#!/usr/bin/env drgn import argparse import os import sys import typing from typing import Any, Callable, Iterator, Optional, Sequence, Union from drgn import FaultError, Object, Program, TypeKind, cast, container_of from drgn.helpers.linux.cpumask import for_each_possible_cpu from drgn.helpers.linux.fs import ( d_path, fget, for_each_file, for_each_mount, inode_path, mount_dst, ) from drgn.helpers.linux.idr import idr_for_each_entry from drgn.helpers.linux.list import ( hlist_for_each_entry, list_empty, list_for_each_entry, ) from drgn.helpers.linux.mm import for_each_vma from drgn.helpers.linux.percpu import per_cpu_ptr from drgn.helpers.linux.pid import find_task, for_each_task from drgn.helpers.linux.plist import plist_for_each_entry from drgn.helpers.linux.rbtree import rbtree_inorder_for_each_entry class warn_on_fault: def __init__(self, message: Union[str, Callable[[], str]]) -> None: self._message = message def __enter__(self) -> None: pass def __exit__(self, exc_type: Any, exc_value: Any, traceback: Any) -> bool: if exc_type is not None and issubclass(exc_type, FaultError): message = ( self._message if isinstance(self._message, str) else self._message() ) if message: print( f"warning: fault while {message}, possibly due to race; results may be incomplete", file=sys.stderr, ) return True return False ignore_fault = warn_on_fault("") format_args = { "dereference": False, "symbolize": False, } if typing.TYPE_CHECKING: class Visitor(typing.Protocol): # novermin def visit_file(self, file: Object) -> Optional[str]: ... def visit_inode(self, inode: Object) -> Optional[str]: ... def visit_path(self, path: Object) -> Optional[str]: ... class InodeVisitor: def __init__(self, inode: Object) -> None: self._inode = inode.read_() def visit_file(self, file: Object) -> Optional[str]: if file.f_inode != self._inode: return None return file.format_(**format_args) def visit_inode(self, inode: Object) -> Optional[str]: if inode != self._inode: return None return inode.format_(**format_args) def visit_path(self, path: Object) -> Optional[str]: if path.dentry.d_inode != self._inode: return None return path.format_(**format_args) def _format_file_with_path(file: Object) -> str: match = file.format_(**format_args) with ignore_fault: match += " " + os.fsdecode(d_path(file.f_path)) return match def _format_inode_with_path(inode: Object) -> str: match = inode.format_(**format_args) with ignore_fault: path = inode_path(inode) if path: match += " " + os.fsdecode(path) return match def _format_path_with_path(path: Object) -> str: match = path.format_(**format_args) with ignore_fault: match += " " + os.fsdecode(d_path(path)) return match class SuperBlockVisitor: def __init__(self, sb: Object) -> None: self._sb = sb.read_() def visit_file(self, file: Object) -> Optional[str]: if file.f_inode.i_sb != self._sb: return None return _format_file_with_path(file) def visit_inode(self, inode: Object) -> Optional[str]: if inode.i_sb != self._sb: return None return _format_inode_with_path(inode) def visit_path(self, path: Object) -> Optional[str]: if path.mnt.mnt_sb != self._sb: return None return _format_path_with_path(path) class BtrfsSubvolumeVisitor: def __init__(self, root: Object) -> None: self._root = root.read_() # We compare the super block first to easily filter out non-Btrfs # inodes. self._sb = self._root.fs_info.sb.read_() def visit_file(self, file: Object) -> Optional[str]: f_inode = file.f_inode.read_() if ( f_inode.i_sb != self._sb or container_of(f_inode, "struct btrfs_inode", "vfs_inode").root != self._root ): return None return _format_file_with_path(file) def visit_inode(self, inode: Object) -> Optional[str]: