From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Paul E. McKenney" Subject: Re: 2.6.39-rc4+: Kernel leaking memory during FS scanning, regression? Date: Wed, 27 Apr 2011 15:02:20 -0700 Message-ID: <20110427220220.GP2135@linux.vnet.ibm.com> References: <20110425191607.GL2468@linux.vnet.ibm.com> <20110425231016.34b4293e@neptune.home> <20110425214933.GO2468@linux.vnet.ibm.com> <20110426081904.0d2b1494@pluto.restena.lu> <20110426112756.GF4308@linux.vnet.ibm.com> <20110426183859.6ff6279b@neptune.home> <20110426185036.GG2135@linux.vnet.ibm.com> Reply-To: paulmck@linux.vnet.ibm.com Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable Cc: Linus Torvalds , Bruno =?iso-8859-1?Q?Pr=E9mont?= , Ingo Molnar , Peter Zijlstra , Mike Frysinger , KOSAKI Motohiro , linux-kernel@vger.kernel.org, linux-mm@kvack.org, linux-fsdevel@vger.kernel.org, "Paul E. McKenney" , Pekka Enberg To: sedat.dilek@gmail.com Return-path: Content-Disposition: inline In-Reply-To: Sender: owner-linux-mm@kvack.org List-Id: linux-fsdevel.vger.kernel.org On Tue, Apr 26, 2011 at 09:17:28PM +0200, Sedat Dilek wrote: > On Tue, Apr 26, 2011 at 8:50 PM, Paul E. McKenney > wrote: > > On Tue, Apr 26, 2011 at 10:12:39AM -0700, Linus Torvalds wrote: > >> On Tue, Apr 26, 2011 at 9:38 AM, Bruno Pr=E9mont > >> wrote: > >> > > >> > Here it comes: > >> > > >> > rcu_kthread (when build processes are STOPped): > >> > [ =A0836.050003] rcu_kthread =A0 =A0 R running =A0 7324 =A0 =A0 6 = =A0 =A0 =A02 0x00000000 > >> > [ =A0836.050003] =A0dd473f28 00000046 5a000240 dd65207c dd407360 d= d651d40 0000035c dd473ed8 > >> > [ =A0836.050003] =A0c10bf8a2 c14d63d8 dd65207c dd473f28 dd445040 d= d445040 dd473eec c10be848 > >> > [ =A0836.050003] =A0dd651d40 dd407360 ddfdca00 dd473f14 c10bfde2 0= 0000000 00000001 000007b6 > >> > [ =A0836.050003] Call Trace: > >> > [ =A0836.050003] =A0[] ? check_object+0x92/0x210 > >> > [ =A0836.050003] =A0[] ? init_object+0x38/0x70 > >> > [ =A0836.050003] =A0[] ? free_debug_processing+0x112/0x1= f0 > >> > [ =A0836.050003] =A0[] ? lock_timer_base+0x2d/0x70 > >> > [ =A0836.050003] =A0[] schedule_timeout+0x137/0x280 > >> > >> Hmm. > >> > >> I'm adding Ingo and Peter to the cc, because this whole "rcu_kthread > >> is running, but never actually running" is starting to smell like a > >> scheduler issue. > >> > >> Peter/Ingo: RCUTINY seems to be broken for Bruno. During any kind of > >> heavy workload, at some point it looks like rcu_kthread simply stops > >> making any progress. It's constantly in runnable state, but it doesn= 't > >> actually use any CPU time, and it's not processing the RCU callbacks= , > >> so the RCU memory freeing isn't happening, and slabs just build up > >> until the machine dies. > >> > >> And it really is RCUTINY, because the thing doesn't happen with the > >> regular tree-RCU. > > > > The difference between TINY_RCU and TREE_RCU is that TREE_RCU still u= ses > > softirq for the core RCU processing. =A0TINY_RCU switched to a kthrea= d > > when I implemented RCU priority boosting. =A0There is a similar chang= e in > > my -rcu tree that makes TREE_RCU use kthreads, and Sedat has been run= ning > > into a very similar problem with that change in place. =A0Which is wh= y I > > do not yet push it to the -next tree. > > > >> This is without CONFIG_RCU_BOOST_PRIO, so we basically have > >> > >> =A0 =A0 =A0 =A0 struct sched_param sp; > >> > >> =A0 =A0 =A0 =A0 rcu_kthread_task =3D kthread_run(rcu_kthread, NULL, = "rcu_kthread"); > >> =A0 =A0 =A0 =A0 sp.sched_priority =3D RCU_BOOST_PRIO; > >> =A0 =A0 =A0 =A0 sched_setscheduler_nocheck(rcu_kthread_task, SCHED_F= IFO, &sp); > >> > >> where RCU_BOOST_PRIO is 1 for the non-boost case. > > > > Good point! =A0Bruno, Sedat, could you please set CONFIG_RCU_BOOST_PR= IO to > > (say) 50, and see if this still happens? =A0(I bet that you do, but..