Received: by 2002:ad5:474a:0:0:0:0:0 with SMTP id i10csp6058598imu; Wed, 30 Jan 2019 08:07:19 -0800 (PST) X-Google-Smtp-Source: ALg8bN5IVAneGEyI5FWyCPPn8KHgruE8XAvMuFsjxTMzpEZsC7deo4rvPEatntwszlrQtMLff/40 X-Received: by 2002:a62:3811:: with SMTP id f17mr31684778pfa.206.1548864439096; Wed, 30 Jan 2019 08:07:19 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1548864439; cv=none; d=google.com; s=arc-20160816; b=BoTwarXxlWQ1JZ54jY1Fjyhn0wnmzLJT0JVTcRPBmnGPP3r1SmmsTccu6tJVmFCnA9 G1XIf7FrPkk7yRnHCRFbtBz3YHUIEmofEfHL5i9VBuoGp7zi6Is1sOZssBH9Q0aOIkJM CKu+hvnhnfu6cNCZPN/ftooz+O3fn430tT4cucqstdu0TNZpg9YK8DvNPsZcGhhoR827 mBCwf4VDqGVxT2Iq7Bfs4GmE9LxSRdpXQLiskXCurAqaSa6tN3NOwsC/pLhTtmKJtHwm BGZm41EJ6D/Dl4aZ3oga54v/69ys8f11O9q7koAgFYJPTY8sW7wkaVptX0oUrW1mOz0D Roug== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:sender:content-transfer-encoding:mime-version :references:in-reply-to:date:cc:to:from:subject:message-id; bh=9CabRwS5VKjd9JYz+akgJBbYn9SjgaFXYMQMSJ6IyZY=; b=bwqMFJWEu9aUPCSGlH0fpSrZdph7MC8wyKhimT5tG6Y8yUe+0B8SR3wQ+pRCMmZo8i eyZ3P3QKl0+r1JoIK/enf+m8efBqfOIVShRf0+kDbIUgvvzrHXkOIHLov+CyW0Pwc/fN Aw0g0l3mwxlAMXkPS+Bh9Xnp/20H/RZ3CttUT/iJOggmFftkerD6Suwn7ve6mU8xLu+1 NEDkCPTU3DFlLsQX8lQ5GgBbVF7n3jGfQRvNLQFaijAVBMg72Pv0hhgHKUIoFgx8W08A Gd7YuPu6HwOZ1D+65vIqs5SkDRkw6YGKB+OPrypUc16ITEaDVfxdcgZAZpxND2zY0Q+C lrbQ== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=intel.com Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id d23si1689034pgm.559.2019.01.30.08.07.03; Wed, 30 Jan 2019 08:07:19 -0800 (PST) Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729063AbfA3QEu (ORCPT + 99 others); Wed, 30 Jan 2019 11:04:50 -0500 Received: from mga01.intel.com ([192.55.52.88]:16741 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726230AbfA3QEt (ORCPT ); Wed, 30 Jan 2019 11:04:49 -0500 X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga006.jf.intel.com ([10.7.209.51]) by fmsmga101.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 30 Jan 2019 08:04:48 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.56,541,1539673200"; d="scan'208";a="112355401" Received: from unknown (HELO localhost.localdomain) ([10.10.37.145]) by orsmga006.jf.intel.com with ESMTP; 30 Jan 2019 08:04:47 -0800 Message-ID: <1548865797.29684.47.camel@intel.com> Subject: Re: [PATCH] mic: vop: Fix broken virtqueues From: Sudeep Dutt To: Vincent Whitchurch Cc: Sudeep Dutt , gregkh@linuxfoundation.org, arnd@arndb.de, ashutosh.dixit@intel.com, linux-kernel@vger.kernel.org, virtualization@lists.linux-foundation.org, tiwei.bie@intel.com, luto@kernel.org, Vincent Whitchurch Date: Wed, 30 Jan 2019 08:29:57 -0800 In-Reply-To: <20190129102207.9577-1-vincent.whitchurch@axis.com> References: <20190129102207.9577-1-vincent.whitchurch@axis.com> Content-Type: text/plain; charset="us-ascii" X-Mailer: Evolution 3.12.11 (3.12.11-15.el7) Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 2019-01-29 at 11:22 +0100, Vincent Whitchurch wrote: > VOP is broken in mainline since commit 1ce9e6055fa0a9043 ("virtio_ring: > introduce packed ring support"); attempting to use the virtqueues leads > to various kernel crashes. I'm testing it with my not-yet-merged > loopback patches, but even the in-tree MIC hardware cannot work. > > The problem is not in the referenced commit per se, but is due to the > following hack in vop_find_vq() which depends on the layout of private > structures in other source files, which that commit happened to change: > > /* > * To reassign the used ring here we are directly accessing > * struct vring_virtqueue which is a private data structure > * in virtio_ring.c. At the minimum, a BUILD_BUG_ON() in > * vring_new_virtqueue() would ensure that > * (&vq->vring == (struct vring *) (&vq->vq + 1)); > */ > vr = (struct vring *)(vq + 1); > vr->used = used; > > Fix vop by using __vring_new_virtqueue() to create the needed vring > layout from the start, instead of attempting to