[PATCH dwarves] btf_encoder: Fix dwarf int type with greater-than-16 byte issue

[PATCH dwarves] btf_encoder: Fix dwarf int type with greater-than-16 byte issue

[Yonghong Song]

Nick Desaulniers and Xin Liu separately reported that int type might
have greater-than-16 byte size ([1] and [2]). More specifically, the
reported int type sizes are 1024 and 64 bytes.

The libbpf and bpf program does not really support any int type greater
than 16 bytes. Therefore, with current pahole, btf encoding will fail
with greater-than-16 byte int types.

Since for now bpf does not support '> 16' bytes int type, the simplest
way is to sanitize such types, similar to existing conditions like
'!byte_sz' and 'byte_sz & (byte_sz - 1)'. This way, pahole won't
call libbpf with an unsupported int type size. The patch [3] was
proposed before. Now I resubmitted this patch as there are another
failure due to the same issue.

  [1] https://github.com/libbpf/libbpf/pull/680
  [2] https://lore.kernel.org/bpf/20240422144538.351722-1-liuxin350@huawei.com/
  [3] https://lore.kernel.org/bpf/20230426055030.3743074-1-yhs@fb.com/

Cc: Xin Liu <liuxin350@huawei.com>
Cc: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
---
 btf_encoder.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/btf_encoder.c b/btf_encoder.c
index e1e3529..19e9d90 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c
@@ -393,7 +393,7 @@ static int32_t btf_encoder__add_base_type(struct btf_encoder *encoder, const str
 	 * these non-regular int types to avoid libbpf/kernel complaints.
 	 */
 	byte_sz = BITS_ROUNDUP_BYTES(bt->bit_size);
-	if (!byte_sz || (byte_sz & (byte_sz - 1))) {
+	if (!byte_sz || (byte_sz & (byte_sz - 1)) || byte_sz > 16) {
 		name = "__SANITIZED_FAKE_INT__";
 		byte_sz = 4;
 	}
-- 
2.43.0

Re: [PATCH dwarves] btf_encoder: Fix dwarf int type with greater-than-16 byte issue

[Alan Maguire]

On 24/04/2024 23:35, Yonghong Song wrote:
> Nick Desaulniers and Xin Liu separately reported that int type might
> have greater-than-16 byte size ([1] and [2]). More specifically, the
> reported int type sizes are 1024 and 64 bytes.
> 
> The libbpf and bpf program does not really support any int type greater
> than 16 bytes. Therefore, with current pahole, btf encoding will fail
> with greater-than-16 byte int types.
> 
> Since for now bpf does not support '> 16' bytes int type, the simplest
> way is to sanitize such types, similar to existing conditions like
> '!byte_sz' and 'byte_sz & (byte_sz - 1)'. This way, pahole won't
> call libbpf with an unsupported int type size. The patch [3] was
> proposed before. Now I resubmitted this patch as there are another
> failure due to the same issue.
> 
>   [1] https://github.com/libbpf/libbpf/pull/680
>   [2] https://lore.kernel.org/bpf/20240422144538.351722-1-liuxin350@huawei.com/
>   [3] https://lore.kernel.org/bpf/20230426055030.3743074-1-yhs@fb.com/
> 
> Cc: Xin Liu <liuxin350@huawei.com>
> Cc: Alan Maguire <alan.maguire@oracle.com>
> Signed-off-by: Yonghong Song <yonghong.song@linux.dev>

Reviewed-by: Alan Maguire <alan.maguire@oracle.com>

> ---
>  btf_encoder.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/btf_encoder.c b/btf_encoder.c
> index e1e3529..19e9d90 100644
> --- a/btf_encoder.c
> +++ b/btf_encoder.c
> @@ -393,7 +393,7 @@ static int32_t btf_encoder__add_base_type(struct btf_encoder *encoder, const str
>  	 * these non-regular int types to avoid libbpf/kernel complaints.
>  	 */
>  	byte_sz = BITS_ROUNDUP_BYTES(bt->bit_size);
> -	if (!byte_sz || (byte_sz & (byte_sz - 1))) {
> +	if (!byte_sz || (byte_sz & (byte_sz - 1)) || byte_sz > 16) {
>  		name = "__SANITIZED_FAKE_INT__";
>  		byte_sz = 4;
>  	}

Re: [PATCH dwarves] btf_encoder: Fix dwarf int type with greater-than-16 byte issue

[Xin Liu]

