@@ -6,6 +6,7 @@ obj-$(CONFIG_PDS_CORE) := pds_core.o
pds_core-y := main.o \
devlink.o \
dev.o \
+ adminq.o \
core.o
pds_core-$(CONFIG_DEBUG_FS) += debugfs.o
new file mode 100644
@@ -0,0 +1,288 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright(c) 2023 Advanced Micro Devices, Inc */
+
+#include <linux/dynamic_debug.h>
+
+#include "core.h"
+
+struct pdsc_wait_context {
+ struct pdsc_qcq *qcq;
+ struct completion wait_completion;
+};
+
+static int pdsc_process_notifyq(struct pdsc_qcq *qcq)
+{
+ union pds_core_notifyq_comp *comp;
+ struct pdsc *pdsc = qcq->pdsc;
+ struct pdsc_cq *cq = &qcq->cq;
+ struct pdsc_cq_info *cq_info;
+ int nq_work = 0;
+ u64 eid;
+
+ cq_info = &cq->info[cq->tail_idx];
+ comp = cq_info->comp;
+ eid = le64_to_cpu(comp->event.eid);
+ while (eid > pdsc->last_eid) {
+ u16 ecode = le16_to_cpu(comp->event.ecode);
+
+ switch (ecode) {
+ case PDS_EVENT_LINK_CHANGE:
+ dev_info(pdsc->dev, "NotifyQ LINK_CHANGE ecode %d eid %lld\n",
+ ecode, eid);
+ break;
+
+ case PDS_EVENT_RESET:
+ dev_info(pdsc->dev, "NotifyQ RESET ecode %d eid %lld\n",
+ ecode, eid);
+ break;
+
+ case PDS_EVENT_XCVR:
+ dev_info(pdsc->dev, "NotifyQ XCVR ecode %d eid %lld\n",
+ ecode, eid);
+ break;
+
+ default:
+ dev_info(pdsc->dev, "NotifyQ ecode %d eid %lld\n",
+ ecode, eid);
+ break;
+ }
+
+ pdsc->last_eid = eid;
+ cq->tail_idx = (cq->tail_idx + 1) & (cq->num_descs - 1);
+ cq_info = &cq->info[cq->tail_idx];
+ comp = cq_info->comp;
+ eid = le64_to_cpu(comp->event.eid);
+
+ nq_work++;
+ }
+
+ qcq->accum_work += nq_work;
+
+ return nq_work;
+}
+
+void pdsc_process_adminq(struct pdsc_qcq *qcq)
+{
+ union pds_core_adminq_comp *comp;
+ struct pdsc_queue *q = &qcq->q;
+ struct pdsc *pdsc = qcq->pdsc;
+ struct pdsc_cq *cq = &qcq->cq;
+ struct pdsc_q_info *q_info;
+ unsigned long irqflags;
+ int nq_work = 0;
+ int aq_work = 0;
+ int credits;
+
+ /* Don't process AdminQ when shutting down */
+ if (pdsc->state & BIT_ULL(PDSC_S_STOPPING_DRIVER)) {
+ dev_err(pdsc->dev, "%s: called while PDSC_S_STOPPING_DRIVER\n",
+ __func__);
+ return;
+ }
+
+ /* Check for NotifyQ event */
+ nq_work = pdsc_process_notifyq(&pdsc->notifyqcq);
+
+ /* Check for empty queue, which can happen if the interrupt was
+ * for a NotifyQ event and there are no new AdminQ completions.
