Re: linux/drivers/cpuidle: cpuidle_enter_state() issue

2018-02-07 Thread Nicholas Piggin 
On Wed, 7 Feb 2018 14:51:56 +0100
"Rafael J. Wysocki"  wrote:

> On 2/7/2018 7:59 AM, Li Wang wrote:
> > Hi Kernel-developers,
> >
> > The flowing call trace was catch from kernel-v4.15, could anyone help
> > to analysis the cpuidle problem?
> > or, if you need any more detail info pls let me know.
> >
> > Test Env:
> > IBM KVM Guest on ibm-p8-kvm-03
> > POWER8E (raw), altivec supported
> > 9216 MB memory, 107 GB disk space
> >
> > 8<
> > [15002.722413] swapper/15: page allocation failure: order:0,
> > mode:0x1080020(GFP_ATOMIC), nodemask=(null)
> > [15002.853793] swapper/15 cpuset=/ mems_allowed=0
> > [15002.853932] CPU: 15 PID: 0 Comm: swapper/15 Not tainted 4.15.0 #1
> > [15002.854019] Call Trace:
> > [15002.854129] [c0023ff77650] [c0940b50]
> > .dump_stack+0xac/0xfc (unreliable)
> > [15002.854285] [c0023ff776e0] [c026c678] .warn_alloc+0xe8/0x180
> > [15002.854376] [c0023ff777a0] [c026d50c]
> > .__alloc_pages_nodemask+0xd6c/0xf90
> > [15002.854490] [c0023ff77980] [c02e9cc0]
> > .alloc_pages_current+0x90/0x120
> > [15002.854624] [c0023ff77a10] [c07990cc]
> > .skb_page_frag_refill+0x8c/0x120
> > [15002.854746] [c0023ff77aa0] [d3a561a8]
> > .try_fill_recv+0x368/0x620 [virtio_net]
> > [15003.422855] [c0023ff77ba0] [d3a568ec]
> > .virtnet_poll+0x25c/0x380 [virtio_net]
> > [15003.423864] [c0023ff77c70] [c07c18d0] 
> > .net_rx_action+0x330/0x4a0
> > [15003.424024] [c0023ff77d90] [c0960d50] 
> > .__do_softirq+0x150/0x3a8
> > [15003.424197] [c0023ff77e90] [c00ff608] .irq_exit+0x198/0x1b0
> > [15003.424342] [c0023ff77f10] [c0015504] .__do_irq+0x94/0x1f0
> > [15003.424485] [c0023ff77f90] [c0026d5c] .call_do_irq+0x14/0x24
> > [15003.424627] [c0023bc63820] [c00156ec] .do_IRQ+0x8c/0x100
> > [15003.424776] [c0023bc638c0] [c0008b34]
> > hardware_interrupt_common+0x114/0x120
> > [15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
> >  LR = .snooze_loop+0x60/0x1c0
> > [15003.425164] [c0023bc63bb0] [c0023bc63c50]
> > 0xc0023bc63c50 (unreliable)
> > [15003.425346] [c0023bc63c30] [c075104c]
> > .cpuidle_enter_state+0xac/0x390
> > [15003.425534] [c0023bc63ce0] [c0157adc] .call_cpuidle+0x3c/0x70
> > [15003.425669] [c0023bc63d50] [c0157e90] .do_idle+0x2a0/0x300
> > [15003.425815] [c0023bc63e20] [c01580ac]
> > .cpu_startup_entry+0x2c/0x40
> > [15003.425995] [c0023bc63ea0] [c0045790]
> > .start_secondary+0x4d0/0x520
> > [15003.426170] [c0023bc63f90] [c000aa70]
> > start_secondary_prolog+0x10/0x14
> > -8<---
> >
> > Any response will be appreciated!
> >  
> 
> I'm not sure if this is a cpuidle problem in the first place.
> 
> It looks like the kernel was unable to allocate a page for something and 
> it was trying to do an atomic allocation, so that's not guaranteed to be 
> successful every time even though the failure here may be unexpected.

Yes you're right. An order-0 failure should be quite rare, but not impossible.
This backtrace does not actually indicate a bug, but it's something to help
kernel developers tune memory allocation.

Is there any negative impact to the running system? Do the messages occur
often?

