All of lore.kernel.org
 help / color / mirror / Atom feed
From: Will Deacon <will@kernel.org>
To: Sean Christopherson <seanjc@google.com>
Cc: Alexandru Elisei <alexandru.elisei@arm.com>,
	Marc Zyngier <maz@kernel.org>,
	linux-arm-kernel@lists.infradead.org, kvm@vger.kernel.org,
	kvmarm@lists.cs.columbia.edu, linux-mm@kvack.org,
	Matthew Wilcox <willy@infradead.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Quentin Perret <qperret@google.com>,
	James Morse <james.morse@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	kernel-team@android.com
Subject: Re: [PATCH 1/5] KVM: arm64: Walk userspace page tables to compute the THP mapping size
Date: Wed, 21 Jul 2021 15:58:29 +0100	[thread overview]
Message-ID: <20210721145828.GA11003@willie-the-truck> (raw)
In-Reply-To: <YPczKoLqlKElLxzb@google.com>

Hey Sean,

On Tue, Jul 20, 2021 at 08:33:46PM +0000, Sean Christopherson wrote:
> On Tue, Jul 20, 2021, Alexandru Elisei wrote:
> > I just can't figure out why having the mmap lock is not needed to walk the
> > userspace page tables. Any hints? Or am I not seeing where it's taken?
> 
> Disclaimer: I'm not super familiar with arm64's page tables, but the relevant KVM
> functionality is common across x86 and arm64.

No need for the disclaimer, there are so many moving parts here that I don't
think it's possible to be familiar with them all! Thanks for taking the time
to write it up so clearly.

> KVM arm64 (and x86) unconditionally registers a mmu_notifier for the mm_struct
> associated with the VM, and disallows calling ioctls from a different process,
> i.e. walking the page tables during KVM_RUN is guaranteed to use the mm for which
> KVM registered the mmu_notifier.  As part of registration, the mmu_notifier
> does mmgrab() and doesn't do mmdrop() until it's unregistered.  That ensures the
> mm_struct itself is live.
> 
> For the page tables liveliness, KVM implements mmu_notifier_ops.release, which is
> invoked at the beginning of exit_mmap(), before the page tables are freed.  In
> its implementation, KVM takes mmu_lock and zaps all its shadow page tables, a.k.a.
> the stage2 tables in KVM arm64.  The flow in question, get_user_mapping_size(),
> also runs under mmu_lock, and so effectively blocks exit_mmap() and thus is
> guaranteed to run with live userspace tables.

Unless I missed a case, exit_mmap() only runs when mm_struct::mm_users drops
to zero, right? The vCPU tasks should hold references to that afaict, so I
don't think it should be possible for exit_mmap() to run while there are
vCPUs running with the corresponding page-table.

> Looking at the arm64 code, one thing I'm not clear on is whether arm64 correctly
> handles the case where exit_mmap() wins the race.  The invalidate_range hooks will
> still be called, so userspace page tables aren't a problem, but
> kvm_arch_flush_shadow_all() -> kvm_free_stage2_pgd() nullifies mmu->pgt without
> any additional notifications that I see.  x86 deals with this by ensuring its
> top-level TDP entry (stage2 equivalent) is valid while the page fault handler is
> running.

But the fact that x86 handles this race has me worried. What am I missing?

I agree that, if the race can occur, we don't appear to handle it in the
arm64 backend.

Cheers,

Will

WARNING: multiple messages have this Message-ID
From: Will Deacon <will@kernel.org>
To: Sean Christopherson <seanjc@google.com>
Cc: kernel-team@android.com, kvm@vger.kernel.org,
	Marc Zyngier <maz@kernel.org>,
	Matthew Wilcox <willy@infradead.org>,
	linux-mm@kvack.org, Paolo Bonzini <pbonzini@redhat.com>,
	kvmarm@lists.cs.columbia.edu,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH 1/5] KVM: arm64: Walk userspace page tables to compute the THP mapping size
Date: Wed, 21 Jul 2021 15:58:29 +0100	[thread overview]
Message-ID: <20210721145828.GA11003@willie-the-truck> (raw)
In-Reply-To: <YPczKoLqlKElLxzb@google.com>

Hey Sean,

On Tue, Jul 20, 2021 at 08:33:46PM +0000, Sean Christopherson wrote:
> On Tue, Jul 20, 2021, Alexandru Elisei wrote:
> > I just can't figure out why having the mmap lock is not needed to walk the
> > userspace page tables. Any hints? Or am I not seeing where it's taken?
> 
> Disclaimer: I'm not super familiar with arm64's page tables, but the relevant KVM
> functionality is common across x86 and arm64.

