Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210730043217.953384-1-aik@ozlabs.ru> <YQklgq4NkL4UToVY@kroah.com>
 <CABgObfb+M9Qeow1EZy+eQwM1jwoZY3zdPJfZW+Q+MoWmkaqcFw@mail.gmail.com> <YQlAl7/GSHWwkzEj@kroah.com>
In-Reply-To: <YQlAl7/GSHWwkzEj@kroah.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
Date:   Tue, 3 Aug 2021 15:29:16 +0200
Message-ID: <CABgObfamvoMai1b9hrPkt1uwgw0kozhU5V_Vvk1__k_sGnx4LA@mail.gmail.com>
Subject: Re: [RFC PATCH kernel] KVM: Stop leaking memory in debugfs
To:     Greg KH <gregkh@linuxfoundation.org>
Cc:     Alexey Kardashevskiy <aik@ozlabs.ru>,
        "Kernel Mailing List, Linux" <linux-kernel@vger.kernel.org>,
        kvm <kvm@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

> So userspace can create kvm resources with duplicate names?  That feels
> wrong to me.

Yes, the name is just the (pid, file descriptor). It's used only for
debugfs, and it's not really likely going to happen unless a program
does it on purpose, but the ugliness/wrongness is one of the reasons
why we now have a non-debugfs mechanism to retrieve the stats.

> But if all that is "duplicate" is the debugfs kvm directory, why not ask
> debugfs if it is already present before trying to create it again?  That
> way you will not have debugfs complain about duplicate
> files/directories.

That would also be racy; it would need a simple mutex around
debugfs_lookup and debugfs_create_dir. But it would indeed avoid the
complaints altogether, so I'll prepare a patch.  Thanks,

Paolo