* [PATCH v4 01/30] baseband/acc100: fix ring availability calculation
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 02/30] baseband/acc100: add function to check AQ availability Hernan Vargas
` (28 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Refactor of the queue availability computation to prevent the
application to dequeue more than what may have been enqueued.
Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 25 +++++++++++++------------
1 file changed, 13 insertions(+), 12 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index e5384223d1..3b0c8e41dc 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -2861,7 +2861,7 @@ acc100_enqueue_enc_cb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i;
union acc_dma_desc *desc;
int ret;
@@ -2899,7 +2899,7 @@ acc100_enqueue_ldpc_enc_cb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i = 0;
union acc_dma_desc *desc;
int ret, desc_idx = 0;
@@ -2949,7 +2949,7 @@ acc100_enqueue_enc_tb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i, enqueued_cbs = 0;
uint8_t cbs_in_tb;
int ret;
@@ -3011,7 +3011,7 @@ acc100_enqueue_dec_cb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_dec_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i;
union acc_dma_desc *desc;
int ret;
@@ -3050,7 +3050,7 @@ acc100_enqueue_ldpc_dec_tb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_dec_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i, enqueued_cbs = 0;
uint8_t cbs_in_tb;
int ret;
@@ -3083,7 +3083,7 @@ acc100_enqueue_ldpc_dec_cb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_dec_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i;
union acc_dma_desc *desc;
int ret;
@@ -3132,7 +3132,7 @@ acc100_enqueue_dec_tb(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_dec_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- int32_t avail = q->sw_ring_depth + q->sw_ring_tail - q->sw_ring_head;
+ int32_t avail = acc_ring_avail_enq(q);
uint16_t i, enqueued_cbs = 0;
uint8_t cbs_in_tb;
int ret;
@@ -3495,12 +3495,13 @@ acc100_dequeue_enc(struct rte_bbdev_queue_data *q_data,
{
struct acc_queue *q = q_data->queue_private;
uint16_t dequeue_num;
- uint32_t avail = q->sw_ring_head - q->sw_ring_tail;
+ uint32_t avail = acc_ring_avail_deq(q);
uint32_t aq_dequeued = 0;
uint16_t i, dequeued_cbs = 0;
struct rte_bbdev_enc_op *op;
int ret;
-
+ if (avail == 0)
+ return 0;
#ifdef RTE_LIBRTE_BBDEV_DEBUG
if (unlikely(ops == NULL || q == NULL)) {
rte_bbdev_log_debug("Unexpected undefined pointer");
@@ -3539,7 +3540,7 @@ acc100_dequeue_ldpc_enc(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- uint32_t avail = q->sw_ring_head - q->sw_ring_tail;
+ uint32_t avail = acc_ring_avail_deq(q);
uint32_t aq_dequeued = 0;
uint16_t dequeue_num, i, dequeued_cbs = 0, dequeued_descs = 0;
int ret;
@@ -3579,7 +3580,7 @@ acc100_dequeue_dec(struct rte_bbdev_queue_data *q_data,
{
struct acc_queue *q = q_data->queue_private;
uint16_t dequeue_num;
- uint32_t avail = q->sw_ring_head - q->sw_ring_tail;
+ uint32_t avail = acc_ring_avail_deq(q);
uint32_t aq_dequeued = 0;
uint16_t i;
uint16_t dequeued_cbs = 0;
@@ -3623,7 +3624,7 @@ acc100_dequeue_ldpc_dec(struct rte_bbdev_queue_data *q_data,
{
struct acc_queue *q = q_data->queue_private;
uint16_t dequeue_num;
- uint32_t avail = q->sw_ring_head - q->sw_ring_tail;
+ uint32_t avail = acc_ring_avail_deq(q);
uint32_t aq_dequeued = 0;
uint16_t i;
uint16_t dequeued_cbs = 0;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 02/30] baseband/acc100: add function to check AQ availability
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 01/30] baseband/acc100: fix ring availability calculation Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 03/30] baseband/acc100: memory leak fix Hernan Vargas
` (27 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
It is possible for some corner case to run more batch enqueue than
supported. A protection is required to avoid that corner case.
Enhance all ACC100 enqueue operations with check to see if there is room
in the atomic queue for enqueueing batches into the queue manager
Check room in AQ for the enqueues batches into Qmgr
Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 18 ++++++++++--------
1 file changed, 10 insertions(+), 8 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 3b0c8e41dc..7fec2283eb 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -2983,7 +2983,8 @@ static uint16_t
acc100_enqueue_enc(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
- if (unlikely(num == 0))
+ int32_t aq_avail = acc_aq_avail(q_data, num);
+ if (unlikely((aq_avail <= 0) || (num == 0)))
return 0;
if (ops[0]->turbo_enc.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
return acc100_enqueue_enc_tb(q_data, ops, num);
@@ -2996,7 +2997,8 @@ static uint16_t
acc100_enqueue_ldpc_enc(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
- if (unlikely(num == 0))
+ int32_t aq_avail = acc_aq_avail(q_data, num);
+ if (unlikely((aq_avail <= 0) || (num == 0)))
return 0;
if (ops[0]->ldpc_enc.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
return acc100_enqueue_enc_tb(q_data, ops, num);
@@ -3164,8 +3166,11 @@ static uint16_t
acc100_enqueue_dec(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_dec_op **ops, uint16_t num)
{
- if (unlikely(num == 0))
+ int32_t aq_avail = acc_aq_avail(q_data, num);
+
+ if (unlikely((aq_avail <= 0) || (num == 0)))
return 0;
+
if (ops[0]->turbo_dec.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
return acc100_enqueue_dec_tb(q_data, ops, num);
else
@@ -3177,11 +3182,9 @@ static uint16_t
acc100_enqueue_ldpc_dec(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_dec_op **ops, uint16_t num)
{
- struct acc_queue *q = q_data->queue_private;
- int32_t aq_avail = q->aq_depth +
- (q->aq_dequeued - q->aq_enqueued) / 128;
+ int32_t aq_avail = acc_aq_avail(q_data, num);
- if (unlikely((aq_avail == 0) || (num == 0)))
+ if (unlikely((aq_avail <= 0) || (num == 0)))
return 0;
if (ops[0]->ldpc_dec.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
@@ -3190,7 +3193,6 @@ acc100_enqueue_ldpc_dec(struct rte_bbdev_queue_data *q_data,
return acc100_enqueue_ldpc_dec_cb(q_data, ops, num);
}
-
/* Dequeue one encode operations from ACC100 device in CB mode */
static inline int
dequeue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 03/30] baseband/acc100: memory leak fix
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 01/30] baseband/acc100: fix ring availability calculation Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 02/30] baseband/acc100: add function to check AQ availability Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 04/30] baseband/acc100: add LDPC encoder padding function Hernan Vargas
` (26 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Move check for undefined device before allocating queue data structure.
Coverity issue: 375803, 375813, 375819, 375827, 375831
Fixes: 060e7672930 ("baseband/acc100: add queue configuration")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 7fec2283eb..7500ef6eb5 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -663,6 +663,10 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
struct acc_queue *q;
int16_t q_idx;
+ if (d == NULL) {
+ rte_bbdev_log(ERR, "Undefined device");
+ return -ENODEV;
+ }
/* Allocate the queue data structure. */
q = rte_zmalloc_socket(dev->device->driver->name, sizeof(*q),
RTE_CACHE_LINE_SIZE, conf->socket);
@@ -670,10 +674,6 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
rte_bbdev_log(ERR, "Failed to allocate queue memory");
return -ENOMEM;
}
- if (d == NULL) {
- rte_bbdev_log(ERR, "Undefined device");
- return -ENODEV;
- }
q->d = d;
q->ring_addr = RTE_PTR_ADD(d->sw_rings, (d->sw_ring_size * queue_id));
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 04/30] baseband/acc100: add LDPC encoder padding function
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (2 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 03/30] baseband/acc100: memory leak fix Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-20 9:58 ` [EXT] " Akhil Goyal
2022-10-19 0:38 ` [PATCH v4 05/30] baseband/acc100: check turbo dec/enc input Hernan Vargas
` (25 subsequent siblings)
29 siblings, 1 reply; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
LDPC Encoder input may need to be padded to avoid small beat for ACC100.
Padding 5GDL input buffer length (BLEN) to avoid case (BLEN % 64) <= 8.
Adding protection for corner case to avoid for 5GDL occurrence of last
beat within the ACC100 fabric with <= 8B which might trigger a fabric
corner case hang issue.
Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 27 ++++++++++++++++++++-------
1 file changed, 20 insertions(+), 7 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 7500ef6eb5..a80ecdca84 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1017,7 +1017,6 @@ acc100_fcw_td_fill(const struct rte_bbdev_dec_op *op, struct acc_fcw_td *fcw)
RTE_BBDEV_TURBO_HALF_ITERATION_EVEN);
}
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
static inline bool
is_acc100(struct acc_queue *q)
@@ -1030,7 +1029,6 @@ validate_op_required(struct acc_queue *q)
{
return is_acc100(q);
}
-#endif
/* Fill in a frame control word for LDPC decoding. */
static inline void
@@ -1355,12 +1353,28 @@ acc100_dma_fill_blk_type_in(struct acc_dma_req_desc *desc,
return next_triplet;
}
+/* May need to pad LDPC Encoder input to avoid small beat for ACC100. */
+static inline uint16_t
+pad_le_in(uint16_t blen, struct acc_queue *q)
+{
+ uint16_t last_beat;
+
+ if (!is_acc100(q))
+ return blen;
+
+ last_beat = blen % 64;
+ if ((last_beat > 0) && (last_beat <= 8))
+ blen += 8;
+
+ return blen;
+}
+
static inline int
acc100_dma_desc_le_fill(struct rte_bbdev_enc_op *op,
struct acc_dma_req_desc *desc, struct rte_mbuf **input,
struct rte_mbuf *output, uint32_t *in_offset,
uint32_t *out_offset, uint32_t *out_length,
- uint32_t *mbuf_total_left, uint32_t *seg_total_left)
+ uint32_t *mbuf_total_left, uint32_t *seg_total_left, struct acc_queue *q)
{
int next_triplet = 1; /* FCW already done */
uint16_t K, in_length_in_bits, in_length_in_bytes;
@@ -1384,8 +1398,7 @@ acc100_dma_desc_le_fill(struct rte_bbdev_enc_op *op,
}
next_triplet = acc100_dma_fill_blk_type_in(desc, input, in_offset,
- in_length_in_bytes,
- seg_total_left, next_triplet);
+ pad_le_in(in_length_in_bytes, q), seg_total_left, next_triplet);
if (unlikely(next_triplet < 0)) {
rte_bbdev_log(ERR,
"Mismatch between data to process and mbuf data length in bbdev_op: %p",
@@ -2035,7 +2048,7 @@ enqueue_ldpc_enc_n_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ops,
acc_header_init(&desc->req);
desc->req.numCBs = num;
- in_length_in_bytes = ops[0]->ldpc_enc.input.data->data_len;
+ in_length_in_bytes = pad_le_in(ops[0]->ldpc_enc.input.data->data_len, q);
out_length = (enc->cb_params.e + 7) >> 3;
desc->req.m2dlen = 1 + num;
desc->req.d2mlen = num;
@@ -2102,7 +2115,7 @@ enqueue_ldpc_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
ret = acc100_dma_desc_le_fill(op, &desc->req, &input, output,
&in_offset, &out_offset, &out_length, &mbuf_total_left,
- &seg_total_left);
+ &seg_total_left, q);
if (unlikely(ret < 0))
return ret;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* RE: [EXT] [PATCH v4 04/30] baseband/acc100: add LDPC encoder padding function
2022-10-19 0:38 ` [PATCH v4 04/30] baseband/acc100: add LDPC encoder padding function Hernan Vargas
@ 2022-10-20 9:58 ` Akhil Goyal
0 siblings, 0 replies; 33+ messages in thread
From: Akhil Goyal @ 2022-10-20 9:58 UTC (permalink / raw)
To: Hernan Vargas, dev, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, stable
> LDPC Encoder input may need to be padded to avoid small beat for ACC100.
> Padding 5GDL input buffer length (BLEN) to avoid case (BLEN % 64) <= 8.
> Adding protection for corner case to avoid for 5GDL occurrence of last
> beat within the ACC100 fabric with <= 8B which might trigger a fabric
> corner case hang issue.
>
> Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
> Cc: stable@dpdk.org
>
> Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Please fix compilation. It has been repeatedly asked to fix individual patch
Compilation for ACC patches. Please take care.
ninja: Entering directory `./build-clang-static'
[1/26] Compiling C object 'drivers/a715181@@tmp_rte_baseband_acc@sta/baseband_acc_rte_acc100_pmd.c.o'.
FAILED: drivers/a715181@@tmp_rte_baseband_acc@sta/baseband_acc_rte_acc100_pmd.c.o
ccache clang -Idrivers/a715181@@tmp_rte_baseband_acc@sta -Idrivers -I../drivers -Idrivers/baseband/acc -I../drivers/baseband/acc -Ilib/bbdev -I../lib/bbdev -I. -I../ -Iconfig -I../config -Ilib/eal/include -I../lib/eal/include -Ilib/eal/linux/include -I../lib/eal/linux/include -Ilib/eal/x86/include -I../lib/eal/x86/include -Ilib/eal/common -I../lib/eal/common -Ilib/eal -I../lib/eal -Ilib/kvargs -I../lib/kvargs -Ilib/telemetry/../metrics -I../lib/telemetry/../metrics -Ilib/telemetry -I../lib/telemetry -Ilib/mbuf -I../lib/mbuf -Ilib/mempool -I../lib/mempool -Ilib/ring -I../lib/ring -Idrivers/bus/pci -I../drivers/bus/pci -I../drivers/bus/pci/linux -Ilib/pci -I../lib/pci -Xclang -fcolor-diagnostics -pipe -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Werror -O2 -g -include rte_config.h -Wcast-qual -Wdeprecated -Wformat -Wformat-nonliteral -Wformat-security -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wpointer-arith -Wsign-compare -Wstrict-prototypes -Wundef -Wwrite-strings -Wno-address-of-packed-member -Wno-missing-field-initializers -D_GNU_SOURCE -fPIC -march=native -DALLOW_EXPERIMENTAL_API -DALLOW_INTERNAL_API -DRTE_LOG_DEFAULT_LOGTYPE=pmd.bb.acc -MD -MQ 'drivers/a715181@@tmp_rte_baseband_acc@sta/baseband_acc_rte_acc100_pmd.c.o' -MF 'drivers/a715181@@tmp_rte_baseband_acc@sta/baseband_acc_rte_acc100_pmd.c.o.d' -o 'drivers/a715181@@tmp_rte_baseband_acc@sta/baseband_acc_rte_acc100_pmd.c.o' -c ../drivers/baseband/acc/rte_acc100_pmd.c
../drivers/baseband/acc/rte_acc100_pmd.c:1028:1: error: unused function 'validate_op_required' [-Werror,-Wunused-function]
validate_op_required(struct acc_queue *q)
^
1 error generated.
ninja: build stopped: subcommand failed.
^ permalink raw reply [flat|nested] 33+ messages in thread
* [PATCH v4 05/30] baseband/acc100: check turbo dec/enc input
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (3 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 04/30] baseband/acc100: add LDPC encoder padding function Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 06/30] baseband/acc100: check for unlikely operation vals Hernan Vargas
` (24 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Add NULL check for the turbo decoder and encoder input length.
