Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:5 as permitted sender) client-ip=2620:137:e000::3:5;
Message-ID: <1ea32c1c-babf-452d-8097-c23777e8018b@bytedance.com>
Date:   Wed, 11 Oct 2023 13:32:02 +0800
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH bpf-next v4 4/8] bpf: Introduce css open-coded iterator
 kfuncs
From:   Chuyi Zhou <zhouchuyi@bytedance.com>
To:     bpf@vger.kernel.org
Cc:     ast@kernel.org, daniel@iogearbox.net, andrii@kernel.org,
        martin.lau@kernel.org, tj@kernel.org, linux-kernel@vger.kernel.org
References: <20231007124522.34834-1-zhouchuyi@bytedance.com>
 <20231007124522.34834-5-zhouchuyi@bytedance.com>
 <f6fd8209-4808-45e1-b5b5-95ea2dcc0ba4@bytedance.com>
In-Reply-To: <f6fd8209-4808-45e1-b5b5-95ea2dcc0ba4@bytedance.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


在 2023/10/11 12:44, Chuyi Zhou 写道:
> 
> 
> 在 2023/10/7 20:45, Chuyi Zhou 写道:
>> This Patch adds kfuncs bpf_iter_css_{new,next,destroy} which allow
>> creation and manipulation of struct bpf_iter_css in open-coded iterator
>> style. These kfuncs actually wrapps css_next_descendant_{pre, post}.
>> css_iter can be used to:
>>
>> 1) iterating a sepcific cgroup tree with pre/post/up order
>>
>> 2) iterating cgroup_subsystem in BPF Prog, like
>> for_each_mem_cgroup_tree/cpuset_for_each_descendant_pre in kernel.
>>
>> The API design is consistent with cgroup_iter. bpf_iter_css_new accepts
>> parameters defining iteration order and starting css. Here we also reuse
>> BPF_CGROUP_ITER_DESCENDANTS_PRE, BPF_CGROUP_ITER_DESCENDANTS_POST,
>> BPF_CGROUP_ITER_ANCESTORS_UP enums.
>>
>> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
>> Acked-by: Tejun Heo <tj@kernel.org>
>> ---
>>   kernel/bpf/cgroup_iter.c                      | 59 +++++++++++++++++++
>>   kernel/bpf/helpers.c                          |  3 +
>>   .../testing/selftests/bpf/bpf_experimental.h  |  6 ++
>>   3 files changed, 68 insertions(+)
>>
>> diff --git a/kernel/bpf/cgroup_iter.c b/kernel/bpf/cgroup_iter.c
>> index 810378f04fbc..9c6ad892ae82 100644
>> --- a/kernel/bpf/cgroup_iter.c
>> +++ b/kernel/bpf/cgroup_iter.c
>> @@ -294,3 +294,62 @@ static int __init bpf_cgroup_iter_init(void)
>>   }
>>   late_initcall(bpf_cgroup_iter_init);
>> +
>> +struct bpf_iter_css {
>> +    __u64 __opaque[3];
>> +} __attribute__((aligned(8)));
>> +
>> +struct bpf_iter_css_kern {
>> +    struct cgroup_subsys_state *start;
>> +    struct cgroup_subsys_state *pos;
>> +    unsigned int flags;
>> +} __attribute__((aligned(8)));
>> +
>> +__bpf_kfunc int bpf_iter_css_new(struct bpf_iter_css *it,
>> +        struct cgroup_subsys_state *start, unsigned int flags)
>> +{
>> +    struct bpf_iter_css_kern *kit = (void *)it;
>> +
>> +    BUILD_BUG_ON(sizeof(struct bpf_iter_css_kern) != sizeof(struct 
>> bpf_iter_css));
>> +    BUILD_BUG_ON(__alignof__(struct bpf_iter_css_kern) != 
>> __alignof__(struct bpf_iter_css));
>> +
> 
> This would cause the fail of netdev/build_32bit CI 
> (https://netdev.bots.linux.dev/static/nipa/790929/13412333/build_32bit/stderr):
> 
> tools/testing/selftests/kvm/settings: warning: ignored by one of the 
> .gitignore files
> ../kernel/bpf/cgroup_iter.c:308:17: warning: no previous prototype for 
> ‘bpf_iter_css_new’ [-Wmissing-prototypes]
>    308 | __bpf_kfunc int bpf_iter_css_new(struct bpf_iter_css *it,
>        |                 ^~~~~~~~~~~~~~~~
> ../kernel/bpf/cgroup_iter.c:332:41: warning: no previous prototype for 
> ‘bpf_iter_css_next’ [-Wmissing-prototypes]
>    332 | __bpf_kfunc struct cgroup_subsys_state 
> *bpf_iter_css_next(struct bpf_iter_css *it)
>        |                                         ^~~~~~~~~~~~~~~~~
> ../kernel/bpf/cgroup_iter.c:353:18: warning: no previous prototype for 
> ‘bpf_iter_css_destroy’ [-Wmissing-prototypes]
>    353 | __bpf_kfunc void bpf_iter_css_destroy(struct bpf_iter_css *it)
>        |                  ^~~~~~~~~~~~~~~~~~~~
> In file included from <command-line>:
> ../kernel/bpf/cgroup_iter.c: In function ‘bpf_iter_css_new’:
> ./../include/linux/compiler_types.h:425:45: error: call to 
> ‘__compiletime_assert_322’ declared with attribute error: BUILD_BUG_ON 
> failed: sizeof(struct bpf_iter_css_kern) != sizeof(struct bpf_iter_css)
>    425 |         _compiletime_assert(condition, msg, 
> __compiletime_assert_, __COUNTER__)
>        |                                             ^
> ./../include/linux/compiler_types.h:406:25: note: in definition of macro 
> ‘__compiletime_assert’
>    406 |                         prefix ## suffix();         \
>        |                         ^~~~~~
> ./../include/linux/compiler_types.h:425:9: note: in expansion of macro 
> ‘_compiletime_assert’
>    425 |         _compiletime_assert(condition, msg, 
> __compiletime_assert_, __COUNTER__)
>        |         ^~~~~~~~~~~~~~~~~~~
> ../include/linux/build_bug.h:39:37: note: in expansion of macro 
> ‘compiletime_assert’
>     39 | #define BUILD_BUG_ON_MSG(cond, msg) compiletime_assert(!(cond), 
> msg)
>        |                                     ^~~~~~~~~~~~~~~~~~
> ../include/linux/build_bug.h:50:9: note: in expansion of macro 
> ‘BUILD_BUG_ON_MSG’
>     50 |         BUILD_BUG_ON_MSG(condition, "BUILD_BUG_ON failed: " 
> #condition)
>        |         ^~~~~~~~~~~~~~~~
> ../kernel/bpf/cgroup_iter.c:313:9: note: in expansion of macro 
> ‘BUILD_BUG_ON’
>    313 |         BUILD_BUG_ON(sizeof(struct bpf_iter_css_kern) != 
> sizeof(struct bpf_iter_css));
> 
> 
> The reason seems on 32-bit machine, sizeof(struct bpf_iter_css) is 24 
> and sizeof(struct bpf_iter_css_kern) is 16.
> 
> I was wondering whether the BUILD_BUG_ON check is necessary. Looking at 
> the struct bpf_list_node and struct bpf_list_node_kern wich are very 
> similay to bpf_iter_css, I didn't see the BUILD_BUG_ON check when 
> convert from (struct bpf_list_node *) to (struct bpf_list_node_kern *)
> 
> /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */
> struct bpf_list_node_kern {
>      struct list_head list_head;
>      void *owner;
> } __attribute__((aligned(8)));
> 
> struct bpf_list_node {
>      __u64 :64;
>      __u64 :64;
>      __u64 :64;
> } __attribute__((aligned(8)));
> 
> __bpf_kfunc int bpf_list_push_back_impl(struct bpf_list_head *head,
>                      struct bpf_list_node *node,
>                      void *meta__ign, u64 off)
> {
>      struct bpf_list_node_kern *n = (void *)node;
> 
> }
> 

