From: David Hildenbrand <david@redhat.com>
To: Lan Tianyu <tianyu.lan@intel.com>
Cc: pbonzini@redhat.com, rkrcmar@redhat.com, kvm@vger.kernel.org,
linux-kernel@vger.kernel.org, dvyukov@google.com,
kernellwp@gmail.com
Subject: Re: [PATCH] KVM/Eventfd: Avoid crash when assign and deassign same eventfd in parallel.
Date: Mon, 18 Dec 2017 09:30:23 +0100 [thread overview]
Message-ID: <d47054f9-16ee-5ef2-4024-935683ce8dfa@redhat.com> (raw)
In-Reply-To: <1513554007-12302-1-git-send-email-tianyu.lan@intel.com>
On 18.12.2017 00:40, Lan Tianyu wrote:
> Syzroot reports crash in kvm_irqfd_assign() is caused by use-after-free.
> Because kvm_irqfd_assign() and kvm_irqfd_deassign() can't run in parallel
> for same eventfd. When assign path hasn't been finished after irqfd
> has been added to kvm->irqfds.items list, another thead may deassign the
> eventfd and free struct kvm_kernel_irqfd(). This causes assign path still
> uses struct kvm_kernel_irqfd freed by deassign path. To avoid such issue,
> add "initialized" flag in the struct kvm_kernel_irqfd and check the flag before
> deactivating irqfd. If irqfd is still in initialization, deassign path
> return fault.>
> Reported-by: Dmitry Vyukov <dvyukov@google.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Radim Krčmář <rkrcmar@redhat.com>
> Cc: Dmitry Vyukov <dvyukov@google.com>
> Cc: Wanpeng Li <kernellwp@gmail.com>
> Signed-off-by: Tianyu Lan <tianyu.lan@intel.com>
> ---
> include/linux/kvm_irqfd.h | 1 +
> virt/kvm/eventfd.c | 11 +++++++++--
> 2 files changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/kvm_irqfd.h b/include/linux/kvm_irqfd.h
> index 76c2fbc..be6b254 100644
> --- a/include/linux/kvm_irqfd.h
> +++ b/include/linux/kvm_irqfd.h
> @@ -66,6 +66,7 @@ struct kvm_kernel_irqfd {
> struct work_struct shutdown;
> struct irq_bypass_consumer consumer;
> struct irq_bypass_producer *producer;
> + u8 initialized:1;
> };
>
> #endif /* __LINUX_KVM_IRQFD_H */
> diff --git a/virt/kvm/eventfd.c b/virt/kvm/eventfd.c
> index a334399..80f06e6 100644
> --- a/virt/kvm/eventfd.c
> +++ b/virt/kvm/eventfd.c
> @@ -421,6 +421,7 @@ int __attribute__((weak)) kvm_arch_update_irqfd_routing(
> }
> #endif
>
> + irqfd->initialized = 1;
The ugly thing in kvm_irqfd_assign() is that we access irqfd without
holding a lock. I think that should rather be fixed than working around
that issue. (e.g. lock() -> lookup again -> verify still in list ->
unlock())
> return 0;
>
> fail:
> @@ -525,6 +526,7 @@ void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
> {
Which tool are you using to generate diffs? git format-patch?
Mentioning, because the indicated function here .... (kvm_irqfd_deassign)
> struct kvm_kernel_irqfd *irqfd, *tmp;
> struct eventfd_ctx *eventfd;
> + int ret = 0;
>
> eventfd = eventfd_ctx_fdget(args->fd);
> if (IS_ERR(eventfd))
> @@ -543,7 +545,12 @@ void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
> write_seqcount_begin(&irqfd->irq_entry_sc);
> irqfd->irq_entry.type = 0;
> write_seqcount_end(&irqfd->irq_entry_sc);
> - irqfd_deactivate(irqfd);
> +
> + if (irqfd->initialized)
> + irqfd_deactivate(irqfd);
> + else
> + ret = -EFAULT;
> +
> }
> }
>
> @@ -557,7 +564,7 @@ void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
> */
and here are wrong and misleading (kvm_irqfd_deassig). (and just noticed
also in the second hunk)
> flush_workqueue(irqfd_cleanup_wq);
>
> - return 0;
> + return ret;
> }
>
> int
>
--
Thanks,
David / dhildenb
next prev parent reply other threads:[~2017-12-18 8:30 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-12-17 23:40 [PATCH] KVM/Eventfd: Avoid crash when assign and deassign same eventfd in parallel Lan Tianyu
2017-12-18 8:30 ` David Hildenbrand [this message]
2017-12-18 8:50 ` Paolo Bonzini
2017-12-18 9:08 ` David Hildenbrand
2017-12-18 9:24 ` Paolo Bonzini
2017-12-19 6:35 ` Lan Tianyu
2017-12-19 10:41 ` Paolo Bonzini
2017-12-20 8:48 ` Lan Tianyu
2017-12-25 8:03 ` Peter Xu
2017-12-19 6:21 ` Lan Tianyu
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=d47054f9-16ee-5ef2-4024-935683ce8dfa@redhat.com \
--to=david@redhat.com \
--cc=dvyukov@google.com \
--cc=kernellwp@gmail.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=rkrcmar@redhat.com \
--cc=tianyu.lan@intel.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