1===============
   2Pathname lookup
   3===============
   4
   5This write-up is based on three articles published at lwn.net:
   6
   7- <https://lwn.net/Articles/649115/> Pathname lookup in Linux
   8- <https://lwn.net/Articles/649729/> RCU-walk: faster pathname lookup in Linux
   9- <https://lwn.net/Articles/650786/> A walk among the symlinks
  10
  11Written by Neil Brown with help from Al Viro and Jon Corbet.
  12It has subsequently been updated to reflect changes in the kernel
  13including:
  14
  15- per-directory parallel name lookup.
  16- ``openat2()`` resolution restriction flags.
  17
  18Introduction to pathname lookup
  19===============================
  20
  21The most obvious aspect of pathname lookup, which very little
  22exploration is needed to discover, is that it is complex.  There are
  23many rules, special cases, and implementation alternatives that all
  24combine to confuse the unwary reader.  Computer science has long been
  25acquainted with such complexity and has tools to help manage it.  One
  26tool that we will make extensive use of is "divide and conquer".  For
  27the early parts of the analysis we will divide off symlinks - leaving
  28them until the final part.  Well before we get to symlinks we have
  29another major division based on the VFS's approach to locking which
  30will allow us to review "REF-walk" and "RCU-walk" separately.  But we
  31are getting ahead of ourselves.  There are some important low level
  32distinctions we need to clarify first.
  33
  34There are two sorts of ...
  35--------------------------
  36
  37.. _openat: http://man7.org/linux/man-pages/man2/openat.2.html
  38
  39Pathnames (sometimes "file names"), used to identify objects in the
  40filesystem, will be familiar to most readers.  They contain two sorts
  41of elements: "slashes" that are sequences of one or more "``/``"
  42characters, and "components" that are sequences of one or more
  43non-"``/``" characters.  These form two kinds of paths.  Those that
  44start with slashes are "absolute" and start from the filesystem root.
  45The others are "relative" and start from the current directory, or
  46from some other location specified by a file descriptor given to
  47"``*at()``" system calls such as `openat() <openat_>`_.
  48
  49.. _execveat: http://man7.org/linux/man-pages/man2/execveat.2.html
  50
  51It is tempting to describe the second kind as starting with a
  52component, but that isn't always accurate: a pathname can lack both
  53slashes and components, it can be empty, in other words.  This is
  54generally forbidden in POSIX, but some of those "``*at()``" system calls
  55in Linux permit it when the ``AT_EMPTY_PATH`` flag is given.  For
  56example, if you have an open file descriptor on an executable file you
  57can execute it by calling `execveat() <execveat_>`_ passing
  58the file descriptor, an empty path, and the ``AT_EMPTY_PATH`` flag.
  59
  60These paths can be divided into two sections: the final component and
  61everything else.  The "everything else" is the easy bit.  In all cases
  62it must identify a directory that already exists, otherwise an error
  63such as ``ENOENT`` or ``ENOTDIR`` will be reported.
  64
  65The final component is not so simple.  Not only do different system
  66calls interpret it quite differently (e.g. some create it, some do
  67not), but it might not even exist: neither the empty pathname nor the
  68pathname that is just slashes have a final component.  If it does
  69exist, it could be "``.``" or "``..``" which are handled quite differently
  70from other components.
  71
  72.. _POSIX: https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_12
  73
  74If a pathname ends with a slash, such as "``/tmp/foo/``" it might be
  75tempting to consider that to have an empty final component.  In many
  76ways that would lead to correct results, but not always.  In
  77particular, ``mkdir()`` and ``rmdir()`` each create or remove a directory named
  78by the final component, and they are required to work with pathnames
  79ending in "``/``".  According to POSIX_:
  80
  81  A pathname that contains at least one non-<slash> character and
  82  that ends with one or more trailing <slash> characters shall not
  83  be resolved successfully unless the last pathname component before
  84  the trailing <slash> characters names an existing directory or a
  85  directory entry that is to be created for a directory immediately
  86  after the pathname is resolved.
  87
  88The Linux pathname walking code (mostly in ``fs/namei.c``) deals with
  89all of these issues: breaking the path into components, handling the
  90"everything else" quite separately from the final component, and
  91checking that the trailing slash is not used where it isn't
  92permitted.  It also addresses the important issue of concurrent
  93access.
  94
  95While one process is looking up a pathname, another might be making
  96changes that affect that lookup.  One fairly extreme case is that if
  97"a/b" were renamed to "a/c/b" while another process were looking up
  98"a/b/..", that process might successfully resolve on "a/c".
  99Most races are much more subtle, and a big part of the task of
 100pathname lookup is to prevent them from having damaging effects.  Many
 101of the possible races are seen most clearly in the context of the
 102"dcache" and an understanding of that is central to understanding
 103pathname lookup.
 104
 105More than just a cache
 106----------------------
 107
 108The "dcache" caches information about names in each filesystem to
 109make them quickly available for lookup.  Each entry (known as a
 110"dentry") contains three significant fields: a component name, a
 111pointer to a parent dentry, and a pointer to the "inode" which
 112contains further information about the object in that parent with
 113the given name.  The inode pointer can be ``NULL`` indicating that the
 114name doesn't exist in the parent.  While there can be linkage in the
 115dentry of a directory to the dentries of the children, that linkage is
 116not used for pathname lookup, and so will not be considered here.
 117
 118The dcache has a number of uses apart from accelerating lookup.  One
 119that will be particularly relevant is that it is closely integrated
 120with the mount table that records which filesystem is mounted where.
 121What the mount table actually stores is which dentry is mounted on top
 122of which other dentry.
 123
 124When considering the dcache, we have another of our "two types"
 125distinctions: there are two types of filesystems.
 126
 127Some filesystems ensure that the information in the dcache is always
 128completely accurate (though not necessarily complete).  This can allow
 129the VFS to determine if a particular file does or doesn't exist
 130without checking with the filesystem, and means that the VFS can
 131protect the filesystem against certain races and other problems.
 132These are typically "local" filesystems such as ext3, XFS, and Btrfs.
 133
 134Other filesystems don't provide that guarantee because they cannot.
 135These are typically filesystems that are shared across a network,
 136whether remote filesystems like NFS and 9P, or cluster filesystems
 137like ocfs2 or cephfs.  These filesystems allow the VFS to revalidate
 138cached information, and must provide their own protection against
 139awkward races.  The VFS can detect these filesystems by the
 140``DCACHE_OP_REVALIDATE`` flag being set in the dentry.
 141
 142REF-walk: simple concurrency management with refcounts and spinlocks
 143--------------------------------------------------------------------
 144
 145With all of those divisions carefully classified, we can now start
 146looking at the actual process of walking along a path.  In particular
 147we will start with the handling of the "everything else" part of a
 148pathname, and focus on the "REF-walk" approach to concurrency
 149management.  This code is found in the ``link_path_walk()`` function, if
 150you ignore all the places that only run when "``LOOKUP_RCU``"
 151(indicating the use of RCU-walk) is set.
 152
 153.. _Meet the Lockers: https://lwn.net/Articles/453685/
 154
 155REF-walk is fairly heavy-handed with locks and reference counts.  Not
 156as heavy-handed as in the old "big kernel lock" days, but certainly not
 157afraid of taking a lock when one is needed.  It uses a variety of
 158different concurrency controls.  A background understanding of the
 159various primitives is assumed, or can be gleaned from elsewhere such
 160as in `Meet the Lockers`_.
 161
 162The locking mechanisms used by REF-walk include:
 163
 164dentry->d_lockref
 165~~~~~~~~~~~~~~~~~
 166
 167This uses the lockref primitive to provide both a spinlock and a
 168reference count.  The special-sauce of this primitive is that the
 169conceptual sequence "lock; inc_ref; unlock;" can often be performed
 170with a single atomic memory operation.
 171
 172Holding a reference on a dentry ensures that the dentry won't suddenly
 173be freed and used for something else, so the values in various fields
 174will behave as expected.  It also protects the ``->d_inode`` reference
 175to the inode to some extent.
 176
 177The association between a dentry and its inode is fairly permanent.
 178For example, when a file is renamed, the dentry and inode move
 179together to the new location.  When a file is created the dentry will
 180initially be negative (i.e. ``d_inode`` is ``NULL``), and will be assigned
 181to the new inode as part of the act of creation.
 182
 183When a file is deleted, this can be reflected in the cache either by
 184setting ``d_inode`` to ``NULL``, or by removing it from the hash table
 185(described shortly) used to look up the name in the parent directory.
 186If the dentry is still in use the second option is used as it is
 187perfectly legal to keep using an open file after it has been deleted
 188and having the dentry around helps.  If the dentry is not otherwise in
 189use (i.e. if the refcount in ``d_lockref`` is one), only then will
 190``d_inode`` be set to ``NULL``.  Doing it this way is more efficient for a
 191very common case.
 192
 193So as long as a counted reference is held to a dentry, a non-``NULL`` ``->d_inode``
 194value will never be changed.
 195
 196dentry->d_lock
 197~~~~~~~~~~~~~~
 198
 199``d_lock`` is a synonym for the spinlock that is part of ``d_lockref`` above.
 200For our purposes, holding this lock protects against the dentry being
 201renamed or unlinked.  In particular, its parent (``d_parent``), and its
 202name (``d_name``) cannot be changed, and it cannot be removed from the
 203dentry hash table.
 204
 205When looking for a name in a directory, REF-walk takes ``d_lock`` on
 206each candidate dentry that it finds in the hash table and then checks
 207that the parent and name are correct.  So it doesn't lock the parent
 208while searching in the cache; it only locks children.
 209
 210When looking for the parent for a given name (to handle "``..``"),
 211REF-walk can take ``d_lock`` to get a stable reference to ``d_parent``,
 212but it first tries a more lightweight approach.  As seen in
 213``dget_parent()``, if a reference can be claimed on the parent, and if
 214subsequently ``d_parent`` can be seen to have not changed, then there is
 215no need to actually take the lock on the child.
 216
 217rename_lock
 218~~~~~~~~~~~
 219
 220Looking up a given name in a given directory involves computing a hash
 221from the two values (the name and the dentry of the directory),
 222accessing that slot in a hash table, and searching the linked list
 223that is found there.
 224
 225When a dentry is renamed, the name and the parent dentry can both
 226change so the hash will almost certainly change too.  This would move the
 227dentry to a different chain in the hash table.  If a filename search
 228happened to be looking at a dentry that was moved in this way,
 229it might end up continuing the search down the wrong chain,
 230and so miss out on part of the correct chain.
 231
 232The name-lookup process (``d_lookup()``) does *not* try to prevent this
 233from happening, but only to detect when it happens.
 234``rename_lock`` is a seqlock that is updated whenever any dentry is
 235renamed.  If ``d_lookup`` finds that a rename happened while it
 236unsuccessfully scanned a chain in the hash table, it simply tries
 237again.
 238
 239``rename_lock`` is also used to detect and defend against potential attacks
 240against ``LOOKUP_BENEATH`` and ``LOOKUP_IN_ROOT`` when resolving ".." (where
 241the parent directory is moved outside the root, bypassing the ``path_equal()``
 242check). If ``rename_lock`` is updated during the lookup and the path encounters
 243a "..", a potential attack occurred and ``handle_dots()`` will bail out with
 244``-EAGAIN``.
 245
 246inode->i_rwsem
 247~~~~~~~~~~~~~~
 248
 249``i_rwsem`` is a read/write semaphore that serializes all changes to a particular
 250directory.  This ensures that, for example, an ``unlink()`` and a ``rename()``
 251cannot both happen at the same time.  It also keeps the directory
 252stable while the filesystem is asked to look up a name that is not
 253currently in the dcache or, optionally, when the list of entries in a
 254directory is being retrieved with ``readdir()``.
 255
 256This has a complementary role to that of ``d_lock``: ``i_rwsem`` on a
 257directory protects all of the names in that directory, while ``d_lock``
 258on a name protects just one name in a directory.  Most changes to the
 259dcache hold ``i_rwsem`` on the relevant directory inode and briefly take
 260``d_lock`` on one or more the dentries while the change happens.  One
 261exception is when idle dentries are removed from the dcache due to
 262memory pressure.  This uses ``d_lock``, but ``i_rwsem`` plays no role.
 263
 264The semaphore affects pathname lookup in two distinct ways.  Firstly it
 265prevents changes during lookup of a name in a directory.  ``walk_component()`` uses
 266``lookup_fast()`` first which, in turn, checks to see if the name is in the cache,
 267using only ``d_lock`` locking.  If the name isn't found, then ``walk_component()``
 268falls back to ``lookup_slow()`` which takes a shared lock on ``i_rwsem``, checks again that
 269the name isn't in the cache, and then calls in to the filesystem to get a
 270definitive answer.  A new dentry will be added to the cache regardless of
 271the result.
 272
 273Secondly, when pathname lookup reaches the final component, it will
 274sometimes need to take an exclusive lock on ``i_rwsem`` before performing the last lookup so
 275that the required exclusion can be achieved.  How path lookup chooses
 276to take, or not take, ``i_rwsem`` is one of the
 277issues addressed in a subsequent section.
 278
 279If two threads attempt to look up the same name at the same time - a
 280name that is not yet in the dcache - the shared lock on ``i_rwsem`` will
 281not prevent them both adding new dentries with the same name.  As this
 282would result in confusion an extra level of interlocking is used,
 283based around a secondary hash table (``in_lookup_hashtable``) and a
 284per-dentry flag bit (``DCACHE_PAR_LOOKUP``).
 285
 286To add a new dentry to the cache while only holding a shared lock on
 287``i_rwsem``, a thread must call ``d_alloc_parallel()``.  This allocates a
 288dentry, stores the required name and parent in it, checks if there
 289is already a matching dentry in the primary or secondary hash
 290tables, and if not, stores the newly allocated dentry in the secondary
 291hash table, with ``DCACHE_PAR_LOOKUP`` set.
 292
 293If a matching dentry was found in the primary hash table then that is
 294returned and the caller can know that it lost a race with some other
 295thread adding the entry.  If no matching dentry is found in either
 296cache, the newly allocated dentry is returned and the caller can
 297detect this from the presence of ``DCACHE_PAR_LOOKUP``.  In this case it
 298knows that it has won any race and now is responsible for asking the
 299filesystem to perform the lookup and find the matching inode.  When
 300the lookup is complete, it must call ``d_lookup_done()`` which clears
 301the flag and does some other house keeping, including removing the
 302dentry from the secondary hash table - it will normally have been
 303added to the primary hash table already.  Note that a ``struct
 304waitqueue_head`` is passed to ``d_alloc_parallel()``, and
 305``d_lookup_done()`` must be called while this ``waitqueue_head`` is still
 306in scope.
 307
 308If a matching dentry is found in the secondary hash table,
 309``d_alloc_parallel()`` has a little more work to do. It first waits for
 310``DCACHE_PAR_LOOKUP`` to be cleared, using a wait_queue that was passed
 311to the instance of ``d_alloc_parallel()`` that won the race and that
 312will be woken by the call to ``d_lookup_done()``.  It then checks to see
 313if the dentry has now been added to the primary hash table.  If it
 314has, the dentry is returned and the caller just sees that it lost any
 315race.  If it hasn't been added to the primary hash table, the most
 316likely explanation is that some other dentry was added instead using
 317``d_splice_alias()``.  In any case, ``d_alloc_parallel()`` repeats all the
 318look ups from the start and will normally return something from the
 319primary hash table.
 320
 321mnt->mnt_count
 322~~~~~~~~~~~~~~
 323
 324``mnt_count`` is a per-CPU reference counter on "``mount``" structures.
 325Per-CPU here means that incrementing the count is cheap as it only
 326uses CPU-local memory, but checking if the count is zero is expensive as
 327it needs to check with every CPU.  Taking a ``mnt_count`` reference
 328prevents the mount structure from disappearing as the result of regular
 329unmount operations, but does not prevent a "lazy" unmount.  So holding
 330``mnt_count`` doesn't ensure that the mount remains in the namespace and,
 331in particular, doesn't stabilize the link to the mounted-on dentry.  It
 332does, however, ensure that the ``mount`` data structure remains coherent,
 333and it provides a reference to the root dentry of the mounted
 334filesystem.  So a reference through ``->mnt_count`` provides a stable
 335reference to the mounted dentry, but not the mounted-on dentry.
 336
 337mount_lock
 338~~~~~~~~~~
 339
 340``mount_lock`` is a global seqlock, a bit like ``rename_lock``.  It can be used to
 341check if any change has been made to any mount points.
 342
 343While walking down the tree (away from the root) this lock is used when
 344crossing a mount point to check that the crossing was safe.  That is,
 345the value in the seqlock is read, then the code finds the mount that
 346is mounted on the current directory, if there is one, and increments
 347the ``mnt_count``.  Finally the value in ``mount_lock`` is checked against
 348the old value.  If there is no change, then the crossing was safe.  If there
 349was a change, the ``mnt_count`` is decremented and the whole process is
 350retried.
