* [PATCH] ntb/ae4dma: added Amd userspace DMA and NTB drivers
@ 2025-07-05 14:10 Bhagyada Modali
0 siblings, 0 replies; only message in thread
From: Bhagyada Modali @ 2025-07-05 14:10 UTC (permalink / raw)
To: dev; +Cc: raghavendra.ningoji, Bhagyada Modali
Added a user-space driver with support for the
AMD EPYC 4th Generation DMA (AE4DMA) offload engine.
AMD Non Transparent Bridge driver.
- Implementation of new user-space driver supporting
NTB and DMA memory copy offload on AMD EYPC 9004 & 8004
systems (Genoa and Siena processors).
- Updated the NTB_FWD application to support AMD NTB driver.
Signed-off-by: Bhagyada Modali <bhagyada.modali@amd.com>
---
app/test-dma-perf/benchmark.c | 29 +-
app/test-dma-perf/config.ini | 134 +++--
app/test-dma-perf/main.c | 7 +-
app/test/test_dmadev.c | 47 +-
drivers/dma/ae4dma/ae4dma_dmadev.c | 704 +++++++++++++++++++++++++
drivers/dma/ae4dma/ae4dma_hw_defs.h | 223 ++++++++
drivers/dma/ae4dma/ae4dma_internal.h | 138 +++++
drivers/dma/ae4dma/meson.build | 7 +
drivers/dma/meson.build | 1 +
drivers/raw/ntb/meson.build | 1 +
drivers/raw/ntb/ntb.c | 371 +++++++++-----
drivers/raw/ntb/ntb.h | 15 +
drivers/raw/ntb/ntb_hw_amd.c | 738 +++++++++++++++++++++++++++
drivers/raw/ntb/ntb_hw_amd.h | 123 +++++
drivers/raw/ntb/ntb_hw_intel.c | 21 +
drivers/raw/ntb/rte_pmd_ntb.h | 2 +
examples/ntb/commands.list | 2 +-
examples/ntb/ntb_fwd.c | 362 +++++++++++--
usertools/dpdk-devbind.py | 15 +-
19 files changed, 2698 insertions(+), 242 deletions(-)
create mode 100644 drivers/dma/ae4dma/ae4dma_dmadev.c
create mode 100644 drivers/dma/ae4dma/ae4dma_hw_defs.h
create mode 100644 drivers/dma/ae4dma/ae4dma_internal.h
create mode 100644 drivers/dma/ae4dma/meson.build
create mode 100644 drivers/raw/ntb/ntb_hw_amd.c
create mode 100644 drivers/raw/ntb/ntb_hw_amd.h
diff --git a/app/test-dma-perf/benchmark.c b/app/test-dma-perf/benchmark.c
index 6d617ea200..7d7e07228c 100644
--- a/app/test-dma-perf/benchmark.c
+++ b/app/test-dma-perf/benchmark.c
@@ -266,6 +266,27 @@ error_exit(int dev_id)
rte_exit(EXIT_FAILURE, "DMA error\n");
}
+static void
+await_hw(int16_t dev_id, uint16_t vchan)
+{
+ enum rte_dma_vchan_status st;
+
+ if (rte_dma_vchan_status(dev_id, vchan, &st) < 0) {
+ /* for drivers that don't support this op, just sleep for 1 us */
+ rte_delay_us_sleep(1);
+ return;
+ }
+
+ /* for those that do, *max* end time is one second from now, but all should be faster */
+ const uint64_t end_cycles = rte_get_timer_cycles() + rte_get_timer_hz();
+
+ while (st == RTE_DMA_VCHAN_ACTIVE && rte_get_timer_cycles() < end_cycles) {
+ rte_pause();
+ rte_dma_vchan_status(dev_id, vchan, &st);
+ }
+}
+
+
static inline void
do_dma_submit_and_poll(uint16_t dev_id, uint64_t *async_cnt,
volatile struct worker_info *worker_info)
@@ -276,7 +297,6 @@ do_dma_submit_and_poll(uint16_t dev_id, uint64_t *async_cnt,
