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, 431 insertions, 0 deletions

diff --git a/fs/orangefs/namei.c b/fs/orangefs/namei.c
new file mode 100644
index 000000000..8ebc34e11
--- /dev/null
+++ b/fs/orangefs/namei.c
@@ -0,0 +1,431 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * (C) 2001 Clemson University and The University of Chicago
+ *
+ * See COPYING in top-level directory.
+ */
+
+/*
+ *  Linux VFS namei operations.
+ */
+
+#include "protocol.h"
+#include "orangefs-kernel.h"
+
+/*
+ * Get a newly allocated inode to go with a negative dentry.
+ */
+static int orangefs_create(struct mnt_idmap *idmap,
+			struct inode *dir,
+			struct dentry *dentry,
+			umode_t mode)
+{
+	struct orangefs_inode_s *parent = ORANGEFS_I(dir);
+	struct orangefs_kernel_op_s *new_op;
+	struct orangefs_object_kref ref;
+	struct inode *inode;
+	struct iattr iattr;
+	int ret;
+
+	gossip_debug(GOSSIP_NAME_DEBUG, "%s: %pd\n",
+		     __func__,
+		     dentry);
+
+	new_op = op_alloc(ORANGEFS_VFS_OP_CREATE);
+	if (!new_op)
+		return -ENOMEM;
+
+	new_op->upcall.req.create.parent_refn = parent->refn;
+
+	fill_default_sys_attrs(new_op->upcall.req.create.attributes, mode);
+
+	strscpy(new_op->upcall.req.create.d_name, dentry->d_name.name);
+
+	ret = service_operation(new_op, __func__, get_interruptible_flag(dir));
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: %pd: handle:%pU: fsid:%d: new_op:%p: ret:%d:\n",
+		     __func__,
+		     dentry,
+		     &new_op->downcall.resp.create.refn.khandle,
+		     new_op->downcall.resp.create.refn.fs_id,
+		     new_op,
+		     ret);
+
+	if (ret < 0)
+		goto out;
+
+	ref = new_op->downcall.resp.create.refn;
+
+	inode = orangefs_new_inode(dir->i_sb, dir, S_IFREG | mode, 0, &ref);
+	if (IS_ERR(inode)) {
+		gossip_err("%s: Failed to allocate inode for file :%pd:\n",
+			   __func__,
+			   dentry);
+		ret = PTR_ERR(inode);
+		goto out;
+	}
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: Assigned inode :%pU: for file :%pd:\n",
+		     __func__,
+		     get_khandle_from_ino(inode),
+		     dentry);
+
+	d_instantiate_new(dentry, inode);
+	orangefs_set_timeout(dentry);
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: dentry instantiated for %pd\n",
+		     __func__,
+		     dentry);
+
+	memset(&iattr, 0, sizeof iattr);
+	iattr.ia_valid |= ATTR_MTIME | ATTR_CTIME;
+	iattr.ia_mtime = iattr.ia_ctime = current_time(dir);
+	__orangefs_setattr(dir, &iattr);
+	ret = 0;
+out:
+	op_release(new_op);
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: %pd: returning %d\n",
+		     __func__,
+		     dentry,
+		     ret);
+	return ret;
+}
+
+/*
+ * Attempt to resolve an object name (dentry->d_name), parent handle, and
+ * fsid into a handle for the object.
