Re: [mm/vmalloc] 5c1f4e690e: BUG:sleeping_function_called_from_invalid_context_at_mm/page_alloc.c

[Linus Torvalds]

[ I've asked for this before, but I'll ask for it again.. ]

On Tue, Jul 13, 2021 at 7:06 AM kernel test robot <oliver.sang@xxxxxxxxx> wrote:
>
>
> [  131.014885] BUG: sleeping function called from invalid context at mm/page_alloc.c:4992
> [  131.019428] in_atomic(): 0, irqs_disabled(): 1, non_block: 0, pid: 2928, name: trinity-c6
> [  131.023996] no locks held by trinity-c6/2928.
> [  131.027987] irq event stamp: 283042
> [  131.031844] hardirqs last  enabled at (283041): [<ffffffff8643af22>] _raw_spin_unlock_irqrestore+0x42/0x50
> [  131.036625] hardirqs last disabled at (283042): [<ffffffff81656d71>] __alloc_pages_bulk+0xae1/0xf20
> [  131.041305] softirqs last  enabled at (281858): [<ffffffff86800623>] __do_softirq+0x623/0x9db
> [  131.046157] softirqs last disabled at (281853): [<ffffffff8116dfee>] irq_exit_rcu+0xce/0xf0
> [  131.050925] CPU: 0 PID: 2928 Comm: trinity-c6 Not tainted 5.13.0+ #1
> [  131.055391] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
> [  131.060133] Call Trace:
> [  131.064110]  dump_stack+0xa5/0xe6
> [  131.068192]  ___might_sleep.cold+0x147/0x177
> [  131.072324]  __alloc_pages+0x462/0x650

Btw, can I ask the kernel test robot to do the same thing I asked
syzbot to do - namely get better backtrace annotations?

It requires debug info in the kernel (but I think DEBUG_INFO_REDUCED
is sufficient, no need for the full one that blows sizes up a lot),
but it makes things _soo_ much easier to read. Not just because of
line numbers, but particularly with inlining information.

It's easy enough to do: just run the kernel output through
"scripts/decode_stacktrace.sh".

That will also do the "Code:" line instruction disassembly, which is
less critical, since that can be done later at any time with no loss
of information, but it can be a convenience.

               Linus