if ( inode.i_sb != self._sb or container_of(inode, "struct btrfs_inode", "vfs_inode").root != self._root ): return None return _format_inode_with_path(inode) def visit_path(self, path: Object) -> Optional[str]: inode = path.dentry.d_inode.read_() if ( inode.i_sb != self._sb or container_of(inode, "struct btrfs_inode", "vfs_inode").root != self._root ): return None return _format_path_with_path(path) def super_block_on_bdev(bdev: Object) -> Optional[Object]: prog = bdev.prog_ try: btrfs_fs_info_type = prog.type("struct btrfs_fs_info *") except LookupError: btrfs_fs_info_type = None for sb in list_for_each_entry( "struct super_block", prog["super_blocks"].address_of_(), "s_list" ): if btrfs_fs_info_type is not None and sb.s_type.name.string_() == b"btrfs": # Btrfs is a special case because of its multi-device support. fs_info = cast(btrfs_fs_info_type, sb.s_fs_info) for device in list_for_each_entry( "struct btrfs_device", fs_info.fs_devices.devices.address_of_(), "dev_list", ): if device.bdev == bdev: return sb elif sb.s_bdev == bdev: return sb return None def visit_tasks( prog: Program, visitor: "Visitor", *, check_mounts: bool, check_tasks: bool ) -> None: check_mounts = check_mounts and isinstance(visitor, SuperBlockVisitor) if check_mounts: init_mnt_ns = prog["init_task"].nsproxy.mnt_ns checked_mnt_ns = {0} with warn_on_fault("iterating tasks"): for task in for_each_task(prog): cached_task_id = None def task_id() -> str: nonlocal cached_task_id if cached_task_id is None: pid = task.pid.value_() comm = os.fsdecode(task.comm.string_()) cached_task_id = f"pid {pid} ({comm})" return cached_task_id def task_fault_warning() -> str: try: return f"checking {task_id()}" except FaultError: return "checking task" with warn_on_fault(task_fault_warning): files: Optional[Object] = task.files.read_() fs: Optional[Object] = task.fs.read_() mm: Optional[Object] = task.mm.read_() # If this task is not the thread group leader, don't bother # checking it again unless it has its own context. group_leader = task.group_leader.read_() if task != group_leader: if files and files == group_leader.files: files = None if fs and fs == group_leader.fs: fs = None if mm and mm == group_leader.mm: mm = None if check_mounts: nsproxy = task.nsproxy.read_() if nsproxy: mnt_ns = nsproxy.mnt_ns.read_() if mnt_ns.value_() not in checked_mnt_ns: for mount in for_each_mount(mnt_ns): with ignore_fault: if mount.mnt.mnt_sb == visitor._sb: # type: ignore [attr-defined] if mnt_ns == init_mnt_ns: mnt_ns_note = "" else: mnt_ns_note = f" (mount namespace {mnt_ns.ns.inum.value_()})" print( f"mount {os.fsdecode(mount_dst(mount))}{mnt_ns_note} " f"{mount.format_(**format_args)}" ) checked_mnt_ns.add(mnt_ns.value_()) if check_tasks: if files: for fd, file in for_each_file(task): with ignore_fault: match = visitor.visit_file(file) if match: print(f"{task_id()} fd {fd} {match}") if fs: with ignore_fault: match = visitor.visit_path(fs.root.address_of_()) if match: print(f"{task_id()} root {match}") with ignore_fault: match = visitor.visit_path(fs.pwd.address_of_()) if match: print(f"{task_id()} cwd {match}") if mm: exe_file = mm.exe_file.read_() if exe_file: match = visitor.visit_file(exe_file) if match: print(f"{task_id()} exe {match}") for vma in for_each_vma(mm): with ignore_fault: file = vma.vm_file.read_() if file: match = visitor.visit_file(file) if match: print( f"{task_id()} vma {hex(vma.vm_start)}-{hex(vma.vm_end)} {match}" ) def visit_binfmt_misc(prog: Program, visitor: "Visitor") -> None: try: Node = prog.type("Node", filename="binfmt_misc.c") except LookupError: # If the Node type doesn't exist, then CONFIG_BINFMT_MISC=n