI_BITS 16384 + +/** + * mpi_read_raw_data - Read a raw byte stream as a positive integer + * @xbuffer: The data to read + * @nbytes: The amount of data to read + */ +MPI mpi_read_raw_data(const void *xbuffer, size_t nbytes) +{ + const uint8_t *buffer = xbuffer; + int i, j; + unsigned nbits, nlimbs; + mpi_limb_t a; + MPI val = NULL; + + while (nbytes > 0 && buffer[0] == 0) { + buffer++; + nbytes--; + } + + nbits = nbytes * 8; + if (nbits > MAX_EXTERN_MPI_BITS) { + pr_info("MPI: mpi too large (%u bits)\n", nbits); + return NULL; + } + if (nbytes > 0) + nbits -= count_leading_zeros(buffer[0]) - (BITS_PER_LONG - 8); + + nlimbs = DIV_ROUND_UP(nbytes, BYTES_PER_MPI_LIMB); + val = mpi_alloc(nlimbs); + if (!val) + return NULL; + val->nbits = nbits; + val->sign = 0; + val->nlimbs = nlimbs; + + if (nbytes > 0) { + i = BYTES_PER_MPI_LIMB - nbytes % BYTES_PER_MPI_LIMB; + i %= BYTES_PER_MPI_LIMB; + for (j = nlimbs; j > 0; j--) { + a = 0; + for (; i < BYTES_PER_MPI_LIMB; i++) { + a <<= 8; + a |= *buffer++; + } + i = 0; + val->d[j - 1] = a; + } + } + return val; +} +EXPORT_SYMBOL_GPL(mpi_read_raw_data); + +MPI mpi_read_from_buffer(const void *xbuffer, unsigned *ret_nread) +{ + const uint8_t *buffer = xbuffer; + unsigned int nbits, nbytes; + MPI val; + + if (*ret_nread < 2) + return ERR_PTR(-EINVAL); + nbits = buffer[0] << 8 | buffer[1]; + + if (nbits > MAX_EXTERN_MPI_BITS) { + pr_info("MPI: mpi too large (%u bits)\n", nbits); + return ERR_PTR(-EINVAL); + } + + nbytes = DIV_ROUND_UP(nbits, 8); + if (nbytes + 2 > *ret_nread) { + pr_info("MPI: mpi larger than buffer nbytes=%u ret_nread=%u\n", + nbytes, *ret_nread); + return ERR_PTR(-EINVAL); + } + + val = mpi_read_raw_data(buffer + 2, nbytes); + if (!val) + return ERR_PTR(-ENOMEM); + + *ret_nread = nbytes + 2; + return val; +} +EXPORT_SYMBOL_GPL(mpi_read_from_buffer); + +static int count_lzeros(MPI a) +{ + mpi_limb_t alimb; + int i, lzeros = 0; + + for (i = a->nlimbs - 1; i >= 0; i--) { + alimb = a->d[i]; + if (alimb == 0) { + lzeros += sizeof(mpi_limb_t); + } else { + lzeros += count_leading_zeros(alimb) / 8; + break; + } + } + return lzeros; +} + +/** + * mpi_read_buffer() - read MPI to a buffer provided by user (msb first) + * + * @a: a multi precision integer + * @buf: buffer to which the output will be written to. Needs to be at + * least mpi_get_size(a) long. + * @buf_len: size of the buf. + * @nbytes: receives the actual length of the data written on success and + * the data to-be-written on -EOVERFLOW in case buf_len was too + * small. + * @sign: if not NULL, it will be set to the sign of a. + * + * Return: 0 on success or error code in case of error + */ +int mpi_read_buffer(MPI a, uint8_t *buf, unsigned buf_len, unsigned *nbytes, + int *sign) +{ + uint8_t *p; +#if BYTES_PER_MPI_LIMB == 4 + __be32 alimb; +#elif BYTES_PER_MPI_LIMB == 8 + __be64 alimb; +#else +#error please implement for this limb size. +#endif + unsigned int n = mpi_get_size(a); + int i, lzeros; + + if (!buf || !nbytes) + return -EINVAL; + + if (sign) + *sign = a->sign; + + lzeros = count_lzeros(a); + + if (buf_len < n - lzeros) { + *nbytes = n - lzeros; + return -EOVERFLOW; + } + + p = buf; + *nbytes = n - lzeros; + + for (i = a->nlimbs - 1 - lzeros / BYTES_PER_MPI_LIMB, + lzeros %= BYTES_PER_MPI_LIMB; + i >= 0; i--) { +#if BYTES_PER_MPI_LIMB == 4 + alimb = cpu_to_be32(a->d[i]); +#elif BYTES_PER_MPI_LIMB == 8 + alimb = cpu_to_be64(a->d[i]); +#else +#error please implement for this limb size. +#endif + memcpy(p, (u8 *)&alimb + lzeros, BYTES_PER_MPI_LIMB - lzeros); + p += BYTES_PER_MPI_LIMB - lzeros; + lzeros = 0; + } + return 0; +} +EXPORT_SYMBOL_GPL(mpi_read_buffer); + +/* + * mpi_get_buffer() - Returns an allocated buffer with the MPI (msb first). + * Caller must free the return string. + * This function does return a 0 byte buffer with nbytes set to zero if the + * value of A is zero. + * + * @a: a multi precision integer. + * @nbytes: receives the length of this buffer. + * @sign: if not NULL, it will be set to the sign of the a. + * + * Return: Pointer to MPI buffer or NULL on error + */ +void *mpi_get_buffer(MPI a, unsigned *nbytes, int *sign) +{ + uint8_t *buf; + unsigned int n; + int ret; + + if (!nbytes) + return NULL; + + n = mpi_get_size(a); + + if (!n) + n++; + + buf = kmalloc(n, GFP_KERNEL); + + if (!buf) + return NULL; + + ret = mpi_read_buffer(a, buf, n, nbytes, sign); + + if (ret) { + kfree(buf); + return NULL; + } + return buf; +} +EXPORT_SYMBOL_GPL(mpi_get_buffer); + +/** + * mpi_write_to_sgl() - Funnction exports MPI to an sgl (msb first) + * + * This function works in the same way as the mpi_read_buffer, but it + * takes an sgl instead of u8 * buf. + * + * @a: a multi precision integer + * @sgl: scatterlist to write to. Needs to be at least + * mpi_get_size(a) long. + * @nbytes: the number of bytes to write. Leading bytes will be + * filled with zero. + * @sign: if not NULL, it will be set to the sign of a. + * + * Return: 0 on success or error code in case of error + */ +int mpi_write_to_sgl(MPI a, struct scatterlist *sgl, unsigned nbytes, + int *sign) +{ + u8 *p, *p2; +#if BYTES_PER_MPI_LIMB == 4 + __be32 alimb; +#elif BYTES_PER_MPI_LIMB == 8 + __be64 alimb; +#else +#error please implement for this limb size. +#endif + unsigned int n = mpi_get_size(a); + struct sg_mapping_iter miter; + int i, x, buf_len; + int nents; + + if (sign) + *sign = a->sign; + + if (nbytes < n) + return -EOVERFLOW; + + nents = sg_nents_for_len(sgl, nbytes); + if (nents < 0) + return -EINVAL; + + sg_miter_start(&miter, sgl, nents, SG_MITER_ATOMIC | SG_MITER_TO_SG); + sg_miter_next(&miter); + buf_len = miter.length; + p2 = miter.addr; + + while (nbytes > n) { + i = min_t(unsigned, nbytes - n, buf_len); + memset(p2, 0, i); + p2 += i; + nbytes -= i; + + buf_len -= i; + if (!buf_len) { + sg_miter_next(&miter); + buf_len = miter.length; + p2 = miter.addr; + } + } + + for (i = a->nlimbs - 1; i >= 0; i--) { +#if BYTES_PER_MPI_LIMB == 4 + alimb = a->d[i] ? cpu_to_be32(a->d[i]) : 0; +#elif BYTES_PER_MPI_LIMB == 8 + alimb = a->d[i] ? cpu_to_be64(a->d[i]) : 0; +#else +#error please implement for this limb size. +#endif + p = (u8 *)&alimb; + + for (x = 0; x < sizeof(alimb); x++) { + *p2++ = *p++; + if (!--buf_len) { + sg_miter_next(&miter); + buf_len = miter.length; + p2 = miter.addr; + } + } + } + + sg_miter_stop(&miter); + return 0; +} +EXPORT_SYMBOL_GPL(mpi_write_to_sgl); + +/* + * mpi_read_raw_from_sgl() - Function allocates an MPI and populates it with + * data from the sgl + * + * This function works in the same way as the mpi_read_raw_data, but it + * takes an sgl instead of void * buffer. i.e. it allocates + * a new MPI and reads the content of the sgl to the MPI. + * + * @sgl: scatterlist to read from + * @nbytes: number of bytes to read + * + * Return: Pointer to a new MPI or NULL on error + */ +MPI mpi_read_raw_from_sgl(struct scatterlist *sgl, unsigned int nbytes) +{ + struct sg_mapping_iter miter; + unsigned int nbits, nlimbs; + int x, j, z, lzeros, ents; + unsigned int len; + const u8 *buff; + mpi_limb_t a; + MPI val = NULL; + + ents = sg_nents_for_len(sgl, nbytes); + if (ents < 0) + return NULL; + + sg_miter_start(&miter, sgl, ents, SG_MITER_ATOMIC | SG_MITER_FROM_SG); + + lzeros = 0; + len = 0; + while (nbytes > 0) { + while (len && !*buff && lzeros < nbytes) { + lzeros++; + len--; + buff++; + } + + if (len && *buff) + break; + + sg_miter_next(&miter); + buff = miter.addr; + len = miter.length; + + nbytes -= lzeros; + lzeros = 0; + } + + miter.consumed = lzeros; + + nbytes -= lzeros; + nbits = nbytes * 8; + if (nbits > MAX_EXTERN_MPI_BITS) { + sg_miter_stop(&miter); + pr_info("MPI: mpi too large (%u bits)\n", nbits); + return NULL; + } + + if (nbytes > 0) + nbits -= count_leading_zeros(*buff) - (BITS_PER_LONG - 8); + + sg_miter_stop(&miter); + + nlimbs = DIV_ROUND_UP(nbytes, BYTES_PER_MPI_LIMB); + val = mpi_alloc(nlimbs); + if (!val) + return NULL; + + val->nbits = nbits; + val->sign = 0; + val->nlimbs = nlimbs; + + if (nbytes == 0) + return val; + + j = nlimbs - 1; + a = 0; + z = BYTES_PER_MPI_LIMB - nbytes % BYTES_PER_MPI_LIMB; + z %= BYTES_PER_MPI_LIMB; + + while (sg_miter_next(&miter)) { + buff = miter.addr; + len = min(miter.length, nbytes); + nbytes -= len; + + for (x = 0; x < len; x++) { + a <<= 8; + a |= *buff++; + if (((z + x + 1) % BYTES_PER_MPI_LIMB) == 0) { + val->d[j--] = a; + a = 0; + } + } + z += x; + } + + return val; +} +EXPORT_SYMBOL_GPL(mpi_read_raw_from_sgl); -- cgit v1.3.1