[CRIU] [PATCH v4 4/6] mem: add dump state of THP_DISABLED prctl

[Pavel Emelyanov]

On 06/28/2017 03:05 PM, Mike Rapoport wrote:
> On Wed, Jun 28, 2017 at 01:56:46PM +0300, Pavel Emelyanov wrote:
>> On 06/28/2017 09:43 AM, Mike Rapoport wrote:
>>> Signed-off-by: Mike Rapoport <rppt at linux.vnet.ibm.com>
>>> ---
>>>  criu/cr-dump.c          | 3 +++
>>>  criu/include/parasite.h | 1 +
>>>  criu/pie/parasite.c     | 1 +
>>>  images/mm.proto         | 1 +
>>>  4 files changed, 6 insertions(+)
>>>
>>> diff --git a/criu/cr-dump.c b/criu/cr-dump.c
>>> index 3f9f8d0..308cc4e 100644
>>> --- a/criu/cr-dump.c
>>> +++ b/criu/cr-dump.c
>>> @@ -507,6 +507,9 @@ static int dump_task_mm(pid_t pid, const struct proc_pid_stat *stat,
>>>  	mme.dumpable = misc->dumpable;
>>>  	mme.has_dumpable = true;
>>>  
>>> +	mme.thp_disabled = misc->thp_disabled;
>>> +	mme.has_thp_disabled = true;
>>> +
>>>  	mme.n_mm_saved_auxv = AT_VECTOR_SIZE;
>>>  	mme.mm_saved_auxv = xmalloc(pb_repeated_size(&mme, mm_saved_auxv));
>>>  	if (!mme.mm_saved_auxv)
>>> diff --git a/criu/include/parasite.h b/criu/include/parasite.h
>>> index cd6d793..115891f 100644
>>> --- a/criu/include/parasite.h
>>> +++ b/criu/include/parasite.h
>>> @@ -124,6 +124,7 @@ struct parasite_dump_misc {
>>>  	u32 umask;
>>>  
>>>  	int dumpable;
>>> +	int thp_disabled;
>>>  };
>>>  
>>>  /*
>>> diff --git a/criu/pie/parasite.c b/criu/pie/parasite.c
>>> index fe08e72..78bbe9e 100644
>>> --- a/criu/pie/parasite.c
>>> +++ b/criu/pie/parasite.c
>>> @@ -177,6 +177,7 @@ static int dump_misc(struct parasite_dump_misc *args)
>>>  	args->umask = sys_umask(0);
>>>  	sys_umask(args->umask); /* never fails */
>>>  	args->dumpable = sys_prctl(PR_GET_DUMPABLE, 0, 0, 0, 0);
>>> +	args->thp_disabled = sys_prctl(PR_GET_THP_DISABLE, 0, 0, 0, 0);
>>
>> How will this work OK on kernels without this prctl fixed?
> 
> The same way as after the fix :)
> The prctl(PR_SET_THP_DISABLE) in any case sets a flag in 'struct
> mm_struct' and the prctl(PR_GET_THP_DISABLE) returns 1 if such flag is set.
> The change in the SET behaviour that was introduced recently is completely
> internal to mm and it only changes the way the flag in 'struct mm_struct'
> is treated by khugepaged and page allocation path.

Ah, I see :) Thanks!