Using RCU lifetime rules to access kernfs_node::name can avoid the
trouble with kernfs_rename_lock in kernfs_name() and kernfs_path_from_node()
if the fs was created with KERNFS_ROOT_INVARIANT_PARENT. This is usefull
as it allows to implement kernfs_path_from_node() only with RCU
protection and avoiding kernfs_rename_lock. The lock is only required if
the __parent node can be changed and the function requires an unchanged
hierarchy while it iterates from the node to its parent.
The change is needed to allow the lookup of the node's path
(kernfs_path_from_node()) from context which runs always with disabled
preemption and or interrutps even on PREEMPT_RT. The problem is that
kernfs_rename_lock becomes a sleeping lock on PREEMPT_RT.
I went through all ::name users and added the required access for the lookup
with a few extensions:
- rdtgroup_pseudo_lock_create() drops all locks and then uses the name
later on. resctrl supports rename with different parents. Here I made
a temporal copy of the name while it is used outside of the lock.
- kernfs_rename_ns() accepts NULL as new_parent. This simplifies
sysfs_move_dir_ns() where it can set NULL in order to reuse the current
name.
- kernfs_rename_ns() is only using kernfs_rename_lock if the parents are
different. All users use either kernfs_rwsem (for stable path view) or
just RCU for the lookup. The ::name uses always RCU free.
Use RCU lifetime guarantees to access kernfs_node::name.
Suggested-by: Tejun Heo <tj@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reported-by: syzbot+6ea37e2e6ffccf41a7e6@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/lkml/67251dc6.050a0220.529b6.015e.GAE@google.com/
Reported-by: Hillf Danton <hdanton@sina.com>
Closes: https://lore.kernel.org/20241102001224.2789-1-hdanton@sina.com
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20250213145023.2820193-7-bigeasy@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernfs_rename_lock is used to obtain stable kernfs_node::{name|parent}
pointer. This is a preparation to access kernfs_node::parent under RCU
and ensure that the pointer remains stable under the RCU lifetime
guarantees.
For a complete path, as it is done in kernfs_path_from_node(), the
kernfs_rename_lock is still required in order to obtain a stable parent
relationship while computing the relevant node depth. This must not
change while the nodes are inspected in order to build the path.
If the kernfs user never moves the nodes (changes the parent) then the
kernfs_rename_lock is not required and the RCU guarantees are
sufficient. This "restriction" can be set with
KERNFS_ROOT_INVARIANT_PARENT. Otherwise the lock is required.
Rename kernfs_node::parent to kernfs_node::__parent to denote the RCU
access and use RCU accessor while accessing the node.
Make cgroup use KERNFS_ROOT_INVARIANT_PARENT since the parent here can
not change.
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Yonghong Song <yonghong.song@linux.dev>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20250213145023.2820193-6-bigeasy@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The BPF helper bpf_cgroup_from_id() calls kernfs_find_and_get_node_by_id()
which acquires kernfs_idr_lock, which is an non-raw non-IRQ-safe lock. This
can lead to deadlocks as bpf_cgroup_from_id() can be called from any BPF
programs including e.g. the ones that attach to functions which are holding
the scheduler rq lock.
Consider the following BPF program:
SEC("fentry/__set_cpus_allowed_ptr_locked")
int BPF_PROG(__set_cpus_allowed_ptr_locked, struct task_struct *p,
struct affinity_context *affn_ctx, struct rq *rq, struct rq_flags *rf)
{
struct cgroup *cgrp = bpf_cgroup_from_id(p->cgroups->dfl_cgrp->kn->id);
if (cgrp) {
bpf_printk("%d[%s] in %s", p->pid, p->comm, cgrp->kn->name);
bpf_cgroup_release(cgrp);
}
return 0;
}
__set_cpus_allowed_ptr_locked() is called with rq lock held and the above
BPF program calls bpf_cgroup_from_id() within leading to the following
lockdep warning:
=====================================================
WARNING: HARDIRQ-safe -> HARDIRQ-unsafe lock order detected
6.7.0-rc3-work-00053-g07124366a1d7-dirty #147 Not tainted
-----------------------------------------------------
repro/1620 [HC0[0]:SC0[0]:HE0:SE1] is trying to acquire:
ffffffff833b3688 (kernfs_idr_lock){+.+.}-{2:2}, at: kernfs_find_and_get_node_by_id+0x1e/0x70
and this task is already holding:
ffff888237ced698 (&rq->__lock){-.-.}-{2:2}, at: task_rq_lock+0x4e/0xf0
which would create a new lock dependency:
(&rq->__lock){-.-.}-{2:2} -> (kernfs_idr_lock){+.+.}-{2:2}
...
