From: Zhenyu Ye <yezhenyu2@huawei.com>
To: Marc Zyngier <maz@kernel.org>
Cc: <will@kernel.org>, <mark.rutland@arm.com>,
<catalin.marinas@arm.com>, <aneesh.kumar@linux.ibm.com>,
<akpm@linux-foundation.org>, <npiggin@gmail.com>,
<peterz@infradead.org>, <arnd@arndb.de>, <rostedt@goodmis.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>, <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>
Subject: Re: [RFC PATCH v4 5/6] arm64: tlb: Use translation level hint in vm_flags
Date: Thu, 26 Mar 2020 15:11:23 +0800 [thread overview]
Message-ID: <6e2fece0-a7d7-a470-6911-5891fd09a140@huawei.com> (raw)
In-Reply-To: <2f4cb3ef52c5589b388225e487651a2b@kernel.org>
Hi Marc,
On 2020/3/25 22:13, Marc Zyngier wrote:
>>>>
>>>> +inline unsigned int get_vma_level(struct vm_area_struct *vma)
>>>> +{
>>>> + unsigned int level = 0;
>>>> + if (vma->vm_flags & VM_LEVEL_PUD)
>>>> + level = 1;
>>>> + else if (vma->vm_flags & VM_LEVEL_PMD)
>>>> + level = 2;
>>>> + else if (vma->vm_flags & VM_LEVEL_PTE)
>>>> + level = 3;
>>>> +
>>>> + vma->vm_flags &= ~(VM_LEVEL_PUD | VM_LEVEL_PMD | VM_LEVEL_PTE);
>>>> + return level;
>>>> +}
>>>> +
>>>> void set_swapper_pgd(pgd_t *pgdp, pgd_t pgd)
>>>> {
>>>> pgd_t *fixmap_pgdp;
>>>
>>>
>>> If feels bizarre a TLBI is now a destructive operation: you've lost the
>>> flags by having cleared them. Even if that's not really a problem in
>>> practice (you issue TLBI because you've unmapped the VMA), it remains
>>> that the act of invalidating TLBs isn't expected to change a kernel
>>> structure (and I'm not even thinking about potential races here).
>>
>> I cleared vm_flags here just out of caution, because every TLBI flush
>> action should set vm_flags themself. As I know, the TLB_LEVEL of an vma
>> will only be changed by transparent hugepage collapse and merge when
>> the page is not mapped, so there may not have potential races.
>>
>> But you are right that TLBI should't change a kernel structure.
>> I will remove the clear action and add some notices here.
>
> More than that. You are changing the VMA flags at TLBI time already,
> when calling get_vma_level(). That is already unacceptable -- I don't
> think (and Peter will certainly correct me if I'm wrong) that you
> are allowed to change the flags on that code path, as you probably
> don't hold the write_lock.
>
Thanks for your review. I will avoid this problem in next version.
>>> If anything, I feel this should be based around the mmu_gather
>>> structure, which already tracks the right level of information and
>>> additionally knows about the P4D level which is missing in your patches
>>> (even if arm64 is so far limited to 4 levels).
>>>
>>> Thanks,
>>>
>>> M.
>>>
>>
>> mmu_gather structure has tracked the level information, but we can only
>> use the info in the tlb_flush interface. If we want to use the info in
>> flush_tlb_range, we may should have to add a parameter to this interface,
>> such as:
>>
>> flush_tlb_range(vma, start, end, flags);
>>
>> However, the flush_tlb_range is a common interface to all architectures,
>> I'm not sure if this is feasible because this feature is only supported
>> by ARM64 currently.
>
> You could always build an on-stack mmu_gather structure and pass it down
> to the TLB invalidation helpers.
>
> And frankly, you are not going to be able to fit such a change in the
> way Linux deals with TLBs by adding hacks at the periphery. You'll need
> to change some of the core abstractions.
>
> Finally, as Peter mentioned separately, there is Power9 which has similar
> instructions, and could make use of it too. So that's yet another reason
> to stop hacking at the arch level.
>
OK, I will try to add struct mmu_gather to flush_tlb_range, such as:
void flush_tlb_range(struct mmu_gather *tlb,
struct vm_area_struct *vma,
unsigned long start, unsigned long end);
This will involve all architectures, I will do it carefully.
>>
>> Or can we ignore the flush_tlb_range and only set the TTL field in
>> tlb_flush? Waiting for your suggestion...
>
> You could, but you could also ignore TTL altogether (what's the point
> in only having half of it?). See my suggestion above.
>
>> For P4D level, the TTL field is limited to 4 bit(2 for translation granule
>> and 2 for page table level), so we can no longer afford more levels :).
>
> You clearly didn't read the bit that said "even if arm64 is so far limited
> to 4 levels". But again, this is Linux, a multi-architecture kernel, and
> not an arm64-special. Changes you make have to work for all architectures,
> and be extensible enough for future changes.
>
Using the struct mmu_gather to pass the TTL value will not have
this problem :).
Thanks,
Zhenyu
next prev parent reply other threads:[~2020-03-26 7:12 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-03-24 13:45 [RFC PATCH v4 0/6] arm64: tlb: add support for TTL feature Zhenyu Ye
2020-03-24 13:45 ` [RFC PATCH v4 1/6] arm64: Detect the ARMv8.4 " Zhenyu Ye
2020-03-24 13:45 ` [RFC PATCH v4 2/6] arm64: Add level-hinted TLB invalidation helper Zhenyu Ye
2020-03-24 13:45 ` [RFC PATCH v4 3/6] arm64: Add level-hinted TLB invalidation helper to tlbi_user Zhenyu Ye
2020-03-24 14:19 ` Marc Zyngier
2020-03-25 2:47 ` Zhenyu Ye
2020-03-24 13:45 ` [RFC PATCH v4 4/6] mm: Add page table level flags to vm_flags Zhenyu Ye
2020-03-24 18:44 ` Steven Rostedt
2020-03-25 2:10 ` yezhenyu (A)
2020-03-24 13:45 ` [RFC PATCH v4 5/6] arm64: tlb: Use translation level hint in vm_flags Zhenyu Ye
2020-03-24 14:45 ` Marc Zyngier
2020-03-25 8:00 ` Zhenyu Ye
2020-03-25 14:13 ` Marc Zyngier
2020-03-26 7:11 ` Zhenyu Ye [this message]
2020-03-24 13:45 ` [RFC PATCH v4 6/6] mm: Set VM_LEVEL flags in some tlb_flush functions Zhenyu Ye
2020-03-24 15:01 ` [RFC PATCH v4 0/6] arm64: tlb: add support for TTL feature Peter Zijlstra
2020-03-25 4:49 ` Zhenyu Ye
2020-03-25 13:32 ` Peter Zijlstra
2020-03-26 7:15 ` Zhenyu Ye
2020-03-25 16:15 ` James Morse
2020-03-25 16:41 ` Peter Zijlstra
2020-03-26 6:45 ` 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=6e2fece0-a7d7-a470-6911-5891fd09a140@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=guohanjun@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=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).