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

[Nicholas Piggin]

On Wed, 7 Feb 2018 14:51:56 +0100
"Rafael J. Wysocki" <rafael.j.wysocki@xxxxxxxxx> 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] [c00000023ff77650] [c000000000940b50]
> > .dump_stack+0xac/0xfc (unreliable)
> > [15002.854285] [c00000023ff776e0] [c00000000026c678] .warn_alloc+0xe8/0x180
> > [15002.854376] [c00000023ff777a0] [c00000000026d50c]
> > .__alloc_pages_nodemask+0xd6c/0xf90
> > [15002.854490] [c00000023ff77980] [c0000000002e9cc0]
> > .alloc_pages_current+0x90/0x120
> > [15002.854624] [c00000023ff77a10] [c0000000007990cc]
> > .skb_page_frag_refill+0x8c/0x120
> > [15002.854746] [c00000023ff77aa0] [d000000003a561a8]
> > .try_fill_recv+0x368/0x620 [virtio_net]
> > [15003.422855] [c00000023ff77ba0] [d000000003a568ec]
> > .virtnet_poll+0x25c/0x380 [virtio_net]
> > [15003.423864] [c00000023ff77c70] [c0000000007c18d0] .net_rx_action+0x330/0x4a0
> > [15003.424024] [c00000023ff77d90] [c000000000960d50] .__do_softirq+0x150/0x3a8
> > [15003.424197] [c00000023ff77e90] [c0000000000ff608] .irq_exit+0x198/0x1b0
> > [15003.424342] [c00000023ff77f10] [c000000000015504] .__do_irq+0x94/0x1f0
> > [15003.424485] [c00000023ff77f90] [c000000000026d5c] .call_do_irq+0x14/0x24
> > [15003.424627] [c00000023bc63820] [c0000000000156ec] .do_IRQ+0x8c/0x100
> > [15003.424776] [c00000023bc638c0] [c000000000008b34]
> > hardware_interrupt_common+0x114/0x120
> > [15003.424963] --- interrupt: 501 at .snooze_loop+0xa4/0x1c0
> >      LR = .snooze_loop+0x60/0x1c0
> > [15003.425164] [c00000023bc63bb0] [c00000023bc63c50]
> > 0xc00000023bc63c50 (unreliable)
> > [15003.425346] [c00000023bc63c30] [c00000000075104c]
> > .cpuidle_enter_state+0xac/0x390
> > [15003.425534] [c00000023bc63ce0] [c000000000157adc] .call_cpuidle+0x3c/0x70
> > [15003.425669] [c00000023bc63d50] [c000000000157e90] .do_idle+0x2a0/0x300
> > [15003.425815] [c00000023bc63e20] [c0000000001580ac]
> > .cpu_startup_entry+0x2c/0x40
> > [15003.425995] [c00000023bc63ea0] [c000000000045790]
> > .start_secondary+0x4d0/0x520
> > [15003.426170] [c00000023bc63f90] [c00000000000aa70]
> > 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