* [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD
@ 2022-05-27 3:40 Chengwen Feng
2022-05-27 3:40 ` [PATCH 1/4] dma/hisilicon: fix return last-idx when no DMA completed Chengwen Feng
` (4 more replies)
0 siblings, 5 replies; 7+ messages in thread
From: Chengwen Feng @ 2022-05-27 3:40 UTC (permalink / raw)
To: dev; +Cc: thomas, kevin.laatz, bruce.richardson
This patch contains four patches for Hisilicon DMA PMD.
Chengwen Feng (4):
dma/hisilicon: fix return last-idx when no DMA completed
app/test: support test last-idx when no DMA completed
dma/hisilicon: enhance robustness of scan CQ
dma/hisilicon: support vchan-status ops
app/test/test_dmadev.c | 15 +++++++++
drivers/dma/hisilicon/hisi_dmadev.c | 50 ++++++++++++++++++++++++++---
drivers/dma/hisilicon/hisi_dmadev.h | 7 +++-
3 files changed, 66 insertions(+), 6 deletions(-)
--
2.33.0
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 1/4] dma/hisilicon: fix return last-idx when no DMA completed
2022-05-27 3:40 [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Chengwen Feng
@ 2022-05-27 3:40 ` Chengwen Feng
2022-05-27 3:40 ` [PATCH 2/4] app/test: support test " Chengwen Feng
` (3 subsequent siblings)
4 siblings, 0 replies; 7+ messages in thread
From: Chengwen Feng @ 2022-05-27 3:40 UTC (permalink / raw)
To: dev; +Cc: thomas, kevin.laatz, bruce.richardson
If no DMA request is completed, the ring_idx of the last completed
operation need returned by last_idx parameter. This patch fixes it.
Fixes: 2db4f0b82360 ("dma/hisilicon: add data path")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
drivers/dma/hisilicon/hisi_dmadev.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/dma/hisilicon/hisi_dmadev.c b/drivers/dma/hisilicon/hisi_dmadev.c
index 9cef2cbfbe..f5c3cd914d 100644
--- a/drivers/dma/hisilicon/hisi_dmadev.c
+++ b/drivers/dma/hisilicon/hisi_dmadev.c
@@ -702,12 +702,12 @@ hisi_dma_completed(void *dev_private,
}
sq_head = (sq_head + 1) & hw->sq_depth_mask;
}
+ *last_idx = hw->cridx + i - 1;
if (i > 0) {
hw->cridx += i;
- *last_idx = hw->cridx - 1;
hw->sq_head = sq_head;
+ hw->completed += i;
}
- hw->completed += i;
return i;
}
@@ -761,12 +761,12 @@ hisi_dma_completed_status(void *dev_private,
hw->status[sq_head] = HISI_DMA_STATUS_SUCCESS;
sq_head = (sq_head + 1) & hw->sq_depth_mask;
}
+ *last_idx = hw->cridx + cpl_num - 1;
if (likely(cpl_num > 0)) {
hw->cridx += cpl_num;
- *last_idx = hw->cridx - 1;
hw->sq_head = sq_head;
+ hw->completed += cpl_num;
}
- hw->completed += cpl_num;
return cpl_num;
}
--
2.33.0
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 2/4] app/test: support test last-idx when no DMA completed
2022-05-27 3:40 [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Chengwen Feng
2022-05-27 3:40 ` [PATCH 1/4] dma/hisilicon: fix return last-idx when no DMA completed Chengwen Feng
@ 2022-05-27 3:40 ` Chengwen Feng
2022-05-30 11:40 ` Kevin Laatz
2022-05-27 3:40 ` [PATCH 3/4] dma/hisilicon: enhance robustness of scan CQ Chengwen Feng
` (2 subsequent siblings)
4 siblings, 1 reply; 7+ messages in thread
From: Chengwen Feng @ 2022-05-27 3:40 UTC (permalink / raw)
To: dev; +Cc: thomas, kevin.laatz, bruce.richardson
If no DMA request is completed, the ring_idx of the last completed
operation need returned by last_idx parameter. This patch adds
