Array `data` in `struct hfi_sfr` is being used as a fake flexible array
at run-time:

drivers/media/platform/qcom/venus/hfi_venus.c:
1033         p = memchr(sfr->data, '\0', sfr->buf_size);
1034         /*
1035          * SFR isn't guaranteed to be NULL terminated since SYS_ERROR indicates
1036          * that Venus is in the process of crashing.
1037          */
1038         if (!p)
1039                 sfr->data[sfr->buf_size - 1] = '\0';
1040
1041         dev_err_ratelimited(dev, "SFR message from FW: %s\n", sfr->data);

Fake flexible arrays are deprecated, and should be replaced by
flexible-array members. So, replace one-element array with a
flexible-array member in `struct hfi_sfr`.

While there, also annotate array `data` with __counted_by() to prepare
for the coming implementation by GCC and Clang of the __counted_by
attribute. Flexible array members annotated with __counted_by can have
their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for
array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family
functions).

This results in no differences in binary output.

This issue was found with the help of Coccinelle, and audited and fixed
manually.

Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---
 drivers/media/platform/qcom/venus/hfi_cmds.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/platform/qcom/venus/hfi_cmds.h b/drivers/media/platform/qcom/venus/hfi_cmds.h
index dd9c5066442d..20acd412ee7b 100644
--- a/drivers/media/platform/qcom/venus/hfi_cmds.h
+++ b/drivers/media/platform/qcom/venus/hfi_cmds.h
@@ -242,7 +242,7 @@ struct hfi_session_parse_sequence_header_pkt {
 
 struct hfi_sfr {
 	u32 buf_size;
-	u8 data[1];
+	u8 data[] __counted_by(buf_size);
 };
 
 struct hfi_sys_test_ssr_pkt {
-- 
2.34.1

On Mon, Oct 09, 2023 at 12:42:05PM -0600, Gustavo A. R. Silva wrote:
> Array `data` in `struct hfi_sfr` is being used as a fake flexible array
> at run-time:
> 
> drivers/media/platform/qcom/venus/hfi_venus.c:
> 1033         p = memchr(sfr->data, '\0', sfr->buf_size);
> 1034         /*
> 1035          * SFR isn't guaranteed to be NULL terminated since SYS_ERROR indicates
> 1036          * that Venus is in the process of crashing.
> 1037          */
> 1038         if (!p)
> 1039                 sfr->data[sfr->buf_size - 1] = '\0';
> 1040
> 1041         dev_err_ratelimited(dev, "SFR message from FW: %s\n", sfr->data);
> 
> Fake flexible arrays are deprecated, and should be replaced by
> flexible-array members. So, replace one-element array with a
> flexible-array member in `struct hfi_sfr`.
> 
> While there, also annotate array `data` with __counted_by() to prepare
> for the coming implementation by GCC and Clang of the __counted_by
> attribute. Flexible array members annotated with __counted_by can have
> their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for
> array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family
> functions).
> 
> This results in no differences in binary output.

Thanks for checking!

> 
> This issue was found with the help of Coccinelle, and audited and fixed
> manually.
> 
> Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

On (23/10/09 12:52), Kees Cook wrote:
> On Mon, Oct 09, 2023 at 12:42:05PM -0600, Gustavo A. R. Silva wrote:
> > Array `data` in `struct hfi_sfr` is being used as a fake flexible array
> > at run-time:
> > 
> > drivers/media/platform/qcom/venus/hfi_venus.c:
> > 1033         p = memchr(sfr->data, '\0', sfr->buf_size);
> > 1034         /*
> > 1035          * SFR isn't guaranteed to be NULL terminated since SYS_ERROR indicates
> > 1036          * that Venus is in the process of crashing.
> > 1037          */
> > 1038         if (!p)
> > 1039                 sfr->data[sfr->buf_size - 1] = '\0';
> > 1040
> > 1041         dev_err_ratelimited(dev, "SFR message from FW: %s\n", sfr->data);
> > 
> > Fake flexible arrays are deprecated, and should be replaced by
> > flexible-array members. So, replace one-element array with a
> > flexible-array member in `struct hfi_sfr`.
> > 
> > While there, also annotate array `data` with __counted_by() to prepare
> > for the coming implementation by GCC and Clang of the __counted_by
> > attribute. Flexible array members annotated with __counted_by can have
> > their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for
> > array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family
> > functions).
> > 
> > This results in no differences in binary output.
> 
> Thanks for checking!

Sorry for shameless plug, a quick question: has any compiler implemented
support for counted_by() at this point?

> Sorry for shameless plug, a quick question: has any compiler implemented
> support for counted_by() at this point?
> 

Not yet. And at least for GCC, it's expected to be released in v15.

Thanks
--
Gustavo

On (24/01/09 07:17), Gustavo A. R. Silva wrote:
> 
> > Sorry for shameless plug, a quick question: has any compiler implemented
> > support for counted_by() at this point?
> > 
> 
> Not yet. And at least for GCC, it's expected to be released in v15.

I see. Thank you.

I got confused by include/linux/compiler_attributes.h comment, as I'm on
clang-18 currently, seems that we need to bump min compilers version.
Oh, and clang link 404-s on me. I'll send a quick patch, I guess.

On 1/9/24 07:28, Sergey Senozhatsky wrote:
> On (24/01/09 07:17), Gustavo A. R. Silva wrote:
>>
>>> Sorry for shameless plug, a quick question: has any compiler implemented
>>> support for counted_by() at this point?
>>>
>>
>> Not yet. And at least for GCC, it's expected to be released in v15.
> 
> I see. Thank you.
> 
> I got confused by include/linux/compiler_attributes.h comment, as I'm on
> clang-18 currently, seems that we need to bump min compilers version.

Ah yes, compiler devs have been running into some issues, and they had to
postpone the release of the attribute.

> Oh, and clang link 404-s on me. I'll send a quick patch, I guess.
> 

You're right, ick!

--
Gustavo