Re: [PATCH v3] thunderbolt: fix bandwidth group reservation indexing

[Mika Westerberg]

Hi,

On Wed, Jun 24, 2026 at 01:07:19PM +0800, raoxu wrote:
> From: Xu Rao <raoxu@xxxxxxxxxxxxx>
> 
> Group ID 0 is reserved, while valid bandwidth groups use IDs 1 through
> 7. tb_consumed_dp_bandwidth() uses the Group ID directly to index
> its group_reserved[] array.
> 
> Currently group_reserved[] has only seven entries, covering indices 0
> through 6. A tunnel in Group ID 7 therefore reads and may write one
> entry past the end of the array, and that group's reserved bandwidth is
> not included in the consumed bandwidth total.
> 
> Include the reserved Group ID 0 in MAX_GROUPS and map tb_cm::groups[]
> directly by Group ID. Initialize every entry with its array index, but
> skip index 0 when allocating a free group or restoring a group reported
> by the hardware. This keeps Group ID 0 reserved while making IDs 1
> through 7 valid indices in both arrays.

I looked at this again and realized that your v1 was almost okay but
instead of the -1 we should do this and just this:

tb_consumed_dp_bandwidth()
{
	 int group_reserved[MAX_GROUPS + 1] = {};
	 ...

keep everything else as is. This should solve the issue, right?