testcase for it.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
app/test/test_dmadev.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/app/test/test_dmadev.c b/app/test/test_dmadev.c
index 2b097e0f47..a7651a486f 100644
--- a/app/test/test_dmadev.c
+++ b/app/test/test_dmadev.c
@@ -177,6 +177,7 @@ do_multi_copies(int16_t dev_id, uint16_t vchan,
static int
test_enqueue_copies(int16_t dev_id, uint16_t vchan)
{
+ enum rte_dma_status_code status;
unsigned int i;
uint16_t id;
@@ -215,6 +216,20 @@ test_enqueue_copies(int16_t dev_id, uint16_t vchan)
ERR_RETURN("Error:incorrect job id received, %u [expected %u]\n",
id, id_count);
+ /* check for completed and id when no job done */
+ if (rte_dma_completed(dev_id, vchan, 1, &id, NULL) != 0)
+ ERR_RETURN("Error with rte_dma_completed when no job done\n");
+ if (id != id_count)
+ ERR_RETURN("Error:incorrect job id received when no job done, %u [expected %u]\n",
+ id, id_count);
+
+ /* check for completed_status and id when no job done */
+ if (rte_dma_completed_status(dev_id, vchan, 1, &id, &status) != 0)
+ ERR_RETURN("Error with rte_dma_completed_status when no job done\n");
+ if (id != id_count)
+ ERR_RETURN("Error:incorrect job id received when no job done, %u [expected %u]\n",
+ id, id_count);
+
rte_pktmbuf_free(src);
rte_pktmbuf_free(dst);
--
2.33.0
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 3/4] dma/hisilicon: enhance robustness of scan CQ
2022-05-27 3:40 [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Chengwen Feng
2022-05-27 3:40 ` [PATCH 1/4] dma/hisilicon: fix return last-idx when no DMA completed Chengwen Feng
2022-05-27 3:40 ` [PATCH 2/4] app/test: support test " Chengwen Feng
@ 2022-05-27 3:40 ` Chengwen Feng
2022-05-27 3:40 ` [PATCH 4/4] dma/hisilicon: support vchan-status ops Chengwen Feng
2022-06-07 10:44 ` [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Thomas Monjalon
4 siblings, 0 replies; 7+ messages in thread
From: Chengwen Feng @ 2022-05-27 3:40 UTC (permalink / raw)
To: dev; +Cc: thomas, kevin.laatz, bruce.richardson
The CQ (completion queue) descriptors were updated by hardware, and then
scanned by driver to retrieve hardware completion status.
This patch enhances robustness by following:
1. replace while (true) with a finite loop to avoid potential dead loop.
2. check the csq_head field in CQ descriptor to avoid status array
overflows.
Fixes: 2db4f0b82360 ("dma/hisilicon: add data path")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
drivers/dma/hisilicon/hisi_dmadev.c | 12 +++++++++++-
1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/drivers/dma/hisilicon/hisi_dmadev.c b/drivers/dma/hisilicon/hisi_dmadev.c
index f5c3cd914d..fbe09284ed 100644
--- a/drivers/dma/hisilicon/hisi_dmadev.c
+++ b/drivers/dma/hisilicon/hisi_dmadev.c
@@ -634,7 +634,7 @@ hisi_dma_scan_cq(struct hisi_dma_dev *hw)
uint16_t count = 0;
uint64_t misc;
- while (true) {
+ while (count < hw->cq_depth) {
cqe = &hw->cqe[cq_head];
misc = cqe->misc;
misc = rte_le_to_cpu_64(misc);
@@ -642,6 +642,16 @@ hisi_dma_scan_cq(struct hisi_dma_dev *hw)
break;
csq_head = FIELD_GET(CQE_SQ_HEAD_MASK, misc);
+ if (unlikely(csq_head > hw->sq_depth_mask)) {
+ /**
+ * Defensive programming to prevent overflow of the
+ * status array indexed by csq_head. Only error logs
+ * are used for prompting.
