librte_eal: ease init in a docker container
Checks
Commit Message
move_pages() is only used to get the numa node id, but this function
is not allowed by default in docker (it needs CAP_SYS_NICE and an update of
the seccomp profile).
get_mempolicy() also requires CAP_SYS_NICE but doesn't need any change in
the default seccomp profile.
Note that the returned value of move_pages() was not checked, thus some
errors could be hidden (if the requested id was 0).
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
Reviewed-by: Didier Pallard <didier.pallard@6wind.com>
---
lib/librte_eal/linux/eal/eal_memalloc.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
Comments
On 22-May-19 4:41 PM, Nicolas Dichtel wrote:
> move_pages() is only used to get the numa node id, but this function
> is not allowed by default in docker (it needs CAP_SYS_NICE and an update of
> the seccomp profile).
> get_mempolicy() also requires CAP_SYS_NICE but doesn't need any change in
> the default seccomp profile.
>
> Note that the returned value of move_pages() was not checked, thus some
> errors could be hidden (if the requested id was 0).
>
> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
> Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
> Reviewed-by: Didier Pallard <didier.pallard@6wind.com>
> ---
I can see the check for move_pages and it's a good fix, but what is the
relation to docker init here? The patch by itself only enables handling
of move_pages() failure and adds nothing else. The commit message
doesn't match the patch in question IMO.
Also, Cc: stable and Fixes: ?
Le 22/05/2019 à 17:57, Burakov, Anatoly a écrit :
> On 22-May-19 4:41 PM, Nicolas Dichtel wrote:
>> move_pages() is only used to get the numa node id, but this function
>> is not allowed by default in docker (it needs CAP_SYS_NICE and an update of
>> the seccomp profile).
>> get_mempolicy() also requires CAP_SYS_NICE but doesn't need any change in
>> the default seccomp profile.
>>
>> Note that the returned value of move_pages() was not checked, thus some
>> errors could be hidden (if the requested id was 0).
>>
>> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
>> Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
>> Reviewed-by: Didier Pallard <didier.pallard@6wind.com>
>> ---
>
> I can see the check for move_pages and it's a good fix, but what is the relation
> to docker init here? The patch by itself only enables handling of move_pages()
> failure and adds nothing else. The commit message doesn't match the patch in
> question IMO.
I'm not sure to understand your comment. The call to move_pages() is replaced by
a call to get_mempolicy().
What am I missing?
Regards,
Nicolas
On 22-May-19 5:08 PM, Nicolas Dichtel wrote:
>
> Le 22/05/2019 à 17:57, Burakov, Anatoly a écrit :
>> On 22-May-19 4:41 PM, Nicolas Dichtel wrote:
>>> move_pages() is only used to get the numa node id, but this function
>>> is not allowed by default in docker (it needs CAP_SYS_NICE and an update of
>>> the seccomp profile).
>>> get_mempolicy() also requires CAP_SYS_NICE but doesn't need any change in
>>> the default seccomp profile.
>>>
>>> Note that the returned value of move_pages() was not checked, thus some
>>> errors could be hidden (if the requested id was 0).
>>>
>>> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
>>> Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
>>> Reviewed-by: Didier Pallard <didier.pallard@6wind.com>
>>> ---
>>
>> I can see the check for move_pages and it's a good fix, but what is the relation
>> to docker init here? The patch by itself only enables handling of move_pages()
>> failure and adds nothing else. The commit message doesn't match the patch in
>> question IMO.
> I'm not sure to understand your comment. The call to move_pages() is replaced by
> a call to get_mempolicy().
> What am I missing?
>
Oh, apologies, i misread the patch. It is i who was missing something :)
>
> Regards,
> Nicolas
>
On 22-May-19 4:41 PM, Nicolas Dichtel wrote:
> move_pages() is only used to get the numa node id, but this function
> is not allowed by default in docker (it needs CAP_SYS_NICE and an update of
> the seccomp profile).
> get_mempolicy() also requires CAP_SYS_NICE but doesn't need any change in
> the default seccomp profile.
>
> Note that the returned value of move_pages() was not checked, thus some
> errors could be hidden (if the requested id was 0).
>
> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
> Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
> Reviewed-by: Didier Pallard <didier.pallard@6wind.com>
> ---
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
Still, should at least specify Cc: stable, if not Fixes tag too (since
ignoring return value of move_pages() is technically a bug).
s/librte_eal/mem/ for the title prefix.
On Thu, May 23, 2019 at 10:56 AM Burakov, Anatoly <anatoly.burakov@intel.com>
wrote:
> On 22-May-19 4:41 PM, Nicolas Dichtel wrote:
> > move_pages() is only used to get the numa node id, but this function
> > is not allowed by default in docker (it needs CAP_SYS_NICE and an update
> of
> > the seccomp profile).
> > get_mempolicy() also requires CAP_SYS_NICE but doesn't need any change in
> > the default seccomp profile.
> >
> > Note that the returned value of move_pages() was not checked, thus some
> > errors could be hidden (if the requested id was 0).
> >
> > Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
> > Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
> > Reviewed-by: Didier Pallard <didier.pallard@6wind.com>
> > ---
>
> Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
>
> Still, should at least specify Cc: stable, if not Fixes tag too (since
> ignoring return value of move_pages() is technically a bug).
>
+1
At first I was wondering if we should separate the fix from the
enhancement, but I suppose backporting both things as one patch is fine too.
@@ -600,9 +600,13 @@ alloc_seg(struct rte_memseg *ms, void *addr, int socket_id,
}
#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES
- move_pages(getpid(), 1, &addr, NULL, &cur_socket_id, 0);
-
- if (cur_socket_id != socket_id) {
+ ret = get_mempolicy(&cur_socket_id, NULL, 0, addr,
+ MPOL_F_NODE | MPOL_F_ADDR);
+ if (ret < 0) {
+ RTE_LOG(DEBUG, EAL, "%s(): get_mempolicy: %s\n",
+ __func__, strerror(errno));
+ goto mapped;
+ } else if (cur_socket_id != socket_id) {
RTE_LOG(DEBUG, EAL,
"%s(): allocation happened on wrong socket (wanted %d, got %d)\n",
__func__, socket_id, cur_socket_id);