linux-kernel - Re: [PATCH 3/4] mm/tlb, x86/mm: Support invalidating TLB caches for RCU_TABLE

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: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <C2D7FE5348E1B147BCA15975FBA23075012B090CA3@us01wembx1.internal.synopsys.com>
Date:   Thu, 30 Aug 2018 00:13:50 +0000
From:   Vineet Gupta <Vineet.Gupta1@...opsys.com>
To:     Peter Zijlstra <peterz@...radead.org>,
        Nicholas Piggin <npiggin@...il.com>
CC:     Will Deacon <will.deacon@....com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Benjamin Herrenschmidt <benh@....ibm.com>,
        Andrew Lutomirski <luto@...nel.org>,
        the arch/x86 maintainers <x86@...nel.org>,
        Borislav Petkov <bp@...en8.de>,
        Rik van Riel <riel@...riel.com>,
        Jann Horn <jannh@...gle.com>,
        Adin Scannell <ascannell@...gle.com>,
        Dave Hansen <dave.hansen@...el.com>,
        "Linux Kernel Mailing List" <linux-kernel@...r.kernel.org>,
        linux-mm <linux-mm@...ck.org>,
        David Miller <davem@...emloft.net>,
        Martin Schwidefsky <schwidefsky@...ibm.com>,
        Michael Ellerman <mpe@...erman.id.au>,
        "jejb@...isc-linux.org" <jejb@...isc-linux.org>
Subject: Re: [PATCH 3/4] mm/tlb, x86/mm: Support invalidating TLB caches for
 RCU_TABLE_FREE

On 08/27/2018 04:00 AM, Peter Zijlstra wrote:
>
> The one obvious thing SH and ARM want is a sensible default for
> tlb_start_vma(). (also: https://lkml.org/lkml/2004/1/15/6 )
>
> The below make tlb_start_vma() default to flush_cache_range(), which
> should be right and sufficient. The only exceptions that I found where
> (oddly):
>
>   - m68k-mmu
>   - sparc64
>   - unicore
>
> Those architectures appear to have a non-NOP flush_cache_range(), but
> their current tlb_start_vma() does not call it.

So indeed we follow the DaveM's insight from 2004 about tlb_{start,end}_vma() and
those are No-ops for ARC for the general case. For the historic VIPT aliasing
dcache they are what they should be per 2004 link above - I presume that is all
hunky dory with you ?

> Furthermore, I think tlb_flush() is broken on arc and parisc; in
> particular they don't appear to have any TLB invalidate for the
> shift_arg_pages() case, where we do not call tlb_*_vma() and fullmm=0.

Care to explain this issue a bit more ?
And that is independent of the current discussion.

> Possibly shift_arg_pages() should be fixed instead.
>
> Some archs (nds32,sparc32) avoid this by having an unconditional
> flush_tlb_mm() in tlb_flush(), which seems somewhat suboptimal if you
> have flush_tlb_range().  TLB_FLUSH_VMA() might be an option, however
> hideous it is.
>
> ---
>
> diff --git a/arch/arc/include/asm/tlb.h b/arch/arc/include/asm/tlb.h
> index a9db5f62aaf3..7af2b373ebe7 100644
> --- a/arch/arc/include/asm/tlb.h
> +++ b/arch/arc/include/asm/tlb.h
> @@ -23,15 +23,6 @@ do {						\
>   *
>   * Note, read http://lkml.org/lkml/2004/1/15/6
>   */
> -#ifndef CONFIG_ARC_CACHE_VIPT_ALIASING
> -#define tlb_start_vma(tlb, vma)
> -#else
> -#define tlb_start_vma(tlb, vma)						\
> -do {									\
> -	if (!tlb->fullmm)						\
> -		flush_cache_range(vma, vma->vm_start, vma->vm_end);	\
> -} while(0)
> -#endif
>  
>  #define tlb_end_vma(tlb, vma)						\
>  do {									\

[snip..]

> 				      \
> diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
> index e811ef7b8350..1d037fd5bb7a 100644
> --- a/include/asm-generic/tlb.h
> +++ b/include/asm-generic/tlb.h
> @@ -181,19 +181,21 @@ static inline void tlb_remove_check_page_size_change(struct mmu_gather *tlb,
>   * the vmas are adjusted to only cover the region to be torn down.
>   */
>  #ifndef tlb_start_vma
> -#define tlb_start_vma(tlb, vma) do { } while (0)
> +#define tlb_start_vma(tlb, vma)						\
> +do {									\
> +	if (!tlb->fullmm)						\
> +		flush_cache_range(vma, vma->vm_start, vma->vm_end);	\
> +} while (0)
>  #endif

So for non aliasing arches to be not affected, this relies on flush_cache_range()
to be no-op ?

> -#define __tlb_end_vma(tlb, vma)					\
> -	do {							\
> -		if (!tlb->fullmm && tlb->end) {			\
> -			tlb_flush(tlb);				\
> -			__tlb_reset_range(tlb);			\
> -		}						\
> -	} while (0)
> -
>  #ifndef tlb_end_vma
> -#define tlb_end_vma	__tlb_end_vma
> +#define tlb_end_vma(tlb, vma)						\
> +	do {								\
> +		if (!tlb->fullmm && tlb->end) {				\
> +			tlb_flush(tlb);					\
> +			__tlb_reset_range(tlb);				\
> +		}							\
> +	} while (0)
>  #endif
>  
>  #ifndef __tlb_remove_tlb_entry

And this one is for shift_arg_pages() but will also cause extraneous flushes for
other cases - not happening currently !

-Vineet