Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 28 Sep 2022 05:31:58 -0400
From:   "Michael S. Tsirkin" <mst@redhat.com>
To:     Stefano Garzarella <sgarzare@redhat.com>
Cc:     Junichi Uekawa <uekawa@chromium.org>,
        Stefan Hajnoczi <stefanha@redhat.com>,
        Jason Wang <jasowang@redhat.com>,
        Eric Dumazet <edumazet@google.com>, davem@davemloft.net,
        netdev@vger.kernel.org, Jakub Kicinski <kuba@kernel.org>,
        virtualization@lists.linux-foundation.org, kvm@vger.kernel.org,
        Paolo Abeni <pabeni@redhat.com>, linux-kernel@vger.kernel.org,
        Bobby Eshleman <bobby.eshleman@gmail.com>
Subject: Re: [PATCH] vhost/vsock: Use kvmalloc/kvfree for larger packets.
Message-ID: <20220928052738-mutt-send-email-mst@kernel.org>
References: <20220928064538.667678-1-uekawa@chromium.org>
 <20220928082823.wyxplop5wtpuurwo@sgarzare-redhat>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220928082823.wyxplop5wtpuurwo@sgarzare-redhat>
Precedence: bulk

On Wed, Sep 28, 2022 at 10:28:23AM +0200, Stefano Garzarella wrote:
> On Wed, Sep 28, 2022 at 03:45:38PM +0900, Junichi Uekawa wrote:
> > When copying a large file over sftp over vsock, data size is usually 32kB,
> > and kmalloc seems to fail to try to allocate 32 32kB regions.
> > 
> > Call Trace:
> >  [<ffffffffb6a0df64>] dump_stack+0x97/0xdb
> >  [<ffffffffb68d6aed>] warn_alloc_failed+0x10f/0x138
> >  [<ffffffffb68d868a>] ? __alloc_pages_direct_compact+0x38/0xc8
> >  [<ffffffffb664619f>] __alloc_pages_nodemask+0x84c/0x90d
> >  [<ffffffffb6646e56>] alloc_kmem_pages+0x17/0x19
> >  [<ffffffffb6653a26>] kmalloc_order_trace+0x2b/0xdb
> >  [<ffffffffb66682f3>] __kmalloc+0x177/0x1f7
> >  [<ffffffffb66e0d94>] ? copy_from_iter+0x8d/0x31d
> >  [<ffffffffc0689ab7>] vhost_vsock_handle_tx_kick+0x1fa/0x301 [vhost_vsock]
> >  [<ffffffffc06828d9>] vhost_worker+0xf7/0x157 [vhost]
> >  [<ffffffffb683ddce>] kthread+0xfd/0x105
> >  [<ffffffffc06827e2>] ? vhost_dev_set_owner+0x22e/0x22e [vhost]
> >  [<ffffffffb683dcd1>] ? flush_kthread_worker+0xf3/0xf3
> >  [<ffffffffb6eb332e>] ret_from_fork+0x4e/0x80
> >  [<ffffffffb683dcd1>] ? flush_kthread_worker+0xf3/0xf3
> > 
> > Work around by doing kvmalloc instead.
> > 
> > Signed-off-by: Junichi Uekawa <uekawa@chromium.org>

My worry here is that this in more of a work around.
It would be better to not allocate memory so aggressively:
if we are so short on memory we should probably process
packets one at a time. Is that very hard to implement?


> > ---
> > 
> > drivers/vhost/vsock.c                   | 2 +-
> > net/vmw_vsock/virtio_transport_common.c | 2 +-
> > 2 files changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/vhost/vsock.c b/drivers/vhost/vsock.c
> > index 368330417bde..5703775af129 100644
> > --- a/drivers/vhost/vsock.c
> > +++ b/drivers/vhost/vsock.c
> > @@ -393,7 +393,7 @@ vhost_vsock_alloc_pkt(struct vhost_virtqueue *vq,
> > 		return NULL;
> > 	}
> > 
> > -	pkt->buf = kmalloc(pkt->len, GFP_KERNEL);
> > +	pkt->buf = kvmalloc(pkt->len, GFP_KERNEL);
> > 	if (!pkt->buf) {
> > 		kfree(pkt);
> > 		return NULL;
> > diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
> > index ec2c2afbf0d0..3a12aee33e92 100644
> > --- a/net/vmw_vsock/virtio_transport_common.c
> > +++ b/net/vmw_vsock/virtio_transport_common.c
> > @@ -1342,7 +1342,7 @@ EXPORT_SYMBOL_GPL(virtio_transport_recv_pkt);
> > 
> > void virtio_transport_free_pkt(struct virtio_vsock_pkt *pkt)
> > {
> > -	kfree(pkt->buf);
> > +	kvfree(pkt->buf);
> 
> virtio_transport_free_pkt() is used also in virtio_transport.c and
> vsock_loopback.c where pkt->buf is allocated with kmalloc(), but IIUC
> kvfree() can be used with that memory, so this should be fine.
> 
> > 	kfree(pkt);
> > }
> > EXPORT_SYMBOL_GPL(virtio_transport_free_pkt);
> > -- 
> > 2.37.3.998.g577e59143f-goog
> > 
> 
> This issue should go away with the Bobby's work about introducing sk_buff
> [1], but we can queue this for now.
> 
> I'm not sure if we should do the same also in the virtio-vsock driver
> (virtio_transport.c). Here in vhost-vsock the buf allocated is only used in
> the host, while in the virtio-vsock driver the buffer is exposed to the
> device emulated in the host, so it should be physically contiguous (if not,
> maybe we need to adjust virtio_vsock_rx_fill()).

More importantly it needs to support DMA API which IIUC kvmalloc
memory does not.

> So for now I think is fine to use kvmalloc only on vhost-vsock (eventually
> we can use it also in vsock_loopback), since the Bobby's patch should rework
> this code:
> 
> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
> 
> [1] https://lore.kernel.org/lkml/65d117ddc530d12a6d47fcc45b38891465a90d9f.1660362668.git.bobby.eshleman@bytedance.com/
> 
> Thanks,
> Stefano