linux-kernel - Re: [PATCH bpf-next 1/4] bpf: enable task local storage for tracing programs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Sat, 16 Jan 2021 01:12:48 +0000
From:   Song Liu <songliubraving@...com>
To:     Yonghong Song <yhs@...com>
CC:     KP Singh <kpsingh@...nel.org>, Martin Lau <kafai@...com>,
        bpf <bpf@...r.kernel.org>, Networking <netdev@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>,
        "mingo@...hat.com" <mingo@...hat.com>,
        "Peter Zijlstra" <peterz@...radead.org>,
        Alexei Starovoitov <ast@...nel.org>,
        "Daniel Borkmann" <daniel@...earbox.net>,
        Andrii Nakryiko <andrii@...nel.org>,
        "John Fastabend" <john.fastabend@...il.com>,
        Kernel Team <Kernel-team@...com>,
        "Hao Luo" <haoluo@...gle.com>, kernel test robot <lkp@...el.com>
Subject: Re: [PATCH bpf-next 1/4] bpf: enable task local storage for tracing
 programs



> On Jan 15, 2021, at 4:55 PM, Yonghong Song <yhs@...com> wrote:
> 
> 
> 
> On 1/15/21 3:34 PM, Song Liu wrote:
>>> On Jan 12, 2021, at 8:53 AM, KP Singh <kpsingh@...nel.org> wrote:
>>> 
>>> On Tue, Jan 12, 2021 at 5:32 PM Yonghong Song <yhs@...com> wrote:
>>>> 
>>>> 
>>>> 
>>>> On 1/11/21 3:45 PM, Song Liu wrote:
>>>>> 
>>>>> 
>>>>>> On Jan 11, 2021, at 1:58 PM, Martin Lau <kafai@...com> wrote:
>>>>>> 
>>>>>> On Mon, Jan 11, 2021 at 10:35:43PM +0100, KP Singh wrote:
>>>>>>> On Mon, Jan 11, 2021 at 7:57 PM Martin KaFai Lau <kafai@...com> wrote:
>>>>>>>> 
>>>>>>>> On Fri, Jan 08, 2021 at 03:19:47PM -0800, Song Liu wrote:
>>>>>>>> 
>>>>>>>> [ ... ]
>>>>>>>> 
>>>>>>>>> diff --git a/kernel/bpf/bpf_local_storage.c b/kernel/bpf/bpf_local_storage.c
>>>>>>>>> index dd5aedee99e73..9bd47ad2b26f1 100644
>>>>>>>>> --- a/kernel/bpf/bpf_local_storage.c
>>>>>>>>> +++ b/kernel/bpf/bpf_local_storage.c
>>> 
>>> [...]
>>> 
>>>>>>>>> +#include <linux/bpf.h>
>>>>>>>>> 
>>>>>>>>> #include <asm/pgalloc.h>
>>>>>>>>> #include <linux/uaccess.h>
>>>>>>>>> @@ -734,6 +735,7 @@ void __put_task_struct(struct task_struct *tsk)
>>>>>>>>>      cgroup_free(tsk);
>>>>>>>>>      task_numa_free(tsk, true);
>>>>>>>>>      security_task_free(tsk);
>>>>>>>>> +     bpf_task_storage_free(tsk);
>>>>>>>>>      exit_creds(tsk);
>>>>>>>> If exit_creds() is traced by a bpf and this bpf is doing
>>>>>>>> bpf_task_storage_get(..., BPF_LOCAL_STORAGE_GET_F_CREATE),
>>>>>>>> new task storage will be created after bpf_task_storage_free().
>>>>>>>> 
>>>>>>>> I recalled there was an earlier discussion with KP and KP mentioned
>>>>>>>> BPF_LSM will not be called with a task that is going away.
>>>>>>>> It seems enabling bpf task storage in bpf tracing will break
>>>>>>>> this assumption and needs to be addressed?
>>>>>>> 
>>>>>>> For tracing programs, I think we will need an allow list where
>>>>>>> task local storage can be used.
>>>>>> Instead of whitelist, can refcount_inc_not_zero(&tsk->usage) be used?
>>>>> 
>>>>> I think we can put refcount_inc_not_zero() in bpf_task_storage_get, like:
>>>>> 
>>>>> diff --git i/kernel/bpf/bpf_task_storage.c w/kernel/bpf/bpf_task_storage.c
>>>>> index f654b56907b69..93d01b0a010e6 100644
>>>>> --- i/kernel/bpf/bpf_task_storage.c
>>>>> +++ w/kernel/bpf/bpf_task_storage.c