ret = rte_dma_submit(dev_id, 0);
if (ret < 0)
error_exit(dev_id);
-
nr_cpl = rte_dma_completed(dev_id, 0, MAX_DMA_CPL_NB, NULL, NULL);
*async_cnt -= nr_cpl;
worker_info->total_cpl += nr_cpl;
@@ -309,14 +329,16 @@ do_dma_plain_mem_copy(void *p)
for (i = 0; i < nr_buf; i++) {
dma_copy:
ret = rte_dma_copy(dev_id, 0, rte_mbuf_data_iova(srcs[i]),
- rte_mbuf_data_iova(dsts[i]), buf_size, 0);
+ rte_mbuf_data_iova(dsts[i]), buf_size, 0);//
if (unlikely(ret < 0)) {
+ await_hw(dev_id, 0);
if (ret == -ENOSPC) {
do_dma_submit_and_poll(dev_id, &async_cnt, worker_info);
goto dma_copy;
} else
error_exit(dev_id);
}
+
async_cnt++;
if ((async_cnt % kick_batch) == 0)
@@ -756,6 +778,7 @@ mem_copy_benchmark(struct test_configure *cfg)
while (1) {
bool ready = true;
+
for (i = 0; i < nb_workers; i++) {
if (lcores[i]->worker_info.ready_flag == false) {
ready = 0;
@@ -786,6 +809,7 @@ mem_copy_benchmark(struct test_configure *cfg)
for (k = 0; k < nb_workers; k++) {
struct rte_mbuf **src_buf = NULL, **dst_buf = NULL;
uint32_t nr_buf_pt = nr_buf / nb_workers;
+
vchan_dev = &cfg->dma_config[k].vchan_dev;
offset = nr_buf / nb_workers * k;
src_buf = srcs + offset;
@@ -871,6 +895,7 @@ mem_copy_benchmark(struct test_configure *cfg)
for (k = 0; k < nb_workers; k++) {
struct rte_mbuf **sbuf = NULL, **dbuf = NULL;
+
vchan_dev = &cfg->dma_config[k].vchan_dev;
offset = nr_buf / nb_workers * k;
m = NULL;
diff --git a/app/test-dma-perf/config.ini b/app/test-dma-perf/config.ini
index 61e49dbae5..4fa8713e89 100644
--- a/app/test-dma-perf/config.ini
+++ b/app/test-dma-perf/config.ini
@@ -61,57 +61,95 @@
[case1]
type=DMA_MEM_COPY
-mem_size=10
-buf_size=64,8192,2,MUL
-dma_ring_size=1024
-kick_batch=32
+mem_size=64
+buf_size=32768
+dma_ring_size=32
+kick_batch=4
src_numa_node=0
dst_numa_node=0
cache_flush=0
test_seconds=2
-lcore_dma0=lcore=10,dev=0000:00:04.1,dir=mem2mem
-lcore_dma1=lcore=11,dev=0000:00:04.2,dir=mem2mem
+lcore_dma0=lcore=4,dev=0000:04:00.1-ch0,dir=mem2mem
+lcore_dma1=lcore=5,dev=0000:04:00.1-ch1,dir=mem2mem
+lcore_dma2=lcore=7,dev=0000:64:00.1-ch0,dir=mem2mem
+lcore_dma3=lcore=8,dev=0000:64:00.1-ch1,dir=mem2mem
+lcore_dma4=lcore=14,dev=0000:41:00.1-ch0,dir=mem2mem
+lcore_dma5=lcore=15,dev=0000:41:00.1-ch1,dir=mem2mem
+lcore_dma6=lcore=17,dev=0000:21:00.1-ch0,dir=mem2mem
+lcore_dma7=lcore=18,dev=0000:21:00.1-ch1,dir=mem2mem
+;lcore_dma0=lcore=13,dev=0000:41:00.1-ch0,dir=mem2mem
+;lcore_dma1=lcore=14,dev=0000:41:00.1-ch1,dir=mem2mem
+;lcore_dma2=lcore=15,dev=0000:41:00.1-ch2,dir=mem2mem
+;lcore_dma3=lcore=16,dev=0000:41:00.1-ch3,dir=mem2mem
+;lcore_dma4=lcore=17,dev=0000:41:00.1-ch4,dir=mem2mem
+;lcore_dma5=lcore=18,dev=0000:41:00.1-ch5,dir=mem2mem
+;lcore_dma6=lcore=19,dev=0000:41:00.1-ch6,dir=mem2mem
+;lcore_dma7=lcore=20,dev=0000:41:00.1-ch7,dir=mem2mem
+;lcore_dma8=lcore=21,dev=0000:41:00.1-ch8,dir=mem2mem