 351
 352When walking up the tree (towards the root) by following a ".." link,
 353a little more care is needed.  In this case the seqlock (which
 354contains both a counter and a spinlock) is fully locked to prevent
 355any changes to any mount points while stepping up.  This locking is
 356needed to stabilize the link to the mounted-on dentry, which the
 357refcount on the mount itself doesn't ensure.
 358
 359``mount_lock`` is also used to detect and defend against potential attacks
 360against ``LOOKUP_BENEATH`` and ``LOOKUP_IN_ROOT`` when resolving ".." (where
 361the parent directory is moved outside the root, bypassing the ``path_equal()``
 362check). If ``mount_lock`` is updated during the lookup and the path encounters
 363a "..", a potential attack occurred and ``handle_dots()`` will bail out with
 364``-EAGAIN``.
 365
 366RCU
 367~~~
 368
 369Finally the global (but extremely lightweight) RCU read lock is held
 370from time to time to ensure certain data structures don't get freed
 371unexpectedly.
 372
 373In particular it is held while scanning chains in the dcache hash
 374table, and the mount point hash table.
 375
 376Bringing it together with ``struct nameidata``
 377----------------------------------------------
 378
 379.. _First edition Unix: https://minnie.tuhs.org/cgi-bin/utree.pl?file=V1/u2.s
 380
 381Throughout the process of walking a path, the current status is stored
 382in a ``struct nameidata``, "namei" being the traditional name - dating
 383all the way back to `First Edition Unix`_ - of the function that
 384converts a "name" to an "inode".  ``struct nameidata`` contains (among
 385other fields):
 386
 387``struct path path``
 388~~~~~~~~~~~~~~~~~~~~
 389
 390A ``path`` contains a ``struct vfsmount`` (which is
 391embedded in a ``struct mount``) and a ``struct dentry``.  Together these
 392record the current status of the walk.  They start out referring to the
 393starting point (the current working directory, the root directory, or some other
 394directory identified by a file descriptor), and are updated on each
 395step.  A reference through ``d_lockref`` and ``mnt_count`` is always
 396held.
 397
 398``struct qstr last``
 399~~~~~~~~~~~~~~~~~~~~
 400
 401This is a string together with a length (i.e. *not* ``nul`` terminated)
 402that is the "next" component in the pathname.
 403
 404``int last_type``
 405~~~~~~~~~~~~~~~~~
 406
 407This is one of ``LAST_NORM``, ``LAST_ROOT``, ``LAST_DOT`` or ``LAST_DOTDOT``.
 408The ``last`` field is only valid if the type is ``LAST_NORM``.
 409
 410``struct path root``
 411~~~~~~~~~~~~~~~~~~~~
 412
 413This is used to hold a reference to the effective root of the
 414filesystem.  Often that reference won't be needed, so this field is
 415only assigned the first time it is used, or when a non-standard root
 416is requested.  Keeping a reference in the ``nameidata`` ensures that
 417only one root is in effect for the entire path walk, even if it races
 418with a ``chroot()`` system call.
 419
 420It should be noted that in the case of ``LOOKUP_IN_ROOT`` or
 421``LOOKUP_BENEATH``, the effective root becomes the directory file descriptor
 422passed to ``openat2()`` (which exposes these ``LOOKUP_`` flags).
 423
 424The root is needed when either of two conditions holds: (1) either the
 425pathname or a symbolic link starts with a "'/'", or (2) a "``..``"
 426component is being handled, since "``..``" from the root must always stay
 427at the root.  The value used is usually the current root directory of
 428the calling process.  An alternate root can be provided as when
 429``sysctl()`` calls ``file_open_root()``, and when NFSv4 or Btrfs call
 430``mount_subtree()``.  In each case a pathname is being looked up in a very
 431specific part of the filesystem, and the lookup must not be allowed to
 432escape that subtree.  It works a bit like a local ``chroot()``.
 433
 434Ignoring the handling of symbolic links, we can now describe the
 435"``link_path_walk()``" function, which handles the lookup of everything
 436except the final component as:
 437
 438   Given a path (``name``) and a nameidata structure (``nd``), check that the
 439   current directory has execute permission and then advance ``name``
 440   over one component while updating ``last_type`` and ``last``.  If that
 441   was the final component, then return, otherwise call
 442   ``walk_component()`` and repeat from the top.
 443
 444``walk_component()`` is even easier.  If the component is ``LAST_DOTS``,
 445it calls ``handle_dots()`` which does the necessary locking as already
 446described.  If it finds a ``LAST_NORM`` component it first calls
 447"``lookup_fast()``" which only looks in the dcache, but will ask the
 448filesystem to revalidate the result if it is that sort of filesystem.
 449If that doesn't get a good result, it calls "``lookup_slow()``" which
 450takes ``i_rwsem``, rechecks the cache, and then asks the filesystem
 451to find a definitive answer.
 452
 453As the last step of walk_component(), step_into() will be called either
 454directly from walk_component() or from handle_dots().  It calls
 455handle_mounts(), to check and handle mount points, in which a new
 456``struct path`` is created containing a counted reference to the new dentry and
 457a reference to the new ``vfsmount`` which is only counted if it is
 458different from the previous ``vfsmount``. Then if there is
 459a symbolic link, step_into() calls pick_link() to deal with it,
 460otherwise it installs the new ``struct path`` in the ``struct nameidata``, and
 461drops the unneeded references.
 462
 463This "hand-over-hand" sequencing of getting a reference to the new
 464dentry before dropping the reference to the previous dentry may
 465seem obvious, but is worth pointing out so that we will recognize its
 466analogue in the "RCU-walk" version.
 467
 468Handling the final component
 469----------------------------
 470
 471``link_path_walk()`` only walks as far as setting ``nd->last`` and
 472``nd->last_type`` to refer to the final component of the path.  It does
 473not call ``walk_component()`` that last time.  Handling that final
 474component remains for the caller to sort out. Those callers are
 475path_lookupat(), path_parentat() and
 476path_openat() each of which handles the differing requirements of
 477different system calls.
 478
 479``path_parentat()`` is clearly the simplest - it just wraps a little bit
 480of housekeeping around ``link_path_walk()`` and returns the parent
 481directory and final component to the caller.  The caller will be either
 482aiming to create a name (via ``filename_create()``) or remove or rename
 483a name (in which case ``user_path_parent()`` is used).  They will use
 484``i_rwsem`` to exclude other changes while they validate and then
 485perform their operation.
 486
 487``path_lookupat()`` is nearly as simple - it is used when an existing
 488object is wanted such as by ``stat()`` or ``chmod()``.  It essentially just
 489calls ``walk_component()`` on the final component through a call to
 490``lookup_last()``.  ``path_lookupat()`` returns just the final dentry.
 491It is worth noting that when flag ``LOOKUP_MOUNTPOINT`` is set,
 492path_lookupat() will unset LOOKUP_JUMPED in nameidata so that in the
 493subsequent path traversal d_weak_revalidate() won't be called.
 494This is important when unmounting a filesystem that is inaccessible, such as
 495one provided by a dead NFS server.
 496
 497Finally ``path_openat()`` is used for the ``open()`` system call; it
 498contains, in support functions starting with "open_last_lookups()", all the
 499complexity needed to handle the different subtleties of O_CREAT (with
 500or without O_EXCL), final "``/``" characters, and trailing symbolic
 501links.  We will revisit this in the final part of this series, which
 502focuses on those symbolic links.  "open_last_lookups()" will sometimes, but
 503not always, take ``i_rwsem``, depending on what it finds.
 504
 505Each of these, or the functions which call them, need to be alert to
 506the possibility that the final component is not ``LAST_NORM``.  If the
 507goal of the lookup is to create something, then any value for
 508``last_type`` other than ``LAST_NORM`` will result in an error.  For
 509example if ``path_parentat()`` reports ``LAST_DOTDOT``, then the caller
 510won't try to create that name.  They also check for trailing slashes
 511by testing ``last.name[last.len]``.  If there is any character beyond
 512the final component, it must be a trailing slash.
 513
 514Revalidation and automounts
 515---------------------------
 516
 517Apart from symbolic links, there are only two parts of the "REF-walk"
 518process not yet covered.  One is the handling of stale cache entries
 519and the other is automounts.
 520
 521On filesystems that require it, the lookup routines will call the
 522``->d_revalidate()`` dentry method to ensure that the cached information
 523is current.  This will often confirm validity or update a few details
 524from a server.  In some cases it may find that there has been change
 525further up the path and that something that was thought to be valid
 526previously isn't really.  When this happens the lookup of the whole
 527path is aborted and retried with the "``LOOKUP_REVAL``" flag set.  This
 528forces revalidation to be more thorough.  We will see more details of
 529this retry process in the next article.
 530
 531Automount points are locations in the filesystem where an attempt to
 532lookup a name can trigger changes to how that lookup should be
 533handled, in particular by mounting a filesystem there.  These are
 534covered in greater detail in autofs.rst in the Linux documentation
 535tree, but a few notes specifically related to path lookup are in order
 536here.
 537
 538The Linux VFS has a concept of "managed" dentries.  There are three
 539potentially interesting things about these dentries corresponding
 540to three different flags that might be set in ``dentry->d_flags``:
 541
 542``DCACHE_MANAGE_TRANSIT``
 543~~~~~~~~~~~~~~~~~~~~~~~~~
 544
 545If this flag has been set, then the filesystem has requested that the
 546``d_manage()`` dentry operation be called before handling any possible
 547mount point.  This can perform two particular services:
 548
 549It can block to avoid races.  If an automount point is being
 550unmounted, the ``d_manage()`` function will usually wait for that
 551process to complete before letting the new lookup proceed and possibly
 552trigger a new automount.
 553
 554It can selectively allow only some processes to transit through a
 555mount point.  When a server process is managing automounts, it may
 556need to access a directory without triggering normal automount
 557processing.  That server process can identify itself to the ``autofs``
 558filesystem, which will then give it a special pass through
 559``d_manage()`` by returning ``-EISDIR``.
 560
 561``DCACHE_MOUNTED``
 562~~~~~~~~~~~~~~~~~~
 563
 564This flag is set on every dentry that is mounted on.  As Linux
 565supports multiple filesystem namespaces, it is possible that the
 566dentry may not be mounted on in *this* namespace, just in some
 567other.  So this flag is seen as a hint, not a promise.
 568
 569If this flag is set, and ``d_manage()`` didn't return ``-EISDIR``,
 570``lookup_mnt()`` is called to examine the mount hash table (honoring the
 571``mount_lock`` described earlier) and possibly return a new ``vfsmount``
 572and a new ``dentry`` (both with counted references).
 573
 574``DCACHE_NEED_AUTOMOUNT``
 575~~~~~~~~~~~~~~~~~~~~~~~~~
 576
 577If ``d_manage()`` allowed us to get this far, and ``lookup_mnt()`` didn't
 578find a mount point, then this flag causes the ``d_automount()`` dentry
 579operation to be called.
 580
 581The ``d_automount()`` operation can be arbitrarily complex and may
 582communicate with server processes etc. but it should ultimately either
 583report that there was an error, that there was nothing to mount, or
 584should provide an updated ``struct path`` with new ``dentry`` and ``vfsmount``.
 585
 586In the latter case, ``finish_automount()`` will be called to safely
 587install the new mount point into the mount table.
 588
 589There is no new locking of import here and it is important that no
 590locks (only counted references) are held over this processing due to
 591the very real possibility of extended delays.
 592This will become more important next time when we examine RCU-walk
 593which is particularly sensitive to delays.
 594
 595RCU-walk - faster pathname lookup in Linux
 596==========================================
 597
 598RCU-walk is another algorithm for performing pathname lookup in Linux.
 599It is in many ways similar to REF-walk and the two share quite a bit
 600of code.  The significant difference in RCU-walk is how it allows for
 601the possibility of concurrent access.
 602
 603We noted that REF-walk is complex because there are numerous details
 604and special cases.  RCU-walk reduces this complexity by simply
 605refusing to handle a number of cases -- it instead falls back to
 606REF-walk.  The difficulty with RCU-walk comes from a different
 607direction: unfamiliarity.  The locking rules when depending on RCU are
 608quite different from traditional locking, so we will spend a little extra
 609time when we come to those.
 610
 611Clear demarcation of roles
 612--------------------------
 613
 614The easiest way to manage concurrency is to forcibly stop any other
 615thread from changing the data structures that a given thread is
 616looking at.  In cases where no other thread would even think of
 617changing the data and lots of different threads want to read at the
 618same time, this can be very costly.  Even when using locks that permit
 619multiple concurrent readers, the simple act of updating the count of
 620the number of current readers can impose an unwanted cost.  So the
 621goal when reading a shared data structure that no other process is
 622changing is to avoid writing anything to memory at all.  Take no
 623locks, increment no counts, leave no footprints.
 624
 625The REF-walk mechanism already described certainly doesn't follow this
 626principle, but then it is really designed to work when there may well
 627be other threads modifying the data.  RCU-walk, in contrast, is
 628designed for the common situation where there are lots of frequent
 629readers and only occasional writers.  This may not be common in all
 630parts of the filesystem tree, but in many parts it will be.  For the
 631other parts it is important that RCU-walk can quickly fall back to
 632using REF-walk.
 633
 634Pathname lookup always starts in RCU-walk mode but only remains there
 635as long as what it is looking for is in the cache and is stable.  It
 636dances lightly down the cached filesystem image, leaving no footprints
 637and carefully watching where it is, to be sure it doesn't trip.  If it
 638notices that something has changed or is changing, or if something
 639isn't in the cache, then it tries to stop gracefully and switch to
 640REF-walk.
 641
 642This stopping requires getting a counted reference on the current
 643``vfsmount`` and ``dentry``, and ensuring that these are still valid -
 644that a path walk with REF-walk would have found the same entries.
 645This is an invariant that RCU-walk must guarantee.  It can only make
 646decisions, such as selecting the next step, that are decisions which
 647REF-walk could also have made if it were walking down the tree at the
 648same time.  If the graceful stop succeeds, the rest of the path is
 649processed with the reliable, if slightly sluggish, REF-walk.  If
 650RCU-walk finds it cannot stop gracefully, it simply gives up and
 651restarts from the top with REF-walk.
 652
 653This pattern of "try RCU-walk, if that fails try REF-walk" can be
 654clearly seen in functions like filename_lookup(),
 655filename_parentat(),
 656do_filp_open(), and do_file_open_root().  These four
 657correspond roughly to the three ``path_*()`` functions we met earlier,
 658each of which calls ``link_path_walk()``.  The ``path_*()`` functions are
 659called using different mode flags until a mode is found which works.
 660They are first called with ``LOOKUP_RCU`` set to request "RCU-walk".  If
 661that fails with the error ``ECHILD`` they are called again with no
 662special flag to request "REF-walk".  If either of those report the
 663error ``ESTALE`` a final attempt is made with ``LOOKUP_REVAL`` set (and no
 664``LOOKUP_RCU``) to ensure that entries found in the cache are forcibly
 665revalidated - normally entries are only revalidated if the filesystem
 666determines that they are too old to trust.
 667
 668The ``LOOKUP_RCU`` attempt may drop that flag internally and switch to
 669REF-walk, but will never then try to switch back to RCU-walk.  Places
 670that trip up RCU-walk are much more likely to be near the leaves and
 671so it is very unlikely that there will be much, if any, benefit from
 672switching back.
 673
 674RCU and seqlocks: fast and light
 675--------------------------------
 676
 677RCU is, unsurprisingly, critical to RCU-walk mode.  The
 678``rcu_read_lock()`` is held for the entire time that RCU-walk is walking
 679down a path.  The particular guarantee it provides is that the key
 680data structures - dentries, inodes, super_blocks, and mounts - will
 681not be freed while the lock is held.  They might be unlinked or
 682invalidated in one way or another, but the memory will not be
 683repurposed so values in various fields will still be meaningful.  This
 684is the only guarantee that RCU provides; everything else is done using
 685seqlocks.
 686
 687As we saw above, REF-walk holds a counted reference to the current
 688dentry and the current vfsmount, and does not release those references
 689before taking references to the "next" dentry or vfsmount.  It also
 690sometimes takes the ``d_lock`` spinlock.  These references and locks are
 691taken to prevent certain changes from happening.  RCU-walk must not
 692take those references or locks and so cannot prevent such changes.
 693Instead, it checks to see if a change has been made, and aborts or
 694retries if it has.
 695
 696To preserve the invariant mentioned above (that RCU-walk may only make
 697decisions that REF-walk could have made), it must make the checks at
 698or near the same places that REF-walk holds the references.  So, when
 699REF-walk increments a reference count or takes a spinlock, RCU-walk
 700samples the status of a seqlock using ``read_seqcount_begin()`` or a
 701similar function.  When REF-walk decrements the count or drops the
 702lock, RCU-walk checks if the sampled status is still valid using
 703``read_seqcount_retry()`` or similar.
 704
 705However, there is a little bit more to seqlocks than that.  If
 706RCU-walk accesses two different fields in a seqlock-protected
 707structure, or accesses the same field twice, there is no a priori
 708guarantee of any consistency between those accesses.  When consistency
 709is needed - which it usually is - RCU-walk must take a copy and then
 710use ``read_seqcount_retry()`` to validate that copy.