No need for the disclaimer, there are so many moving parts here that I don't
think it's possible to be familiar with them all! Thanks for taking the time
to write it up so clearly.

> KVM arm64 (and x86) unconditionally registers a mmu_notifier for the mm_struct
> associated with the VM, and disallows calling ioctls from a different process,
> i.e. walking the page tables during KVM_RUN is guaranteed to use the mm for which
> KVM registered the mmu_notifier.  As part of registration, the mmu_notifier
> does mmgrab() and doesn't do mmdrop() until it's unregistered.  That ensures the
> mm_struct itself is live.
> 
> For the page tables liveliness, KVM implements mmu_notifier_ops.release, which is
> invoked at the beginning of exit_mmap(), before the page tables are freed.  In
> its implementation, KVM takes mmu_lock and zaps all its shadow page tables, a.k.a.
> the stage2 tables in KVM arm64.  The flow in question, get_user_mapping_size(),
> also runs under mmu_lock, and so effectively blocks exit_mmap() and thus is
> guaranteed to run with live userspace tables.

Unless I missed a case, exit_mmap() only runs when mm_struct::mm_users drops
to zero, right? The vCPU tasks should hold references to that afaict, so I
don't think it should be possible for exit_mmap() to run while there are
vCPUs running with the corresponding page-table.

> Looking at the arm64 code, one thing I'm not clear on is whether arm64 correctly
> handles the case where exit_mmap() wins the race.  The invalidate_range hooks will
> still be called, so userspace page tables aren't a problem, but
> kvm_arch_flush_shadow_all() -> kvm_free_stage2_pgd() nullifies mmu->pgt without
> any additional notifications that I see.  x86 deals with this by ensuring its
> top-level TDP entry (stage2 equivalent) is valid while the page fault handler is
> running.

But the fact that x86 handles this race has me worried. What am I missing?

I agree that, if the race can occur, we don't appear to handle it in the
arm64 backend.

Cheers,

Will
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

WARNING: multiple messages have this Message-ID
From: Will Deacon <will@kernel.org>
To: Sean Christopherson <seanjc@google.com>
Cc: Alexandru Elisei <alexandru.elisei@arm.com>,
	Marc Zyngier <maz@kernel.org>,
	linux-arm-kernel@lists.infradead.org, kvm@vger.kernel.org,
	kvmarm@lists.cs.columbia.edu, linux-mm@kvack.org,
	Matthew Wilcox <willy@infradead.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Quentin Perret <qperret@google.com>,
	James Morse <james.morse@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	kernel-team@android.com
Subject: Re: [PATCH 1/5] KVM: arm64: Walk userspace page tables to compute the THP mapping size
Date: Wed, 21 Jul 2021 15:58:29 +0100	[thread overview]
Message-ID: <20210721145828.GA11003@willie-the-truck> (raw)
In-Reply-To: <YPczKoLqlKElLxzb@google.com>

Hey Sean,

On Tue, Jul 20, 2021 at 08:33:46PM +0000, Sean Christopherson wrote:
> On Tue, Jul 20, 2021, Alexandru Elisei wrote:
> > I just can't figure out why having the mmap lock is not needed to walk the
> > userspace page tables. Any hints? Or am I not seeing where it's taken?
> 
> Disclaimer: I'm not super familiar with arm64's page tables, but the relevant KVM
> functionality is common across x86 and arm64.

No need for the disclaimer, there are so many moving parts here that I don't
think it's possible to be familiar with them all! Thanks for taking the time
to write it up so clearly.

> KVM arm64 (and x86) unconditionally registers a mmu_notifier for the mm_struct
> associated with the VM, and disallows calling ioctls from a different process,
> i.e. walking the page tables during KVM_RUN is guaranteed to use the mm for which
> KVM registered the mmu_notifier.  As part of registration, the mmu_notifier
> does mmgrab() and doesn't do mmdrop() until it's unregistered.  That ensures the
> mm_struct itself is live.
> 
> For the page tables liveliness, KVM implements mmu_notifier_ops.release, which is
> invoked at the beginning of exit_mmap(), before the page tables are freed.  In
> its implementation, KVM takes mmu_lock and zaps all its shadow page tables, a.k.a.
> the stage2 tables in KVM arm64.  The flow in question, get_user_mapping_size(),
> also runs under mmu_lock, and so effectively blocks exit_mmap() and thus is
> guaranteed to run with live userspace tables.