Fixes: 3bfc5f60403 ("baseband/acc100: add debug function to validate input")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 23 ++++++++++++++++++-----
1 file changed, 18 insertions(+), 5 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index a80ecdca84..bf0d7d77e7 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1763,6 +1763,11 @@ validate_enc_op(struct rte_bbdev_enc_op *op, struct acc_queue *q)
return -1;
}
+ if (unlikely(turbo_enc->input.length == 0)) {
+ rte_bbdev_log(ERR, "input length null");
+ return -1;
+ }
+
if (turbo_enc->code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK) {
tb = &turbo_enc->tb_params;
if ((tb->k_neg < RTE_BBDEV_TURBO_MIN_CB_SIZE
@@ -1782,11 +1787,12 @@ validate_enc_op(struct rte_bbdev_enc_op *op, struct acc_queue *q)
RTE_BBDEV_TURBO_MAX_CB_SIZE);
return -1;
}
- if (tb->c_neg > (RTE_BBDEV_TURBO_MAX_CODE_BLOCKS - 1))
+ if (unlikely(tb->c_neg > 0)) {
rte_bbdev_log(ERR,
- "c_neg (%u) is out of range 0 <= value <= %u",
- tb->c_neg,
- RTE_BBDEV_TURBO_MAX_CODE_BLOCKS - 1);
+ "c_neg (%u) expected to be null",
+ tb->c_neg);
+ return -1;
+ }
if (tb->c < 1 || tb->c > RTE_BBDEV_TURBO_MAX_CODE_BLOCKS) {
rte_bbdev_log(ERR,
"c (%u) is out of range 1 <= value <= %u",
@@ -2280,6 +2286,11 @@ validate_dec_op(struct rte_bbdev_dec_op *op, struct acc_queue *q)
return -1;
}
+ if (unlikely(turbo_dec->input.length == 0)) {
+ rte_bbdev_log(ERR, "input length null");
+ return -1;
+ }
+
if (turbo_dec->code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK) {
tb = &turbo_dec->tb_params;
if ((tb->k_neg < RTE_BBDEV_TURBO_MIN_CB_SIZE
@@ -2300,11 +2311,13 @@ validate_dec_op(struct rte_bbdev_dec_op *op, struct acc_queue *q)
RTE_BBDEV_TURBO_MAX_CB_SIZE);
return -1;
}
- if (tb->c_neg > (RTE_BBDEV_TURBO_MAX_CODE_BLOCKS - 1))
+ if (unlikely(tb->c_neg > (RTE_BBDEV_TURBO_MAX_CODE_BLOCKS - 1))) {
rte_bbdev_log(ERR,
"c_neg (%u) is out of range 0 <= value <= %u",
tb->c_neg,
RTE_BBDEV_TURBO_MAX_CODE_BLOCKS - 1);
+ return -1;
+ }
if (tb->c < 1 || tb->c > RTE_BBDEV_TURBO_MAX_CODE_BLOCKS) {
rte_bbdev_log(ERR,
"c (%u) is out of range 1 <= value <= %u",
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 06/30] baseband/acc100: check for unlikely operation vals
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (4 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 05/30] baseband/acc100: check turbo dec/enc input Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 07/30] baseband/acc100: enforce additional check on FCW Hernan Vargas
` (23 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Add unlikely checks for NULL operation values.
Fixes: f404dfe35cc ("baseband/acc100: support 4G processing")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index bf0d7d77e7..50c1536dee 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -2179,6 +2179,10 @@ enqueue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
r = op->turbo_enc.tb_params.r;
while (mbuf_total_left > 0 && r < c) {
+ if (unlikely(input == NULL)) {
+ rte_bbdev_log(ERR, "Not enough input segment");
+ return -EINVAL;
+ }
seg_total_left = rte_pktmbuf_data_len(input) - in_offset;
/* Set up DMA descriptor */
desc = acc_desc(q, total_enqueued_cbs);
@@ -3096,6 +3100,8 @@ acc100_enqueue_ldpc_dec_tb(struct rte_bbdev_queue_data *q_data,
break;
enqueued_cbs += ret;
}
+ if (unlikely(enqueued_cbs == 0))
+ return 0; /* Nothing to enqueue */
acc_dma_enqueue(q, enqueued_cbs, &q_data->queue_stats);
@@ -3624,6 +3630,8 @@ acc100_dequeue_dec(struct rte_bbdev_queue_data *q_data,
for (i = 0; i < dequeue_num; ++i) {
op = acc_op_tail(q, dequeued_cbs);
+ if (unlikely(op == NULL))
+ break;
if (op->turbo_dec.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
ret = dequeue_dec_one_op_tb(q, &ops[i], dequeued_cbs,
&aq_dequeued);
@@ -3669,6 +3677,8 @@ acc100_dequeue_ldpc_dec(struct rte_bbdev_queue_data *q_data,
for (i = 0; i < dequeue_num; ++i) {
op = (q->ring_addr + ((q->sw_ring_tail + dequeued_cbs)
& q->sw_ring_wrap_mask))->req.op_addr;
+ if (unlikely(op == NULL))
+ break;
if (op->ldpc_dec.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
ret = dequeue_dec_one_op_tb(q, &ops[i], dequeued_cbs,
&aq_dequeued);
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 07/30] baseband/acc100: enforce additional check on FCW
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (5 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 06/30] baseband/acc100: check for unlikely operation vals Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 08/30] baseband/acc100: allocate ring/queue mem when NULL Hernan Vargas
` (22 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Enforce additional check on Frame Control Word validity and add stronger
alignment for decompression mode.
Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/acc100_pmd.h | 1 +
drivers/baseband/acc/acc_common.h | 1 +
drivers/baseband/acc/rte_acc100_pmd.c | 71 ++++++++++++++++++++++-----
3 files changed, 62 insertions(+), 11 deletions(-)
diff --git a/drivers/baseband/acc/acc100_pmd.h b/drivers/baseband/acc/acc100_pmd.h
index 9486e98521..eb6349c85a 100644
--- a/drivers/baseband/acc/acc100_pmd.h
+++ b/drivers/baseband/acc/acc100_pmd.h
@@ -87,6 +87,7 @@
#define ACC100_HARQ_DDR (512 * 1)
#define ACC100_PRQ_DDR_VER 0x10092020
#define ACC100_DDR_TRAINING_MAX (5000)
+#define ACC100_HARQ_ALIGN_COMP 256
struct acc100_registry_addr {
unsigned int dma_ring_dl5g_hi;
diff --git a/drivers/baseband/acc/acc_common.h b/drivers/baseband/acc/acc_common.h
index 6f141c95ce..97d10b8b40 100644
--- a/drivers/baseband/acc/acc_common.h
+++ b/drivers/baseband/acc/acc_common.h
@@ -120,6 +120,7 @@
#define ACC_ALGO_SPA 0
#define ACC_ALGO_MSA 1
+#define ACC_HARQ_ALIGN_64B 64
/* Helper macro for logging */
#define rte_acc_log(level, fmt, ...) \
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 50c1536dee..1c83d591e3 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1039,6 +1039,7 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
uint16_t harq_index;
uint32_t l;
bool harq_prun = false;
+ uint32_t max_hc_in;
fcw->qm = op->ldpc_dec.q_m;
fcw->nfiller = op->ldpc_dec.n_filler;
@@ -1088,13 +1089,22 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
harq_in_length = op->ldpc_dec.harq_combined_input.length;
if (fcw->hcin_decomp_mode > 0)
harq_in_length = harq_in_length * 8 / 6;
- harq_in_length = RTE_ALIGN(harq_in_length, 64);
- if ((harq_layout[harq_index].offset > 0) & harq_prun) {
+
+ harq_in_length = RTE_MIN(harq_in_length, op->ldpc_dec.n_cb
+ - op->ldpc_dec.n_filler);
+
+ /* Alignment on next 64B - Already enforced from HC output */
+ harq_in_length = RTE_ALIGN_FLOOR(harq_in_length, ACC_HARQ_ALIGN_64B);
+
+ /* Stronger alignment requirement when in decompression mode */
+ if (fcw->hcin_decomp_mode > 0)
+ harq_in_length = RTE_ALIGN_FLOOR(harq_in_length, ACC100_HARQ_ALIGN_COMP);
+
+ if ((harq_layout[harq_index].offset > 0) && harq_prun) {
rte_bbdev_log_debug("HARQ IN offset unexpected for now\n");
fcw->hcin_size0 = harq_layout[harq_index].size0;
fcw->hcin_offset = harq_layout[harq_index].offset;
- fcw->hcin_size1 = harq_in_length -
- harq_layout[harq_index].offset;
+ fcw->hcin_size1 = harq_in_length - harq_layout[harq_index].offset;
} else {
fcw->hcin_size0 = harq_in_length;
fcw->hcin_offset = 0;
@@ -1106,6 +1116,21 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
fcw->hcin_size1 = 0;
}
+ /* Enforce additional check on FCW validity */
+ max_hc_in = RTE_ALIGN_CEIL(fcw->ncb - fcw->nfiller, ACC_HARQ_ALIGN_64B);
+ if ((fcw->hcin_size0 > max_hc_in) ||
+ (fcw->hcin_size1 + fcw->hcin_offset > max_hc_in) ||
+ ((fcw->hcin_size0 > fcw->hcin_offset) &&
+ (fcw->hcin_size1 != 0))) {
+ rte_bbdev_log(ERR, " Invalid FCW : HCIn %d %d %d, Ncb %d F %d",
+ fcw->hcin_size0, fcw->hcin_size1,
+ fcw->hcin_offset,
+ fcw->ncb, fcw->nfiller);
+ /* Disable HARQ input in that case to carry forward */
+ op->ldpc_dec.op_flags ^= RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE;
+ fcw->hcin_en = 0;
+ }
+
fcw->itmax = op->ldpc_dec.iter_max;
fcw->itstop = check_bit(op->ldpc_dec.op_flags,
RTE_BBDEV_LDPC_ITERATION_STOP_ENABLE);
@@ -1130,15 +1155,27 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
if (fcw->hcout_en > 0) {
parity_offset = (op->ldpc_dec.basegraph == 1 ? 20 : 8)
* op->ldpc_dec.z_c - op->ldpc_dec.n_filler;
- k0_p = (fcw->k0 > parity_offset) ?
- fcw->k0 - op->ldpc_dec.n_filler : fcw->k0;
+ k0_p = (fcw->k0 > parity_offset) ? fcw->k0 - op->ldpc_dec.n_filler : fcw->k0;
ncb_p = fcw->ncb - op->ldpc_dec.n_filler;
- l = k0_p + fcw->rm_e;
+ l = RTE_MIN(k0_p + fcw->rm_e, INT16_MAX);
harq_out_length = (uint16_t) fcw->hcin_size0;
- harq_out_length = RTE_MIN(RTE_MAX(harq_out_length, l), ncb_p);
- harq_out_length = (harq_out_length + 0x3F) & 0xFFC0;
- if ((k0_p > fcw->hcin_size0 + ACC_HARQ_OFFSET_THRESHOLD) &&
- harq_prun) {
+ harq_out_length = RTE_MAX(harq_out_length, l);
+
+ /* Stronger alignment when in compression mode */
+ if (fcw->hcout_comp_mode > 0)
+ harq_out_length = RTE_ALIGN_CEIL(harq_out_length, ACC100_HARQ_ALIGN_COMP);
+
+ /* Cannot exceed the pruned Ncb circular buffer */
+ harq_out_length = RTE_MIN(harq_out_length, ncb_p);
+
+ /* Alignment on next 64B */
+ harq_out_length = RTE_ALIGN_CEIL(harq_out_length, ACC_HARQ_ALIGN_64B);
+
+ /* Stronger alignment when in compression mode enforced again */
+ if (fcw->hcout_comp_mode > 0)
+ harq_out_length = RTE_ALIGN_FLOOR(harq_out_length, ACC100_HARQ_ALIGN_COMP);
+
+ if ((k0_p > fcw->hcin_size0 + ACC_HARQ_OFFSET_THRESHOLD) && harq_prun) {
fcw->hcout_size0 = (uint16_t) fcw->hcin_size0;
fcw->hcout_offset = k0_p & 0xFFC0;
fcw->hcout_size1 = harq_out_length - fcw->hcout_offset;
@@ -1147,6 +1184,14 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
fcw->hcout_size1 = 0;
fcw->hcout_offset = 0;
}
+
+ if (fcw->hcout_size0 == 0) {
+ rte_bbdev_log(ERR, " Invalid FCW : HCout %d",
+ fcw->hcout_size0);
+ op->ldpc_dec.op_flags ^= RTE_BBDEV_LDPC_HQ_COMBINE_OUT_ENABLE;
+ fcw->hcout_en = 0;
+ }
+
harq_layout[harq_index].offset = fcw->hcout_offset;
harq_layout[harq_index].size0 = fcw->hcout_size0;
} else {
@@ -1187,6 +1232,10 @@ acc101_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
/* Disable HARQ input in that case to carry forward */
op->ldpc_dec.op_flags ^= RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE;
}
+ if (unlikely(fcw->rm_e == 0)) {
+ rte_bbdev_log(WARNING, "Null E input provided");
+ fcw->rm_e = 2;
+ }
fcw->hcin_en = check_bit(op->ldpc_dec.op_flags,
RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE);
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 08/30] baseband/acc100: allocate ring/queue mem when NULL
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (6 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 07/30] baseband/acc100: enforce additional check on FCW Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 09/30] baseband/acc100: reduce input length for CRC24B Hernan Vargas
` (21 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Allocate info ring, tail pointers and HARQ layout memory for a device
only if it hasn't already been allocated.
Fixes: 06531464151 ("baseband/acc100: support interrupt")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 30 ++++++++++++++++++---------
1 file changed, 20 insertions(+), 10 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 1c83d591e3..00d02ce90d 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -408,9 +408,9 @@ allocate_info_ring(struct rte_bbdev *dev)
reg_addr = &vf_reg_addr;
/* Allocate InfoRing */
d->info_ring = rte_zmalloc_socket("Info Ring",
- ACC_INFO_RING_NUM_ENTRIES *
- sizeof(*d->info_ring), RTE_CACHE_LINE_SIZE,
- dev->data->socket_id);
+ ACC_INFO_RING_NUM_ENTRIES *
+ sizeof(*d->info_ring), RTE_CACHE_LINE_SIZE,
+ dev->data->socket_id);
if (d->info_ring == NULL) {
rte_bbdev_log(ERR,
"Failed to allocate Info Ring for %s:%u",
@@ -499,7 +499,8 @@ acc100_setup_queues(struct rte_bbdev *dev, uint16_t num_queues, int socket_id)
acc_reg_write(d, reg_addr->ring_size, value);
/* Configure tail pointer for use when SDONE enabled */
- d->tail_ptrs = rte_zmalloc_socket(
+ if (d->tail_ptrs == NULL)
+ d->tail_ptrs = rte_zmalloc_socket(
dev->device->driver->name,
ACC100_NUM_QGRPS * ACC100_NUM_AQS * sizeof(uint32_t),
RTE_CACHE_LINE_SIZE, socket_id);
@@ -507,8 +508,8 @@ acc100_setup_queues(struct rte_bbdev *dev, uint16_t num_queues, int socket_id)
rte_bbdev_log(ERR, "Failed to allocate tail ptr for %s:%u",
dev->device->driver->name,
dev->data->dev_id);
- rte_free(d->sw_rings);
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto free_sw_rings;
}
d->tail_ptr_iova = rte_malloc_virt2iova(d->tail_ptrs);
@@ -531,15 +532,16 @@ acc100_setup_queues(struct rte_bbdev *dev, uint16_t num_queues, int socket_id)
/* Continue */
}
- d->harq_layout = rte_zmalloc_socket("HARQ Layout",
+ if (d->harq_layout == NULL)
+ d->harq_layout = rte_zmalloc_socket("HARQ Layout",
ACC_HARQ_LAYOUT * sizeof(*d->harq_layout),
RTE_CACHE_LINE_SIZE, dev->data->socket_id);
if (d->harq_layout == NULL) {
rte_bbdev_log(ERR, "Failed to allocate harq_layout for %s:%u",
dev->device->driver->name,
dev->data->dev_id);
- rte_free(d->sw_rings);
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto free_tail_ptrs;
}
/* Mark as configured properly */
@@ -548,8 +550,16 @@ acc100_setup_queues(struct rte_bbdev *dev, uint16_t num_queues, int socket_id)
rte_bbdev_log_debug(
"ACC100 (%s) configured sw_rings = %p, sw_rings_iova = %#"
PRIx64, dev->data->name, d->sw_rings, d->sw_rings_iova);
-
return 0;
+
+free_tail_ptrs:
+ rte_free(d->tail_ptrs);
+ d->tail_ptrs = NULL;
+free_sw_rings:
+ rte_free(d->sw_rings_base);
+ d->sw_rings = NULL;
+
+ return ret;
}
static int
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 09/30] baseband/acc100: reduce input length for CRC24B
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (7 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 08/30] baseband/acc100: allocate ring/queue mem when NULL Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 10/30] baseband/acc100: fix clearing PF IR outside handler Hernan Vargas
` (20 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Input length should be reduced only for CRC24B.
Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 00d02ce90d..b79a251e9f 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1443,8 +1443,7 @@ acc100_dma_desc_le_fill(struct rte_bbdev_enc_op *op,
K = (enc->basegraph == 1 ? 22 : 10) * enc->z_c;
in_length_in_bits = K - enc->n_filler;
- if ((enc->op_flags & RTE_BBDEV_LDPC_CRC_24A_ATTACH) ||
- (enc->op_flags & RTE_BBDEV_LDPC_CRC_24B_ATTACH))
+ if (enc->op_flags & RTE_BBDEV_LDPC_CRC_24B_ATTACH)
in_length_in_bits -= 24;
in_length_in_bytes = in_length_in_bits >> 3;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 10/30] baseband/acc100: fix clearing PF IR outside handler
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (8 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 09/30] baseband/acc100: reduce input length for CRC24B Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:38 ` [PATCH v4 11/30] baseband/acc100: set device min alignment to 1 Hernan Vargas
` (19 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Clearing of PF info ring outside of handler may cause interrupt to be
missed.
A condition in the ACC100 PMD implementation may cause an interrupt
functional handler call to be missed due to related bit being cleared
when checking PF info ring status.
Fixes: 06531464151 ("baseband/acc100: support interrupt")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index b79a251e9f..f892b402ca 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -262,11 +262,12 @@ acc100_check_ir(struct acc_device *acc100_dev)
while (ring_data->valid) {
if ((ring_data->int_nb < ACC100_PF_INT_DMA_DL_DESC_IRQ) || (
ring_data->int_nb >
- ACC100_PF_INT_DMA_DL5G_DESC_IRQ))
+ ACC100_PF_INT_DMA_DL5G_DESC_IRQ)) {
rte_bbdev_log(WARNING, "InfoRing: ITR:%d Info:0x%x",
ring_data->int_nb, ring_data->detailed_info);
- /* Initialize Info Ring entry and move forward */
- ring_data->val = 0;
+ /* Initialize Info Ring entry and move forward */
+ ring_data->val = 0;
+ }
info_ring_head++;
ring_data = acc100_dev->info_ring +
(info_ring_head & ACC_INFO_RING_MASK);
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 11/30] baseband/acc100: set device min alignment to 1
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (9 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 10/30] baseband/acc100: fix clearing PF IR outside handler Hernan Vargas
@ 2022-10-19 0:38 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 12/30] baseband/acc100: add protection for NULL HARQ input Hernan Vargas
` (18 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:38 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Historical mistakes, there should be no 64B alignment requirement for
the buffer being processed. Any 1B alignment is sufficient.
Fixes: 9200ffa5cd5 ("baseband/acc100: add info get function")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index f892b402ca..8d11699c0c 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -938,7 +938,7 @@ acc100_dev_info_get(struct rte_bbdev *dev,
d->acc_conf.q_ul_4g.num_qgroups - 1;
dev_info->default_queue_conf = default_queue_conf;
dev_info->cpu_flag_reqs = NULL;
- dev_info->min_alignment = 64;
+ dev_info->min_alignment = 1;
dev_info->capabilities = bbdev_capabilities;
#ifdef ACC100_EXT_MEM
dev_info->harq_buffer_size = d->ddr_size;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 12/30] baseband/acc100: add protection for NULL HARQ input
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (10 preceding siblings ...)
2022-10-19 0:38 ` [PATCH v4 11/30] baseband/acc100: set device min alignment to 1 Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 13/30] baseband/acc100: reset pointer after rte_free Hernan Vargas
` (17 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
It is possible to cause an invalid HW operation in case the user
provides the BBDEV API and HARQ operation with input enabled and zero
input. Adding protection for that case.
Fixes: 5ad5060f8f7 ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 8d11699c0c..ed4a08ea60 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1067,6 +1067,14 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
op->ldpc_dec.tb_params.ea :
op->ldpc_dec.tb_params.eb;
+ if (unlikely(check_bit(op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE) &&
+ (op->ldpc_dec.harq_combined_input.length == 0))) {
+ rte_bbdev_log(WARNING, "Null HARQ input size provided");
+ /* Disable HARQ input in that case to carry forward. */
+ op->ldpc_dec.op_flags ^= RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE;
+ }
+
fcw->hcin_en = check_bit(op->ldpc_dec.op_flags,
RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE);
fcw->hcout_en = check_bit(op->ldpc_dec.op_flags,
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 13/30] baseband/acc100: reset pointer after rte_free
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (11 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 12/30] baseband/acc100: add protection for NULL HARQ input Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 14/30] baseband/acc100: fix debug print for LDPC FCW Hernan Vargas
` (16 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas, stable
Set local pointer to NULL after rte_free.
This needs to be set explicitly since logic may check for null pointers.
Fixes: 060e7672930 ("baseband/acc100: add queue configuration")
Cc: stable@dpdk.org
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index ed4a08ea60..13eefaad80 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -618,6 +618,9 @@ acc100_dev_close(struct rte_bbdev *dev)
rte_free(d->info_ring);
rte_free(d->sw_rings_base);
d->sw_rings_base = NULL;
+ d->tail_ptrs = NULL;
+ d->info_ring = NULL;
+ d->harq_layout = NULL;
}
/* Ensure all in flight HW transactions are completed */
usleep(ACC_LONG_WAIT);
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 14/30] baseband/acc100: fix debug print for LDPC FCW
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (12 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 13/30] baseband/acc100: reset pointer after rte_free Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 15/30] baseband/acc100: add enqueue status Hernan Vargas
` (15 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Print full size of FCW LDPC structure on debug messages.
This is just a cosmetic fix, no need to fix on previous code base.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 13eefaad80..fc2fbe5511 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -2756,7 +2756,7 @@ enqueue_ldpc_dec_one_op_cb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
#ifdef RTE_LIBRTE_BBDEV_DEBUG
rte_memdump(stderr, "FCW", &desc->req.fcw_ld,
- sizeof(desc->req.fcw_ld) - 8);
+ sizeof(desc->req.fcw_ld));
rte_memdump(stderr, "Req Desc.", desc, sizeof(*desc));
#endif
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 15/30] baseband/acc100: add enqueue status
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (13 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 14/30] baseband/acc100: fix debug print for LDPC FCW Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 16/30] baseband/acc100: add scatter-gather support Hernan Vargas
` (14 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Add enqueue status as part of rte_bbdev_queue_data.
This is a new feature to update queue status and indicate the reason why
a previous enqueue may or may not have consumed all requested operations.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 56 ++++++++++++++++++++-------
1 file changed, 42 insertions(+), 14 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index fc2fbe5511..ca92ae906f 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -2968,13 +2968,17 @@ acc100_enqueue_enc_cb(struct rte_bbdev_queue_data *q_data,
for (i = 0; i < num; ++i) {
/* Check if there are available space for further processing */
- if (unlikely(avail - 1 < 0))
+ if (unlikely(avail - 1 < 0)) {
+ acc_enqueue_ring_full(q_data);
break;
+ }
avail -= 1;
ret = enqueue_enc_one_op_cb(q, ops[i], i);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
}
if (unlikely(i == 0))
@@ -3006,20 +3010,26 @@ acc100_enqueue_ldpc_enc_cb(struct rte_bbdev_queue_data *q_data,
int16_t enq, left = num;
while (left > 0) {
- if (unlikely(avail < 1))
+ if (unlikely(avail < 1)) {
+ acc_enqueue_ring_full(q_data);
break;
+ }
avail--;
enq = RTE_MIN(left, ACC_MUX_5GDL_DESC);
if (check_mux(&ops[i], enq)) {
ret = enqueue_ldpc_enc_n_op_cb(q, &ops[i],
desc_idx, enq);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
i += enq;
} else {
ret = enqueue_ldpc_enc_one_op_cb(q, ops[i], desc_idx);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
i++;
}
desc_idx++;
@@ -3057,13 +3067,17 @@ acc100_enqueue_enc_tb(struct rte_bbdev_queue_data *q_data,
for (i = 0; i < num; ++i) {
cbs_in_tb = get_num_cbs_in_tb_enc(&ops[i]->turbo_enc);
/* Check if there are available space for further processing */
- if (unlikely(avail - cbs_in_tb < 0))
+ if (unlikely(avail - cbs_in_tb < 0)) {
+ acc_enqueue_ring_full(q_data);
break;
+ }
avail -= cbs_in_tb;
ret = enqueue_enc_one_op_tb(q, ops[i], enqueued_cbs, cbs_in_tb);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
enqueued_cbs += ret;
}
if (unlikely(enqueued_cbs == 0))
@@ -3120,13 +3134,17 @@ acc100_enqueue_dec_cb(struct rte_bbdev_queue_data *q_data,
for (i = 0; i < num; ++i) {
/* Check if there are available space for further processing */
- if (unlikely(avail - 1 < 0))
+ if (unlikely(avail - 1 < 0)) {
+ acc_enqueue_ring_full(q_data);
break;
+ }
avail -= 1;
ret = enqueue_dec_one_op_cb(q, ops[i], i);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
}
if (unlikely(i == 0))
@@ -3166,8 +3184,10 @@ acc100_enqueue_ldpc_dec_tb(struct rte_bbdev_queue_data *q_data,
ret = enqueue_ldpc_dec_one_op_tb(q, ops[i],
enqueued_cbs, cbs_in_tb);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
enqueued_cbs += ret;
}
if (unlikely(enqueued_cbs == 0))
@@ -3194,8 +3214,10 @@ acc100_enqueue_ldpc_dec_cb(struct rte_bbdev_queue_data *q_data,
bool same_op = false;
for (i = 0; i < num; ++i) {
/* Check if there are available space for further processing */
- if (unlikely(avail < 1))
+ if (unlikely(avail < 1)) {
+ acc_enqueue_ring_full(q_data);
break;
+ }
avail -= 1;
if (i > 0)
@@ -3208,8 +3230,10 @@ acc100_enqueue_ldpc_dec_cb(struct rte_bbdev_queue_data *q_data,
ops[i]->ldpc_dec.n_filler, ops[i]->ldpc_dec.cb_params.e,
same_op);
ret = enqueue_ldpc_dec_one_op_cb(q, ops[i], i, same_op);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
}
if (unlikely(i == 0))
@@ -3244,13 +3268,17 @@ acc100_enqueue_dec_tb(struct rte_bbdev_queue_data *q_data,
for (i = 0; i < num; ++i) {
cbs_in_tb = get_num_cbs_in_tb_dec(&ops[i]->turbo_dec);
/* Check if there are available space for further processing */
- if (unlikely(avail - cbs_in_tb < 0))
+ if (unlikely(avail - cbs_in_tb < 0)) {
+ acc_enqueue_ring_full(q_data);
break;
+ }
avail -= cbs_in_tb;
ret = enqueue_dec_one_op_tb(q, ops[i], enqueued_cbs, cbs_in_tb);
- if (ret < 0)
+ if (ret < 0) {
+ acc_enqueue_invalid(q_data);
break;
+ }
enqueued_cbs += ret;
}
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 16/30] baseband/acc100: add scatter-gather support
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (14 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 15/30] baseband/acc100: add enqueue status Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 17/30] baseband/acc100: add HARQ index helper function Hernan Vargas
` (13 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Add flag to support scatter-gather for the mbuf
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 38 ++++++++++++++++++---------
1 file changed, 25 insertions(+), 13 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index ca92ae906f..0921e9a44d 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1363,6 +1363,8 @@ acc101_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
* Store information about device capabilities
* @param next_triplet
* Index for ACC100 DMA Descriptor triplet
+ * @param scattergather
+ * Flag to support scatter-gather for the mbuf
*
* @return
* Returns index of next triplet on success, other value if lengths of
@@ -1372,12 +1374,16 @@ acc101_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
static inline int
acc100_dma_fill_blk_type_in(struct acc_dma_req_desc *desc,
struct rte_mbuf **input, uint32_t *offset, uint32_t cb_len,
- uint32_t *seg_total_left, int next_triplet)
+ uint32_t *seg_total_left, int next_triplet,
+ bool scattergather)
{
uint32_t part_len;
struct rte_mbuf *m = *input;
- part_len = (*seg_total_left < cb_len) ? *seg_total_left : cb_len;
+ if (scattergather)
+ part_len = (*seg_total_left < cb_len) ? *seg_total_left : cb_len;
+ else
+ part_len = cb_len;
cb_len -= part_len;
*seg_total_left -= part_len;
@@ -1468,7 +1474,7 @@ acc100_dma_desc_le_fill(struct rte_bbdev_enc_op *op,
}
next_triplet = acc100_dma_fill_blk_type_in(desc, input, in_offset,
- pad_le_in(in_length_in_bytes, q), seg_total_left, next_triplet);
+ pad_le_in(in_length_in_bytes, q), seg_total_left, next_triplet, false);
if (unlikely(next_triplet < 0)) {
rte_bbdev_log(ERR,
"Mismatch between data to process and mbuf data length in bbdev_op: %p",
@@ -1556,7 +1562,9 @@ acc100_dma_desc_td_fill(struct rte_bbdev_dec_op *op,
}
next_triplet = acc100_dma_fill_blk_type_in(desc, input, in_offset, kw,
- seg_total_left, next_triplet);
+ seg_total_left, next_triplet,
+ check_bit(op->turbo_dec.op_flags,
+ RTE_BBDEV_TURBO_DEC_SCATTER_GATHER));
if (unlikely(next_triplet < 0)) {
rte_bbdev_log(ERR,
"Mismatch between data to process and mbuf data length in bbdev_op: %p",
@@ -1658,7 +1666,9 @@ acc100_dma_desc_ld_fill(struct rte_bbdev_dec_op *op,
next_triplet = acc100_dma_fill_blk_type_in(desc, input,
in_offset, input_length,
- seg_total_left, next_triplet);
+ seg_total_left, next_triplet,
+ check_bit(op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_DEC_SCATTER_GATHER));
if (unlikely(next_triplet < 0)) {
rte_bbdev_log(ERR,
@@ -2726,10 +2736,9 @@ enqueue_ldpc_dec_one_op_cb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
fcw = &desc->req.fcw_ld;
q->d->fcw_ld_fill(op, fcw, harq_layout);
- /* Special handling when overusing mbuf */
- if (fcw->rm_e < ACC_MAX_E_MBUF)
- seg_total_left = rte_pktmbuf_data_len(input)
- - in_offset;
+ /* Special handling when using mbuf or not */
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_DEC_SCATTER_GATHER))
+ seg_total_left = rte_pktmbuf_data_len(input) - in_offset;
else
seg_total_left = fcw->rm_e;
@@ -2804,9 +2813,10 @@ enqueue_ldpc_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
r = op->ldpc_dec.tb_params.r;
while (mbuf_total_left > 0 && r < c) {
-
- seg_total_left = rte_pktmbuf_data_len(input) - in_offset;
-
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_DEC_SCATTER_GATHER))
+ seg_total_left = rte_pktmbuf_data_len(input) - in_offset;
+ else
+ seg_total_left = op->ldpc_dec.input.length;
/* Set up DMA descriptor */
desc = acc_desc(q, total_enqueued_cbs);
desc->req.data_ptrs[0].address = q->ring_addr_iova + fcw_offset;
@@ -2831,7 +2841,9 @@ enqueue_ldpc_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
rte_memdump(stderr, "Req Desc.", desc, sizeof(*desc));
#endif
- if (seg_total_left == 0) {
+ if (check_bit(op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_DEC_SCATTER_GATHER)
+ && (seg_total_left == 0)) {
/* Go to the next mbuf */
input = input->next;
in_offset = 0;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 17/30] baseband/acc100: add HARQ index helper function
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (15 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 16/30] baseband/acc100: add scatter-gather support Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 18/30] baseband/acc100: enable input validation by default Hernan Vargas
` (12 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Refactor code to use the HARQ index helper function and make harq_idx
uint32.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 36 +++++++++++++--------------
1 file changed, 17 insertions(+), 19 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 0921e9a44d..d0c98ced32 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1050,7 +1050,7 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
union acc_harq_layout_data *harq_layout)
{
uint16_t harq_out_length, harq_in_length, ncb_p, k0_p, parity_offset;
- uint16_t harq_index;
+ uint32_t harq_index;
uint32_t l;
bool harq_prun = false;
uint32_t max_hc_in;
@@ -1098,8 +1098,7 @@ acc100_fcw_ld_fill(struct rte_bbdev_dec_op *op, struct acc_fcw_ld *fcw,
RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION);
fcw->llr_pack_mode = check_bit(op->ldpc_dec.op_flags,
RTE_BBDEV_LDPC_LLR_COMPRESSION);
- harq_index = op->ldpc_dec.harq_combined_output.offset /
- ACC_HARQ_OFFSET;
+ harq_index = hq_index(op->ldpc_dec.harq_combined_output.offset);
#ifdef ACC100_EXT_MEM
/* Limit cases when HARQ pruning is valid */
harq_prun = ((op->ldpc_dec.harq_combined_output.offset %
@@ -1777,20 +1776,17 @@ acc100_dma_desc_ld_update(struct rte_bbdev_dec_op *op,
*h_out_length = desc->data_ptrs[next_triplet].blen;
next_triplet++;
- if (check_bit(op->ldpc_dec.op_flags,
- RTE_BBDEV_LDPC_HQ_COMBINE_OUT_ENABLE)) {
- desc->data_ptrs[next_triplet].address =
- op->ldpc_dec.harq_combined_output.offset;
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HQ_COMBINE_OUT_ENABLE)) {
+ struct rte_bbdev_dec_op *prev_op;
+ uint32_t harq_idx, prev_harq_idx;
+ desc->data_ptrs[next_triplet].address = op->ldpc_dec.harq_combined_output.offset;
/* Adjust based on previous operation */
- struct rte_bbdev_dec_op *prev_op = desc->op_addr;
+ prev_op = desc->op_addr;
op->ldpc_dec.harq_combined_output.length =
prev_op->ldpc_dec.harq_combined_output.length;
- int16_t hq_idx = op->ldpc_dec.harq_combined_output.offset /
- ACC_HARQ_OFFSET;
- int16_t prev_hq_idx =
- prev_op->ldpc_dec.harq_combined_output.offset
- / ACC_HARQ_OFFSET;
- harq_layout[hq_idx].val = harq_layout[prev_hq_idx].val;
+ harq_idx = hq_index(op->ldpc_dec.harq_combined_output.offset);
+ prev_harq_idx = hq_index(prev_op->ldpc_dec.harq_combined_output.offset);
+ harq_layout[harq_idx].val = harq_layout[prev_harq_idx].val;
#ifndef ACC100_EXT_MEM
struct rte_bbdev_op_data ho =
op->ldpc_dec.harq_combined_output;
@@ -2534,6 +2530,9 @@ harq_loopback(struct acc_queue *q, struct rte_bbdev_dec_op *op,
struct rte_mbuf *hq_output_head, *hq_output;
uint16_t harq_dma_length_in, harq_dma_length_out;
uint16_t harq_in_length = op->ldpc_dec.harq_combined_input.length;
+ bool ddr_mem_in;
+ union acc_harq_layout_data *harq_layout;
+ uint32_t harq_index;
if (harq_in_length == 0) {
rte_bbdev_log(ERR, "Loopback of invalid null size\n");
@@ -2553,13 +2552,12 @@ harq_loopback(struct acc_queue *q, struct rte_bbdev_dec_op *op,
}
harq_dma_length_out = harq_dma_length_in;
- bool ddr_mem_in = check_bit(op->ldpc_dec.op_flags,
+ ddr_mem_in = check_bit(op->ldpc_dec.op_flags,
RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_IN_ENABLE);
- union acc_harq_layout_data *harq_layout = q->d->harq_layout;
- uint16_t harq_index = (ddr_mem_in ?
