[bug report] scsi: ufs: host: mediatek: Set IRQ affinity policy for MCQ mode

[bug report] scsi: ufs: host: mediatek: Set IRQ affinity policy for MCQ mode

[Dan Carpenter]

Hello Peter Wang,

Commit 66e26a4b8a77 ("scsi: ufs: host: mediatek: Set IRQ affinity
policy for MCQ mode") from Jul 22, 2025 (linux-next), leads to the
following *UNPUBLISHED* Smatch static checker warning:

	drivers/ufs/host/ufs-mediatek.c:827 ufs_mtk_mcq_get_irq()
	warn: array off by one? 'host->mcq_intr_info[q_index]'

drivers/ufs/host/ufs-mediatek.c
    812 static u32 ufs_mtk_mcq_get_irq(struct ufs_hba *hba, unsigned int cpu)
    813 {
    814         struct ufs_mtk_host *host = ufshcd_get_variant(hba);
    815         struct blk_mq_tag_set *tag_set = &hba->host->tag_set;
    816         struct blk_mq_queue_map        *map = &tag_set->map[HCTX_TYPE_DEFAULT];
    817         unsigned int nr = map->nr_queues;
    818         unsigned int q_index;
    819 
    820         q_index = map->mq_map[cpu];
    821         if (q_index > nr) {

This really looks like it should be ">= nr" instead of "> nr" but I'm not
certain enough to send a patch for it.  Could you take a look?

    822                 dev_err(hba->dev, "hwq index %d exceed %d\n",
    823                         q_index, nr);
    824                 return MTK_MCQ_INVALID_IRQ;
    825         }
    826 
--> 827         return host->mcq_intr_info[q_index].irq;
    828 }

regards,
dan carpenter

Re: [bug report] scsi: ufs: host: mediatek: Set IRQ affinity policy for MCQ mode

[Peter Wang (王信友)]

On Fri, 2025-08-01 at 16:18 +0300, Dan Carpenter wrote:
> 
> 
>     821         if (q_index > nr) {
> 
> This really looks like it should be ">= nr" instead of "> nr" but I'm
> not
> certain enough to send a patch for it.  Could you take a look?
> 
>     822                 dev_err(hba->dev, "hwq index %d exceed %d\n",
>     823                         q_index, nr);
>     824                 return MTK_MCQ_INVALID_IRQ;
>     825         }
>     826
> --> 827         return host->mcq_intr_info[q_index].irq;
>     828 }
> 
> regards,
> dan carpenter

Hi Dan,

Yes, this is indeed a coding defect that needs to be fixed.
I will correct it.

Thanks.
Peter