patch in the used ring > later. __vring_new_virtqueue() was added way back in commit > 2a2d1382fe9dcc ("virtio: Add improved queue allocation API") in order to > address mic's usecase, according to the commit message. > Thank you for fixing this up Vincent. Reviewed-by: Sudeep Dutt > Signed-off-by: Vincent Whitchurch > --- > drivers/misc/mic/vop/vop_main.c | 60 +++++++++++++++++++-------------- > 1 file changed, 34 insertions(+), 26 deletions(-) > > diff --git a/drivers/misc/mic/vop/vop_main.c b/drivers/misc/mic/vop/vop_main.c > index d2b9782eee87..fef45bf6d519 100644 > --- a/drivers/misc/mic/vop/vop_main.c > +++ b/drivers/misc/mic/vop/vop_main.c > @@ -284,6 +284,26 @@ static void vop_del_vqs(struct virtio_device *dev) > vop_del_vq(vq, idx++); > } > > +static struct virtqueue *vop_new_virtqueue(unsigned int index, > + unsigned int num, > + struct virtio_device *vdev, > + bool context, > + void *pages, > + bool (*notify)(struct virtqueue *vq), > + void (*callback)(struct virtqueue *vq), > + const char *name, > + void *used) > +{ > + bool weak_barriers = false; > + struct vring vring; > + > + vring_init(&vring, num, pages, MIC_VIRTIO_RING_ALIGN); > + vring.used = used; > + > + return __vring_new_virtqueue(index, vring, vdev, weak_barriers, context, > + notify, callback, name); > +} > + > /* > * This routine will assign vring's allocated in host/io memory. Code in > * virtio_ring.c however continues to access this io memory as if it were local > @@ -303,7 +323,6 @@ static struct virtqueue *vop_find_vq(struct virtio_device *dev, > struct _mic_vring_info __iomem *info; > void *used; > int vr_size, _vr_size, err, magic; > - struct vring *vr; > u8 type = ioread8(&vdev->desc->type); > > if (index >= ioread8(&vdev->desc->num_vq)) > @@ -322,17 +341,7 @@ static struct virtqueue *vop_find_vq(struct virtio_device *dev, > return ERR_PTR(-ENOMEM); > vdev->vr[index] = va; > memset_io(va, 0x0, _vr_size); > - vq = vring_new_virtqueue( > - index, > - le16_to_cpu(config.num), MIC_VIRTIO_RING_ALIGN, > - dev, > - false, > - ctx, > - (void __force *)va, vop_notify, callback, name); > - if (!vq) { > - err = -ENOMEM; > - goto unmap; > - } > + > info = va + _vr_size; > magic = ioread32(&info->magic); > > @@ -341,7 +350,6 @@ static struct virtqueue *vop_find_vq(struct virtio_device *dev, > goto unmap; > } > > - /* Allocate and reassign used ring now */ > vdev->used_size[index] = PAGE_ALIGN(sizeof(__u16) * 3 + > sizeof(struct vring_used_elem) * > le16_to_cpu(config.num)); > @@ -351,8 +359,17 @@ static struct virtqueue *vop_find_vq(struct virtio_device *dev, > err = -ENOMEM; > dev_err(_vop_dev(vdev), "%s %d err %d\n", > __func__, __LINE__, err); > - goto del_vq; > + goto unmap; > + } > + > + vq = vop_new_virtqueue(index, le16_to_cpu(config.num), dev, ctx, > + (void __force *)va, vop_notify, callback, > + name, used); > + if (!vq) { > + err = -ENOMEM; > + goto free_used; > } > + > vdev->used[index] = dma_map_single(&vpdev->dev, used, > vdev->used_size[index], > DMA_BIDIRECTIONAL); > @@ -360,26 +377,17 @@ static struct virtqueue *vop_find_vq(struct virtio_device *dev, > err = -ENOMEM; > dev_err(_vop_dev(vdev), "%s %d err %d\n", > __func__, __LINE__, err); > - goto free_used; > + goto del_vq; > } > writeq(vdev->used[index], &vqconfig->used_address); > - /* > - * To reassign the used ring here we are directly accessing > - * struct vring_virtqueue which is a private data structure > - * in virtio_ring.c. At the minimum, a BUILD_BUG_ON() in > - * vring_new_virtqueue() would ensure that > - * (&vq->vring == (struct vring *) (&vq->vq + 1)); > - */ > - vr = (struct vring *)(vq + 1); > - vr->used = used; > > vq->priv = vdev; > return vq; > +del_vq: > + vring_del_virtqueue(vq); > free_used: > free_pages((unsigned long)used, > get_order(vdev->used_size[index])); > -del_vq: > - vring_del_virtqueue(vq); > unmap: > vpdev->hw_ops->unmap(vpdev, vdev->vr[index]); > return ERR_PTR(err);