Message-ID: <40287575-53e5-afd4-a198-3c3e4d6f6a37@maine.edu>
Date: Wed, 3 Jan 2024 14:13:14 -0500 (EST)
From: Vince Weaver <vincent.weaver@...ne.edu>
To: Ian Rogers <irogers@...gle.com>
cc: peterz@...radead.org, mingo@...hat.com, acme@...nel.org, 
    mark.rutland@....com, 
    Alexander Shishkin <alexander.shishkin@...ux.intel.com>, jolsa@...nel.org, 
    namhyung@...nel.org, adrian.hunter@...el.com, linux-kernel@...r.kernel.org, 
    linux-perf-users <linux-perf-users@...r.kernel.org>
Subject: Re: [perf] vsyscall, possible circular locking dependency detected

On Wed, 3 Jan 2024, Ian Rogers wrote:

> On Wed, Jan 3, 2024 at 10:08 AM Ian Rogers <irogers@...gle.com> wrote:
> >
> > On Wed, Dec 13, 2023 at 7:08 AM Vince Weaver <vincent.weaver@...ne.edu> wrote:
> > >
> > > Hello
> > >
> > > I recently added vsyscall fuzzing support to the perf_fuzzer and it gave
> > > me this dump when fuzzing on a haswell machine running current git
> > > (6.7.0-rc5+)
> > >
> > > [12559.226693] ======================================================
> > > [12559.232906] WARNING: possible circular locking dependency detected
> > > [12559.239116] 6.7.0-rc5+ #1 Tainted: G        W
> > > [12559.244286] ------------------------------------------------------
> > > [12559.250493] perf_fuzzer/284551 is trying to acquire lock:
> > > [12559.255923] ffff88800aff65d8 (&sighand->siglock){-.-.}-{2:2}, at: force_sig_info_to_task+0x41/0x260
> > > [12559.265034]
> > >                but task is already holding lock:
> > > [12559.270891] ffff8880d08bd4d8 (&rq->__lock){-.-.}-{2:2}, at: raw_spin_rq_lock_nested+0x26/0xb0
> > > [12559.279463]
> > >                which lock already depends on the new lock.
> > >
> > > [12559.287670]
> > >                the existing dependency chain (in reverse order) is:
> > > [12559.295182]
> > >                -> #1 (&rq->__lock){-.-.}-{2:2}:
> > > [12559.300966]        _raw_spin_lock_nested+0x32/0x40
> > > [12559.305789]        raw_spin_rq_lock_nested+0x26/0xb0
> > > [12559.310783]        _raw_spin_rq_lock_irqsave+0x23/0x60
> > > [12559.315959]        sched_mm_cid_before_execve+0xcd/0x3d0
> > > [12559.321306]        exit_signals+0x1a3/0x450
> > > [12559.325521]        do_exit+0x24b/0x1100
> > > [12559.329389]        __x64_sys_exit+0x2c/0x30
> > > [12559.333600]        do_syscall_64+0x41/0x120
> > > [12559.337814]        entry_SYSCALL_64_after_hwframe+0x6e/0x76
> > > [12559.343423]
> > >                -> #0 (&sighand->siglock){-.-.}-{2:2}:
> > > [12559.349732]        __lock_acquire+0x1d8d/0x31f0
> > > [12559.354296]        lock_acquire+0x16c/0x3e0
> > > [12559.358506]        _raw_spin_lock_irqsave+0x47/0x70
> > > [12559.363416]        force_sig_info_to_task+0x41/0x260
> > > [12559.368418]        force_sig_fault+0xaa/0xf0
> > > [12559.372722]        exc_page_fault+0x61/0xf0
> > > [12559.376932]        asm_exc_page_fault+0x22/0x30
> > > [12559.381494]        __get_user_handle_exception+0x0/0x20
> > > [12559.386755]        perf_callchain_user+0x2d3/0x390
> > > [12559.391581]        get_perf_callchain+0x21a/0x2d0
> > > [12559.396312]        perf_callchain+0xc3/0xe0
> > > [12559.400526]        perf_prepare_sample+0x4f1/0xc70
