at 12:04 AM, Keith Busch <kbu...@kernel.org> wrote:
> On Thu, Jun 30, 2022 at 11:22:31AM +0800, Jinhao Fan wrote:
>> +static int nvme_init_sq_ioeventfd(NvmeSQueue *sq)
>> +{
>> + NvmeCtrl *n = sq->ctrl;
>> + uint16_t offset = sq->sqid << 3;
>> + int ret;
>> +
>> + ret = event_notifier_init(&sq->notifier, 0);
>> + if (ret < 0) {
>> + return ret;
>> + }
>> +
>> + event_notifier_set_handler(&sq->notifier, nvme_sq_notifier);
>> + memory_region_add_eventfd(&n->iomem,
>> + 0x1000 + offset, 4, false, 0, &sq->notifier);
>> +
>> + return 0;
>> +}
>> +
>> static void nvme_free_sq(NvmeSQueue *sq, NvmeCtrl *n)
>> {
>> n->sq[sq->sqid] = NULL;
>> timer_free(sq->timer);
>> + event_notifier_cleanup(&sq->notifier);
>> g_free(sq->io_req);
>> if (sq->sqid) {
>> g_free(sq);
>
> I believe there needs to be a corresponding memory_region_del_eventfd() when
> deleting the queue before event_notifier_cleanup(), otherwise you'll leak
> invalid listeners.
Thanks. Fixed this in v3.
