x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

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) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 599 insertions, 0 deletions

diff --git a/fs/vboxsf/utils.c b/fs/vboxsf/utils.c
new file mode 100644
index 000000000..298bfc932
--- /dev/null
+++ b/fs/vboxsf/utils.c
@@ -0,0 +1,599 @@
+// SPDX-License-Identifier: MIT
+/*
+ * VirtualBox Guest Shared Folders support: Utility functions.
+ * Mainly conversion from/to VirtualBox/Linux data structures.
+ *
+ * Copyright (C) 2006-2018 Oracle Corporation
+ */
+
+#include <linux/namei.h>
+#include <linux/nls.h>
+#include <linux/sizes.h>
+#include <linux/pagemap.h>
+#include <linux/vfs.h>
+#include <linux/fileattr.h>
+#include "vfsmod.h"
+
+struct inode *vboxsf_new_inode(struct super_block *sb)
+{
+	struct vboxsf_sbi *sbi = VBOXSF_SBI(sb);
+	struct inode *inode;
+	unsigned long flags;
+	int cursor, ret;
+	u32 gen;
+
+	inode = new_inode(sb);
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+
+	idr_preload(GFP_KERNEL);
+	spin_lock_irqsave(&sbi->ino_idr_lock, flags);
+	cursor = idr_get_cursor(&sbi->ino_idr);
+	ret = idr_alloc_cyclic(&sbi->ino_idr, inode, 1, 0, GFP_ATOMIC);
+	if (ret >= 0 && ret < cursor)
+		sbi->next_generation++;
+	gen = sbi->next_generation;
+	spin_unlock_irqrestore(&sbi->ino_idr_lock, flags);
+	idr_preload_end();
+
+	if (ret < 0) {
+		iput(inode);
+		return ERR_PTR(ret);
+	}
+
+	inode->i_ino = ret;
+	inode->i_generation = gen;
+	return inode;
+}
+
+/* set [inode] attributes based on [info], uid/gid based on [sbi] */
+int vboxsf_init_inode(struct vboxsf_sbi *sbi, struct inode *inode,
+		       const struct shfl_fsobjinfo *info, bool reinit)
+{
+	const struct shfl_fsobjattr *attr;
+	s64 allocated;
+	umode_t mode;
+
+	attr = &info->attr;
+
+#define mode_set(r) ((attr->mode & (SHFL_UNIX_##r)) ? (S_##r) : 0)
+
+	mode = mode_set(IRUSR);
+	mode |= mode_set(IWUSR);
+	mode |= mode_set(IXUSR);
+
+	mode |= mode_set(IRGRP);
+	mode |= mode_set(IWGRP);
+	mode |= mode_set(IXGRP);
+
+	mode |= mode_set(IROTH);
+	mode |= mode_set(IWOTH);
+	mode |= mode_set(IXOTH);
+
+#undef mode_set
+
+	/* We use the host-side values for these */
+	inode->i_flags |= S_NOATIME | S_NOCMTIME;
+	inode->i_mapping->a_ops = &vboxsf_reg_aops;
+
+	if (SHFL_IS_DIRECTORY(attr->mode)) {
+		if (sbi->o.dmode_set)
+			mode = sbi->o.dmode;
+		mode &= ~sbi->o.dmask;
+		mode |= S_IFDIR;
+		if (!reinit) {
+			inode->i_op = &vboxsf_dir_iops;
+			inode->i_fop = &vboxsf_dir_fops;
+			/*
+			 * XXX: this probably should be set to the number of entries
+			 * in the directory plus two (. ..)