>>>>> @@ -216,6 +216,9 @@ BPF_CALL_4(bpf_task_storage_get, struct bpf_map *, map, struct task_struct *,
>>>>>          * by an RCU read-side critical section.
>>>>>          */
>>>>>         if (flags & BPF_LOCAL_STORAGE_GET_F_CREATE) {
>>>>> +               if (!refcount_inc_not_zero(&task->usage))
>>>>> +                       return -EBUSY;
>>>>> +
>>>>>                 sdata = bpf_local_storage_update(
>>>>>                         task, (struct bpf_local_storage_map *)map, value,
>>>>>                         BPF_NOEXIST);
>>>>> 
>>>>> But where shall we add the refcount_dec()? IIUC, we cannot add it to
>>>>> __put_task_struct().
>>>> 
>>>> Maybe put_task_struct()?
>>> 
>>> Yeah, something like, or if you find a more elegant alternative :)
>>> 
>>> --- a/include/linux/sched/task.h
>>> +++ b/include/linux/sched/task.h
>>> @@ -107,13 +107,20 @@ extern void __put_task_struct(struct task_struct *t);
>>> 
>>> static inline void put_task_struct(struct task_struct *t)
>>> {
>>> -       if (refcount_dec_and_test(&t->usage))
>>> +
>>> +       if (rcu_access_pointer(t->bpf_storage)) {
>>> +               if (refcount_sub_and_test(2, &t->usage))
>>> +                       __put_task_struct(t);
>>> +       } else if (refcount_dec_and_test(&t->usage))
>>>                __put_task_struct(t);
>>> }
>>> 
>>> static inline void put_task_struct_many(struct task_struct *t, int nr)
>>> {
>>> -       if (refcount_sub_and_test(nr, &t->usage))
>>> +       if (rcu_access_pointer(t->bpf_storage)) {
>>> +               if (refcount_sub_and_test(nr + 1, &t->usage))
>>> +                       __put_task_struct(t);
>>> +       } else if (refcount_sub_and_test(nr, &t->usage))
>>>                __put_task_struct(t);
>>> }
>> It is not ideal to leak bpf_storage here. How about we only add the
>> following:
>> diff --git i/kernel/bpf/bpf_task_storage.c w/kernel/bpf/bpf_task_storage.c
>> index f654b56907b69..2811b9fc47233 100644
>> --- i/kernel/bpf/bpf_task_storage.c
>> +++ w/kernel/bpf/bpf_task_storage.c
>> @@ -216,6 +216,10 @@ BPF_CALL_4(bpf_task_storage_get, struct bpf_map *, map, struct task_struct *,
>>          * by an RCU read-side critical section.
>>          */
>>         if (flags & BPF_LOCAL_STORAGE_GET_F_CREATE) {
>> +               /* the task_struct is being freed, fail over*/
>> +               if (!refcount_read(&task->usage))
>> +                       return -EBUSY;
> 
> This may not work? Even we check here and task->usage is not 0, it could still become 0 immediately after the above refcount_read, right?

We call bpf_task_storage_get() with "task" that has valid BTF, so "task"
should not go away during the BPF program? Whatever mechanism that 
triggers the BPF program should either hold a reference to task (usage > 0)
or be the only one owning it (usage == 0, in __put_task_struct). Did I miss
anything?

Thanks,
Song

> 
>> +
>>                 sdata = bpf_local_storage_update(
>>                         task, (struct bpf_local_storage_map *)map, value,
>>                         BPF_NOEXIST);
>>> 
>>> 
>>> I may be missing something but shouldn't bpf_storage be an __rcu
>>> member like we have for sk_bpf_storage?
>> Good catch! I will fix this in v2.
>> Thanks,
>> Song