> > > [12559.405346]        perf_event_output_forward+0xbf/0x270
> > > [12559.410599]        __perf_event_overflow+0x1ad/0x3c0
> > > [12559.415597]        ___perf_sw_event+0x1f7/0x390
> > > [12559.420154]        __schedule+0x12d3/0x2560
> > > [12559.424367]        preempt_schedule_irq+0x47/0x90
> > > [12559.429098]        irqentry_exit+0x31/0x80
> > > [12559.433223]        asm_sysvec_apic_timer_interrupt+0x16/0x20
> > > [12559.438912]        seqcount_lockdep_reader_access.constprop.0+0x88/0x90
> > > [12559.445562]        ktime_get_real_ts64+0x28/0x120
> > > [12559.450297]        __x64_sys_gettimeofday+0x86/0x150
> > > [12559.455292]        emulate_vsyscall+0x258/0x590
> > > [12559.459858]        do_user_addr_fault+0x346/0x8c0
> > > [12559.464588]        exc_page_fault+0x61/0xf0
> > > [12559.468798]        asm_exc_page_fault+0x22/0x30
> > > [12559.473360]        _end+0x7a5da000/0x0
> > > [12559.477135]
> > >                other info that might help us debug this:
> > >
> > > [12559.485163]  Possible unsafe locking scenario:
> > >
> > > [12559.491115]        CPU0                    CPU1
> > > [12559.495670]        ----                    ----
> > > [12559.500228]   lock(&rq->__lock);
> > > [12559.503492]                                lock(&sighand->siglock);
> > > [12559.509791]                                lock(&rq->__lock);
> > > [12559.515569]   lock(&sighand->siglock);
> > > [12559.519352]
> > >                 *** DEADLOCK ***
> > >
> > > [12559.525303] 3 locks held by perf_fuzzer/284551:
> > > [12559.529867]  #0: ffff8880d08bd4d8 (&rq->__lock){-.-.}-{2:2}, at: raw_spin_rq_lock_nested+0x26/0xb0
> > > [12559.538873]  #1: ffffffff831a3ec0 (rcu_read_lock){....}-{1:2}, at: ___perf_sw_event+0x10f/0x390
> > > [12559.547618]  #2: ffffffff831a3ec0 (rcu_read_lock){....}-{1:2}, at: perf_event_output_forward+0x8f/0x270
> > > [12559.557058]
> > >                stack backtrace:
> > > [12559.561439] CPU: 5 PID: 284551 Comm: perf_fuzzer Tainted: G        W          6.7.0-rc5+ #1
> > > [12559.569823] Hardware name: LENOVO 10AM000AUS/SHARKBAY, BIOS FBKT72AUS 01/26/2014
> > > [12559.577248] Call Trace:
> > > [12559.579720]  <TASK>
> > > [12559.581846]  dump_stack_lvl+0x57/0x90
> > > [12559.585539]  check_noncircular+0x24f/0x290
> > > [12559.589668]  ? __pfx_check_noncircular+0x10/0x10
> > > [12559.594327]  ? lockdep_lock+0xb3/0x170
> > > [12559.598106]  ? __pfx_lockdep_lock+0x10/0x10
> > > [12559.602323]  ? add_chain_block+0x2dc/0x300
> > > [12559.606456]  __lock_acquire+0x1d8d/0x31f0
> > > [12559.610508]  ? __pfx___lock_acquire+0x10/0x10
> > > [12559.614896]  ? __pfx_mark_lock+0x10/0x10
> > > [12559.618849]  lock_acquire+0x16c/0x3e0
> > > [12559.622548]  ? force_sig_info_to_task+0x41/0x260
> > > [12559.627197]  ? __pfx_lock_acquire+0x10/0x10
> > > [12559.631414]  ? __pfx_mark_lock+0x10/0x10
> > > [12559.635370]  ? mark_lock+0x10b/0xd40
> > > [12559.638981]  ? mark_lock+0x10b/0xd40
> > > [12559.642586]  _raw_spin_lock_irqsave+0x47/0x70
> > > [12559.646971]  ? force_sig_info_to_task+0x41/0x260
> > > [12559.651618]  force_sig_info_to_task+0x41/0x260
> > > [12559.656101]  force_sig_fault+0xaa/0xf0
> > > [12559.659882]  ? __pfx_force_sig_fault+0x10/0x10