+ harq_layout = q->d->harq_layout;
+ harq_index = hq_index(ddr_mem_in ?
op->ldpc_dec.harq_combined_input.offset :
- op->ldpc_dec.harq_combined_output.offset)
- / ACC_HARQ_OFFSET;
+ op->ldpc_dec.harq_combined_output.offset);
desc = acc_desc(q, total_enqueued_cbs);
fcw = &desc->req.fcw_ld;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 18/30] baseband/acc100: enable input validation by default
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (16 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 17/30] baseband/acc100: add HARQ index helper function Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 19/30] baseband/acc100: added LDPC transport block support Hernan Vargas
` (11 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Enable validation functions by default and provide a new flag
RTE_LIBRTE_SKIP_VALIDATE if the user wants to run without
validating input to save cycles.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 36 +++++++++++++--------------
1 file changed, 18 insertions(+), 18 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index d0c98ced32..c75a426d38 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1800,7 +1800,7 @@ acc100_dma_desc_ld_update(struct rte_bbdev_dec_op *op,
desc->op_addr = op;
}
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validates turbo encoder parameters */
static inline int
validate_enc_op(struct rte_bbdev_enc_op *op, struct acc_queue *q)
@@ -2063,10 +2063,10 @@ enqueue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
seg_total_left;
struct rte_mbuf *input, *output_head, *output;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_enc_op(op, q) == -1) {
- rte_bbdev_log(ERR, "Turbo encoder validation failed");
+ rte_bbdev_log(ERR, "Turbo encoder validation rejected");
return -EINVAL;
}
#endif
@@ -2115,10 +2115,10 @@ enqueue_ldpc_enc_n_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ops,
uint16_t in_length_in_bytes;
struct rte_bbdev_op_ldpc_enc *enc = &ops[0]->ldpc_enc;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_ldpc_enc_op(ops[0], q) == -1) {
- rte_bbdev_log(ERR, "LDPC encoder validation failed");
+ rte_bbdev_log(ERR, "LDPC encoder validation rejected");
return -EINVAL;
}
#endif
@@ -2175,10 +2175,10 @@ enqueue_ldpc_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
seg_total_left;
struct rte_mbuf *input, *output_head, *output;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_ldpc_enc_op(op, q) == -1) {
- rte_bbdev_log(ERR, "LDPC encoder validation failed");
+ rte_bbdev_log(ERR, "LDPC encoder validation rejected");
return -EINVAL;
}
#endif
@@ -2231,10 +2231,10 @@ enqueue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
uint16_t desc_idx, current_enqueued_cbs = 0;
uint64_t fcw_offset;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_enc_op(op, q) == -1) {
- rte_bbdev_log(ERR, "Turbo encoder validation failed");
+ rte_bbdev_log(ERR, "Turbo encoder validation rejected");
return -EINVAL;
}
#endif
@@ -2305,7 +2305,7 @@ enqueue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
return current_enqueued_cbs;
}
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validates turbo decoder parameters */
static inline int
validate_dec_op(struct rte_bbdev_dec_op *op, struct acc_queue *q)
@@ -2463,10 +2463,10 @@ enqueue_dec_one_op_cb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
struct rte_mbuf *input, *h_output_head, *h_output,
*s_output_head, *s_output;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_dec_op(op, q) == -1) {
- rte_bbdev_log(ERR, "Turbo decoder validation failed");
+ rte_bbdev_log(ERR, "Turbo decoder validation rejected");
return -EINVAL;
}
#endif
@@ -2686,10 +2686,10 @@ enqueue_ldpc_dec_one_op_cb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
return ret;
}
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_ldpc_dec_op(op, q) == -1) {
- rte_bbdev_log(ERR, "LDPC decoder validation failed");
+ rte_bbdev_log(ERR, "LDPC decoder validation rejected");
return -EINVAL;
}
#endif
@@ -2787,10 +2787,10 @@ enqueue_ldpc_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
uint64_t fcw_offset;
union acc_harq_layout_data *harq_layout;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_ldpc_dec_op(op, q) == -1) {
- rte_bbdev_log(ERR, "LDPC decoder validation failed");
+ rte_bbdev_log(ERR, "LDPC decoder validation rejected");
return -EINVAL;
}
#endif
@@ -2879,10 +2879,10 @@ enqueue_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
uint16_t desc_idx, current_enqueued_cbs = 0;
uint64_t fcw_offset;
-#ifdef RTE_LIBRTE_BBDEV_DEBUG
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
if (validate_dec_op(op, q) == -1) {
- rte_bbdev_log(ERR, "Turbo decoder validation failed");
+ rte_bbdev_log(ERR, "Turbo decoder validation rejected");
return -EINVAL;
}
#endif
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 19/30] baseband/acc100: added LDPC transport block support
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (17 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 18/30] baseband/acc100: enable input validation by default Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 20/30] baseband/acc100: update validate LDPC enc/dec Hernan Vargas
` (10 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Added LDPC enqueue functions to handle transport blocks.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 164 +++++++++++++++++++++++++-
1 file changed, 163 insertions(+), 1 deletion(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index c75a426d38..580780c274 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -2164,6 +2164,56 @@ enqueue_ldpc_enc_n_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ops,
return num;
}
+/* Enqueue one encode operations for ACC100 device for a partial TB
+ * all codes blocks have same configuration multiplexed on the same descriptor.
+ */
+static inline void
+enqueue_ldpc_enc_part_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
+ uint16_t total_enqueued_descs, int16_t num_cbs, uint32_t e,
+ uint16_t in_len_bytes, uint32_t out_len_bytes, uint32_t *in_offset,
+ uint32_t *out_offset)
+{
+ union acc_dma_desc *desc = NULL;
+ struct rte_mbuf *output_head, *output;
+ int i, next_triplet;
+ struct rte_bbdev_op_ldpc_enc *enc = &op->ldpc_enc;
+ uint16_t desc_idx = ((q->sw_ring_head + total_enqueued_descs) & q->sw_ring_wrap_mask);
+
+ desc = q->ring_addr + desc_idx;
+ acc_fcw_le_fill(op, &desc->req.fcw_le, num_cbs, e);
+
+ /* This could be done at polling. */
+ acc_header_init(&desc->req);
+ desc->req.numCBs = num_cbs;
+
+ desc->req.m2dlen = 1 + num_cbs;
+ desc->req.d2mlen = num_cbs;
+ next_triplet = 1;
+
+ for (i = 0; i < num_cbs; i++) {
+ desc->req.data_ptrs[next_triplet].address =
+ rte_pktmbuf_iova_offset(enc->input.data, *in_offset);
+ *in_offset += in_len_bytes;
+ desc->req.data_ptrs[next_triplet].blen = in_len_bytes;
+ next_triplet++;
+ desc->req.data_ptrs[next_triplet].address =
+ rte_pktmbuf_iova_offset(enc->output.data, *out_offset);
+ *out_offset += out_len_bytes;
+ desc->req.data_ptrs[next_triplet].blen = out_len_bytes;
+ next_triplet++;
+ enc->output.length += out_len_bytes;
+ output_head = output = enc->output.data;
+ mbuf_append(output_head, output, out_len_bytes);
+ }
+
+#ifdef RTE_LIBRTE_BBDEV_DEBUG
+ rte_memdump(stderr, "FCW", &desc->req.fcw_le,
+ sizeof(desc->req.fcw_le) - 8);
+ rte_memdump(stderr, "Req Desc.", desc, sizeof(*desc));
+#endif
+
+}
+
/* Enqueue one encode operations for ACC100 device in CB mode */
static inline int
enqueue_ldpc_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
@@ -2305,6 +2355,76 @@ enqueue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
return current_enqueued_cbs;
}
+/* Enqueue one encode operations for ACC100 device in TB mode.
+ * returns the number of descs used.
+ */
+static inline int
+enqueue_ldpc_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
+ uint16_t enq_descs, uint8_t cbs_in_tb)
+{
+#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
+ if (validate_ldpc_enc_op(op, q) == -1) {
+ rte_bbdev_log(ERR, "LDPC encoder validation failed");
+ return -EINVAL;
+ }
+#endif
+ uint8_t num_a, num_b;
+ uint16_t desc_idx;
+ uint8_t r = op->ldpc_enc.tb_params.r;
+ uint8_t cab = op->ldpc_enc.tb_params.cab;
+ union acc_dma_desc *desc;
+ uint16_t init_enq_descs = enq_descs;
+ uint16_t input_len_B = ((op->ldpc_enc.basegraph == 1 ? 22 : 10) *
+ op->ldpc_enc.z_c - op->ldpc_enc.n_filler) >> 3;
+ uint32_t in_offset = 0, out_offset = 0;
+ uint16_t return_descs;
+
+ if (check_bit(op->ldpc_enc.op_flags, RTE_BBDEV_LDPC_CRC_24B_ATTACH))
+ input_len_B -= 3;
+
+ if (r < cab) {
+ num_a = cab - r;
+ num_b = cbs_in_tb - cab;
+ } else {
+ num_a = 0;
+ num_b = cbs_in_tb - r;
+ }
+
+ while (num_a > 0) {
+ uint32_t e = op->ldpc_enc.tb_params.ea;
+ uint32_t out_len_bytes = (e + 7) >> 3;
+ uint8_t enq = RTE_MIN(num_a, ACC_MUX_5GDL_DESC);
+ num_a -= enq;
+ enqueue_ldpc_enc_part_tb(q, op, enq_descs, enq, e, input_len_B,
+ out_len_bytes, &in_offset, &out_offset);
+ enq_descs++;
+ }
+ while (num_b > 0) {
+ uint32_t e = op->ldpc_enc.tb_params.eb;
+ uint32_t out_len_bytes = (e + 7) >> 3;
+ uint8_t enq = RTE_MIN(num_b, ACC_MUX_5GDL_DESC);
+ num_b -= enq;
+ enqueue_ldpc_enc_part_tb(q, op, enq_descs, enq, e, input_len_B,
+ out_len_bytes, &in_offset, &out_offset);
+ enq_descs++;
+ }
+
+ return_descs = enq_descs - init_enq_descs;
+ /* Keep total number of CBs in first TB. */
+ desc_idx = ((q->sw_ring_head + init_enq_descs) & q->sw_ring_wrap_mask);
+ desc = q->ring_addr + desc_idx;
+ desc->req.cbs_in_tb = return_descs; /** Actual number of descriptors. */
+ desc->req.op_addr = op;
+
+ /* Set SDone on last CB descriptor for TB mode. */
+ desc_idx = ((q->sw_ring_head + enq_descs - 1) & q->sw_ring_wrap_mask);
+ desc = q->ring_addr + desc_idx;
+ desc->req.sdone_enable = 1;
+ desc->req.irq_enable = q->irq_enable;
+ desc->req.op_addr = op;
+ return return_descs;
+}
+
#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validates turbo decoder parameters */
static inline int
@@ -2881,6 +3001,10 @@ enqueue_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
/* Validate op structure */
+ if (cbs_in_tb == 0) {
+ rte_bbdev_log(ERR, "Turbo decoder invalid number of CBs");
+ return -EINVAL;
+ }
if (validate_dec_op(op, q) == -1) {
rte_bbdev_log(ERR, "Turbo decoder validation rejected");
return -EINVAL;
@@ -3102,6 +3226,44 @@ acc100_enqueue_enc_tb(struct rte_bbdev_queue_data *q_data,
return i;
}
+/* Enqueue LDPC encode operations for ACC100 device in TB mode. */
+static uint16_t
+acc100_enqueue_ldpc_enc_tb(struct rte_bbdev_queue_data *q_data,
+ struct rte_bbdev_enc_op **ops, uint16_t num)
+{
+ struct acc_queue *q = q_data->queue_private;
+ int32_t avail = acc_ring_avail_enq(q);
+ uint16_t i, enqueued_descs = 0;
+ uint8_t cbs_in_tb;
+ int descs_used;
+
+ for (i = 0; i < num; ++i) {
+ cbs_in_tb = get_num_cbs_in_tb_ldpc_enc(&ops[i]->ldpc_enc);
+ /* Check if there are available space for further processing. */
+ if (unlikely(avail - cbs_in_tb < 0)) {
+ acc_enqueue_ring_full(q_data);
+ break;
+ }
+ descs_used = enqueue_ldpc_enc_one_op_tb(q, ops[i], enqueued_descs, cbs_in_tb);
+ if (descs_used < 0) {
+ acc_enqueue_invalid(q_data);
+ break;
+ }
+ enqueued_descs += descs_used;
+ avail -= descs_used;
+ }
+ if (unlikely(enqueued_descs == 0))
+ return 0; /* Nothing to enqueue. */
+
+ acc_dma_enqueue(q, enqueued_descs, &q_data->queue_stats);
+
+ /* Update stats. */
+ q_data->queue_stats.enqueued_count += i;
+ q_data->queue_stats.enqueue_err_count += num - i;
+
+ return i;
+}
+
/* Enqueue encode operations for ACC100 device. */
static uint16_t
acc100_enqueue_enc(struct rte_bbdev_queue_data *q_data,
@@ -3125,7 +3287,7 @@ acc100_enqueue_ldpc_enc(struct rte_bbdev_queue_data *q_data,
if (unlikely((aq_avail <= 0) || (num == 0)))
return 0;
if (ops[0]->ldpc_enc.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
- return acc100_enqueue_enc_tb(q_data, ops, num);
+ return acc100_enqueue_ldpc_enc_tb(q_data, ops, num);
else
return acc100_enqueue_ldpc_enc_cb(q_data, ops, num);
}
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 20/30] baseband/acc100: update validate LDPC enc/dec
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (18 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 19/30] baseband/acc100: added LDPC transport block support Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 21/30] baseband/acc100: implement configurable queue depth Hernan Vargas
` (9 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Update validate functions to check for valid LDPC parameters to avoid
any HW issues.
