Re: PROVE_RCU_LIST + /proc/lockdep warning

From: Boqun Feng
Date: Wed Mar 11 2020 - 20:48:36 EST

Next message: Dave Chinner: "Re: [PATCH V5 00/12] Enable per-file/per-directory DAX operations V5"
Previous message: Ashish Kalra: "Re: [PATCH v2] swiotlb: Adjust SWIOTBL bounce buffer size for SEV guests."
In reply to: Qian Cai: "Re: PROVE_RCU_LIST + /proc/lockdep warning"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Wed, Mar 11, 2020 at 09:24:50AM -0400, Qian Cai wrote:
> On Wed, 2020-03-11 at 11:30 +0800, Boqun Feng wrote:
> > On Tue, Mar 10, 2020 at 02:42:24PM -0400, Qian Cai wrote:
> > > On Mon, 2020-03-09 at 09:40 +0800, Boqun Feng wrote:
> > > > On Sat, Mar 07, 2020 at 12:16:18PM -0500, Joel Fernandes wrote:
> > > > > On Thu, Mar 05, 2020 at 11:06:10PM -0500, Qian Cai wrote:
> > > > > > Since the linux-next commit c9af03c14bfd (âDefault enable RCU list lockdep debugging with PROVE_RCUâ),
> > > > > > read /proc/lockdep will trigger a warning with this config below. Reverted the commit fixed the issue
> > > > > > right away.
> > > > > >
> > > > > > https://raw.githubusercontent.com/cailca/linux-mm/master/x86.config
> > > > >
> > > > > Hmm, since Peter hates the list-RCU checking patches and rejected a patch by
> > > > > Amol to fix this (;-)), the easiest way to resolve it would be to just bypass
> > > > > the check in lockdep code:
> > > > >
> > > > > Peter, this should be the last of the list-RCU changes and thank you for the
> > > > > patience.
> > > > >
> > > > > Should I or Amol send a patch for this?
> > > > >
> > > >
> > > > Hmm.. IIUC, the warning got triggered here is because
> > > > lockdep_count_forward_deps() didn't set up the ->lockdep_recursion, as a
> > > > result, __bfs() was called without ->lockdep_recursion being 1, which
> > > > introduced the inconsistency of lockdep status. So how about the
> > > > following (untested) fix?:
> > > >
> > > > Thoughts?
> > > >
> > > > Regards,
> > > > Boqun
> > > >
> > > > -------------------------
> > > > diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> > > > index 32406ef0d6a2..a258640edace 100644
> > > > --- a/kernel/locking/lockdep.c
> > > > +++ b/kernel/locking/lockdep.c
> > > > @@ -1720,7 +1720,9 @@ unsigned long lockdep_count_forward_deps(struct lock_class *class)
> > > >
> > > > raw_local_irq_save(flags);
> > > > arch_spin_lock(&lockdep_lock);
> > > > + current->lockdep_recursion = 1;
> > > > ret = __lockdep_count_forward_deps(&this);
> > > > + current->lockdep_recursion = 0;
> > > > arch_spin_unlock(&lockdep_lock);
> > > > raw_local_irq_restore(flags);
> > >
> > > This does not work. Still the same splat.
> > >
> >
> > Hmm.. interesting, I tried this myself. I could trigger the
> > "DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)" warning after a "cat
> > /proc/lockdep" with the linux-next tree, and after I applied the changes
> > I proposed above, "cat /proc/lockdep" did not trigger the same warning.
> >
> > Is this the typical "works on my machine" situation? ;-)
> >
> > Just to be clear, you mean after you added the ->lockdep_recursion pair
> > for both lockdep_count_forward_deps() and lockdep_count_backward_deps(),
> > the same splat ("DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)" still
> > got trigger, after you did a read on /proc/lockdep? Thanks!
>
> Ah, my fault. I missed the part for lockdep_count_backward_deps(). Anyway, it
> works fine here. Care to post the patch?
>

OK, I will post a patch shortly.