+ */
+static struct dentry *orangefs_lookup(struct inode *dir, struct dentry *dentry,
+				   unsigned int flags)
+{
+	struct orangefs_inode_s *parent = ORANGEFS_I(dir);
+	struct orangefs_kernel_op_s *new_op;
+	struct inode *inode;
+	int ret = -EINVAL;
+
+	/*
+	 * in theory we could skip a lookup here (if the intent is to
+	 * create) in order to avoid a potentially failed lookup, but
+	 * leaving it in can skip a valid lookup and try to create a file
+	 * that already exists (e.g. the vfs already handles checking for
+	 * -EEXIST on O_EXCL opens, which is broken if we skip this lookup
+	 * in the create path)
+	 */
+	gossip_debug(GOSSIP_NAME_DEBUG, "%s called on %pd\n",
+		     __func__, dentry);
+
+	if (dentry->d_name.len > (ORANGEFS_NAME_MAX - 1))
+		return ERR_PTR(-ENAMETOOLONG);
+
+	new_op = op_alloc(ORANGEFS_VFS_OP_LOOKUP);
+	if (!new_op)
+		return ERR_PTR(-ENOMEM);
+
+	new_op->upcall.req.lookup.sym_follow = ORANGEFS_LOOKUP_LINK_NO_FOLLOW;
+
+	gossip_debug(GOSSIP_NAME_DEBUG, "%s:%s:%d using parent %pU\n",
+		     __FILE__,
+		     __func__,
+		     __LINE__,
+		     &parent->refn.khandle);
+	new_op->upcall.req.lookup.parent_refn = parent->refn;
+
+	strscpy(new_op->upcall.req.lookup.d_name, dentry->d_name.name);
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: doing lookup on %s under %pU,%d\n",
+		     __func__,
+		     new_op->upcall.req.lookup.d_name,
+		     &new_op->upcall.req.lookup.parent_refn.khandle,
+		     new_op->upcall.req.lookup.parent_refn.fs_id);
+
+	ret = service_operation(new_op, __func__, get_interruptible_flag(dir));
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Lookup Got %pU, fsid %d (ret=%d)\n",
+		     &new_op->downcall.resp.lookup.refn.khandle,
+		     new_op->downcall.resp.lookup.refn.fs_id,
+		     ret);
+
+	if (ret == 0) {
+		orangefs_set_timeout(dentry);
+		inode = orangefs_iget(dir->i_sb, &new_op->downcall.resp.lookup.refn);
+	} else if (ret == -ENOENT) {
+		inode = NULL;
+	} else {
+		/* must be a non-recoverable error */
+		inode = ERR_PTR(ret);
+	}
+
+	op_release(new_op);
+	return d_splice_alias(inode, dentry);
+}
+
+/* return 0 on success; non-zero otherwise */
+static int orangefs_unlink(struct inode *dir, struct dentry *dentry)
+{
+	struct inode *inode = dentry->d_inode;
+	struct orangefs_inode_s *parent = ORANGEFS_I(dir);
+	struct orangefs_kernel_op_s *new_op;
+	struct iattr iattr;
+	int ret;
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: called on %pd\n"
+		     "  (inode %pU): Parent is %pU | fs_id %d\n",
+		     __func__,
+		     dentry,
+		     get_khandle_from_ino(inode),
+		     &parent->refn.khandle,
+		     parent->refn.fs_id);
+
+	new_op = op_alloc(ORANGEFS_VFS_OP_REMOVE);
+	if (!new_op)
+		return -ENOMEM;
+
+	new_op->upcall.req.remove.parent_refn = parent->refn;
+	strscpy(new_op->upcall.req.remove.d_name, dentry->d_name.name);
+
+	ret = service_operation(new_op, "orangefs_unlink",
+				get_interruptible_flag(inode));
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "%s: service_operation returned:%d:\n",
+		     __func__,
+		     ret);
+
+	op_release(new_op);
+
+	if (!ret) {
+		drop_nlink(inode);
+
+		memset(&iattr, 0, sizeof iattr);
+		iattr.ia_valid |= ATTR_MTIME | ATTR_CTIME;
+		iattr.ia_mtime = iattr.ia_ctime = current_time(dir);
+		__orangefs_setattr(dir, &iattr);
+	}
+	return ret;
+}
+
+static int orangefs_symlink(struct mnt_idmap *idmap,
+		         struct inode *dir,
+			 struct dentry *dentry,
+			 const char *symname)
+{
+	struct orangefs_inode_s *parent = ORANGEFS_I(dir);
+	struct orangefs_kernel_op_s *new_op;
+	struct orangefs_object_kref ref;
+	struct inode *inode;
+	struct iattr iattr;
+	int mode = 0755;
+	int ret;
+
+	gossip_debug(GOSSIP_NAME_DEBUG, "%s: called\n", __func__);
