Re: [PATCH v4 bpf 0/4] vmalloc: bpf: introduce VM_ALLOW_HUGE_VMAP

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Nicholas Piggin <npiggin@gmail.com>
Cc: Mike Rapoport <rppt@kernel.org>,
	"akpm@linux-foundation.org" <akpm@linux-foundation.org>,
	"ast@kernel.org" <ast@kernel.org>, "bp@alien8.de" <bp@alien8.de>,
	"bpf@vger.kernel.org" <bpf@vger.kernel.org>,
	"daniel@iogearbox.net" <daniel@iogearbox.net>,
	"dborkman@redhat.com" <dborkman@redhat.com>,
	"edumazet@google.com" <edumazet@google.com>,
	"hch@infradead.org" <hch@infradead.org>,
	"hpa@zytor.com" <hpa@zytor.com>,
	"imbrenda@linux.ibm.com" <imbrenda@linux.ibm.com>,
	Kernel Team <Kernel-team@fb.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-mm@kvack.org" <linux-mm@kvack.org>,
	"mbenes@suse.cz" <mbenes@suse.cz>,
	"mcgrof@kernel.org" <mcgrof@kernel.org>,
	"pmladek@suse.com" <pmladek@suse.com>,
	"Edgecombe, Rick P" <rick.p.edgecombe@intel.com>,
	"song@kernel.org" <song@kernel.org>,
	Song Liu <songliubraving@fb.com>
Subject: Re: [PATCH v4 bpf 0/4] vmalloc: bpf: introduce VM_ALLOW_HUGE_VMAP
Date: Wed, 20 Apr 2022 22:48:20 -0700	[thread overview]
Message-ID: <CAHk-=wiQ5=S3m2+xRbm-1H8fuQwWfQxnO7tHhKg8FjegxzdVaQ@mail.gmail.com> (raw)
In-Reply-To: <1650511496.iys9nxdueb.astroid@bobo.none>

On Wed, Apr 20, 2022 at 8:25 PM Nicholas Piggin <npiggin@gmail.com> wrote:
>
> Why not just revert fac54e2bfb5b ?

That would be stupid, with no sane way forward.

The fact is, HUGE_VMALLOC was badly misdesigned, and enabling it on
x86 only ended up showing the problems.

It wasn't fac54e2bfb5b that was the fundamental issue. It was the
whole "oh, we should never have done it that way to begin with".

The whole initial notion that HAVE_ARCH_HUGE_VMALLOC means that there
must be no PAGE_SIZE pte assumptions was simply broken. There were
actual real cases that had those assumptions, and the whole "let's
just change vmalloc behavior by default and then people who don't like
it can opt out" was just fundamentally a really bad idea.

Power had that random "oh, we don't want to do this for module_alloc",
which you had a comment about "more testing" for.

And s390 had a case of hardware limitations where it didn't work for some cases.

And then enabling it on x86 turned up more issues.

So yes, commit fac54e2bfb5b _exposed_ things to a much larger
audience. But all it just made clear was that your original notion of
"let's change behavior and randomly disable it as things turn up" was
just broken.

Including "small" details like the fact that apparently
VM_FLUSH_RESET_PERMS didn't work correctly any more for this, which
caused issues for bpf, and that [PATCH 4/4]. And yes, there was a
half-arsed comment ("may require extra work") to that effect in the
powerpc __module_alloc() function, but it had been left to others to
notice separately.

So no. We're not going back to that completely broken model. The
lagepage thing needs to be opt-in, and needs a lot more care.

               Linus