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, 289 insertions, 0 deletions

diff --git a/fs/afs/dir_silly.c b/fs/afs/dir_silly.c
new file mode 100644
index 000000000..982bb6ec1
--- /dev/null
+++ b/fs/afs/dir_silly.c
@@ -0,0 +1,289 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* AFS silly rename handling
+ *
+ * Copyright (C) 2019 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ * - Derived from NFS's sillyrename.
+ */
+
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/namei.h>
+#include <linux/fsnotify.h>
+#include "internal.h"
+
+static void afs_silly_rename_success(struct afs_operation *op)
+{
+	_enter("op=%08x", op->debug_id);
+
+	afs_check_dir_conflict(op, &op->file[0]);
+	afs_vnode_commit_status(op, &op->file[0]);
+}
+
+static void afs_silly_rename_edit_dir(struct afs_operation *op)
+{
+	struct afs_vnode_param *dvp = &op->file[0];
+	struct afs_vnode *dvnode = dvp->vnode;
+	struct afs_vnode *vnode = AFS_FS_I(d_inode(op->dentry));
+	struct dentry *old = op->dentry;
+	struct dentry *new = op->dentry_2;
+
+	spin_lock(&old->d_lock);
+	old->d_flags |= DCACHE_NFSFS_RENAMED;
+	spin_unlock(&old->d_lock);
+	if (dvnode->silly_key != op->key) {
+		key_put(dvnode->silly_key);
+		dvnode->silly_key = key_get(op->key);
+	}
+
+	down_write(&dvnode->validate_lock);
+	if (test_bit(AFS_VNODE_DIR_VALID, &dvnode->flags) &&
+	    dvnode->status.data_version == dvp->dv_before + dvp->dv_delta) {
+		afs_edit_dir_remove(dvnode, &old->d_name,
+				    afs_edit_dir_for_silly_0);
+		afs_edit_dir_add(dvnode, &new->d_name,
+				 &vnode->fid, afs_edit_dir_for_silly_1);
+	}
+	up_write(&dvnode->validate_lock);
+}
+
+static const struct afs_operation_ops afs_silly_rename_operation = {
+	.issue_afs_rpc	= afs_fs_rename,
+	.issue_yfs_rpc	= yfs_fs_rename,
+	.success	= afs_silly_rename_success,
+	.edit_dir	= afs_silly_rename_edit_dir,
+};
+
+/*
+ * Actually perform the silly rename step.
+ */
+static int afs_do_silly_rename(struct afs_vnode *dvnode, struct afs_vnode *vnode,
+			       struct dentry *old, struct dentry *new,
+			       struct key *key)
+{
+	struct afs_operation *op;
+
+	_enter("%pd,%pd", old, new);
+
+	op = afs_alloc_operation(key, dvnode->volume);
+	if (IS_ERR(op))
+		return PTR_ERR(op);
+
+	op->more_files = kvzalloc_objs(struct afs_vnode_param, 2);
+	if (!op->more_files) {
+		afs_put_operation(op);
+		return -ENOMEM;
+	}
+
+	afs_op_set_vnode(op, 0, dvnode);
+	afs_op_set_vnode(op, 1, dvnode);
+	op->file[0].dv_delta = 1;
+	op->file[1].dv_delta = 1;
+	op->file[0].modification = true;
+	op->file[1].modification = true;
+	op->file[0].update_ctime = true;
+	op->file[1].update_ctime = true;
+	op->more_files[0].vnode		= AFS_FS_I(d_inode(old));
+	op->more_files[0].speculative	= true;
+	op->more_files[1].vnode		= AFS_FS_I(d_inode(new));
+	op->more_files[1].speculative	= true;
+	op->nr_files = 4;
+
+	op->dentry		= old;
+	op->dentry_2		= new;
+	op->ops			= &afs_silly_rename_operation;
+
+	trace_afs_silly_rename(vnode, false);
+	return afs_do_sync_operation(op);
+}
+
+/*
+ * Perform silly-rename of a dentry.
+ *
+ * AFS is stateless and the server doesn't know when the client is holding a
+ * file open.  To prevent application problems when a file is unlinked while
+ * it's still open, the client performs a "silly-rename".  That is, it renames
+ * the file to a hidden file in the same directory, and only performs the
+ * unlink once the last reference to it is put.
+ *
+ * The final cleanup is done during dentry_iput.
+ */
+int afs_sillyrename(struct afs_vnode *dvnode, struct afs_vnode *vnode,
+		    struct dentry *dentry, struct key *key)
+{
+	static unsigned int sillycounter;
+	struct dentry *sdentry = NULL;
+	unsigned char silly[16];
+	int ret = -EBUSY;
+
+	_enter("");
+
+	/* We don't allow a dentry to be silly-renamed twice. */
+	if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
+		return -EBUSY;
+
+	sdentry = NULL;
+	do {
+		dput(sdentry);
+		sillycounter++;
+
+		/* Create a silly name.  Note that the ".__afs" prefix is
+		 * understood by the salvager and must not be changed.
+		 */
+		scnprintf(silly, sizeof(silly), ".__afs%04X", sillycounter);
+		sdentry = lookup_noperm(&QSTR(silly), dentry->d_parent);
+
+		/* N.B. Better to return EBUSY here ... it could be dangerous
+		 * to delete the file while it's in use.
