linux-kernel - Re: [RFC PATCH] arch: arm64: introduce RODATA_FULL_USE_PTE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAMj1kXGvVkMtZrKYpr8SP8YSvUuWnyQ2i67riHBdQgfDy_0e1A@mail.gmail.com>
Date:   Mon, 22 Nov 2021 09:52:08 +0100
From:   Ard Biesheuvel <ardb@...nel.org>
To:     Huangzhaoyang <huangzhaoyang@...il.com>
Cc:     Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>,
        Anshuman Khandual <anshuman.khandual@....com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Nicholas Piggin <npiggin@...il.com>,
        Mike Rapoport <rppt@...nel.org>,
        Pavel Tatashin <pasha.tatashin@...een.com>,
        Christophe Leroy <christophe.leroy@...roup.eu>,
        Jonathan Marek <jonathan@...ek.ca>,
        Zhaoyang Huang <zhaoyang.huang@...soc.com>,
        Linux Memory Management List <linux-mm@...ck.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [RFC PATCH] arch: arm64: introduce RODATA_FULL_USE_PTE_CONT

On Mon, 22 Nov 2021 at 06:29, Huangzhaoyang <huangzhaoyang@...il.com> wrote:
>
> From: Zhaoyang Huang <zhaoyang.huang@...soc.com>
>
> Kernel linear mapping will be split to the smallest granularity when
> RODATA_FULL applied, which could lead to TLB pressure. Introduce a method
> to apply PTE_CONT on pte.
>
> Signed-off-by: Zhaoyang Huang <zhaoyang.huang@...soc.com>

How would this lead to TLB pressure, and how does the use of
contiguous mappings mitigate that? The linear mapping of the kernel is
rarely used, as all normal accesses to it go via the vmalloc region,
so in which case would TLB entries be allocated for this region in a
way that could cause a measurable performance impact?


> ---
>  arch/arm64/Kconfig       |  9 +++++++++
>  arch/arm64/mm/mmu.c      | 10 ++++++++--
>  arch/arm64/mm/pageattr.c |  9 +++++++++
>  3 files changed, 26 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index fee914c..3f8fbf0 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -1198,6 +1198,15 @@ config RODATA_FULL_DEFAULT_ENABLED
>           This requires the linear region to be mapped down to pages,
>           which may adversely affect performance in some cases.
>
> +config RODATA_FULL_USE_PTE_CONT
> +       bool "Apply PTE_CONT when RODATA_FULL_DEFAULT_ENABLED enabled"
> +       depends on RODATA_FULL_DEFAULT_ENABLED
> +       default y
> +       help
> +         Apply PTE_CONT on linear mapping as much as we can when
> +         RODATA_FULL_DEFAULT_ENABLED enabled which could decrease the
> +         impaction on performance by small pte granularity.
> +
>  config ARM64_SW_TTBR0_PAN
>         bool "Emulate Privileged Access Never using TTBR0_EL1 switching"
>         help
> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> index cfd9deb..8017b17 100644
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -124,15 +124,21 @@ static bool pgattr_change_is_safe(u64 old, u64 new)
>          * The following mapping attributes may be updated in live
>          * kernel mappings without the need for break-before-make.
>          */
> +#ifndef CONFIG_RODATA_FULL_USE_PTE_CONT
>         pteval_t mask = PTE_PXN | PTE_RDONLY | PTE_WRITE | PTE_NG;
> +#else
> +       pteval_t mask = PTE_PXN | PTE_RDONLY | PTE_WRITE | PTE_NG | PTE_CONT;
> +#endif
>
>         /* creating or taking down mappings is always safe */
>         if (old == 0 || new == 0)
>                 return true;
>
>         /* live contiguous mappings may not be manipulated at all */
> -       if ((old | new) & PTE_CONT)
> +#ifndef CONFIG_RODATA_FULL_USE_PTE_CONT
> +       if (old | new) & PTE_CONT)
>                 return false;
> +#endif
>
>         /* Transitioning from Non-Global to Global is unsafe */
>         if (old & ~new & PTE_NG)
> @@ -206,7 +212,7 @@ static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr,
>
>                 /* use a contiguous mapping if the range is suitably aligned */
>                 if ((((addr | next | phys) & ~CONT_PTE_MASK) == 0) &&
> -                   (flags & NO_CONT_MAPPINGS) == 0)
> +                   (IS_ENABLED(CONFIG_RODATA_FULL_USE_PTE_CONT) || (flags & NO_CONT_MAPPINGS) == 0))
>                         __prot = __pgprot(pgprot_val(prot) | PTE_CONT);
>
>                 init_pte(pmdp, addr, next, phys, __prot);
> diff --git a/arch/arm64/mm/pageattr.c b/arch/arm64/mm/pageattr.c
> index a3bacd7..88a87eb 100644
> --- a/arch/arm64/mm/pageattr.c
> +++ b/arch/arm64/mm/pageattr.c
> @@ -99,6 +99,15 @@ static int change_memory_common(unsigned long addr, int numpages,
>         if (rodata_full && (pgprot_val(set_mask) == PTE_RDONLY ||
>                             pgprot_val(clear_mask) == PTE_RDONLY)) {
>                 for (i = 0; i < area->nr_pages; i++) {
> +#ifdef CONFIG_RODATA_FULL_USE_PTE_CONT
> +                       unsigned long cont_pte_low_bound;
> +                       unsigned long addr;
> +
> +                       addr = (u64)page_address(area->pages[i]);
> +                       cont_pte_low_bound = addr & CONT_PTE_MASK;
> +                       __change_memory_common(cont_pte_low_bound,
> +                                              (~CONT_PTE_MASK + 1), __pgprot(0) , __pgprot(PTE_CONT));
> +#endif
>                         __change_memory_common((u64)page_address(area->pages[i]),
>                                                PAGE_SIZE, set_mask, clear_mask);
>                 }
> --
> 1.9.1
>