+ */
+ HISI_DMA_ERR(hw, "invalid csq_head:%u!\n", csq_head);
+ count = 0;
+ break;
+ }
if (unlikely(misc & CQE_STATUS_MASK))
hw->status[csq_head] = FIELD_GET(CQE_STATUS_MASK,
misc);
--
2.33.0
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 4/4] dma/hisilicon: support vchan-status ops
2022-05-27 3:40 [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Chengwen Feng
` (2 preceding siblings ...)
2022-05-27 3:40 ` [PATCH 3/4] dma/hisilicon: enhance robustness of scan CQ Chengwen Feng
@ 2022-05-27 3:40 ` Chengwen Feng
2022-06-07 10:44 ` [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Thomas Monjalon
4 siblings, 0 replies; 7+ messages in thread
From: Chengwen Feng @ 2022-05-27 3:40 UTC (permalink / raw)
To: dev; +Cc: thomas, kevin.laatz, bruce.richardson
This patch adds support for vchan-status ops.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
drivers/dma/hisilicon/hisi_dmadev.c | 30 +++++++++++++++++++++++++++++
drivers/dma/hisilicon/hisi_dmadev.h | 7 ++++++-
2 files changed, 36 insertions(+), 1 deletion(-)
diff --git a/drivers/dma/hisilicon/hisi_dmadev.c b/drivers/dma/hisilicon/hisi_dmadev.c
index fbe09284ed..9494b60779 100644
--- a/drivers/dma/hisilicon/hisi_dmadev.c
+++ b/drivers/dma/hisilicon/hisi_dmadev.c
@@ -461,6 +461,27 @@ hisi_dma_stats_reset(struct rte_dma_dev *dev, uint16_t vchan)
return 0;
}
+static int
+hisi_dma_vchan_status(const struct rte_dma_dev *dev, uint16_t vchan,
+ enum rte_dma_vchan_status *status)
+{
+ struct hisi_dma_dev *hw = dev->data->dev_private;
+ uint32_t val;
+
+ RTE_SET_USED(vchan);
+
+ val = hisi_dma_read_queue(hw, HISI_DMA_QUEUE_FSM_REG);
+ val = FIELD_GET(HISI_DMA_QUEUE_FSM_STS_M, val);
+ if (val == HISI_DMA_STATE_RUN)
+ *status = RTE_DMA_VCHAN_ACTIVE;
+ else if (val == HISI_DMA_STATE_CPL)
+ *status = RTE_DMA_VCHAN_IDLE;
+ else
+ *status = RTE_DMA_VCHAN_HALTED_ERROR;
+
+ return 0;
+}
+
static void
hisi_dma_dump_range(struct hisi_dma_dev *hw, FILE *f, uint32_t start,
uint32_t end)
@@ -816,6 +837,14 @@ hisi_dma_gen_dev_name(const struct rte_pci_device *pci_dev,
* dev_stop| |
* | v
* ------------------
+ * | CPL |
+ * ------------------
+ * ^ |
+ * hardware | |
+ * completed all| |dev_submit
+ * descriptors | |
+ * | |
+ * ------------------
* | RUN |
* ------------------
*
@@ -829,6 +858,7 @@ static const struct rte_dma_dev_ops hisi_dmadev_ops = {
.vchan_setup = hisi_dma_vchan_setup,
.stats_get = hisi_dma_stats_get,
.stats_reset = hisi_dma_stats_reset,
+ .vchan_status = hisi_dma_vchan_status,
.dev_dump = hisi_dma_dump,
};
diff --git a/drivers/dma/hisilicon/hisi_dmadev.h b/drivers/dma/hisilicon/hisi_dmadev.h
index 90b85322ca..deb1357eea 100644
--- a/drivers/dma/hisilicon/hisi_dmadev.h
+++ b/drivers/dma/hisilicon/hisi_dmadev.h
@@ -132,11 +132,16 @@ enum {
/**
* In fact, there are multiple states, but it need to pay attention to
- * the following two states for the driver:
+ * the following three states for the driver:
*/
enum {
HISI_DMA_STATE_IDLE = 0,
HISI_DMA_STATE_RUN,
+ /**
+ * All of the submitted descriptor are finished, and the queue
+ * is waiting for new descriptors.
