This is a note to let you know that I've just added the patch titled
Revert "kernfs: convert kernfs_idr_lock to an irq safe raw spinlock"
to the 6.7-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary
The filename of the patch is:
revert-kernfs-convert-kernfs_idr_lock-to-an-irq-safe-raw-spinlock.patch
and it can be found in the queue-6.7 subdirectory.
If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@xxxxxxxxxxxxxxx> know about it.
>From e3977e0609a07d86406029fceea0fd40d7849368 Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@xxxxxxxxxx>
Date: Tue, 9 Jan 2024 11:48:02 -1000
Subject: Revert "kernfs: convert kernfs_idr_lock to an irq safe raw spinlock"
From: Tejun Heo <tj@xxxxxxxxxx>
commit e3977e0609a07d86406029fceea0fd40d7849368 upstream.
This reverts commit dad3fb67ca1cbef87ce700e83a55835e5921ce8a.
The commit converted kernfs_idr_lock to an IRQ-safe raw_spinlock because it
could be acquired while holding an rq lock through bpf_cgroup_from_id().
However, kernfs_idr_lock is held while doing GPF_NOWAIT allocations which
involves acquiring an non-IRQ-safe and non-raw lock leading to the following
lockdep warning:
=============================
[ BUG: Invalid wait context ]
6.7.0-rc5-kzm9g-00251-g655022a45b1c #578 Not tainted
-----------------------------
swapper/0/0 is trying to lock:
dfbcd488 (&c->lock){....}-{3:3}, at: local_lock_acquire+0x0/0xa4
other info that might help us debug this:
context-{5:5}
2 locks held by swapper/0/0:
#0: dfbc9c60 (lock){+.+.}-{3:3}, at: local_lock_acquire+0x0/0xa4
#1: c0c012a8 (kernfs_idr_lock){....}-{2:2}, at: __kernfs_new_node.constprop.0+0x68/0x258
stack backtrace:
CPU: 0 PID: 0 Comm: swapper/0 Not tainted 6.7.0-rc5-kzm9g-00251-g655022a45b1c #578
Hardware name: Generic SH73A0 (Flattened Device Tree)
unwind_backtrace from show_stack+0x10/0x14
show_stack from dump_stack_lvl+0x68/0x90
dump_stack_lvl from __lock_acquire+0x3cc/0x168c
__lock_acquire from lock_acquire+0x274/0x30c
lock_acquire from local_lock_acquire+0x28/0xa4
local_lock_acquire from ___slab_alloc+0x234/0x8a8
___slab_alloc from __slab_alloc.constprop.0+0x30/0x44
__slab_alloc.constprop.0 from kmem_cache_alloc+0x7c/0x148
kmem_cache_alloc from radix_tree_node_alloc.constprop.0+0x44/0xdc
radix_tree_node_alloc.constprop.0 from idr_get_free+0x110/0x2b8
idr_get_free from idr_alloc_u32+0x9c/0x108
idr_alloc_u32 from idr_alloc_cyclic+0x50/0xb8
idr_alloc_cyclic from __kernfs_new_node.constprop.0+0x88/0x258
__kernfs_new_node.constprop.0 from kernfs_create_root+0xbc/0x154
kernfs_create_root from sysfs_init+0x18/0x5c
sysfs_init from mnt_init+0xc4/0x220
mnt_init from vfs_caches_init+0x6c/0x88
vfs_caches_init from start_kernel+0x474/0x528
start_kernel from 0x0
Let's rever the commit. It's undesirable to spread out raw spinlock usage
anyway and the problem can be solved by protecting the lookup path with RCU
instead.