Thanks,
Nick

Re: linux/drivers/cpuidle: cpuidle_enter_state() issue

2018-02-07 Thread Nicholas Piggin 
On Wed, 7 Feb 2018 14:51:56 +0100
"Rafael J. Wysocki"  wrote:

> On 2/7/2018 7:59 AM, Li Wang wrote:
> > Hi Kernel-developers,
> >
> > The flowing call trace was catch from kernel-v4.15, could anyone help
> > to analysis the cpuidle problem?
> > or, if you need any more detail info pls let me know.
> >
> > Test Env:
> > IBM KVM Guest on ibm-p8-kvm-03
> > POWER8E (raw), altivec supported
> > 9216 MB memory, 107 GB disk space
> >
> > 8<
> > [15002.722413] swapper/15: page allocation failure: order:0,
> > mode:0x1080020(GFP_ATOMIC), nodemask=(null)
> > [15002.853793] swapper/15 cpuset=/ mems_allowed=0
> > [15002.853932] CPU: 15 PID: 0 Comm: swapper/15 Not tainted 4.15.0 #1
> > [15002.854019] Call Trace:
> > [15002.854129] [c0023ff77650] [c0940b50]
> > .dump_stack+0xac/0xfc (unreliable)
> > [15002.854285] [c0023ff776e0] [c026c678] .warn_alloc+0xe8/0x180
> > [15002.854376] [c0023ff777a0] [c026d50c]
> > .__alloc_pages_nodemask+0xd6c/0xf90
> > [15002.854490] [c0023ff77980] [c02e9cc0]
> > .alloc_pages_current+0x90/0x120
> > [15002.854624] [c0023ff77a10] [c07990cc]
> > .skb_page_frag_refill+0x8c/0x120
> > [15002.854746] [c0023ff77aa0] [d3a561a8]
> > .try_fill_recv+0x368/0x620 [virtio_net]
> > [15003.422855] [c0023ff77ba0] [d3a568ec]
> > .virtnet_poll+0x25c/0x380 [virtio_net]
> > [15003.423864] [c0023ff77c70] [c07c18d0] 
> > .net_rx_action+0x330/0x4a0
> > [15003.424024] [c0023ff77d90] [c0960d50] 
> > .__do_softirq+0x150/0x3a8
> > [15003.424197] [c0023ff77e90] [c00ff608] .irq_exit+0x198/0x1b0
> > [15003.424342] [c0023ff77f10] [c0015504] .__do_irq+0x94/0x1f0
> > [15003.424485] [c0023ff77f90] [c0026d5c] .call_do_irq+0x14/0x24
> > [15003.424627] [c0023bc63820] [c00156ec] .do_IRQ+0x8c/0x100
> > [15003.424776] [c0023bc638c0] [c0008b34]
> > hardware_interrupt_common+0x114/0x120
> > [15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
> >  LR = .snooze_loop+0x60/0x1c0
> > [15003.425164] [c0023bc63bb0] [c0023bc63c50]
> > 0xc0023bc63c50 (unreliable)
> > [15003.425346] [c0023bc63c30] [c075104c]
> > .cpuidle_enter_state+0xac/0x390
> > [15003.425534] [c0023bc63ce0] [c0157adc] .call_cpuidle+0x3c/0x70
> > [15003.425669] [c0023bc63d50] [c0157e90] .do_idle+0x2a0/0x300
> > [15003.425815] [c0023bc63e20] [c01580ac]
> > .cpu_startup_entry+0x2c/0x40
> > [15003.425995] [c0023bc63ea0] [c0045790]
> > .start_secondary+0x4d0/0x520
> > [15003.426170] [c0023bc63f90] [c000aa70]
> > start_secondary_prolog+0x10/0x14
> > -8<---
> >
> > Any response will be appreciated!
> >  
> 
> I'm not sure if this is a cpuidle problem in the first place.
> 
> It looks like the kernel was unable to allocate a page for something and 
> it was trying to do an atomic allocation, so that's not guaranteed to be 
> successful every time even though the failure here may be unexpected.

Yes you're right. An order-0 failure should be quite rare, but not impossible.
This backtrace does not actually indicate a bug, but it's something to help
kernel developers tune memory allocation.

Is there any negative impact to the running system? Do the messages occur
often?

Thanks,
Nick

Re: linux/drivers/cpuidle: cpuidle_enter_state() issue

2018-02-07 Thread Rafael J. Wysocki 

On 2/7/2018 7:59 AM, Li Wang wrote:

Hi Kernel-developers,

The flowing call trace was catch from kernel-v4.15, could anyone help
to analysis the cpuidle problem?
or, if you need any more detail info pls let me know.

