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/coda/inode.c | 402 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 402 insertions(+) create mode 100644 fs/coda/inode.c (limited to 'fs/coda/inode.c') diff --git a/fs/coda/inode.c b/fs/coda/inode.c new file mode 100644 index 000000000..40b43866e --- /dev/null +++ b/fs/coda/inode.c @@ -0,0 +1,402 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Super block/filesystem wide operations + * + * Copyright (C) 1996 Peter J. Braam and + * Michael Callahan + * + * Rewritten for Linux 2.1. Peter Braam + * Copyright (C) Carnegie Mellon University + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include "coda_psdev.h" +#include "coda_linux.h" +#include "coda_cache.h" + +#include "coda_int.h" + +/* VFS super_block ops */ +static void coda_evict_inode(struct inode *); +static void coda_put_super(struct super_block *); +static int coda_statfs(struct dentry *dentry, struct kstatfs *buf); + +static struct kmem_cache * coda_inode_cachep; + +static struct inode *coda_alloc_inode(struct super_block *sb) +{ + struct coda_inode_info *ei; + ei = alloc_inode_sb(sb, coda_inode_cachep, GFP_KERNEL); + if (!ei) + return NULL; + memset(&ei->c_fid, 0, sizeof(struct CodaFid)); + ei->c_flags = 0; + ei->c_uid = GLOBAL_ROOT_UID; + ei->c_cached_perm = 0; + spin_lock_init(&ei->c_lock); + return &ei->vfs_inode; +} + +static void coda_free_inode(struct inode *inode) +{ + kmem_cache_free(coda_inode_cachep, ITOC(inode)); +} + +static void init_once(void *foo) +{ + struct coda_inode_info *ei = (struct coda_inode_info *) foo; + + inode_init_once(&ei->vfs_inode); +} + +int __init coda_init_inodecache(void) +{ + coda_inode_cachep = kmem_cache_create("coda_inode_cache", + sizeof(struct coda_inode_info), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT, + init_once); + if (coda_inode_cachep == NULL) + return -ENOMEM; + return 0; +} + +void coda_destroy_inodecache(void) +{ + /* + * Make sure all delayed rcu free inodes are flushed before we + * destroy cache. + */ + rcu_barrier(); + kmem_cache_destroy(coda_inode_cachep); +} + +static int coda_reconfigure(struct fs_context *fc) +{ + sync_filesystem(fc->root->d_sb); + fc->sb_flags |= SB_NOATIME; + return 0; +} + +/* exported operations */ +static const struct super_operations coda_super_operations = +{ + .alloc_inode = coda_alloc_inode, + .free_inode = coda_free_inode, + .evict_inode = coda_evict_inode, + .put_super = coda_put_super, + .statfs = coda_statfs, +}; + +struct coda_fs_context { + int idx; +}; + +enum { + Opt_fd, +}; + +static const struct fs_parameter_spec coda_param_specs[] = { + fsparam_fd ("fd", Opt_fd), + {} +}; + +static int coda_set_idx(struct fs_context *fc, struct file *file) +{ + struct coda_fs_context *ctx = fc->fs_private; + struct inode *inode; + int idx; + + inode = file_inode(file); + if (!S_ISCHR(inode->i_mode) || imajor(inode) != CODA_PSDEV_MAJOR) { + return invalf(fc, "coda: Not coda psdev"); + } + idx = iminor(inode); + if (idx < 0 || idx >= MAX_CODADEVS) + return invalf(fc, "coda: Bad minor number"); + ctx->idx = idx; + return 0; +} + +static int coda_parse_fd(struct fs_context *fc, struct fs_parameter *param, + struct fs_parse_result *result) +{ + struct file *file; + int err; + + if (param->type == fs_value_is_file) { + file = param->file; + param->file = NULL; + } else { + file = fget(result->uint_32); + } + if (!file) + return -EBADF; + + err = coda_set_idx(fc, file); + fput(file); + return err; +} + +static int coda_parse_param(struct fs_context *fc, struct fs_parameter *param) +{ + struct fs_parse_result result; + int opt; + + opt = fs_parse(fc, coda_param_specs, param, &result); + if (opt < 0) + return opt; + + switch (opt) { + case Opt_fd: + return coda_parse_fd(fc, param, &result); + } + + return 0; +} + +/* + * Parse coda's binary mount data form. We ignore any errors and go with index + * 0 if we get one for backward compatibility. + */ +static int coda_parse_monolithic(struct fs_context *fc, void *_data) +{ + struct file *file; + struct coda_mount_data *data = _data; + + if (!data) + return invalf(fc, "coda: Bad mount data"); + + if (data->version != CODA_MOUNT_VERSION) + return invalf(fc, "coda: Bad mount version"); + + file = fget(data->fd); + if (file) { + coda_set_idx(fc, file); + fput(file); + } + return 0; +} + +static int coda_fill_super(struct super_block *sb, struct fs_context *fc) +{ + struct coda_fs_context *ctx = fc->fs_private; + struct inode *root = NULL; + struct venus_comm *vc; + struct CodaFid