bpf: Let verifier consider {task,cgroup} is trusted in bpf_iter_reg
authorChuyi Zhou <zhouchuyi@bytedance.com>
Tue, 7 Nov 2023 13:22:03 +0000 (21:22 +0800)
committerMartin KaFai Lau <martin.lau@kernel.org>
Tue, 7 Nov 2023 23:24:25 +0000 (15:24 -0800)
commit0de4f50de25af79c2a46db55d70cdbd8f985c6d1
tree02dcfa685f6f84672356ac0c8f1fe8716160eb77
parentd84b139f53e8fa8048f16814c6b2a53d7bc15c3d
bpf: Let verifier consider {task,cgroup} is trusted in bpf_iter_reg

BTF_TYPE_SAFE_TRUSTED(struct bpf_iter__task) in verifier.c wanted to
teach BPF verifier that bpf_iter__task -> task is a trusted ptr. But it
doesn't work well.

The reason is, bpf_iter__task -> task would go through btf_ctx_access()
which enforces the reg_type of 'task' is ctx_arg_info->reg_type, and in
task_iter.c, we actually explicitly declare that the
ctx_arg_info->reg_type is PTR_TO_BTF_ID_OR_NULL.

Actually we have a previous case like this[1] where PTR_TRUSTED is added to
the arg flag for map_iter.

This patch sets ctx_arg_info->reg_type is PTR_TO_BTF_ID_OR_NULL |
PTR_TRUSTED in task_reg_info.

Similarly, bpf_cgroup_reg_info -> cgroup is also PTR_TRUSTED since we are
under the protection of cgroup_mutex and we would check cgroup_is_dead()
in __cgroup_iter_seq_show().

This patch is to improve the user experience of the newly introduced
bpf_iter_css_task kfunc before hitting the mainline. The Fixes tag is
pointing to the commit introduced the bpf_iter_css_task kfunc.

Link[1]:https://lore.kernel.org/all/20230706133932.45883-3-aspsk@isovalent.com/

Fixes: 9c66dc94b62a ("bpf: Introduce css_task open-coded iterator kfuncs")
Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20231107132204.912120-2-zhouchuyi@bytedance.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
kernel/bpf/cgroup_iter.c
kernel/bpf/task_iter.c