or the # binfmt_misc module isn't loaded. return with warn_on_fault("iterating binfmt_misc instances"): for sb in hlist_for_each_entry( "struct super_block", prog["bm_fs_type"].fs_supers, "s_instances" ): # Since Linux kernel commit 21ca59b365c0 ("binfmt_misc: enable # sandboxed mounts") (in v6.7), each user namespace can have its # own binfmt_misc instance. Before that, there is one global # instance. user_ns = cast("struct user_namespace *", sb.s_fs_info) try: binfmt_misc = user_ns.binfmt_misc except AttributeError: entries = prog.object("entries", filename="binfmt_misc.c") have_user_ns = False else: entries = binfmt_misc.entries have_user_ns = True for node in list_for_each_entry(Node, entries.address_of_(), "list"): with ignore_fault: match = visitor.visit_file(node.interp_file) if match: if have_user_ns and user_ns.level: user_ns_note = ( f" (user namespace {user_ns.ns.inum.value_()})" ) else: user_ns_note = "" print( f"binfmt_misc{user_ns_note} {os.fsdecode(node.name.string_())} {node.format_(**format_args)} {match}" ) def visit_loop_devices(prog: Program, visitor: "Visitor") -> None: try: loop_index_idr = prog["loop_index_idr"] except KeyError: # If loop_index_idr doesn't exist, then CONFIG_BLK_DEV_LOOP=n or the # loop module isn't loaded. return with warn_on_fault("iterating loop devices"): for i, lo in idr_for_each_entry( loop_index_idr.address_of_(), "struct loop_device" ): with ignore_fault: file = lo.lo_backing_file.read_() if file: match = visitor.visit_file(file) if match: print(f"loop device {i} {lo.format_(**format_args)} {match}") def visit_swap_files(prog: Program, visitor: "Visitor") -> None: try: swap_active_head = prog["swap_active_head"] except KeyError: # If swap_active_head doesn't exist, then CONFIG_SWAP=n. return with warn_on_fault("iterating swap files"): for swap_info in plist_for_each_entry( "struct swap_info_struct", swap_active_head.address_of_(), "list" ): with ignore_fault: match = visitor.visit_file(swap_info.swap_file) if match: print(f"swap file {swap_info.format_(**format_args)} {match}") # call was moved from struct trace_probe to struct trace_event Linux kernel # commit 60d53e2c3b75 ("tracing/probe: Split trace_event related data from # trace_probe") (in v5.4). def _trace_probe_call(tp: Object) -> Object: try: event = tp.event except AttributeError: return tp.call return event.call def trace_probe_group_name(tp: Object) -> str: return os.fsdecode(_trace_probe_call(tp).member_("class").system.string_()) def trace_probe_name(tp: Object) -> str: prog = tp.prog_ call = _trace_probe_call(tp).read_() # TRACE_EVENT_FL_CUSTOM was added in Linux kernel commit 3a73333fb370 # ("tracing: Add TRACE_CUSTOM_EVENT() macro") (in v5.18). try: TRACE_EVENT_FL_CUSTOM = prog["TRACE_EVENT_FL_CUSTOM"] except KeyError: pass else: if call.flags & TRACE_EVENT_FL_CUSTOM: return os.fsdecode(call.name.string_()) if call.flags & prog["TRACE_EVENT_FL_TRACEPOINT"]: tracepoint = call.tp.read_() return os.fsdecode(tracepoint.name.string_()) if tracepoint else "" else: return os.fsdecode(call.name.string_()) def visit_uprobes(prog: Program, visitor: "Visitor") -> None: try: uprobes_tree = prog["uprobes_tree"] except KeyError: # If uprobes_tree doesn't exist, then CONFIG_UPROBES=n. return try: uprobe_dispatcher = prog["uprobe_dispatcher"] except KeyError: # uprobe_dispatcher only exists if CONFIG_UPROBE_EVENTS=y, which is # theoretically separate from CONFIG_UPROBES, although as of Linux 6.8 # they will always be the same. uprobe_dispatcher = None with warn_on_fault("iterating uprobes"): for uprobe in rbtree_inorder_for_each_entry( "struct uprobe", uprobes_tree.address_of_(), "rb_node" ): try: match = visitor.visit_inode(uprobe.inode) except FaultError: continue if not match: continue found_consumer = False with warn_on_fault("iterating uprobe consumers"): consumer_list = uprobe.consumers # Since Linux kernel commit cc01bd044e6a ("uprobes: travers # uprobe's consumer list locklessly under SRCU protection") (in # v6.12), struct uprobe::consumers is a struct list_head. # Before that, it was a hand-rolled singly-linked list. if consumer_list.type_.kind == TypeKind.STRUCT: consumers = list_for_each_entry( "struct uprobe_consumer", consumer_list.address_of_(), "cons_node", ) else: def for_each_consumer(first: Object) -> Iterator[Object]: consumer = first.read_() while consumer: yield consumer consumer = consumer.next.read_() consumers = for_each_consumer(consumer_list) for consumer in consumers: handler = consumer.handler.read_() if handler == uprobe_dispatcher: tu = container_of(consumer, "struct trace_uprobe", "consumer") # uprobe events created through tracefs are in a list # anchored on devent.list since Linux kernel commit # 0597c49c69d5 ("tracing/uprobes: Use dyn_event # framework for uprobe events") (in v5.0) and list # before that. try: event_list = tu.devent.list except AttributeError: event_list = tu.list if list_empty(event_list.address_of_()): found_perf_event = False with ignore_fault: call = _trace_probe_call(tu.tp) # uprobes created with perf_event_open have a # struct perf_event in call.perf_events, which # only exists if CONFIG_PERF_EVENTS=y. try: perf_events = call.perf_events except AttributeError: pass else: for cpu in for_each_possible_cpu(prog): for perf_event in hlist_for_each_entry( "struct perf_event", per_cpu_ptr(perf_events, cpu), "hlist_entry", ): owner = perf_event.owner.read_() if owner: owner_pid = owner.pid.value_() owner_comm = os.fsdecode( owner.comm.string_() ) print( f"perf uprobe (owned by pid {owner_pid} ({owner_comm})) {perf_event.format_(**format_args)} {match}" ) else: print( f"perf uprobe (no owner) {perf_event.format_(**format_args)} {match}" ) found_perf_event = True if not found_perf_event: print( f"unknown trace uprobe {tu.format_(**format_args)} {match}" ) else: c = "r" if tu.consumer.ret_handler else "p" group_name = trace_probe_group_name(tu.tp) event_name = trace_probe_name(tu.tp) print( f"uprobe event {c}:{group_name}/{event_name} {tu.format_(**format_args)} {match}" ) else: print( f"unknown uprobe consumer {consumer.format_(**format_args)}" ) found_consumer = True if not found_consumer: print(f"unknown uprobe {uprobe.format_(**format_args)} {match}") def hexint(x: str) -> int: return int(x, 16) def main(prog: Program, argv: Sequence[str]) -> None: parser = argparse.ArgumentParser( description="find what is referencing a filesystem object" ) parser.add_argument( "-L", "--dereference", action="store_true", help="if the given path is a symbolic link, follow it", ) object_group = parser.add_argument_group( title="filesystem object selection" ).add_mutually_exclusive_group(required=True) object_group.add_argument( "--inode", metavar="PATH", help="find references to the inode at the given path" ) object_group.add_argument( "--inode-pointer", metavar="ADDRESS", type=hexint, help="find references to the given struct inode pointer", ) object_group.add_argument( "--super-block", metavar="PATH", help="find references to the filesystem (super block) containing the given path", ) object_group.add_argument( "--super-block-on-block-device", metavar="PATH", help="find references to the filesystem (super block) on the block device at the given path", ) object_group.add_argument( "--super-block-pointer", metavar="ADDRESS", type=hexint, help="find