+;lcore_dma9=lcore=22,dev=0000:41:00.1-ch9,dir=mem2mem
+;lcore_dma10=lcore=23,dev=0000:41:00.1-ch10,dir=mem2mem
+;lcore_dma11=lcore=24,dev=0000:41:00.1-ch11,dir=mem2mem
+;lcore_dma12=lcore=25,dev=0000:41:00.1-ch12,dir=mem2mem
+;lcore_dma13=lcore=26,dev=0000:41:00.1-ch13,dir=mem2mem
+;lcore_dma14=lcore=27,dev=0000:41:00.1-ch14,dir=mem2mem
+;lcore_dma15=lcore=28,dev=0000:41:00.1-ch15,dir=mem2mem
+;lcore_dma16=lcore=32,dev=0000:21:00.1-ch0,dir=mem2mem
+;lcore_dma17=lcore=33,dev=0000:21:00.1-ch1,dir=mem2mem
+;lcore_dma18=lcore=34,dev=0000:21:00.1-ch2,dir=mem2mem
+;lcore_dma19=lcore=35,dev=0000:21:00.1-ch3,dir=mem2mem
+;lcore_dma20=lcore=36,dev=0000:21:00.1-ch4,dir=mem2mem
+;lcore_dma21=lcore=37,dev=0000:21:00.1-ch5,dir=mem2mem
+;lcore_dma22=lcore=38,dev=0000:21:00.1-ch6,dir=mem2mem
+;lcore_dma23=lcore=39,dev=0000:21:00.1-ch7,dir=mem2mem
+;lcore_dma24=lcore=40,dev=0000:21:00.1-ch8,dir=mem2mem
+;lcore_dma25=lcore=41,dev=0000:21:00.1-ch9,dir=mem2mem
+;lcore_dma26=lcore=42,dev=0000:21:00.1-ch10,dir=mem2mem
+;lcore_dma27=lcore=43,dev=0000:21:00.1-ch11,dir=mem2mem
+;lcore_dma28=lcore=44,dev=0000:21:00.1-ch12,dir=mem2mem
+;lcore_dma29=lcore=45,dev=0000:21:00.1-ch13,dir=mem2mem
+;lcore_dma30=lcore=46,dev=0000:21:00.1-ch14,dir=mem2mem
+;lcore_dma31=lcore=47,dev=0000:21:00.1-ch15,dir=mem2mem
eal_args=--in-memory --file-prefix=test
-[case2]
-type=DMA_MEM_COPY
-mem_size=10
-buf_size=64,8192,2,MUL
-dma_ring_size=1024
-dma_src_sge=4
-dma_dst_sge=1
-kick_batch=32
-src_numa_node=0
-dst_numa_node=0
-cache_flush=0
-test_seconds=2
-lcore_dma0=lcore=10,dev=0000:00:04.1,dir=mem2mem
-lcore_dma1=lcore=11,dev=0000:00:04.2,dir=mem2mem
-eal_args=--in-memory --file-prefix=test
-
-[case3]
-skip=1
-type=DMA_MEM_COPY
-mem_size=10
-buf_size=64,4096,2,MUL
-dma_ring_size=1024
-kick_batch=32
-src_numa_node=0
-dst_numa_node=0
-cache_flush=0
-test_seconds=2
-lcore_dma0=lcore=10,dev=0000:00:04.1,dir=mem2mem
-lcore_dma1=lcore=11,dev=0000:00:04.2,dir=dev2mem,raddr=0x200000000,coreid=1,pfid=2,vfid=3
-lcore_dma2=lcore=12,dev=0000:00:04.3,dir=mem2dev,raddr=0x300000000,coreid=3,pfid=2,vfid=1
-eal_args=--in-memory --file-prefix=test
-
-[case4]
-type=CPU_MEM_COPY
-mem_size=10
-buf_size=64,8192,2,MUL
-src_numa_node=0
-dst_numa_node=1
-cache_flush=0
-test_seconds=2
-lcore = 3, 4
-eal_args=--in-memory --no-pci
+;[case2]
+;type=DMA_MEM_COPY
+;mem_size=10
+;buf_size=64,8192,2,MUL
+;dma_ring_size=1024
+;dma_src_sge=4
+;dma_dst_sge=1
+;kick_batch=32
+;src_numa_node=0
+;dst_numa_node=0
+;cache_flush=0
+;test_seconds=2
+;lcore_dma0=lcore=10,dev=0000:00:04.1,dir=mem2mem
+;lcore_dma1=lcore=11,dev=0000:00:04.2,dir=mem2mem
+;eal_args=--in-memory --file-prefix=test
+;
+;[case3]
+;skip=1
+;type=DMA_MEM_COPY
+;mem_size=10
+;buf_size=64,4096,2,MUL
+;dma_ring_size=1024
+;kick_batch=32
+;src_numa_node=0
+;dst_numa_node=0
+;cache_flush=0
+;test_seconds=2
