Re: dd: page allocation failure: order:0, mode:0x1080020(GFP_ATOMIC), nodemask=(null)

[Fengguang Wu <fengguang.wu@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1936 bytes --]

On Thu, Nov 30, 2017 at 10:08:04PM +0800, Fengguang Wu wrote:
>>>> [   78.848629] dd: page allocation failure: order:0, mode:0x1080020(GFP_ATOMIC), nodemask=(null)
>>>> [   78.857841] dd cpuset=/ mems_allowed=0-1
>>>> [   78.862502] CPU: 0 PID: 6131 Comm: dd Tainted: G           O     4.15.0-rc1 #1
>>>> [   78.870437] Call Trace:
>>>> [   78.873610]  <IRQ>
>>>> [   78.876342]  dump_stack+0x5c/0x7b
>>>> [   78.880414]  warn_alloc+0xbe/0x150
>>>> [   78.884550]  __alloc_pages_slowpath+0xda7/0xdf0
>>>> [   78.889822]  ? xhci_urb_enqueue+0x23d/0x580
>>>> [   78.894713]  __alloc_pages_nodemask+0x269/0x280
>>>> [   78.899891]  page_frag_alloc+0x11c/0x150
>>>> [   78.904471]  __netdev_alloc_skb+0xa0/0x110
>>>> [   78.909277]  rx_submit+0x3b/0x2e0
>>>> [   78.913256]  rx_complete+0x196/0x2d0
>>>> [   78.917560]  __usb_hcd_giveback_urb+0x86/0x100
>>>> [   78.922681]  xhci_giveback_urb_in_irq+0x86/0x100
>>>> [   78.928769]  ? ip_rcv+0x261/0x390
>>>> [   78.932739]  xhci_td_cleanup+0xe7/0x170
>>>> [   78.937308]  handle_tx_event+0x297/0x1190
>>>> [   78.941990]  xhci_irq+0x300/0xb80
>>>> [   78.945968]  ? pciehp_isr+0x46/0x320
>>>> [   78.950870]  __handle_irq_event_percpu+0x3a/0x1a0
>>>> [   78.956311]  handle_irq_event_percpu+0x20/0x50
>>>> [   78.961466]  handle_irq_event+0x3d/0x60
>>>> [   78.965962]  handle_edge_irq+0x71/0x190
>>>> [   78.970480]  handle_irq+0xa5/0x100
>>>> [   78.974565]  do_IRQ+0x41/0xc0
>>>> [   78.978206]  ? pagevec_move_tail_fn+0x350/0x350
>>>> [   78.983412]  common_interrupt+0x96/0x96
>>>
>>>Unfortunatelly we are missing the most imporatant information, the
>>>meminfo. We cannot tell much without it. Maybe collecting /proc/vmstat
>>>during the test will tell us more.
>>
>>Attached the JSON format per-second vmstat records.
>>It feels more readable than the raw dumps.
>
>And here is the meminfo lines.

Here are the raw proc file dumps during every second of the test.

Thanks,
Fengguang

[-- Attachment #2: meminfo.gz --]
[-- Type: application/gzip, Size: 5796 bytes --]

[-- Attachment #3: proc-vmstat.gz --]
[-- Type: application/gzip, Size: 13705 bytes --]