[PATCH] An attempt to have an unlimitedly extendable sys_clone

Cedric Le Goater clg at fr.ibm.com
Tue Jan 15 07:56:39 PST 2008

Oren Laadan wrote:
> Pavel Emelyanov wrote:
>> We have one bit in the clone_flags left, so we won't be
>> able to create more namespaces after we make it busy.
>> Besides, for checkpoint/restart jobs we might want to
>> create tasks with pre-defined pids (virtual of course).
>> What else might be required from clone() - nobody knows.
>> This is an attempt to create a extendable API for clone.
>> I use the last bit in the clone_flags for CLONE_NEWCLONE.
> how about "CLONE_EXTEND" ?
>> When set it will denote that the child_tidptr is not a
>> pointer on the tid storage, but the pointer on the struct
>> long_clone_struct which currently looks like this:
>> struct long_clone_arg {
>> 	int size;
>> };
> how about "ext_clone_arg" ?
> (both suggestion make the use more explicit and are more
> consistent with each other; but definitely a nit ...)

yeah I agree. The naming can be improved but let's just wait
for the patch to be sent on lkml at . I'm sure we will have 
plenty of feedback.

however, the last clone flag name should be consistent with the 
structure name. CLONE_NEWCLONE is not.


More information about the Containers mailing list