+;lcore_dma0=lcore=10,dev=0000:00:04.1,dir=mem2mem
+;lcore_dma1=lcore=11,dev=0000:00:04.2,dir=dev2mem,raddr=0x200000000,coreid=1,pfid=2,vfid=3
+;lcore_dma2=lcore=12,dev=0000:00:04.3,dir=mem2dev,raddr=0x300000000,coreid=3,pfid=2,vfid=1
+;eal_args=--in-memory --file-prefix=test
+;
+;[case4]
+;type=CPU_MEM_COPY
+;mem_size=10
+;buf_size=64,8192,2,MUL
+;src_numa_node=0
+;dst_numa_node=1
+;cache_flush=0
+;test_seconds=2
+;lcore = 3, 4
+;eal_args=--in-memory --no-pci
diff --git a/app/test-dma-perf/main.c b/app/test-dma-perf/main.c
index 0586b3e1d0..66fb1ee0d9 100644
--- a/app/test-dma-perf/main.c
+++ b/app/test-dma-perf/main.c
@@ -192,6 +192,7 @@ parse_lcore(struct test_configure *test_case, const char *value)
strlcpy(input, value, len + 1);
char *token = strtok(input, ", ");
+
while (token != NULL) {
lcore_dma_map = &(test_case->dma_config[test_case->num_worker++].lcore_dma_map);
memset(lcore_dma_map, 0, sizeof(struct lcore_dma_map_t));
@@ -201,6 +202,7 @@ parse_lcore(struct test_configure *test_case, const char *value)
}
uint16_t lcore_id = atoi(token);
+
lcore_dma_map->lcore = lcore_id;
token = strtok(NULL, ", ");
@@ -428,6 +430,7 @@ load_configs(const char *path)
char lc_dma[RTE_DEV_NAME_MAX_LEN];
int i = 0;
+
while (1) {
snprintf(lc_dma, RTE_DEV_NAME_MAX_LEN, "lcore_dma%d", i);
lcore_dma = rte_cfgfile_get_entry(cfgfile, section_name, lc_dma);
@@ -460,6 +463,7 @@ load_configs(const char *path)
} else {
lcore_dma = rte_cfgfile_get_entry(cfgfile, section_name, "lcore");
int lcore_ret = parse_lcore(test_case, lcore_dma);
+
if (lcore_ret < 0) {
printf("parse lcore error in case %d.\n", i + 1);
test_case->is_valid = false;
@@ -551,6 +555,7 @@ main(int argc, char *argv[])
if (rst_path_ptr == NULL) {
strlcpy(rst_path, cfg_path_ptr, PATH_MAX);
char *token = strtok(basename(rst_path), ".");
+
if (token == NULL) {
printf("Config file error.\n");
return -1;
@@ -566,7 +571,6 @@ main(int argc, char *argv[])
return -1;
}
fclose(fd);
-
printf("Running cases...\n");
for (i = 0; i < case_nb; i++) {
if (test_cases[i].is_skip) {
@@ -644,7 +648,6 @@ main(int argc, char *argv[])
printf("Case process unknown terminated.\n\n");
}
}
-
printf("Bye...\n");
return 0;
}
diff --git a/app/test/test_dmadev.c b/app/test/test_dmadev.c
index 143e1bcd68..52e5d33dbe 100644
--- a/app/test/test_dmadev.c
+++ b/app/test/test_dmadev.c
@@ -4,6 +4,7 @@
*/
#include <inttypes.h>
+#include <rte_malloc.h>
#include <rte_dmadev.h>
#include <rte_mbuf.h>
@@ -19,9 +20,10 @@
#define ERR_RETURN(...) do { print_err(__func__, __LINE__, __VA_ARGS__); return -1; } while (0)
#define TEST_NAME_MAX_LEN 80
-#define TEST_RINGSIZE 512
+#define TEST_RINGSIZE 32
+//#define TEST_RINGSIZE 512
#define COPY_LEN 2048
-
+#define ALIGN_4K 4096
static struct rte_dma_info info;
static struct rte_mempool *pool;
static bool check_err_stats;
@@ -122,6 +124,7 @@ await_hw(int16_t dev_id, uint16_t vchan)
