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 --- fs/kernel_read_file.c | 180 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 180 insertions(+) create mode 100644 fs/kernel_read_file.c (limited to 'fs/kernel_read_file.c') diff --git a/fs/kernel_read_file.c b/fs/kernel_read_file.c new file mode 100644 index 000000000..9c2ba9240 --- /dev/null +++ b/fs/kernel_read_file.c @@ -0,0 +1,180 @@ +// SPDX-License-Identifier: GPL-2.0-only +#include +#include +#include +#include +#include + +/** + * kernel_read_file() - read file contents into a kernel buffer + * + * @file: file to read from + * @offset: where to start reading from (see below). + * @buf: pointer to a "void *" buffer for reading into (if + * *@buf is NULL, a buffer will be allocated, and + * @buf_size will be ignored) + * @buf_size: size of buf, if already allocated. If @buf not + * allocated, this is the largest size to allocate. + * @file_size: if non-NULL, the full size of @file will be + * written here. + * @id: the kernel_read_file_id identifying the type of + * file contents being read (for LSMs to examine) + * + * @offset must be 0 unless both @buf and @file_size are non-NULL + * (i.e. the caller must be expecting to read partial file contents + * via an already-allocated @buf, in at most @buf_size chunks, and + * will be able to determine when the entire file was read by + * checking @file_size). This isn't a recommended way to read a + * file, though, since it is possible that the contents might + * change between calls to kernel_read_file(). + * + * Returns number of bytes read (no single read will be bigger + * than SSIZE_MAX), or negative on error. + * + */ +ssize_t kernel_read_file(struct file *file, loff_t offset, void **buf, + size_t buf_size, size_t *file_size, + enum kernel_read_file_id id) +{ + loff_t i_size, pos; + ssize_t copied; + void *allocated = NULL; + bool whole_file; + int ret; + + if (offset != 0 && (!*buf || !file_size)) + return -EINVAL; + + if (!S_ISREG(file_inode(file)->i_mode)) + return -EINVAL; + + ret = deny_write_access(file); + if (ret) + return ret; + + i_size = i_size_read(file_inode(file)); + if (i_size <= 0) { + ret = -EINVAL; + goto out; + } + /* The file is too big for sane activities. */ + if (i_size > SSIZE_MAX) { + ret = -EFBIG; + goto out; + } + /* The entire file cannot be read in one buffer. */ + if (!file_size && offset == 0 && i_size > buf_size) { + ret = -EFBIG; + goto out; + } + + whole_file = (offset == 0 && i_size <= buf_size); + ret = security_kernel_read_file(file, id, whole_file); + if (ret) + goto out; + + if (file_size) + *file_size = i_size; + + if (!*buf) + *buf = allocated = vmalloc(i_size); + if (!*buf) { + ret = -ENOMEM; + goto out; + } + + pos = offset; + copied = 0; + while (copied < buf_size) { + ssize_t bytes; + size_t wanted = min_t(size_t, buf_size - copied, + i_size - pos); + + bytes = kernel_read(file, *buf + copied, wanted, &pos); + if (bytes < 0) { + ret = bytes; + goto out_free; + } + + if (bytes == 0) + break; + copied += bytes; + } + + if (whole_file) { + if (pos != i_size) { + ret = -EIO; + goto out_free; + } + + ret = security_kernel_post_read_file(file, *buf, i_size, id); + } + +out_free: + if (ret < 0) { + if (allocated) { + vfree(*buf); + *buf = NULL; + } + } + +out: + allow_write_access(file); + return ret == 0 ? copied : ret; +} +EXPORT_SYMBOL_GPL(kernel_read_file); + +ssize_t kernel_read_file_from_path(const char *path, loff_t offset, void **buf, + size_t buf_size, size_t *file_size, + enum kernel_read_file_id id) +{ + struct file *file; + ssize_t ret; + + if (!path || !*path) + return -EINVAL; + + file = filp_open(path, O_RDONLY, 0); + if (IS_ERR(file)) + return PTR_ERR(file); + + ret = kernel_read_file(file, offset, buf, buf_size, file_size, id); + fput(file); + return ret; +} +EXPORT_SYMBOL_GPL(kernel_read_file_from_path); + +ssize_t kernel_read_file_from_path_initns(const char *path, loff_t offset, + void **buf, size_t buf_size, + size_t *file_size, + enum kernel_read_file_id id) +{ + struct file *file; + ssize_t ret; + + if (!path || !*path) + return -EINVAL; + + scoped_with_init_fs() + file = filp_open(path, O_RDONLY, 0); + if (IS_ERR(file)) + return PTR_ERR(file); + + ret = kernel_read_file(file, offset, buf, buf_size, file_size, id); + fput(file); + return ret; +} +EXPORT_SYMBOL_GPL(kernel_read_file_from_path_initns); + +ssize_t kernel_read_file_from_fd(int fd, loff_t offset, void **buf, + size_t buf_size, size_t *file_size, + enum kernel_read_file_id id) +{ + CLASS(fd, f)(fd); + + if (fd_empty(f) || !(fd_file(f)->f_mode & FMODE_READ)) + return -EBADF; + + return kernel_read_file(fd_file(f), offset, buf, buf_size, file_size, id); +} +EXPORT_SYMBOL_GPL(kernel_read_file_from_fd); -- cgit v1.3.1