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/bfs/dir.c | 356 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 356 insertions(+) create mode 100644 fs/bfs/dir.c (limited to 'fs/bfs/dir.c') diff --git a/fs/bfs/dir.c b/fs/bfs/dir.c new file mode 100644 index 000000000..91a4871fa --- /dev/null +++ b/fs/bfs/dir.c @@ -0,0 +1,356 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * fs/bfs/dir.c + * BFS directory operations. + * Copyright (C) 1999-2018 Tigran Aivazian + * Made endianness-clean by Andrew Stribblehill 2005 + */ + +#include +#include +#include +#include +#include +#include "bfs.h" + +#undef DEBUG + +#ifdef DEBUG +#define dprintf(x...) printf(x) +#else +#define dprintf(x...) +#endif + +static int bfs_add_entry(struct inode *dir, const struct qstr *child, int ino); +static struct buffer_head *bfs_find_entry(struct inode *dir, + const struct qstr *child, + struct bfs_dirent **res_dir); + +static int bfs_readdir(struct file *f, struct dir_context *ctx) +{ + struct inode *dir = file_inode(f); + struct buffer_head *bh; + struct bfs_dirent *de; + unsigned int offset; + int block; + + if (ctx->pos & (BFS_DIRENT_SIZE - 1)) { + printf("Bad f_pos=%08lx for %s:%08llx\n", + (unsigned long)ctx->pos, + dir->i_sb->s_id, dir->i_ino); + return -EINVAL; + } + + while (ctx->pos < dir->i_size) { + offset = ctx->pos & (BFS_BSIZE - 1); + block = BFS_I(dir)->i_sblock + (ctx->pos >> BFS_BSIZE_BITS); + bh = sb_bread(dir->i_sb, block); + if (!bh) { + ctx->pos += BFS_BSIZE - offset; + continue; + } + do { + de = (struct bfs_dirent *)(bh->b_data + offset); + if (de->ino) { + int size = strnlen(de->name, BFS_NAMELEN); + if (!dir_emit(ctx, de->name, size, + le16_to_cpu(de->ino), + DT_UNKNOWN)) { + brelse(bh); + return 0; + } + } + offset += BFS_DIRENT_SIZE; + ctx->pos += BFS_DIRENT_SIZE; + } while ((offset < BFS_BSIZE) && (ctx->pos < dir->i_size)); + brelse(bh); + } + return 0; +} + +const struct file_operations bfs_dir_operations = { + .read = generic_read_dir, + .iterate_shared = bfs_readdir, + .fsync = simple_fsync, + .llseek = generic_file_llseek, +}; + +static int bfs_create(struct mnt_idmap *idmap, struct inode *dir, + struct dentry *dentry, umode_t mode) +{ + int err; + struct inode *inode; + struct super_block *s = dir->i_sb; + struct bfs_sb_info *info = BFS_SB(s); + unsigned long ino; + + inode = new_inode(s); + if (!inode) + return -ENOMEM; + mutex_lock(&info->bfs_lock); + ino = find_first_zero_bit(info->si_imap, info->si_lasti + 1); + if (ino > info->si_lasti) { + mutex_unlock(&info->bfs_lock); + iput(inode); + return -ENOSPC; + } + set_bit(ino, info->si_imap); + info->si_freei--; + inode_init_owner(&nop_mnt_idmap, inode, dir, mode); + simple_inode_init_ts(inode); + inode->i_blocks = 0; + inode->i_op = &bfs_file_inops; + inode->i_fop = &bfs_file_operations; + inode->i_mapping->a_ops = &bfs_aops; + inode->i_ino = ino; + BFS_I(inode)->i_dsk_ino = ino; + BFS_I(inode)->i_sblock = 0; + BFS_I(inode)->i_eblock = 0; + insert_inode_hash(inode); + mark_inode_dirty(inode); + bfs_dump_imap("create", s); + + err = bfs_add_entry(dir, &dentry->d_name, inode->i_ino); + if (err) { + inode_dec_link_count(inode); + mutex_unlock(&info->bfs_lock); + iput(inode); + return err; + } + mutex_unlock(&info->bfs_lock); + d_instantiate(dentry, inode); + return 0; +} + +static struct dentry *bfs_lookup(struct inode *dir, struct dentry *dentry, + unsigned int flags) +{ + struct inode *inode = NULL; + struct buffer_head *bh; + struct bfs_dirent *de; + struct bfs_sb_info *info = BFS_SB(dir->i_sb); + + if (dentry->d_name.len > BFS_NAMELEN) + return ERR_PTR(-ENAMETOOLONG); + + mutex_lock(&info->bfs_lock); + bh = bfs_find_entry(dir, &dentry->d_name, &de); + if (bh) { + unsigned long ino = (unsigned long)le16_to_cpu(de->ino); + brelse(bh); + inode = bfs_iget(dir->i_sb, ino); + } + mutex_unlock(&info->bfs_lock); + return d_splice_alias(inode, dentry); +} + +static int bfs_link(struct dentry *old, struct inode *dir, + struct dentry *new) +{ + struct inode *inode = d_inode(old); + struct bfs_sb_info *info = BFS_SB(inode->i_sb); + int err; + + mutex_lock(&info->bfs_lock); + err = bfs_add_entry(dir, &new->d_name, inode->i_ino); + if (err) { + mutex_unlock(&info->bfs_lock); + return err; + } + inc_nlink(inode); + inode_set_ctime_current(inode); + mark_inode_dirty(inode); + ihold(inode); + d_instantiate(new, inode); + mutex_unlock(&info->bfs_lock); + return 0; +} + +static int bfs_unlink(struct inode *dir, struct dentry *dentry) +{ + int error = -ENOENT; + struct inode *inode = d_inode(dentry); + struct buffer_head *bh; + struct bfs_dirent *de; + struct bfs_sb_info *info = BFS_SB(inode->i_sb); + + mutex_lock(&info->bfs_lock); + bh = bfs_find_entry(dir, &dentry->d_name, &de); + if (!bh || (le16_to_cpu(de->ino) != inode->i_ino)) + goto out_brelse; + + if (!inode->i_nlink) { + printf("unlinking non-existent file %s:%llu (nlink=%d)\n", + inode->i_sb->s_id, inode->i_ino, + inode->i_nlink); + set_nlink(inode, 1); + } + de->ino = 0; + mmb_mark_buffer_dirty(bh, &BFS_I(dir)->i_metadata_bhs); + inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir)); + mark_inode_dirty(dir); + inode_set_ctime_to_ts(inode, inode_get_ctime(dir)); + inode_dec_link_count(inode); + error = 0; + +out_brelse: + brelse(bh); + mutex_unlock(&info->bfs_lock); + return error; +} + +static int bfs_rename(struct mnt_idmap *idmap, struct inode *old_dir, + struct dentry *old_dentry, struct inode *new_dir, + struct dentry *new_dentry, unsigned int flags) +{ + struct inode *old_inode, *new_inode; + struct buffer_head *old_bh, *new_bh; + struct bfs_dirent *old_de, *new_de; + struct bfs_sb_info *info; + int error = -ENOENT; + + if (flags & ~RENAME_NOREPLACE) + return -EINVAL; + + old_bh = new_bh = NULL; + old_inode = d_inode(old_dentry); + if (S_ISDIR(old_inode->i_mode)) + return -EINVAL; + + info = BFS_SB(old_inode->i_sb); + + mutex_lock(&info->bfs_lock); + old_bh = bfs_find_entry(old_dir, &old_dentry->d_name, &old_de); + + if (!old_bh || (le16_to_cpu(old_de->ino) != old_inode->i_ino)) + goto end_rename; + + error = -EPERM; + new_inode = d_inode(new_dentry); + new_bh = bfs_find_entry(new_dir, &new_dentry->d_name, &new_de); + + if (new_bh && !new_inode) { + brelse(new_bh); + new_bh = NULL; + } + if (!new_bh) { + error = bfs_add_entry(new_dir, &new_dentry->d_name, + old_inode->i_ino); + if (error) + goto end_rename; + } + old_de->ino = 0; + inode_set_mtime_to_ts(old_dir, inode_set_ctime_current(old_dir)); + mark_inode_dirty(old_dir); + if (new_inode) { + inode_set_ctime_current(new_inode); + inode_dec_link_count(new_inode); + } + mmb_mark_buffer_dirty(old_bh, &BFS_I(old_dir)->i_metadata_bhs); + error = 0; + +end_rename: + mutex_unlock(&info->bfs_lock); + brelse(old_bh); + brelse(new_bh); + return error; +} + +const struct inode_operations bfs_dir_inops = { + .create = bfs_create, + .lookup = bfs_lookup, + .link = bfs_link, + .unlink = bfs_unlink, + .rename = bfs_rename, +}; + +static int bfs_add_entry(struct inode *dir, const struct qstr *child, int ino) +{ + const unsigned char *name = child->name; + int namelen = child->len; + struct buffer_head *bh; + struct bfs_dirent *de; + int block, sblock, eblock, off, pos; + int i; + + dprintf("name=%s, namelen=%d\n", name, namelen); + + sblock = BFS_I(dir)->i_sblock; + eblock = BFS_I(dir)->i_eblock; + for (block = sblock; block <= eblock; block++) { + bh = sb_bread(dir->i_sb, block); + if (!bh) + return -EIO; + for (off = 0; off < BFS_BSIZE; off += BFS_DIRENT_SIZE) { + de = (struct bfs_dirent *)(bh->b_data + off); + if (!de->ino) { + pos = (block - sblock) * BFS_BSIZE + off; + if (pos >= dir->i_size) { + dir->i_size += BFS_DIRENT_SIZE; + inode_set_ctime_current(dir); + } + inode_set_mtime_to_ts(dir, + inode_set_ctime_current(dir)); + mark_inode_dirty(dir); + de->ino = cpu_to_le16((u16)ino); + for (i = 0; i < BFS_NAMELEN; i++) + de->name[i] = + (i < namelen) ? name[i] : 0; + mmb_mark_buffer_dirty(bh, + &BFS_I(dir)->i_metadata_bhs); + brelse(bh); + return 0; + } + } + brelse(bh); + } + return -ENOSPC; +} + +static inline int bfs_namecmp(int len, const unsigned char *name, + const char *buffer) +{ + if ((len < BFS_NAMELEN) && buffer[len]) + return 0; + return !memcmp(name, buffer, len); +} + +static struct buffer_head *bfs_find_entry(struct inode *dir, + const struct qstr *child, + struct bfs_dirent **res_dir) +{ + unsigned long block = 0, offset = 0; + struct buffer_head *bh = NULL; + struct bfs_dirent *de; + const unsigned char *name = child->name; + int namelen = child->len; + + *res_dir = NULL; + if (namelen > BFS_NAMELEN) + return NULL; + + while (block * BFS_BSIZE + offset < dir->i_size) { + if (!bh) { + bh = sb_bread(dir->i_sb, BFS_I(dir)->i_sblock + block); + if (!bh) { + block++; + continue; + } + } + de = (struct bfs_dirent *)(bh->b_data + offset); + offset += BFS_DIRENT_SIZE; + if (le16_to_cpu(de->ino) && + bfs_namecmp(namelen, name, de->name)) { + *res_dir = de; + return bh; + } + if (offset < bh->b_size) + continue; + brelse(bh); + bh = NULL; + offset = 0; + block++; + } + brelse(bh); + return NULL; +} -- cgit v1.3.1