On 01/11/2011 12:51 PM, Konrad Rzeszutek Wilk wrote:
> On Tue, Jan 11, 2011 at 07:45:34AM -0500, Daniel De Graaf wrote:
>> On 01/10/2011 04:52 PM, Konrad Rzeszutek Wilk wrote:
>>>> static long gntdev_ioctl(struct file *flip,
>>>> unsigned int cmd, unsigned long arg)
>>>> @@ -555,9 +538,6 @@ static long gntdev_ioctl(struct file *flip,
>>>> case IOCTL_GNTDEV_GET_OFFSET_FOR_VADDR:
>>>> return gntdev_ioctl_get_offset_for_vaddr(priv, ptr);
>>>> - case IOCTL_GNTDEV_SET_MAX_GRANTS:
>>>> - return gntdev_ioctl_set_max_grants(priv, ptr);
>>> Would it make sense to return -EPNOTSUPPORTED? Or does it not really
>>> matter as nobody has been using this ioctl call?
>> Does this produce a clearer error message than the default -ENOIOCTLCMD?
>> It's possible that some people use it, since it was exposed as an API.
> Looking at the Xen tools the user of this is:
> xc_gnttab_set_max_grants which would end up returning whatever the
> error is. I don't see any users of this in the Xen tools, thought there might
> be some in the XCP code. Lets stay with your ENOIOCTLCMD.
> However, I was wondering if you are going to submit a patch to the Xen
> tool stack so that it can utlize the SysFS interface to set the limits
> for that API call?
No, because the semantics of what the limit is covering have changed. The
new limit is per-domain, and if there was any existing code that set the
limit, it would have been a per-open value and probably too low. I think
it was suggested that the call be removed from the Xen API; I can submit
a patch to do that, if you want.
The new value is probably best set in modprobe.conf if gntdev is a module;
the sysfs interface is useful for runtime adjustment or if it is builtin.
Daniel De Graaf
National Security Agency
Xen-devel mailing list