+ */
+ HISI_DMA_STATE_CPL,
};
/**
--
2.33.0
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/4] app/test: support test last-idx when no DMA completed
2022-05-27 3:40 ` [PATCH 2/4] app/test: support test " Chengwen Feng
@ 2022-05-30 11:40 ` Kevin Laatz
0 siblings, 0 replies; 7+ messages in thread
From: Kevin Laatz @ 2022-05-30 11:40 UTC (permalink / raw)
To: Chengwen Feng, dev; +Cc: thomas, bruce.richardson
Tested with IOAT and IDXD, both pass unit tests.
Tested-by: Kevin Laatz <kevin.laatz@intel.com>
On 27/05/2022 04:40, Chengwen Feng wrote:
> If no DMA request is completed, the ring_idx of the last completed
> operation need returned by last_idx parameter. This patch adds
> testcase for it.
>
> Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
> ---
> app/test/test_dmadev.c | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
> diff --git a/app/test/test_dmadev.c b/app/test/test_dmadev.c
> index 2b097e0f47..a7651a486f 100644
> --- a/app/test/test_dmadev.c
> +++ b/app/test/test_dmadev.c
> @@ -177,6 +177,7 @@ do_multi_copies(int16_t dev_id, uint16_t vchan,
> static int
> test_enqueue_copies(int16_t dev_id, uint16_t vchan)
> {
> + enum rte_dma_status_code status;
> unsigned int i;
> uint16_t id;
>
> @@ -215,6 +216,20 @@ test_enqueue_copies(int16_t dev_id, uint16_t vchan)
> ERR_RETURN("Error:incorrect job id received, %u [expected %u]\n",
> id, id_count);
>
> + /* check for completed and id when no job done */
> + if (rte_dma_completed(dev_id, vchan, 1, &id, NULL) != 0)
> + ERR_RETURN("Error with rte_dma_completed when no job done\n");
> + if (id != id_count)
> + ERR_RETURN("Error:incorrect job id received when no job done, %u [expected %u]\n",
> + id, id_count);
> +
> + /* check for completed_status and id when no job done */
> + if (rte_dma_completed_status(dev_id, vchan, 1, &id, &status) != 0)
> + ERR_RETURN("Error with rte_dma_completed_status when no job done\n");
> + if (id != id_count)
> + ERR_RETURN("Error:incorrect job id received when no job done, %u [expected %u]\n",
> + id, id_count);
> +
> rte_pktmbuf_free(src);
> rte_pktmbuf_free(dst);
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD
2022-05-27 3:40 [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Chengwen Feng
` (3 preceding siblings ...)
2022-05-27 3:40 ` [PATCH 4/4] dma/hisilicon: support vchan-status ops Chengwen Feng
@ 2022-06-07 10:44 ` Thomas Monjalon
4 siblings, 0 replies; 7+ messages in thread
From: Thomas Monjalon @ 2022-06-07 10:44 UTC (permalink / raw)
To: Chengwen Feng; +Cc: dev, kevin.laatz, bruce.richardson
27/05/2022 05:40, Chengwen Feng:
> This patch contains four patches for Hisilicon DMA PMD.
>
> Chengwen Feng (4):
> dma/hisilicon: fix return last-idx when no DMA completed
> app/test: support test last-idx when no DMA completed
> dma/hisilicon: enhance robustness of scan CQ
> dma/hisilicon: support vchan-status ops
Series applied, thanks.
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2022-06-07 10:44 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-05-27 3:40 [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Chengwen Feng
2022-05-27 3:40 ` [PATCH 1/4] dma/hisilicon: fix return last-idx when no DMA completed Chengwen Feng
2022-05-27 3:40 ` [PATCH 2/4] app/test: support test " Chengwen Feng
2022-05-30 11:40 ` Kevin Laatz
2022-05-27 3:40 ` [PATCH 3/4] dma/hisilicon: enhance robustness of scan CQ Chengwen Feng
2022-05-27 3:40 ` [PATCH 4/4] dma/hisilicon: support vchan-status ops Chengwen Feng
2022-06-07 10:44 ` [PATCH 0/4] bugfix and feature for Hisilicon DMA PMD Thomas Monjalon
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).