On 2016-11-09 14:25:39 [+0800], kernel test robot wrote:
>
> FYI, we noticed the following commit:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
> commit a7410f28ceb566bda840b4afc278747c63383fb6 ("rcu: update: Make
> RCU_EXPEDITE_BOOT be the default")
>
> in testcase: boot
>
> on test machine: qemu-system-x86_64 -enable-kvm -cpu host -smp 2 -m 1G
>
> caused below changes:
>
>
> +-----------------------------------------------------------------------------------+------------+------------+
> |
> | 9aa96e24ec | a7410f28ce |
> +-----------------------------------------------------------------------------------+------------+------------+
> | boot_successes
> | 3 | 7 |
> | boot_failures
> | 17 | 17 |
> | INFO:rcu_sched_self-detected_stall_on_CPU
> | 17 | 14 |
> | calltrace:mark_rodata_ro
> | 17 | 14 |
> | INFO:rcu_sched_detected_stalls_on_CPUs/tasks
> | 15 | 14 |
> |
> INFO:rcu_sched_detected_expedited_stalls_on_CPUs/tasks:{#-...}#jiffies_s:#root:#/
> | 0 | 15 |
> +-----------------------------------------------------------------------------------+------------+------------+
Is it correct to assume that there are RCU stalls with and without my
patch but with my patch there are also "expedited" stalls?
> [ 38.504100] Write protecting the kernel read-only data: 47104k
> [ 38.511102] Freeing unused kernel memory: 720K (ffff88000314c000 -
> ffff880003200000)
> [ 38.547480] Freeing unused kernel memory: 1032K (ffff880003cfe000 -
> ffff880003e00000)
> [ 149.710631] INFO: rcu_sched detected expedited stalls on CPUs/tasks: {
> 0-... } 25089 jiffies s: 9 root: 0x1/.
> [ 149.720353] blocking rcu_node structures:
The other testing robot reported something similar against rcu/next.
With my patch I get:
[ 14.260226] Freeing unused kernel memory: 536K (ffff88000277a000 -
ffff880002800000)
[ 14.263114] Freeing unused kernel memory: 252K (ffff8800031c1000 -
ffff880003200000)
[ 35.256735] INFO: rcu_sched self-detected stall on CPU
[ 35.257565] 1-...: (6299 ticks this GP) idle=0e2/140000000000001/0
softirq=90/90 fqs=2100 last_accelerate: 0000/c973, nonlazy_posted: 0, .D
[ 35.259410] (t=6300 jiffies g=-252 c=-253 q=0)
[ 35.260106] Task dump for CPU 1:
[ 35.260608] swapper/0 R running task 0 1 0 0x00000008
[ 35.261698] ffff880036107cc0 ffffffff81181de2 0000000000021780
ffff880036121780
[ 35.262835] 0000000000000001 ffffffff832d7fc0 ffff880036107ce0
ffffffff8118616d
[ 35.263995] ffffffff832d7fc0 0000000000000001 ffff880036107d30
ffffffff8128b3b8
[ 35.265131] Call Trace:
[ 35.265512] <IRQ> [ 35.265826] [<ffffffff81181de2>]
sched_show_task+0x102/0x180
[ 35.266701] [<ffffffff8118616d>] dump_cpu_task+0x4d/0x60
[ 35.267526] [<ffffffff8128b3b8>] rcu_dump_cpu_stacks+0xee/0x11b
…
[ 35.289870] INFO: rcu_sched detected stalls on CPUs/tasks:
[ 35.290500] 1-...: (6301 ticks this GP) idle=0e2/140000000000000/0
softirq=90/90 fqs=2100 last_accelerate: 0000/c97d, nonlazy_posted: 0, .D
[ 35.291834] (detected by 0, t=6310 jiffies, g=-252, c=-253, q=0)
[ 35.292505] Task dump for CPU 1:
…
[ 44.260103] INFO: rcu_sched detected expedited stalls on CPUs/tasks: { 1-...
} 6784 jiffies s: 299 root: 0x2/.
[ 44.261911] blocking rcu_node structures:
[ 44.262627] Task dump for CPU 1:
[ 44.263177] swapper/0 R running task 0 1 0 0x00000008
[ 44.264379] 0000000041b58ab3 ffffffff81249eab ffffffff81b6f150
ffffffff819468d7
[ 44.265640] 0000000041b58ab3 ffffffff830f76b0 0000000000000297
ffffffff83d87aa0
[ 44.266915] 000000000000c6e0 0000000000000058 0000000000000000
ffff880035547be8
[ 44.268153] Call Trace:
[ 44.268603] [<ffffffff81249eab>] ? trace_hardirqs_on_caller+0xdb/0x1e0
[ 44.269708] [<ffffffff81b6f150>] ? serial8250_set_defaults+0x1a0/0x1a0
and without it:
[ 15.837956] Freeing unused kernel memory: 536K (ffff88000277a000 -
ffff880002800000)
[ 15.841396] Freeing unused kernel memory: 252K (ffff8800031c1000 -
ffff880003200000)
[ 43.070453] INFO: rcu_sched self-detected stall on CPU
[ 43.071215] 1-...: (6299 ticks this GP) idle=daa/140000000000001/0
softirq=211/211 fqs=2100 last_accelerate: 0000/d29c, nonlazy_posted: 0, .D
[ 43.072807] (t=6300 jiffies g=-91 c=-92 q=0)
[ 43.073385] Task dump for CPU 1:
[ 43.073800] swapper/0 R running task 0 1 0 0x00000008
[ 43.074734] ffff880036107cc0 ffffffff81181de2 0000000000021780
ffff880036121780
[ 43.075715] 0000000000000001 ffffffff832d7f80 ffff880036107ce0
ffffffff8118616d
[ 43.076691] ffffffff832d7f80 0000000000000001 ffff880036107d30
ffffffff8128b3b8
[ 43.077668] Call Trace:
[ 43.077983] <IRQ> [ 43.078265] [<ffffffff81181de2>]
sched_show_task+0x102/0x180
[ 43.079006] [<ffffffff8118616d>] dump_cpu_task+0x4d/0x60
…
[ 43.097862] INFO: rcu_sched detected stalls on CPUs/tasks:
[ 43.098442] 1-...: (6300 ticks this GP) idle=daa/140000000000000/0
softirq=211/211 fqs=2100 last_accelerate: 0000/d2a4, nonlazy_posted: 0, .D
[ 43.099613] (detected by 0, t=6308 jiffies, g=-91, c=-92, q=0)
[ 43.100180] Task dump for CPU 1:
[ 43.100487] swapper/0 R running task 0 1 0 0x00000008
[ 43.101171] ffffffff83d86c40 1ffff10006aa8f67 0000000500000058
0000000041b58ab3
[ 43.101912] ffffffff8316ae78 ffffffff81b6f150 ffffffff819468d7
0000000041b58ab3
[ 43.102657] ffffffff830f76b0 ffffffff81946830 ffff103000000020
ffff880000000020
[ 43.103395] Call Trace:
[ 43.103647] [<ffffffff81b6f150>] ? serial8250_set_defaults+0x1a0/0x1a0
[ 43.104272] [<ffffffff819468d7>] ? sprintf+0xa7/0xd0
So it looks to me that the problem has been always there but the
additional "expedited" part in the INFO message was the reason to create
the report.
Sebastian
