Re: [PATCH RFC v2 13/20] x86: mm: carve out the generic compile-time folded pgtable case in effective_prot()

From: Yeoreum Yun <yeoreum.yun@arm.com>
To: Dave Hansen <dave.hansen@intel.com>
Cc: Yeoreum Yun <yeoreum.yun@arm.com>,
	Russell King <linux@armlinux.org.uk>,
	Huacai Chen <chenhuacai@kernel.org>,
	WANG Xuerui <kernel@xen0n.name>,
	Thomas Bogendoerfer <tsbogend@alpha.franken.de>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>, Arnd Bergmann <arnd@arndb.de>,
	Andrew Morton <akpm@linux-foundation.org>,
	Kairui Song <kasong@tencent.com>, Qi Zheng <qi.zheng@linux.dev>,
	Shakeel Butt <shakeel.butt@linux.dev>,
	Barry Song <baohua@kernel.org>,
	Axel Rasmussen <axelrasmussen@google.com>,
	Yuanchu Xie <yuanchu@google.com>, Wei Xu <weixugc@google.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	David Hildenbrand <david@kernel.org>,
	Michal Hocko <mhocko@kernel.org>,
	Lorenzo Stoakes <ljs@kernel.org>,
	Tianrui Zhao <zhaotianrui@loongson.cn>,
	Bibo Mao <maobibo@loongson.cn>, Anup Patel <anup@brainfault.org>,
	Atish Patra <atish.patra@linux.dev>,
	Paul Walmsley <pjw@kernel.org>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Alexandre Ghiti <alex@ghiti.fr>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Andy Lutomirski <luto@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@kernel.org>, Ingo Molnar <mingo@redhat.com>,
	Borislav Petkov <bp@alien8.de>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
	"Liam R. Howlett" <liam@infradead.org>,
	Vlastimil Babka <vbabka@kernel.org>,
	Mike Rapoport <rppt@kernel.org>,
	Suren Baghdasaryan <surenb@google.com>,
	Michal Hocko <mhocko@suse.com>, Jonas Bonn <jonas@southpole.se>,
	Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>,
	Stafford Horne <shorne@gmail.com>,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org, loongarch@lists.linux.dev,
	linux-mips@vger.kernel.org, linux-arch@vger.kernel.org,
	linux-mm@kvack.org, kvm@vger.kernel.org,
	kvm-riscv@lists.infradead.org, linux-riscv@lists.infradead.org,
	linux-openrisc@vger.kernel.org
Subject: Re: [PATCH RFC v2 13/20] x86: mm: carve out the generic compile-time folded pgtable case in effective_prot()
Date: Wed, 22 Jul 2026 17:28:18 +0100	[thread overview]
Message-ID: <amDvonIJezy_YQZA@e129823.arm.com> (raw)
In-Reply-To: <d16a8412-acee-4c21-a126-f570d6306983@intel.com>

Hi Dave,

> On 7/22/26 08:30, Yeoreum Yun wrote:
> >  struct pg_state {
> >  	struct ptdump_state ptdump;
> > +	int first_level;
> >  	int level;
> >  	pgprotval_t current_prot;
> >  	pgprotval_t effective_prot;
> 
> There is zero reason on x86 to hold this value in a data structure. It
> is 100% determined at boot time or earlier. Pre-5-level-paging it's a
> compile-time constant.
> 
> On x86, it never varies per-mm.

Yes once ptdump_pt_level_first() is converted with static inline.
I'll change this.

> > --- a/mm/ptdump.c
> > +++ b/mm/ptdump.c
> > @@ -197,6 +197,17 @@ void ptdump_walk_pgd(struct ptdump_state *st, struct mm_struct *mm, pgd_t *pgd)
> >  	st->note_page_flush(st);
> >  }
> >  
> > +int ptdump_pt_level_first(struct mm_struct *mm)
> > +{
> > +	if (mm_pmd_folded(mm))
> > +		return 3;
> > +	if (mm_pud_folded(mm))
> > +		return 2;
> > +	if (mm_p4d_folded(mm))
> > +		return 1;
> > +	return 0;
> > +}
> 
> Having this in a .c file may not be the best design decision. On a lot
> of architectures and configs, this is 100% a compile-time constant.
> Hiding it in a .c file and forcing a call is kinda silly.
> 
> In the worst-case scenario on x86, this ends up being an "ALTERNATIVES"
> patched check. So it's technically variable, but patched at boot-time to
> one possibility or the other.
> 
> If something is making the function larger than a few instructions,
> something is wrong and needs to get fixed.
> 
> IOW, this should be a static inline in a header, not a full CALL'd
> function in a .c file. *That* will make the value visible to the
> compiler at all call sites and let it do all the smart things compilers
> can do with build time constants.
> 
> Also, my original objection was "this function has zero to do with x86".
> Well, it also really has zero to do with ptdump, either. I'd probably
> just stick it in one of the generic page table headers.

Agree. I'll locate in the linux/pgtable.h

-- 
Sincerely,
Yeoreum Yun

-- 
kvm-riscv mailing list
kvm-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kvm-riscv