fid; + int error; + + infof(fc, "coda: device index: %i\n", ctx->idx); + + vc = &coda_comms[ctx->idx]; + mutex_lock(&vc->vc_mutex); + + if (!vc->vc_inuse) { + errorf(fc, "coda: No pseudo device"); + error = -EINVAL; + goto unlock_out; + } + + if (vc->vc_sb) { + errorf(fc, "coda: Device already mounted"); + error = -EBUSY; + goto unlock_out; + } + + vc->vc_sb = sb; + mutex_unlock(&vc->vc_mutex); + + sb->s_fs_info = vc; + sb->s_flags |= SB_NOATIME; + sb->s_blocksize = 4096; /* XXXXX what do we put here?? */ + sb->s_blocksize_bits = 12; + sb->s_magic = CODA_SUPER_MAGIC; + sb->s_op = &coda_super_operations; + set_default_d_op(sb, &coda_dentry_operations); + sb->s_time_gran = 1; + sb->s_time_min = S64_MIN; + sb->s_time_max = S64_MAX; + + error = super_setup_bdi(sb); + if (error) + goto error; + + /* get root fid from Venus: this needs the root inode */ + error = venus_rootfid(sb, &fid); + if ( error ) { + pr_warn("%s: coda_get_rootfid failed with %d\n", + __func__, error); + goto error; + } + pr_info("%s: rootfid is %s\n", __func__, coda_f2s(&fid)); + + /* make root inode */ + root = coda_cnode_make(&fid, sb); + if (IS_ERR(root)) { + error = PTR_ERR(root); + pr_warn("Failure of coda_cnode_make for root: error %d\n", + error); + goto error; + } + + pr_info("%s: rootinode is %llu dev %s\n", + __func__, root->i_ino, root->i_sb->s_id); + sb->s_root = d_make_root(root); + if (!sb->s_root) { + error = -EINVAL; + goto error; + } + return 0; + +error: + mutex_lock(&vc->vc_mutex); + vc->vc_sb = NULL; + sb->s_fs_info = NULL; +unlock_out: + mutex_unlock(&vc->vc_mutex); + return error; +} + +static void coda_put_super(struct super_block *sb) +{ + struct venus_comm *vcp = coda_vcp(sb); + mutex_lock(&vcp->vc_mutex); + vcp->vc_sb = NULL; + sb->s_fs_info = NULL; + mutex_unlock(&vcp->vc_mutex); + mutex_destroy(&vcp->vc_mutex); + + pr_info("Bye bye.\n"); +} + +static void coda_evict_inode(struct inode *inode) +{ + truncate_inode_pages_final(&inode->i_data); + clear_inode(inode); + coda_cache_clear_inode(inode); +} + +int coda_getattr(struct mnt_idmap *idmap, const struct path *path, + struct kstat *stat, u32 request_mask, unsigned int flags) +{ + int err = coda_revalidate_inode(d_inode(path->dentry)); + if (!err) + generic_fillattr(&nop_mnt_idmap, request_mask, + d_inode(path->dentry), stat); + return err; +} + +int coda_setattr(struct mnt_idmap *idmap, struct dentry *de, + struct iattr *iattr) +{ + struct inode *inode = d_inode(de); + struct coda_vattr vattr; + int error; + + memset(&vattr, 0, sizeof(vattr)); + + inode_set_ctime_current(inode); + coda_iattr_to_vattr(iattr, &vattr); + vattr.va_type = C_VNON; /* cannot set type */ + + /* Venus is responsible for truncating the container-file!!! */ + error = venus_setattr(inode->i_sb, coda_i2f(inode), &vattr); + + if (!error) { + coda_vattr_to_iattr(inode, &vattr); + coda_cache_clear_inode(inode); + } + return error; +} + +const struct inode_operations coda_file_inode_operations = { + .permission = coda_permission, + .getattr = coda_getattr, + .setattr = coda_setattr, +}; + +static int coda_statfs(struct dentry *dentry, struct kstatfs *buf) +{ + int error; + + error = venus_statfs(dentry, buf); + + if (error) { + /* fake something like AFS does */ + buf->f_blocks = 9000000; + buf->f_bfree = 9000000; + buf->f_bavail = 9000000; + buf->f_files = 9000000; + buf->f_ffree = 9000000; + } + + /* and fill in the rest */ + buf->f_type = CODA_SUPER_MAGIC; + buf->f_bsize = 4096; + buf->f_namelen = CODA_MAXNAMLEN; + + return 0; +} + +static int coda_get_tree(struct fs_context *fc) +{ + if (task_active_pid_ns(current) != &init_pid_ns) + return -EINVAL; + + return get_tree_nodev(fc, coda_fill_super); +} + +static void coda_free_fc(struct fs_context *fc) +{ + kfree(fc->fs_private); +} + +static const struct fs_context_operations coda_context_ops = { + .free = coda_free_fc, + .parse_param = coda_parse_param, + .parse_monolithic = coda_parse_monolithic, + .get_tree = coda_get_tree, + .reconfigure = coda_reconfigure, +}; + +static int coda_init_fs_context(struct fs_context *fc) +{ + struct coda_fs_context *ctx; + + ctx = kzalloc_obj(struct coda_fs_context); + if (!ctx) + return -ENOMEM; + + fc->fs_private = ctx; + fc->ops = &coda_context_ops; + return 0; +} + +struct file_system_type coda_fs_type = { + .owner = THIS_MODULE, + .name = "coda", + .init_fs_context = coda_init_fs_context, + .parameters = coda_param_specs, + .kill_sb = kill_anon_super, + .fs_flags = FS_BINARY_MOUNTDATA, +}; +MODULE_ALIAS_FS("coda"); + -- cgit v1.3.1