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/autofs/inode.c | 457 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 457 insertions(+) create mode 100644 fs/autofs/inode.c (limited to 'fs/autofs/inode.c') diff --git a/fs/autofs/inode.c b/fs/autofs/inode.c new file mode 100644 index 000000000..6b15a3717 --- /dev/null +++ b/fs/autofs/inode.c @@ -0,0 +1,457 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright 1997-1998 Transmeta Corporation -- All Rights Reserved + * Copyright 2005-2006 Ian Kent + */ + +#include +#include + +#include "autofs_i.h" + +struct autofs_info *autofs_new_ino(struct autofs_sb_info *sbi) +{ + struct autofs_info *ino; + + ino = kzalloc_obj(*ino); + if (ino) { + INIT_LIST_HEAD(&ino->active); + INIT_LIST_HEAD(&ino->expiring); + ino->last_used = jiffies; + ino->sbi = sbi; + ino->exp_timeout = -1; + ino->count = 1; + } + return ino; +} + +void autofs_clean_ino(struct autofs_info *ino) +{ + ino->uid = GLOBAL_ROOT_UID; + ino->gid = GLOBAL_ROOT_GID; + ino->exp_timeout = -1; + ino->last_used = jiffies; +} + +void autofs_free_ino(struct autofs_info *ino) +{ + kfree_rcu(ino, rcu); +} + +void autofs_kill_sb(struct super_block *sb) +{ + struct autofs_sb_info *sbi = autofs_sbi(sb); + + /* + * In the event of a failure in get_sb_nodev the superblock + * info is not present so nothing else has been setup, so + * just call kill_anon_super when we are called from + * deactivate_super. + */ + if (sbi) { + /* Free wait queues, close pipe */ + autofs_catatonic_mode(sbi); + put_pid(sbi->oz_pgrp); + } + + pr_debug("shutting down\n"); + kill_anon_super(sb); + if (sbi) + kfree_rcu(sbi, rcu); +} + +static int autofs_show_options(struct seq_file *m, struct dentry *root) +{ + struct autofs_sb_info *sbi = autofs_sbi(root->d_sb); + struct inode *root_inode = d_inode(root->d_sb->s_root); + + if (!sbi) + return 0; + + seq_printf(m, ",fd=%d", sbi->pipefd); + if (!uid_eq(root_inode->i_uid, GLOBAL_ROOT_UID)) + seq_printf(m, ",uid=%u", + from_kuid_munged(&init_user_ns, root_inode->i_uid)); + if (!gid_eq(root_inode->i_gid, GLOBAL_ROOT_GID)) + seq_printf(m, ",gid=%u", + from_kgid_munged(&init_user_ns, root_inode->i_gid)); + seq_printf(m, ",pgrp=%d", pid_vnr(sbi->oz_pgrp)); + seq_printf(m, ",timeout=%lu", sbi->exp_timeout/HZ); + seq_printf(m, ",minproto=%d", sbi->min_proto); + seq_printf(m, ",maxproto=%d", sbi->max_proto); + + if (autofs_type_offset(sbi->type)) + seq_puts(m, ",offset"); + else if (autofs_type_direct(sbi->type)) + seq_puts(m, ",direct"); + else + seq_puts(m, ",indirect"); + if (sbi->flags & AUTOFS_SBI_STRICTEXPIRE) + seq_puts(m, ",strictexpire"); + if (sbi->flags & AUTOFS_SBI_IGNORE) + seq_puts(m, ",ignore"); +#ifdef CONFIG_CHECKPOINT_RESTORE + if (sbi->pipe) + seq_printf(m, ",pipe_ino=%llu", file_inode(sbi->pipe)->i_ino); + else + seq_puts(m, ",pipe_ino=-1"); +#endif + return 0; +} + +static void autofs_evict_inode(struct inode *inode) +{ + clear_inode(inode); + kfree(inode->i_private); +} + +static const struct super_operations autofs_sops = { + .statfs = simple_statfs, + .show_options = autofs_show_options, + .evict_inode = autofs_evict_inode, +}; + +enum { + Opt_direct, + Opt_fd, + Opt_gid, + Opt_ignore, + Opt_indirect, + Opt_maxproto, + Opt_minproto, + Opt_offset, + Opt_pgrp, + Opt_strictexpire, + Opt_uid, +}; + +const struct fs_parameter_spec autofs_param_specs[] = { + fsparam_flag ("direct", Opt_direct), + fsparam_fd ("fd", Opt_fd), + fsparam_gid ("gid", Opt_gid), + fsparam_flag ("ignore", Opt_ignore), + fsparam_flag ("indirect", Opt_indirect), + fsparam_u32 ("maxproto", Opt_maxproto), + fsparam_u32 ("minproto", Opt_minproto), + fsparam_flag ("offset", Opt_offset), + fsparam_u32 ("pgrp", Opt_pgrp), + fsparam_flag ("strictexpire", Opt_strictexpire), + fsparam_uid ("uid", Opt_uid), + {} +}; + +struct autofs_fs_context { + kuid_t uid; + kgid_t gid; + int pgrp; + bool pgrp_set; +}; + +/* + * Open the fd. We do it here rather than in get_tree so that it's done in the + * context of the system call that passed the data and not the one that + * triggered the superblock creation, lest the fd gets reassigned. + */ +static int autofs_parse_fd(struct fs_context *fc, struct autofs_sb_info *sbi, + struct fs_parameter *param, + struct fs_parse_result *result) +{ + struct file *pipe; + int ret; + + if (param->type == fs_value_is_file) { + /* came through the new api */ + pipe = param->file; + param->file = NULL; + } else { + pipe = fget(result->uint_32); + } + if (!pipe) { + errorf(fc, "could not open pipe file descriptor"); + return -EBADF; + } + + ret = autofs_check_pipe(pipe); + if (ret < 0) { + errorf(fc, "Invalid/unusable pipe"); + fput(pipe); + return -EBADF; + } + + autofs_set_packet_pipe_flags(pipe); + + if (sbi->pipe) + fput(sbi->pipe); + + sbi->pipefd = result->uint_32; + sbi->pipe = pipe; + + return 0; +} + +static int autofs_parse_param(struct fs_context *fc, struct fs_parameter *param) +{ + struct autofs_fs_context *ctx = fc->fs_private; + struct autofs_sb_info *sbi = fc->s_fs_info; + struct fs_parse_result result; + int opt; + + opt = fs_parse(fc, autofs_param_specs, param, &result); + if (opt < 0) + return opt; + + switch (opt) { + case Opt_fd: + return autofs_parse_fd(fc, sbi, param, &result); + case Opt_uid: + ctx->uid = result.uid; + break; + case Opt_gid: + ctx->gid = result.gid; + break; + case Opt_pgrp: + ctx->pgrp = result.uint_32; + ctx->pgrp_set = true; + break; + case Opt_minproto: + sbi->min_proto = result.uint_32; + break; + case Opt_maxproto: + sbi->max_proto = result.uint_32; + break; + case Opt_indirect: + set_autofs_type_indirect(&sbi->type); + break; + case Opt_direct: + set_autofs_type_direct(&sbi->type); + break; + case Opt_offset: + set_autofs_type_offset(&sbi->type); + break; + case Opt_strictexpire: + sbi->flags |= AUTOFS_SBI_STRICTEXPIRE; + break; + case Opt_ignore: + sbi->flags |= AUTOFS_SBI_IGNORE; + } + + return 0; +} + +static struct autofs_sb_info *autofs_alloc_sbi(void) +{ + struct autofs_sb_info *sbi; + + sbi = kzalloc_obj(*sbi); + if (!sbi) + return NULL; + + sbi->magic = AUTOFS_SBI_MAGIC; + sbi->flags = AUTOFS_SBI_CATATONIC; + sbi->min_proto = AUTOFS_MIN_PROTO_VERSION; + sbi->max_proto = AUTOFS_MAX_PROTO_VERSION; + sbi->pipefd = -1; + sbi->mnt_ns_id = to_ns_common(current->nsproxy->mnt_ns)->ns_id; + + set_autofs_type_indirect(&sbi->type); + mutex_init(&sbi->wq_mutex); + mutex_init(&sbi->pipe_mutex); + spin_lock_init(&sbi->fs_lock); + spin_lock_init(&sbi->lookup_lock); + INIT_LIST_HEAD(&sbi->active_list); + INIT_LIST_HEAD(&sbi->expiring_list); + + return sbi; +} + +static int autofs_validate_protocol(struct fs_context *fc) +{ + struct autofs_sb_info *sbi = fc->s_fs_info; + + /* Test versions first */ + if (sbi->max_proto < AUTOFS_MIN_PROTO_VERSION || + sbi->min_proto > AUTOFS_MAX_PROTO_VERSION) { + errorf(fc, "kernel does not match daemon version " + "daemon (%d, %d) kernel (%d, %d)\n", + sbi->min_proto, sbi->max_proto, + AUTOFS_MIN_PROTO_VERSION, AUTOFS_MAX_PROTO_VERSION); + return -EINVAL; + } + + /* Establish highest kernel protocol version */ + if (sbi->max_proto > AUTOFS_MAX_PROTO_VERSION) + sbi->version = AUTOFS_MAX_PROTO_VERSION; + else + sbi->version = sbi->max_proto; + + switch (sbi->version) { + case 4: + sbi->sub_version = 7; + break; + case 5: + sbi->sub_version = AUTOFS_PROTO_SUBVERSION; + break; + default: + sbi->sub_version = 0; + } + + return 0; +} + +static int autofs_fill_super(struct super_block *s, struct fs_context *fc) +{ + struct autofs_fs_context *ctx = fc->fs_private; + struct autofs_sb_info *sbi = s->s_fs_info; + struct inode *root_inode; + struct autofs_info *ino; + + pr_debug("starting up, sbi = %p\n", sbi); + + sbi->sb = s; + s->s_blocksize = 1024; + s->s_blocksize_bits = 10; + s->s_magic = AUTOFS_SUPER_MAGIC; + s->s_op = &autofs_sops; + set_default_d_op(s, &autofs_dentry_operations); + s->s_time_gran = 1; + + /* + * Get the root inode and dentry, but defer checking for errors. + */ + ino = autofs_new_ino(sbi); + if (!ino) + return -ENOMEM; + + root_inode = autofs_get_inode(s, S_IFDIR | 0755); + if (!root_inode) { + autofs_free_ino(ino); + return -ENOMEM; + } + + root_inode->i_uid = ctx->uid; + root_inode->i_gid = ctx->gid; + root_inode->i_fop = &autofs_root_operations; + root_inode->i_op = &autofs_dir_inode_operations; + + s->s_root = d_make_root(root_inode); + if (unlikely(!s->s_root)) { + autofs_free_ino(ino); + return -ENOMEM; + } + s->s_root->d_fsdata = ino; + + if (ctx->pgrp_set) { + sbi->oz_pgrp = find_get_pid(ctx->pgrp); + if (!sbi->oz_pgrp) + return invalf(fc, "Could not find process group %d", + ctx->pgrp); + } else + sbi->oz_pgrp = get_task_pid(current, PIDTYPE_PGID); + + if (autofs_type_trigger(sbi->type)) + /* s->s_root won't be contended so there's little to + * be gained by not taking the d_lock when setting + * d_flags, even when a lot mounts are being done. + */ + managed_dentry_set_managed(s->s_root); + + pr_debug("pipe fd = %d, pgrp = %u\n", + sbi->pipefd, pid_nr(sbi->oz_pgrp)); + + sbi->flags &= ~AUTOFS_SBI_CATATONIC; + return 0; +} + +/* + * Validate the parameters and then request a superblock. + */ +static int autofs_get_tree(struct fs_context *fc) +{ + struct autofs_sb_info *sbi = fc->s_fs_info; + int ret; + + ret = autofs_validate_protocol(fc); + if (ret) + return ret; + + if (sbi->pipefd < 0) + return invalf(fc, "No control pipe specified"); + + return get_tree_nodev(fc, autofs_fill_super); +} + +static void autofs_free_fc(struct fs_context *fc) +{ + struct autofs_fs_context *ctx = fc->fs_private; + struct autofs_sb_info *sbi = fc->s_fs_info; + + if (sbi) { + if (sbi->pipe) + fput(sbi->pipe); + kfree(sbi); + } + kfree(ctx); +} + +static const struct fs_context_operations autofs_context_ops = { + .free = autofs_free_fc, + .parse_param = autofs_parse_param, + .get_tree = autofs_get_tree, +}; + +/* + * Set up the filesystem mount context. + */ +int autofs_init_fs_context(struct fs_context *fc) +{ + struct autofs_fs_context *ctx; + struct autofs_sb_info *sbi; + + ctx = kzalloc_obj(struct autofs_fs_context); + if (!ctx) + goto nomem; + + ctx->uid = current_uid(); + ctx->gid = current_gid(); + + sbi = autofs_alloc_sbi(); + if (!sbi) + goto nomem_ctx; + + fc->fs_private = ctx; + fc->s_fs_info = sbi; + fc->ops = &autofs_context_ops; + return 0; + +nomem_ctx: + kfree(ctx); +nomem: + return -ENOMEM; +} + +struct inode *autofs_get_inode(struct super_block *sb, umode_t mode) +{ + struct inode *inode = new_inode(sb); + + if (inode == NULL) + return NULL; + + inode->i_mode = mode; + if (sb->s_root) { + inode->i_uid = d_inode(sb->s_root)->i_uid; + inode->i_gid = d_inode(sb->s_root)->i_gid; + } + simple_inode_init_ts(inode); + inode->i_ino = get_next_ino(); + + if (S_ISDIR(mode)) { + set_nlink(inode, 2); + inode->i_op = &autofs_dir_inode_operations; + inode->i_fop = &autofs_dir_operations; + } else if (S_ISLNK(mode)) { + inode->i_op = &autofs_symlink_inode_operations; + } else + WARN_ON(1); + + return inode; +} -- cgit v1.3.1