 711
 712``read_seqcount_retry()`` not only checks the sequence number, but also
 713imposes a memory barrier so that no memory-read instruction from
 714*before* the call can be delayed until *after* the call, either by the
 715CPU or by the compiler.  A simple example of this can be seen in
 716``slow_dentry_cmp()`` which, for filesystems which do not use simple
 717byte-wise name equality, calls into the filesystem to compare a name
 718against a dentry.  The length and name pointer are copied into local
 719variables, then ``read_seqcount_retry()`` is called to confirm the two
 720are consistent, and only then is ``->d_compare()`` called.  When
 721standard filename comparison is used, ``dentry_cmp()`` is called
 722instead.  Notably it does *not* use ``read_seqcount_retry()``, but
 723instead has a large comment explaining why the consistency guarantee
 724isn't necessary.  A subsequent ``read_seqcount_retry()`` will be
 725sufficient to catch any problem that could occur at this point.
 726
 727With that little refresher on seqlocks out of the way we can look at
 728the bigger picture of how RCU-walk uses seqlocks.
 729
 730``mount_lock`` and ``nd->m_seq``
 731~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 732
 733We already met the ``mount_lock`` seqlock when REF-walk used it to
 734ensure that crossing a mount point is performed safely.  RCU-walk uses
 735it for that too, but for quite a bit more.
 736
 737Instead of taking a counted reference to each ``vfsmount`` as it
 738descends the tree, RCU-walk samples the state of ``mount_lock`` at the
 739start of the walk and stores this initial sequence number in the
 740``struct nameidata`` in the ``m_seq`` field.  This one lock and one
 741sequence number are used to validate all accesses to all ``vfsmounts``,
 742and all mount point crossings.  As changes to the mount table are
 743relatively rare, it is reasonable to fall back on REF-walk any time
 744that any "mount" or "unmount" happens.
 745
 746``m_seq`` is checked (using ``read_seqretry()``) at the end of an RCU-walk
 747sequence, whether switching to REF-walk for the rest of the path or
 748when the end of the path is reached.  It is also checked when stepping
 749down over a mount point (in ``__follow_mount_rcu()``) or up (in
 750``follow_dotdot_rcu()``).  If it is ever found to have changed, the
 751whole RCU-walk sequence is aborted and the path is processed again by
 752REF-walk.
 753
 754If RCU-walk finds that ``mount_lock`` hasn't changed then it can be sure
 755that, had REF-walk taken counted references on each vfsmount, the
 756results would have been the same.  This ensures the invariant holds,
 757at least for vfsmount structures.
 758
 759``dentry->d_seq`` and ``nd->seq``
 760~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 761
 762In place of taking a count or lock on ``d_reflock``, RCU-walk samples
 763the per-dentry ``d_seq`` seqlock, and stores the sequence number in the
 764``seq`` field of the nameidata structure, so ``nd->seq`` should always be
 765the current sequence number of ``nd->dentry``.  This number needs to be
 766revalidated after copying, and before using, the name, parent, or
 767inode of the dentry.
 768
 769The handling of the name we have already looked at, and the parent is
 770only accessed in ``follow_dotdot_rcu()`` which fairly trivially follows
 771the required pattern, though it does so for three different cases.
 772
 773When not at a mount point, ``d_parent`` is followed and its ``d_seq`` is
 774collected.  When we are at a mount point, we instead follow the
 775``mnt->mnt_mountpoint`` link to get a new dentry and collect its
 776``d_seq``.  Then, after finally finding a ``d_parent`` to follow, we must
 777check if we have landed on a mount point and, if so, must find that
 778mount point and follow the ``mnt->mnt_root`` link.  This would imply a
 779somewhat unusual, but certainly possible, circumstance where the
 780starting point of the path lookup was in part of the filesystem that
 781was mounted on, and so not visible from the root.
 782
 783The inode pointer, stored in ``->d_inode``, is a little more
 784interesting.  The inode will always need to be accessed at least
 785twice, once to determine if it is NULL and once to verify access
 786permissions.  Symlink handling requires a validated inode pointer too.
 787Rather than revalidating on each access, a copy is made on the first
 788access and it is stored in the ``inode`` field of ``nameidata`` from where
 789it can be safely accessed without further validation.
 790
 791``lookup_fast()`` is the only lookup routine that is used in RCU-mode,
 792``lookup_slow()`` being too slow and requiring locks.  It is in
 793``lookup_fast()`` that we find the important "hand over hand" tracking
 794of the current dentry.
 795
 796The current ``dentry`` and current ``seq`` number are passed to
 797``__d_lookup_rcu()`` which, on success, returns a new ``dentry`` and a
 798new ``seq`` number.  ``lookup_fast()`` then copies the inode pointer and
 799revalidates the new ``seq`` number.  It then validates the old ``dentry``
 800with the old ``seq`` number one last time and only then continues.  This
 801process of getting the ``seq`` number of the new dentry and then
 802checking the ``seq`` number of the old exactly mirrors the process of
 803getting a counted reference to the new dentry before dropping that for
 804the old dentry which we saw in REF-walk.
 805
 806No ``inode->i_rwsem`` or even ``rename_lock``
 807~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 808
 809A semaphore is a fairly heavyweight lock that can only be taken when it is
 810permissible to sleep.  As ``rcu_read_lock()`` forbids sleeping,
 811``inode->i_rwsem`` plays no role in RCU-walk.  If some other thread does
 812take ``i_rwsem`` and modifies the directory in a way that RCU-walk needs
 813to notice, the result will be either that RCU-walk fails to find the
 814dentry that it is looking for, or it will find a dentry which
 815``read_seqretry()`` won't validate.  In either case it will drop down to
 816REF-walk mode which can take whatever locks are needed.
 817
 818Though ``rename_lock`` could be used by RCU-walk as it doesn't require
 819any sleeping, RCU-walk doesn't bother.  REF-walk uses ``rename_lock`` to
 820protect against the possibility of hash chains in the dcache changing
 821while they are being searched.  This can result in failing to find
 822something that actually is there.  When RCU-walk fails to find
 823something in the dentry cache, whether it is really there or not, it
 824already drops down to REF-walk and tries again with appropriate
 825locking.  This neatly handles all cases, so adding extra checks on
 826rename_lock would bring no significant value.
 827
 828``unlazy walk()`` and ``complete_walk()``
 829-----------------------------------------
 830
 831That "dropping down to REF-walk" typically involves a call to
 832``unlazy_walk()``, so named because "RCU-walk" is also sometimes
 833referred to as "lazy walk".  ``unlazy_walk()`` is called when
 834following the path down to the current vfsmount/dentry pair seems to
 835have proceeded successfully, but the next step is problematic.  This
 836can happen if the next name cannot be found in the dcache, if
 837permission checking or name revalidation couldn't be achieved while
 838the ``rcu_read_lock()`` is held (which forbids sleeping), if an
 839automount point is found, or in a couple of cases involving symlinks.
 840It is also called from ``complete_walk()`` when the lookup has reached
 841the final component, or the very end of the path, depending on which
 842particular flavor of lookup is used.
 843
 844Other reasons for dropping out of RCU-walk that do not trigger a call
 845to ``unlazy_walk()`` are when some inconsistency is found that cannot be
 846handled immediately, such as ``mount_lock`` or one of the ``d_seq``
 847seqlocks reporting a change.  In these cases the relevant function
 848will return ``-ECHILD`` which will percolate up until it triggers a new
 849attempt from the top using REF-walk.
 850
 851For those cases where ``unlazy_walk()`` is an option, it essentially
 852takes a reference on each of the pointers that it holds (vfsmount,
 853dentry, and possibly some symbolic links) and then verifies that the
 854relevant seqlocks have not been changed.  If there have been changes,
 855it, too, aborts with ``-ECHILD``, otherwise the transition to REF-walk
 856has been a success and the lookup process continues.
 857
 858Taking a reference on those pointers is not quite as simple as just
 859incrementing a counter.  That works to take a second reference if you
 860already have one (often indirectly through another object), but it
 861isn't sufficient if you don't actually have a counted reference at
 862all.  For ``dentry->d_lockref``, it is safe to increment the reference
 863counter to get a reference unless it has been explicitly marked as
 864"dead" which involves setting the counter to ``-128``.
 865``lockref_get_not_dead()`` achieves this.
 866
 867For ``mnt->mnt_count`` it is safe to take a reference as long as
 868``mount_lock`` is then used to validate the reference.  If that
 869validation fails, it may *not* be safe to just drop that reference in
 870the standard way of calling ``mnt_put()`` - an unmount may have
 871progressed too far.  So the code in ``legitimize_mnt()``, when it
 872finds that the reference it got might not be safe, checks the
 873``MNT_SYNC_UMOUNT`` flag to determine if a simple ``mnt_put()`` is
 874correct, or if it should just decrement the count and pretend none of
 875this ever happened.
 876
 877Taking care in filesystems
 878--------------------------
 879
 880RCU-walk depends almost entirely on cached information and often will
 881not call into the filesystem at all.  However there are two places,
 882besides the already-mentioned component-name comparison, where the
 883file system might be included in RCU-walk, and it must know to be
 884careful.
 885
 886If the filesystem has non-standard permission-checking requirements -
 887such as a networked filesystem which may need to check with the server
 888- the ``i_op->permission`` interface might be called during RCU-walk.
 889In this case an extra "``MAY_NOT_BLOCK``" flag is passed so that it
 890knows not to sleep, but to return ``-ECHILD`` if it cannot complete
 891promptly.  ``i_op->permission`` is given the inode pointer, not the
 892dentry, so it doesn't need to worry about further consistency checks.
 893However if it accesses any other filesystem data structures, it must
 894ensure they are safe to be accessed with only the ``rcu_read_lock()``
 895held.  This typically means they must be freed using ``kfree_rcu()`` or
 896similar.
 897
 898.. _READ_ONCE: https://lwn.net/Articles/624126/
 899
 900If the filesystem may need to revalidate dcache entries, then
 901``d_op->d_revalidate`` may be called in RCU-walk too.  This interface
 902*is* passed the dentry but does not have access to the ``inode`` or the
 903``seq`` number from the ``nameidata``, so it needs to be extra careful
 904when accessing fields in the dentry.  This "extra care" typically
 905involves using  `READ_ONCE() <READ_ONCE_>`_ to access fields, and verifying the
 906result is not NULL before using it.  This pattern can be seen in
 907``nfs_lookup_revalidate()``.
 908
 909A pair of patterns
 910------------------
 911
 912In various places in the details of REF-walk and RCU-walk, and also in
 913the big picture, there are a couple of related patterns that are worth
 914being aware of.
 915
 916The first is "try quickly and check, if that fails try slowly".  We
 917can see that in the high-level approach of first trying RCU-walk and
 918then trying REF-walk, and in places where ``unlazy_walk()`` is used to
 919switch to REF-walk for the rest of the path.  We also saw it earlier
 920in ``dget_parent()`` when following a "``..``" link.  It tries a quick way
 921to get a reference, then falls back to taking locks if needed.
 922
 923The second pattern is "try quickly and check, if that fails try
 924again - repeatedly".  This is seen with the use of ``rename_lock`` and
 925``mount_lock`` in REF-walk.  RCU-walk doesn't make use of this pattern -
 926if anything goes wrong it is much safer to just abort and try a more
 927sedate approach.
 928
 929The emphasis here is "try quickly and check".  It should probably be
 930"try quickly *and carefully*, then check".  The fact that checking is
 931needed is a reminder that the system is dynamic and only a limited
 932number of things are safe at all.  The most likely cause of errors in
 933this whole process is assuming something is safe when in reality it
 934isn't.  Careful consideration of what exactly guarantees the safety of
 935each access is sometimes necessary.
 936
 937A walk among the symlinks
 938=========================
 939
 940There are several basic issues that we will examine to understand the
 941handling of symbolic links:  the symlink stack, together with cache
 942lifetimes, will help us understand the overall recursive handling of
 943symlinks and lead to the special care needed for the final component.
 944Then a consideration of access-time updates and summary of the various
 945flags controlling lookup will finish the story.
 946
 947The symlink stack
 948-----------------
 949
 950There are only two sorts of filesystem objects that can usefully
 951appear in a path prior to the final component: directories and symlinks.
 952Handling directories is quite straightforward: the new directory
 953simply becomes the starting point at which to interpret the next
 954component on the path.  Handling symbolic links requires a bit more
 955work.
 956
 957Conceptually, symbolic links could be handled by editing the path.  If
 958a component name refers to a symbolic link, then that component is
 959replaced by the body of the link and, if that body starts with a '/',
 960then all preceding parts of the path are discarded.  This is what the
 961"``readlink -f``" command does, though it also edits out "``.``" and
 962"``..``" components.
 963
 964Directly editing the path string is not really necessary when looking
 965up a path, and discarding early components is pointless as they aren't
 966looked at anyway.  Keeping track of all remaining components is
 967important, but they can of course be kept separately; there is no need
 968to concatenate them.  As one symlink may easily refer to another,
 969which in turn can refer to a third, we may need to keep the remaining
 970components of several paths, each to be processed when the preceding
 971ones are completed.  These path remnants are kept on a stack of
 972limited size.
 973
 974There are two reasons for placing limits on how many symlinks can
 975occur in a single path lookup.  The most obvious is to avoid loops.
 976If a symlink referred to itself either directly or through
 977intermediaries, then following the symlink can never complete
 978successfully - the error ``ELOOP`` must be returned.  Loops can be
 979detected without imposing limits, but limits are the simplest solution
 980and, given the second reason for restriction, quite sufficient.
 981
 982.. _outlined recently: http://thread.gmane.org/gmane.linux.kernel/1934390/focus=1934550
 983
 984The second reason was `outlined recently`_ by Linus:
 985
 986   Because it's a latency and DoS issue too. We need to react well to
 987   true loops, but also to "very deep" non-loops. It's not about memory
 988   use, it's about users triggering unreasonable CPU resources.
 989
 990Linux imposes a limit on the length of any pathname: ``PATH_MAX``, which
 991is 4096.  There are a number of reasons for this limit; not letting the
 992kernel spend too much time on just one path is one of them.  With
 993symbolic links you can effectively generate much longer paths so some
 994sort of limit is needed for the same reason.  Linux imposes a limit of
 995at most 40 (MAXSYMLINKS) symlinks in any one path lookup.  It previously imposed
 996a further limit of eight on the maximum depth of recursion, but that was
 997raised to 40 when a separate stack was implemented, so there is now
 998just the one limit.
 999
1000The ``nameidata`` structure that we met in an earlier article contains a
1001small stack that can be used to store the remaining part of up to two
1002symlinks.  In many cases this will be sufficient.  If it isn't, a
1003separate stack is allocated with room for 40 symlinks.  Pathname
1004lookup will never exceed that stack as, once the 40th symlink is
1005detected, an error is returned.
1006
1007It might seem that the name remnants are all that needs to be stored on
1008this stack, but we need a bit more.  To see that, we need to move on to
1009cache lifetimes.
1010
1011Storage and lifetime of cached symlinks
1012---------------------------------------
1013
1014Like other filesystem resources, such as inodes and directory
1015entries, symlinks are cached by Linux to avoid repeated costly access
1016to external storage.  It is particularly important for RCU-walk to be
1017able to find and temporarily hold onto these cached entries, so that
1018it doesn't need to drop down into REF-walk.
1019
1020.. _object-oriented design pattern: https://lwn.net/Articles/446317/
1021
1022While each filesystem is free to make its own choice, symlinks are
1023typically stored in one of two places.  Short symlinks are often
1024stored directly in the inode.  When a filesystem allocates a ``struct
1025inode`` it typically allocates extra space to store private data (a
1026common `object-oriented design pattern`_ in the kernel).  This will
1027sometimes include space for a symlink.  The other common location is
1028in the page cache, which normally stores the content of files.  The
1029pathname in a symlink can be seen as the content of that symlink and
1030can easily be stored in the page cache just like file content.
1031
1032When neither of these is suitable, the next most likely scenario is
1033that the filesystem will allocate some temporary memory and copy or
1034construct the symlink content into that memory whenever it is needed.
1035
1036When the symlink is stored in the inode, it has the same lifetime as
1037the inode which, itself, is protected by RCU or by a counted reference
1038on the dentry.  This means that the mechanisms that pathname lookup
1039uses to access the dcache and icache (inode cache) safely are quite
1040sufficient for accessing some cached symlinks safely.  In these cases,
1041the ``i_link`` pointer in the inode is set to point to wherever the
1042symlink is stored and it can be accessed directly whenever needed.
1043
1044When the symlink is stored in the page cache or elsewhere, the
1045situation is not so straightforward.  A reference on a dentry or even
1046on an inode does not imply any reference on cached pages of that
1047inode, and even an ``rcu_read_lock()`` is not sufficient to ensure that
1048a page will not disappear.  So for these symlinks the pathname lookup
1049code needs to ask the filesystem to provide a stable reference and,
1050significantly, needs to release that reference when it is finished
1051with it.
