Re: [PATCH 06/17] virt: acrn: Introduce VM management interfaces

From: Greg Kroah-Hartman
Date: Sun Aug 30 2020 - 03:24:09 EST


On Sat, Aug 29, 2020 at 07:04:36PM +0800, Shuo A Liu wrote:
> Hi Greg,
>
> On Fri 28.Aug'20 at 12:27:38 +0200, Greg Kroah-Hartman wrote:
> > On Tue, Aug 25, 2020 at 10:45:06AM +0800, shuo.a.liu@xxxxxxxxx wrote:
> > > + default:
> > > + pr_warn("Unknown IOCTL 0x%x!\n", cmd);
> > > + ret = -EINVAL;
> >
> > Wrong error value here, right?
>
> Right, it should be -ENOIOCTLCMD.

It could, but really, just return the correct error for this, to prevent
the core from having to do the conversion.

The reviewers at Intel who should have read this before submitting it,
know the correct value to return for an illegal ioctl, please go ask
them.

> However, i found many instances in kernel drivers return -EINVAL for no
> ioctl command support. :)

Then they too are wrong. No need to add known bugs before the code is
accepted.

See the comments above the is_unrecognized_ioctl() in block/ioctl.c for
all of the details and why -EINVAL is not the correct thing to do here.

thanks,

greg k-h