Test Env:
IBM KVM Guest on ibm-p8-kvm-03
POWER8E (raw), altivec supported
9216 MB memory, 107 GB disk space

8<
[15002.722413] swapper/15: page allocation failure: order:0,
mode:0x1080020(GFP_ATOMIC), nodemask=(null)
[15002.853793] swapper/15 cpuset=/ mems_allowed=0
[15002.853932] CPU: 15 PID: 0 Comm: swapper/15 Not tainted 4.15.0 #1
[15002.854019] Call Trace:
[15002.854129] [c0023ff77650] [c0940b50]
.dump_stack+0xac/0xfc (unreliable)
[15002.854285] [c0023ff776e0] [c026c678] .warn_alloc+0xe8/0x180
[15002.854376] [c0023ff777a0] [c026d50c]
.__alloc_pages_nodemask+0xd6c/0xf90
[15002.854490] [c0023ff77980] [c02e9cc0]
.alloc_pages_current+0x90/0x120
[15002.854624] [c0023ff77a10] [c07990cc]
.skb_page_frag_refill+0x8c/0x120
[15002.854746] [c0023ff77aa0] [d3a561a8]
.try_fill_recv+0x368/0x620 [virtio_net]
[15003.422855] [c0023ff77ba0] [d3a568ec]
.virtnet_poll+0x25c/0x380 [virtio_net]
[15003.423864] [c0023ff77c70] [c07c18d0] .net_rx_action+0x330/0x4a0
[15003.424024] [c0023ff77d90] [c0960d50] .__do_softirq+0x150/0x3a8
[15003.424197] [c0023ff77e90] [c00ff608] .irq_exit+0x198/0x1b0
[15003.424342] [c0023ff77f10] [c0015504] .__do_irq+0x94/0x1f0
[15003.424485] [c0023ff77f90] [c0026d5c] .call_do_irq+0x14/0x24
[15003.424627] [c0023bc63820] [c00156ec] .do_IRQ+0x8c/0x100
[15003.424776] [c0023bc638c0] [c0008b34]
hardware_interrupt_common+0x114/0x120
[15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
 LR = .snooze_loop+0x60/0x1c0
[15003.425164] [c0023bc63bb0] [c0023bc63c50]
0xc0023bc63c50 (unreliable)
[15003.425346] [c0023bc63c30] [c075104c]
.cpuidle_enter_state+0xac/0x390
[15003.425534] [c0023bc63ce0] [c0157adc] .call_cpuidle+0x3c/0x70
[15003.425669] [c0023bc63d50] [c0157e90] .do_idle+0x2a0/0x300
[15003.425815] [c0023bc63e20] [c01580ac]
.cpu_startup_entry+0x2c/0x40
[15003.425995] [c0023bc63ea0] [c0045790]
.start_secondary+0x4d0/0x520
[15003.426170] [c0023bc63f90] [c000aa70]
start_secondary_prolog+0x10/0x14
-8<---

Any response will be appreciated!



I'm not sure if this is a cpuidle problem in the first place.

It looks like the kernel was unable to allocate a page for something and 
it was trying to do an atomic allocation, so that's not guaranteed to be 
successful every time even though the failure here may be unexpected.


Thanks,

Rafael

Re: linux/drivers/cpuidle: cpuidle_enter_state() issue

2018-02-07 Thread Rafael J. Wysocki 

On 2/7/2018 7:59 AM, Li Wang wrote:

Hi Kernel-developers,

The flowing call trace was catch from kernel-v4.15, could anyone help
to analysis the cpuidle problem?
or, if you need any more detail info pls let me know.

Test Env:
IBM KVM Guest on ibm-p8-kvm-03
POWER8E (raw), altivec supported
9216 MB memory, 107 GB disk space

