Cedric Le Goater wrote:
>>>  * possible direction for C/R user api
>>>    	    . checkpoint/restart syscalls
>>>            . C/R file systems 
>>>	      	  solves the set id issue 
>>>	      	  elegant but exposes too much the ABI
>>
>>I vote for the filesystem :) I'd add more details as we did on mini-summit.
>>
>>tasks
>> `- <pid1>
>>    `- <tid1>
>>       ...
>>       <tidN>
>>       files
>>        `- 1 -> /* made as a symlink */
>>           2 -> /* if socket point to net/ objects */
>>       memory
>>        `- <vma1> -> /* symlink to mm objects */
>>    <pid2>
>>    ...
>>    <pidN>
>>mm 
>>ipc
>>network
>>
>>and so on and so forth.
> 
> 
> We need to dig on this idea. RFC ?

1. resource interrelashionships are much more complicated then a tree.
   e.g. pid can be owned by a number of processes, threads, terminals, etc.
   So I'm not a fan of the idea.
2. exposing such a low-level information to the user-space can be undesirable:
   a) it allows to create non-GPL checkpointing
   b) significantly hits the performance of checkpoint/restore
   c) BTW, how do you plan to restore via filesystem? mkdir? create? :)

Thanks,
Kirill

_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers