Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.37 as permitted sender) client-ip=23.128.96.37;
Date:   Tue, 26 Sep 2023 13:13:51 -0400
From:   "Michael S. Tsirkin" <mst@redhat.com>
To:     Halil Pasic <pasic@linux.ibm.com>
Cc:     "Gonglei (Arei)" <arei.gonglei@huawei.com>,
        Herbert Xu <herbert@gondor.apana.org.au>,
        "linux-crypto@vger.kernel.org" <linux-crypto@vger.kernel.org>,
        Marc Hartmayer <mhartmay@linux.ibm.com>,
        Jason Wang <jasowang@redhat.com>,
        "virtualization@lists.linux-foundation.org" 
        <virtualization@lists.linux-foundation.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "pizhenwei@bytedance.com" <pizhenwei@bytedance.com>,
        Cornelia Huck <cohuck@redhat.com>
Subject: Re: [PATCH] crypto: virtio-crypto: call finalize with bh disabled
Message-ID: <20230926130521-mutt-send-email-mst@kernel.org>
References: <1914739e2de14ed396e5674aa2d4766c@huawei.com>
 <20230926184158.4ca2c0c3.pasic@linux.ibm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230926184158.4ca2c0c3.pasic@linux.ibm.com>
Precedence: bulk

On Tue, Sep 26, 2023 at 06:41:58PM +0200, Halil Pasic wrote:
> [..]
> > --- a/drivers/crypto/virtio/virtio_crypto_akcipher_algs.c
> > +++ b/drivers/crypto/virtio/virtio_crypto_akcipher_algs.c
> > @@ -61,8 +61,9 @@ static void virtio_crypto_akcipher_finalize_req(
> >  	vc_akcipher_req->src_buf = NULL;
> >  	vc_akcipher_req->dst_buf = NULL;
> >  	virtcrypto_clear_request(&vc_akcipher_req->base);
> > -
> > +	local_bh_disable();
> >  	crypto_finalize_akcipher_request(vc_akcipher_req->base.dataq->engine, req, err);
> > +	local_bh_enable();
> 
> Thanks Gonglei!
> 
> I did this a quick spin, and it does not seem to be sufficient on s390x.
> Which does not come as a surprise to me, because 
> 
> #define lockdep_assert_in_softirq()                                     \
> do {                                                                    \
>         WARN_ON_ONCE(__lockdep_enabled                  &&              \
>                      (!in_softirq() || in_irq() || in_nmi()));          \
> } while (0)
> 
> will still warn because  in_irq() still evaluates to true (your patch
> addresses the !in_softirq() part).
> 
> I don't have any results on x86 yet. My current understanding is that the
> virtio-pci transport code disables interrupts locally somewhere in the
> call chain (actually in vp_vring_interrupt() via spin_lock_irqsave())
> and then x86 would be fine. But I will get that verified.
> 
> On the other hand virtio_airq_handler() calls vring_interrupt() with
> interrupts enabled. (While vring_interrupt() is called in a (read)
> critical section in virtio_airq_handler() we use read_lock() and
> not read_lock_irqsave() to grab the lock. Whether that is correct in
> it self (i.e. disregarding the crypto problem) or not I'm not sure right
> now. Will think some more about it tomorrow.) If the way to go forward
> is disabling interrupts in virtio-ccw before vring_interrupt() is
> called, I would be glad to spin a patch for that.
> 
> Copying Conny, as she may have an opinion on this (if I'm not wrong she
> authored that code).
> 
> Regards,
> Halil

On a related note, config change callback is also handled incorrectly
in this driver, it takes a mutex from interrupt context.