1052
1053Taking a reference to a cache page is often possible even in RCU-walk
1054mode.  It does require making changes to memory, which is best avoided,
1055but that isn't necessarily a big cost and it is better than dropping
1056out of RCU-walk mode completely.  Even filesystems that allocate
1057space to copy the symlink into can use ``GFP_ATOMIC`` to often successfully
1058allocate memory without the need to drop out of RCU-walk.  If a
1059filesystem cannot successfully get a reference in RCU-walk mode, it
1060must return ``-ECHILD`` and ``unlazy_walk()`` will be called to return to
1061REF-walk mode in which the filesystem is allowed to sleep.
1062
1063The place for all this to happen is the ``i_op->get_link()`` inode
1064method. This is called both in RCU-walk and REF-walk. In RCU-walk the
1065``dentry*`` argument is NULL, ``->get_link()`` can return -ECHILD to drop out of
1066RCU-walk.  Much like the ``i_op->permission()`` method we
1067looked at previously, ``->get_link()`` would need to be careful that
1068all the data structures it references are safe to be accessed while
1069holding no counted reference, only the RCU lock. A callback
1070``struct delayed_called`` will be passed to ``->get_link()``:
1071file systems can set their own put_link function and argument through
1072set_delayed_call(). Later on, when VFS wants to put link, it will call
1073do_delayed_call() to invoke that callback function with the argument.
1074
1075In order for the reference to each symlink to be dropped when the walk completes,
1076whether in RCU-walk or REF-walk, the symlink stack needs to contain,
1077along with the path remnants:
1078
1079- the ``struct path`` to provide a reference to the previous path
1080- the ``const char *`` to provide a reference to the to previous name
1081- the ``seq`` to allow the path to be safely switched from RCU-walk to REF-walk
1082- the ``struct delayed_call`` for later invocation.
1083
1084This means that each entry in the symlink stack needs to hold five
1085pointers and an integer instead of just one pointer (the path
1086remnant).  On a 64-bit system, this is about 40 bytes per entry;
1087with 40 entries it adds up to 1600 bytes total, which is less than
1088half a page.  So it might seem like a lot, but is by no means
1089excessive.
1090
1091Note that, in a given stack frame, the path remnant (``name``) is not
1092part of the symlink that the other fields refer to.  It is the remnant
1093to be followed once that symlink has been fully parsed.
1094
1095Following the symlink
1096---------------------
1097
1098The main loop in ``link_path_walk()`` iterates seamlessly over all
1099components in the path and all of the non-final symlinks.  As symlinks
1100are processed, the ``name`` pointer is adjusted to point to a new
1101symlink, or is restored from the stack, so that much of the loop
1102doesn't need to notice.  Getting this ``name`` variable on and off the
1103stack is very straightforward; pushing and popping the references is
1104a little more complex.
1105
1106When a symlink is found, walk_component() calls pick_link() via step_into()
1107which returns the link from the filesystem.
1108Providing that operation is successful, the old path ``name`` is placed on the
1109stack, and the new value is used as the ``name`` for a while.  When the end of
1110the path is found (i.e. ``*name`` is ``'\0'``) the old ``name`` is restored
1111off the stack and path walking continues.
1112
1113Pushing and popping the reference pointers (inode, cookie, etc.) is more
1114complex in part because of the desire to handle tail recursion.  When
1115the last component of a symlink itself points to a symlink, we
1116want to pop the symlink-just-completed off the stack before pushing
1117the symlink-just-found to avoid leaving empty path remnants that would
1118just get in the way.
1119
1120It is most convenient to push the new symlink references onto the
1121stack in ``walk_component()`` immediately when the symlink is found;
1122``walk_component()`` is also the last piece of code that needs to look at the
1123old symlink as it walks that last component.  So it is quite
1124convenient for ``walk_component()`` to release the old symlink and pop
1125the references just before pushing the reference information for the
1126new symlink.  It is guided in this by three flags: ``WALK_NOFOLLOW`` which
1127forbids it from following a symlink if it finds one, ``WALK_MORE``
1128which indicates that it is yet too early to release the
1129current symlink, and ``WALK_TRAILING`` which indicates that it is on the final
1130component of the lookup, so we will check userspace flag ``LOOKUP_FOLLOW`` to
1131decide whether follow it when it is a symlink and call ``may_follow_link()`` to
1132check if we have privilege to follow it.
1133
1134Symlinks with no final component
1135~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1136
1137A pair of special-case symlinks deserve a little further explanation.
1138Both result in a new ``struct path`` (with mount and dentry) being set
1139up in the ``nameidata``, and result in pick_link() returning ``NULL``.
1140
1141The more obvious case is a symlink to "``/``".  All symlinks starting
1142with "``/``" are detected in pick_link() which resets the ``nameidata``
1143to point to the effective filesystem root.  If the symlink only
1144contains "``/``" then there is nothing more to do, no components at all,
1145so ``NULL`` is returned to indicate that the symlink can be released and
1146the stack frame discarded.
1147
1148The other case involves things in ``/proc`` that look like symlinks but
1149aren't really (and are therefore commonly referred to as "magic-links")::
1150
1151     $ ls -l /proc/self/fd/1
1152     lrwx------ 1 neilb neilb 64 Jun 13 10:19 /proc/self/fd/1 -> /dev/pts/4
1153
1154Every open file descriptor in any process is represented in ``/proc`` by
1155something that looks like a symlink.  It is really a reference to the
1156target file, not just the name of it.  When you ``readlink`` these
1157objects you get a name that might refer to the same file - unless it
1158has been unlinked or mounted over.  When ``walk_component()`` follows
1159one of these, the ``->get_link()`` method in "procfs" doesn't return
1160a string name, but instead calls nd_jump_link() which updates the
1161``nameidata`` in place to point to that target.  ``->get_link()`` then
1162returns ``NULL``.  Again there is no final component and pick_link()
1163returns ``NULL``.
1164
1165Following the symlink in the final component
1166--------------------------------------------
1167
1168All this leads to ``link_path_walk()`` walking down every component, and
1169following all symbolic links it finds, until it reaches the final
1170component.  This is just returned in the ``last`` field of ``nameidata``.
1171For some callers, this is all they need; they want to create that
1172``last`` name if it doesn't exist or give an error if it does.  Other
1173callers will want to follow a symlink if one is found, and possibly
1174apply special handling to the last component of that symlink, rather
1175than just the last component of the original file name.  These callers
1176potentially need to call ``link_path_walk()`` again and again on
1177successive symlinks until one is found that doesn't point to another
1178symlink.
1179
1180This case is handled by relevant callers of link_path_walk(), such as
1181path_lookupat(), path_openat() using a loop that calls link_path_walk(),
1182and then handles the final component by calling open_last_lookups() or
1183lookup_last(). If it is a symlink that needs to be followed,
1184open_last_lookups() or lookup_last() will set things up properly and
1185return the path so that the loop repeats, calling
1186link_path_walk() again.  This could loop as many as 40 times if the last
1187component of each symlink is another symlink.
1188
1189Of the various functions that examine the final component, 
1190open_last_lookups() is the most interesting as it works in tandem
1191with do_open() for opening a file.  Part of open_last_lookups() runs
1192with ``i_rwsem`` held and this part is in a separate function: lookup_open().
1193
1194Explaining open_last_lookups() and do_open() completely is beyond the scope
1195of this article, but a few highlights should help those interested in exploring
1196the code.
1197
11981. Rather than just finding the target file, do_open() is used after
1199   open_last_lookup() to open
1200   it.  If the file was found in the dcache, then ``vfs_open()`` is used for
1201   this.  If not, then ``lookup_open()`` will either call ``atomic_open()`` (if
1202   the filesystem provides it) to combine the final lookup with the open, or
1203   will perform the separate ``i_op->lookup()`` and ``i_op->create()`` steps
1204   directly.  In the later case the actual "open" of this newly found or
1205   created file will be performed by vfs_open(), just as if the name
1206   were found in the dcache.
1207
12082. vfs_open() can fail with ``-EOPENSTALE`` if the cached information
1209   wasn't quite current enough.  If it's in RCU-walk ``-ECHILD`` will be returned
1210   otherwise ``-ESTALE`` is returned.  When ``-ESTALE`` is returned, the caller may
1211   retry with ``LOOKUP_REVAL`` flag set.
1212
12133. An open with O_CREAT **does** follow a symlink in the final component,
1214   unlike other creation system calls (like ``mkdir``).  So the sequence::
1215
1216          ln -s bar /tmp/foo
1217          echo hello > /tmp/foo
1218
1219   will create a file called ``/tmp/bar``.  This is not permitted if
1220   ``O_EXCL`` is set but otherwise is handled for an O_CREAT open much
1221   like for a non-creating open: lookup_last() or open_last_lookup()
1222   returns a non ``NULL`` value, and link_path_walk() gets called and the
1223   open process continues on the symlink that was found.
1224
1225Updating the access time
1226------------------------
1227
1228We previously said of RCU-walk that it would "take no locks, increment
1229no counts, leave no footprints."  We have since seen that some
1230"footprints" can be needed when handling symlinks as a counted
1231reference (or even a memory allocation) may be needed.  But these
1232footprints are best kept to a minimum.
1233
1234One other place where walking down a symlink can involve leaving
1235footprints in a way that doesn't affect directories is in updating access times.
1236In Unix (and Linux) every filesystem object has a "last accessed
1237time", or "``atime``".  Passing through a directory to access a file
1238within is not considered to be an access for the purposes of
1239``atime``; only listing the contents of a directory can update its ``atime``.
1240Symlinks are different it seems.  Both reading a symlink (with ``readlink()``)
1241and looking up a symlink on the way to some other destination can
1242update the atime on that symlink.
1243
1244.. _clearest statement: https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_08
1245
1246It is not clear why this is the case; POSIX has little to say on the
1247subject.  The `clearest statement`_ is that, if a particular implementation
1248updates a timestamp in a place not specified by POSIX, this must be
1249documented "except that any changes caused by pathname resolution need
1250not be documented".  This seems to imply that POSIX doesn't really
1251care about access-time updates during pathname lookup.
1252
1253.. _Linux 1.3.87: https://git.kernel.org/cgit/linux/kernel/git/history/history.git/diff/fs/ext2/symlink.c?id=f806c6db77b8eaa6e00dcfb6b567706feae8dbb8
1254
1255An examination of history shows that prior to `Linux 1.3.87`_, the ext2
1256filesystem, at least, didn't update atime when following a link.
1257Unfortunately we have no record of why that behavior was changed.
1258
1259In any case, access time must now be updated and that operation can be
1260quite complex.  Trying to stay in RCU-walk while doing it is best
1261avoided.  Fortunately it is often permitted to skip the ``atime``
1262update.  Because ``atime`` updates cause performance problems in various
1263areas, Linux supports the ``relatime`` mount option, which generally
1264limits the updates of ``atime`` to once per day on files that aren't
1265being changed (and symlinks never change once created).  Even without
1266``relatime``, many filesystems record ``atime`` with a one-second
1267granularity, so only one update per second is required.
1268
1269It is easy to test if an ``atime`` update is needed while in RCU-walk
1270mode and, if it isn't, the update can be skipped and RCU-walk mode
1271continues.  Only when an ``atime`` update is actually required does the
1272path walk drop down to REF-walk.  All of this is handled in the
1273``get_link()`` function.
1274
1275A few flags
1276-----------
1277
1278A suitable way to wrap up this tour of pathname walking is to list
1279the various flags that can be stored in the ``nameidata`` to guide the
1280lookup process.  Many of these are only meaningful on the final
1281component, others reflect the current state of the pathname lookup, and some
1282apply restrictions to all path components encountered in the path lookup.
1283
1284And then there is ``LOOKUP_EMPTY``, which doesn't fit conceptually with
1285the others.  If this is not set, an empty pathname causes an error
1286very early on.  If it is set, empty pathnames are not considered to be
1287an error.
1288
1289Global state flags
1290~~~~~~~~~~~~~~~~~~
1291
1292We have already met two global state flags: ``LOOKUP_RCU`` and
1293``LOOKUP_REVAL``.  These select between one of three overall approaches
1294to lookup: RCU-walk, REF-walk, and REF-walk with forced revalidation.
1295
1296``LOOKUP_PARENT`` indicates that the final component hasn't been reached
1297yet.  This is primarily used to tell the audit subsystem the full
1298context of a particular access being audited.
1299
1300``ND_ROOT_PRESET`` indicates that the ``root`` field in the ``nameidata`` was
1301provided by the caller, so it shouldn't be released when it is no
1302longer needed.
1303
1304``ND_JUMPED`` means that the current dentry was chosen not because
1305it had the right name but for some other reason.  This happens when
1306following "``..``", following a symlink to ``/``, crossing a mount point
1307or accessing a "``/proc/$PID/fd/$FD``" symlink (also known as a "magic
1308link"). In this case the filesystem has not been asked to revalidate the
1309name (with ``d_revalidate()``).  In such cases the inode may still need
1310to be revalidated, so ``d_op->d_weak_revalidate()`` is called if
1311``ND_JUMPED`` is set when the look completes - which may be at the
1312final component or, when creating, unlinking, or renaming, at the penultimate component.
1313
1314Resolution-restriction flags
1315~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1316
1317In order to allow userspace to protect itself against certain race conditions
1318and attack scenarios involving changing path components, a series of flags are
1319available which apply restrictions to all path components encountered during
1320path lookup. These flags are exposed through ``openat2()``'s ``resolve`` field.
1321
1322``LOOKUP_NO_SYMLINKS`` blocks all symlink traversals (including magic-links).
1323This is distinctly different from ``LOOKUP_FOLLOW``, because the latter only
1324relates to restricting the following of trailing symlinks.
1325
1326``LOOKUP_NO_MAGICLINKS`` blocks all magic-link traversals. Filesystems must
1327ensure that they return errors from ``nd_jump_link()``, because that is how
1328``LOOKUP_NO_MAGICLINKS`` and other magic-link restrictions are implemented.
1329
1330``LOOKUP_NO_XDEV`` blocks all ``vfsmount`` traversals (this includes both
1331bind-mounts and ordinary mounts). Note that the ``vfsmount`` which contains the
1332lookup is determined by the first mountpoint the path lookup reaches --
1333absolute paths start with the ``vfsmount`` of ``/``, and relative paths start
1334with the ``dfd``'s ``vfsmount``. Magic-links are only permitted if the
1335``vfsmount`` of the path is unchanged.
1336
1337``LOOKUP_BENEATH`` blocks any path components which resolve outside the
1338starting point of the resolution. This is done by blocking ``nd_jump_root()``
1339as well as blocking ".." if it would jump outside the starting point.
1340``rename_lock`` and ``mount_lock`` are used to detect attacks against the
1341resolution of "..". Magic-links are also blocked.
1342
1343``LOOKUP_IN_ROOT`` resolves all path components as though the starting point
1344were the filesystem root. ``nd_jump_root()`` brings the resolution back to
1345the starting point, and ".." at the starting point will act as a no-op. As with
1346``LOOKUP_BENEATH``, ``rename_lock`` and ``mount_lock`` are used to detect
1347attacks against ".." resolution. Magic-links are also blocked.
1348
1349Final-component flags
1350~~~~~~~~~~~~~~~~~~~~~
1351
1352Some of these flags are only set when the final component is being
1353considered.  Others are only checked for when considering that final
1354component.
1355
1356``LOOKUP_AUTOMOUNT`` ensures that, if the final component is an automount
1357point, then the mount is triggered.  Some operations would trigger it
1358anyway, but operations like ``stat()`` deliberately don't.  ``statfs()``
1359needs to trigger the mount but otherwise behaves a lot like ``stat()``, so
1360it sets ``LOOKUP_AUTOMOUNT``, as does "``quotactl()``" and the handling of
1361"``mount --bind``".
1362
1363``LOOKUP_FOLLOW`` has a similar function to ``LOOKUP_AUTOMOUNT`` but for
1364symlinks.  Some system calls set or clear it implicitly, while
1365others have API flags such as ``AT_SYMLINK_FOLLOW`` and
1366``UMOUNT_NOFOLLOW`` to control it.  Its effect is similar to
1367``WALK_GET`` that we already met, but it is used in a different way.
1368
1369``LOOKUP_DIRECTORY`` insists that the final component is a directory.
1370Various callers set this and it is also set when the final component
1371is found to be followed by a slash.
1372
1373Finally ``LOOKUP_OPEN``, ``LOOKUP_CREATE``, ``LOOKUP_EXCL``, and
1374``LOOKUP_RENAME_TARGET`` are not used directly by the VFS but are made
1375available to the filesystem and particularly the ``->d_revalidate()``
1376method.  A filesystem can choose not to bother revalidating too hard
1377if it knows that it will be asked to open or create the file soon.
1378These flags were previously useful for ``->lookup()`` too but with the
1379introduction of ``->atomic_open()`` they are less relevant there.
1380
1381End of the road
1382---------------
1383
1384Despite its complexity, all this pathname lookup code appears to be
1385in good shape - various parts are certainly easier to understand now
1386than even a couple of releases ago.  But that doesn't mean it is
1387"finished".   As already mentioned, RCU-walk currently only follows
1388symlinks that are stored in the inode so, while it handles many ext4
1389symlinks, it doesn't help with NFS, XFS, or Btrfs.  That support
1390is not likely to be long delayed.

