Re: + c-r-prctl-add-ability-to-set-new-mm_struct-exe_file-update-after-mm- num_exe_file_vmas-removal.patch added to -mm tree

[Konstantin Khlebnikov <khlebnikov@openvz.org>]

Oleg Nesterov wrote:
> On 04/20, Cyrill Gorcunov wrote:
>>
>> On Thu, Apr 19, 2012 at 09:20:05PM +0200, Oleg Nesterov wrote:
>>> On 04/19, Andrew Morton wrote:
>>>>
>>>> From: Konstantin Khlebnikov<khlebnikov@openvz.org>
>>>> Subject: c/r: prctl: update prctl_set_mm_exe_file() after mm->num_exe_file_vmas removal
>>>>
>>>> [ fix for "c-r-prctl-add-ability-to-set-new-mm_struct-exe_file-v2" from mm tree ]
>>>>
>>>> After removing mm->num_exe_file_vmas kernel keeps mm->exe_file until final
>>>> mmput(), it never becomes NULL while task is alive.
>>>>
>>>> We can check for other mapped files in mm instead of checking
>>>> mm->num_exe_file_vmas, and mark mm with flag MMF_EXE_FILE_CHANGED in order
>>>> to forbid second changing of mm->exe_file.
>>>
>>> I lost the track a long ago.
>>>
>>> Just one question, what does this "forbid second changing" actually mean?
>>
>> Heh :) Oleg, it was actually your idea to make this feature "one-shot".
>
> Heh, no ;)
>
> IIRC, I only asked you what do you actually want,
>
> 	Just one note for the record, prctl_set_mm_exe_file() does
>
> 		if (mm->num_exe_file_vmas)
> 			return -EBUSY;
>
> 	We could do
>
> 		if (mm->exe_file)
> 			return -EBUSY;
>
> 	This way "because this feature is a special to C/R" becomes
> 	really true. IOW, you can't do PR_SET_MM_EXE_FILE twice.
>
> 	I am fine either way, just I want to ensure you really want
> 	the current version.
>
> and only because it was documented as "feature is a special to C/R".
>
>> Once exe-file changed to a new value, it can't be changed again. The
>> reason was to bring at least minimum disturbance in sysadmins life.
>
> You misunderstood. I am not arguing with "one-shot", I do not really
> care.
>
> My question is: unless I missed something "it can't be changed again"
> is not actually true. A task does PR_SET_MM_EXE_FILE, then it forks
> the new child. The child can do PR_SET_MM_EXE_FILE again. Is this
> by design?
>
> Oleg.
>

I found more weird case: child thread (with CLONE_THREAD and without CLONE_VM)
changes its exe_file...