From: Zhenyu Ye <yezhenyu2@huawei.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: <mark.rutland@arm.com>, <will@kernel.org>,
<catalin.marinas@arm.com>, <aneesh.kumar@linux.ibm.com>,
<akpm@linux-foundation.org>, <npiggin@gmail.com>, <arnd@arndb.de>,
<rostedt@goodmis.org>, <maz@kernel.org>, <suzuki.poulose@arm.com>,
<tglx@linutronix.de>, <yuzhao@google.com>, <Dave.Martin@arm.com>,
<steven.price@arm.com>, <broonie@kernel.org>,
<guohanjun@huawei.com>, <corbet@lwn.net>, <vgupta@synopsys.com>,
<tony.luck@intel.com>, <linux-arm-kernel@lists.infradead.org>,
<linux-kernel@vger.kernel.org>, <linux-arch@vger.kernel.org>,
<linux-mm@kvack.org>, <arm@kernel.org>, <xiexiangyou@huawei.com>,
<prime.zeng@hisilicon.com>, <zhangshaokun@hisilicon.com>,
<kuhn.chenqun@huawei.com>
Subject: Re: [RFC PATCH v5 4/8] mm: tlb: Pass struct mmu_gather to flush_pmd_tlb_range
Date: Thu, 2 Apr 2020 19:24:04 +0800 [thread overview]
Message-ID: <53675fb9-21c7-5309-07b8-1bbc1e775f9b@huawei.com> (raw)
In-Reply-To: <20200401122004.GE20713@hirez.programming.kicks-ass.net>
Hi Peter,
On 2020/4/1 20:20, Peter Zijlstra wrote:
> On Wed, Apr 01, 2020 at 04:51:15PM +0800, Zhenyu Ye wrote:
>> On 2020/3/31 23:13, Peter Zijlstra wrote:
>
>>> Instead of trying to retro-fit flush_*tlb_range() to take an mmu_gather
>>> parameter, please replace them out-right.
>>>
>>
>> I'm sorry that I'm not sure what "replace them out-right" means. Do you
>> mean that I should define flush_*_tlb_range like this?
>>
>> #define flush_pmd_tlb_range(vma, addr, end) \
>> do { \
>> struct mmu_gather tlb; \
>> tlb_gather_mmu(&tlb, (vma)->vm_mm, addr, end); \
>> tlba.cleared_pmds = 1; \
>> flush_tlb_range(&tlb, vma, addr, end); \
>> tlb_finish_mmu(&tlb, addr, end); \
>> } while (0)
>>
>
> I was thinking to remove flush_*tlb_range() entirely (from generic
> code).
>
> And specifically to not use them like the above; instead extend the
> mmu_gather API.
>
> Specifically, if you wanted to express flush_pmd_tlb_range() in mmu
> gather, you'd write it like:
>
> static inline void flush_pmd_tlb_range(struct vm_area_struct *vma, unsigned long addr, unsigned long end)
> {
> struct mmu_gather tlb;
>
> tlb_gather_mmu(&tlb, vma->vm_mm, addr, end);
> tlb_start_vma(&tlb, vma);
> tlb.cleared_pmds = 1;
> __tlb_adjust_range(addr, end - addr);
> tlb_end_vma(&tlb, vma);
> tlb_finish_mmu(&tlb, addr, end);
> }
>
> Except of course, that the code between start_vma and end_vma is not a
> proper mmu_gather API.
>
> So maybe add:
>
> tlb_flush_{pte,pmd,pud,p4d}_range()
>
> Then we can write:
>
> static inline void flush_XXX_tlb_range(struct vm_area_struct *vma, unsigned long addr, unsigned long end)
> {
> struct mmu_gather tlb;
>
> tlb_gather_mmu(&tlb, vma->vm_mm, addr, end);
> tlb_start_vma(&tlb, vma);
> tlb_flush_XXX_range(&tlb, addr, end - addr);
> tlb_end_vma(&tlb, vma);
> tlb_finish_mmu(&tlb, addr, end);
> }
>
> But when I look at the output of:
>
> git grep flush_.*tlb_range -- :^arch/
>
> I doubt it makes sense to provide wrappers like the above.
>
Thanks for your detailed explanation. I notice that you used
`tlb_end_vma` replace `flush_tlb_range`, which will call `tlb_flush`,
then finally call `flush_tlb_range` in generic code. However, some
architectures define tlb_end_vma|tlb_flush|flush_tlb_range themselves,
so this may cause problems.
For example, in s390, it defines:
#define tlb_end_vma(tlb, vma) do { } while (0)
And it doesn't define it's own flush_pmd_tlb_range(). So there will be
a mistake if we changed flush_pmd_tlb_range() using tlb_end_vma().
Is this really a problem or something I understand wrong ?
If true, I think there are three ways to solve this problem:
1. use `flush_tlb_range` rather than `tlb_end_vma` in flush_XXX_tlb_range;
In this way, we still need retro-fit `flush_tlb_range` to take an mmu_gather
parameter.
2. use `tlb_flush` rather than `tlb_end_vma`.
There is a constraint such like:
#ifndef tlb_flush
#if defined(tlb_start_vma) || defined(tlb_end_vma)
#error Default tlb_flush() relies on default tlb_start_vma() and tlb_end_vma()
#endif
So all architectures that define tlb_{start|end}_vma have defined tlb_flush.
Also, we can add a constraint to flush_XXX_tlb_range such like:
#ifndef flush_XXX_tlb_range
#if defined(tlb_start_vma) || defined(tlb_end_vma)
#error Default flush_XXX_tlb_range() relies on default tlb_start/end_vma()
#endif
3. Define flush_XXX_tlb_range() architecture-self, and keep original define in
generic code, such as:
In arm64:
#define flush_XXX_tlb_range flush_XXX_tlb_range
In generic:
#ifndef flush_XXX_tlb_range
#define flush_XXX_tlb_range flush_tlb_range
Which do you think is more appropriate?
> ( Also, we should probably remove the (addr, end) arguments from
> tlb_finish_mmu(), Will? )
>
This can be changed quickly. If you want I can do this with a
separate patch.
> ---
> diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
> index f391f6b500b4..be5452a8efaa 100644
> --- a/include/asm-generic/tlb.h
> +++ b/include/asm-generic/tlb.h
> @@ -511,6 +511,34 @@ static inline void tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vm
> }
> #endif
>
> +static inline void tlb_flush_pte_range(struct mmu_gather *tlb,
> + unsigned long address, unsigned long size)
> +{
> + __tlb_adjust_range(tlb, address, size);
> + tlb->cleared_ptes = 1;
> +}
> +
> +static inline void tlb_flush_pmd_range(struct mmu_gather *tlb,
> + unsigned long address, unsigned long size)
> +{
> + __tlb_adjust_range(tlb, address, size);
> + tlb->cleared_pmds = 1;
> +}
> +
> +static inline void tlb_flush_pud_range(struct mmu_gather *tlb,
> + unsigned long address, unsigned long size)
> +{
> + __tlb_adjust_range(tlb, address, size);
> + tlb->cleared_puds = 1;
> +}
> +
> +static inline void tlb_flush_p4d_range(struct mmu_gather *tlb,
> + unsigned long address, unsigned long size)
> +{
> + __tlb_adjust_range(tlb, address, size);
> + tlb->cleared_p4ds = 1;
> +}
> +
By the way, I think the name of tlb_set_XXX_range() is more suitable, because
we don't do actual flush there.