   1===============
   2Pathname lookup
   3===============
   4
   5This write-up is based on three articles published at lwn.net:
   6
   7- <https://lwn.net/Articles/649115/> Pathname lookup in Linux
   8- <https://lwn.net/Articles/649729/> RCU-walk: faster pathname lookup in Linux
   9- <https://lwn.net/Articles/650786/> A walk among the symlinks
  10
  11Written by Neil Brown with help from Al Viro and Jon Corbet.
  12It has subsequently been updated to reflect changes in the kernel
  13including:
  14
  15- per-directory parallel name lookup.
  16- ``openat2()`` resolution restriction flags.
  17
  18Introduction to pathname lookup
  19===============================
  20
  21The most obvious aspect of pathname lookup, which very little
  22exploration is needed to discover, is that it is complex.  There are
  23many rules, special cases, and implementation alternatives that all
  24combine to confuse the unwary reader.  Computer science has long been
  25acquainted with such complexity and has tools to help manage it.  One
  26tool that we will make extensive use of is "divide and conquer".  For
  27the early parts of the analysis we will divide off symlinks - leaving
  28them until the final part.  Well before we get to symlinks we have
  29another major division based on the VFS's approach to locking which
  30will allow us to review "REF-walk" and "RCU-walk" separately.  But we
  31are getting ahead of ourselves.  There are some important low level
  32distinctions we need to clarify first.
  33
  34There are two sorts of ...
  35--------------------------
  36
  37.. _openat: http://man7.org/linux/man-pages/man2/openat.2.html
  38
  39Pathnames (sometimes "file names"), used to identify objects in the
  40filesystem, will be familiar to most readers.  They contain two sorts
  41of elements: "slashes" that are sequences of one or more "``/``"
  42characters, and "components" that are sequences of one or more
  43non-"``/``" characters.  These form two kinds of paths.  Those that
  44start with slashes are "absolute" and start from the filesystem root.
  45The others are "relative" and start from the current directory, or
  46from some other location specified by a file descriptor given to
  47"``*at()``" system calls such as `openat() <openat_>`_.
  48
  49.. _execveat: http://man7.org/linux/man-pages/man2/execveat.2.html
  50
  51It is tempting to describe the second kind as starting with a
  52component, but that isn't always accurate: a pathname can lack both
  53slashes and components, it can be empty, in other words.  This is
  54generally forbidden in POSIX, but some of those "``*at()``" system calls
  55in Linux permit it when the ``AT_EMPTY_PATH`` flag is given.  For
  56example, if you have an open file descriptor on an executable file you
  57can execute it by calling `execveat() <execveat_>`_ passing
  58the file descriptor, an empty path, and the ``AT_EMPTY_PATH`` flag.
  59
  60These paths can be divided into two sections: the final component and
  61everything else.  The "everything else" is the easy bit.  In all cases
  62it must identify a directory that already exists, otherwise an error
  63such as ``ENOENT`` or ``ENOTDIR`` will be reported.
  64
  65The final component is not so simple.  Not only do different system
  66calls interpret it quite differently (e.g. some create it, some do
  67not), but it might not even exist: neither the empty pathname nor the
  68pathname that is just slashes have a final component.  If it does
  69exist, it could be "``.``" or "``..``" which are handled quite differently
  70from other components.
  71
  72.. _POSIX: https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_12
  73
  74If a pathname ends with a slash, such as "``/tmp/foo/``" it might be
  75tempting to consider that to have an empty final component.  In many
  76ways that would lead to correct results, but not always.  In
  77particular, ``mkdir()`` and ``rmdir()`` each create or remove a directory named
  78by the final component, and they are required to work with pathnames
  79ending in "``/``".  According to POSIX_:
  80
  81  A pathname that contains at least one non-<slash> character and
  82  that ends with one or more trailing <slash> characters shall not
  83  be resolved successfully unless the last pathname component before
  84  the trailing <slash> characters names an existing directory or a
  85  directory entry that is to be created for a directory immediately
  86  after the pathname is resolved.
  87
  88The Linux pathname walking code (mostly in ``fs/namei.c``) deals with
  89all of these issues: breaking the path into components, handling the
  90"everything else" quite separately from the final component, and
  91checking that the trailing slash is not used where it isn't
  92permitted.  It also addresses the important issue of concurrent
  93access.
  94
  95While one process is looking up a pathname, another might be making
  96changes that affect that lookup.  One fairly extreme case is that if
  97"a/b" were renamed to "a/c/b" while another process were looking up
  98"a/b/..", that process might successfully resolve on "a/c".
  99Most races are much more subtle, and a big part of the task of
 100pathname lookup is to prevent them from having damaging effects.  Many
 101of the possible races are seen most clearly in the context of the
 102"dcache" and an understanding of that is central to understanding
 103pathname lookup.
 104
 105More than just a cache
 106----------------------
 107
 108The "dcache" caches information about names in each filesystem to
 109make them quickly available for lookup.  Each entry (known as a
 110"dentry") contains three significant fields: a component name, a
 111pointer to a parent dentry, and a pointer to the "inode" which
 112contains further information about the object in that parent with
 113the given name.  The inode pointer can be ``NULL`` indicating that the
 114name doesn't exist in the parent.  While there can be linkage in the
 115dentry of a directory to the dentries of the children, that linkage is
 116not used for pathname lookup, and so will not be considered here.
 117
 118The dcache has a number of uses apart from accelerating lookup.  One
 119that will be particularly relevant is that it is closely integrated
 120with the mount table that records which filesystem is mounted where.
 121What the mount table actually stores is which dentry is mounted on top
 122of which other dentry.
 123
 124When considering the dcache, we have another of our "two types"
 125distinctions: there are two types of filesystems.
 126
 127Some filesystems ensure that the information in the dcache is always
 128completely accurate (though not necessarily complete).  This can allow
 129the VFS to determine if a particular file does or doesn't exist
 130without checking with the filesystem, and means that the VFS can
 131protect the filesystem against certain races and other problems.
 132These are typically "local" filesystems such as ext3, XFS, and Btrfs.
 133
 134Other filesystems don't provide that guarantee because they cannot.
 135These are typically filesystems that are shared across a network,
 136whether remote filesystems like NFS and 9P, or cluster filesystems
 137like ocfs2 or cephfs.  These filesystems allow the VFS to revalidate
 138cached information, and must provide their own protection against
 139awkward races.  The VFS can detect these filesystems by the
 140``DCACHE_OP_REVALIDATE`` flag being set in the dentry.
 141
 142REF-walk: simple concurrency management with refcounts and spinlocks
 143--------------------------------------------------------------------
 144
 145With all of those divisions carefully classified, we can now start
 146looking at the actual process of walking along a path.  In particular
 147we will start with the handling of the "everything else" part of a
 148pathname, and focus on the "REF-walk" approach to concurrency
 149management.  This code is found in the ``link_path_walk()`` function, if
 150you ignore all the places that only run when "``LOOKUP_RCU``"
 151(indicating the use of RCU-walk) is set.
 152
 153.. _Meet the Lockers: https://lwn.net/Articles/453685/
 154
 155REF-walk is fairly heavy-handed with locks and reference counts.  Not
 156as heavy-handed as in the old "big kernel lock" days, but certainly not
 157afraid of taking a lock when one is needed.  It uses a variety of
 158different concurrency controls.  A background understanding of the
 159various primitives is assumed, or can be gleaned from elsewhere such
 160as in `Meet the Lockers`_.
 161
 162The locking mechanisms used by REF-walk include:
 163
 164dentry->d_lockref
 165~~~~~~~~~~~~~~~~~
 166
 167This uses the lockref primitive to provide both a spinlock and a
 168reference count.  The special-sauce of this primitive is that the
 169conceptual sequence "lock; inc_ref; unlock;" can often be performed
 170with a single atomic memory operation.
 171
 172Holding a reference on a dentry ensures that the dentry won't suddenly
 173be freed and used for something else, so the values in various fields
 174will behave as expected.  It also protects the ``->d_inode`` reference
 175to the inode to some extent.
 176
 177The association between a dentry and its inode is fairly permanent.
 178For example, when a file is renamed, the dentry and inode move
 179together to the new location.  When a file is created the dentry will
 180initially be negative (i.e. ``d_inode`` is ``NULL``), and will be assigned
 181to the new inode as part of the act of creation.
 182
 183When a file is deleted, this can be reflected in the cache either by
 184setting ``d_inode`` to ``NULL``, or by removing it from the hash table
 185(described shortly) used to look up the name in the parent directory.
 186If the dentry is still in use the second option is used as it is
 187perfectly legal to keep using an open file after it has been deleted
 188and having the dentry around helps.  If the dentry is not otherwise in
 189use (i.e. if the refcount in ``d_lockref`` is one), only then will
 190``d_inode`` be set to ``NULL``.  Doing it this way is more efficient for a
 191very common case.
 192
 193So as long as a counted reference is held to a dentry, a non-``NULL`` ``->d_inode``
 194value will never be changed.
 195
 196dentry->d_lock
 197~~~~~~~~~~~~~~
 198
 199``d_lock`` is a synonym for the spinlock that is part of ``d_lockref`` above.
 200For our purposes, holding this lock protects against the dentry being
 201renamed or unlinked.  In particular, its parent (``d_parent``), and its
 202name (``d_name``) cannot be changed, and it cannot be removed from the
 203dentry hash table.
 204
 205When looking for a name in a directory, REF-walk takes ``d_lock`` on
 206each candidate dentry that it finds in the hash table and then checks
 207that the parent and name are correct.  So it doesn't lock the parent
 208while searching in the cache; it only locks children.
 209
 210When looking for the parent for a given name (to handle "``..``"),
 211REF-walk can take ``d_lock`` to get a stable reference to ``d_parent``,
 212but it first tries a more lightweight approach.  As seen in
 213``dget_parent()``, if a reference can be claimed on the parent, and if
 214subsequently ``d_parent`` can be seen to have not changed, then there is
 215no need to actually take the lock on the child.
 216
 217rename_lock
 218~~~~~~~~~~~
 219
 220Looking up a given name in a given directory involves computing a hash
 221from the two values (the name and the dentry of the directory),
 222accessing that slot in a hash table, and searching the linked list
 223that is found there.
 224
 225When a dentry is renamed, the name and the parent dentry can both
 226change so the hash will almost certainly change too.  This would move the
 227dentry to a different chain in the hash table.  If a filename search
 228happened to be looking at a dentry that was moved in this way,
 229it might end up continuing the search down the wrong chain,
 230and so miss out on part of the correct chain.
 231
 232The name-lookup process (``d_lookup()``) does *not* try to prevent this
 233from happening, but only to detect when it happens.
 234``rename_lock`` is a seqlock that is updated whenever any dentry is
 235renamed.  If ``d_lookup`` finds that a rename happened while it
 236unsuccessfully scanned a chain in the hash table, it simply tries
 237again.
 238
 239``rename_lock`` is also used to detect and defend against potential attacks
 240against ``LOOKUP_BENEATH`` and ``LOOKUP_IN_ROOT`` when resolving ".." (where
 241the parent directory is moved outside the root, bypassing the ``path_equal()``
 242check). If ``rename_lock`` is updated during the lookup and the path encounters
 243a "..", a potential attack occurred and ``handle_dots()`` will bail out with
 244``-EAGAIN``.
 245
 246inode->i_rwsem
 247~~~~~~~~~~~~~~
 248
 249``i_rwsem`` is a read/write semaphore that serializes all changes to a particular
 250directory.  This ensures that, for example, an ``unlink()`` and a ``rename()``
 251cannot both happen at the same time.  It also keeps the directory
 252stable while the filesystem is asked to look up a name that is not
 253currently in the dcache or, optionally, when the list of entries in a
 254directory is being retrieved with ``readdir()``.
 255
 256This has a complementary role to that of ``d_lock``: ``i_rwsem`` on a
 257directory protects all of the names in that directory, while ``d_lock``
 258on a name protects just one name in a directory.  Most changes to the
 259dcache hold ``i_rwsem`` on the relevant directory inode and briefly take
 260``d_lock`` on one or more the dentries while the change happens.  One
 261exception is when idle dentries are removed from the dcache due to
 262memory pressure.  This uses ``d_lock``, but ``i_rwsem`` plays no role.
 263
 264The semaphore affects pathname lookup in two distinct ways.  Firstly it
 265prevents changes during lookup of a name in a directory.  ``walk_component()`` uses
 266``lookup_fast()`` first which, in turn, checks to see if the name is in the cache,
 267using only ``d_lock`` locking.  If the name isn't found, then ``walk_component()``
 268falls back to ``lookup_slow()`` which takes a shared lock on ``i_rwsem``, checks again that
 269the name isn't in the cache, and then calls in to the filesystem to get a
 270definitive answer.  A new dentry will be added to the cache regardless of
 271the result.
 272
 273Secondly, when pathname lookup reaches the final component, it will
 274sometimes need to take an exclusive lock on ``i_rwsem`` before performing the last lookup so
 275that the required exclusion can be achieved.  How path lookup chooses
 276to take, or not take, ``i_rwsem`` is one of the
 277issues addressed in a subsequent section.
 278
 279If two threads attempt to look up the same name at the same time - a
 280name that is not yet in the dcache - the shared lock on ``i_rwsem`` will
 281not prevent them both adding new dentries with the same name.  As this
 282would result in confusion an extra level of interlocking is used,
 283based around a secondary hash table (``in_lookup_hashtable``) and a
 284per-dentry flag bit (``DCACHE_PAR_LOOKUP``).
 285
 286To add a new dentry to the cache while only holding a shared lock on
 287``i_rwsem``, a thread must call ``d_alloc_parallel()``.  This allocates a
 288dentry, stores the required name and parent in it, checks if there
 289is already a matching dentry in the primary or secondary hash
 290tables, and if not, stores the newly allocated dentry in the secondary
 291hash table, with ``DCACHE_PAR_LOOKUP`` set.
 292
 293If a matching dentry was found in the primary hash table then that is
 294returned and the caller can know that it lost a race with some other
 295thread adding the entry.  If no matching dentry is found in either
 296cache, the newly allocated dentry is returned and the caller can
 297detect this from the presence of ``DCACHE_PAR_LOOKUP``.  In this case it
 298knows that it has won any race and now is responsible for asking the
 299filesystem to perform the lookup and find the matching inode.  When
 300the lookup is complete, it must call ``d_lookup_done()`` which clears
 301the flag and does some other house keeping, including removing the
 302dentry from the secondary hash table - it will normally have been
 303added to the primary hash table already.  Note that a ``struct
 304waitqueue_head`` is passed to ``d_alloc_parallel()``, and
 305``d_lookup_done()`` must be called while this ``waitqueue_head`` is still
 306in scope.
 307
 308If a matching dentry is found in the secondary hash table,
 309``d_alloc_parallel()`` has a little more work to do. It first waits for
 310``DCACHE_PAR_LOOKUP`` to be cleared, using a wait_queue that was passed
 311to the instance of ``d_alloc_parallel()`` that won the race and that
 312will be woken by the call to ``d_lookup_done()``.  It then checks to see
 313if the dentry has now been added to the primary hash table.  If it
 314has, the dentry is returned and the caller just sees that it lost any
 315race.  If it hasn't been added to the primary hash table, the most
 316likely explanation is that some other dentry was added instead using
 317``d_splice_alias()``.  In any case, ``d_alloc_parallel()`` repeats all the
 318look ups from the start and will normally return something from the
 319primary hash table.
 320
 321mnt->mnt_count
 322~~~~~~~~~~~~~~
 323
 324``mnt_count`` is a per-CPU reference counter on "``mount``" structures.
 325Per-CPU here means that incrementing the count is cheap as it only
 326uses CPU-local memory, but checking if the count is zero is expensive as
 327it needs to check with every CPU.  Taking a ``mnt_count`` reference
 328prevents the mount structure from disappearing as the result of regular
 329unmount operations, but does not prevent a "lazy" unmount.  So holding
 330``mnt_count`` doesn't ensure that the mount remains in the namespace and,
 331in particular, doesn't stabilize the link to the mounted-on dentry.  It
 332does, however, ensure that the ``mount`` data structure remains coherent,
 333and it provides a reference to the root dentry of the mounted
 334filesystem.  So a reference through ``->mnt_count`` provides a stable
 335reference to the mounted dentry, but not the mounted-on dentry.
 336
 337mount_lock
 338~~~~~~~~~~
 339
 340``mount_lock`` is a global seqlock, a bit like ``rename_lock``.  It can be used to
 341check if any change has been made to any mount points.
 342
 343While walking down the tree (away from the root) this lock is used when
 344crossing a mount point to check that the crossing was safe.  That is,
 345the value in the seqlock is read, then the code finds the mount that
 346is mounted on the current directory, if there is one, and increments
 347the ``mnt_count``.  Finally the value in ``mount_lock`` is checked against
 348the old value.  If there is no change, then the crossing was safe.  If there
 349was a change, the ``mnt_count`` is decremented and the whole process is
 350retried.
