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/hostfs/hostfs_user.c | 421 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 421 insertions(+) create mode 100644 fs/hostfs/hostfs_user.c (limited to 'fs/hostfs/hostfs_user.c') diff --git a/fs/hostfs/hostfs_user.c b/fs/hostfs/hostfs_user.c new file mode 100644 index 000000000..3bcd9f35e --- /dev/null +++ b/fs/hostfs/hostfs_user.c @@ -0,0 +1,421 @@ +/* + * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + * Licensed under the GPL + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "hostfs.h" +#include + +static void statx_to_hostfs(const struct statx *buf, struct hostfs_stat *p) +{ + p->ino = buf->stx_ino; + p->mode = buf->stx_mode; + p->nlink = buf->stx_nlink; + p->uid = buf->stx_uid; + p->gid = buf->stx_gid; + p->size = buf->stx_size; + p->atime.tv_sec = buf->stx_atime.tv_sec; + p->atime.tv_nsec = buf->stx_atime.tv_nsec; + p->ctime.tv_sec = buf->stx_ctime.tv_sec; + p->ctime.tv_nsec = buf->stx_ctime.tv_nsec; + p->mtime.tv_sec = buf->stx_mtime.tv_sec; + p->mtime.tv_nsec = buf->stx_mtime.tv_nsec; + if (buf->stx_mask & STATX_BTIME) { + p->btime.tv_sec = buf->stx_btime.tv_sec; + p->btime.tv_nsec = buf->stx_btime.tv_nsec; + } else { + memset(&p->btime, 0, sizeof(p->btime)); + } + p->blksize = buf->stx_blksize; + p->blocks = buf->stx_blocks; + p->rdev.maj = buf->stx_rdev_major; + p->rdev.min = buf->stx_rdev_minor; + p->dev.maj = buf->stx_dev_major; + p->dev.min = buf->stx_dev_minor; +} + +int stat_file(const char *path, struct hostfs_stat *p, int fd) +{ + struct statx buf; + int flags = AT_SYMLINK_NOFOLLOW; + + if (fd >= 0) { + flags |= AT_EMPTY_PATH; + path = ""; + } + + if ((statx(fd, path, flags, STATX_BASIC_STATS | STATX_BTIME, &buf)) < 0) + return -errno; + + statx_to_hostfs(&buf, p); + return 0; +} + +int access_file(char *path, int r, int w, int x) +{ + int mode = 0; + + if (r) + mode = R_OK; + if (w) + mode |= W_OK; + if (x) + mode |= X_OK; + if (access(path, mode) != 0) + return -errno; + else return 0; +} + +int open_file(char *path, int r, int w, int append) +{ + int mode = 0, fd; + + if (r && !w) + mode = O_RDONLY; + else if (!r && w) + mode = O_WRONLY; + else if (r && w) + mode = O_RDWR; + else panic("Impossible mode in open_file"); + + if (append) + mode |= O_APPEND; + fd = open64(path, mode); + if (fd < 0) + return -errno; + else return fd; +} + +void *open_dir(char *path, int *err_out) +{ + DIR *dir; + + dir = opendir(path); + *err_out = errno; + + return dir; +} + +void seek_dir(void *stream, unsigned long long pos) +{ + DIR *dir = stream; + + seekdir(dir, pos); +} + +char *read_dir(void *stream, unsigned long long *pos_out, + unsigned long long *ino_out, int *len_out, + unsigned int *type_out) +{ + DIR *dir = stream; + struct dirent *ent; + + ent = readdir(dir); + if (ent == NULL) + return NULL; + *len_out = strlen(ent->d_name); + *ino_out = ent->d_ino; + *type_out = ent->d_type; + *pos_out = ent->d_off; + return ent->d_name; +} + +int read_file(int fd, unsigned long long *offset, char *buf, int len) +{ + int n; + + n = pread64(fd, buf, len, *offset); + if (n < 0) + return -errno; + *offset += n; + return n; +} + +int write_file(int fd, unsigned long long *offset, const char *buf, int len) +{ + int n; + + n = pwrite64(fd, buf, len, *offset); + if (n < 0) + return -errno; + *offset += n; + return n; +} + +int lseek_file(int fd, long long offset, int whence) +{ + int ret; + + ret = lseek64(fd, offset, whence); + if (ret < 0) + return -errno; + return 0; +} + +int fsync_file(int fd, int datasync) +{ + int ret; + if (datasync) + ret = fdatasync(fd); + else + ret = fsync(fd); + + if (ret < 0) + return -errno; + return 0; +} + +int replace_file(int oldfd, int fd) +{ + return dup2(oldfd, fd); +} + +void close_file(void *stream) +{ + close(*((int *) stream)); +} + +void close_dir(void *stream) +{ + closedir(stream); +} + +int file_create(char *name, int mode) +{ + int fd; + + fd = open64(name, O_CREAT | O_RDWR, mode); + if (fd < 0) + return -errno; + return fd; +} + +int set_attr(const char *file, struct hostfs_iattr *attrs, int fd) +{ + struct hostfs_stat st; + struct timeval times[2]; + int err, ma; + + if (attrs->ia_valid & HOSTFS_ATTR_MODE) { + if (fd >= 0) { + if (fchmod(fd, attrs->ia_mode) != 0) + return -errno; + } else if (chmod(file, attrs->ia_mode) != 0) { + return -errno; + } + } + if (attrs->ia_valid & HOSTFS_ATTR_UID) { + if (fd >= 0) { + if (fchown(fd, attrs->ia_uid, -1)) + return -errno; + } else if (chown(file, attrs->ia_uid, -1)) { + return -errno; + } + } + if (attrs->ia_valid & HOSTFS_ATTR_GID) { + if (fd >= 0) { + if (fchown(fd, -1, attrs->ia_gid)) + return -errno; + } else if (chown(file, -1, attrs->ia_gid)) { + return -errno; + } + } + if (attrs->ia_valid & HOSTFS_ATTR_SIZE) { + if (fd >= 0) { + if (ftruncate(fd, attrs->ia_size)) + return -errno; + } else if (truncate(file, attrs->ia_size)) { + return -errno; + } + } + + /* + * Update accessed and/or modified time, in two parts: first set + * times according to the changes to perform, and then call futimes() + * or utimes() to apply them. + */ + ma = (HOSTFS_ATTR_ATIME_SET | HOSTFS_ATTR_MTIME_SET); + if (attrs->ia_valid & ma) { + err = stat_file(file, &st, fd); + if (err != 0) + return err; + + times[0].tv_sec = st.atime.tv_sec; + times[0].tv_usec = st.atime.tv_nsec / 1000; + times[1].tv_sec = st.mtime.tv_sec; + times[1].tv_usec = st.mtime.tv_nsec / 1000; + + if (attrs->ia_valid & HOSTFS_ATTR_ATIME_SET) { + times[0].tv_sec = attrs->ia_atime.tv_sec; + times[0].tv_usec = attrs->ia_atime.tv_nsec / 1000; + } + if (attrs->ia_valid & HOSTFS_ATTR_MTIME_SET) { + times[1].tv_sec = attrs->ia_mtime.tv_sec; + times[1].tv_usec = attrs->ia_mtime.tv_nsec / 1000; + } + + if (fd >= 0) { + if (futimes(fd, times) != 0) + return -errno; + } else if (utimes(file, times) != 0) { + return -errno; + } + } + + /* Note: ctime is not handled */ + if (attrs->ia_valid & (HOSTFS_ATTR_ATIME | HOSTFS_ATTR_MTIME)) { + err = stat_file(file, &st, fd); + attrs->ia_atime = st.atime; + attrs->ia_mtime = st.mtime; + if (err != 0) + return err; + } + return 0; +} + +int make_symlink(const char *from, const char *to) +{ + int err; + + err = symlink(to, from); + if (err) + return -errno; + return 0; +} + +int unlink_file(const char *file) +{ + int err; + + err = unlink(file); + if (err) + return -errno; + return 0; +} + +int do_mkdir(const char *file, int mode) +{ + int err; + + err = mkdir(file, mode); + if (err) + return -errno; + return 0; +} + +int hostfs_do_rmdir(const char *file) +{ + int err; + + err = rmdir(file); + if (err) + return -errno; + return 0; +} + +int do_mknod(const char *file, int mode, unsigned int major, unsigned int minor) +{ + int err; + + err = mknod(file, mode, os_makedev(major, minor)); + if (err) + return -errno; + return 0; +} + +int link_file(const char *to, const char *from) +{ + int err; + + err = link(to, from); + if (err) + return -errno; + return 0; +} + +int hostfs_do_readlink(char *file, char *buf, int size) +{ + int n; + + n = readlink(file, buf, size); + if (n < 0) + return -errno; + if (n < size) + buf[n] = '\0'; + return n; +} + +int rename_file(char *from, char *to) +{ + int err; + + err = rename(from, to); + if (err < 0) + return -errno; + return 0; +} + +int rename2_file(char *from, char *to, unsigned int flags) +{ + int err; + +#ifndef SYS_renameat2 +# ifdef __x86_64__ +# define SYS_renameat2 316 +# endif +# ifdef __i386__ +# define SYS_renameat2 353 +# endif +#endif + +#ifdef SYS_renameat2 + err = syscall(SYS_renameat2, AT_FDCWD, from, AT_FDCWD, to, flags); + if (err < 0) { + if (errno != ENOSYS) + return -errno; + else + return -EINVAL; + } + return 0; +#else + return -EINVAL; +#endif +} + +int do_statfs(char *root, long *bsize_out, long long *blocks_out, + long long *bfree_out, long long *bavail_out, + long long *files_out, long long *ffree_out, + void *fsid_out, int fsid_size, long *namelen_out) +{ + struct statfs64 buf; + int err; + + err = statfs64(root, &buf); + if (err < 0) + return -errno; + + *bsize_out = buf.f_bsize; + *blocks_out = buf.f_blocks; + *bfree_out = buf.f_bfree; + *bavail_out = buf.f_bavail; + *files_out = buf.f_files; + *ffree_out = buf.f_ffree; + memcpy(fsid_out, &buf.f_fsid, + sizeof(buf.f_fsid) > fsid_size ? fsid_size : + sizeof(buf.f_fsid)); + *namelen_out = buf.f_namelen; + + return 0; +} -- cgit v1.3.1