2020-08-14 10:01:17

by Miaohe Lin

Subject: [PATCH] bpf: Convert to use the preferred fallthrough macro

Convert the uses of fallthrough comments to fallthrough macro.

Signed-off-by: Miaohe Lin <[email protected]>
---
kernel/bpf/cgroup.c | 2 +-
kernel/bpf/cpumap.c | 2 +-
kernel/bpf/syscall.c | 2 +-
kernel/bpf/verifier.c | 6 +++---
4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
index 83ff127ef7ae..e21de4f1754c 100644
--- a/kernel/bpf/cgroup.c
+++ b/kernel/bpf/cgroup.c
@@ -1794,7 +1794,7 @@ static bool cg_sockopt_is_valid_access(int off, int size,
return prog->expected_attach_type ==
BPF_CGROUP_GETSOCKOPT;
case offsetof(struct bpf_sockopt, optname):
- /* fallthrough */
+ fallthrough;
case offsetof(struct bpf_sockopt, level):
if (size != size_default)
return false;
diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index f1c46529929b..6386b7bb98f2 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -279,7 +279,7 @@ static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu,
break;
default:
bpf_warn_invalid_xdp_action(act);
- /* fallthrough */
+ fallthrough;
case XDP_DROP:
xdp_return_frame(xdpf);
stats->drop++;
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 86299a292214..1bf960aa615c 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -2029,7 +2029,7 @@ bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
case BPF_PROG_TYPE_EXT:
if (expected_attach_type)
return -EINVAL;
- /* fallthrough */
+ fallthrough;
default:
return 0;
}
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index ef938f17b944..1e7f34663f86 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -2639,7 +2639,7 @@ static bool may_access_direct_pkt_data(struct bpf_verifier_env *env,
case BPF_PROG_TYPE_CGROUP_SKB:
if (t == BPF_WRITE)
return false;
- /* fallthrough */
+ fallthrough;

/* Program types with direct read + write access go here! */
case BPF_PROG_TYPE_SCHED_CLS:
@@ -5236,7 +5236,7 @@ static int adjust_ptr_min_max_vals(struct bpf_verifier_env *env,
off_reg == dst_reg ? dst : src);
return -EACCES;
}
- /* fall-through */
+ fallthrough;
default:
break;
}
@@ -10988,7 +10988,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
default:
if (!prog_extension)
return -EINVAL;
- /* fallthrough */
+ fallthrough;
case BPF_MODIFY_RETURN:
case BPF_LSM_MAC:
case BPF_TRACE_FENTRY:
--
2.19.1

2020-08-14 16:58:57

by Yonghong Song

Subject: Re: [PATCH] bpf: Convert to use the preferred fallthrough macro

On 8/14/20 2:16 AM, Miaohe Lin wrote:
> Convert the uses of fallthrough comments to fallthrough macro.
>
> Signed-off-by: Miaohe Lin <[email protected]>

This is not a bug fix but rather an enhancement so not sure whether
this should push to bpf tree or wait until bpf-next.

It may be worthwhile to mention Commit 294f69e662d1
("compiler_attributes.h: Add 'fallthrough' pseudo keyword for
switch/case use") so people can understand why this patch is
needed.