> > > [12559.664360]  ? kernelmode_fixup_or_oops+0x10a/0x150
> > > [12559.669272]  exc_page_fault+0x61/0xf0
> > > [12559.672964]  asm_exc_page_fault+0x22/0x30
> > > [12559.677007] RIP: 0010:__get_user_handle_exception+0x0/0x20
> > > [12559.682524] Code: 90 90 90 90 90 90 90 90 90 90 0f 1f 00 0f ae e8 48 8b 10 31 c0 0f 1f 00 c3 cc cc cc cc 66 66 2e 0f 1f 84 00 00 00 00 00 66 90 <0f> 1f 00 31 d2 48 c7 c0 f2 ff ff ff c3 cc cc cc cc cc cc cc cc cc
> > > [12559.701336] RSP: 0000:ffffc90021bef5d0 EFLAGS: 00010006
> > > [12559.706592] RAX: 0000000000000005 RBX: ffffc90021bef680 RCX: ffffffff8100a5a3
> > > [12559.713764] RDX: 0000000000000011 RSI: ffffffff8100a55d RDI: ffff8880138d8090
> > > [12559.720931] RBP: 0000000000000005 R08: 0000000000000000 R09: fffffbfff0799b5c
> > > [12559.728098] R10: ffffffff83ccdae1 R11: 0000000000000000 R12: ffffc90021bef68c
> > > [12559.735266] R13: ffffc90021bef688 R14: 0000000000000005 R15: ffff8880138d8000
> > > [12559.742436]  ? perf_callchain_user+0x353/0x390
> > > [12559.746914]  ? perf_callchain_user+0x30d/0x390
> > > [12559.751391]  perf_callchain_user+0x2d3/0x390
> > > [12559.755699]  get_perf_callchain+0x21a/0x2d0
> > > [12559.759917]  ? __pfx_get_perf_callchain+0x10/0x10
> > > [12559.764662]  ? lock_is_held_type+0xe3/0x140
> > > [12559.768882]  ? preempt_schedule_irq+0x47/0x90
> > > [12559.773269]  perf_callchain+0xc3/0xe0
> > > [12559.776961]  perf_prepare_sample+0x4f1/0xc70
> > > [12559.781263]  ? __pfx_perf_prepare_sample+0x10/0x10
> > > [12559.786084]  ? __pfx___lock_acquire+0x10/0x10
> > > [12559.790478]  perf_event_output_forward+0xbf/0x270
> > > [12559.795214]  ? __pfx_perf_event_output_forward+0x10/0x10
> > > [12559.800554]  ? lock_acquire+0x16c/0x3e0
> > > [12559.804420]  ? lock_acquire+0x17c/0x3e0
> > > [12559.808295]  ? __pfx_lock_acquire+0x10/0x10
> > > [12559.812511]  ? ___perf_sw_event+0x22f/0x390
> > > [12559.816727]  ? __pfx_lock_release+0x10/0x10
> > > [12559.820951]  __perf_event_overflow+0x1ad/0x3c0
> > > [12559.825428]  ___perf_sw_event+0x1f7/0x390
> > > [12559.829467]  ? __pfx____perf_sw_event+0x10/0x10
> > > [12559.834037]  ? preempt_schedule_irq+0x47/0x90
> > > [12559.838426]  ? __update_load_avg_se+0x86/0x5c0
> > > [12559.842900]  ? __update_load_avg_cfs_rq+0x70/0x570
> > > [12559.847726]  ? lock_acquire+0x16c/0x3e0
> > > [12559.851590]  ? __update_load_avg_se+0x15f/0x5c0
> > > [12559.856154]  ? lock_is_held_type+0xe3/0x140
> > > [12559.860372]  ? update_load_avg+0x154/0xdc0
> > > [12559.864499]  ? lock_is_held_type+0xe3/0x140
> > > [12559.868721]  ? set_next_entity+0xec/0x1e0
> > > [12559.872764]  ? preempt_schedule_irq+0x47/0x90
> > > [12559.877149]  ? __schedule+0x12d3/0x2560
> > > [12559.881012]  __schedule+0x12d3/0x2560
> > > [12559.884717]  ? __pfx___schedule+0x10/0x10
> > > [12559.888754]  ? __pfx___lock_acquire+0x10/0x10
> > > [12559.893139]  ? lock_is_held_type+0xe3/0x140
> > > [12559.897353]  ? mark_held_locks+0x24/0x90
> > > [12559.901316]  preempt_schedule_irq+0x47/0x90
> > > [12559.905531]  irqentry_exit+0x31/0x80
> > > [12559.909136]  asm_sysvec_apic_timer_interrupt+0x16/0x20
> > > [12559.914305] RIP: 0010:seqcount_lockdep_reader_access.constprop.0+0x88/0x90