+			 */
+			set_nlink(inode, 1);
+		} else if (!S_ISDIR(inode->i_mode))
+			return -ESTALE;
+		inode->i_mode = mode;
+	} else if (SHFL_IS_SYMLINK(attr->mode)) {
+		if (sbi->o.fmode_set)
+			mode = sbi->o.fmode;
+		mode &= ~sbi->o.fmask;
+		mode |= S_IFLNK;
+		if (!reinit) {
+			inode->i_op = &vboxsf_lnk_iops;
+			set_nlink(inode, 1);
+		} else if (!S_ISLNK(inode->i_mode))
+			return -ESTALE;
+		inode->i_mode = mode;
+	} else {
+		if (sbi->o.fmode_set)
+			mode = sbi->o.fmode;
+		mode &= ~sbi->o.fmask;
+		mode |= S_IFREG;
+		if (!reinit) {
+			inode->i_op = &vboxsf_reg_iops;
+			inode->i_fop = &vboxsf_reg_fops;
+			set_nlink(inode, 1);
+		} else if (!S_ISREG(inode->i_mode))
+			return -ESTALE;
+		inode->i_mode = mode;
+	}
+
+	inode->i_uid = sbi->o.uid;
+	inode->i_gid = sbi->o.gid;
+
+	inode->i_size = info->size;
+	inode->i_blkbits = 12;
+	/* i_blocks always in units of 512 bytes! */
+	allocated = info->allocated + 511;
+	do_div(allocated, 512);
+	inode->i_blocks = allocated;
+
+	inode_set_atime_to_ts(inode,
+			      ns_to_timespec64(info->access_time.ns_relative_to_unix_epoch));
+	inode_set_ctime_to_ts(inode,
+			      ns_to_timespec64(info->change_time.ns_relative_to_unix_epoch));
+	inode_set_mtime_to_ts(inode,
+			      ns_to_timespec64(info->modification_time.ns_relative_to_unix_epoch));
+	return 0;
+}
+
+int vboxsf_create_at_dentry(struct dentry *dentry,
+			    struct shfl_createparms *params)
+{
+	struct vboxsf_sbi *sbi = VBOXSF_SBI(dentry->d_sb);
+	struct shfl_string *path;
+	int err;
+
+	path = vboxsf_path_from_dentry(sbi, dentry);
+	if (IS_ERR(path))
+		return PTR_ERR(path);
+
+	err = vboxsf_create(sbi->root, path, params);
+	__putname(path);
+
+	return err;
+}
+
+int vboxsf_stat(struct vboxsf_sbi *sbi, struct shfl_string *path,
+		struct shfl_fsobjinfo *info)
+{
+	struct shfl_createparms params = {};
+	int err;
+
+	params.handle = SHFL_HANDLE_NIL;
+	params.create_flags = SHFL_CF_LOOKUP | SHFL_CF_ACT_FAIL_IF_NEW;
+
+	err = vboxsf_create(sbi->root, path, &params);
+	if (err)
+		return err;
+
+	if (params.result != SHFL_FILE_EXISTS)
+		return -ENOENT;
+
+	if (info)
+		*info = params.info;
+
+	return 0;
+}
+
+int vboxsf_stat_dentry(struct dentry *dentry, struct shfl_fsobjinfo *info)
+{
+	struct vboxsf_sbi *sbi = VBOXSF_SBI(dentry->d_sb);
+	struct shfl_string *path;
+	int err;
+
+	path = vboxsf_path_from_dentry(sbi, dentry);
+	if (IS_ERR(path))
+		return PTR_ERR(path);
+
+	err = vboxsf_stat(sbi, path, info);
+	__putname(path);
+	return err;
+}
+
+int vboxsf_inode_revalidate(struct dentry *dentry)
+{
+	struct vboxsf_sbi *sbi;
+	struct vboxsf_inode *sf_i;
+	struct shfl_fsobjinfo info;
+	struct timespec64 mtime, prev_mtime;
+	struct inode *inode;
+	int err;
+
+	if (!dentry || !d_really_is_positive(dentry))
+		return -EINVAL;
+
+	inode = d_inode(dentry);
+	prev_mtime = inode_get_mtime(inode);
+	sf_i = VBOXSF_I(inode);
+	sbi = VBOXSF_SBI(dentry->d_sb);
+	if (!sf_i->force_restat) {
+		if (time_before(jiffies, dentry->d_time + sbi->o.ttl))
+			return 0;
+	}
+
+	err = vboxsf_stat_dentry(dentry, &info);
+	if (err)
+		return err;
+
+	dentry->d_time = jiffies;
+	sf_i->force_restat = 0;
+	err = vboxsf_init_inode(sbi, inode, &info, true);
+	if (err)
+		return err;
+
+	/*
+	 * If the file was changed on the host side we need to invalidate the
+	 * page-cache for it.  Note this also gets triggered by our own writes,
+	 * this is unavoidable.
