2023-10-01 13:27:07

by Christophe JAILLET

Subject: [PATCH net-next 2/2] net: openvswitch: Annotate struct mask_array with __counted_byUse struct_size()

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 checking via CONFIG_UBSAN_BOUNDS
(for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family
functions).

Signed-off-by: Christophe JAILLET <[email protected]>
---
This patch is part of a work done in parallel of what is currently worked
on by Kees Cook.

My patches are only related to corner cases that do NOT match the
semantic of his Coccinelle script[1].

In this case, in tbl_mask_array_alloc(), several things are allocated with
a single allocation. Then, some pointer arithmetic computes the address of
the memory after the flex-array.

[1] https://github.com/kees/kernel-tools/blob/trunk/coccinelle/examples/counted_by.cocci
---
net/openvswitch/flow_table.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/openvswitch/flow_table.h b/net/openvswitch/flow_table.h
index 9e659db78c05..8d9e83b4d62c 100644
--- a/net/openvswitch/flow_table.h
+++ b/net/openvswitch/flow_table.h
@@ -48,7 +48,7 @@ struct mask_array {
int count, max;
struct mask_array_stats __percpu *masks_usage_stats;
u64 *masks_usage_zero_cntr;
- struct sw_flow_mask __rcu *masks[];
+ struct sw_flow_mask __rcu *masks[] __counted_by(size);
};

struct table_instance {
--
2.34.1

2023-10-02 17:04:45

by Ilya Maximets

Subject: Re: [ovs-dev] [PATCH net-next 2/2] net: openvswitch: Annotate struct mask_array with __counted_byUse struct_size()

On 10/1/23 13:07, Christophe JAILLET wrote:
> 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 checking via CONFIG_UBSAN_BOUNDS
> (for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family
> functions).
>
> Signed-off-by: Christophe JAILLET <[email protected]>
> ---
> This patch is part of a work done in parallel of what is currently worked
> on by Kees Cook.
>
> My patches are only related to corner cases that do NOT match the
> semantic of his Coccinelle script[1].
>
> In this case, in tbl_mask_array_alloc(), several things are allocated with
> a single allocation. Then, some pointer arithmetic computes the address of
> the memory after the flex-array.
>
> [1] https://github.com/kees/kernel-tools/blob/trunk/coccinelle/examples/counted_by.cocci
> ---
> net/openvswitch/flow_table.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/openvswitch/flow_table.h b/net/openvswitch/flow_table.h
> index 9e659db78c05..8d9e83b4d62c 100644
> --- a/net/openvswitch/flow_table.h
> +++ b/net/openvswitch/flow_table.h
> @@ -48,7 +48,7 @@ struct mask_array {
> int count, max;
> struct mask_array_stats __percpu *masks_usage_stats;
> u64 *masks_usage_zero_cntr;
> - struct sw_flow_mask __rcu *masks[];
> + struct sw_flow_mask __rcu *masks[] __counted_by(size);

Did you mean 'max'? There is no 'size' in the structure.

Also, the patch subject is messed up a bit.

Best regards, Ilya Maximets.

2023-10-10 05:28:25

by Christophe JAILLET

Subject: Re: [ovs-dev] [PATCH net-next 2/2] net: openvswitch: Annotate struct mask_array with __counted_byUse struct_size()

Le 02/10/2023 à 18:51, Ilya Maximets a écrit :
> On 10/1/23 13:07, Christophe JAILLET wrote:
>> 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 checking via CONFIG_UBSAN_BOUNDS
>> (for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family
>> functions).
>>
>> Signed-off-by: Christophe JAILLET <[email protected]>
>> ---
>> This patch is part of a work done in parallel of what is currently worked
>> on by Kees Cook.
>>
>> My patches are only related to corner cases that do NOT match the
>> semantic of his Coccinelle script[1].
>>
>> In this case, in tbl_mask_array_alloc(), several things are allocated with
>> a single allocation. Then, some pointer arithmetic computes the address of
>> the memory after the flex-array.
>>
>> [1] https://github.com/kees/kernel-tools/blob/trunk/coccinelle/examples/counted_by.cocci
>> ---
>> net/openvswitch/flow_table.h | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/net/openvswitch/flow_table.h b/net/openvswitch/flow_table.h
>> index 9e659db78c05..8d9e83b4d62c 100644
>> --- a/net/openvswitch/flow_table.h
>> +++ b/net/openvswitch/flow_table.h
>> @@ -48,7 +48,7 @@ struct mask_array {
>> int count, max;
>> struct mask_array_stats __percpu *masks_usage_stats;
>> u64 *masks_usage_zero_cntr;
>> - struct sw_flow_mask __rcu *masks[];
>> + struct sw_flow_mask __rcu *masks[] __counted_by(size);
>
> Did you mean 'max'? There is no 'size' in the structure.

Hi,

Of courtse, yes. I'll resend.

'size' is the name of the variable that is written in mask_array->max in
tbl_mask_array_alloc()

>
> Also, the patch subject is messed up a bit.

Yes.
Will fix it as well.

CJ

>
> Best regards, Ilya Maximets.
>