> > > [12559.921217] Code: 31 f6 48 c7 c7 48 7c 1c 83 e8 84 25 f8 ff 48 8b 74 24 08 48 c7 c7 48 7c 1c 83 e8 93 1c f8 ff e8 6e d2 0c 00 fb 0f 1f 44 00 00 <5a> c3 cc cc cc cc 66 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90
> > > [12559.940025] RSP: 0000:ffffc90021befdb8 EFLAGS: 00000202
> > > [12559.945281] RAX: 000000000182a8d9 RBX: ffffc90021befe18 RCX: dffffc0000000000
> > > [12559.952453] RDX: 0000000000000000 RSI: ffffffff826804e0 RDI: ffffffff828c8620
> > > [12559.959619] RBP: 00007ffcf0f36fd0 R08: 0000000000000001 R09: 0000000000000001
> > > [12559.966786] R10: ffffffff8371fd27 R11: 0000000000000000 R12: 0000000000000000
> > > [12559.973953] R13: 0000000000000000 R14: ffff88800aa55340 R15: ffffc90021befff0
> > > [12559.981129]  ? seqcount_lockdep_reader_access.constprop.0+0x82/0x90
> > > [12559.987426]  ? ktime_get_real_ts64+0x28/0x120
> > > [12559.991820]  ktime_get_real_ts64+0x28/0x120
> > > [12559.996032]  __x64_sys_gettimeofday+0x86/0x150
> > > [12560.000512]  ? __pfx___x64_sys_gettimeofday+0x10/0x10
> > > [12560.005602]  ? do_user_addr_fault+0x3c5/0x8c0
> > > [12560.009994]  emulate_vsyscall+0x258/0x590
> > > [12560.014034]  ? 0xffffffffff600000
> > > [12560.017376]  do_user_addr_fault+0x346/0x8c0
> > > [12560.021587]  ? rcu_is_watching+0x34/0x60
> > > [12560.025541]  ? 0xffffffffff600000
> > > [12560.028882]  exc_page_fault+0x61/0xf0
> > > [12560.032578]  asm_exc_page_fault+0x22/0x30
> > > [12560.036616] RIP: 0033:_end+0x7a5da000/0x0
> > > [12560.040653] Code: Unable to access opcode bytes at 0xffffffffff5fffd6.
> > > [12560.047210] RSP: 002b:00007ffcf0f36fb8 EFLAGS: 00010246
> > > [12560.052470] RAX: ffffffffffffffda RBX: 000000000000000d RCX: 000000000000000a
> > > [12560.059632] RDX: 0000000000000010 RSI: 0000000000000000 RDI: 00007ffcf0f36fd0
> > > [12560.066799] RBP: 00007ffcf0f36fc0 R08: 00007f482684b274 R09: 00007f482684b280
> > > [12560.073966] R10: 00007f4826675a10 R11: 0000000000000246 R12: 0000000000000000
> > > [12560.081132] R13: 00007ffcf0f39448 R14: 000055f2641d2dd8 R15: 00007f48268a3000
> > > [12560.088304]  </TASK>
> >
> > Thanks Vince, the potential deadlock is bad. I notice you updated your tests at:
> > https://github.com/deater/perf_event_tests
> > is there a simple reproduction for this failure? Has there been any
> > other progress on this issue?
> 
> Apologies, I just saw the syzkaller reproduction and your comments:
> 
>https://lore.kernel.org/lkml/CABOYnLxoG=YsE1XyOjfF0O+2UrdTyK1JMir3xPapGsRzh -XFAA@...l.gmail.com/

I think that is a separate bug.
The warning from the link you provide is
	 WARNING: CPU: 1 PID: 8266 at kernel/events/core.c:1950 __do_sys_perf_event_open0
and the perf_fuzzer can trigger this within seconds (and apparently 
trinity and syzkaller all triggered it as well).

The deadlock mentioned in the e-mail you are replying to here is different 
and took many hours before triggering, and I think is vsyscall related.
I only recently added vsyscall support to the fuzzer.

I don't necessarily have a reproducer for it.  I can try to trigger it 
again and get one together.  There was a power outage so my fuzzing 
machines are all down and I need to get them running again.

Vince 

Powered by blists - more mailing lists