lkml.org 
[lkml]   [2013]   [Jun]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [IPC] INFO: suspicious RCU usage.
From
Date
On Mon, 2013-06-10 at 15:56 -0700, Andrew Morton wrote:
> On Thu, 06 Jun 2013 10:35:22 -0700 Davidlohr Bueso <davidlohr.bueso@hp.com> wrote:
>
> > > [ 51.524946]
> > > [ 51.525983] ===============================
> > > [ 51.532875] [ INFO: suspicious RCU usage. ]
> > > [ 51.535385] 3.10.0-rc4-next-20130606 #6 Not tainted
> > > [ 51.538304] -------------------------------
> > > [ 51.540937] /c/kernel-tests/src/stable/include/linux/rcupdate.h:471 Illegal context switch in RCU read-side critical section!
> > > [ 51.548110]
> > > [ 51.548110] other info that might help us debug this:
> > > [ 51.548110]
> > > [ 51.553055]
> > > [ 51.553055] rcu_scheduler_active = 1, debug_locks = 1
> > > [ 51.557199] 2 locks held by trinity/1107:
> > > [ 51.560168] #0: (&ids->rw_mutex){+.+.+.}, at: [<ffffffff811e15ee>] ipcget+0x38/0x2b3
> > > [ 51.566465] #1: (rcu_read_lock){.+.+..}, at: [<ffffffff811e7698>] newseg+0x19d/0x3fd
> > > [ 51.572413]
> > > [ 51.572413] stack backtrace:
> > > [ 51.574761] CPU: 0 PID: 1107 Comm: trinity Not tainted 3.10.0-rc4-next-20130606 #6
> > > [ 51.579331] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
> > > [ 51.583068] 0000000000000001 ffff880004a07d88 ffffffff817b1f5c ffff880004a07db8
> > > [ 51.592119] ffffffff810f2f1d ffffffff81b78569 00000000000001a8 0000000000000000
> > > [ 51.596726] 0000000000000000 ffff880004a07de8 ffffffff810ded5e ffff880004a07fd8
> > > [ 51.605189] Call Trace:
> > > [ 51.606409] [<ffffffff817b1f5c>] dump_stack+0x19/0x1b
> > > [ 51.609632] [<ffffffff810f2f1d>] lockdep_rcu_suspicious+0xeb/0xf4
> > > [ 51.612905] [<ffffffff810ded5e>] __might_sleep+0x59/0x1dc
> > > [ 51.618614] [<ffffffff81238623>] idr_preload+0x9b/0x142
> > > [ 51.621939] [<ffffffff811e0e56>] ipc_addid+0x3d/0x193
> > > [ 51.624373] [<ffffffff811e771c>] newseg+0x221/0x3fd
> > > [ 51.626596] [<ffffffff811e7698>] ? newseg+0x19d/0x3fd
> > > [ 51.630177] [<ffffffff811e1774>] ipcget+0x1be/0x2b3
> > > [ 51.633174] [<ffffffff817bc094>] ? retint_swapgs+0x13/0x1b
> > > [ 51.636356] [<ffffffff811e7a5a>] SyS_shmget+0x59/0x5d
> > > [ 51.639576] [<ffffffff811e74fb>] ? shm_try_destroy_orphaned+0xbf/0xbf
> > > [ 51.643673] [<ffffffff811e6ce5>] ? shm_get_unmapped_area+0x20/0x20
> > > [ 51.647321] [<ffffffff811e6cf0>] ? shm_security+0xb/0xb
> > > [ 51.650831] [<ffffffff817bcb27>] system_call_fastpath+0x16/0x1b
> >
> > I suspect this is caused because now we call idr_preload() in ipc_addid
> > with the rcu lock held by the caller. So, we can either have a two level
> > rcu locking or a two level idr_preload/idr_preload_end.
>
> I'm not sure what to suggest really, apart from the use of explosives.
>
> ipc_addid():
>
> spin_lock_init(&new->lock);
> new->deleted = 0;
> spin_lock(&new->lock);
>
> this makes no sense. If we can run spin_lock_init() against a lock
> then it had darn well better be the case that no other thread is able
> to access that lock. And if no other thread can access that lock then
> there's no need to lock it!

Good point, and yet that's been working for years.

>
> Presumably at some point in the future, other threads can look up this
> object and then the lock becomes useful. Perhaps that's the
> rcu_read_unlock() after a successful idr_alloc() - it's unclear from a
> quick read.
>
>
> Also, ipc_addid() undoes its caller's rcu_read_lock() if idr_alloc()
> failed. This is strange from an interface point of view, is not
> documented in the ipc_addid() interface description and will cause
> newseg() (at least) to perform a double rcu_read_unlock().

Yep, I missed that in the original patch.

>
> As for this particular trace: I'd view the putting of rcu_read_lock()
> around the ipc_addid() call as being the core mistake. By its very
> nature, ipc_addid() allocates memory and hence should be called in
> GFP_KERNEL context.

I completely agree. The patch below restores the rcu locking in
ipc_addid() and will also take care of that idiotic double rcu
unlocking. Just like before, now newseg, newary and newque are in charge
of explicitly only calling rcu_read_unlock.

Fengguang, Sasha, Valdis, does this take care of suspicious RCU usage?