 351
 352When walking up the tree (towards the root) by following a ".." link,
 353a little more care is needed.  In this case the seqlock (which
 354contains both a counter and a spinlock) is fully locked to prevent
 355any changes to any mount points while stepping up.  This locking is
 356needed to stabilize the link to the mounted-on dentry, which the
 357refcount on the mount itself doesn't ensure.
 358
 359``mount_lock`` is also used to detect and defend against potential attacks
 360against ``LOOKUP_BENEATH`` and ``LOOKUP_IN_ROOT`` when resolving ".." (where
 361the parent directory is moved outside the root, bypassing the ``path_equal()``
 362check). If ``mount_lock`` is updated during the lookup and the path encounters
 363a "..", a potential attack occurred and ``handle_dots()`` will bail out with
 364``-EAGAIN``.
 365
 366RCU
 367~~~
 368
 369Finally the global (but extremely lightweight) RCU read lock is held
 370from time to time to ensure certain data structures don't get freed
 371unexpectedly.
 372
 373In particular it is held while scanning chains in the dcache hash
 374table, and the mount point hash table.
 375
 376Bringing it together with ``struct nameidata``
 377----------------------------------------------
 378
 379.. _First edition Unix: https://minnie.tuhs.org/cgi-bin/utree.pl?file=V1/u2.s
 380
 381Throughout the process of walking a path, the current status is stored
 382in a ``struct nameidata``, "namei" being the traditional name - dating
 383all the way back to `First Edition Unix`_ - of the function that
 384converts a "name" to an "inode".  ``struct nameidata`` contains (among
 385other fields):
 386
 387``struct path path``
 388~~~~~~~~~~~~~~~~~~~~
 389
 390A ``path`` contains a ``struct vfsmount`` (which is
 391embedded in a ``struct mount``) and a ``struct dentry``.  Together these
 392record the current status of the walk.  They start out referring to the
 393starting point (the current working directory, the root directory, or some other
 394directory identified by a file descriptor), and are updated on each
 395step.  A reference through ``d_lockref`` and ``mnt_count`` is always
 396held.
 397
 398``struct qstr last``
 399~~~~~~~~~~~~~~~~~~~~
 400
 401This is a string together with a length (i.e. *not* ``nul`` terminated)
 402that is the "next" component in the pathname.
 403
 404``int last_type``
 405~~~~~~~~~~~~~~~~~
 406
 407This is one of ``LAST_NORM``, ``LAST_ROOT``, ``LAST_DOT`` or ``LAST_DOTDOT``.
 408The ``last`` field is only valid if the type is ``LAST_NORM``.
 409
 410``struct path root``
 411~~~~~~~~~~~~~~~~~~~~
 412
 413This is used to hold a reference to the effective root of the
 414filesystem.  Often that reference won't be needed, so this field is
 415only assigned the first time it is used, or when a non-standard root
 416is requested.  Keeping a reference in the ``nameidata`` ensures that
 417only one root is in effect for the entire path walk, even if it races
 418with a ``chroot()`` system call.
 419
 420It should be noted that in the case of ``LOOKUP_IN_ROOT`` or
 421``LOOKUP_BENEATH``, the effective root becomes the directory file descriptor
 422passed to ``openat2()`` (which exposes these ``LOOKUP_`` flags).
 423
 424The root is needed when either of two conditions holds: (1) either the
 425pathname or a symbolic link starts with a "'/'", or (2) a "``..``"
 426component is being handled, since "``..``" from the root must always stay
 427at the root.  The value used is usually the current root directory of
 428the calling process.  An alternate root can be provided as when
 429``sysctl()`` calls ``file_open_root()``, and when NFSv4 or Btrfs call
 430``mount_subtree()``.  In each case a pathname is being looked up in a very
 431specific part of the filesystem, and the lookup must not be allowed to
 432escape that subtree.  It works a bit like a local ``chroot()``.
 433
 434Ignoring the handling of symbolic links, we can now describe the
 435"``link_path_walk()``" function, which handles the lookup of everything
 436except the final component as:
 437
 438   Given a path (``name``) and a nameidata structure (``nd``), check that the
 439   current directory has execute permission and then advance ``name``
 440   over one component while updating ``last_type`` and ``last``.  If that
 441   was the final component, then return, otherwise call
 442   ``walk_component()`` and repeat from the top.
 443
 444``walk_component()`` is even easier.  If the component is ``LAST_DOTS``,
 445it calls ``handle_dots()`` which does the necessary locking as already
 446described.  If it finds a ``LAST_NORM`` component it first calls
 447"``lookup_fast()``" which only looks in the dcache, but will ask the
 448filesystem to revalidate the result if it is that sort of filesystem.
 449If that doesn't get a good result, it calls "``lookup_slow()``" which
 450takes ``i_rwsem``, rechecks the cache, and then asks the filesystem
 451to find a definitive answer.
 452
 453As the last step of walk_component(), step_into() will be called either
 454directly from walk_component() or from handle_dots().  It calls
 455handle_mounts(), to check and handle mount points, in which a new
 456``struct path`` is created containing a counted reference to the new dentry and
 457a reference to the new ``vfsmount`` which is only counted if it is
 458different from the previous ``vfsmount``. Then if there is
 459a symbolic link, step_into() calls pick_link() to deal with it,
 460otherwise it installs the new ``struct path`` in the ``struct nameidata``, and
 461drops the unneeded references.
 462
 463This "hand-over-hand" sequencing of getting a reference to the new
 464dentry before dropping the reference to the previous dentry may
 465seem obvious, but is worth pointing out so that we will recognize its
 466analogue in the "RCU-walk" version.
 467
 468Handling the final component
 469----------------------------
 470
 471``link_path_walk()`` only walks as far as setting ``nd->last`` and
 472``nd->last_type`` to refer to the final component of the path.  It does
 473not call ``walk_component()`` that last time.  Handling that final
 474component remains for the caller to sort out. Those callers are
 475path_lookupat(), path_parentat() and
 476path_openat() each of which handles the differing requirements of
 477different system calls.
 478
 479``path_parentat()`` is clearly the simplest - it just wraps a little bit
 480of housekeeping around ``link_path_walk()`` and returns the parent
 481directory and final component to the caller.  The caller will be either
 482aiming to create a name (via ``filename_create()``) or remove or rename
 483a name (in which case ``user_path_parent()`` is used).  They will use
 484``i_rwsem`` to exclude other changes while they validate and then
 485perform their operation.
 486
 487``path_lookupat()`` is nearly as simple - it is used when an existing
 488object is wanted such as by ``stat()`` or ``chmod()``.  It essentially just
 489calls ``walk_component()`` on the final component through a call to
 490``lookup_last()``.  ``path_lookupat()`` returns just the final dentry.
 491It is worth noting that when flag ``LOOKUP_MOUNTPOINT`` is set,
 492path_lookupat() will unset LOOKUP_JUMPED in nameidata so that in the
 493subsequent path traversal d_weak_revalidate() won't be called.
 494This is important when unmounting a filesystem that is inaccessible, such as
 495one provided by a dead NFS server.
 496
 497Finally ``path_openat()`` is used for the ``open()`` system call; it
 498contains, in support functions starting with "open_last_lookups()", all the
 499complexity needed to handle the different subtleties of O_CREAT (with
 500or without O_EXCL), final "``/``" characters, and trailing symbolic
 501links.  We will revisit this in the final part of this series, which
 502focuses on those symbolic links.  "open_last_lookups()" will sometimes, but
 503not always, take ``i_rwsem``, depending on what it finds.
 504
 505Each of these, or the functions which call them, need to be alert to
 506the possibility that the final component is not ``LAST_NORM``.  If the
 507goal of the lookup is to create something, then any value for
 508``last_type`` other than ``LAST_NORM`` will result in an error.  For
 509example if ``path_parentat()`` reports ``LAST_DOTDOT``, then the caller
 510won't try to create that name.  They also check for trailing slashes
 511by testing ``last.name[last.len]``.  If there is any character beyond
 512the final component, it must be a trailing slash.
 513
 514Revalidation and automounts
 515---------------------------
 516
 517Apart from symbolic links, there are only two parts of the "REF-walk"
 518process not yet covered.  One is the handling of stale cache entries
 519and the other is automounts.
 520
 521On filesystems that require it, the lookup routines will call the
 522``->d_revalidate()`` dentry method to ensure that the cached information
 523is current.  This will often confirm validity or update a few details
 524from a server.  In some cases it may find that there has been change
 525further up the path and that something that was thought to be valid
 526previously isn't really.  When this happens the lookup of the whole
 527path is aborted and retried with the "``LOOKUP_REVAL``" flag set.  This
 528forces revalidation to be more thorough.  We will see more details of
 529this retry process in the next article.
 530
 531Automount points are locations in the filesystem where an attempt to
 532lookup a name can trigger changes to how that lookup should be
 533handled, in particular by mounting a filesystem there.  These are
 534covered in greater detail in autofs.txt in the Linux documentation
 535tree, but a few notes specifically related to path lookup are in order
 536here.
 537
 538The Linux VFS has a concept of "managed" dentries.  There are three
 539potentially interesting things about these dentries corresponding
 540to three different flags that might be set in ``dentry->d_flags``:
 541
 542``DCACHE_MANAGE_TRANSIT``
 543~~~~~~~~~~~~~~~~~~~~~~~~~
 544
 545If this flag has been set, then the filesystem has requested that the
 546``d_manage()`` dentry operation be called before handling any possible
 547mount point.  This can perform two particular services:
 548
 549It can block to avoid races.  If an automount point is being
 550unmounted, the ``d_manage()`` function will usually wait for that
 551process to complete before letting the new lookup proceed and possibly
 552trigger a new automount.
 553
 554It can selectively allow only some processes to transit through a
 555mount point.  When a server process is managing automounts, it may
 556need to access a directory without triggering normal automount
 557processing.  That server process can identify itself to the ``autofs``
 558filesystem, which will then give it a special pass through
 559``d_manage()`` by returning ``-EISDIR``.
 560
 561``DCACHE_MOUNTED``
 562~~~~~~~~~~~~~~~~~~
 563
 564This flag is set on every dentry that is mounted on.  As Linux
 565supports multiple filesystem namespaces, it is possible that the
 566dentry may not be mounted on in *this* namespace, just in some
 567other.  So this flag is seen as a hint, not a promise.
 568
 569If this flag is set, and ``d_manage()`` didn't return ``-EISDIR``,
 570``lookup_mnt()`` is called to examine the mount hash table (honoring the
 571``mount_lock`` described earlier) and possibly return a new ``vfsmount``
 572and a new ``dentry`` (both with counted references).
 573
 574``DCACHE_NEED_AUTOMOUNT``
 575~~~~~~~~~~~~~~~~~~~~~~~~~
 576
 577If ``d_manage()`` allowed us to get this far, and ``lookup_mnt()`` didn't
 578find a mount point, then this flag causes the ``d_automount()`` dentry
 579operation to be called.
 580
 581The ``d_automount()`` operation can be arbitrarily complex and may
 582communicate with server processes etc. but it should ultimately either
 583report that there was an error, that there was nothing to mount, or
 584should provide an updated ``struct path`` with new ``dentry`` and ``vfsmount``.
 585
 586In the latter case, ``finish_automount()`` will be called to safely
 587install the new mount point into the mount table.
 588
 589There is no new locking of import here and it is important that no
 590locks (only counted references) are held over this processing due to
 591the very real possibility of extended delays.
 592This will become more important next time when we examine RCU-walk
 593which is particularly sensitive to delays.
 594
 595RCU-walk - faster pathname lookup in Linux
 596==========================================
 597
 598RCU-walk is another algorithm for performing pathname lookup in Linux.
 599It is in many ways similar to REF-walk and the two share quite a bit
 600of code.  The significant difference in RCU-walk is how it allows for
 601the possibility of concurrent access.
 602
 603We noted that REF-walk is complex because there are numerous details
 604and special cases.  RCU-walk reduces this complexity by simply
 605refusing to handle a number of cases -- it instead falls back to
 606REF-walk.  The difficulty with RCU-walk comes from a different
 607direction: unfamiliarity.  The locking rules when depending on RCU are
 608quite different from traditional locking, so we will spend a little extra
 609time when we come to those.
 610
 611Clear demarcation of roles
 612--------------------------
 613
 614The easiest way to manage concurrency is to forcibly stop any other
 615thread from changing the data structures that a given thread is
 616looking at.  In cases where no other thread would even think of
 617changing the data and lots of different threads want to read at the
 618same time, this can be very costly.  Even when using locks that permit
 619multiple concurrent readers, the simple act of updating the count of
 620the number of current readers can impose an unwanted cost.  So the
 621goal when reading a shared data structure that no other process is
 622changing is to avoid writing anything to memory at all.  Take no
 623locks, increment no counts, leave no footprints.
 624
 625The REF-walk mechanism already described certainly doesn't follow this
 626principle, but then it is really designed to work when there may well
 627be other threads modifying the data.  RCU-walk, in contrast, is
 628designed for the common situation where there are lots of frequent
 629readers and only occasional writers.  This may not be common in all
 630parts of the filesystem tree, but in many parts it will be.  For the
 631other parts it is important that RCU-walk can quickly fall back to
 632using REF-walk.
 633
 634Pathname lookup always starts in RCU-walk mode but only remains there
 635as long as what it is looking for is in the cache and is stable.  It
 636dances lightly down the cached filesystem image, leaving no footprints
 637and carefully watching where it is, to be sure it doesn't trip.  If it
 638notices that something has changed or is changing, or if something
 639isn't in the cache, then it tries to stop gracefully and switch to
 640REF-walk.
 641
 642This stopping requires getting a counted reference on the current
 643``vfsmount`` and ``dentry``, and ensuring that these are still valid -
 644that a path walk with REF-walk would have found the same entries.
 645This is an invariant that RCU-walk must guarantee.  It can only make
 646decisions, such as selecting the next step, that are decisions which
 647REF-walk could also have made if it were walking down the tree at the
 648same time.  If the graceful stop succeeds, the rest of the path is
 649processed with the reliable, if slightly sluggish, REF-walk.  If
 650RCU-walk finds it cannot stop gracefully, it simply gives up and
 651restarts from the top with REF-walk.
 652
 653This pattern of "try RCU-walk, if that fails try REF-walk" can be
 654clearly seen in functions like filename_lookup(),
 655filename_parentat(),
 656do_filp_open(), and do_file_open_root().  These four
 657correspond roughly to the three ``path_*()`` functions we met earlier,
 658each of which calls ``link_path_walk()``.  The ``path_*()`` functions are
 659called using different mode flags until a mode is found which works.
 660They are first called with ``LOOKUP_RCU`` set to request "RCU-walk".  If
 661that fails with the error ``ECHILD`` they are called again with no
 662special flag to request "REF-walk".  If either of those report the
 663error ``ESTALE`` a final attempt is made with ``LOOKUP_REVAL`` set (and no
 664``LOOKUP_RCU``) to ensure that entries found in the cache are forcibly
 665revalidated - normally entries are only revalidated if the filesystem
 666determines that they are too old to trust.
 667
 668The ``LOOKUP_RCU`` attempt may drop that flag internally and switch to
 669REF-walk, but will never then try to switch back to RCU-walk.  Places
 670that trip up RCU-walk are much more likely to be near the leaves and
 671so it is very unlikely that there will be much, if any, benefit from
 672switching back.
 673
 674RCU and seqlocks: fast and light
 675--------------------------------
 676
 677RCU is, unsurprisingly, critical to RCU-walk mode.  The
 678``rcu_read_lock()`` is held for the entire time that RCU-walk is walking
 679down a path.  The particular guarantee it provides is that the key
 680data structures - dentries, inodes, super_blocks, and mounts - will
 681not be freed while the lock is held.  They might be unlinked or
 682invalidated in one way or another, but the memory will not be
 683repurposed so values in various fields will still be meaningful.  This
 684is the only guarantee that RCU provides; everything else is done using
 685seqlocks.
 686
 687As we saw above, REF-walk holds a counted reference to the current
 688dentry and the current vfsmount, and does not release those references
 689before taking references to the "next" dentry or vfsmount.  It also
 690sometimes takes the ``d_lock`` spinlock.  These references and locks are
 691taken to prevent certain changes from happening.  RCU-walk must not
 692take those references or locks and so cannot prevent such changes.
 693Instead, it checks to see if a change has been made, and aborts or
 694retries if it has.
 695
 696To preserve the invariant mentioned above (that RCU-walk may only make
 697decisions that REF-walk could have made), it must make the checks at
 698or near the same places that REF-walk holds the references.  So, when
 699REF-walk increments a reference count or takes a spinlock, RCU-walk
 700samples the status of a seqlock using ``read_seqcount_begin()`` or a
 701similar function.  When REF-walk decrements the count or drops the
 702lock, RCU-walk checks if the sampled status is still valid using
 703``read_seqcount_retry()`` or similar.
 704
 705However, there is a little bit more to seqlocks than that.  If
 706RCU-walk accesses two different fields in a seqlock-protected
 707structure, or accesses the same field twice, there is no a priori
 708guarantee of any consistency between those accesses.  When consistency
 709is needed - which it usually is - RCU-walk must take a copy and then
 710use ``read_seqcount_retry()`` to validate that copy.