+ */
+ if (q->tail_idx == q->head_idx)
+ goto credits;
+
+ /* Find the first completion to clean,
+ * run the callback in the related q_info,
+ * and continue while we still match done color
+ */
+ spin_lock_irqsave(&pdsc->adminq_lock, irqflags);
+ comp = cq->info[cq->tail_idx].comp;
+ while (pdsc_color_match(comp->color, cq->done_color)) {
+ q_info = &q->info[q->tail_idx];
+ q->tail_idx = (q->tail_idx + 1) & (q->num_descs - 1);
+
+ /* Copy out the completion data */
+ memcpy(q_info->dest, comp, sizeof(*comp));
+
+ complete_all(&q_info->wc->wait_completion);
+
+ if (cq->tail_idx == cq->num_descs - 1)
+ cq->done_color = !cq->done_color;
+ cq->tail_idx = (cq->tail_idx + 1) & (cq->num_descs - 1);
+ comp = cq->info[cq->tail_idx].comp;
+
+ aq_work++;
+ }
+ spin_unlock_irqrestore(&pdsc->adminq_lock, irqflags);
+
+ qcq->accum_work += aq_work;
+
+credits:
+ /* Return the interrupt credits, one for each completion */
+ credits = nq_work + aq_work;
+ if (credits)
+ pds_core_intr_credits(&pdsc->intr_ctrl[qcq->intx],
+ credits,
+ PDS_CORE_INTR_CRED_REARM);
+}
+
+void pdsc_work_thread(struct work_struct *work)
+{
+ struct pdsc_qcq *qcq = container_of(work, struct pdsc_qcq, work);
+
+ pdsc_process_adminq(qcq);
+}
+
+irqreturn_t pdsc_adminq_isr(int irq, void *data)
+{
+ struct pdsc_qcq *qcq = data;
+ struct pdsc *pdsc = qcq->pdsc;
+
+ /* Don't process AdminQ when shutting down */
+ if (pdsc->state & BIT_ULL(PDSC_S_STOPPING_DRIVER)) {
+ dev_err(pdsc->dev, "%s: called while PDSC_S_STOPPING_DRIVER\n",
+ __func__);
+ return IRQ_HANDLED;
+ }
+
+ queue_work(pdsc->wq, &qcq->work);
+ pds_core_intr_mask(&pdsc->intr_ctrl[irq], PDS_CORE_INTR_MASK_CLEAR);
+
+ return IRQ_HANDLED;
+}
+
+static int __pdsc_adminq_post(struct pdsc *pdsc,
+ struct pdsc_qcq *qcq,
+ union pds_core_adminq_cmd *cmd,
+ union pds_core_adminq_comp *comp,
+ struct pdsc_wait_context *wc)
+{
+ struct pdsc_queue *q = &qcq->q;
+ struct pdsc_q_info *q_info;
+ unsigned long irqflags;
+ unsigned int avail;
+ int index;
+ int ret;
+
+ spin_lock_irqsave(&pdsc->adminq_lock, irqflags);
+
+ /* Check for space in the queue */
+ avail = q->tail_idx;
+ if (q->head_idx >= avail)
+ avail += q->num_descs - q->head_idx - 1;
+ else
+ avail -= q->head_idx + 1;
+ if (!avail) {
+ ret = -ENOSPC;
+ goto err_out_unlock;
+ }
+
+ /* Check that the FW is running */
+ if (!pdsc_is_fw_running(pdsc)) {
+ u8 fw_status = ioread8(&pdsc->info_regs->fw_status);
+
+ dev_info(pdsc->dev, "%s: post failed - fw not running %#02x:\n",
+ __func__, fw_status);
+ ret = -ENXIO;
+
+ goto err_out_unlock;
+ }
+
+ /* Post the request */
+ index = q->head_idx;
+ q_info = &q->info[index];
+ q_info->wc = wc;
+ q_info->dest = comp;
+ memcpy(q_info->desc, cmd, sizeof(*cmd));
+
+ dev_dbg(pdsc->dev, "head_idx %d tail_idx %d\n",
+ q->head_idx, q->tail_idx);
+ dev_dbg(pdsc->dev, "post admin queue command:\n");
+ dynamic_hex_dump("cmd ", DUMP_PREFIX_OFFSET, 16, 1,
+ cmd, sizeof(*cmd), true);
+
+ q->head_idx = (q->head_idx + 1) & (q->num_descs - 1);
+
+ pds_core_dbell_ring(pdsc->kern_dbpage,
+ q->hw_type, q->dbval | q->head_idx);
+ ret = index;
+
+err_out_unlock:
+ spin_unlock_irqrestore(&pdsc->adminq_lock, irqflags);
+ return ret;
+}
+