Adding protection for null corner case and for HARQ inbound size out
of range.
HARQ input size from application may be invalid and causing HW issue.
Add checks to ensure that if HARQ is invalid, set to some valid size to
ensure HW issues do not occur.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/acc_common.h | 1 +
drivers/baseband/acc/rte_acc100_pmd.c | 285 +++++++++++++++++++++++---
2 files changed, 255 insertions(+), 31 deletions(-)
diff --git a/drivers/baseband/acc/acc_common.h b/drivers/baseband/acc/acc_common.h
index 97d10b8b40..eae7eab4e9 100644
--- a/drivers/baseband/acc/acc_common.h
+++ b/drivers/baseband/acc/acc_common.h
@@ -121,6 +121,7 @@
#define ACC_ALGO_SPA 0
#define ACC_ALGO_MSA 1
#define ACC_HARQ_ALIGN_64B 64
+#define ACC_MAX_ZC 384
/* Helper macro for logging */
#define rte_acc_log(level, fmt, ...) \
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 580780c274..94863d7afd 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1953,14 +1953,11 @@ static inline int
validate_ldpc_enc_op(struct rte_bbdev_enc_op *op, struct acc_queue *q)
{
struct rte_bbdev_op_ldpc_enc *ldpc_enc = &op->ldpc_enc;
+ int K, N, q_m, crc24;
if (!validate_op_required(q))
return 0;
- if (op->mempool == NULL) {
- rte_bbdev_log(ERR, "Invalid mempool pointer");
- return -1;
- }
if (ldpc_enc->input.data == NULL) {
rte_bbdev_log(ERR, "Invalid input pointer");
return -1;
@@ -1969,17 +1966,12 @@ validate_ldpc_enc_op(struct rte_bbdev_enc_op *op, struct acc_queue *q)
rte_bbdev_log(ERR, "Invalid output pointer");
return -1;
}
- if (ldpc_enc->input.length >
- RTE_BBDEV_LDPC_MAX_CB_SIZE >> 3) {
- rte_bbdev_log(ERR, "CB size (%u) is too big, max: %d",
- ldpc_enc->input.length,
- RTE_BBDEV_LDPC_MAX_CB_SIZE);
+ if (ldpc_enc->input.length == 0) {
+ rte_bbdev_log(ERR, "CB size (%u) is null", ldpc_enc->input.length);
return -1;
}
if ((ldpc_enc->basegraph > 2) || (ldpc_enc->basegraph == 0)) {
- rte_bbdev_log(ERR,
- "BG (%u) is out of range 1 <= value <= 2",
- ldpc_enc->basegraph);
+ rte_bbdev_log(ERR, "BG (%u) is out of range 1 <= value <= 2", ldpc_enc->basegraph);
return -1;
}
if (ldpc_enc->rv_index > 3) {
@@ -1994,13 +1986,89 @@ validate_ldpc_enc_op(struct rte_bbdev_enc_op *op, struct acc_queue *q)
ldpc_enc->code_block_mode);
return -1;
}
- int K = (ldpc_enc->basegraph == 1 ? 22 : 10) * ldpc_enc->z_c;
- if (ldpc_enc->n_filler >= K) {
- rte_bbdev_log(ERR,
- "K and F are not compatible %u %u",
- K, ldpc_enc->n_filler);
+ if (ldpc_enc->z_c > ACC_MAX_ZC) {
+ rte_bbdev_log(ERR, "Zc (%u) is out of range", ldpc_enc->z_c);
+ return -1;
+ }
+
+ K = (ldpc_enc->basegraph == 1 ? 22 : 10) * ldpc_enc->z_c;
+ N = (ldpc_enc->basegraph == 1 ? ACC_N_ZC_1 : ACC_N_ZC_2) * ldpc_enc->z_c;
+ q_m = ldpc_enc->q_m;
+ crc24 = 0;
+
+ if (check_bit(op->ldpc_enc.op_flags,
+ RTE_BBDEV_LDPC_CRC_24A_ATTACH) ||
+ check_bit(op->ldpc_enc.op_flags,
+ RTE_BBDEV_LDPC_CRC_24B_ATTACH))
+ crc24 = 24;
+ if ((K - ldpc_enc->n_filler) % 8 > 0) {
+ rte_bbdev_log(ERR, "K - F not byte aligned %u", K - ldpc_enc->n_filler);
+ return -1;
+ }
+ if (ldpc_enc->n_filler > (K - 2 * ldpc_enc->z_c)) {
+ rte_bbdev_log(ERR, "K - F invalid %u %u", K, ldpc_enc->n_filler);
+ return -1;
+ }
+ if ((ldpc_enc->n_cb > N) || (ldpc_enc->n_cb <= K)) {
+ rte_bbdev_log(ERR, "Ncb (%u) is out of range K %d N %d", ldpc_enc->n_cb, K, N);
return -1;
}
+ if (!check_bit(op->ldpc_enc.op_flags,
+ RTE_BBDEV_LDPC_INTERLEAVER_BYPASS) &&
+ ((q_m == 0) || ((q_m > 2) && ((q_m % 2) == 1))
+ || (q_m > 8))) {
+ rte_bbdev_log(ERR, "Qm (%u) is out of range", ldpc_enc->q_m);
+ return -1;
+ }
+ if (ldpc_enc->code_block_mode == RTE_BBDEV_CODE_BLOCK) {
+ if (ldpc_enc->cb_params.e == 0) {
+ rte_bbdev_log(ERR, "E is null");
+ return -1;
+ }
+ if (q_m > 0) {
+ if (ldpc_enc->cb_params.e % q_m > 0) {
+ rte_bbdev_log(ERR, "E not multiple of qm %d", q_m);
+ return -1;
+ }
+ }
+ if ((ldpc_enc->z_c <= 11) && (ldpc_enc->cb_params.e > 3456)) {
+ rte_bbdev_log(ERR, "E too large for small block");
+ return -1;
+ }
+ if (ldpc_enc->input.length >
+ RTE_BBDEV_LDPC_MAX_CB_SIZE >> 3) {
+ rte_bbdev_log(ERR, "CB size (%u) is too big, max: %d",
+ ldpc_enc->input.length,
+ RTE_BBDEV_LDPC_MAX_CB_SIZE);
+ return -1;
+ }
+ if (K < (int) (ldpc_enc->input.length * 8 + ldpc_enc->n_filler) + crc24) {
+ rte_bbdev_log(ERR,
+ "K and F not matching input size %u %u %u",
+ K, ldpc_enc->n_filler,
+ ldpc_enc->input.length);
+ return -1;
+ }
+ } else {
+ if ((ldpc_enc->tb_params.c == 0) ||
+ (ldpc_enc->tb_params.ea == 0) ||
+ (ldpc_enc->tb_params.eb == 0)) {
+ rte_bbdev_log(ERR, "TB parameter is null");
+ return -1;
+ }
+ if (q_m > 0) {
+ if ((ldpc_enc->tb_params.ea % q_m > 0) ||
+ (ldpc_enc->tb_params.eb % q_m > 0)) {
+ rte_bbdev_log(ERR, "E not multiple of qm %d", q_m);
+ return -1;
+ }
+ }
+ if ((ldpc_enc->z_c <= 11) && (RTE_MAX(ldpc_enc->tb_params.ea,
+ ldpc_enc->tb_params.eb) > 3456)) {
+ rte_bbdev_log(ERR, "E too large for small block");
+ return -1;
+ }
+ }
return 0;
}
@@ -2009,24 +2077,30 @@ static inline int
validate_ldpc_dec_op(struct rte_bbdev_dec_op *op, struct acc_queue *q)
{
struct rte_bbdev_op_ldpc_dec *ldpc_dec = &op->ldpc_dec;
+ int K, N, q_m;
+ uint32_t min_harq_input;
if (!validate_op_required(q))
return 0;
- if (op->mempool == NULL) {
- rte_bbdev_log(ERR, "Invalid mempool pointer");
+ if (ldpc_dec->input.data == NULL) {
+ rte_bbdev_log(ERR, "Invalid input pointer");
+ return -1;
+ }
+ if (ldpc_dec->hard_output.data == NULL) {
+ rte_bbdev_log(ERR, "Invalid output pointer");
+ return -1;
+ }
+ if (ldpc_dec->input.length == 0) {
+ rte_bbdev_log(ERR, "input is null");
return -1;
}
if ((ldpc_dec->basegraph > 2) || (ldpc_dec->basegraph == 0)) {
- rte_bbdev_log(ERR,
- "BG (%u) is out of range 1 <= value <= 2",
- ldpc_dec->basegraph);
+ rte_bbdev_log(ERR, "BG (%u) is out of range 1 <= value <= 2", ldpc_dec->basegraph);
return -1;
}
if (ldpc_dec->iter_max == 0) {
- rte_bbdev_log(ERR,
- "iter_max (%u) is equal to 0",
- ldpc_dec->iter_max);
+ rte_bbdev_log(ERR, "iter_max (%u) is equal to 0", ldpc_dec->iter_max);
return -1;
}
if (ldpc_dec->rv_index > 3) {
@@ -2041,13 +2115,162 @@ validate_ldpc_dec_op(struct rte_bbdev_dec_op *op, struct acc_queue *q)
ldpc_dec->code_block_mode);
return -1;
}
- int K = (ldpc_dec->basegraph == 1 ? 22 : 10) * ldpc_dec->z_c;
- if (ldpc_dec->n_filler >= K) {
- rte_bbdev_log(ERR,
- "K and F are not compatible %u %u",
- K, ldpc_dec->n_filler);
+ /* Check Zc is valid value. */
+ if ((ldpc_dec->z_c > ACC_MAX_ZC) || (ldpc_dec->z_c < 2)) {
+ rte_bbdev_log(ERR, "Zc (%u) is out of range", ldpc_dec->z_c);
+ return -1;
+ }
+ if (ldpc_dec->z_c > 256) {
+ if ((ldpc_dec->z_c % 32) != 0) {
+ rte_bbdev_log(ERR, "Invalid Zc %d", ldpc_dec->z_c);
+ return -1;
+ }
+ } else if (ldpc_dec->z_c > 128) {
+ if ((ldpc_dec->z_c % 16) != 0) {
+ rte_bbdev_log(ERR, "Invalid Zc %d", ldpc_dec->z_c);
+ return -1;
+ }
+ } else if (ldpc_dec->z_c > 64) {
+ if ((ldpc_dec->z_c % 8) != 0) {
+ rte_bbdev_log(ERR, "Invalid Zc %d", ldpc_dec->z_c);
+ return -1;
+ }
+ } else if (ldpc_dec->z_c > 32) {
+ if ((ldpc_dec->z_c % 4) != 0) {
+ rte_bbdev_log(ERR, "Invalid Zc %d", ldpc_dec->z_c);
+ return -1;
+ }
+ } else if (ldpc_dec->z_c > 16) {
+ if ((ldpc_dec->z_c % 2) != 0) {
+ rte_bbdev_log(ERR, "Invalid Zc %d", ldpc_dec->z_c);
+ return -1;
+ }
+ }
+
+ K = (ldpc_dec->basegraph == 1 ? 22 : 10) * ldpc_dec->z_c;
+ N = (ldpc_dec->basegraph == 1 ? ACC_N_ZC_1 : ACC_N_ZC_2) * ldpc_dec->z_c;
+ q_m = ldpc_dec->q_m;
+
+ if (ldpc_dec->n_filler >= K - 2 * ldpc_dec->z_c) {
+ rte_bbdev_log(ERR, "K and F are not compatible %u %u", K, ldpc_dec->n_filler);
+ return -1;
+ }
+ if ((ldpc_dec->n_cb > N) || (ldpc_dec->n_cb <= K)) {
+ rte_bbdev_log(ERR, "Ncb (%u) is out of range K %d N %d", ldpc_dec->n_cb, K, N);
+ return -1;
+ }
+ if (((q_m == 0) || ((q_m > 2) && ((q_m % 2) == 1))
+ || (q_m > 8))) {
+ rte_bbdev_log(ERR, "Qm (%u) is out of range", ldpc_dec->q_m);
+ return -1;
+ }
+ if (ldpc_dec->code_block_mode == RTE_BBDEV_CODE_BLOCK) {
+ if (ldpc_dec->cb_params.e == 0) {
+ rte_bbdev_log(ERR, "E is null");
+ return -1;
+ }
+ if (ldpc_dec->cb_params.e % q_m > 0) {
+ rte_bbdev_log(ERR, "E not multiple of qm %d", q_m);
+ return -1;
+ }
+ if (ldpc_dec->cb_params.e > 512 * ldpc_dec->z_c) {
+ rte_bbdev_log(ERR, "E too high");
+ return -1;
+ }
+ } else {
+ if ((ldpc_dec->tb_params.c == 0) ||
+ (ldpc_dec->tb_params.ea == 0) ||
+ (ldpc_dec->tb_params.eb == 0)) {
+ rte_bbdev_log(ERR, "TB parameter is null");
+ return -1;
+ }
+ if ((ldpc_dec->tb_params.ea % q_m > 0) ||
+ (ldpc_dec->tb_params.eb % q_m > 0)) {
+ rte_bbdev_log(ERR, "E not multiple of qm %d", q_m);
+ return -1;
+ }
+ if ((ldpc_dec->tb_params.ea > 512 * ldpc_dec->z_c) ||
+ (ldpc_dec->tb_params.eb > 512 * ldpc_dec->z_c)) {
+ rte_bbdev_log(ERR, "E too high");
+ return -1;
+ }
+ }
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_DECODE_BYPASS)) {
+ rte_bbdev_log(ERR, "Avoid LDPC Decode bypass");
+ return -1;
+ }
+
+ /* Avoid HARQ compression for small block size */
+ if ((check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION)) && (K < 2048))
+ op->ldpc_dec.op_flags ^= RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION;
+
+ min_harq_input = check_bit(op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION) ? 256 : 64;
+ if (check_bit(op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE) &&
+ ldpc_dec->harq_combined_input.length <
+ min_harq_input) {
+ rte_bbdev_log(ERR, "HARQ input size is too small %d < %d",
+ ldpc_dec->harq_combined_input.length,
+ min_harq_input);
return -1;
}
+
+ /* Enforce in-range HARQ input size */
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE)) {
+ uint32_t max_harq_input = RTE_ALIGN_CEIL(ldpc_dec->n_cb - ldpc_dec->n_filler, 64);
+
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION))
+ max_harq_input = max_harq_input * 3 / 4;
+
+ if (ldpc_dec->harq_combined_input.length > max_harq_input) {
+ rte_bbdev_log(ERR,
+ "HARQ input size out of range %d > %d, Ncb %d F %d K %d N %d",
+ ldpc_dec->harq_combined_input.length,
+ max_harq_input, ldpc_dec->n_cb,
+ ldpc_dec->n_filler, K, N);
+ /* Fallback to flush HARQ combine */
+ ldpc_dec->harq_combined_input.length = 0;
+
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE))
+ op->ldpc_dec.op_flags ^= RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE;
+ }
+ }
+
+#ifdef ACC100_EXT_MEM
+ /* Enforce in-range HARQ offset */
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE)) {
+ if ((op->ldpc_dec.harq_combined_input.offset >> 10) >= q->d->ddr_size) {
+ rte_bbdev_log(ERR,
+ "HARQin offset out of range %d > %d",
+ op->ldpc_dec.harq_combined_input.offset,
+ q->d->ddr_size);
+ return -1;
+ }
+ if ((op->ldpc_dec.harq_combined_input.offset & 0x3FF) > 0) {
+ rte_bbdev_log(ERR,
+ "HARQin offset not aligned on 1kB %d",
+ op->ldpc_dec.harq_combined_input.offset);
+ return -1;
+ }
+ }
+ if (check_bit(op->ldpc_dec.op_flags, RTE_BBDEV_LDPC_HQ_COMBINE_OUT_ENABLE)) {
+ if ((op->ldpc_dec.harq_combined_output.offset >> 10) >= q->d->ddr_size) {
+ rte_bbdev_log(ERR,
+ "HARQout offset out of range %d > %d",
+ op->ldpc_dec.harq_combined_output.offset,
+ q->d->ddr_size);
+ return -1;
+ }
+ if ((op->ldpc_dec.harq_combined_output.offset & 0x3FF) > 0) {
+ rte_bbdev_log(ERR,
+ "HARQout offset not aligned on 1kB %d",
+ op->ldpc_dec.harq_combined_output.offset);
+ return -1;
+ }
+ }
+#endif
+
return 0;
}
#endif
@@ -2685,7 +2908,7 @@ harq_loopback(struct acc_queue *q, struct rte_bbdev_dec_op *op,
memset(fcw, 0, sizeof(struct acc_fcw_ld));
fcw->FCWversion = ACC_FCW_VER;
fcw->qm = 2;
- fcw->Zc = 384;
+ fcw->Zc = ACC_MAX_ZC;
if (harq_in_length < 16 * ACC_N_ZC_1)
fcw->Zc = 16;
fcw->ncb = fcw->Zc * ACC_N_ZC_1;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 21/30] baseband/acc100: implement configurable queue depth
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (19 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 20/30] baseband/acc100: update validate LDPC enc/dec Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 22/30] baseband/acc100: add queue stop operation Hernan Vargas
` (8 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Implement new feature to make queue depth configurable based on decode
or encode mode.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 12 +++++++++---
1 file changed, 9 insertions(+), 3 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 94863d7afd..a734be8553 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -776,9 +776,15 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
q->qgrp_id = (q_idx >> ACC100_GRP_ID_SHIFT) & 0xF;
q->vf_id = (q_idx >> ACC100_VF_ID_SHIFT) & 0x3F;
q->aq_id = q_idx & 0xF;
- q->aq_depth = (conf->op_type == RTE_BBDEV_OP_TURBO_DEC) ?