/* for those that do, *max* end time is one second from now, but all should be faster */
const uint64_t end_cycles = rte_get_timer_cycles() + rte_get_timer_hz();
+
while (st == RTE_DMA_VCHAN_ACTIVE && rte_get_timer_cycles() < end_cycles) {
rte_pause();
rte_dma_vchan_status(dev_id, vchan, &st);
@@ -135,8 +138,8 @@ do_multi_copies(int16_t dev_id, uint16_t vchan,
int split_completions, /* gather 2 x 16 or 1 x 32 completions */
int use_completed_status) /* use completed or completed_status function */
{
- struct rte_mbuf *srcs[32], *dsts[32];
- enum rte_dma_status_code sc[32];
+ struct rte_mbuf *srcs[16], *dsts[16];
+ enum rte_dma_status_code sc[16];
unsigned int i, j;
bool dma_err = false;
@@ -159,6 +162,7 @@ do_multi_copies(int16_t dev_id, uint16_t vchan,
if (rte_dma_copy(dev_id, vchan, rte_mbuf_data_iova(srcs[i]),
rte_mbuf_data_iova(dsts[i]), COPY_LEN, 0) != id_count++)
ERR_RETURN("Error with rte_dma_copy for buffer %u\n", i);
+ id_count %= 32;
}
rte_dma_submit(dev_id, vchan);
@@ -168,6 +172,7 @@ do_multi_copies(int16_t dev_id, uint16_t vchan,
/* gather completions in two halves */
uint16_t half_len = RTE_DIM(srcs) / 2;
int ret = rte_dma_completed(dev_id, vchan, half_len, NULL, &dma_err);
+
if (ret != half_len || dma_err)
ERR_RETURN("Error with rte_dma_completed - first half. ret = %d, expected ret = %u, dma_err = %d\n",
ret, half_len, dma_err);
@@ -182,11 +187,13 @@ do_multi_copies(int16_t dev_id, uint16_t vchan,
*/
if (!use_completed_status) {
int n = rte_dma_completed(dev_id, vchan, RTE_DIM(srcs), NULL, &dma_err);
+
if (n != RTE_DIM(srcs) || dma_err)
ERR_RETURN("Error with rte_dma_completed, %u [expected: %zu], dma_err = %d\n",
n, RTE_DIM(srcs), dma_err);
} else {
int n = rte_dma_completed_status(dev_id, vchan, RTE_DIM(srcs), NULL, sc);
+
if (n != RTE_DIM(srcs))
ERR_RETURN("Error with rte_dma_completed_status, %u [expected: %zu]\n",
n, RTE_DIM(srcs));
@@ -229,8 +236,10 @@ test_single_copy(int16_t dev_id, uint16_t vchan)
struct rte_mbuf *src, *dst;
char *src_data, *dst_data;
+
src = rte_pktmbuf_alloc(pool);
dst = rte_pktmbuf_alloc(pool);
+
src_data = rte_pktmbuf_mtod(src, char *);
dst_data = rte_pktmbuf_mtod(dst, char *);
@@ -250,6 +259,10 @@ test_single_copy(int16_t dev_id, uint16_t vchan)
if (dst_data[i] != src_data[i])
ERR_RETURN("Data mismatch at char %u [Got %02x not %02x]\n", i,
dst_data[i], src_data[i]);
+/* else
+ printf("Data MATCHED! at char %u [ %02x == %02x]\n", i,
+ dst_data[i], src_data[i]);
+*/
/* now check completion works */
id = ~id;
@@ -284,6 +297,7 @@ test_single_copy(int16_t dev_id, uint16_t vchan)
ERR_RETURN("Error with rte_dma_completed in empty check\n");
id_count++;
+ id_count %= 32;
return 0;
}
@@ -296,17 +310,17 @@ test_enqueue_copies(int16_t dev_id, uint16_t vchan)
/* test doing a single copy */
if (test_single_copy(dev_id, vchan) < 0)
return -1;
-
