linux-kernel - Re: [PATCH v4 7/8] vhost: Add new UAPI to support change to task mode

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <2on4eblmkzkhecpyiwtauel6hxw6upnlh6wunfxgxvfp45cej3@6z5lzdermzeg>
Date: Tue, 10 Dec 2024 18:54:54 +0100
From: Stefano Garzarella <sgarzare@...hat.com>
To: Cindy Lu <lulu@...hat.com>
Cc: jasowang@...hat.com, mst@...hat.com, michael.christie@...cle.com, 
	linux-kernel@...r.kernel.org, virtualization@...ts.linux-foundation.org, 
	netdev@...r.kernel.org
Subject: Re: [PATCH v4 7/8] vhost: Add new UAPI to support change to task mode

On Wed, Dec 11, 2024 at 12:41:46AM +0800, Cindy Lu wrote:
>Add a new UAPI to enable setting the vhost device to task mode.
>The userspace application can use VHOST_SET_INHERIT_FROM_OWNER
>to configure the mode if necessary.
>This setting must be applied before VHOST_SET_OWNER, as the worker
>will be created in the VHOST_SET_OWNER function
>
>Signed-off-by: Cindy Lu <lulu@...hat.com>
>---
> drivers/vhost/vhost.c      | 22 +++++++++++++++++++++-
> include/uapi/linux/vhost.h | 18 ++++++++++++++++++
> 2 files changed, 39 insertions(+), 1 deletion(-)
>
>diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
>index 3e9cb99da1b5..12c3bf3d1ed4 100644
>--- a/drivers/vhost/vhost.c
>+++ b/drivers/vhost/vhost.c
>@@ -2257,15 +2257,35 @@ long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
> {
> 	struct eventfd_ctx *ctx;
> 	u64 p;
>-	long r;
>+	long r = 0;
> 	int i, fd;
>+	u8 inherit_owner;
>
> 	/* If you are not the owner, you can become one */
> 	if (ioctl == VHOST_SET_OWNER) {
> 		r = vhost_dev_set_owner(d);
> 		goto done;
> 	}
>+	if (ioctl == VHOST_SET_INHERIT_FROM_OWNER) {
>+		/*inherit_owner can only be modified before owner is set*/
>+		if (vhost_dev_has_owner(d)) {
>+			r = -EBUSY;
>+			goto done;
>+		}
>+		if (copy_from_user(&inherit_owner, argp, sizeof(u8))) {
>+			r = -EFAULT;
>+			goto done;
>+		}
>+		/* Validate the inherit_owner value, ensuring it is either 0 or 1 */
>+		if (inherit_owner > 1) {
>+			r = -EINVAL;
>+			goto done;
>+		}
>+
>+		d->inherit_owner = (bool)inherit_owner;
>
>+		goto done;
>+	}
> 	/* You must be the owner to do anything else */
> 	r = vhost_dev_check_owner(d);
> 	if (r)
>diff --git a/include/uapi/linux/vhost.h b/include/uapi/linux/vhost.h
>index b95dd84eef2d..d7564d62b76d 100644
>--- a/include/uapi/linux/vhost.h
>+++ b/include/uapi/linux/vhost.h
>@@ -235,4 +235,22 @@
>  */
> #define VHOST_VDPA_GET_VRING_SIZE	_IOWR(VHOST_VIRTIO, 0x82,	
> \
> 					      struct vhost_vring_state)
>+
>+/**
>+ * VHOST_SET_INHERIT_FROM_OWNER - Set the inherit_owner flag for the vhost device
>+ *
>+ * @param inherit_owner: An 8-bit value that determines the vhost thread mode
>+ *
>+ * When inherit_owner is set to 1:
>+ *   - The VHOST worker threads inherit its values/checks from
>+ *     the thread that owns the VHOST device, The vhost threads will
>+ *     be counted in the nproc rlimits.

We should mention that this is the default behaviour, so the user does 
not need to call VHOST_SET_INHERIT_FROM_OWNER if the default is okay.

>+ *
>+ * When inherit_owner is set to 0:
>+ *   - The VHOST worker threads will use the traditional kernel thread (kthread)
>+ *     implementation, which may be preferred by older userspace applications that
>+ *     do not utilize the newer vhost_task concept.
>+ */
>+#define VHOST_SET_INHERIT_FROM_OWNER _IOW(VHOST_VIRTIO, 0x83, __u8)

Do we really need a parameter? I mean could we just have an IOCTL to set 
the old behavior, since the new one is enabled by default?

Not a strong opinion on that, but just an idea to reduce confusion in 
the user. Anyway, if we want the parameter, maybe we can use int instead 
of u8, since we don't particularly care about the length.

Thanks,
Stefano

>+
> #endif
>-- 
>2.45.0
>