Regards,
Boqun

> >
> > Regards,
> > Boqun
> >
> > > >
> > > > > ---8<-----------------------
> > > > >
> > > > > diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> > > > > index 32406ef0d6a2d..d47643d8081b2 100644
> > > > > --- a/kernel/locking/lockdep.c
> > > > > +++ b/kernel/locking/lockdep.c
> > > > > @@ -1493,7 +1493,7 @@ static int __bfs(struct lock_list *source_entry,
> > > > >
> > > > > DEBUG_LOCKS_WARN_ON(!irqs_disabled());
> > > > >
> > > > > - list_for_each_entry_rcu(entry, head, entry) {
> > > > > + list_for_each_entry_rcu(entry, head, entry, true) {
> > > > > if (!lock_accessed(entry)) {
> > > > > unsigned int cq_depth;
> > > > > mark_lock_accessed(entry, lock);
> > > > >
> > > > > thanks,
> > > > >
> > > > > - Joel
> > > > >
> > > > > >
> > > > > > [26405.676199][ T3548] DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)
> > > > > > [26405.676239][ T3548] WARNING: CPU: 11 PID: 3548 at kernel/locking/lockdep.c:4637 check_flags.part.28+0x218/0x220
> > > > > > [26405.756287][ T3548] Modules linked in: kvm_intel nls_iso8859_1 nls_cp437 kvm vfat fat irqbypass intel_cstate intel_uncore intel_rapl_perf dax_pmem dax_pmem_core efivars ip_tables x_tables xfs sd_mod bnx2x hpsa mdio scsi_transport_sas firmware_class dm_mirror dm_region_hash dm_log dm_mod efivarfs
> > > > > > [26405.881899][ T3548] CPU: 11 PID: 3548 Comm: cat Not tainted 5.6.0-rc4-next-20200305+ #8
> > > > > > [26405.920091][ T3548] Hardware name: HP ProLiant BL660c Gen9, BIOS I38 10/17/2018
> > > > > > [26405.955370][ T3548] RIP: 0010:check_flags.part.28+0x218/0x220
> > > > > > [26405.983016][ T3548] Code: 13 8a e8 2b 3f 29 00 44 8b 15 84 df ba 01 45 85 d2 0f 85 c7 94 00 00 48 c7 c6 40 2b 47 89 48 c7 c7 40 04 47 89 e8 49 e3 f3 ff <0f> 0b e9 ad 94 00 00 90 55 48 89 e5 41 57 4d 89 cf 41 56 45 89 c6
> > > > > > [26406.076147][ T3548] RSP: 0018:ffffc9000695f848 EFLAGS: 00010086
> > > > > > [26406.104215][ T3548] RAX: 0000000000000000 RBX: ffff888fe6184040 RCX: ffffffff8858cecf
> > > > > > [26406.140856][ T3548] RDX: 0000000000000007 RSI: dffffc0000000000 RDI: 0000000000000000
> > > > > > [26406.178465][ T3548] RBP: ffffc9000695f850 R08: fffffbfff1377355 R09: fffffbfff1377355
> > > > > > [26406.217995][ T3548] R10: ffffffff89bb9aa3 R11: fffffbfff1377354 R12: 0000000000000000
> > > > > > [26406.256760][ T3548] R13: ffffffff8aa55ee0 R14: 0000000000000046 R15: ffffffff8aa55ec0
> > > > > > [26406.293708][ T3548] FS: 00007f58cf3a3540(0000) GS:ffff88905fa80000(0000) knlGS:0000000000000000
> > > > > > [26406.335252][ T3548] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > > > [26406.366331][ T3548] CR2: 00007f58cf326000 CR3: 0000000f1ba38006 CR4: 00000000001606e0
> > > > > > [26406.402408][ T3548] Call Trace:
> > > > > > [26406.416739][ T3548] lock_is_held_type+0x5d/0x150
> > > > > > [26406.438262][ T3548] ? rcu_lockdep_current_cpu_online+0x64/0x80