Unless I missed a case, exit_mmap() only runs when mm_struct::mm_users drops
to zero, right? The vCPU tasks should hold references to that afaict, so I
don't think it should be possible for exit_mmap() to run while there are
vCPUs running with the corresponding page-table.

> Looking at the arm64 code, one thing I'm not clear on is whether arm64 correctly
> handles the case where exit_mmap() wins the race.  The invalidate_range hooks will
> still be called, so userspace page tables aren't a problem, but
> kvm_arch_flush_shadow_all() -> kvm_free_stage2_pgd() nullifies mmu->pgt without
> any additional notifications that I see.  x86 deals with this by ensuring its
> top-level TDP entry (stage2 equivalent) is valid while the page fault handler is
> running.

But the fact that x86 handles this race has me worried. What am I missing?

I agree that, if the race can occur, we don't appear to handle it in the
arm64 backend.

Cheers,

Will

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

  reply	other threads:[~2021-07-21 14:58 UTC|newest]

Thread overview: 54+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-17  9:55 [PATCH 0/5] KVM: Remove kvm_is_transparent_hugepage() and friends Marc Zyngier
2021-07-17  9:55 ` Marc Zyngier
2021-07-17  9:55 ` Marc Zyngier
2021-07-17  9:55 ` [PATCH 1/5] KVM: arm64: Walk userspace page tables to compute the THP mapping size Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-19  6:31   ` Paolo Bonzini
2021-07-19  6:31     ` Paolo Bonzini
2021-07-19  6:31     ` Paolo Bonzini
2021-07-19  9:31     ` Marc Zyngier
2021-07-19  9:31       ` Marc Zyngier
2021-07-19  9:31       ` Marc Zyngier
2021-07-20 17:23   ` Alexandru Elisei
2021-07-20 17:23     ` Alexandru Elisei
2021-07-20 17:23     ` Alexandru Elisei
2021-07-20 20:33     ` Sean Christopherson
2021-07-20 20:33       ` Sean Christopherson
2021-07-20 20:33       ` Sean Christopherson
2021-07-21 14:58       ` Will Deacon [this message]
2021-07-21 14:58         ` Will Deacon
2021-07-21 14:58         ` Will Deacon
2021-07-21 15:56         ` Sean Christopherson
2021-07-21 15:56           ` Sean Christopherson
2021-07-21 15:56           ` Sean Christopherson
2021-07-21 16:37       ` Alexandru Elisei
2021-07-21 16:37         ` Alexandru Elisei
2021-07-21 16:37         ` Alexandru Elisei
2021-07-23  8:48     ` Marc Zyngier
2021-07-23  8:48       ` Marc Zyngier
2021-07-23  8:48       ` Marc Zyngier
2021-07-17  9:55 ` [PATCH 2/5] KVM: arm64: Avoid mapping size adjustment on permission fault Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-23 15:55   ` Alexandru Elisei
2021-07-23 15:55     ` Alexandru Elisei
2021-07-23 15:55     ` Alexandru Elisei
2021-07-23 16:18     ` Marc Zyngier
2021-07-23 16:18       ` Marc Zyngier
2021-07-23 16:18       ` Marc Zyngier
2021-07-17  9:55 ` [PATCH 3/5] KVM: Remove kvm_is_transparent_hugepage() and PageTransCompoundMap() Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-19  6:31   ` Paolo Bonzini
2021-07-19  6:31     ` Paolo Bonzini
2021-07-19  6:31     ` Paolo Bonzini
2021-07-17  9:55 ` [PATCH 4/5] KVM: arm64: Use get_page() instead of kvm_get_pfn() Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-17  9:55 ` [PATCH 5/5] KVM: Get rid " Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-17  9:55   ` Marc Zyngier
2021-07-19  6:31   ` Paolo Bonzini
2021-07-19  6:31     ` Paolo Bonzini
2021-07-19  6:31     ` Paolo Bonzini

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=20210721145828.GA11003@willie-the-truck \
    --to=will@kernel.org \
    --cc=alexandru.elisei@arm.com \
    --cc=james.morse@arm.com \
    --cc=kernel-team@android.com \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-mm@kvack.org \
    --cc=maz@kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=qperret@google.com \
    --cc=seanjc@google.com \
    --cc=suzuki.poulose@arm.com \
    --cc=willy@infradead.org \
    --subject='Re: [PATCH 1/5] KVM: arm64: Walk userspace page tables to compute the THP mapping size' \
    /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

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.