+	 */
+	mtime = inode_get_mtime(inode);
+	if (timespec64_compare(&mtime, &prev_mtime) > 0)
+		invalidate_inode_pages2(inode->i_mapping);
+
+	return 0;
+}
+
+int vboxsf_getattr(struct mnt_idmap *idmap, const struct path *path,
+		   struct kstat *kstat, u32 request_mask, unsigned int flags)
+{
+	int err;
+	struct dentry *dentry = path->dentry;
+	struct inode *inode = d_inode(dentry);
+	struct vboxsf_inode *sf_i = VBOXSF_I(inode);
+
+	switch (flags & AT_STATX_SYNC_TYPE) {
+	case AT_STATX_DONT_SYNC:
+		err = 0;
+		break;
+	case AT_STATX_FORCE_SYNC:
+		sf_i->force_restat = 1;
+		fallthrough;
+	default:
+		err = vboxsf_inode_revalidate(dentry);
+	}
+	if (err)
+		return err;
+
+	generic_fillattr(&nop_mnt_idmap, request_mask, d_inode(dentry), kstat);
+	return 0;
+}
+
+int vboxsf_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
+		   struct iattr *iattr)
+{
+	struct vboxsf_inode *sf_i = VBOXSF_I(d_inode(dentry));
+	struct vboxsf_sbi *sbi = VBOXSF_SBI(dentry->d_sb);
+	struct shfl_createparms params = {};
+	struct shfl_fsobjinfo info = {};
+	u32 buf_len;
+	int err;
+
+	params.handle = SHFL_HANDLE_NIL;
+	params.create_flags = SHFL_CF_ACT_OPEN_IF_EXISTS |
+			      SHFL_CF_ACT_FAIL_IF_NEW |
+			      SHFL_CF_ACCESS_ATTR_WRITE;
+
+	/* this is at least required for Posix hosts */
+	if (iattr->ia_valid & ATTR_SIZE)
+		params.create_flags |= SHFL_CF_ACCESS_WRITE;
+
+	err = vboxsf_create_at_dentry(dentry, &params);
+	if (err || params.result != SHFL_FILE_EXISTS)
+		return err ? err : -ENOENT;
+
+#define mode_set(r) ((iattr->ia_mode & (S_##r)) ? SHFL_UNIX_##r : 0)
+
+	/*
+	 * Setting the file size and setting the other attributes has to
+	 * be handled separately.
+	 */
+	if (iattr->ia_valid & (ATTR_MODE | ATTR_ATIME | ATTR_MTIME)) {
+		if (iattr->ia_valid & ATTR_MODE) {
+			info.attr.mode = mode_set(IRUSR);
+			info.attr.mode |= mode_set(IWUSR);
+			info.attr.mode |= mode_set(IXUSR);
+			info.attr.mode |= mode_set(IRGRP);
+			info.attr.mode |= mode_set(IWGRP);
+			info.attr.mode |= mode_set(IXGRP);
+			info.attr.mode |= mode_set(IROTH);
+			info.attr.mode |= mode_set(IWOTH);
+			info.attr.mode |= mode_set(IXOTH);
+
+			if (iattr->ia_mode & S_IFDIR)
+				info.attr.mode |= SHFL_TYPE_DIRECTORY;
+			else
+				info.attr.mode |= SHFL_TYPE_FILE;
+		}
+
+		if (iattr->ia_valid & ATTR_ATIME)
+			info.access_time.ns_relative_to_unix_epoch =
+					    timespec64_to_ns(&iattr->ia_atime);
+
+		if (iattr->ia_valid & ATTR_MTIME)
+			info.modification_time.ns_relative_to_unix_epoch =
+					    timespec64_to_ns(&iattr->ia_mtime);
+
+		/*
+		 * Ignore ctime (inode change time) as it can't be set
+		 * from userland anyway.
