Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:1 as permitted sender) client-ip=2620:137:e000::3:1;
Date:   Tue, 10 Oct 2023 01:37:46 +0100
From:   Al Viro <viro@zeniv.linux.org.uk>
To:     Sean Christopherson <seanjc@google.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH gmem FIXUP] kvm: guestmem: do not use a file system
Message-ID: <20231010003746.GN800259@ZenIV>
References: <20230928180651.1525674-1-pbonzini@redhat.com>
 <169595365500.1386813.6579237770749312873.b4-ty@google.com>
 <20231009022248.GD800259@ZenIV>
 <ZSQO4fHaAxDkbGyz@google.com>
 <20231009200608.GJ800259@ZenIV>
 <ZSRgdgQe3fseEQpf@google.com>
 <20231009204037.GK800259@ZenIV>
 <ZSRwDItBbsn2IfWl@google.com>
 <20231010000910.GM800259@ZenIV>
 <ZSSaWPc5wjU9k1Kw@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZSSaWPc5wjU9k1Kw@google.com>
Sender: Al Viro <viro@ftp.linux.org.uk>
Precedence: bulk

On Mon, Oct 09, 2023 at 05:27:04PM -0700, Sean Christopherson wrote:

> If the last reference is effectively held by guest_memfd, it would be:
> 
>   kvm_gmem_release(), a.k.a. file_operations.release()
>   |
>   -> kvm_put_kvm()
>      |
>      -> kvm_destroy_vm()
>         |
>         -> module_put(kvm_chardev_ops.owner);

... and now your thread gets preempted and loses CPU; before you get
it back, some joker calls delete_module(), and page of code containing
kvm_gmem_release() is unmapped.  Even though an address within that
page is stored as return address in a frame on your thread's stack.
That thread gets the timeslice again and proceeds to return into
unmapped page.  Oops...