+
+	if (!symname)
+		return -EINVAL;
+
+	if (strlen(symname)+1 > ORANGEFS_NAME_MAX)
+		return -ENAMETOOLONG;
+
+	new_op = op_alloc(ORANGEFS_VFS_OP_SYMLINK);
+	if (!new_op)
+		return -ENOMEM;
+
+	new_op->upcall.req.sym.parent_refn = parent->refn;
+
+	fill_default_sys_attrs(new_op->upcall.req.sym.attributes, mode);
+
+	strscpy(new_op->upcall.req.sym.entry_name, dentry->d_name.name);
+	strscpy(new_op->upcall.req.sym.target, symname);
+
+	ret = service_operation(new_op, __func__, get_interruptible_flag(dir));
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Symlink Got ORANGEFS handle %pU on fsid %d (ret=%d)\n",
+		     &new_op->downcall.resp.sym.refn.khandle,
+		     new_op->downcall.resp.sym.refn.fs_id, ret);
+
+	if (ret < 0) {
+		gossip_debug(GOSSIP_NAME_DEBUG,
+			    "%s: failed with error code %d\n",
+			    __func__, ret);
+		goto out;
+	}
+
+	ref = new_op->downcall.resp.sym.refn;
+
+	inode = orangefs_new_inode(dir->i_sb, dir, S_IFLNK | mode, 0, &ref);
+	if (IS_ERR(inode)) {
+		gossip_err
+		    ("*** Failed to allocate orangefs symlink inode\n");
+		ret = PTR_ERR(inode);
+		goto out;
+	}
+	/*
+	 * This is necessary because orangefs_inode_getattr will not
+	 * re-read symlink size as it is impossible for it to change.
+	 * Invalidating the cache does not help.  orangefs_new_inode
+	 * does not set the correct size (it does not know symname).
+	 */
+	inode->i_size = strlen(symname);
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Assigned symlink inode new number of %pU\n",
+		     get_khandle_from_ino(inode));
+
+	d_instantiate_new(dentry, inode);
+	orangefs_set_timeout(dentry);
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Inode (Symlink) %pU -> %pd\n",
+		     get_khandle_from_ino(inode),
+		     dentry);
+
+	memset(&iattr, 0, sizeof iattr);
+	iattr.ia_valid |= ATTR_MTIME | ATTR_CTIME;
+	iattr.ia_mtime = iattr.ia_ctime = current_time(dir);
+	__orangefs_setattr(dir, &iattr);
+	ret = 0;
+out:
+	op_release(new_op);
+	return ret;
+}
+
+static struct dentry *orangefs_mkdir(struct mnt_idmap *idmap, struct inode *dir,
+				     struct dentry *dentry, umode_t mode)
+{
+	struct orangefs_inode_s *parent = ORANGEFS_I(dir);
+	struct orangefs_kernel_op_s *new_op;
+	struct orangefs_object_kref ref;
+	struct inode *inode;
+	struct iattr iattr;
+	int ret;
+
+	new_op = op_alloc(ORANGEFS_VFS_OP_MKDIR);
+	if (!new_op)
+		return ERR_PTR(-ENOMEM);
+
+	new_op->upcall.req.mkdir.parent_refn = parent->refn;
+
+	fill_default_sys_attrs(new_op->upcall.req.mkdir.attributes, mode);
+
+	strscpy(new_op->upcall.req.mkdir.d_name, dentry->d_name.name);
+
+	ret = service_operation(new_op, __func__, get_interruptible_flag(dir));
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Mkdir Got ORANGEFS handle %pU on fsid %d\n",
+		     &new_op->downcall.resp.mkdir.refn.khandle,
+		     new_op->downcall.resp.mkdir.refn.fs_id);
+
+	if (ret < 0) {
+		gossip_debug(GOSSIP_NAME_DEBUG,
+			     "%s: failed with error code %d\n",
+			     __func__, ret);
+		goto out;
+	}
+
+	ref = new_op->downcall.resp.mkdir.refn;
+
+	inode = orangefs_new_inode(dir->i_sb, dir, mode, 0, &ref);
+	if (IS_ERR(inode)) {
+		gossip_err("*** Failed to allocate orangefs dir inode\n");
+		ret = PTR_ERR(inode);
+		goto out;
+	}
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Assigned dir inode new number of %pU\n",
+		     get_khandle_from_ino(inode));
+
+	d_instantiate_new(dentry, inode);
+	orangefs_set_timeout(dentry);
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "Inode (Directory) %pU -> %pd\n",
+		     get_khandle_from_ino(inode),
+		     dentry);
+
+	/*
+	 * NOTE: we have no good way to keep nlink consistent for directories
+	 * across clients; keep constant at 1.