With above suggestions,
Acked-by: Yonghong Song <[email protected]>

> ---
> kernel/bpf/cgroup.c | 2 +-
> kernel/bpf/cpumap.c | 2 +-
> kernel/bpf/syscall.c | 2 +-
> kernel/bpf/verifier.c | 6 +++---
> 4 files changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
> index 83ff127ef7ae..e21de4f1754c 100644
> --- a/kernel/bpf/cgroup.c
> +++ b/kernel/bpf/cgroup.c
> @@ -1794,7 +1794,7 @@ static bool cg_sockopt_is_valid_access(int off, int size,
> return prog->expected_attach_type ==
> BPF_CGROUP_GETSOCKOPT;
> case offsetof(struct bpf_sockopt, optname):
> - /* fallthrough */
> + fallthrough;
> case offsetof(struct bpf_sockopt, level):
> if (size != size_default)
> return false;
> diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
> index f1c46529929b..6386b7bb98f2 100644
> --- a/kernel/bpf/cpumap.c
> +++ b/kernel/bpf/cpumap.c
> @@ -279,7 +279,7 @@ static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu,
> break;
> default:
> bpf_warn_invalid_xdp_action(act);
> - /* fallthrough */
> + fallthrough;
> case XDP_DROP:
> xdp_return_frame(xdpf);
> stats->drop++;
> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> index 86299a292214..1bf960aa615c 100644
> --- a/kernel/bpf/syscall.c
> +++ b/kernel/bpf/syscall.c
> @@ -2029,7 +2029,7 @@ bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
> case BPF_PROG_TYPE_EXT:
> if (expected_attach_type)
> return -EINVAL;
> - /* fallthrough */
> + fallthrough;
> default:
> return 0;
> }
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index ef938f17b944..1e7f34663f86 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -2639,7 +2639,7 @@ static bool may_access_direct_pkt_data(struct bpf_verifier_env *env,
> case BPF_PROG_TYPE_CGROUP_SKB:
> if (t == BPF_WRITE)
> return false;
> - /* fallthrough */
> + fallthrough;
>
> /* Program types with direct read + write access go here! */
> case BPF_PROG_TYPE_SCHED_CLS:
> @@ -5236,7 +5236,7 @@ static int adjust_ptr_min_max_vals(struct bpf_verifier_env *env,
> off_reg == dst_reg ? dst : src);
> return -EACCES;
> }
> - /* fall-through */
> + fallthrough;
> default:
> break;
> }
> @@ -10988,7 +10988,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
> default:
> if (!prog_extension)
> return -EINVAL;
> - /* fallthrough */
> + fallthrough;
> case BPF_MODIFY_RETURN:
> case BPF_LSM_MAC:
> case BPF_TRACE_FENTRY:
>

2020-08-14 17:13:05

by Andrii Nakryiko

Subject: Re: [PATCH] bpf: Convert to use the preferred fallthrough macro

On Fri, Aug 14, 2020 at 2:58 AM Miaohe Lin <[email protected]> wrote:
>
> Convert the uses of fallthrough comments to fallthrough macro.
>
> Signed-off-by: Miaohe Lin <[email protected]>
> ---
> kernel/bpf/cgroup.c | 2 +-
> kernel/bpf/cpumap.c | 2 +-
> kernel/bpf/syscall.c | 2 +-
> kernel/bpf/verifier.c | 6 +++---
> 4 files changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
> index 83ff127ef7ae..e21de4f1754c 100644
> --- a/kernel/bpf/cgroup.c
> +++ b/kernel/bpf/cgroup.c
> @@ -1794,7 +1794,7 @@ static bool cg_sockopt_is_valid_access(int off, int size,
> return prog->expected_attach_type ==
> BPF_CGROUP_GETSOCKOPT;
> case offsetof(struct bpf_sockopt, optname):
> - /* fallthrough */
> + fallthrough;

this fallthrough is not event necessary, let's drop it instead?

> case offsetof(struct bpf_sockopt, level):
> if (size != size_default)
> return false;

[...]

2020-08-16 07:31:50

by Miaohe Lin

Subject: Re: [PATCH] bpf: Convert to use the preferred fallthrough macro

Andrii Nakryiko <[email protected]> wrote:
On Fri, Aug 14, 2020 at 2:58 AM Miaohe Lin <[email protected]> wrote:
>>
>> Convert the uses of fallthrough comments to fallthrough macro.
>> @@ -1794,7 +1794,7 @@ static bool cg_sockopt_is_valid_access(int off, int size,
>> return prog->expected_attach_type ==
>> BPF_CGROUP_GETSOCKOPT;
>> case offsetof(struct bpf_sockopt, optname):
>> - /* fallthrough */
>> + fallthrough;
>
>this fallthrough is not event necessary, let's drop it instead?

Will do. Many thanks.

>
>> case offsetof(struct bpf_sockopt, level):
>> if (size != size_default)
>> return false;
>
> [...]

2020-08-16 09:11:01

by Miaohe Lin

Subject: Re: [PATCH] bpf: Convert to use the preferred fallthrough macro

Yonghong Song <[email protected]> wrote:
>On 8/14/20 2:16 AM, Miaohe Lin wrote:
>> Convert the uses of fallthrough comments to fallthrough macro.
>>
>> Signed-off-by: Miaohe Lin <[email protected]>
>
>This is not a bug fix but rather an enhancement so not sure whether this should push to bpf tree or wait until bpf-next.
>
>It may be worthwhile to mention Commit 294f69e662d1
> ("compiler_attributes.h: Add 'fallthrough' pseudo keyword for switch/case use") so people can understand why this patch is needed.
>

Will do. Many thanks.

>With above suggestions,
>Acked-by: Yonghong Song <[email protected]>
>