On 5/29/24 2:18 AM, Li RongQing wrote:
> list_for_each_entry_safe() should be used when the descriptor will be
> freed in idxd_desc_complete(), Otherwise the freed descriptor will be
> dereferenced to get its next, and always deletes freed descriptor from
> list firstly
>
> Fixes: 16e19e11228b ("dmaengine: idxd: Fix list corruption in description completion")
> Signed-off-by: Li RongQing <lirongqing@xxxxxxxxx>
> ---
> drivers/dma/idxd/irq.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/dma/idxd/irq.c b/drivers/dma/idxd/irq.c
> index 8dc029c..5571de1 100644
> --- a/drivers/dma/idxd/irq.c
> +++ b/drivers/dma/idxd/irq.c
> @@ -573,6 +573,8 @@ static void irq_process_pending_llist(struct idxd_irq_entry *irq_entry)
> * Check against the original status as ABORT is software defined
> * and 0xff, which DSA_COMP_STATUS_MASK can mask out.
> */
> + list_del(&desc->list);
I don't think this is needed. The iterator is working over a detached llist and nothing to do with 'struct list_head'. Otherwise everything else look ok.
> +
> if (unlikely(desc->completion->status == IDXD_COMP_DESC_ABORT)) {
> idxd_desc_complete(desc, IDXD_COMPLETE_ABORT, true);
> continue;
> @@ -611,11 +613,13 @@ static void irq_process_work_list(struct idxd_irq_entry *irq_entry)
>
> spin_unlock(&irq_entry->list_lock);
>
> - list_for_each_entry(desc, &flist, list) {
> + list_for_each_entry_safe(desc, n, &flist, list) {
> /*
> * Check against the original status as ABORT is software defined
> * and 0xff, which DSA_COMP_STATUS_MASK can mask out.
> */
> + list_del(&desc->list);
> +
> if (unlikely(desc->completion->status == IDXD_COMP_DESC_ABORT)) {
> idxd_desc_complete(desc, IDXD_COMPLETE_ABORT, true);
> continue;
