dmaengine: idxd: support reporting of halt interrupt
Hello Dave Jiang,
The patch 5b0c68c473a1: "dmaengine: idxd: support reporting of halt interrupt" from Apr 20, 2021, leads to the following Smatch static checker warning:
drivers/dma/idxd/device.c:431 idxd_wq_quiesce() warn: sleeping in atomic context
drivers/dma/idxd/device.c
428 void idxd_wq_quiesce(struct idxd_wq *wq)
429 {
430 percpu_ref_kill(&wq->wq_active);
--> 431 wait_for_completion(&wq->wq_dead);
432 percpu_ref_exit(&wq->wq_active);
433 }The call tree is:
process_misc_interrupts() <- disables preempt
-> idxd_wqs_quiesce()
-> idxd_wq_quiesce()
drivers/dma/idxd/irq.c
124 gensts.bits = ioread32(idxd->reg_base + IDXD_GENSTATS_OFFSET);
125 if (gensts.state == IDXD_DEVICE_STATE_HALT) {
126 idxd->state = IDXD_DEV_HALTED;
127 if (gensts.reset_type == IDXD_DEVICE_RESET_SOFTWARE) {
128 /*
129 * If we need a software reset, we will throw the work
130 * on a system workqueue in order to allow interrupts
131 * for the device command completions.
132 */
133 INIT_WORK(&idxd->work, idxd_device_reinit);
134 queue_work(idxd->wq, &idxd->work);
135 } else {
136 spin_lock_bh(&idxd->dev_lock);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^Maybe take this spinlock later?
137 idxd_wqs_quiesce(idxd);
^^^^^^^^^^^^^^^^^^^^^^Sleeps.
138 idxd_wqs_unmap_portal(idxd); 139 idxd_device_clear_state(idxd); 140 dev_err(&idxd->pdev->dev, 141 "idxd halted, need %s.\n", 142 gensts.reset_type == IDXD_DEVICE_RESET_FLR ? 143 "FLR" : "system reset");
144 spin_unlock_bh(&idxd->dev_lock);
145 return -ENXIO;
146 }
147 }
148
149 return 0;
regards,
dan carpenter