or we can change the BUILD_BUG_ON check, like bpf_timer_kern in 
bpf_timer_init:

--- a/kernel/bpf/cgroup_iter.c
+++ b/kernel/bpf/cgroup_iter.c
@@ -310,7 +310,7 @@ __bpf_kfunc int bpf_iter_css_new(struct bpf_iter_css 
*it,
  {
         struct bpf_iter_css_kern *kit = (void *)it;

-       BUILD_BUG_ON(sizeof(struct bpf_iter_css_kern) != sizeof(struct 
bpf_iter_css));
+       BUILD_BUG_ON(sizeof(struct bpf_iter_css_kern) > sizeof(struct 
bpf_iter_css));
         BUILD_BUG_ON(__alignof__(struct bpf_iter_css_kern) != 
__alignof__(struct bpf_iter_css));

         kit->start = NULL;
diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
index 773be9a221f5..0772545568f1 100644
--- a/kernel/bpf/task_iter.c
+++ b/kernel/bpf/task_iter.c
@@ -877,7 +877,7 @@ __bpf_kfunc int bpf_iter_task_new(struct 
bpf_iter_task *it,
  {
         struct bpf_iter_task_kern *kit = (void *)it;

-       BUILD_BUG_ON(sizeof(struct bpf_iter_task_kern) != sizeof(struct 
bpf_iter_task));
+       BUILD_BUG_ON(sizeof(struct bpf_iter_task_kern) > sizeof(struct 
bpf_iter_task));
         BUILD_BUG_ON(__alignof__(struct bpf_iter_task_kern) !=
                                         __alignof__(struct bpf_iter_task));