8<
[15002.722413] swapper/15: page allocation failure: order:0,
mode:0x1080020(GFP_ATOMIC), nodemask=(null)
[15002.853793] swapper/15 cpuset=/ mems_allowed=0
[15002.853932] CPU: 15 PID: 0 Comm: swapper/15 Not tainted 4.15.0 #1
[15002.854019] Call Trace:
[15002.854129] [c0023ff77650] [c0940b50]
.dump_stack+0xac/0xfc (unreliable)
[15002.854285] [c0023ff776e0] [c026c678] .warn_alloc+0xe8/0x180
[15002.854376] [c0023ff777a0] [c026d50c]
.__alloc_pages_nodemask+0xd6c/0xf90
[15002.854490] [c0023ff77980] [c02e9cc0]
.alloc_pages_current+0x90/0x120
[15002.854624] [c0023ff77a10] [c07990cc]
.skb_page_frag_refill+0x8c/0x120
[15002.854746] [c0023ff77aa0] [d3a561a8]
.try_fill_recv+0x368/0x620 [virtio_net]
[15003.422855] [c0023ff77ba0] [d3a568ec]
.virtnet_poll+0x25c/0x380 [virtio_net]
[15003.423864] [c0023ff77c70] [c07c18d0] .net_rx_action+0x330/0x4a0
[15003.424024] [c0023ff77d90] [c0960d50] .__do_softirq+0x150/0x3a8
[15003.424197] [c0023ff77e90] [c00ff608] .irq_exit+0x198/0x1b0
[15003.424342] [c0023ff77f10] [c0015504] .__do_irq+0x94/0x1f0
[15003.424485] [c0023ff77f90] [c0026d5c] .call_do_irq+0x14/0x24
[15003.424627] [c0023bc63820] [c00156ec] .do_IRQ+0x8c/0x100
[15003.424776] [c0023bc638c0] [c0008b34]
hardware_interrupt_common+0x114/0x120
[15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
 LR = .snooze_loop+0x60/0x1c0
[15003.425164] [c0023bc63bb0] [c0023bc63c50]
0xc0023bc63c50 (unreliable)
[15003.425346] [c0023bc63c30] [c075104c]
.cpuidle_enter_state+0xac/0x390
[15003.425534] [c0023bc63ce0] [c0157adc] .call_cpuidle+0x3c/0x70
[15003.425669] [c0023bc63d50] [c0157e90] .do_idle+0x2a0/0x300
[15003.425815] [c0023bc63e20] [c01580ac]
.cpu_startup_entry+0x2c/0x40
[15003.425995] [c0023bc63ea0] [c0045790]
.start_secondary+0x4d0/0x520
[15003.426170] [c0023bc63f90] [c000aa70]
start_secondary_prolog+0x10/0x14
-8<---

Any response will be appreciated!



I'm not sure if this is a cpuidle problem in the first place.

It looks like the kernel was unable to allocate a page for something and 
it was trying to do an atomic allocation, so that's not guaranteed to be 
successful every time even though the failure here may be unexpected.


Thanks,

Rafael

linux/drivers/cpuidle: cpuidle_enter_state() issue

2018-02-06 Thread Li Wang 
Hi Kernel-developers,

The flowing call trace was catch from kernel-v4.15, could anyone help
to analysis the cpuidle problem?
or, if you need any more detail info pls let me know.

Test Env:
IBM KVM Guest on ibm-p8-kvm-03
POWER8E (raw), altivec supported
9216 MB memory, 107 GB disk space

8<
[15002.722413] swapper/15: page allocation failure: order:0,
mode:0x1080020(GFP_ATOMIC), nodemask=(null)
[15002.853793] swapper/15 cpuset=/ mems_allowed=0
[15002.853932] CPU: 15 PID: 0 Comm: swapper/15 Not tainted 4.15.0 #1
[15002.854019] Call Trace:
[15002.854129] [c0023ff77650] [c0940b50]
.dump_stack+0xac/0xfc (unreliable)
[15002.854285] [c0023ff776e0] [c026c678] .warn_alloc+0xe8/0x180
[15002.854376] [c0023ff777a0] [c026d50c]
.__alloc_pages_nodemask+0xd6c/0xf90
[15002.854490] [c0023ff77980] [c02e9cc0]
.alloc_pages_current+0x90/0x120
[15002.854624] [c0023ff77a10] [c07990cc]
.skb_page_frag_refill+0x8c/0x120
[15002.854746] [c0023ff77aa0] [d3a561a8]
.try_fill_recv+0x368/0x620 [virtio_net]
[15003.422855] [c0023ff77ba0] [d3a568ec]
.virtnet_poll+0x25c/0x380 [virtio_net]
[15003.423864] [c0023ff77c70] [c07c18d0] .net_rx_action+0x330/0x4a0
[15003.424024] [c0023ff77d90] [c0960d50] .__do_softirq+0x150/0x3a8
[15003.424197] [c0023ff77e90] [c00ff608] .irq_exit+0x198/0x1b0
[15003.424342] [c0023ff77f10] [c0015504] .__do_irq+0x94/0x1f0
[15003.424485] [c0023ff77f90] [c0026d5c] .call_do_irq+0x14/0x24
[15003.424627] [c0023bc63820] [c00156ec] .do_IRQ+0x8c/0x100
[15003.424776] [c0023bc638c0] [c0008b34]
hardware_interrupt_common+0x114/0x120
[15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
LR = .snooze_loop+0x60/0x1c0
[15003.425164] [c0023bc63bb0] [c0023bc63c50]
0xc0023bc63c50 (unreliable)
[15003.425346] [c0023bc63c30] [c075104c]
.cpuidle_enter_state+0xac/0x390
[15003.425534] [c0023bc63ce0] [c0157adc] .call_cpuidle+0x3c/0x70
[15003.425669] [c0023bc63d50] [c0157e90] .do_idle+0x2a0/0x300
[15003.425815] [c0023bc63e20] [c01580ac]
.cpu_startup_entry+0x2c/0x40
[15003.425995] [c0023bc63ea0] [c0045790]
.start_secondary+0x4d0/0x520
[15003.426170] [c0023bc63f90] [c000aa70]
start_secondary_prolog+0x10/0x14
-8<---