On Wed, 24 Apr 2024 15:35:38 -0700 Yonghong Song <yonghong.song@linux.dev> wrote:
> Nick Desaulniers and Xin Liu separately reported that int type might
> have greater-than-16 byte size ([1] and [2]). More specifically, the
> reported int type sizes are 1024 and 64 bytes.
> 
> The libbpf and bpf program does not really support any int type greater
> than 16 bytes. Therefore, with current pahole, btf encoding will fail
> with greater-than-16 byte int types.
> 
> Since for now bpf does not support '> 16' bytes int type, the simplest
> way is to sanitize such types, similar to existing conditions like
> '!byte_sz' and 'byte_sz & (byte_sz - 1)'. This way, pahole won't
> call libbpf with an unsupported int type size. The patch [3] was
> proposed before. Now I resubmitted this patch as there are another
> failure due to the same issue.
> 
>   [1] https://github.com/libbpf/libbpf/pull/680
>   [2] https://lore.kernel.org/bpf/20240422144538.351722-1-liuxin350@huawei.com/
>   [3] https://lore.kernel.org/bpf/20230426055030.3743074-1-yhs@fb.com/
> 
> Cc: Xin Liu <liuxin350@huawei.com>
> Cc: Alan Maguire <alan.maguire@oracle.com>
> Signed-off-by: Yonghong Song <yonghong.song@linux.dev>

Reviewed-by: Xin Liu <liuxin350@huawei.com>

> ---
>  btf_encoder.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/btf_encoder.c b/btf_encoder.c
> index e1e3529..19e9d90 100644
> --- a/btf_encoder.c
> +++ b/btf_encoder.c
> @@ -393,7 +393,7 @@ static int32_t btf_encoder__add_base_type(struct btf_encoder *encoder, const str
>  	 * these non-regular int types to avoid libbpf/kernel complaints.
>  	 */
>  	byte_sz = BITS_ROUNDUP_BYTES(bt->bit_size);
> -	if (!byte_sz || (byte_sz & (byte_sz - 1))) {
> +	if (!byte_sz || (byte_sz & (byte_sz - 1)) || byte_sz > 16) {
>  		name = "__SANITIZED_FAKE_INT__";
>  		byte_sz = 4;
>  	}

Re: [PATCH dwarves] btf_encoder: Fix dwarf int type with greater-than-16 byte issue

[Arnaldo Carvalho de Melo]

On Thu, Apr 25, 2024 at 09:43:40PM +0800, Xin Liu wrote:
> On Wed, 24 Apr 2024 15:35:38 -0700 Yonghong Song <yonghong.song@linux.dev> wrote:
> > Nick Desaulniers and Xin Liu separately reported that int type might
> > have greater-than-16 byte size ([1] and [2]). More specifically, the
> > reported int type sizes are 1024 and 64 bytes.
> > 
> > The libbpf and bpf program does not really support any int type greater
> > than 16 bytes. Therefore, with current pahole, btf encoding will fail
> > with greater-than-16 byte int types.
> > 
> > Since for now bpf does not support '> 16' bytes int type, the simplest
> > way is to sanitize such types, similar to existing conditions like
> > '!byte_sz' and 'byte_sz & (byte_sz - 1)'. This way, pahole won't
> > call libbpf with an unsupported int type size. The patch [3] was
> > proposed before. Now I resubmitted this patch as there are another
> > failure due to the same issue.
> > 
> >   [1] https://github.com/libbpf/libbpf/pull/680
> >   [2] https://lore.kernel.org/bpf/20240422144538.351722-1-liuxin350@huawei.com/
> >   [3] https://lore.kernel.org/bpf/20230426055030.3743074-1-yhs@fb.com/
> > 
> > Cc: Xin Liu <liuxin350@huawei.com>
> > Cc: Alan Maguire <alan.maguire@oracle.com>
> > Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
> 
> Reviewed-by: Xin Liu <liuxin350@huawei.com>

Thanks, applied to next,

- Arnaldo
 
> > ---
> >  btf_encoder.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/btf_encoder.c b/btf_encoder.c
> > index e1e3529..19e9d90 100644
> > --- a/btf_encoder.c
> > +++ b/btf_encoder.c
> > @@ -393,7 +393,7 @@ static int32_t btf_encoder__add_base_type(struct btf_encoder *encoder, const str
> >  	 * these non-regular int types to avoid libbpf/kernel complaints.
> >  	 */
> >  	byte_sz = BITS_ROUNDUP_BYTES(bt->bit_size);
> > -	if (!byte_sz || (byte_sz & (byte_sz - 1))) {
> > +	if (!byte_sz || (byte_sz & (byte_sz - 1)) || byte_sz > 16) {
> >  		name = "__SANITIZED_FAKE_INT__";
> >  		byte_sz = 4;
> >  	}