- (1 << d->acc_conf.q_ul_4g.aq_depth_log2) :
- (1 << d->acc_conf.q_dl_4g.aq_depth_log2);
+ q->aq_depth = 0;
+ if (conf->op_type == RTE_BBDEV_OP_TURBO_DEC)
+ q->aq_depth = (1 << d->acc_conf.q_ul_4g.aq_depth_log2);
+ else if (conf->op_type == RTE_BBDEV_OP_TURBO_ENC)
+ q->aq_depth = (1 << d->acc_conf.q_dl_4g.aq_depth_log2);
+ else if (conf->op_type == RTE_BBDEV_OP_LDPC_DEC)
+ q->aq_depth = (1 << d->acc_conf.q_ul_5g.aq_depth_log2);
+ else if (conf->op_type == RTE_BBDEV_OP_LDPC_ENC)
+ q->aq_depth = (1 << d->acc_conf.q_dl_5g.aq_depth_log2);
q->mmio_reg_enqueue = RTE_PTR_ADD(d->mmio_base,
queue_offset(d->pf_device,
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 22/30] baseband/acc100: add queue stop operation
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (20 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 21/30] baseband/acc100: implement configurable queue depth Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 23/30] baseband/acc100: update uplink CB input length Hernan Vargas
` (7 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Implement new feature to stop queue operation.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 60 +++++++++++++++++++++++++++
1 file changed, 60 insertions(+)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index a734be8553..7d392bfba7 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -799,6 +799,65 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
return 0;
}
+static inline void
+acc100_print_op(struct rte_bbdev_dec_op *op, enum rte_bbdev_op_type op_type,
+ uint16_t index)
+{
+ if (op == NULL)
+ return;
+ if (op_type == RTE_BBDEV_OP_LDPC_DEC)
+ rte_bbdev_log(DEBUG,
+ " Op 5GUL %d %d %d %d %d %d %d %d %d %d %d %d",
+ index,
+ op->ldpc_dec.basegraph, op->ldpc_dec.z_c,
+ op->ldpc_dec.n_cb, op->ldpc_dec.q_m,
+ op->ldpc_dec.n_filler, op->ldpc_dec.cb_params.e,
+ op->ldpc_dec.op_flags, op->ldpc_dec.rv_index,
+ op->ldpc_dec.iter_max, op->ldpc_dec.iter_count,
+ op->ldpc_dec.harq_combined_input.length
+ );
+ else if (op_type == RTE_BBDEV_OP_LDPC_ENC) {
+ struct rte_bbdev_enc_op *op_dl = (struct rte_bbdev_enc_op *) op;
+ rte_bbdev_log(DEBUG,
+ " Op 5GDL %d %d %d %d %d %d %d %d %d",
+ index,
+ op_dl->ldpc_enc.basegraph, op_dl->ldpc_enc.z_c,
+ op_dl->ldpc_enc.n_cb, op_dl->ldpc_enc.q_m,
+ op_dl->ldpc_enc.n_filler, op_dl->ldpc_enc.cb_params.e,
+ op_dl->ldpc_enc.op_flags, op_dl->ldpc_enc.rv_index
+ );
+ }
+}
+
+static int
+acc100_queue_stop(struct rte_bbdev *dev, uint16_t queue_id)
+{
+ struct acc_queue *q;
+ struct rte_bbdev_dec_op *op;
+ uint16_t i;
+
+ q = dev->data->queues[queue_id].queue_private;
+ rte_bbdev_log(INFO, "Queue Stop %d H/T/D %d %d %x OpType %d",
+ queue_id, q->sw_ring_head, q->sw_ring_tail,
+ q->sw_ring_depth, q->op_type);
+ for (i = 0; i < q->sw_ring_depth; ++i) {
+ op = (q->ring_addr + i)->req.op_addr;
+ acc100_print_op(op, q->op_type, i);
+ }
+ /* ignore all operations in flight and clear counters */
+ q->sw_ring_tail = q->sw_ring_head;
+ q->aq_enqueued = 0;
+ q->aq_dequeued = 0;
+ dev->data->queues[queue_id].queue_stats.enqueued_count = 0;
+ dev->data->queues[queue_id].queue_stats.dequeued_count = 0;
+ dev->data->queues[queue_id].queue_stats.enqueue_err_count = 0;
+ dev->data->queues[queue_id].queue_stats.dequeue_err_count = 0;
+ dev->data->queues[queue_id].queue_stats.enqueue_warn_count = 0;
+ dev->data->queues[queue_id].queue_stats.dequeue_warn_count = 0;
+
+ return 0;
+}
+
/* Release ACC100 queue */
static int
acc100_queue_release(struct rte_bbdev *dev, uint16_t q_id)
@@ -991,6 +1050,7 @@ static const struct rte_bbdev_ops acc100_bbdev_ops = {
.info_get = acc100_dev_info_get,
.queue_setup = acc100_queue_setup,
.queue_release = acc100_queue_release,
+ .queue_stop = acc100_queue_stop,
.queue_intr_enable = acc100_queue_intr_enable,
.queue_intr_disable = acc100_queue_intr_disable
};
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 23/30] baseband/acc100: update uplink CB input length
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (21 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 22/30] baseband/acc100: add queue stop operation Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 24/30] baseband/acc100: update log messages Hernan Vargas
` (6 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Use the FCW E parameter for rate matching as the code block input
length.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 7d392bfba7..c94fe14439 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -1713,7 +1713,7 @@ acc100_dma_desc_ld_fill(struct rte_bbdev_dec_op *op,
crc24_overlap = 24;
/* Compute some LDPC BG lengths */
- input_length = dec->cb_params.e;
+ input_length = fcw->rm_e;
if (check_bit(op->ldpc_dec.op_flags,
RTE_BBDEV_LDPC_LLR_COMPRESSION))
input_length = (input_length * 3 + 3) / 4;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 24/30] baseband/acc100: update log messages
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (22 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 23/30] baseband/acc100: update uplink CB input length Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 25/30] baseband/acc100: store FCW from first CB descriptor Hernan Vargas
` (5 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Add extra values for some log messages. No functional impact.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index c94fe14439..e00a636d9f 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -979,6 +979,7 @@ acc100_dev_info_get(struct rte_bbdev *dev,
/* Read and save the populated config from ACC100 registers */
fetch_acc100_config(dev);
+ /* Check the status of device */
dev_info->device_status = RTE_BBDEV_DEV_NOT_SUPPORTED;
/* Expose number of queues */
@@ -2653,7 +2654,7 @@ enqueue_ldpc_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
{
#ifndef RTE_LIBRTE_BBDEV_SKIP_VALIDATE
if (validate_ldpc_enc_op(op, q) == -1) {
- rte_bbdev_log(ERR, "LDPC encoder validation failed");
+ rte_bbdev_log(ERR, "LDPC encoder validation rejected");
return -EINVAL;
}
#endif
@@ -3876,8 +3877,9 @@ dequeue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
atom_desc.atom_hdr = __atomic_load_n((uint64_t *)desc,
__ATOMIC_RELAXED);
rsp.val = atom_desc.rsp.val;
- rte_bbdev_log_debug("Resp. desc %p: %x", desc,
- rsp.val);
+ rte_bbdev_log_debug("Resp. desc %p: %x r %d c %d\n",
+ desc, rsp.val,
+ cb_idx, cbs_in_tb);
op->status |= ((rsp.input_err)
? (1 << RTE_BBDEV_DATA_ERROR) : 0);
@@ -3973,6 +3975,7 @@ dequeue_ldpc_dec_one_op_cb(struct rte_bbdev_queue_data *q_data,
return -1;
rsp.val = atom_desc.rsp.val;
+ rte_bbdev_log_debug("Resp. desc %p: %x\n", desc, rsp.val);
/* Dequeue */
op = desc->req.op_addr;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 25/30] baseband/acc100: store FCW from first CB descriptor
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (23 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 24/30] baseband/acc100: update log messages Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 26/30] baseband/acc100: update device info Hernan Vargas
` (4 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Store the descriptor from the first code block from a transport block.