+		 */
+		if (IS_ERR(sdentry))
+			goto out;
+	} while (!d_is_negative(sdentry));
+
+	ihold(&vnode->netfs.inode);
+
+	ret = afs_do_silly_rename(dvnode, vnode, dentry, sdentry, key);
+	switch (ret) {
+	case 0:
+		/* The rename succeeded. */
+		set_bit(AFS_VNODE_SILLY_DELETED, &vnode->flags);
+		d_move(dentry, sdentry);
+		break;
+	case -ERESTARTSYS:
+		/* The result of the rename is unknown. Play it safe by forcing
+		 * a new lookup.
+		 */
+		d_drop(dentry);
+		d_drop(sdentry);
+	}
+
+	iput(&vnode->netfs.inode);
+	dput(sdentry);
+out:
+	_leave(" = %d", ret);
+	return ret;
+}
+
+static void afs_silly_unlink_success(struct afs_operation *op)
+{
+	_enter("op=%08x", op->debug_id);
+	afs_check_dir_conflict(op, &op->file[0]);
+	afs_vnode_commit_status(op, &op->file[0]);
+	afs_vnode_commit_status(op, &op->file[1]);
+	afs_update_dentry_version(op, &op->file[0], op->dentry);
+}
+
+static void afs_silly_unlink_edit_dir(struct afs_operation *op)
+{
+	struct afs_vnode_param *dvp = &op->file[0];
+	struct afs_vnode *dvnode = dvp->vnode;
+
+	_enter("op=%08x", op->debug_id);
+	down_write(&dvnode->validate_lock);
+	if (test_bit(AFS_VNODE_DIR_VALID, &dvnode->flags) &&
+	    dvnode->status.data_version == dvp->dv_before + dvp->dv_delta)
+		afs_edit_dir_remove(dvnode, &op->dentry->d_name,
+				    afs_edit_dir_for_unlink);
+	up_write(&dvnode->validate_lock);
+}
+
+static const struct afs_operation_ops afs_silly_unlink_operation = {
+	.issue_afs_rpc	= afs_fs_remove_file,
+	.issue_yfs_rpc	= yfs_fs_remove_file,
+	.success	= afs_silly_unlink_success,
+	.aborted	= afs_check_for_remote_deletion,
+	.edit_dir	= afs_silly_unlink_edit_dir,
+};
+
+/*
+ * Tell the server to remove a sillyrename file.
+ */
+static int afs_do_silly_unlink(struct afs_vnode *dvnode, struct afs_vnode *vnode,
+			       struct dentry *dentry, struct key *key)
+{
+	struct afs_operation *op;
+
+	_enter("");
+
+	op = afs_alloc_operation(NULL, dvnode->volume);
+	if (IS_ERR(op))
+		return PTR_ERR(op);
+
+	afs_op_set_vnode(op, 0, dvnode);
+	afs_op_set_vnode(op, 1, vnode);
+	op->file[0].dv_delta = 1;
+	op->file[0].modification = true;
+	op->file[0].update_ctime = true;
+	op->file[1].op_unlinked = true;
+	op->file[1].update_ctime = true;
+
+	op->dentry	= dentry;
+	op->ops		= &afs_silly_unlink_operation;
+
+	trace_afs_silly_rename(vnode, true);
+	afs_begin_vnode_operation(op);
+	afs_wait_for_operation(op);
+
+	/* If there was a conflict with a third party, check the status of the
+	 * unlinked vnode.
+	 */
+	if (op->cumul_error.error == 0 && (op->flags & AFS_OPERATION_DIR_CONFLICT)) {
+		op->file[1].update_ctime = false;
+		op->fetch_status.which = 1;
+		op->ops = &afs_fetch_status_operation;
+		afs_begin_vnode_operation(op);
+		afs_wait_for_operation(op);
+	}
+
+	return afs_put_operation(op);
+}
+
+/*
+ * Remove sillyrename file on iput.
+ */
+int afs_silly_iput(struct dentry *dentry, struct inode *inode)
+{
+	struct afs_vnode *dvnode = AFS_FS_I(d_inode(dentry->d_parent));
+	struct afs_vnode *vnode = AFS_FS_I(inode);
+	struct dentry *alias;
+	int ret;
+
+	_enter("%p{%pd},%llx", dentry, dentry, vnode->fid.vnode);
+
+	down_read(&dvnode->rmdir_lock);
+
+	alias = d_alloc_parallel(dentry->d_parent, &dentry->d_name);
+	if (IS_ERR(alias)) {
+		up_read(&dvnode->rmdir_lock);
+		return 0;
+	}
+
+	if (!d_in_lookup(alias)) {
+		/* We raced with lookup...  See if we need to transfer the
+		 * sillyrename information to the aliased dentry.
+		 */
+		ret = 0;
+		spin_lock(&alias->d_lock);
+		if (d_really_is_positive(alias) &&
+		    !(alias->d_flags & DCACHE_NFSFS_RENAMED)) {
+			alias->d_flags |= DCACHE_NFSFS_RENAMED;
+			ret = 1;
+		}
+		spin_unlock(&alias->d_lock);
+		up_read(&dvnode->rmdir_lock);
+		dput(alias);
+		return ret;
+	}
+
+	/* Stop lock-release from complaining. */
+	spin_lock(&vnode->lock);
+	vnode->lock_state = AFS_VNODE_LOCK_DELETED;
+	trace_afs_flock_ev(vnode, NULL, afs_flock_silly_delete, 0);
+	spin_unlock(&vnode->lock);
+
+	afs_do_silly_unlink(dvnode, vnode, dentry, dvnode->silly_key);
+	up_read(&dvnode->rmdir_lock);
+	d_lookup_done(alias);
+	dput(alias);
+	return 1;
+}