Possible interrupt unsafe locking scenario:
CPU0 CPU1
---- ----
lock(kernfs_idr_lock);
local_irq_disable();
lock(&rq->__lock);
lock(kernfs_idr_lock);
<Interrupt>
lock(&rq->__lock);
*** DEADLOCK ***
...
Call Trace:
dump_stack_lvl+0x55/0x70
dump_stack+0x10/0x20
__lock_acquire+0x781/0x2a40
lock_acquire+0xbf/0x1f0
_raw_spin_lock+0x2f/0x40
kernfs_find_and_get_node_by_id+0x1e/0x70
cgroup_get_from_id+0x21/0x240
bpf_cgroup_from_id+0xe/0x20
bpf_prog_98652316e9337a5a___set_cpus_allowed_ptr_locked+0x96/0x11a
bpf_trampoline_6442545632+0x4f/0x1000
__set_cpus_allowed_ptr_locked+0x5/0x5a0
sched_setaffinity+0x1b3/0x290
__x64_sys_sched_setaffinity+0x4f/0x60
do_syscall_64+0x40/0xe0
entry_SYSCALL_64_after_hwframe+0x46/0x4e
Let's fix it by protecting kernfs_node and kernfs_root with RCU and making
kernfs_find_and_get_node_by_id() acquire rcu_read_lock() instead of
kernfs_idr_lock.
This adds an rcu_head to kernfs_node making it larger by 16 bytes on 64bit.
Combined with the preceding rearrange patch, the net increase is 8 bytes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Andrea Righi <andrea.righi@canonical.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/r/20240109214828.252092-4-tj@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Moving .flags and .mode right below .hash makes kernfs_node smaller by 8
bytes on 64bit. To avoid creating a hole from 8 bytes alignment on 32bit
archs, .priv is moved below so that there are two 32bit pointers after the
64bit .id field.
v2: Updated to avoid size increase on 32bit noticed by Geert.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/r/ZZ7hwA18nfmFjYpj@slm.duckdns.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
As of now, seeking in sysfs files is handled by generic_file_llseek().
There are situations where one may want to customize seeking logic:
- Many sysfs entries are fixed files while generic_file_llseek() accepts
past-the-end positions. Not only being useless by itself, this
also means a bug in userspace code will trigger not at lseek(), but at
some later point making debugging harder.
- generic_file_llseek() relies on f_mapping->host to get the file size
which might not be correct for all sysfs entries.
See commit 636b21b501 ("PCI: Revoke mappings like devmem") as an example.
Implement llseek method to override this behavior at sysfs attribute
level. The method is optional, and if it is absent,
generic_file_llseek() is called to preserve backwards compatibility.
Signed-off-by: Valentine Sinitsyn <valesini@yandex-team.ru>
Link: https://lore.kernel.org/r/20230925084013.309399-1-valesini@yandex-team.ru
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In some randconfig builds, kernfs ends up being disabled, so there is no prototype
for kernfs_generic_poll()
In file included from kernel/sched/build_utility.c:97:
kernel/sched/psi.c:1479:3: error: implicit declaration of function 'kernfs_generic_poll' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
kernfs_generic_poll(t->of, wait);
^
Add a stub helper for it, as we have it for other kernfs functions.
