Re: [syzbot ci] Re: make vmalloc gfp flags usage more apparent

[Vishal Moola (Oracle)]

On Wed, Nov 12, 2025 at 01:02:03PM +0100, Uladzislau Rezki wrote:
> On Tue, Nov 11, 2025 at 11:07:29PM -0800, Christoph Hellwig wrote:
> > On Tue, Nov 11, 2025 at 09:21:06PM +0100, Uladzislau Rezki wrote:
> > > > Unexpected gfp: 0x100000 (__GFP_HARDWALL). Fixing up to gfp: 0xdc0 (GFP_KERNEL|__GFP_ZERO). Fix your code!
> > > >
> > > It looks like we need to add __GFP_HARDWALL to the white-list-mask.
> > 
> > __GFP_HARDWALL is part of GFP_USER.  Doing GFP_USER vmalloc sounds like
> > a bit of an odd idea to me, but there are a few users mostly in bpf
> > and drm code (why do these always show up for odd API usage patterns?).
> > 
> > So I guess yes, we'll need to allow it for now, but I'd like to start
> > a discussion if it really makes much sense.
> > 
> <snip>
> 	/* plain bpf_prog allocation */
> 	prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER);
> 	if (!prog) {
> <snip>
> 
> I assume that was the place that triggered the splat.
> 
> Vishal, will you send the patch adding GFP_USER to address the splat?

Yes, I'll send a new version including __GFP_HARDWALL in the mask and
update the comment accordingly.