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

[Uladzislau Rezki]

On Tue, Jul 13, 2021 at 07:26:53PM +0100, Matthew Wilcox wrote:
> On Tue, Jul 13, 2021 at 11:19:29AM -0700, Linus Torvalds wrote:
> > Does anybody see what the problem is there?
> > 
> > There's an odd report _after_ the warning:
> > 
> > [  131.345319] raw_local_irq_restore() called with IRQs enabled
> > [  131.366561] RIP: 0010:warn_bogus_irq_restore+0x1d/0x20
> > [  131.433334]  __alloc_pages_bulk+0xbb8/0xf20
> 
> That's the key -- __alloc_pages_bulk has interrupts disabled and then
> page_owner allocates memory for the stack dump.  Mel has a fix; I think
> we're just waiting for it to hit your tree.
> 
I was thinking about how we came to the step when a sleeping check is fired
somewhere deep in the "page-bulk" allocator. If vmalloc was invoked from
non-sleepin context we would see it earlier, at least in alloc_vmap_area().

I think, the bulk allocator disables interrupts and does some sleeping
things.

Matthew, Could you please point to the fix?

--
Vlad Rezki