+		 */
+
+		buf_len = sizeof(info);
+		err = vboxsf_fsinfo(sbi->root, params.handle,
+				    SHFL_INFO_SET | SHFL_INFO_FILE, &buf_len,
+				    &info);
+		if (err) {
+			vboxsf_close(sbi->root, params.handle);
+			return err;
+		}
+
+		/* the host may have given us different attr then requested */
+		sf_i->force_restat = 1;
+	}
+
+#undef mode_set
+
+	if (iattr->ia_valid & ATTR_SIZE) {
+		memset(&info, 0, sizeof(info));
+		info.size = iattr->ia_size;
+		buf_len = sizeof(info);
+		err = vboxsf_fsinfo(sbi->root, params.handle,
+				    SHFL_INFO_SET | SHFL_INFO_SIZE, &buf_len,
+				    &info);
+		if (err) {
+			vboxsf_close(sbi->root, params.handle);
+			return err;
+		}
+
+		/* the host may have given us different attr then requested */
+		sf_i->force_restat = 1;
+	}
+
+	vboxsf_close(sbi->root, params.handle);
+
+	/* Update the inode with what the host has actually given us. */
+	if (sf_i->force_restat)
+		vboxsf_inode_revalidate(dentry);
+
+	return 0;
+}
+
+/*
+ * [dentry] contains string encoded in coding system that corresponds
+ * to [sbi]->nls, we must convert it to UTF8 here.
+ * Returns a shfl_string allocated through __getname (must be freed using
+ * __putname), or an ERR_PTR on error.
+ */
+struct shfl_string *vboxsf_path_from_dentry(struct vboxsf_sbi *sbi,
+					    struct dentry *dentry)
+{
+	struct shfl_string *shfl_path;
+	int path_len, out_len, nb;
+	char *buf, *path;
+	wchar_t uni;
+	u8 *out;
+
+	buf = __getname();
+	if (!buf)
+		return ERR_PTR(-ENOMEM);
+
+	path = dentry_path_raw(dentry, buf, PATH_MAX);
+	if (IS_ERR(path)) {
+		__putname(buf);
+		return ERR_CAST(path);
+	}
+	path_len = strlen(path);
+
+	if (sbi->nls) {
+		shfl_path = __getname();
+		if (!shfl_path) {
+			__putname(buf);
+			return ERR_PTR(-ENOMEM);
+		}
+
+		out = shfl_path->string.utf8;
+		out_len = PATH_MAX - SHFLSTRING_HEADER_SIZE - 1;
+
+		while (path_len) {
+			nb = sbi->nls->char2uni(path, path_len, &uni);
+			if (nb < 0) {
+				__putname(shfl_path);
+				__putname(buf);
+				return ERR_PTR(-EINVAL);
+			}
+			path += nb;
+			path_len -= nb;
+
+			nb = utf32_to_utf8(uni, out, out_len);
+			if (nb < 0) {
+				__putname(shfl_path);
+				__putname(buf);
+				return ERR_PTR(-ENAMETOOLONG);
+			}
+			out += nb;
+			out_len -= nb;
+		}
+		*out = 0;
+		shfl_path->length = out - shfl_path->string.utf8;
+		shfl_path->size = shfl_path->length + 1;
+		__putname(buf);
+	} else {
+		if ((SHFLSTRING_HEADER_SIZE + path_len + 1) > PATH_MAX) {
+			__putname(buf);
+			return ERR_PTR(-ENAMETOOLONG);
+		}
+		/*
+		 * dentry_path stores the name at the end of buf, but the
+		 * shfl_string string we return must be properly aligned.