> #ifndef __tlb_remove_tlb_entry
> #define __tlb_remove_tlb_entry(tlb, ptep, address) do { } while (0)
> #endif
> @@ -524,8 +552,7 @@ static inline void tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vm
> */
> #define tlb_remove_tlb_entry(tlb, ptep, address) \
> do { \
> - __tlb_adjust_range(tlb, address, PAGE_SIZE); \
> - tlb->cleared_ptes = 1; \
> + tlb_flush_pte_range(tlb, address, PAGE_SIZE); \
> __tlb_remove_tlb_entry(tlb, ptep, address); \
> } while (0)
>
> @@ -550,8 +577,7 @@ static inline void tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vm
>
> #define tlb_remove_pmd_tlb_entry(tlb, pmdp, address) \
> do { \
> - __tlb_adjust_range(tlb, address, HPAGE_PMD_SIZE); \
> - tlb->cleared_pmds = 1; \
> + tlb_flush_pmd_range(tlb, address, HPAGE_PMD_SIZE); \
> __tlb_remove_pmd_tlb_entry(tlb, pmdp, address); \
> } while (0)
>
> @@ -565,8 +591,7 @@ static inline void tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vm
>
> #define tlb_remove_pud_tlb_entry(tlb, pudp, address) \
> do { \
> - __tlb_adjust_range(tlb, address, HPAGE_PUD_SIZE); \
> - tlb->cleared_puds = 1; \
> + tlb_flush_pud_range(tlb, address, HPAGE_PUD_SIZE); \
> __tlb_remove_pud_tlb_entry(tlb, pudp, address); \
> } while (0)
>
>
> .
>
next prev parent reply other threads:[~2020-04-02 11:24 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-03-31 14:29 [RFC PATCH v5 0/8] arm64: tlb: add support for TTL feature Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 1/8] arm64: Detect the ARMv8.4 " Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 2/8] arm64: Add level-hinted TLB invalidation helper Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 3/8] arm64: Add tlbi_user_level " Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 4/8] mm: tlb: Pass struct mmu_gather to flush_pmd_tlb_range Zhenyu Ye
2020-03-31 15:13 ` Peter Zijlstra
2020-04-01 8:51 ` Zhenyu Ye
2020-04-01 12:20 ` Peter Zijlstra
2020-04-02 11:24 ` Zhenyu Ye [this message]
2020-04-02 16:38 ` Peter Zijlstra
2020-04-03 5:14 ` Zhenyu Ye
2020-04-08 9:00 ` Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 5/8] mm: tlb: Pass struct mmu_gather to flush_pud_tlb_range Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 6/8] mm: tlb: Pass struct mmu_gather to flush_hugetlb_tlb_range Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 7/8] mm: tlb: Pass struct mmu_gather to flush_tlb_range Zhenyu Ye
2020-03-31 14:29 ` [RFC PATCH v5 8/8] arm64: tlb: Set the TTL field in flush_tlb_range Zhenyu Ye
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=53675fb9-21c7-5309-07b8-1bbc1e775f9b@huawei.com \
--to=yezhenyu2@huawei.com \
--cc=Dave.Martin@arm.com \
--cc=akpm@linux-foundation.org \
--cc=aneesh.kumar@linux.ibm.com \
--cc=arm@kernel.org \
--cc=arnd@arndb.de \
--cc=broonie@kernel.org \
--cc=catalin.marinas@arm.com \
--cc=corbet@lwn.net \
--cc=guohanjun@huawei.com \
--cc=kuhn.chenqun@huawei.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mark.rutland@arm.com \
--cc=maz@kernel.org \
--cc=npiggin@gmail.com \
--cc=peterz@infradead.org \
--cc=prime.zeng@hisilicon.com \
--cc=rostedt@goodmis.org \
--cc=steven.price@arm.com \
--cc=suzuki.poulose@arm.com \
--cc=tglx@linutronix.de \
--cc=tony.luck@intel.com \
--cc=vgupta@synopsys.com \
--cc=will@kernel.org \
--cc=xiexiangyou@huawei.com \
--cc=yuzhao@google.com \
--cc=zhangshaokun@hisilicon.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).