Re: [PATCH v2 02/12] x86: pgtable: Always use pte_free_kernel()

From: Dave Hansen <dave.hansen@intel.com>
To: Matthew Wilcox <willy@infradead.org>
Cc: Kevin Brodsky <kevin.brodsky@arm.com>,
	linux-mm@kvack.org, linux-kernel@vger.kernel.org,
	Albert Ou <aou@eecs.berkeley.edu>,
	Andreas Larsson <andreas@gaisler.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	"David S. Miller" <davem@davemloft.net>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	Linus Walleij <linus.walleij@linaro.org>,
	Madhavan Srinivasan <maddy@linux.ibm.com>,
	Mark Rutland <mark.rutland@arm.com>,
	Michael Ellerman <mpe@ellerman.id.au>,
	"Mike Rapoport (IBM)" <rppt@kernel.org>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Qi Zheng <zhengqi.arch@bytedance.com>,
	Ryan Roberts <ryan.roberts@arm.com>,
	Will Deacon <will@kernel.org>,
	Yang Shi <yang@os.amperecomputing.com>,
	linux-arch@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
	linux-csky@vger.kernel.org, linux-m68k@lists.linux-m68k.org,
	linux-openrisc@vger.kernel.org, linux-riscv@lists.infradead.org,
	linux-s390@vger.kernel.org, linuxppc-dev@lists.ozlabs.org,
	sparclinux@vger.kernel.org, x86@kernel.org
Subject: Re: [PATCH v2 02/12] x86: pgtable: Always use pte_free_kernel()
Date: Tue, 8 Apr 2025 09:54:42 -0700	[thread overview]
Message-ID: <9247436d-ae01-4eb8-bd5d-370b2fb2eebc@intel.com> (raw)
In-Reply-To: <Z_VQxyqkU8DV7QGy@casper.infradead.org>

On 4/8/25 09:37, Matthew Wilcox wrote:
> On Tue, Apr 08, 2025 at 08:22:47AM -0700, Dave Hansen wrote:
>> Are there any tests for folio_test_pgtable() at free_page() time? If we
>> had that, it would make it less likely that another free_page() user
>> could sneak in without calling the destructor.
> It's hidden, but yes:
> 
> static inline bool page_expected_state(struct page *page,
>                                         unsigned long check_flags)
> {
>         if (unlikely(atomic_read(&page->_mapcount) != -1))
>                 return false;
> 
> PageTable uses page_type which aliases with mapcount, so this check
> covers "PageTable is still set when the last refcount to it is put".

Huh, so shouldn't we have ended up in bad_page() for these, other than:

        pagetable_dtor(virt_to_ptdesc(pmd));
        free_page((unsigned long)pmd);

?