> > > > > > [26406.466463][ T3548] rcu_read_lock_any_held+0xac/0x100
> > > > > > [26406.490105][ T3548] ? rcu_read_lock_held+0xc0/0xc0
> > > > > > [26406.513258][ T3548] ? __slab_free+0x421/0x540
> > > > > > [26406.535012][ T3548] ? kasan_kmalloc+0x9/0x10
> > > > > > [26406.555901][ T3548] ? __kmalloc_node+0x1d7/0x320
> > > > > > [26406.578668][ T3548] ? kvmalloc_node+0x6f/0x80
> > > > > > [26406.599872][ T3548] __bfs+0x28a/0x3c0
> > > > > > [26406.617075][ T3548] ? class_equal+0x30/0x30
> > > > > > [26406.637524][ T3548] lockdep_count_forward_deps+0x11a/0x1a0
> > > > > > [26406.664134][ T3548] ? check_noncircular+0x2e0/0x2e0
> > > > > > [26406.688191][ T3548] ? __kasan_check_read+0x11/0x20
> > > > > > [26406.713581][ T3548] ? check_chain_key+0x1df/0x2e0
> > > > > > [26406.738044][ T3548] ? seq_vprintf+0x4e/0xb0
> > > > > > [26406.758241][ T3548] ? seq_printf+0x9b/0xd0
> > > > > > [26406.778169][ T3548] ? seq_vprintf+0xb0/0xb0
> > > > > > [26406.798172][ T3548] l_show+0x1c4/0x380
> > > > > > [26406.816474][ T3548] ? print_name+0xb0/0xb0
> > > > > > [26406.836393][ T3548] seq_read+0x56b/0x750
> > > > > > [26406.855346][ T3548] proc_reg_read+0x1b4/0x200
> > > > > > [26406.876737][ T3548] ? proc_reg_unlocked_ioctl+0x1e0/0x1e0
> > > > > > [26406.903030][ T3548] ? check_chain_key+0x1df/0x2e0
> > > > > > [26406.926531][ T3548] ? find_held_lock+0xca/0xf0
> > > > > > [26406.948291][ T3548] __vfs_read+0x50/0xa0
> > > > > > [26406.967391][ T3548] vfs_read+0xcb/0x1e0
> > > > > > [26406.986102][ T3548] ksys_read+0xc6/0x160
> > > > > > [26407.005405][ T3548] ? kernel_write+0xc0/0xc0
> > > > > > [26407.026076][ T3548] ? do_syscall_64+0x79/0xaec
> > > > > > [26407.047448][ T3548] ? do_syscall_64+0x79/0xaec
> > > > > > [26407.068650][ T3548] __x64_sys_read+0x43/0x50
> > > > > > [26407.089132][ T3548] do_syscall_64+0xcc/0xaec
> > > > > > [26407.109939][ T3548] ? trace_hardirqs_on_thunk+0x1a/0x1c
> > > > > > [26407.134924][ T3548] ? syscall_return_slowpath+0x580/0x580
> > > > > > [26407.160854][ T3548] ? entry_SYSCALL_64_after_hwframe+0x3e/0xbe
> > > > > > [26407.188943][ T3548] ? trace_hardirqs_off_caller+0x3a/0x150
> > > > > > [26407.216692][ T3548] ? trace_hardirqs_off_thunk+0x1a/0x1c
> > > > > > [26407.243534][ T3548] entry_SYSCALL_64_after_hwframe+0x49/0xbe
> > > > > > [26407.272720][ T3548] RIP: 0033:0x7f58ceeafd75
> > > > > > [26407.293162][ T3548] Code: fe ff ff 50 48 8d 3d 4a dc 09 00 e8 25 0e 02 00 0f 1f 44 00 00 f3 0f 1e fa 48 8d 05 a5 59 2d 00 8b 00 85 c0 75 0f 31 c0 0f 05 <48> 3d 00 f0 ff ff 77 53 c3 66 90 41 54 49 89 d4 55 48 89 f5 53 89
> > > > > > [26407.386043][ T3548] RSP: 002b:00007ffc115111a8 EFLAGS: 00000246 ORIG_RAX: 0000000000000000
> > > > > > [26407.425283][ T3548] RAX: ffffffffffffffda RBX: 0000000000020000 RCX: 00007f58ceeafd75
> > > > > > [26407.462717][ T3548] RDX: 0000000000020000 RSI: 00007f58cf327000 RDI: 0000000000000003
