Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:4 as permitted sender) client-ip=2620:137:e000::3:4;
Date:   Fri, 13 Oct 2023 09:55:33 -0700
In-Reply-To: <7c4b1c78-de74-5fff-7327-0863f403eb7e@redhat.com>
Mime-Version: 1.0
References: <ZSQO4fHaAxDkbGyz@google.com> <20231009200608.GJ800259@ZenIV>
 <ZSRgdgQe3fseEQpf@google.com> <20231009204037.GK800259@ZenIV>
 <ZSRwDItBbsn2IfWl@google.com> <20231010000910.GM800259@ZenIV>
 <ZSSaWPc5wjU9k1Kw@google.com> <20231010003746.GN800259@ZenIV>
 <ZSXeipdJcWZjLx8k@google.com> <7c4b1c78-de74-5fff-7327-0863f403eb7e@redhat.com>
Message-ID: <ZSl2hdfF8XSXss3h@google.com>
Subject: Re: [PATCH gmem FIXUP] kvm: guestmem: do not use a file system
From:   Sean Christopherson <seanjc@google.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Al Viro <viro@zeniv.linux.org.uk>, linux-kernel@vger.kernel.org,
        kvm@vger.kernel.org
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On Wed, Oct 11, 2023, Paolo Bonzini wrote:
> Your patch 2 looks good, but perhaps instead of setting the owner we could
> stash the struct module* in a global, and try_get/put it from open and
> release respectively?  That is, .owner keeps the kvm module alive and the
> kvm module keeps kvm-intel/kvm-amd alive.  That would subsume patches 1 and 3.

I don't think that would be a net positive.  We'd have to implement .open() for
several file types just to get a reference to the sub-module.  At that point, the
odds of forgetting to implement .open() are about the same as forgetting to set
.owner when adding a new file type, e.g. guest_memfd.