Copy the LDPC FCW from the first descriptor into the rest of the CBs in
that TB.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index e00a636d9f..ed5ee63950 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -3188,6 +3188,7 @@ enqueue_ldpc_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
uint16_t total_enqueued_cbs, uint8_t cbs_in_tb)
{
union acc_dma_desc *desc = NULL;
+ union acc_dma_desc *desc_first = NULL;
int ret;
uint8_t r, c;
uint32_t in_offset, h_out_offset,
@@ -3207,6 +3208,7 @@ enqueue_ldpc_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
desc_idx = acc_desc_idx(q, total_enqueued_cbs);
desc = q->ring_addr + desc_idx;
+ desc_first = desc;
fcw_offset = (desc_idx << 8) + ACC_DESC_FCW_OFFSET;
harq_layout = q->d->harq_layout;
q->d->fcw_ld_fill(op, &desc->req.fcw_ld, harq_layout);
@@ -3229,6 +3231,7 @@ enqueue_ldpc_dec_one_op_tb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
desc = acc_desc(q, total_enqueued_cbs);
desc->req.data_ptrs[0].address = q->ring_addr_iova + fcw_offset;
desc->req.data_ptrs[0].blen = ACC_FCW_LD_BLEN;
+ rte_memcpy(&desc->req.fcw_ld, &desc_first->req.fcw_ld, ACC_FCW_LD_BLEN);
ret = acc100_dma_desc_ld_fill(op, &desc->req, &input,
h_output, &in_offset, &h_out_offset,
&h_out_length,
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 26/30] baseband/acc100: update device info
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (24 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 25/30] baseband/acc100: store FCW from first CB descriptor Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 27/30] baseband/acc100: add ring companion address Hernan Vargas
` (3 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Remove unused capabilities, use dummy operation as start count for
number of queues.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index ed5ee63950..45783a7a78 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -885,7 +885,6 @@ acc100_dev_info_get(struct rte_bbdev *dev,
{
struct acc_device *d = dev->data->dev_private;
int i;
-
static const struct rte_bbdev_op_cap bbdev_capabilities[] = {
{
.type = RTE_BBDEV_OP_TURBO_DEC,
@@ -897,7 +896,6 @@ acc100_dev_info_get(struct rte_bbdev *dev,
RTE_BBDEV_TURBO_EARLY_TERMINATION |
RTE_BBDEV_TURBO_DEC_INTERRUPTS |
RTE_BBDEV_TURBO_NEG_LLR_1_BIT_IN |
- RTE_BBDEV_TURBO_MAP_DEC |
RTE_BBDEV_TURBO_DEC_TB_CRC_24B_KEEP |
RTE_BBDEV_TURBO_DEC_CRC_24B_DROP |
RTE_BBDEV_TURBO_DEC_SCATTER_GATHER,
@@ -992,12 +990,13 @@ acc100_dev_info_get(struct rte_bbdev *dev,
d->acc_conf.q_ul_5g.num_qgroups;
dev_info->num_queues[RTE_BBDEV_OP_LDPC_ENC] = d->acc_conf.q_dl_5g.num_aqs_per_groups *
d->acc_conf.q_dl_5g.num_qgroups;
+ dev_info->num_queues[RTE_BBDEV_OP_FFT] = 0;
dev_info->queue_priority[RTE_BBDEV_OP_TURBO_DEC] = d->acc_conf.q_ul_4g.num_qgroups;
dev_info->queue_priority[RTE_BBDEV_OP_TURBO_ENC] = d->acc_conf.q_dl_4g.num_qgroups;
dev_info->queue_priority[RTE_BBDEV_OP_LDPC_DEC] = d->acc_conf.q_ul_5g.num_qgroups;
dev_info->queue_priority[RTE_BBDEV_OP_LDPC_ENC] = d->acc_conf.q_dl_5g.num_qgroups;
dev_info->max_num_queues = 0;
- for (i = RTE_BBDEV_OP_TURBO_DEC; i <= RTE_BBDEV_OP_LDPC_ENC; i++)
+ for (i = RTE_BBDEV_OP_NONE; i <= RTE_BBDEV_OP_LDPC_ENC; i++)
dev_info->max_num_queues += dev_info->num_queues[i];
dev_info->queue_size_lim = ACC_MAX_QUEUE_DEPTH;
dev_info->hardware_accelerated = true;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 27/30] baseband/acc100: add ring companion address
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (25 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 26/30] baseband/acc100: update device info Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 28/30] baseband/acc100: add workaround for deRM corner cases Hernan Vargas
` (2 subsequent siblings)
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Store the virtual address of companion ring as part of queue
information. Use this address to calculate the op address.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/rte_acc100_pmd.c | 179 +++++++++++++++++---------
1 file changed, 116 insertions(+), 63 deletions(-)
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 45783a7a78..87756839ef 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -676,6 +676,7 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
struct acc_device *d = dev->data->dev_private;
struct acc_queue *q;
int16_t q_idx;
+ int ret;
if (d == NULL) {
rte_bbdev_log(ERR, "Undefined device");
@@ -734,8 +735,8 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
RTE_CACHE_LINE_SIZE, conf->socket);
if (q->lb_in == NULL) {
rte_bbdev_log(ERR, "Failed to allocate lb_in memory");
- rte_free(q);
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto free_q;
}
q->lb_in_addr_iova = rte_malloc_virt2iova(q->lb_in);
q->lb_out = rte_zmalloc_socket(dev->device->driver->name,
@@ -743,11 +744,18 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
RTE_CACHE_LINE_SIZE, conf->socket);
if (q->lb_out == NULL) {
rte_bbdev_log(ERR, "Failed to allocate lb_out memory");
- rte_free(q->lb_in);
- rte_free(q);
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto free_lb_in;
}
q->lb_out_addr_iova = rte_malloc_virt2iova(q->lb_out);
+ q->companion_ring_addr = rte_zmalloc_socket(dev->device->driver->name,
+ d->sw_ring_max_depth * sizeof(*q->companion_ring_addr),
+ RTE_CACHE_LINE_SIZE, conf->socket);
+ if (q->companion_ring_addr == NULL) {
+ rte_bbdev_log(ERR, "Failed to allocate companion_ring memory");
+ ret = -ENOMEM;
+ goto free_lb_out;
+ }
/*
* Software queue ring wraps synchronously with the HW when it reaches
@@ -767,10 +775,8 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
q_idx = acc100_find_free_queue_idx(dev, conf);
if (q_idx == -1) {
- rte_free(q->lb_in);
- rte_free(q->lb_out);
- rte_free(q);
- return -1;
+ ret = -EINVAL;
+ goto free_companion_ring_addr;
}
q->qgrp_id = (q_idx >> ACC100_GRP_ID_SHIFT) & 0xF;
@@ -797,6 +803,21 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
dev->data->queues[queue_id].queue_private = q;
return 0;
+
+free_companion_ring_addr:
+ rte_free(q->companion_ring_addr);
+ q->companion_ring_addr = NULL;
+free_lb_out:
+ rte_free(q->lb_out);
+ q->lb_out = NULL;
+free_lb_in:
+ rte_free(q->lb_in);
+ q->lb_in = NULL;
+free_q:
+ rte_free(q);
+ q = NULL;
+
+ return ret;
}
static inline void
@@ -869,6 +890,7 @@ acc100_queue_release(struct rte_bbdev *dev, uint16_t q_id)
/* Mark the Queue as un-assigned */
d->q_assigned_bit_map[q->qgrp_id] &= (0xFFFFFFFFFFFFFFFF -
(uint64_t) (1 << q->aq_id));
+ rte_free(q->companion_ring_addr);
rte_free(q->lb_in);
rte_free(q->lb_out);
rte_free(q);
@@ -2395,7 +2417,7 @@ enqueue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op *op,
/* Enqueue one encode operations for ACC100 device in CB mode */
static inline int
enqueue_ldpc_enc_n_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ops,
- uint16_t total_enqueued_cbs, int16_t num)
+ uint16_t total_enqueued_descs, int16_t num)
{
union acc_dma_desc *desc = NULL;
uint32_t out_length;
@@ -2412,7 +2434,7 @@ enqueue_ldpc_enc_n_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ops,
}
#endif
- desc = acc_desc(q, total_enqueued_cbs);
+ desc = acc_desc(q, total_enqueued_descs);
acc_fcw_le_fill(ops[0], &desc->req.fcw_le, num, 0);
/** This could be done at polling */
@@ -2442,6 +2464,11 @@ enqueue_ldpc_enc_n_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ops,
}
desc->req.op_addr = ops[0];
+ /* Keep track of pointers even when multiplexed in single descriptor */
+ struct acc_ptrs *context_ptrs = q->companion_ring_addr
+ + acc_desc_idx(q, total_enqueued_descs);
+ for (i = 0; i < num; i++)
+ context_ptrs->ptr[i].op_addr = ops[i];
#ifdef RTE_LIBRTE_BBDEV_DEBUG
rte_memdump(stderr, "FCW", &desc->req.fcw_le,
@@ -3790,7 +3817,8 @@ acc100_enqueue_ldpc_dec(struct rte_bbdev_queue_data *q_data,
/* Dequeue one encode operations from ACC100 device in CB mode */
static inline int
dequeue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
- uint16_t total_dequeued_cbs, uint32_t *aq_dequeued)
+ uint16_t *dequeued_ops, uint32_t *aq_dequeued,
+ uint16_t *dequeued_descs)
{
union acc_dma_desc *desc, atom_desc;
union acc_dma_rsp_desc rsp;
@@ -3798,7 +3826,7 @@ dequeue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
int i;
uint16_t desc_idx;
- desc_idx = acc_desc_idx_tail(q, total_dequeued_cbs);
+ desc_idx = acc_desc_idx_tail(q, *dequeued_descs);
desc = q->ring_addr + desc_idx;
atom_desc.atom_hdr = __atomic_load_n((uint64_t *)desc,
__ATOMIC_RELAXED);
@@ -3808,7 +3836,7 @@ dequeue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
return -1;
rsp.val = atom_desc.rsp.val;
- rte_bbdev_log_debug("Resp. desc %p: %x", desc, rsp.val);
+ rte_bbdev_log_debug("Resp. desc %p: %x num %d\n", desc, rsp.val, desc->req.numCBs);
/* Dequeue */
op = desc->req.op_addr;
@@ -3828,27 +3856,35 @@ dequeue_enc_one_op_cb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
desc->rsp.add_info_0 = 0; /*Reserved bits */
desc->rsp.add_info_1 = 0; /*Reserved bits */
- /* Flag that the muxing cause loss of opaque data */
- op->opaque_data = (void *)-1;
- for (i = 0 ; i < desc->req.numCBs; i++)
- ref_op[i] = op;
+ ref_op[0] = op;
+ struct acc_ptrs *context_ptrs = q->companion_ring_addr + desc_idx;
+ for (i = 1 ; i < desc->req.numCBs; i++)
+ ref_op[i] = context_ptrs->ptr[i].op_addr;
+
+ /* One CB (op) was successfully dequeued */
+ /* One op was successfully dequeued */
+ (*dequeued_descs)++;
+ *dequeued_ops += desc->req.numCBs;
/* One CB (op) was successfully dequeued */
return desc->req.numCBs;
}
-/* Dequeue one encode operations from ACC100 device in TB mode */
+/* Dequeue one LDPC encode operations from ACC100 device in TB mode
+ * That operation may cover multiple descriptors
+ */
static inline int
dequeue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
- uint16_t total_dequeued_cbs, uint32_t *aq_dequeued)
+ uint16_t *dequeued_ops, uint32_t *aq_dequeued,
+ uint16_t *dequeued_descs)
{
union acc_dma_desc *desc, *last_desc, atom_desc;
union acc_dma_rsp_desc rsp;
struct rte_bbdev_enc_op *op;
uint8_t i = 0;
- uint16_t current_dequeued_cbs = 0, cbs_in_tb;
+ uint16_t current_dequeued_descs = 0, descs_in_tb;
- desc = acc_desc_tail(q, total_dequeued_cbs);
+ desc = acc_desc_tail(q, *dequeued_descs);
atom_desc.atom_hdr = __atomic_load_n((uint64_t *)desc,
__ATOMIC_RELAXED);
@@ -3857,9 +3893,9 @@ dequeue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
return -1;
/* Get number of CBs in dequeued TB */
- cbs_in_tb = desc->req.cbs_in_tb;
+ descs_in_tb = desc->req.cbs_in_tb;
/* Get last CB */
- last_desc = acc_desc_tail(q, total_dequeued_cbs + cbs_in_tb - 1);
+ last_desc = acc_desc_tail(q, *dequeued_descs + descs_in_tb - 1);
/* Check if last CB in TB is ready to dequeue (and thus
* the whole TB) - checking sdone bit. If not return.
*/
@@ -3874,14 +3910,13 @@ dequeue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
/* Clearing status, it will be set based on response */
op->status = 0;
- while (i < cbs_in_tb) {
- desc = acc_desc_tail(q, total_dequeued_cbs);
+ while (i < descs_in_tb) {
+ desc = acc_desc_tail(q, *dequeued_descs);
atom_desc.atom_hdr = __atomic_load_n((uint64_t *)desc,
__ATOMIC_RELAXED);
rsp.val = atom_desc.rsp.val;
- rte_bbdev_log_debug("Resp. desc %p: %x r %d c %d\n",
- desc, rsp.val,
- cb_idx, cbs_in_tb);
+ rte_bbdev_log_debug("Resp. desc %p: %x descs %d cbs %d\n",
+ desc, rsp.val, descs_in_tb, desc->req.numCBs);
op->status |= ((rsp.input_err)
? (1 << RTE_BBDEV_DATA_ERROR) : 0);
@@ -3895,14 +3930,15 @@ dequeue_enc_one_op_tb(struct acc_queue *q, struct rte_bbdev_enc_op **ref_op,
desc->rsp.val = ACC_DMA_DESC_TYPE;
desc->rsp.add_info_0 = 0;
desc->rsp.add_info_1 = 0;
- total_dequeued_cbs++;
- current_dequeued_cbs++;
+ (*dequeued_descs)++;
+ current_dequeued_descs++;
i++;
}
*ref_op = op;
- return current_dequeued_cbs;
+ (*dequeued_ops)++;
+ return current_dequeued_descs;
}
/* Dequeue one decode operation from ACC100 device in CB mode */
@@ -4092,12 +4128,12 @@ acc100_dequeue_enc(struct rte_bbdev_queue_data *q_data,
struct rte_bbdev_enc_op **ops, uint16_t num)
{
struct acc_queue *q = q_data->queue_private;
- uint16_t dequeue_num;
uint32_t avail = acc_ring_avail_deq(q);
uint32_t aq_dequeued = 0;
- uint16_t i, dequeued_cbs = 0;
+ uint16_t i, dequeued_ops = 0, dequeued_descs = 0;
+ int ret, cbm;
struct rte_bbdev_enc_op *op;
- int ret;
+
if (avail == 0)
return 0;
#ifdef RTE_LIBRTE_BBDEV_DEBUG
@@ -4106,30 +4142,36 @@ acc100_dequeue_enc(struct rte_bbdev_queue_data *q_data,
return 0;
}
#endif
+ op = (q->ring_addr + (q->sw_ring_tail &
+ q->sw_ring_wrap_mask))->req.op_addr;
+ if (unlikely(ops == NULL || op == NULL))
+ return 0;
+ cbm = op->turbo_enc.code_block_mode;
- dequeue_num = (avail < num) ? avail : num;
-
- for (i = 0; i < dequeue_num; ++i) {
- op = acc_op_tail(q, dequeued_cbs);
- if (op->turbo_enc.code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK)
- ret = dequeue_enc_one_op_tb(q, &ops[i], dequeued_cbs,
- &aq_dequeued);
+ for (i = 0; i < num; i++) {
+ if (cbm == RTE_BBDEV_TRANSPORT_BLOCK)
+ ret = dequeue_enc_one_op_tb(q, &ops[dequeued_ops],
+ &dequeued_ops, &aq_dequeued,
+ &dequeued_descs);
else
- ret = dequeue_enc_one_op_cb(q, &ops[i], dequeued_cbs,
- &aq_dequeued);
+ ret = dequeue_enc_one_op_cb(q, &ops[dequeued_ops],
+ &dequeued_ops, &aq_dequeued,
+ &dequeued_descs);
if (ret < 0)
break;
- dequeued_cbs += ret;
+
+ if (dequeued_ops >= num)
+ break;
}
q->aq_dequeued += aq_dequeued;
- q->sw_ring_tail += dequeued_cbs;
+ q->sw_ring_tail += dequeued_descs;
/* Update enqueue stats */
- q_data->queue_stats.dequeued_count += i;
+ q_data->queue_stats.dequeued_count += dequeued_ops;
- return i;
+ return dequeued_ops;
}
/* Dequeue LDPC encode operations from ACC100 device. */
@@ -4140,24 +4182,36 @@ acc100_dequeue_ldpc_enc(struct rte_bbdev_queue_data *q_data,
struct acc_queue *q = q_data->queue_private;
uint32_t avail = acc_ring_avail_deq(q);
uint32_t aq_dequeued = 0;
- uint16_t dequeue_num, i, dequeued_cbs = 0, dequeued_descs = 0;
- int ret;
+ uint16_t i, dequeued_ops = 0, dequeued_descs = 0;
+ int ret, cbm;
+ struct rte_bbdev_enc_op *op;
+ union acc_dma_desc *desc;
+ if (q == NULL)
+ return 0;
#ifdef RTE_LIBRTE_BBDEV_DEBUG
- if (unlikely(ops == 0 && q == NULL))
+ if (unlikely(ops == 0))
return 0;
#endif
-
- dequeue_num = RTE_MIN(avail, num);
-
- for (i = 0; i < dequeue_num; i++) {
- ret = dequeue_enc_one_op_cb(q, &ops[dequeued_cbs],
- dequeued_descs, &aq_dequeued);
+ desc = q->ring_addr + (q->sw_ring_tail & q->sw_ring_wrap_mask);
+ if (unlikely(desc == NULL))
+ return 0;
+ op = desc->req.op_addr;
+ if (unlikely(ops == NULL || op == NULL))
+ return 0;
+ cbm = op->ldpc_enc.code_block_mode;
+ for (i = 0; i < avail; i++) {
+ if (cbm == RTE_BBDEV_TRANSPORT_BLOCK)
+ ret = dequeue_enc_one_op_tb(q, &ops[dequeued_ops],
+ &dequeued_ops, &aq_dequeued,
+ &dequeued_descs);
+ else
+ ret = dequeue_enc_one_op_cb(q, &ops[dequeued_ops],
+ &dequeued_ops, &aq_dequeued,
+ &dequeued_descs);
if (ret < 0)
break;
- dequeued_cbs += ret;
- dequeued_descs++;
- if (dequeued_cbs >= num)
+ if (dequeued_ops >= num)
break;
}
@@ -4165,12 +4219,11 @@ acc100_dequeue_ldpc_enc(struct rte_bbdev_queue_data *q_data,
q->sw_ring_tail += dequeued_descs;
/* Update enqueue stats */
- q_data->queue_stats.dequeued_count += dequeued_cbs;
+ q_data->queue_stats.dequeued_count += dequeued_ops;
- return dequeued_cbs;
+ return dequeued_ops;
}
-
/* Dequeue decode operations from ACC100 device. */
static uint16_t
acc100_dequeue_dec(struct rte_bbdev_queue_data *q_data,
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 28/30] baseband/acc100: add workaround for deRM corner cases
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (26 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 27/30] baseband/acc100: add ring companion address Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 29/30] baseband/acc100: configure PMON control registers Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 30/30] baseband/acc100: update guide docs Hernan Vargas
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Add function to support de-ratematch pre-processing for SW corner cases.
Some specific 5GUL FEC corner cases may cause unintended back pressure
and in some cases potential stability issue on the ACC100.
To be able to avoid completly such potential issue, the PMD can preempt
such code block configuration so that to process the first level deRM
in SW using the SDK libraries prior to running the rest of the FEC
decoding in HW using an amended code block configuration.