= .) > > >=20 > What's with CONFIG_RCU_BOOST_DELAY setting? CONFIG_RCU_BOOST_DELAY controls how long preemptible RCU lets a grace period run before boosting the priority of any blocked RCU readers. It is completely irrelevant if the rcu_kthread task isn't getting a chance to run, though. This is because it is the rcu_kthread task that does the boosting. > Are those values OK? >=20 > $ egrep 'M486|M686|X86_UP|CONFIG_SMP|NR_CPUS|PREEMPT|_RCU|_HIGHMEM|PAE'= .config > CONFIG_TREE_PREEMPT_RCU=3Dy > CONFIG_PREEMPT_RCU=3Dy > CONFIG_RCU_TRACE=3Dy > CONFIG_RCU_FANOUT=3D32 > # CONFIG_RCU_FANOUT_EXACT is not set > CONFIG_TREE_RCU_TRACE=3Dy > CONFIG_RCU_BOOST=3Dy I suggest CONFIG_RCU_BOOST=3Dn to keep things simple for the moment, but CONFIG_RCU_BOOST=3Dy should be OK too. > CONFIG_RCU_BOOST_PRIO=3D50 > CONFIG_RCU_BOOST_DELAY=3D500 > CONFIG_SMP=3Dy > # CONFIG_M486 is not set > CONFIG_M686=3Dy I don't have an opinion on CONFIG_M486 vs. CONFIG_M686. > CONFIG_NR_CPUS=3D32 > # CONFIG_PREEMPT_NONE is not set > # CONFIG_PREEMPT_VOLUNTARY is not set > CONFIG_PREEMPT=3Dy > CONFIG_HIGHMEM4G=3Dy > # CONFIG_HIGHMEM64G is not set > CONFIG_HIGHMEM=3Dy > CONFIG_DEBUG_OBJECTS_RCU_HEAD=3Dy > CONFIG_DEBUG_PREEMPT=3Dy The above two could be left out, but shouldn't hurt. > # CONFIG_SPARSE_RCU_POINTER is not set > # CONFIG_DEBUG_HIGHMEM is not set > CONFIG_RCU_TORTURE_TEST=3Dm > CONFIG_RCU_CPU_STALL_TIMEOUT=3D60 > CONFIG_RCU_CPU_STALL_VERBOSE=3Dy > CONFIG_PREEMPT_TRACER=3Dy So they look fine to me, the ones that I understand, anyway. ;-) Thanx, Paul >=20 > - Sedat - >=20 > >> Is that so low that even the idle thread will take priority? It's a = UP > >> config with PREEMPT_VOLUNTARY. So pretty much _all_ the stars are > >> aligned for odd scheduling behavior. > >> > >> Other users of SCHED_FIFO tend to set the priority really high (eg > >> "MAX_RT_PRIO-1" is clearly the default one - softirq's, watchdog), b= ut > >> "1" is not unheard of either (touchscreen/ucb1400_ts and > >> mmc/core/sdio_irq), and there are some other random choises out tere= . > >> > >> Any ideas? > > > > I have found one bug so far in my code, but it only affects TREE_RCU > > in my -rcu tree, and even then only if HOTPLUG_CPU is enabled. =A0I a= m > > testing a fix, but I expect Sedat's tests to still break. > > > > I gave Sedat a patch that make rcu_kthread() run at normal (non-realt= ime) > > priority, and he did not see the failure. =A0So running non-realtime = at > > least greatly reduces the probability of failure. > > > > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0= =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Thanx, Paul > > -- > > To unsubscribe from this list: send the line "unsubscribe linux-fsdev= el" in > > the body of a message to majordomo@vger.kernel.org > > More majordomo info at =A0http://vger.kernel.org/majordomo-info.html > > -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter= .ca/ Don't email: email@kvack.org