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:   Tue, 6 Dec 2022 12:00:40 +0100
From:   Stefano Garzarella <sgarzare@redhat.com>
To:     "Longpeng(Mike)" <longpeng2@huawei.com>
Cc:     stefanha@redhat.com, mst@redhat.com, jasowang@redhat.com,
        eperezma@redhat.com, cohuck@redhat.com, arei.gonglei@huawei.com,
        yechuan@huawei.com, huangzhichao@huawei.com,
        virtualization@lists.linux-foundation.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] vp_vdpa: harden the logic of set status
Message-ID: <20221206110040.q5i2k7ypojuira2e@sgarzare-redhat>
References: <20221206021321.2400-1-longpeng2@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
In-Reply-To: <20221206021321.2400-1-longpeng2@huawei.com>
Precedence: bulk

On Tue, Dec 06, 2022 at 10:13:21AM +0800, Longpeng(Mike) wrote:
>From: Longpeng <longpeng2@huawei.com>
>
>1. We should not set status to 0 when invoking vp_vdpa_set_status(),
>   trigger a warning in that case.
>
>2. The driver MUST wait for a read of device_status to return 0 before
>   reinitializing the device. But we also don't want to keep us in an
>   infinite loop forever, so wait for 5s if we try to reset the device.
>
>Signed-off-by: Longpeng <longpeng2@huawei.com>
>---
>Changes v1->v2:
>  - use WARN_ON instead of BUG_ON. [Stefano]
>  - use "warning + failed" instead of "infinite loop". [Jason, Stefano]
>  - use usleep_range instead of msleep (checkpatch). [Longpeng]
>
>---
> drivers/vdpa/virtio_pci/vp_vdpa.c | 22 ++++++++++++++++++++++
> 1 file changed, 22 insertions(+)
>
>diff --git a/drivers/vdpa/virtio_pci/vp_vdpa.c b/drivers/vdpa/virtio_pci/vp_vdpa.c
>index 13701c2a1963..a2d3b13ac646 100644
>--- a/drivers/vdpa/virtio_pci/vp_vdpa.c
>+++ b/drivers/vdpa/virtio_pci/vp_vdpa.c
>@@ -214,6 +214,9 @@ static void vp_vdpa_set_status(struct vdpa_device *vdpa, u8 status)
> 	struct virtio_pci_modern_device *mdev = vp_vdpa_to_mdev(vp_vdpa);
> 	u8 s = vp_vdpa_get_status(vdpa);
>
>+	/* We should never be setting status to 0. */
>+	WARN_ON(status == 0);
>+
> 	if (status & VIRTIO_CONFIG_S_DRIVER_OK &&
> 	    !(s & VIRTIO_CONFIG_S_DRIVER_OK)) {
> 		vp_vdpa_request_irq(vp_vdpa);
>@@ -222,14 +225,33 @@ static void vp_vdpa_set_status(struct vdpa_device *vdpa, u8 status)
> 	vp_modern_set_status(mdev, status);
> }
>
>+#define VP_VDPA_RESET_TMOUT_MS 5000 /* 5s */

What about moving this define on top of this file?
Near the other macro.

And I would use TIMEOUT entirely.

>+
> static int vp_vdpa_reset(struct vdpa_device *vdpa, bool clear)
> {
> 	struct vp_vdpa *vp_vdpa = vdpa_to_vp(vdpa);
> 	struct virtio_pci_modern_device *mdev = vp_vdpa_to_mdev(vp_vdpa);
> 	u8 s = vp_vdpa_get_status(vdpa);
>+	unsigned long timeout;
>
> 	vp_modern_set_status(mdev, 0);
>
>+	/*
>+	 * As the virtio v1.1 spec (4.1.4.3.2) says: After writing 0 to

I think we can refer to the lates v1.2 (the section should be the same).

>+	 * device_status, the driver MUST wait for a read of device_status
>+	 * to return 0 before reinitializing the device.
>+	 * To avoid keep us here forever, we only wait for 5 seconds.
>+	 */
>+	timeout = jiffies + msecs_to_jiffies(VP_VDPA_RESET_TMOUT_MS);
>+	while (vp_modern_get_status(mdev)) {
>+		usleep_range(1000, 1500);
>+		if (time_after(jiffies, timeout)) {
>+			dev_err(&mdev->pci_dev->dev,
>+				"vp_vdpa: fail to reset device\n");
>+			return -ETIMEDOUT;
>+		}
>+	}
>+
> 	if (s & VIRTIO_CONFIG_S_DRIVER_OK)
> 		vp_vdpa_free_irq(vp_vdpa);
>
>-- 
>2.23.0
>

The rest LGTM!

Thanks,
Stefano