Any response will be appreciated!

-- 
Regards,
Li Wang
Email: wangli.a...@gmail.com

linux/drivers/cpuidle: cpuidle_enter_state() issue

2018-02-06 Thread Li Wang 
Hi Kernel-developers,

The flowing call trace was catch from kernel-v4.15, could anyone help
to analysis the cpuidle problem?
or, if you need any more detail info pls let me know.

Test Env:
IBM KVM Guest on ibm-p8-kvm-03
POWER8E (raw), altivec supported
9216 MB memory, 107 GB disk space

8<
[15002.722413] swapper/15: page allocation failure: order:0,
mode:0x1080020(GFP_ATOMIC), nodemask=(null)
[15002.853793] swapper/15 cpuset=/ mems_allowed=0
[15002.853932] CPU: 15 PID: 0 Comm: swapper/15 Not tainted 4.15.0 #1
[15002.854019] Call Trace:
[15002.854129] [c0023ff77650] [c0940b50]
.dump_stack+0xac/0xfc (unreliable)
[15002.854285] [c0023ff776e0] [c026c678] .warn_alloc+0xe8/0x180
[15002.854376] [c0023ff777a0] [c026d50c]
.__alloc_pages_nodemask+0xd6c/0xf90
[15002.854490] [c0023ff77980] [c02e9cc0]
.alloc_pages_current+0x90/0x120
[15002.854624] [c0023ff77a10] [c07990cc]
.skb_page_frag_refill+0x8c/0x120
[15002.854746] [c0023ff77aa0] [d3a561a8]
.try_fill_recv+0x368/0x620 [virtio_net]
[15003.422855] [c0023ff77ba0] [d3a568ec]
.virtnet_poll+0x25c/0x380 [virtio_net]
[15003.423864] [c0023ff77c70] [c07c18d0] .net_rx_action+0x330/0x4a0
[15003.424024] [c0023ff77d90] [c0960d50] .__do_softirq+0x150/0x3a8
[15003.424197] [c0023ff77e90] [c00ff608] .irq_exit+0x198/0x1b0
[15003.424342] [c0023ff77f10] [c0015504] .__do_irq+0x94/0x1f0
[15003.424485] [c0023ff77f90] [c0026d5c] .call_do_irq+0x14/0x24
[15003.424627] [c0023bc63820] [c00156ec] .do_IRQ+0x8c/0x100
[15003.424776] [c0023bc638c0] [c0008b34]
hardware_interrupt_common+0x114/0x120
[15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
LR = .snooze_loop+0x60/0x1c0
[15003.425164] [c0023bc63bb0] [c0023bc63c50]
0xc0023bc63c50 (unreliable)
[15003.425346] [c0023bc63c30] [c075104c]
.cpuidle_enter_state+0xac/0x390
[15003.425534] [c0023bc63ce0] [c0157adc] .call_cpuidle+0x3c/0x70
[15003.425669] [c0023bc63d50] [c0157e90] .do_idle+0x2a0/0x300
[15003.425815] [c0023bc63e20] [c01580ac]
.cpu_startup_entry+0x2c/0x40
[15003.425995] [c0023bc63ea0] [c0045790]
.start_secondary+0x4d0/0x520
[15003.426170] [c0023bc63f90] [c000aa70]
start_secondary_prolog+0x10/0x14
-8<---

Any response will be appreciated!

-- 
Regards,
Li Wang
Email: wangli.a...@gmail.com