From mboxrd@z Thu Jan  1 00:00:00 1970
From: jan.glauber@caviumnetworks.com (Jan Glauber)
Date: Mon, 29 Jan 2018 18:30:54 +0100
Subject: [BUG] perf, arm64, acpi: sleeping function called from invalid context
Message-ID: <20180129173054.GD794@hc>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

Hi Will & Mark,

I'm seeing the following warning with 4.15 (and earlier) using ACPI & perf:

[   34.823577] BUG: sleeping function called from invalid context at mm/slab.h:419
[   34.830881] in_atomic(): 0, irqs_disabled(): 128, pid: 14, name: cpuhp/0
[   34.837574] 1 lock held by cpuhp/0/14:
[   34.841314]  #0:  (cpuhp_state-up){....}, at: [<00000000f44ba116>] cpuhp_thread_fun+0x148/0x290
[   34.850032] CPU: 0 PID: 14 Comm: cpuhp/0 Not tainted 4.15.0-rc9-jang+ #13
[   34.856810] Hardware name: Default string Cavium ThunderX2/Default string, BIOS 5.13 12/18/2017
[   34.865499] Call trace:
[   34.867941]  dump_backtrace+0x0/0x160
[   34.871595]  show_stack+0x24/0x30
[   34.874905]  dump_stack+0x9c/0xd0
[   34.878214]  ___might_sleep+0x140/0x1a0
[   34.882042]  __might_sleep+0x58/0x90
[   34.885610]  kmem_cache_alloc_trace+0x2c4/0x320
[   34.890134]  armpmu_alloc+0x38/0x1b0
[   34.893701]  arm_pmu_acpi_cpu_starting+0x10c/0x138
[   34.898484]  cpuhp_invoke_callback+0x120/0xaa8
[   34.902920]  cpuhp_thread_fun+0xec/0x290
[   34.906834]  smpboot_thread_fn+0x21c/0x2b8
[   34.910923]  kthread+0x10c/0x138
[   34.914143]  ret_from_fork+0x10/0x18

Changing the allocations in arm_pmu_alloc() to GFP_ATOMIC didn't help,
as the interrupt request is also not happy in this context.

Would it be possible to init the PMUs later?

--Jan