> > > > > > [26407.500428][ T3548] RBP: 00007f58cf327000 R08: 00000000ffffffff R09: 0000000000000000
> > > > > > [26407.538473][ T3548] R10: 0000000000000022 R11: 0000000000000246 R12: 00007f58cf327000
> > > > > > [26407.575743][ T3548] R13: 0000000000000003 R14: 0000000000000fff R15: 0000000000020000
> > > > > > [26407.613112][ T3548] irq event stamp: 7161
> > > > > > [26407.632089][ T3548] hardirqs last enabled at (7161): [<ffffffff88ea2684>] _raw_spin_unlock_irqrestore+0x44/0x50
> > > > > > [26407.680094][ T3548] hardirqs last disabled at (7160): [<ffffffff88ea2418>] _raw_spin_lock_irqsave+0x18/0x50
> > > > > > [26407.727273][ T3548] softirqs last enabled at (5898): [<ffffffff89200447>] __do_softirq+0x447/0x766
> > > > > > [26407.774000][ T3548] softirqs last disabled at (5889): [<ffffffff884d20e6>] irq_exit+0xd6/0xf0
> > > > > > [26407.814407][ T3548] ---[ end trace 1026d00df66af83e ]---
> > > > > > [26407.839742][ T3548] possible reason: unannotated irqs-off.
> > > > > > [26407.866243][ T3548] irq event stamp: 7161
> > > > > > [26407.885407][ T3548] hardirqs last enabled at (7161): [<ffffffff88ea2684>] _raw_spin_unlock_irqrestore+0x44/0x50
> > > > > > [26407.933602][ T3548] hardirqs last disabled at (7160): [<ffffffff88ea2418>] _raw_spin_lock_irqsave+0x18/0x50
> > > > > > [26407.980432][ T3548] softirqs last enabled at (5898): [<ffffffff89200447>] __do_softirq+0x447/0x766
> > > > > > [26408.022826][ T3548] softirqs last disabled at (5889): [<ffffffff884d20e6>] irq_exit+0xd6/0xf0
> > > > > >
> > > > > > On a side note, it likely to hit another bug in next-20200305 (not such problem on 0304) where it
> > > > > > will stuck during boot, but the reverting does not help there. Rebooting a few times could pass.
> > > > > >
> > > > > > [ 0.013514][ C0] NMI watchdog: Watchdog detected hard LOCKUP on cpu 0
> > > > > > [ 0.013514][ C0] Modules linked in:
> > > > > > [ 0.013514][ C0] irq event stamp: 64186318
> > > > > > [ 0.013514][ C0] hardirqs last enabled at (64186317): [<ffffffff84c9b107>] _raw_spin_unlock_irq+0x27/0x40
> > > > > > [ 0.013514][ C0] hardirqs last disabled at (64186318): [<ffffffff84c8f384>] __schedule+0x214/0x1070
> > > > > > [ 0.013514][ C0] softirqs last enabled at (267904): [<ffffffff85000447>] __do_softirq+0x447/0x766
> > > > > > [ 0.013514][ C0] softirqs last disabled at (267897): [<ffffffff842d1f16>] irq_exit+0xd6/0xf0
> > > > > > [ 0.013514][ C0] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.6.0-rc4-next-20200305+ #6
> > > > > > [ 0.013514][ C0] Hardware name: HP ProLiant BL660c Gen9, BIOS I38 10/17/2018
> > > > > > [ 0.013514][ C0] RIP: 0010:lock_is_held_type+0x12a/0x150
> > > > > > [ 0.013514][ C0] Code: 41 0f 94 c4 65 48 8b 1c 25 40 0f 02 00 48 8d bb 74 08 00 00 e8 77 c0 28 00 c7 83 74 08 00 00 00 00 00 00 41 56 9d 48 83 c4 18 <44> 89 e0 5b 41 5c 41 5d 41 5e 41 5f 5d c3 45 31 e4 eb c7 41 bc 01
> > > > > > [ 0.013514][ C0] RSP: 0000:ffffc9000628f9f8 EFLAGS: 00000082
