Re: BUG: sleeping function called from atomic due to "Balance initial LPI affinity across CPUs"

[John Garry]

On 20/05/2020 10:51, Marc Zyngier wrote:
> Hi John,
>
> On 2020-05-20 09:43, John Garry wrote:
> > On 19/05/2020 23:09, Qian Cai wrote:
> > > Reverted the linux-next commit f068a62c548c ("irqchip/gic-v3-its:
> > > Balance initial LPI affinity across CPUs") fixed these warnings during
> > > boot,
> >
> > Thanks for the notice. So we need the following set to see this:
> > CONFIG_CPUMASK_OFFSTACK=y
> > CONFIG_DEBUG_ATOMIC_SLEEP=y
> > CONFIG_DEBUG_PER_CPU_MAPS=y
>
> Ah, thanks for pointing this out.
>
> > > its_select_cpu at drivers/irqchip/irq-gic-v3-its.c:1572
> > >
> > > [Â 332.819381][ T3359] BUG: sleeping function called from invalid
> > > context at mm/slab.h:568
> > > [Â 332.827405][ T3359] in_atomic(): 1, irqs_disabled(): 128,
> > > non_block: 0, pid: 3359, name: irqbalance
> > > [Â 332.836455][ T3359] INFO: lockdep is turned off.
> > > [Â 332.841076][ T3359] irq event stamp: 0
> > > [ 332.844836][ T3359] hardirqs last enabled at (0): 
> > > [<0000000000000000>] 0x0
> > > [Â 332.851828][ T3359] hardirqs last disabled at (0):
> > > [<ffff9000101ea65c>] copy_process+0x98c/0x1f34
> > > [ 332.860710][ T3359] softirqs last enabled at (0):
> > > [<ffff9000101ea690>] copy_process+0x9c0/0x1f34
> > > [Â 332.869586][ T3359] softirqs last disabled at (0): 
> > > [<0000000000000000>] 0x0
> > > [Â 332.876560][ T3359] CPU: 155 PID: 3359 Comm: irqbalance Tainted: G
> > > ÂÂÂÂÂÂ WÂÂÂ LÂÂÂ 5.7.0-rc6-next-20200519 #1
> > > [Â 332.886563][ T3359] Hardware name: HPE Apollo 70
> > > /C01_APACHE_MBÂÂÂÂÂÂÂÂ , BIOS L50_5.13_1.11 06/18/2019
> > > [Â 332.897000][ T3359] Call trace:
> > > [Â 332.900151][ T3359]Â dump_backtrace+0x0/0x22c
> > > [Â 332.904514][ T3359]Â show_stack+0x28/0x34
> > > [Â 332.908543][ T3359]Â dump_stack+0x104/0x194
> > > [Â 332.912738][ T3359]Â ___might_sleep+0x314/0x328
> > > [Â 332.917274][ T3359]Â __might_sleep+0x7c/0xe0
> > > [Â 332.921563][ T3359]Â slab_pre_alloc_hook+0x44/0x8c
> > > [Â 332.926360][ T3359]Â __kmalloc_node+0xb0/0x618
> > > [Â 332.930811][ T3359]Â alloc_cpumask_var_node+0x48/0x94
> >
> > We could use GFP_ATOMIC flag at the callsite here, but maybe there is
> > a better solution.
>
> I don't see one, and I doubt it is worth the hassle to have anything
> but GFP_ATOMIC. The default arm64 config is to have on-stack cpumasks,
> and only DEBUG_PER_CPU_MAPS allows this to be changed.

JFYI, I was able to recreate on my D05:

[11.951739] CPU: 21 PID: 1 Comm: swapper/0 Tainted: G 
W5.7.0-rc6-next-20200519-dirty #16
[11.961262] Hardware name: Huawei Taishan 2280 /D05, BIOS Hisilicon D05 
IT21 Nemo 2.0 RC0 04/18/2018
[11.970435] Call trace:
[11.972890]  dump_backtrace+0x0/0x1b0
[11.976561]  show_stack+0x14/0x1c
[11.979884]  dump_stack+0xc0/0xf0
[11.983207]  ___might_sleep+0x10c/0x12c
[11.987052]  __might_sleep+0x4c/0x80
[11.990637]  __kmalloc_node+0x1ac/0x2dc
[11.994483]  alloc_cpumask_var_node+0x28/0x60
[11.998852]  alloc_cpumask_var+0x10/0x18
[12.002787]  its_select_cpu+0x2c/0x19c
[12.006546]  its_irq_domain_activate+0x38/0xf0
[12.011004]  __irq_domain_activate_irq+0x6c/0xac
[12.011108] sd 0:0:7:0: [sdh] Attached SCSI disk
[12.015636]  __irq_domain_activate_irq+0x34/0xac
[12.015638]  __irq_domain_activate_irq+0x34/0xac
[12.015640]  irq_domain_activate_irq+0x3c/0x4c
[12.015643]  irq_activate+0x20/0x30
[12.037482]  __setup_irq+0x538/0x6ec
[12.041065]  request_threaded_irq+0xdc/0x18c
[12.045350]  usb_add_hcd+0x4b0/0x6c0
[12.048934]  ehci_platform_probe+0x1f4/0x438
[12.053218]  platform_drv_probe+0x4c/0xa0
[12.057239]  really_probe+0xf4/0x35c
[12.060823]  driver_probe_device+0x54/0xb0
[12.064931]  device_driver_attach+0x68/0x70
[12.069126]  __driver_attach+0x9c/0xf8
[12.072884]  bus_for_each_dev+0x50/0xa0
[12.076730]  driver_attach+0x20/0x28
[12.080313]  bus_add_driver+0x148/0x1fc
[12.084159]  driver_register+0x6c/0x124
[12.088005]  __platform_driver_register+0x48/0x50
[12.092727]  ehci_platform_init+0x54/0x64
[12.096747]  do_one_initcall+0x50/0x194
[12.100593]  kernel_init_freeable+0x1e4/0x250
[12.104964]  kernel_init+0x10/0x108
[12.108460]  ret_from_fork+0x10/0x18
[12.112077] ehci-platform PNP0D20:00: irq 361, io mem 0xa7020000

And setting GFP_ATOMIC removes that BUG.

Cheers,
John