[vbox-dev] VirtualBox 5.0.0 r101573 shows "scheduling while atomic" splats

[Larry Finger]

On 07/14/2015 08:51 AM, Larry Finger wrote:
> On Linux kernels with lockdep testing enabled, version 5.0.0 issues many of the
> following splats in the logs:
>
> BUG: sleeping function called from invalid context at kernel/cpu.c:94
> in_atomic(): 1, irqs_disabled(): 0, pid: 5748, name: EMT-0
> INFO: lockdep is turned off.
> CPU: 1 PID: 5748 Comm: EMT-0 Tainted: G        W  O    4.2.0-wd-next-wl+ #73
> Hardware name: TOSHIBA TECRA A50-A/TECRA A50-A, BIOS Version 4.20   04/17/2014
>    ffffffff81a39004 ffff88000d287c28 ffffffff817486dd 0000000000000000
>    ffff88015c314d80 ffff88000d287c58 ffffffff81090117 ffff88000d287c68
>    ffffffff81a39004 000000000000005e 0000000000000000 ffff88000d287c88
> Call Trace:
>    [<ffffffff817486dd>] dump_stack+0x4c/0x6e
>    [<ffffffff81090117>] ___might_sleep+0x147/0x1f0
>    [<ffffffff8109020d>] __might_sleep+0x4d/0x90
>    [<ffffffff81065300>] get_online_cpus+0x20/0x80
>    [<ffffffff810653aa>] ? put_online_cpus+0x4a/0x90
>    [<ffffffff8100919f>] arch_jump_label_transform+0x1f/0x100
>    [<ffffffff81180fd8>] __jump_label_update+0x58/0x70
>    [<ffffffff81181093>] jump_label_update+0xa3/0xc0
>    [<ffffffff811812f5>] static_key_slow_inc+0xb5/0xc0
>    [<ffffffff8108fccd>] preempt_notifier_register+0x1d/0x60
>    [<ffffffffa099166f>] VBoxHost_RTThreadCtxHookEnable+0x2f/0x50 [vboxdrv]
>    [<ffffffffa097e5f1>] ? supdrvIOCtlFast+0x71/0x90 [vboxdrv]
>    [<ffffffffa097b40c>] ? VBoxDrvLinuxIOCtl_5_0_0+0x4c/0x270 [vboxdrv]
>    [<ffffffff8121f2f5>] ? __fget+0x5/0x1f0
>    [<ffffffff81213aad>] ? do_vfs_ioctl+0x30d/0x580
>    [<ffffffff8121f2f5>] ? __fget+0x5/0x1f0
>    [<ffffffff811dfff5>] ? kfree+0x35/0x340
>    [<ffffffff8121f54a>] ? __fget_light+0x2a/0x90
>    [<ffffffff81011ad6>] ? do_audit_syscall_entry+0x66/0x70
>    [<ffffffff81213d99>] ? SyS_ioctl+0x79/0x90
>    [<ffffffff81751ed7>] ? entry_SYSCALL_64_fastpath+0x12/0x6f
>
> I thought that I had found this in Beta/RC testing, but I can find no evidence
> of a fix or having reported it.

This scheduling while atomic splat apparently is caused by same problem that 
leads to the lockups. When I fixed that one, these were gone.

Larry