Signed-off-by: Tejun Heo <tj@xxxxxxxxxx>
Cc: Andrea Righi <andrea.righi@xxxxxxxxxxxxx>
Reported-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
Link: http://lkml.kernel.org/r/CAMuHMdV=AKt+mwY7svEq5gFPx41LoSQZ_USME5_MEdWQze13ww@xxxxxxxxxxxxxx
Link: https://lore.kernel.org/r/20240109214828.252092-2-tj@xxxxxxxxxx
Tested-by: Andrea Righi <andrea.righi@xxxxxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
---
fs/kernfs/dir.c | 23 ++++++++++-------------
1 file changed, 10 insertions(+), 13 deletions(-)
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -27,7 +27,7 @@ static DEFINE_RWLOCK(kernfs_rename_lock)
*/
static DEFINE_SPINLOCK(kernfs_pr_cont_lock);
static char kernfs_pr_cont_buf[PATH_MAX]; /* protected by pr_cont_lock */
-static DEFINE_RAW_SPINLOCK(kernfs_idr_lock); /* root->ino_idr */
+static DEFINE_SPINLOCK(kernfs_idr_lock); /* root->ino_idr */
#define rb_to_kn(X) rb_entry((X), struct kernfs_node, rb)
@@ -539,7 +539,6 @@ void kernfs_put(struct kernfs_node *kn)
{
struct kernfs_node *parent;
struct kernfs_root *root;
- unsigned long flags;
if (!kn || !atomic_dec_and_test(&kn->count))
return;
@@ -564,9 +563,9 @@ void kernfs_put(struct kernfs_node *kn)
simple_xattrs_free(&kn->iattr->xattrs, NULL);
kmem_cache_free(kernfs_iattrs_cache, kn->iattr);
}
- raw_spin_lock_irqsave(&kernfs_idr_lock, flags);
+ spin_lock(&kernfs_idr_lock);
idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
- raw_spin_unlock_irqrestore(&kernfs_idr_lock, flags);
+ spin_unlock(&kernfs_idr_lock);
kmem_cache_free(kernfs_node_cache, kn);
kn = parent;
@@ -608,7 +607,6 @@ static struct kernfs_node *__kernfs_new_
struct kernfs_node *kn;
u32 id_highbits;
int ret;
- unsigned long irqflags;
name = kstrdup_const(name, GFP_KERNEL);
if (!name)
@@ -619,13 +617,13 @@ static struct kernfs_node *__kernfs_new_
goto err_out1;
idr_preload(GFP_KERNEL);
- raw_spin_lock_irqsave(&kernfs_idr_lock, irqflags);
+ spin_lock(&kernfs_idr_lock);
ret = idr_alloc_cyclic(&root->ino_idr, kn, 1, 0, GFP_ATOMIC);
if (ret >= 0 && ret < root->last_id_lowbits)
root->id_highbits++;
id_highbits = root->id_highbits;
root->last_id_lowbits = ret;
- raw_spin_unlock_irqrestore(&kernfs_idr_lock, irqflags);
+ spin_unlock(&kernfs_idr_lock);
idr_preload_end();
if (ret < 0)
goto err_out2;
@@ -661,9 +659,9 @@ static struct kernfs_node *__kernfs_new_
return kn;
err_out3:
- raw_spin_lock_irqsave(&kernfs_idr_lock, irqflags);
+ spin_lock(&kernfs_idr_lock);
idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
- raw_spin_unlock_irqrestore(&kernfs_idr_lock, irqflags);
+ spin_unlock(&kernfs_idr_lock);
err_out2:
kmem_cache_free(kernfs_node_cache, kn);
err_out1:
@@ -704,9 +702,8 @@ struct kernfs_node *kernfs_find_and_get_
struct kernfs_node *kn;
ino_t ino = kernfs_id_ino(id);
u32 gen = kernfs_id_gen(id);
- unsigned long flags;
- raw_spin_lock_irqsave(&kernfs_idr_lock, flags);
+ spin_lock(&kernfs_idr_lock);
kn = idr_find(&root->ino_idr, (u32)ino);
if (!kn)
@@ -730,10 +727,10 @@ struct kernfs_node *kernfs_find_and_get_
if (unlikely(!__kernfs_active(kn) || !atomic_inc_not_zero(&kn->count)))
goto err_unlock;
- raw_spin_unlock_irqrestore(&kernfs_idr_lock, flags);
+ spin_unlock(&kernfs_idr_lock);
return kn;
err_unlock:
- raw_spin_unlock_irqrestore(&kernfs_idr_lock, flags);
+ spin_unlock(&kernfs_idr_lock);
return NULL;
}
Patches currently in stable-queue which might be from tj@xxxxxxxxxx are
queue-6.7/blk-cgroup-fix-rcu-lockdep-warning-in-blkg_lookup.patch
queue-6.7/revert-kernfs-convert-kernfs_idr_lock-to-an-irq-safe-raw-spinlock.patch
queue-6.7/kernfs-convert-kernfs_idr_lock-to-an-irq-safe-raw-spinlock.patch
