Re: [RFC V2 PATCH] kernfs: create raw version kernfs_path_len and kernfs_path

From: Shi, Yang
Date: Fri Feb 26 2016 - 18:05:21 EST


On 2/26/2016 3:01 PM, Greg KH wrote:
On Fri, Feb 26, 2016 at 01:47:39PM -0800, Yang Shi wrote:
commit 5634cc2aa9aebc77bc862992e7805469dcf83dac ("writeback: update writeback
tracepoints to report cgroup") made writeback tracepoints report cgroup
writeback, but it may trigger the below bug on -rt kernel since kernfs_path
and kernfs_path_len are called by tracepoints, which acquire sleeping lock.

BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:930
in_atomic(): 1, irqs_disabled(): 0, pid: 625, name: kworker/u16:3
INFO: lockdep is turned off.
Preemption disabled at:[<ffffffc000374a5c>] wb_writeback+0xec/0x830

CPU: 7 PID: 625 Comm: kworker/u16:3 Not tainted 4.4.1-rt5 #20
Hardware name: Freescale Layerscape 2085a RDB Board (DT)
Workqueue: writeback wb_workfn (flush-7:0)
Call trace:
[<ffffffc00008d708>] dump_backtrace+0x0/0x200
[<ffffffc00008d92c>] show_stack+0x24/0x30
[<ffffffc0007b0f40>] dump_stack+0x88/0xa8
[<ffffffc000127d74>] ___might_sleep+0x2ec/0x300
[<ffffffc000d5d550>] rt_spin_lock+0x38/0xb8
[<ffffffc0003e0548>] kernfs_path_len+0x30/0x90
[<ffffffc00036b360>] trace_event_raw_event_writeback_work_class+0xe8/0x2e8
[<ffffffc000374f90>] wb_writeback+0x620/0x830
[<ffffffc000376224>] wb_workfn+0x61c/0x950
[<ffffffc000110adc>] process_one_work+0x3ac/0xb30
[<ffffffc0001112fc>] worker_thread+0x9c/0x7a8
[<ffffffc00011a9e8>] kthread+0x190/0x1b0
[<ffffffc000086ca0>] ret_from_fork+0x10/0x30

Since kernfs_* functions are heavily used by cgroup, so it sounds not
reasonable to convert kernfs_rename_lock to raw lock.

Call synchronize_sched() when kernfs_node is updated since tracepoints are
protected by rcu_read_lock_sched.

Create raw version kernfs_path, kernfs_path_len and cgroup_path, which don't
acquire lock and are used by tracepoints only.

Signed-off-by: Yang Shi <yang.shi@xxxxxxxxxx>
---
It should be applicable to both mainline and -rt kernel.
The change survives ftrace stress test in ltp.

v2:
* Adopted Steven's suggestion to call synchronize_sched in kernfs_rename_ns.

fs/kernfs/dir.c | 30 +++++++++++++++++++++++++++---
include/linux/cgroup.h | 10 ++++++++++
include/linux/kernfs.h | 10 ++++++++++
include/trace/events/writeback.h | 4 ++--
4 files changed, 49 insertions(+), 5 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index 996b774..70a9687 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -44,7 +44,7 @@ static int kernfs_name_locked(struct kernfs_node *kn, char *buf, size_t buflen)
return strlcpy(buf, kn->parent ? kn->name : "/", buflen);
}

-static char * __must_check kernfs_path_locked(struct kernfs_node *kn, char *buf,
+static char * __must_check __kernfs_path(struct kernfs_node *kn, char *buf,
size_t buflen)
{
char *p = buf + buflen;
@@ -131,12 +131,33 @@ char *kernfs_path(struct kernfs_node *kn, char *buf, size_t buflen)
char *p;

spin_lock_irqsave(&kernfs_rename_lock, flags);
- p = kernfs_path_locked(kn, buf, buflen);
+ p = __kernfs_path(kn, buf, buflen);
spin_unlock_irqrestore(&kernfs_rename_lock, flags);
return p;
}
EXPORT_SYMBOL_GPL(kernfs_path);

+/* Raw version. Used by tracepoints only without acquiring lock */
+size_t _kernfs_path_len(struct kernfs_node *kn)
+{
+ size_t len = 0;
+
+ do {
+ len += strlen(kn->name) + 1;
+ kn = kn->parent;
+ } while (kn && kn->parent);
+
+ return len;
+}
+
+char *_kernfs_path(struct kernfs_node *kn, char *buf, size_t buflen)
+{
+ char *p;
+
+ p = __kernfs_path(kn, buf, buflen);
+ return p;
+}
+
/**
* pr_cont_kernfs_name - pr_cont name of a kernfs_node
* @kn: kernfs_node of interest
@@ -168,7 +189,7 @@ void pr_cont_kernfs_path(struct kernfs_node *kn)

spin_lock_irqsave(&kernfs_rename_lock, flags);

- p = kernfs_path_locked(kn, kernfs_pr_cont_buf,
+ p = __kernfs_path(kn, kernfs_pr_cont_buf,
sizeof(kernfs_pr_cont_buf));
if (p)
pr_cont("%s", p);
@@ -1397,6 +1418,9 @@ int kernfs_rename_ns(struct kernfs_node *kn, struct kernfs_node *new_parent,
kn->hash = kernfs_name_hash(kn->name, kn->ns);
kernfs_link_sibling(kn);

+ /* Tracepoints are protected by rcu_read_lock_sched */
+ synchronize_sched();
+
kernfs_put(old_parent);
kfree_const(old_name);

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 2162dca..bbd88d3 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -538,6 +538,16 @@ static inline char * __must_check cgroup_path(struct cgroup *cgrp, char *buf,
return kernfs_path(cgrp->kn, buf, buflen);
}

+/*
+ * Without acquiring kernfs_rename_lock in _kernfs_path.
+ * Used by tracepoints only.
+ */
+static inline char * __must_check _cgroup_path(struct cgroup *cgrp, char *buf,
+ size_t buflen)
+{
+ return _kernfs_path(cgrp->kn, buf, buflen);
+}
+
static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
{
pr_cont_kernfs_name(cgrp->kn);
diff --git a/include/linux/kernfs.h b/include/linux/kernfs.h
index af51df3..14c01cdc 100644
--- a/include/linux/kernfs.h
+++ b/include/linux/kernfs.h
@@ -267,8 +267,11 @@ static inline bool kernfs_ns_enabled(struct kernfs_node *kn)

int kernfs_name(struct kernfs_node *kn, char *buf, size_t buflen);
size_t kernfs_path_len(struct kernfs_node *kn);
+size_t _kernfs_path_len(struct kernfs_node *kn);
char * __must_check kernfs_path(struct kernfs_node *kn, char *buf,
size_t buflen);
+char * __must_check _kernfs_path(struct kernfs_node *kn, char *buf,
+ size_t buflen);

I despise functions with just a _ in the front of them, you don't give
anyone a clue as to what they are for, or why to use one vs. the other.
Trust me, you will not remember it either in 1 month, let alone 5 years.

Please change the whole function name to be obvious as to what to use,
or not use.

How about kernfs_*_unlocked() for raw version? And, keep lock version name intact?

Thanks,
Yang


thanks,

greg k-h