> > > > > > [ 0.013514][ C0] RAX: 0000000000000000 RBX: ffff889880efc040 RCX: ffffffff8438b449
> > > > > > [ 0.013514][ C0] RDX: 0000000000000007 RSI: dffffc0000000000 RDI: ffff889880efc8b4
> > > > > > [ 0.013514][ C0] RBP: ffffc9000628fa20 R08: ffffed1108588a24 R09: ffffed1108588a24
> > > > > > [ 0.013514][ C0] R10: ffff888842c4511b R11: 0000000000000000 R12: 0000000000000000
> > > > > > [ 0.013514][ C0] R13: ffff889880efc908 R14: 0000000000000046 R15: 0000000000000003
> > > > > > [ 0.013514][ C0] FS: 0000000000000000(0000) GS:ffff888842c00000(0000) knlGS:0000000000000000
> > > > > > [ 0.013514][ C0] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > > > [ 0.013514][ C0] CR2: ffff88a0707ff000 CR3: 0000000b72012001 CR4: 00000000001606f0
> > > > > > [ 0.013514][ C0] Call Trace:
> > > > > > [ 0.013514][ C0] rcu_read_lock_sched_held+0xac/0xe0
> > > > > > lock_is_held at include/linux/lockdep.h:361
> > > > > > (inlined by) rcu_read_lock_sched_held at kernel/rcu/update.c:121
> > > > > > [ 0.013514][ C0] ? rcu_read_lock_bh_held+0xc0/0xc0
> > > > > > [ 0.013514][ C0] rcu_note_context_switcx186/0x3b0
> > > > > > [ 0.013514][ C0] __schedule+0x21f/0x1070
> > > > > > [ 0.013514][ C0] ? __sched_text_start+0x8/0x8
> > > > > > [ 0.013514][ C0] schedule+0x95/0x160
> > > > > > [ 0.013514][ C0] do_boot_cpu+0x58c/0xaf0
> > > > > > [ 0.013514][ C0] native_cpu_up+0x298/0x430
> > > > > > [ 0.013514][ C0] ? common_cpu_up+0x150/0x150
> > > > > > [ 0.013514][ C0] bringup_cpu+0x44/0x310
> > > > > > [ 0.013514][ C0] ? timers_prepare_cpu+0x114/0x190
> > > > > > [ 0.013514][ C0] ? takedown_cpu+0x2e0/0x2e0
> > > > > > [ 0.013514][ C0] cpuhp_invoke_callback+0x197/0x1120
> > > > > > [ 0.013514][ C0] ? ring_buffer_record_is_set_on+0x40/0x40
> > > > > > [ 0.013514][ C0] _cpu_up+0x171/0x280
> > > > > > [ 0.013514][ C0] do_cpu_up+0xb1/0x120
> > > > > > [ 0.013514][ C0] cpu_up+0x13/0x20
> > > > > > [ 0.013514][ C0] smp_init+0x91/0x118
> > > > > > [ 0.013514][ C0] kernel_init_freeable+0x221/0x4f8
> > > > > > [ 0.013514][ C0] ? mark_held_locks+0x34/0xb0
> > > > > > [ 0.013514][ C0] ? _raw_spin_unlock_irq+0x27/0x40
> > > > > > [ 0.013514][ C0] ? start_kernel+0x876/0x876
> > > > > > [ 0.013514][ C0] ? lockdep_hardirqs_on+0x1b0/0x2a0
> > > > > > [ 0.013514][ C0] ? _raw_spin_unlock_irq+0x27/0x40
> > > > > > [ 0.013514][ C0] ? rest_init+0x307/0x307
> > > > > > [ 0.013514][ C0] kernel_init+0x 0.013514][ C0] ? rest_init+0x307/0x307
> > > > > > [ 0.013514][ C0] ret_from_fork+0x3a/0x50
> > > > > >
> > > > > >
> > > > >
> > > > >

Next message: Dave Chinner: "Re: [PATCH V5 00/12] Enable per-file/per-directory DAX operations V5"
Previous message: Ashish Kalra: "Re: [PATCH v2] swiotlb: Adjust SWIOTBL bounce buffer size for SEV guests."
In reply to: Qian Cai: "Re: PROVE_RCU_LIST + /proc/lockdep warning"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]