+	 */
+	memset(&iattr, 0, sizeof iattr);
+	iattr.ia_valid |= ATTR_MTIME | ATTR_CTIME;
+	iattr.ia_mtime = iattr.ia_ctime = current_time(dir);
+	__orangefs_setattr(dir, &iattr);
+out:
+	op_release(new_op);
+	return ret ? ERR_PTR(ret) : NULL;
+}
+
+static int orangefs_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 orangefs_kernel_op_s *new_op;
+	struct iattr iattr;
+	int ret;
+
+	if (flags)
+		return -EINVAL;
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "orangefs_rename: called (%pd2 => %pd2) ct=%d\n",
+		     old_dentry, new_dentry, d_count(new_dentry));
+
+	memset(&iattr, 0, sizeof iattr);
+	iattr.ia_valid |= ATTR_MTIME | ATTR_CTIME;
+	iattr.ia_mtime = iattr.ia_ctime = current_time(new_dir);
+	__orangefs_setattr(new_dir, &iattr);
+
+	new_op = op_alloc(ORANGEFS_VFS_OP_RENAME);
+	if (!new_op)
+		return -EINVAL;
+
+	new_op->upcall.req.rename.old_parent_refn = ORANGEFS_I(old_dir)->refn;
+	new_op->upcall.req.rename.new_parent_refn = ORANGEFS_I(new_dir)->refn;
+
+	strscpy(new_op->upcall.req.rename.d_old_name, old_dentry->d_name.name);
+	strscpy(new_op->upcall.req.rename.d_new_name, new_dentry->d_name.name);
+
+	ret = service_operation(new_op,
+				"orangefs_rename",
+				get_interruptible_flag(old_dentry->d_inode));
+
+	gossip_debug(GOSSIP_NAME_DEBUG,
+		     "orangefs_rename: got downcall status %d\n",
+		     ret);
+
+	if (new_dentry->d_inode)
+		inode_set_ctime_current(d_inode(new_dentry));
+
+	op_release(new_op);
+	return ret;
+}
+
+/* ORANGEFS implementation of VFS inode operations for directories */
+const struct inode_operations orangefs_dir_inode_operations = {
+	.lookup = orangefs_lookup,
+	.get_inode_acl = orangefs_get_acl,
+	.set_acl = orangefs_set_acl,
+	.create = orangefs_create,
+	.unlink = orangefs_unlink,
+	.symlink = orangefs_symlink,
+	.mkdir = orangefs_mkdir,
+	.rmdir = orangefs_unlink,
+	.rename = orangefs_rename,
+	.setattr = orangefs_setattr,
+	.getattr = orangefs_getattr,
+	.listxattr = orangefs_listxattr,
+	.permission = orangefs_permission,
+	.update_time = orangefs_update_time,
+};