+		 */
+		shfl_path = (struct shfl_string *)buf;
+		memmove(shfl_path->string.utf8, path, path_len);
+		shfl_path->string.utf8[path_len] = 0;
+		shfl_path->length = path_len;
+		shfl_path->size = path_len + 1;
+	}
+
+	return shfl_path;
+}
+
+int vboxsf_nlscpy(struct vboxsf_sbi *sbi, char *name, size_t name_bound_len,
+		  const unsigned char *utf8_name, size_t utf8_len)
+{
+	const char *in;
+	char *out;
+	size_t out_bound_len;
+	size_t in_bound_len;
+
+	in = utf8_name;
+	in_bound_len = utf8_len;
+
+	out = name;
+	/* Reserve space for terminating 0 */
+	out_bound_len = name_bound_len - 1;
+
+	while (in_bound_len) {
+		int nb;
+		unicode_t uni;
+
+		nb = utf8_to_utf32(in, in_bound_len, &uni);
+		if (nb < 0)
+			return -EINVAL;
+
+		in += nb;
+		in_bound_len -= nb;
+
+		nb = sbi->nls->uni2char(uni, out, out_bound_len);
+		if (nb < 0)
+			return nb;
+
+		out += nb;
+		out_bound_len -= nb;
+	}
+
+	*out = 0;
+
+	return 0;
+}
+
+static struct vboxsf_dir_buf *vboxsf_dir_buf_alloc(struct list_head *list)
+{
+	struct vboxsf_dir_buf *b;
+
+	b = kmalloc_obj(*b);
+	if (!b)
+		return NULL;
+
+	b->buf = kmalloc(DIR_BUFFER_SIZE, GFP_KERNEL);
+	if (!b->buf) {
+		kfree(b);
+		return NULL;
+	}
+
+	b->entries = 0;
+	b->used = 0;
+	b->free = DIR_BUFFER_SIZE;
+	list_add(&b->head, list);
+
+	return b;
+}
+
+static void vboxsf_dir_buf_free(struct vboxsf_dir_buf *b)
+{
+	list_del(&b->head);
+	kfree(b->buf);
+	kfree(b);
+}
+
+struct vboxsf_dir_info *vboxsf_dir_info_alloc(void)
+{
+	struct vboxsf_dir_info *p;
+
+	p = kmalloc_obj(*p);
+	if (!p)
+		return NULL;
+
+	INIT_LIST_HEAD(&p->info_list);
+	return p;
+}
+
+void vboxsf_dir_info_free(struct vboxsf_dir_info *p)
+{
+	struct list_head *list, *pos, *tmp;
+
+	list = &p->info_list;
+	list_for_each_safe(pos, tmp, list) {
+		struct vboxsf_dir_buf *b;
+
+		b = list_entry(pos, struct vboxsf_dir_buf, head);
+		vboxsf_dir_buf_free(b);
+	}
+	kfree(p);
+}
+
+int vboxsf_dir_read_all(struct vboxsf_sbi *sbi, struct vboxsf_dir_info *sf_d,
+			u64 handle)
+{
+	struct vboxsf_dir_buf *b;
+	u32 entries, size;
+	int err = 0;
+	void *buf;
+
+	/* vboxsf_dirinfo returns 1 on end of dir */
+	while (err == 0) {
+		b = vboxsf_dir_buf_alloc(&sf_d->info_list);
+		if (!b) {
+			err = -ENOMEM;
+			break;
+		}
+
+		buf = b->buf;
+		size = b->free;
+
+		err = vboxsf_dirinfo(sbi->root, handle, NULL, 0, 0,
+				     &size, buf, &entries);
+		if (err < 0)
+			break;
+
+		b->entries += entries;
+		b->free -= size;
+		b->used += size;
+	}
+
+	if (b && b->used == 0)
+		vboxsf_dir_buf_free(b);
+
+	/* -EILSEQ means the host could not translate a filename, ignore */
+	if (err > 0 || err == -EILSEQ)
+		err = 0;
+
+	return err;
+}
+
+int vboxsf_query_case_sensitive(struct vboxsf_sbi *sbi)
+{
+	struct shfl_volinfo volinfo = {};
+	u32 buf_len;
+	int err;
+
+	buf_len = sizeof(volinfo);
+	err = vboxsf_fsinfo(sbi->root, 0, SHFL_INFO_GET | SHFL_INFO_VOLUME,
+			    &buf_len, &volinfo);
+	if (err)
+		return err;
+	if (buf_len < sizeof(volinfo))
+		return 0;
+
+	sbi->case_insensitive = !volinfo.properties.case_sensitive;
+	return 0;
+}
+
+int vboxsf_fileattr_get(struct dentry *dentry, struct file_kattr *fa)
+{
+	struct vboxsf_sbi *sbi = VBOXSF_SBI(dentry->d_sb);
+
+	if (sbi->case_insensitive) {
+		fa->fsx_xflags |= FS_XFLAG_CASEFOLD;
+		fa->flags |= FS_CASEFOLD_FL;
+	}
+	return 0;
+}