+int pdsc_adminq_post(struct pdsc *pdsc,
+ union pds_core_adminq_cmd *cmd,
+ union pds_core_adminq_comp *comp,
+ bool fast_poll)
+{
+ struct pdsc_wait_context wc = {
+ .wait_completion =
+ COMPLETION_INITIALIZER_ONSTACK(wc.wait_completion),
+ };
+ unsigned long poll_interval = 1;
+ unsigned long poll_jiffies;
+ unsigned long time_limit;
+ unsigned long time_start;
+ unsigned long time_done;
+ unsigned long remaining;
+ int err = 0;
+ int index;
+
+ wc.qcq = &pdsc->adminqcq;
+ index = __pdsc_adminq_post(pdsc, &pdsc->adminqcq, cmd, comp, &wc);
+ if (index < 0) {
+ err = index;
+ goto err_out;
+ }
+
+ time_start = jiffies;
+ time_limit = time_start + HZ * pdsc->devcmd_timeout;
+ do {
+ /* Timeslice the actual wait to catch IO errors etc early */
+ poll_jiffies = msecs_to_jiffies(poll_interval);
+ remaining = wait_for_completion_timeout(&wc.wait_completion,
+ poll_jiffies);
+ if (remaining)
+ break;
+
+ if (!pdsc_is_fw_running(pdsc)) {
+ u8 fw_status = ioread8(&pdsc->info_regs->fw_status);
+
+ dev_dbg(pdsc->dev, "%s: post wait failed - fw not running %#02x:\n",
+ __func__, fw_status);
+ err = -ENXIO;
+ break;
+ }
+
+ /* When fast_poll is not requested, prevent aggressive polling
+ * on failures due to timeouts by doing exponential back off.
+ */
+ if (!fast_poll && poll_interval < PDSC_ADMINQ_MAX_POLL_INTERVAL)
+ poll_interval <<= 1;
+ } while (time_before(jiffies, time_limit));
+ time_done = jiffies;
+ dev_dbg(pdsc->dev, "%s: elapsed %d msecs\n",
+ __func__, jiffies_to_msecs(time_done - time_start));
+
+ /* Check the results */
+ if (time_after_eq(time_done, time_limit))
+ err = -ETIMEDOUT;
+
+ dev_dbg(pdsc->dev, "read admin queue completion idx %d:\n", index);
+ dynamic_hex_dump("comp ", DUMP_PREFIX_OFFSET, 16, 1,
+ comp, sizeof(*comp), true);
+
+ if (remaining && comp->status)
+ err = pdsc_err_to_errno(comp->status);
+
+err_out:
+ if (err) {
+ dev_dbg(pdsc->dev, "%s: opcode %d status %d err %pe\n",
+ __func__, cmd->opcode, comp->status, ERR_PTR(err));
+ if (err == -ENXIO || err == -ETIMEDOUT)
+ queue_work(pdsc->wq, &pdsc->health_work);
+ }
+
+ return err;
+}
+EXPORT_SYMBOL_GPL(pdsc_adminq_post);
@@ -6,17 +6,6 @@
#include "core.h"
-void pdsc_work_thread(struct work_struct *work)
-{
- /* stub */
-}
-
-irqreturn_t pdsc_adminq_isr(int irq, void *data)
-{
- /* stub */
- return IRQ_HANDLED;
-}
-
void pdsc_intr_free(struct pdsc *pdsc, int index)
{
struct pdsc_intr_info *intr_info;
@@ -4,6 +4,8 @@
#ifndef _PDS_CORE_ADMINQ_H_
#define _PDS_CORE_ADMINQ_H_
+#define PDSC_ADMINQ_MAX_POLL_INTERVAL 256
+
enum pds_core_adminq_flags {
PDS_AQ_FLAG_FASTPOLL = BIT(1), /* completion poll at 1ms */
};
@@ -631,8 +633,15 @@ static_assert(sizeof(union pds_core_notifyq_comp) == 64);
* where the meaning alternates between '1' and '0' for alternating
* passes through the completion descriptor ring.
*/
-static inline u8 pdsc_color_match(u8 color, u8 done_color)
+static inline bool pdsc_color_match(u8 color, bool done_color)
{
return (!!(color & PDS_COMP_COLOR_MASK)) == done_color;
}
+
+struct pdsc;
+int pdsc_adminq_post(struct pdsc *pdsc,
+ union pds_core_adminq_cmd *cmd,
+ union pds_core_adminq_comp *comp,
+ bool fast_poll);
+
#endif /* _PDS_CORE_ADMINQ_H_ */