references to the given struct super_block pointer", ) object_group.add_argument( "--btrfs-subvolume", metavar="PATH", help="find references to the Btrfs subvolume containing the given path", ) object_group.add_argument( "--btrfs-subvolume-pointer", metavar="ADDRESS", type=hexint, help="find references to the given struct btrfs_root pointer", ) CHECKS = [ "binfmt_misc", "loop", "mounts", "swap", "tasks", "uprobes", ] check_group = parser.add_argument_group( title="check selection" ).add_mutually_exclusive_group() check_group.add_argument( "--check", choices=CHECKS, action="append", help="only check for references from the given source; may be given multiple times (default: all)", ) check_group.add_argument( "--no-check", choices=CHECKS, action="append", help="don't check for references from the given source; may be given multiple times", ) args = parser.parse_args(argv) visitor: "Visitor" if args.inode is not None: fd = os.open(args.inode, os.O_PATH | (0 if args.dereference else os.O_NOFOLLOW)) try: visitor = InodeVisitor(fget(find_task(prog, os.getpid()), fd).f_inode) finally: os.close(fd) elif args.inode_pointer is not None: visitor = InodeVisitor(Object(prog, "struct inode *", args.inode_pointer)) elif args.super_block is not None: fd = os.open( args.super_block, os.O_PATH | (0 if args.dereference else os.O_NOFOLLOW) ) try: visitor = SuperBlockVisitor( fget(find_task(prog, os.getpid()), fd).f_inode.i_sb ) finally: os.close(fd) elif args.super_block_on_block_device: fd = os.open( args.super_block_on_block_device, # O_PATH doesn't look up the bdev inode, so we can't use it here. os.O_RDONLY | (0 if args.dereference else os.O_NOFOLLOW), ) try: file = fget(find_task(prog, os.getpid()), fd) if file.f_op != prog["def_blk_fops"].address_of_(): sys.exit( f"{args.super_block_on_block_device} is not a block device. Do you need --dereference?" ) sb = super_block_on_bdev( container_of( file.f_mapping.host, "struct bdev_inode", "vfs_inode" ).bdev.address_of_() ) if sb is None: sys.exit(f"no filesystem found on {args.super_block_on_block_device}") visitor = SuperBlockVisitor(sb) finally: os.close(fd) elif args.super_block_pointer is not None: visitor = SuperBlockVisitor( Object(prog, "struct super_block *", args.super_block_pointer) ) elif args.btrfs_subvolume is not None: fd = os.open( args.btrfs_subvolume, os.O_PATH | (0 if args.dereference else os.O_NOFOLLOW) ) try: inode = fget(find_task(prog, os.getpid()), fd).f_inode.read_() # It'd be better to check the filesystem type directly from # userspace using fstatfs(2), but Python doesn't provide # {,f}statfs() (and os.statvfs() doesn't include f_type). if inode.i_sb.s_type.name.string_() != b"btrfs": sys.exit(f"{args.btrfs_subvolume} is not on Btrfs") visitor = BtrfsSubvolumeVisitor( container_of(inode, "struct btrfs_inode", "vfs_inode").root ) finally: os.close(fd) elif args.btrfs_subvolume_pointer is not None: visitor = BtrfsSubvolumeVisitor( Object(prog, "struct btrfs_root *", args.btrfs_subvolume_pointer) ) else: assert False if args.check: enabled_checks = set(args.check) else: enabled_checks = set(CHECKS) if args.no_check: enabled_checks -= set(args.no_check) if "mounts" in enabled_checks or "tasks" in enabled_checks: visit_tasks( prog, visitor, check_mounts="mounts" in enabled_checks, check_tasks="tasks" in enabled_checks, ) if "binfmt_misc" in enabled_checks: visit_binfmt_misc(prog, visitor) if "loop" in enabled_checks: visit_loop_devices(prog, visitor) if "swap" in enabled_checks: visit_swap_files(prog, visitor) if "uprobes" in enabled_checks: visit_uprobes(prog, visitor) if __name__ == "__main__": prog: Program main(prog, sys.argv[1:])