---8<---
From: Davidlohr Bueso <davidlohr.bueso@hp.com>
Subject: [PATCH] ipc: restore rcu locking in ipc_addid

Fengguang reported the following trinity triggered issue:

[ 51.524946]
[ 51.525983] ===============================
[ 51.532875] [ INFO: suspicious RCU usage. ]
[ 51.535385] 3.10.0-rc4-next-20130606 #6 Not tainted
[ 51.538304] -------------------------------
[ 51.540937] /c/kernel-tests/src/stable/include/linux/rcupdate.h:471 Illegal context switch in RCU read-side critical section!
[ 51.548110]
[ 51.548110] other info that might help us debug this:
[ 51.548110]
[ 51.553055]
[ 51.553055] rcu_scheduler_active = 1, debug_locks = 1
[ 51.557199] 2 locks held by trinity/1107:
[ 51.560168] #0: (&ids->rw_mutex){+.+.+.}, at: [<ffffffff811e15ee>] ipcget+0x38/0x2b3
[ 51.566465] #1: (rcu_read_lock){.+.+..}, at: [<ffffffff811e7698>] newseg+0x19d/0x3fd
[ 51.572413]
[ 51.572413] stack backtrace:
[ 51.574761] CPU: 0 PID: 1107 Comm: trinity Not tainted 3.10.0-rc4-next-20130606 #6
[ 51.579331] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
[ 51.583068] 0000000000000001 ffff880004a07d88 ffffffff817b1f5c ffff880004a07db8
[ 51.592119] ffffffff810f2f1d ffffffff81b78569 00000000000001a8 0000000000000000
[ 51.596726] 0000000000000000 ffff880004a07de8 ffffffff810ded5e ffff880004a07fd8
[ 51.605189] Call Trace:
[ 51.606409] [<ffffffff817b1f5c>] dump_stack+0x19/0x1b
[ 51.609632] [<ffffffff810f2f1d>] lockdep_rcu_suspicious+0xeb/0xf4
[ 51.612905] [<ffffffff810ded5e>] __might_sleep+0x59/0x1dc
[ 51.618614] [<ffffffff81238623>] idr_preload+0x9b/0x142
[ 51.621939] [<ffffffff811e0e56>] ipc_addid+0x3d/0x193
[ 51.624373] [<ffffffff811e771c>] newseg+0x221/0x3fd
[ 51.626596] [<ffffffff811e7698>] ? newseg+0x19d/0x3fd
[ 51.630177] [<ffffffff811e1774>] ipcget+0x1be/0x2b3
[ 51.633174] [<ffffffff817bc094>] ? retint_swapgs+0x13/0x1b
[ 51.636356] [<ffffffff811e7a5a>] SyS_shmget+0x59/0x5d
[ 51.639576] [<ffffffff811e74fb>] ? shm_try_destroy_orphaned+0xbf/0xbf
[ 51.643673] [<ffffffff811e6ce5>] ? shm_get_unmapped_area+0x20/0x20
[ 51.647321] [<ffffffff811e6cf0>] ? shm_security+0xb/0xb
[ 51.650831] [<ffffffff817bcb27>] system_call_fastpath+0x16/0x1b

The issue was caused because we were allocating memory in GFP_KERNEL context after
calling rcu_read_lock. This patch restores the rcu_read_lock call into ipc_addid()
and thus maintains the original behavior.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
---
ipc/msg.c | 2 --
ipc/sem.c | 2 --
ipc/shm.c | 2 --
ipc/util.c | 3 ++-
4 files changed, 2 insertions(+), 7 deletions(-)

diff --git a/ipc/msg.c b/ipc/msg.c
index 3b7b4b5..a1cf70e 100644
--- a/ipc/msg.c
+++ b/ipc/msg.c
@@ -196,10 +196,8 @@ static int newque(struct ipc_namespace *ns, struct ipc_params *params)
}

/* ipc_addid() locks msq upon success. */
- rcu_read_lock();
id = ipc_addid(&msg_ids(ns), &msq->q_perm, ns->msg_ctlmni);
if (id < 0) {
- rcu_read_unlock();
security_msg_queue_free(msq);
ipc_rcu_putref(msq);
return id;
diff --git a/ipc/sem.c b/ipc/sem.c
index fad2da5..94ffe72 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -461,10 +461,8 @@ static int newary(struct ipc_namespace *ns, struct ipc_params *params)
return retval;
}

- rcu_read_lock();
id = ipc_addid(&sem_ids(ns), &sma->sem_perm, ns->sc_semmni);
if (id < 0) {
- rcu_read_unlock();
security_sem_free(sma);
ipc_rcu_putref(sma);
return id;
diff --git a/ipc/shm.c b/ipc/shm.c
index 202e014..c6b4ad5 100644
--- a/ipc/shm.c
+++ b/ipc/shm.c
@@ -521,11 +521,9 @@ static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
if (IS_ERR(file))
goto no_file;

- rcu_read_lock();
id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
if (id < 0) {
error = id;
- rcu_read_unlock();
goto no_id;
}

diff --git a/ipc/util.c b/ipc/util.c
index a746abb..a0c139f 100644
--- a/ipc/util.c
+++ b/ipc/util.c
@@ -246,7 +246,7 @@ int ipc_get_maxid(struct ipc_ids *ids)
* is returned. The 'new' entry is returned in a locked state on success.
* On failure the entry is not locked and a negative err-code is returned.
*
- * Called with RCU read lock and writer ipc_ids.rw_mutex held.
+ * Called with writer ipc_ids.rw_mutex held.
*/
int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
{
@@ -265,6 +265,7 @@ int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)

spin_lock_init(&new->lock);
new->deleted = 0;
+ rcu_read_lock();
spin_lock(&new->lock);

id = idr_alloc(&ids->ipcs_idr, new,
--
1.7.11.7




\
 
 \ /
  Last update: 2013-06-12 02:41    [W:0.065 / U:2.736 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site