In case meson build system doesn't find such SDK libraries, the fall
method is to run in HW as is with a warning.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
drivers/baseband/acc/acc_common.h | 8 ++
drivers/baseband/acc/meson.build | 21 +++++
drivers/baseband/acc/rte_acc100_pmd.c | 108 +++++++++++++++++++++++++-
3 files changed, 134 insertions(+), 3 deletions(-)
diff --git a/drivers/baseband/acc/acc_common.h b/drivers/baseband/acc/acc_common.h
index eae7eab4e9..5e8972b40a 100644
--- a/drivers/baseband/acc/acc_common.h
+++ b/drivers/baseband/acc/acc_common.h
@@ -123,6 +123,14 @@
#define ACC_HARQ_ALIGN_64B 64
#define ACC_MAX_ZC 384
+/* De-ratematch code rate limitation when padding is required */
+#define ACC_LIM_03 2 /* 0.03 */
+#define ACC_LIM_09 6 /* 0.09 */
+#define ACC_LIM_14 9 /* 0.14 */
+#define ACC_LIM_21 14 /* 0.21 */
+#define ACC_LIM_31 20 /* 0.31 */
+#define ACC_MAX_E (128 * 1024 - 2)
+
/* Helper macro for logging */
#define rte_acc_log(level, fmt, ...) \
rte_log(RTE_LOG_ ## level, RTE_LOG_NOTICE, fmt "\n", \
diff --git a/drivers/baseband/acc/meson.build b/drivers/baseband/acc/meson.build
index 77c393b533..a5fc4fed01 100644
--- a/drivers/baseband/acc/meson.build
+++ b/drivers/baseband/acc/meson.build
@@ -1,6 +1,27 @@
# SPDX-License-Identifier: BSD-3-Clause
# Copyright(c) 2020 Intel Corporation
+# Check for FlexRAN SDK libraries
+dep_dec5g = dependency('flexran_sdk_ldpc_decoder_5gnr', required: false)
+
+if dep_dec5g.found()
+ ext_deps += cc.find_library('libstdc++', required: true)
+ ext_deps += cc.find_library('libirc', required: true)
+ ext_deps += cc.find_library('libimf', required: true)
+ ext_deps += cc.find_library('libipps', required: true)
+ ext_deps += cc.find_library('libsvml', required: true)
+ ext_deps += dep_dec5g
+ ext_deps += dependency('flexran_sdk_ldpc_encoder_5gnr', required: true)
+ ext_deps += dependency('flexran_sdk_LDPC_ratematch_5gnr', required: true)
+ ext_deps += dependency('flexran_sdk_rate_dematching_5gnr', required: true)
+ ext_deps += dependency('flexran_sdk_turbo', required: true)
+ ext_deps += dependency('flexran_sdk_crc', required: true)
+ ext_deps += dependency('flexran_sdk_rate_matching', required: true)
+ ext_deps += dependency('flexran_sdk_common', required: true)
+ cflags += ['-DRTE_BBDEV_SDK_AVX2']
+ cflags += ['-DRTE_BBDEV_SDK_AVX512']
+endif
+
deps += ['bbdev', 'bus_pci']
sources = files('rte_acc100_pmd.c', 'rte_acc200_pmd.c')
diff --git a/drivers/baseband/acc/rte_acc100_pmd.c b/drivers/baseband/acc/rte_acc100_pmd.c
index 87756839ef..0c2454995a 100644
--- a/drivers/baseband/acc/rte_acc100_pmd.c
+++ b/drivers/baseband/acc/rte_acc100_pmd.c
@@ -25,6 +25,10 @@
#include "acc101_pmd.h"
#include "acc200_cfg.h"
+#ifdef RTE_BBDEV_SDK_AVX512
+#include <phy_rate_dematching_5gnr.h>
+#endif
+
#ifdef RTE_LIBRTE_BBDEV_DEBUG
RTE_LOG_REGISTER_DEFAULT(acc100_logtype, DEBUG);
#else
@@ -756,6 +760,14 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
ret = -ENOMEM;
goto free_lb_out;
}
+ q->derm_buffer = rte_zmalloc_socket(dev->device->driver->name,
+ RTE_BBDEV_TURBO_MAX_CB_SIZE * 10,
+ RTE_CACHE_LINE_SIZE, conf->socket);
+ if (q->derm_buffer == NULL) {
+ rte_bbdev_log(ERR, "Failed to allocate derm_buffer memory");
+ ret = -ENOMEM;
+ goto free_companion_ring_addr;
+ }
/*
* Software queue ring wraps synchronously with the HW when it reaches
@@ -776,7 +788,7 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
q_idx = acc100_find_free_queue_idx(dev, conf);
if (q_idx == -1) {
ret = -EINVAL;
- goto free_companion_ring_addr;
+ goto free_derm_buffer;
}
q->qgrp_id = (q_idx >> ACC100_GRP_ID_SHIFT) & 0xF;
@@ -804,6 +816,9 @@ acc100_queue_setup(struct rte_bbdev *dev, uint16_t queue_id,
dev->data->queues[queue_id].queue_private = q;
return 0;
+free_derm_buffer:
+ rte_free(q->derm_buffer);
+ q->derm_buffer = NULL;
free_companion_ring_addr:
rte_free(q->companion_ring_addr);
q->companion_ring_addr = NULL;
@@ -890,6 +905,7 @@ acc100_queue_release(struct rte_bbdev *dev, uint16_t q_id)
/* Mark the Queue as un-assigned */
d->q_assigned_bit_map[q->qgrp_id] &= (0xFFFFFFFFFFFFFFFF -
(uint64_t) (1 << q->aq_id));
+ rte_free(q->derm_buffer);
rte_free(q->companion_ring_addr);
rte_free(q->lb_in);
rte_free(q->lb_out);
@@ -3110,10 +3126,44 @@ harq_loopback(struct acc_queue *q, struct rte_bbdev_dec_op *op,
return 1;
}
+/** Assess whether a work around is required for the deRM corner cases */
+static inline bool
+derm_workaround_required(struct rte_bbdev_op_ldpc_dec *ldpc_dec, struct acc_queue *q)
+{
+ if (!is_acc100(q))
+ return false;
+ int32_t e = ldpc_dec->cb_params.e;
+ int q_m = ldpc_dec->q_m;
+ int z_c = ldpc_dec->z_c;
+ int K = (ldpc_dec->basegraph == 1 ? ACC_K_ZC_1 : ACC_K_ZC_2)
+ * z_c;
+
+ bool required = false;
+ if (ldpc_dec->basegraph == 1) {
+ if ((q_m == 4) && (z_c >= 320) && (e * ACC_LIM_31 > K * 64))
+ required = true;
+ else if ((e * ACC_LIM_21 > K * 64))
+ required = true;
+ } else {
+ if (q_m <= 2) {
+ if ((z_c >= 208) && (e * ACC_LIM_09 > K * 64))
+ required = true;
+ else if ((z_c < 208) && (e * ACC_LIM_03 > K * 64))
+ required = true;
+ } else if (e * ACC_LIM_14 > K * 64)
+ required = true;
+ }
+ if (required)
+ rte_bbdev_log(INFO, "Running deRM pre-processing in SW");
+
+ return required;
+}
+
/** Enqueue one decode operations for ACC100 device in CB mode */
static inline int
enqueue_ldpc_dec_one_op_cb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
- uint16_t total_enqueued_cbs, bool same_op)
+ uint16_t total_enqueued_cbs, bool same_op,
+ struct rte_bbdev_queue_data *q_data)
{
int ret;
if (unlikely(check_bit(op->ldpc_dec.op_flags,
@@ -3167,6 +3217,58 @@ enqueue_ldpc_dec_one_op_cb(struct acc_queue *q, struct rte_bbdev_dec_op *op,
} else {
struct acc_fcw_ld *fcw;
uint32_t seg_total_left;
+
+ if (derm_workaround_required(&op->ldpc_dec, q)) {
+ #ifdef RTE_BBDEV_SDK_AVX512
+ struct rte_bbdev_op_ldpc_dec *dec = &op->ldpc_dec;
+ struct bblib_rate_dematching_5gnr_request derm_req;
+ struct bblib_rate_dematching_5gnr_response derm_resp;
+ uint8_t *in;
+
+ /* Checking input size is matching with E */
+ if (dec->input.data->data_len < dec->cb_params.e) {
+ rte_bbdev_log(ERR, "deRM: Input size mismatch");
+ return -EFAULT;
+ }
+ /* Run first deRM processing in SW */
+ in = rte_pktmbuf_mtod_offset(dec->input.data, uint8_t *, in_offset);
+ derm_req.p_in = (int8_t *) in;
+ derm_req.p_harq = (int8_t *) q->derm_buffer;
+ derm_req.base_graph = dec->basegraph;
+ derm_req.zc = dec->z_c;
+ derm_req.ncb = dec->n_cb;
+ derm_req.e = dec->cb_params.e;
+ if (derm_req.e > ACC_MAX_E) {
+ rte_bbdev_log(WARNING,
+ "deRM: E %d > %d max",
+ derm_req.e, ACC_MAX_E);
+ derm_req.e = ACC_MAX_E;
+ }
+ derm_req.k0 = 0; /* Actual output from SDK */
+ derm_req.isretx = false;
+ derm_req.rvid = dec->rv_index;
+ derm_req.modulation_order = dec->q_m;
+ derm_req.start_null_index =
+ (dec->basegraph == 1 ? 22 : 10)
+ * dec->z_c - 2 * dec->z_c
+ - dec->n_filler;
+ derm_req.num_of_null = dec->n_filler;
+ bblib_rate_dematching_5gnr(&derm_req, &derm_resp);
+ /* Force back the HW DeRM */
+ dec->q_m = 1;
+ dec->cb_params.e = dec->n_cb - dec->n_filler;
+ dec->rv_index = 0;
+ rte_memcpy(in, q->derm_buffer, dec->cb_params.e);
+ /* Capture counter when pre-processing is used */
+ q_data->queue_stats.enqueue_warn_count++;
+ #else
+ RTE_SET_USED(q_data);
+ rte_bbdev_log(WARNING,
+ "Corner case may require deRM pre-processing in SDK"
+ );
+ #endif
+ }
+
fcw = &desc->req.fcw_ld;
q->d->fcw_ld_fill(op, fcw, harq_layout);
@@ -3720,7 +3822,7 @@ acc100_enqueue_ldpc_dec_cb(struct rte_bbdev_queue_data *q_data,
ops[i]->ldpc_dec.n_cb, ops[i]->ldpc_dec.q_m,
ops[i]->ldpc_dec.n_filler, ops[i]->ldpc_dec.cb_params.e,
same_op);
- ret = enqueue_ldpc_dec_one_op_cb(q, ops[i], i, same_op);
+ ret = enqueue_ldpc_dec_one_op_cb(q, ops[i], i, same_op, q_data);
if (ret < 0) {
acc_enqueue_invalid(q_data);
break;
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 29/30] baseband/acc100: configure PMON control registers
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (27 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 28/30] baseband/acc100: add workaround for deRM corner cases Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-19 0:39 ` [PATCH v4 30/30] baseband/acc100: update guide docs Hernan Vargas
29 siblings, 0 replies; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Enable performance monitor control registers.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/baseband/acc/acc100_pmd.h | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/baseband/acc/acc100_pmd.h b/drivers/baseband/acc/acc100_pmd.h
index eb6349c85a..8c0aec5ed8 100644
--- a/drivers/baseband/acc/acc100_pmd.h
+++ b/drivers/baseband/acc/acc100_pmd.h
@@ -115,6 +115,8 @@ struct acc100_registry_addr {
unsigned int depth_log1_offset;
unsigned int qman_group_func;
unsigned int ddr_range;
+ unsigned int pmon_ctrl_a;
+ unsigned int pmon_ctrl_b;
};
/* Structure holding registry addresses for PF */
@@ -144,6 +146,8 @@ static const struct acc100_registry_addr pf_reg_addr = {
.depth_log1_offset = HWPfQmgrGrpDepthLog21Vf,
.qman_group_func = HWPfQmgrGrpFunction0,
.ddr_range = HWPfDmaVfDdrBaseRw,
+ .pmon_ctrl_a = HWVfPmACntrlRegVf,
+ .pmon_ctrl_b = HWVfPmBCntrlRegVf,
};
/* Structure holding registry addresses for VF */
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH v4 30/30] baseband/acc100: update guide docs
2022-10-19 0:38 [PATCH v4 00/30] baseband/acc100: changes for 22.11 Hernan Vargas
` (28 preceding siblings ...)
2022-10-19 0:39 ` [PATCH v4 29/30] baseband/acc100: configure PMON control registers Hernan Vargas
@ 2022-10-19 0:39 ` Hernan Vargas
2022-10-20 9:59 ` [EXT] " Akhil Goyal
29 siblings, 1 reply; 33+ messages in thread
From: Hernan Vargas @ 2022-10-19 0:39 UTC (permalink / raw)
To: dev, gakhil, trix, maxime.coquelin
Cc: nicolas.chautru, qi.z.zhang, Hernan Vargas
Add new flags ACC100_EXT_MEM and RTE_LIBRTE_BBDEV_SKIP_VALIDATE to the
documentation.
Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
---
doc/guides/bbdevs/acc100.rst | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/doc/guides/bbdevs/acc100.rst b/doc/guides/bbdevs/acc100.rst
index 8b29b92a9d..b0ac2526de 100644
--- a/doc/guides/bbdevs/acc100.rst
+++ b/doc/guides/bbdevs/acc100.rst
@@ -65,6 +65,15 @@ ACC100 and ACC101 5G/4G FEC PMDs support the following BBDEV capabilities:
- ``RTE_BBDEV_TURBO_DEC_SCATTER_GATHER`` : supports scatter-gather for input/output data
- ``RTE_BBDEV_TURBO_HALF_ITERATION_EVEN`` : set half iteration granularity
+* PMD Specific build flags:
+ The ACC100 PMD includes some optional build flags which may be used for troubleshooting.
+ Recommended build configuration is for these to be kept as default.
+ - ``ACC100_EXT_MEM`` : default option with memory external to CPU on the PCie card DDR itself.
+ Alternative build option will use CPU memory (not recommended).
+ - ``RTE_LIBRTE_BBDEV_SKIP_VALIDATE`` : option to skip API input validation.
+ Recommended value is to keep the validation enable by default as a protection for negative
+ scenarios at a cost of some cycles spent to enforce these checks.
+
Installation
------------
--
2.37.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* RE: [EXT] [PATCH v4 30/30] baseband/acc100: update guide docs
2022-10-19 0:39 ` [PATCH v4 30/30] baseband/acc100: update guide docs Hernan Vargas
@ 2022-10-20 9:59 ` Akhil Goyal
0 siblings, 0 replies; 33+ messages in thread
From: Akhil Goyal @ 2022-10-20 9:59 UTC (permalink / raw)
To: Hernan Vargas, dev, trix, maxime.coquelin; +Cc: nicolas.chautru, qi.z.zhang
> Add new flags ACC100_EXT_MEM and RTE_LIBRTE_BBDEV_SKIP_VALIDATE to
> the
> documentation.
>
> Signed-off-by: Hernan Vargas <hernan.vargas@intel.com>
Squash the patch where these flags were introduced.
> ---
> doc/guides/bbdevs/acc100.rst | 9 +++++++++
> 1 file changed, 9 insertions(+)
>
> diff --git a/doc/guides/bbdevs/acc100.rst b/doc/guides/bbdevs/acc100.rst
> index 8b29b92a9d..b0ac2526de 100644
> --- a/doc/guides/bbdevs/acc100.rst
> +++ b/doc/guides/bbdevs/acc100.rst
> @@ -65,6 +65,15 @@ ACC100 and ACC101 5G/4G FEC PMDs support the
> following BBDEV capabilities:
> - ``RTE_BBDEV_TURBO_DEC_SCATTER_GATHER`` : supports scatter-gather
> for input/output data
> - ``RTE_BBDEV_TURBO_HALF_ITERATION_EVEN`` : set half iteration
> granularity
>
> +* PMD Specific build flags:
> + The ACC100 PMD includes some optional build flags which may be used for
> troubleshooting.
> + Recommended build configuration is for these to be kept as default.
> + - ``ACC100_EXT_MEM`` : default option with memory external to CPU on the
> PCie card DDR itself.
> + Alternative build option will use CPU memory (not recommended).
> + - ``RTE_LIBRTE_BBDEV_SKIP_VALIDATE`` : option to skip API input validation.
> + Recommended value is to keep the validation enable by default as a
> protection for negative
> + scenarios at a cost of some cycles spent to enforce these checks.
> +
> Installation
> ------------
>
> --
> 2.37.1
^ permalink raw reply [flat|nested] 33+ messages in thread