Fixes: aff037078e ("sched/psi: use kernfs polling functions for PSI trigger polling")
Fixes: 147e1a97c4 ("fs: kernfs: add poll file operation")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Chengming Zhou <zhouchengming@bytedance.com>
Link: https://lore.kernel.org/r/20230724121823.1357562-1-arnd@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, kernfs nodes can be created hidden and activated later by calling
kernfs_activate() to allow creation of multiple nodes to succeed or fail as
a unit. This is an one-way one-time-only transition. This patch introduces
kernfs_show() which can toggle visibility dynamically.
As the currently proposed use - toggling the cgroup pressure files - only
requires operating on leaf nodes, for the sake of simplicity, restrict it as
such for now.
Hiding uses the same mechanism as deactivation and likewise guarantees that
there are no in-flight operations on completion. KERNFS_ACTIVATED and
KERNFS_HIDDEN are used to manage the interactions between activations and
show/hide operations. A node is visible iff both activated & !hidden.
Cc: Chengming Zhou <zhouchengming@bytedance.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220828050440.734579-9-tj@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
KERNFS_ACTIVATED tracks whether a given node has ever been activated. As a
node was only deactivated on removal, this was used for
1. Drain optimization (removed by the previous patch).
2. To hide !activated nodes
3. To avoid double activations
4. Reject adding children to a node being removed
5. Skip activaing a node which is being removed.
We want to decouple deactivation from removal so that nodes can be
deactivated and hidden dynamically, which makes KERNFS_ACTIVATED useless for
all of the above purposes.
#1 is already gone. #2 and #3 can instead test whether the node is currently
active. A new flag KERNFS_REMOVING is added to explicitly mark nodes which
are being removed for #4 and #5.
While this leaves KERNFS_ACTIVATED with no users, leave it be as it will be
used in a following patch.
Cc: Chengming Zhou <zhouchengming@bytedance.com>
Tested-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220828050440.734579-7-tj@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In current kernfs design a single mutex, kernfs_open_file_mutex, protects
the list of kernfs_open_file instances corresponding to a sysfs attribute.
So even if different tasks are opening or closing different sysfs files
they can contend on osq_lock of this mutex. The contention is more apparent
in large scale systems with few hundred CPUs where most of the CPUs have
running tasks that are opening, accessing or closing sysfs files at any
point of time.
Using hashed mutexes in place of a single global mutex, can significantly
reduce contention around global mutex and hence can provide better
scalability. Moreover as these hashed mutexes are not part of kernfs_node
objects we will not see any singnificant change in memory utilization of
kernfs based file systems like sysfs, cgroupfs etc.
Modify interface introduced in previous patch to make use of hashed
mutexes. Use kernfs_node address as hashing key.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Imran Khan <imran.f.khan@oracle.com>
Link: https://lore.kernel.org/r/20220615021059.862643-5-imran.f.khan@oracle.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The kernfs implementation has big lock granularity(kernfs_rwsem) so
every kernfs-based(e.g., sysfs, cgroup) fs are able to compete the
lock. It makes trouble for some cases to wait the global lock
for a long time even though they are totally independent contexts
each other.
A general example is process A goes under direct reclaim with holding
the lock when it accessed the file in sysfs and process B is waiting
the lock with exclusive mode and then process C is waiting the lock
until process B could finish the job after it gets the lock from
process A.
This patch switches the global kernfs_rwsem to per-fs lock, which
put the rwsem into kernfs_root.
Suggested-by: Tejun Heo <tj@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Link: https://lore.kernel.org/r/20211118230008.2679780-1-minchan@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add a revision counter to kernfs directory nodes so it can be used
to detect if a directory node has changed during negative dentry
revalidation.
There's an assumption that sizeof(unsigned long) <= sizeof(pointer)
on all architectures and as far as I know that assumption holds.