 711
 712``read_seqcount_retry()`` not only checks the sequence number, but also
 713imposes a memory barrier so that no memory-read instruction from
 714*before* the call can be delayed until *after* the call, either by the
 715CPU or by the compiler.  A simple example of this can be seen in
 716``slow_dentry_cmp()`` which, for filesystems which do not use simple
 717byte-wise name equality, calls into the filesystem to compare a name
 718against a dentry.  The length and name pointer are copied into local
 719variables, then ``read_seqcount_retry()`` is called to confirm the two
 720are consistent, and only then is ``->d_compare()`` called.  When
 721standard filename comparison is used, ``dentry_cmp()`` is called
 722instead.  Notably it does *not* use ``read_seqcount_retry()``, but
 723instead has a large comment explaining why the consistency guarantee
 724isn't necessary.  A subsequent ``read_seqcount_retry()`` will be
 725sufficient to catch any problem that could occur at this point.
 726
 727With that little refresher on seqlocks out of the way we can look at
 728the bigger picture of how RCU-walk uses seqlocks.
 729
 730``mount_lock`` and ``nd->m_seq``
 731~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 732
 733We already met the ``mount_lock`` seqlock when REF-walk used it to
 734ensure that crossing a mount point is performed safely.  RCU-walk uses
 735it for that too, but for quite a bit more.
 736
 737Instead of taking a counted reference to each ``vfsmount`` as it
 738descends the tree, RCU-walk samples the state of ``mount_lock`` at the
 739start of the walk and stores this initial sequence number in the
 740``struct nameidata`` in the ``m_seq`` field.  This one lock and one
 741sequence number are used to validate all accesses to all ``vfsmounts``,
 742and all mount point crossings.  As changes to the mount table are
 743relatively rare, it is reasonable to fall back on REF-walk any time
 744that any "mount" or "unmount" happens.
 745
 746``m_seq`` is checked (using ``read_seqretry()``) at the end of an RCU-walk
 747sequence, whether switching to REF-walk for the rest of the path or
 748when the end of the path is reached.  It is also checked when stepping
 749down over a mount point (in ``__follow_mount_rcu()``) or up (in
 750``follow_dotdot_rcu()``).  If it is ever found to have changed, the
 751whole RCU-walk sequence is aborted and the path is processed again by
 752REF-walk.
 753
 754If RCU-walk finds that ``mount_lock`` hasn't changed then it can be sure
 755that, had REF-walk taken counted references on each vfsmount, the
 756results would have been the same.  This ensures the invariant holds,
 757at least for vfsmount structures.
 758
 759``dentry->d_seq`` and ``nd->seq``
 760~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 761
 762In place of taking a count or lock on ``d_reflock``, RCU-walk samples
 763the per-dentry ``d_seq`` seqlock, and stores the sequence number in the
 764``seq`` field of the nameidata structure, so ``nd->seq`` should always be
 765the current sequence number of ``nd->dentry``.  This number needs to be
 766revalidated after copying, and before using, the name, parent, or
 767inode of the dentry.
 768
 769The handling of the name we have already looked at, and the parent is
 770only accessed in ``follow_dotdot_rcu()`` which fairly trivially follows
 771the required pattern, though it does so for three different cases.
 772
 773When not at a mount point, ``d_parent`` is followed and its ``d_seq`` is
 774collected.  When we are at a mount point, we instead follow the
 775``mnt->mnt_mountpoint`` link to get a new dentry and collect its
 776``d_seq``.  Then, after finally finding a ``d_parent`` to follow, we must
 777check if we have landed on a mount point and, if so, must find that
 778mount point and follow the ``mnt->mnt_root`` link.  This would imply a
 779somewhat unusual, but certainly possible, circumstance where the
 780starting point of the path lookup was in part of the filesystem that
 781was mounted on, and so not visible from the root.
 782
 783The inode pointer, stored in ``->d_inode``, is a little more
 784interesting.  The inode will always need to be accessed at least
 785twice, once to determine if it is NULL and once to verify access
 786permissions.  Symlink handling requires a validated inode pointer too.
 787Rather than revalidating on each access, a copy is made on the first
 788access and it is stored in the ``inode`` field of ``nameidata`` from where
 789it can be safely accessed without further validation.
 790
 791``lookup_fast()`` is the only lookup routine that is used in RCU-mode,
 792``lookup_slow()`` being too slow and requiring locks.  It is in
 793``lookup_fast()`` that we find the important "hand over hand" tracking
 794of the current dentry.
 795
 796The current ``dentry`` and current ``seq`` number are passed to
 797``__d_lookup_rcu()`` which, on success, returns a new ``dentry`` and a
 798new ``seq`` number.  ``lookup_fast()`` then copies the inode pointer and
 799revalidates the new ``seq`` number.  It then validates the old ``dentry``
 800with the old ``seq`` number one last time and only then continues.  This
 801process of getting the ``seq`` number of the new dentry and then
 802checking the ``seq`` number of the old exactly mirrors the process of
 803getting a counted reference to the new dentry before dropping that for
 804the old dentry which we saw in REF-walk.
 805
 806No ``inode->i_rwsem`` or even ``rename_lock``
 807~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 808
 809A semaphore is a fairly heavyweight lock that can only be taken when it is
 810permissible to sleep.  As ``rcu_read_lock()`` forbids sleeping,
 811``inode->i_rwsem`` plays no role in RCU-walk.  If some other thread does
 812take ``i_rwsem`` and modifies the directory in a way that RCU-walk needs
 813to notice, the result will be either that RCU-walk fails to find the
 814dentry that it is looking for, or it will find a dentry which
 815``read_seqretry()`` won't validate.  In either case it will drop down to
 816REF-walk mode which can take whatever locks are needed.
 817
 818Though ``rename_lock`` could be used by RCU-walk as it doesn't require
 819any sleeping, RCU-walk doesn't bother.  REF-walk uses ``rename_lock`` to
 820protect against the possibility of hash chains in the dcache changing
 821while they are being searched.  This can result in failing to find
 822something that actually is there.  When RCU-walk fails to find
 823something in the dentry cache, whether it is really there or not, it
 824already drops down to REF-walk and tries again with appropriate
 825locking.  This neatly handles all cases, so adding extra checks on
 826rename_lock would bring no significant value.
 827
 828``unlazy walk()`` and ``complete_walk()``
 829-----------------------------------------
 830
 831That "dropping down to REF-walk" typically involves a call to
 832``unlazy_walk()``, so named because "RCU-walk" is also sometimes
 833referred to as "lazy walk".  ``unlazy_walk()`` is called when
 834following the path down to the current vfsmount/dentry pair seems to
 835have proceeded successfully, but the next step is problematic.  This
 836can happen if the next name cannot be found in the dcache, if
 837permission checking or name revalidation couldn't be achieved while
 838the ``rcu_read_lock()`` is held (which forbids sleeping), if an
 839automount point is found, or in a couple of cases involving symlinks.
 840It is also called from ``complete_walk()`` when the lookup has reached
 841the final component, or the very end of the path, depending on which
 842particular flavor of lookup is used.
 843
 844Other reasons for dropping out of RCU-walk that do not trigger a call
 845to ``unlazy_walk()`` are when some inconsistency is found that cannot be
 846handled immediately, such as ``mount_lock`` or one of the ``d_seq``
 847seqlocks reporting a change.  In these cases the relevant function
 848will return ``-ECHILD`` which will percolate up until it triggers a new
 849attempt from the top using REF-walk.
 850
 851For those cases where ``unlazy_walk()`` is an option, it essentially
 852takes a reference on each of the pointers that it holds (vfsmount,
 853dentry, and possibly some symbolic links) and then verifies that the
 854relevant seqlocks have not been changed.  If there have been changes,
 855it, too, aborts with ``-ECHILD``, otherwise the transition to REF-walk
 856has been a success and the lookup process continues.
 857
 858Taking a reference on those pointers is not quite as simple as just
 859incrementing a counter.  That works to take a second reference if you
 860already have one (often indirectly through another object), but it
 861isn't sufficient if you don't actually have a counted reference at
 862all.  For ``dentry->d_lockref``, it is safe to increment the reference
 863counter to get a reference unless it has been explicitly marked as
 864"dead" which involves setting the counter to ``-128``.
 865``lockref_get_not_dead()`` achieves this.
 866
 867For ``mnt->mnt_count`` it is safe to take a reference as long as
 868``mount_lock`` is then used to validate the reference.  If that
 869validation fails, it may *not* be safe to just drop that reference in
 870the standard way of calling ``mnt_put()`` - an unmount may have
 871progressed too far.  So the code in ``legitimize_mnt()``, when it
 872finds that the reference it got might not be safe, checks the
 873``MNT_SYNC_UMOUNT`` flag to determine if a simple ``mnt_put()`` is
 874correct, or if it should just decrement the count and pretend none of
 875this ever happened.
 876
 877Taking care in filesystems
 878--------------------------
 879
 880RCU-walk depends almost entirely on cached information and often will
 881not call into the filesystem at all.  However there are two places,
 882besides the already-mentioned component-name comparison, where the
 883file system might be included in RCU-walk, and it must know to be
 884careful.
 885
 886If the filesystem has non-standard permission-checking requirements -
 887such as a networked filesystem which may need to check with the server
 888- the ``i_op->permission`` interface might be called during RCU-walk.
 889In this case an extra "``MAY_NOT_BLOCK``" flag is passed so that it
 890knows not to sleep, but to return ``-ECHILD`` if it cannot complete
 891promptly.  ``i_op->permission`` is given the inode pointer, not the
 892dentry, so it doesn't need to worry about further consistency checks.
 893However if it accesses any other filesystem data structures, it must
 894ensure they are safe to be accessed with only the ``rcu_read_lock()``
 895held.  This typically means they must be freed using ``kfree_rcu()`` or
 896similar.
 897
 898.. _READ_ONCE: https://lwn.net/Articles/624126/
 899
 900If the filesystem may need to revalidate dcache entries, then
 901``d_op->d_revalidate`` may be called in RCU-walk too.  This interface
 902*is* passed the dentry but does not have access to the ``inode`` or the
 903``seq`` number from the ``nameidata``, so it needs to be extra careful
 904when accessing fields in the dentry.  This "extra care" typically
 905involves using  `READ_ONCE() <READ_ONCE_>`_ to access fields, and verifying the
 906result is not NULL before using it.  This pattern can be seen in
 907``nfs_lookup_revalidate()``.
 908
 909A pair of patterns
 910------------------
 911
 912In various places in the details of REF-walk and RCU-walk, and also in
 913the big picture, there are a couple of related patterns that are worth
 914being aware of.
 915
 916The first is "try quickly and check, if that fails try slowly".  We
 917can see that in the high-level approach of first trying RCU-walk and
 918then trying REF-walk, and in places where ``unlazy_walk()`` is used to
 919switch to REF-walk for the rest of the path.  We also saw it earlier
 920in ``dget_parent()`` when following a "``..``" link.  It tries a quick way
 921to get a reference, then falls back to taking locks if needed.
 922
 923The second pattern is "try quickly and check, if that fails try
 924again - repeatedly".  This is seen with the use of ``rename_lock`` and
 925``mount_lock`` in REF-walk.  RCU-walk doesn't make use of this pattern -
 926if anything goes wrong it is much safer to just abort and try a more
 927sedate approach.
 928
 929The emphasis here is "try quickly and check".  It should probably be
 930"try quickly *and carefully*, then check".  The fact that checking is
 931needed is a reminder that the system is dynamic and only a limited
 932number of things are safe at all.  The most likely cause of errors in
 933this whole process is assuming something is safe when in reality it
 934isn't.  Careful consideration of what exactly guarantees the safety of
 935each access is sometimes necessary.
 936
 937A walk among the symlinks
 938=========================
 939
 940There are several basic issues that we will examine to understand the
 941handling of symbolic links:  the symlink stack, together with cache
 942lifetimes, will help us understand the overall recursive handling of
 943symlinks and lead to the special care needed for the final component.
 944Then a consideration of access-time updates and summary of the various
 945flags controlling lookup will finish the story.
 946
 947The symlink stack
 948-----------------
 949
 950There are only two sorts of filesystem objects that can usefully
 951appear in a path prior to the final component: directories and symlinks.
 952Handling directories is quite straightforward: the new directory
 953simply becomes the starting point at which to interpret the next
 954component on the path.  Handling symbolic links requires a bit more
 955work.
 956
 957Conceptually, symbolic links could be handled by editing the path.  If
 958a component name refers to a symbolic link, then that component is
 959replaced by the body of the link and, if that body starts with a '/',
 960then all preceding parts of the path are discarded.  This is what the
 961"``readlink -f``" command does, though it also edits out "``.``" and
 962"``..``" components.
 963
 964Directly editing the path string is not really necessary when looking
 965up a path, and discarding early components is pointless as they aren't
 966looked at anyway.  Keeping track of all remaining components is
 967important, but they can of course be kept separately; there is no need
 968to concatenate them.  As one symlink may easily refer to another,
 969which in turn can refer to a third, we may need to keep the remaining
 970components of several paths, each to be processed when the preceding
 971ones are completed.  These path remnants are kept on a stack of
 972limited size.
 973
 974There are two reasons for placing limits on how many symlinks can
 975occur in a single path lookup.  The most obvious is to avoid loops.
 976If a symlink referred to itself either directly or through
 977intermediaries, then following the symlink can never complete
 978successfully - the error ``ELOOP`` must be returned.  Loops can be
 979detected without imposing limits, but limits are the simplest solution
 980and, given the second reason for restriction, quite sufficient.
 981
 982.. _outlined recently: http://thread.gmane.org/gmane.linux.kernel/1934390/focus=1934550
 983
 984The second reason was `outlined recently`_ by Linus:
 985
 986   Because it's a latency and DoS issue too. We need to react well to
 987   true loops, but also to "very deep" non-loops. It's not about memory
 988   use, it's about users triggering unreasonable CPU resources.
 989
 990Linux imposes a limit on the length of any pathname: ``PATH_MAX``, which
 991is 4096.  There are a number of reasons for this limit; not letting the
 992kernel spend too much time on just one path is one of them.  With
 993symbolic links you can effectively generate much longer paths so some
 994sort of limit is needed for the same reason.  Linux imposes a limit of
 995at most 40 (MAXSYMLINKS) symlinks in any one path lookup.  It previously imposed
 996a further limit of eight on the maximum depth of recursion, but that was
 997raised to 40 when a separate stack was implemented, so there is now
 998just the one limit.
 999
1000The ``nameidata`` structure that we met in an earlier article contains a
1001small stack that can be used to store the remaining part of up to two
1002symlinks.  In many cases this will be sufficient.  If it isn't, a
1003separate stack is allocated with room for 40 symlinks.  Pathname
1004lookup will never exceed that stack as, once the 40th symlink is
1005detected, an error is returned.
1006
1007It might seem that the name remnants are all that needs to be stored on
1008this stack, but we need a bit more.  To see that, we need to move on to
1009cache lifetimes.
1010
1011Storage and lifetime of cached symlinks
1012---------------------------------------
1013
1014Like other filesystem resources, such as inodes and directory
1015entries, symlinks are cached by Linux to avoid repeated costly access
1016to external storage.  It is particularly important for RCU-walk to be
1017able to find and temporarily hold onto these cached entries, so that
1018it doesn't need to drop down into REF-walk.
1019
1020.. _object-oriented design pattern: https://lwn.net/Articles/446317/
1021
1022While each filesystem is free to make its own choice, symlinks are
1023typically stored in one of two places.  Short symlinks are often
1024stored directly in the inode.  When a filesystem allocates a ``struct
1025inode`` it typically allocates extra space to store private data (a
1026common `object-oriented design pattern`_ in the kernel).  This will
1027sometimes include space for a symlink.  The other common location is
1028in the page cache, which normally stores the content of files.  The
1029pathname in a symlink can be seen as the content of that symlink and
1030can easily be stored in the page cache just like file content.
1031
1032When neither of these is suitable, the next most likely scenario is
1033that the filesystem will allocate some temporary memory and copy or
1034construct the symlink content into that memory whenever it is needed.
1035
1036When the symlink is stored in the inode, it has the same lifetime as
1037the inode which, itself, is protected by RCU or by a counted reference
1038on the dentry.  This means that the mechanisms that pathname lookup
1039uses to access the dcache and icache (inode cache) safely are quite
1040sufficient for accessing some cached symlinks safely.  In these cases,
1041the ``i_link`` pointer in the inode is set to point to wherever the
1042symlink is stored and it can be accessed directly whenever needed.
1043
1044When the symlink is stored in the page cache or elsewhere, the
1045situation is not so straightforward.  A reference on a dentry or even
1046on an inode does not imply any reference on cached pages of that
1047inode, and even an ``rcu_read_lock()`` is not sufficient to ensure that
1048a page will not disappear.  So for these symlinks the pathname lookup
1049code needs to ask the filesystem to provide a stable reference and,
1050significantly, needs to release that reference when it is finished
1051with it.
