Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1269608

[PATCH 3.2 26/60] sched/core: Fix TASK_DEAD race in finish_task_switch()

From Ben Hutchings <ben@decadent.org.uk>
Newsgroups linux.kernel
Subject [PATCH 3.2 26/60] sched/core: Fix TASK_DEAD race in finish_task_switch()
Date 2015-11-15 03:20 +0100
Message-ID <quV59-3JS-41@gated-at.bofh.it> (permalink)
References <quUVr-3GF-3@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw


3.2.73-rc1 review patch.  If anyone has any objections, please let me know.

------------------

From: Peter Zijlstra <peterz@infradead.org>

commit 95913d97914f44db2b81271c2e2ebd4d2ac2df83 upstream.

So the problem this patch is trying to address is as follows:

        CPU0                            CPU1

        context_switch(A, B)
                                        ttwu(A)
                                          LOCK A->pi_lock
                                          A->on_cpu == 0
        finish_task_switch(A)
          prev_state = A->state  <-.
          WMB                      |
          A->on_cpu = 0;           |
          UNLOCK rq0->lock         |
                                   |    context_switch(C, A)
                                   `--  A->state = TASK_DEAD
          prev_state == TASK_DEAD
            put_task_struct(A)
                                        context_switch(A, C)
                                        finish_task_switch(A)
                                          A->state == TASK_DEAD
                                            put_task_struct(A)

The argument being that the WMB will allow the load of A->state on CPU0
to cross over and observe CPU1's store of A->state, which will then
result in a double-drop and use-after-free.

Now the comment states (and this was true once upon a long time ago)
that we need to observe A->state while holding rq->lock because that
will order us against the wakeup; however the wakeup will not in fact
acquire (that) rq->lock; it takes A->pi_lock these days.

We can obviously fix this by upgrading the WMB to an MB, but that is
expensive, so we'd rather avoid that.

The alternative this patch takes is: smp_store_release(&A->on_cpu, 0),
which avoids the MB on some archs, but not important ones like ARM.

Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Cc: manfred@colorfullife.com
Cc: will.deacon@arm.com
Fixes: e4a52bcb9a18 ("sched: Remove rq->lock from the first half of ttwu()")
Link: http://lkml.kernel.org/r/20150929124509.GG3816@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.2:
 - Adjust filename
 - As smp_store_release() is not defined, use smp_mb()]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
---
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -1016,8 +1016,10 @@ static inline void finish_lock_switch(st
 	 * After ->on_cpu is cleared, the task can be moved to a different CPU.
 	 * We must ensure this doesn't happen until the switch is completely
 	 * finished.
+	 *
+	 * Pairs with the control dependency and rmb in try_to_wake_up().
 	 */
-	smp_wmb();
+	smp_mb();
 	prev->on_cpu = 0;
 #endif
 #ifdef CONFIG_DEBUG_SPINLOCK
@@ -3191,11 +3192,11 @@ static void finish_task_switch(struct rq
 	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
 	 * schedule one last time. The schedule call will never return, and
 	 * the scheduled task must drop that reference.
-	 * The test for TASK_DEAD must occur while the runqueue locks are
-	 * still held, otherwise prev could be scheduled on another cpu, die
-	 * there before we look at prev->state, and then the reference would
-	 * be dropped twice.
-	 *		Manfred Spraul <manfred@colorfullife.com>
+	 *
+	 * We must observe prev->state before clearing prev->on_cpu (in
+	 * finish_lock_switch), otherwise a concurrent wakeup can get prev
+	 * running on another CPU and we could rave with its RUNNING -> DEAD
+	 * transition, resulting in a double drop.
 	 */
 	prev_state = prev->state;
 	finish_arch_switch(prev);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Back to linux.kernel | Previous | NextPrevious in thread | Next in thread | Find similar | Unroll thread


Thread

[PATCH 3.2 00/60] 3.2.73-rc1 review Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 43/60] dm btree remove: fix a bug when rebalancing  nodes after removal Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 27/60] 3w-9xxx: don't unmap bounce buffered commands Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 12/60] genirq: Fix race in register_irq_proc() Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 31/60] iommu/vt-d: fix range computation when making  room for large pages Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 28/60] xen-blkfront: check for null drvdata in  blkback_changed (XenbusStateClosing) Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 32/60] xhci: don't finish a TD if we get a short  transfer event mid TD Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 46/60] md/raid1: don't clear bitmap bit when  bad-block-list write fails. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 24/60] iwlwifi: dvm: fix D3 firmware PN programming Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 04/60] regmap: debugfs: Don't bother actually printing  when calculating max length Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:10 +0100
  [PATCH 3.2 55/60] asix: Don't reset PHY on if_up for ASIX 88772 Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 08/60] UBI: Validate data_size Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 53/60] skbuff: Fix skb checksum partial check. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 38/60] IB/cm: Fix rb-tree duplicate free and  use-after-free Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 20/60] usb: Add device quirk for Logitech PTZ cameras Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 33/60] xhci: handle no ping response error properly Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 19/60] USB: Add reset-resume quirk for two Plantronics  usb headphones. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 44/60] dm btree: fix leak of bufio-backed block in  btree_split_beneath error path Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 40/60] powerpc/rtas: Validate rtas.entry before  calling enter_rtas() Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 34/60] xhci: Switch Intel Lynx Point LP ports to EHCI  on shutdown. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 07/60] x86/xen: Do not clip xen_e820_map to  xen_e820_map_entries when sanitizing map Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 03/60] regmap: debugfs: Ensure we don't underflow when  printing access masks Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 51/60] net: add length argument to  skb_copy_and_csum_datagram_iovec Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 05/60] ath9k: declare required extra tx headroom Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 39/60] drm/nouveau/gem: return only valid domain when  there's only one Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 06/60] m68k: Define asmlinkage_protect Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 26/60] sched/core: Fix TASK_DEAD race in  finish_task_switch() Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 45/60] md/raid1: ensure device failure recorded before  write request returns. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 15/60] md/raid0: update queue parameter in a safer  location. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 59/60] KEYS: Fix race between key destruction and  finding a keyring by name Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 23/60] ppp: don't override sk->sk_state in  pppoe_flush_dev() Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 49/60] mvsas: Fix NULL pointer dereference in  mvs_slot_task_free Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 58/60] KVM: x86: work around infinite loop in  microcode when #AC is delivered Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:20 +0100
  [PATCH 3.2 42/60] ppp: fix pppoe_dev deletion condition in  pppoe_release() Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 10/60] MIPS: dma-default: Fix 32-bit fall back to GFP_DMA Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 50/60] sched: declare pid_alive as inline Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 17/60] clocksource: Fix abs() usage w/ 64bit values Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 36/60] crypto: api - Only abort operations on fatal signal Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 16/60] md/raid0: apply base queue limits *before*  disk_stack_limits Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 47/60] md/raid10: ensure device failure recorded  before write request returns. Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 57/60] Failing to send a CLOSE if file is opened  WRONLY and server reboots on a 4.x mount Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  [PATCH 3.2 14/60] [SMB3] Do not fall back to SMBWriteX in  set_file_size error cases Ben Hutchings <ben@decadent.org.uk> - 2015-11-15 03:30 +0100
  Re: [PATCH 3.2 00/60] 3.2.73-rc1 review Guenter Roeck <linux@roeck-us.net> - 2015-11-15 14:50 +0100
    Re: [PATCH 3.2 00/60] 3.2.73-rc1 review Ben Hutchings <ben@decadent.org.uk> - 2015-11-16 12:20 +0100

csiph-web