So adding a revision counter to the struct kernfs_elem_dir variant of
the kernfs_node type union won't increase the size of the kernfs_node
struct. This is because struct kernfs_elem_dir is at least
sizeof(pointer) smaller than the largest union variant. It's tempting
to make the revision counter a u64 but that would increase the size of
kernfs_node on archs where sizeof(pointer) is smaller than the revision
counter.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Ian Kent <raven@themaw.net>
Link: https://lore.kernel.org/r/162642769895.63632.8356662784964509867.stgit@web.messagingengine.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
User extended attributes are useful as metadata storage for kernfs
consumers like cgroups. Especially in the case of cgroups, it is useful
to have a central metadata store that multiple processes/services can
use to coordinate actions.
A concrete example is for userspace out of memory killers. We want to
let delegated cgroup subtree owners (running as non-root) to be able to
say "please avoid killing this cgroup". This is especially important for
desktop linux as delegated subtrees owners are less likely to run as
root.
This patch introduces a new flag, KERNFS_ROOT_SUPPORT_USER_XATTR, that
lets kernfs consumers enable user xattr support. An initial limit of 128
entries or 128KB -- whichever is hit first -- is placed per cgroup
because xattrs come from kernel memory and we don't want to let
unprivileged users accidentally eat up too much kernel memory.
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Acked-by: Chris Down <chris@chrisdown.name>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Each kernfs_node is identified with a 64bit ID. The low 32bit is
exposed as ino and the high gen. While this already allows using inos
as keys by looking up with wildcard generation number of 0, it's
adding unnecessary complications for 64bit ino archs which can
directly use kernfs_node IDs as inos to uniquely identify each cgroup
instance.
This patch exposes IDs directly as inos on 64bit ino archs. The
conversion is mostly straight-forward.
* 32bit ino archs behave the same as before. 64bit ino archs now use
the whole 64bit ID as ino and the generation number is fixed at 1.
* 64bit inos still use the same idr allocator which gurantees that the
lower 32bits identify the current live instance uniquely and the
high 32bits are incremented whenever the low bits wrap. As the
upper 32bits are no longer used as gen and we don't wanna start ino
allocation with 33rd bit set, the initial value for highbits
allocation is changed to 0 on 64bit ino archs.
* blktrace exposes two 32bit numbers - (INO,GEN) pair - to identify
the issuing cgroup. Userland builds FILEID_INO32_GEN fids from
these numbers to look up the cgroups. To remain compatible with the
behavior, always output (LOW32,HIGH32) which will be constructed
back to the original 64bit ID by __kernfs_fh_to_dentry().
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
kernfs_find_and_get_node_by_ino() looks the kernfs_node matching the
specified ino. On top of that, kernfs_get_node_by_id() and
kernfs_fh_get_inode() implement full ID matching by testing the rest
of ID.
On surface, confusingly, the two are slightly different in that the
latter uses 0 gen as wildcard while the former doesn't - does it mean
that the latter can't uniquely identify inodes w/ 0 gen? In practice,
this is a distinction without a difference because generation number
starts at 1. There are no actual IDs with 0 gen, so it can always
safely used as wildcard.
Let's simplify the code by renaming kernfs_find_and_get_node_by_ino()
to kernfs_find_and_get_node_by_id(), moving all lookup logics into it,
and removing now unnecessary kernfs_get_node_by_id().
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernfs_node->id is currently a union kernfs_node_id which represents
either a 32bit (ino, gen) pair or u64 value. I can't see much value
in the usage of the union - all that's needed is a 64bit ID which the
current code is already limited to. Using a union makes the code
unnecessarily complicated and prevents using 64bit ino without adding
practical benefits.
This patch drops union kernfs_node_id and makes kernfs_node->id a u64.
ino is stored in the lower 32bits and gen upper. Accessors -
kernfs[_id]_ino() and kernfs[_id]_gen() - are added to retrieve the
ino and gen. This simplifies ID handling less cumbersome and will
allow using 64bit inos on supported archs.
This patch doesn't make any functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Alexei Starovoitov <ast@kernel.org>
Sebastian Andrzej Siewior
Tejun Heo
Valentine Sinitsyn
Arnd Bergmann
Imran Khan
Greg Kroah-Hartman
Andy Shevchenko
Minchan Kim
Christoph Hellwig
Ian Kent
Willem de Bruijn
Daniel Xu