1052
1053Taking a reference to a cache page is often possible even in RCU-walk
1054mode.  It does require making changes to memory, which is best avoided,
1055but that isn't necessarily a big cost and it is better than dropping
1056out of RCU-walk mode completely.  Even filesystems that allocate
1057space to copy the symlink into can use ``GFP_ATOMIC`` to often successfully
1058allocate memory without the need to drop out of RCU-walk.  If a
1059filesystem cannot successfully get a reference in RCU-walk mode, it
1060must return ``-ECHILD`` and ``unlazy_walk()`` will be called to return to
1061REF-walk mode in which the filesystem is allowed to sleep.
1062
1063The place for all this to happen is the ``i_op->get_link()`` inode
1064method. This is called both in RCU-walk and REF-walk. In RCU-walk the
1065``dentry*`` argument is NULL, ``->get_link()`` can return -ECHILD to drop out of
1066RCU-walk.  Much like the ``i_op->permission()`` method we
1067looked at previously, ``->get_link()`` would need to be careful that
1068all the data structures it references are safe to be accessed while
1069holding no counted reference, only the RCU lock. A callback
1070``struct delayed_called`` will be passed to ``->get_link()``:
1071file systems can set their own put_link function and argument through
1072set_delayed_call(). Later on, when VFS wants to put link, it will call
1073do_delayed_call() to invoke that callback function with the argument.
1074
1075In order for the reference to each symlink to be dropped when the walk completes,
1076whether in RCU-walk or REF-walk, the symlink stack needs to contain,
1077along with the path remnants:
1078
1079- the ``struct path`` to provide a reference to the previous path
1080- the ``const char *`` to provide a reference to the to previous name
1081- the ``seq`` to allow the path to be safely switched from RCU-walk to REF-walk
1082- the ``struct delayed_call`` for later invocation.
1083
1084This means that each entry in the symlink stack needs to hold five
1085pointers and an integer instead of just one pointer (the path
1086remnant).  On a 64-bit system, this is about 40 bytes per entry;
1087with 40 entries it adds up to 1600 bytes total, which is less than
1088half a page.  So it might seem like a lot, but is by no means
1089excessive.
1090
1091Note that, in a given stack frame, the path remnant (``name``) is not
1092part of the symlink that the other fields refer to.  It is the remnant
1093to be followed once that symlink has been fully parsed.
1094
1095Following the symlink
1096---------------------
1097
1098The main loop in ``link_path_walk()`` iterates seamlessly over all
1099components in the path and all of the non-final symlinks.  As symlinks
1100are processed, the ``name`` pointer is adjusted to point to a new
1101symlink, or is restored from the stack, so that much of the loop
1102doesn't need to notice.  Getting this ``name`` variable on and off the
1103stack is very straightforward; pushing and popping the references is
1104a little more complex.
1105
1106When a symlink is found, walk_component() calls pick_link() via step_into()
1107which returns the link from the filesystem.
1108Providing that operation is successful, the old path ``name`` is placed on the
1109stack, and the new value is used as the ``name`` for a while.  When the end of
1110the path is found (i.e. ``*name`` is ``'\0'``) the old ``name`` is restored
1111off the stack and path walking continues.
1112
1113Pushing and popping the reference pointers (inode, cookie, etc.) is more
1114complex in part because of the desire to handle tail recursion.  When
1115the last component of a symlink itself points to a symlink, we
1116want to pop the symlink-just-completed off the stack before pushing
1117the symlink-just-found to avoid leaving empty path remnants that would
1118just get in the way.
1119
1120It is most convenient to push the new symlink references onto the
1121stack in ``walk_component()`` immediately when the symlink is found;
1122``walk_component()`` is also the last piece of code that needs to look at the
1123old symlink as it walks that last component.  So it is quite
1124convenient for ``walk_component()`` to release the old symlink and pop
1125the references just before pushing the reference information for the
1126new symlink.  It is guided in this by three flags: ``WALK_NOFOLLOW`` which
1127forbids it from following a symlink if it finds one, ``WALK_MORE``
1128which indicates that it is yet too early to release the
1129current symlink, and ``WALK_TRAILING`` which indicates that it is on the final
1130component of the lookup, so we will check userspace flag ``LOOKUP_FOLLOW`` to
1131decide whether follow it when it is a symlink and call ``may_follow_link()`` to
1132check if we have privilege to follow it.
1133
1134Symlinks with no final component
1135~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1136
1137A pair of special-case symlinks deserve a little further explanation.
1138Both result in a new ``struct path`` (with mount and dentry) being set
1139up in the ``nameidata``, and result in pick_link() returning ``NULL``.
1140
1141The more obvious case is a symlink to "``/``".  All symlinks starting
1142with "``/``" are detected in pick_link() which resets the ``nameidata``
1143to point to the effective filesystem root.  If the symlink only
1144contains "``/``" then there is nothing more to do, no components at all,
1145so ``NULL`` is returned to indicate that the symlink can be released and
1146the stack frame discarded.
1147
1148The other case involves things in ``/proc`` that look like symlinks but
1149aren't really (and are therefore commonly referred to as "magic-links")::
1150
1151     $ ls -l /proc/self/fd/1
1152     lrwx------ 1 neilb neilb 64 Jun 13 10:19 /proc/self/fd/1 -> /dev/pts/4
1153
1154Every open file descriptor in any process is represented in ``/proc`` by
1155something that looks like a symlink.  It is really a reference to the
1156target file, not just the name of it.  When you ``readlink`` these
1157objects you get a name that might refer to the same file - unless it
1158has been unlinked or mounted over.  When ``walk_component()`` follows
1159one of these, the ``->get_link()`` method in "procfs" doesn't return
1160a string name, but instead calls nd_jump_link() which updates the
1161``nameidata`` in place to point to that target.  ``->get_link()`` then
1162returns ``NULL``.  Again there is no final component and pick_link()
1163returns ``NULL``.
1164
1165Following the symlink in the final component
1166--------------------------------------------
1167
1168All this leads to ``link_path_walk()`` walking down every component, and
1169following all symbolic links it finds, until it reaches the final
1170component.  This is just returned in the ``last`` field of ``nameidata``.
1171For some callers, this is all they need; they want to create that
1172``last`` name if it doesn't exist or give an error if it does.  Other
1173callers will want to follow a symlink if one is found, and possibly
1174apply special handling to the last component of that symlink, rather
1175than just the last component of the original file name.  These callers
1176potentially need to call ``link_path_walk()`` again and again on
1177successive symlinks until one is found that doesn't point to another
1178symlink.
1179
1180This case is handled by relevant callers of link_path_walk(), such as
1181path_lookupat(), path_openat() using a loop that calls link_path_walk(),
1182and then handles the final component by calling open_last_lookups() or
1183lookup_last(). If it is a symlink that needs to be followed,
1184open_last_lookups() or lookup_last() will set things up properly and
1185return the path so that the loop repeats, calling
1186link_path_walk() again.  This could loop as many as 40 times if the last
1187component of each symlink is another symlink.
1188
1189Of the various functions that examine the final component, 
1190open_last_lookups() is the most interesting as it works in tandem
1191with do_open() for opening a file.  Part of open_last_lookups() runs
1192with ``i_rwsem`` held and this part is in a separate function: lookup_open().
1193
1194Explaining open_last_lookups() and do_open() completely is beyond the scope
1195of this article, but a few highlights should help those interested in exploring
1196the code.
1197
11981. Rather than just finding the target file, do_open() is used after
1199   open_last_lookup() to open
1200   it.  If the file was found in the dcache, then ``vfs_open()`` is used for
1201   this.  If not, then ``lookup_open()`` will either call ``atomic_open()`` (if
1202   the filesystem provides it) to combine the final lookup with the open, or
1203   will perform the separate ``i_op->lookup()`` and ``i_op->create()`` steps
1204   directly.  In the later case the actual "open" of this newly found or
1205   created file will be performed by vfs_open(), just as if the name
1206   were found in the dcache.
1207
12082. vfs_open() can fail with ``-EOPENSTALE`` if the cached information
1209   wasn't quite current enough.  If it's in RCU-walk ``-ECHILD`` will be returned
1210   otherwise ``-ESTALE`` is returned.  When ``-ESTALE`` is returned, the caller may
1211   retry with ``LOOKUP_REVAL`` flag set.
1212
12133. An open with O_CREAT **does** follow a symlink in the final component,
1214   unlike other creation system calls (like ``mkdir``).  So the sequence::
1215
1216          ln -s bar /tmp/foo
1217          echo hello > /tmp/foo
1218
1219   will create a file called ``/tmp/bar``.  This is not permitted if
1220   ``O_EXCL`` is set but otherwise is handled for an O_CREAT open much
1221   like for a non-creating open: lookup_last() or open_last_lookup()
1222   returns a non ``NULL`` value, and link_path_walk() gets called and the
1223   open process continues on the symlink that was found.
1224
1225Updating the access time
1226------------------------
1227
1228We previously said of RCU-walk that it would "take no locks, increment
1229no counts, leave no footprints."  We have since seen that some
1230"footprints" can be needed when handling symlinks as a counted
1231reference (or even a memory allocation) may be needed.  But these
1232footprints are best kept to a minimum.
1233
1234One other place where walking down a symlink can involve leaving
1235footprints in a way that doesn't affect directories is in updating access times.
1236In Unix (and Linux) every filesystem object has a "last accessed
1237time", or "``atime``".  Passing through a directory to access a file
1238within is not considered to be an access for the purposes of
1239``atime``; only listing the contents of a directory can update its ``atime``.
1240Symlinks are different it seems.  Both reading a symlink (with ``readlink()``)
1241and looking up a symlink on the way to some other destination can
1242update the atime on that symlink.
1243
1244.. _clearest statement: https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_08
1245
1246It is not clear why this is the case; POSIX has little to say on the
1247subject.  The `clearest statement`_ is that, if a particular implementation
1248updates a timestamp in a place not specified by POSIX, this must be
1249documented "except that any changes caused by pathname resolution need
1250not be documented".  This seems to imply that POSIX doesn't really
1251care about access-time updates during pathname lookup.
1252
1253.. _Linux 1.3.87: https://git.kernel.org/cgit/linux/kernel/git/history/history.git/diff/fs/ext2/symlink.c?id=f806c6db77b8eaa6e00dcfb6b567706feae8dbb8
1254
1255An examination of history shows that prior to `Linux 1.3.87`_, the ext2
1256filesystem, at least, didn't update atime when following a link.
1257Unfortunately we have no record of why that behavior was changed.
1258
1259In any case, access time must now be updated and that operation can be
1260quite complex.  Trying to stay in RCU-walk while doing it is best
1261avoided.  Fortunately it is often permitted to skip the ``atime``
1262update.  Because ``atime`` updates cause performance problems in various
1263areas, Linux supports the ``relatime`` mount option, which generally
1264limits the updates of ``atime`` to once per day on files that aren't
1265being changed (and symlinks never change once created).  Even without
1266``relatime``, many filesystems record ``atime`` with a one-second
1267granularity, so only one update per second is required.
1268
1269It is easy to test if an ``atime`` update is needed while in RCU-walk
1270mode and, if it isn't, the update can be skipped and RCU-walk mode
1271continues.  Only when an ``atime`` update is actually required does the
1272path walk drop down to REF-walk.  All of this is handled in the
1273``get_link()`` function.
1274
1275A few flags
1276-----------
1277
1278A suitable way to wrap up this tour of pathname walking is to list
1279the various flags that can be stored in the ``nameidata`` to guide the
1280lookup process.  Many of these are only meaningful on the final
1281component, others reflect the current state of the pathname lookup, and some
1282apply restrictions to all path components encountered in the path lookup.
1283
1284And then there is ``LOOKUP_EMPTY``, which doesn't fit conceptually with
1285the others.  If this is not set, an empty pathname causes an error
1286very early on.  If it is set, empty pathnames are not considered to be
1287an error.
1288
1289Global state flags
1290~~~~~~~~~~~~~~~~~~
1291
1292We have already met two global state flags: ``LOOKUP_RCU`` and
1293``LOOKUP_REVAL``.  These select between one of three overall approaches
1294to lookup: RCU-walk, REF-walk, and REF-walk with forced revalidation.
1295
1296``LOOKUP_PARENT`` indicates that the final component hasn't been reached
1297yet.  This is primarily used to tell the audit subsystem the full
1298context of a particular access being audited.
1299
1300``ND_ROOT_PRESET`` indicates that the ``root`` field in the ``nameidata`` was
1301provided by the caller, so it shouldn't be released when it is no
1302longer needed.
1303
1304``ND_JUMPED`` means that the current dentry was chosen not because
1305it had the right name but for some other reason.  This happens when
1306following "``..``", following a symlink to ``/``, crossing a mount point
1307or accessing a "``/proc/$PID/fd/$FD``" symlink (also known as a "magic
1308link"). In this case the filesystem has not been asked to revalidate the
1309name (with ``d_revalidate()``).  In such cases the inode may still need
1310to be revalidated, so ``d_op->d_weak_revalidate()`` is called if
1311``ND_JUMPED`` is set when the look completes - which may be at the
1312final component or, when creating, unlinking, or renaming, at the penultimate component.
1313
1314Resolution-restriction flags
1315~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1316
1317In order to allow userspace to protect itself against certain race conditions
1318and attack scenarios involving changing path components, a series of flags are
1319available which apply restrictions to all path components encountered during
1320path lookup. These flags are exposed through ``openat2()``'s ``resolve`` field.
1321
1322``LOOKUP_NO_SYMLINKS`` blocks all symlink traversals (including magic-links).
1323This is distinctly different from ``LOOKUP_FOLLOW``, because the latter only
1324relates to restricting the following of trailing symlinks.
1325
1326``LOOKUP_NO_MAGICLINKS`` blocks all magic-link traversals. Filesystems must
1327ensure that they return errors from ``nd_jump_link()``, because that is how
1328``LOOKUP_NO_MAGICLINKS`` and other magic-link restrictions are implemented.
1329
1330``LOOKUP_NO_XDEV`` blocks all ``vfsmount`` traversals (this includes both
1331bind-mounts and ordinary mounts). Note that the ``vfsmount`` which contains the
1332lookup is determined by the first mountpoint the path lookup reaches --
1333absolute paths start with the ``vfsmount`` of ``/``, and relative paths start
1334with the ``dfd``'s ``vfsmount``. Magic-links are only permitted if the
1335``vfsmount`` of the path is unchanged.
1336
1337``LOOKUP_BENEATH`` blocks any path components which resolve outside the
1338starting point of the resolution. This is done by blocking ``nd_jump_root()``
1339as well as blocking ".." if it would jump outside the starting point.
1340``rename_lock`` and ``mount_lock`` are used to detect attacks against the
1341resolution of "..". Magic-links are also blocked.
1342
1343``LOOKUP_IN_ROOT`` resolves all path components as though the starting point
1344were the filesystem root. ``nd_jump_root()`` brings the resolution back to
1345the starting point, and ".." at the starting point will act as a no-op. As with
1346``LOOKUP_BENEATH``, ``rename_lock`` and ``mount_lock`` are used to detect
1347attacks against ".." resolution. Magic-links are also blocked.
1348
1349Final-component flags
1350~~~~~~~~~~~~~~~~~~~~~
1351
1352Some of these flags are only set when the final component is being
1353considered.  Others are only checked for when considering that final
1354component.
1355
1356``LOOKUP_AUTOMOUNT`` ensures that, if the final component is an automount
1357point, then the mount is triggered.  Some operations would trigger it
1358anyway, but operations like ``stat()`` deliberately don't.  ``statfs()``
1359needs to trigger the mount but otherwise behaves a lot like ``stat()``, so
1360it sets ``LOOKUP_AUTOMOUNT``, as does "``quotactl()``" and the handling of
1361"``mount --bind``".
1362
1363``LOOKUP_FOLLOW`` has a similar function to ``LOOKUP_AUTOMOUNT`` but for
1364symlinks.  Some system calls set or clear it implicitly, while
1365others have API flags such as ``AT_SYMLINK_FOLLOW`` and
1366``UMOUNT_NOFOLLOW`` to control it.  Its effect is similar to
1367``WALK_GET`` that we already met, but it is used in a different way.
1368
1369``LOOKUP_DIRECTORY`` insists that the final component is a directory.
1370Various callers set this and it is also set when the final component
1371is found to be followed by a slash.
1372
1373Finally ``LOOKUP_OPEN``, ``LOOKUP_CREATE``, ``LOOKUP_EXCL``, and
1374``LOOKUP_RENAME_TARGET`` are not used directly by the VFS but are made
1375available to the filesystem and particularly the ``->d_revalidate()``
1376method.  A filesystem can choose not to bother revalidating too hard
1377if it knows that it will be asked to open or create the file soon.
1378These flags were previously useful for ``->lookup()`` too but with the
1379introduction of ``->atomic_open()`` they are less relevant there.
1380
1381End of the road
1382---------------
1383
1384Despite its complexity, all this pathname lookup code appears to be
1385in good shape - various parts are certainly easier to understand now
1386than even a couple of releases ago.  But that doesn't mean it is
1387"finished".   As already mentioned, RCU-walk currently only follows
1388symlinks that are stored in the inode so, while it handles many ext4
1389symlinks, it doesn't help with NFS, XFS, or Btrfs.  That support
1390is not likely to be long delayed.