/* test doing a multiple single copies */
do {
uint16_t id;
- const uint16_t max_ops = 4;
+ const uint16_t max_ops = 28;
struct rte_mbuf *src, *dst;
char *src_data, *dst_data;
uint16_t count;
src = rte_pktmbuf_alloc(pool);
dst = rte_pktmbuf_alloc(pool);
+
src_data = rte_pktmbuf_mtod(src, char *);
dst_data = rte_pktmbuf_mtod(dst, char *);
@@ -314,12 +328,14 @@ test_enqueue_copies(int16_t dev_id, uint16_t vchan)
src_data[i] = rte_rand() & 0xFF;
/* perform the same copy <max_ops> times */
- for (i = 0; i < max_ops; i++)
+ for (i = 0; i < max_ops; i++) {
if (rte_dma_copy(dev_id, vchan,
rte_pktmbuf_iova(src),
rte_pktmbuf_iova(dst),
COPY_LEN, RTE_DMA_OP_FLAG_SUBMIT) != id_count++)
ERR_RETURN("Error with rte_dma_copy\n");
+ id_count %= 32;
+ }
await_hw(dev_id, vchan);
@@ -328,7 +344,7 @@ test_enqueue_copies(int16_t dev_id, uint16_t vchan)
ERR_RETURN("Error with rte_dma_completed, got %u not %u\n",
count, max_ops);
- if (id != id_count - 1)
+ if (id != (id_count - 1 + 32) % 32)
ERR_RETURN("Error, incorrect job id returned: got %u not %u\n",
id, id_count - 1);
@@ -339,8 +355,8 @@ test_enqueue_copies(int16_t dev_id, uint16_t vchan)
rte_pktmbuf_free(src);
rte_pktmbuf_free(dst);
} while (0);
-
/* test doing multiple copies */
+ return 0;
return do_multi_copies(dev_id, vchan, 0, 0, 0) /* enqueue and complete 1 batch at a time */
/* enqueue 2 batches and then complete both */
|| do_multi_copies(dev_id, vchan, 1, 0, 0)
@@ -634,6 +650,7 @@ test_individual_status_query_with_failure(int16_t dev_id, uint16_t vchan, bool f
/* use regular "completed" until we hit error */
while (!error) {
uint16_t n = rte_dma_completed(dev_id, vchan, 1, &idx, &error);
+
count += n;
if (n > 1 || count >= COMP_BURST_SZ)
ERR_RETURN("Error - too many completions got\n");
@@ -883,6 +900,7 @@ test_enqueue_fill(int16_t dev_id, uint16_t vchan)
/* check the data from the fill operation is correct */
for (j = 0; j < lengths[i]; j++) {
char pat_byte = ((char *)&pattern)[j % 8];
+
if (dst_data[j] != pat_byte)
ERR_RETURN("Error with fill operation (lengths = %u): got (%x), not (%x)\n",
lengths[i], dst_data[j], pat_byte);
@@ -1161,7 +1179,7 @@ test_dmadev_setup(void)
if (rte_dma_stats_get(dev_id, vchan, &stats) != 0)
ERR_RETURN("Error with rte_dma_stats_get()\n");
- if (rte_dma_burst_capacity(dev_id, vchan) < 32)
+ if (rte_dma_burst_capacity(dev_id, vchan) < 2)
ERR_RETURN("Error: Device does not have sufficient burst capacity to run tests");
if (stats.completed != 0 || stats.submitted != 0 || stats.errors != 0)
@@ -1211,7 +1229,7 @@ test_dmadev_instance(int16_t dev_id)
};
static struct runtest_param param[] = {
- {"copy", test_enqueue_copies, 640},
+ {"copy", test_enqueue_copies, 10000},
{"sg_copy", test_enqueue_sg_copies, 1},
{"stop_start", test_stop_start, 1},
{"burst_capacity", test_burst_capacity, 1},
@@ -1246,6 +1264,8 @@ test_dmadev_instance(int16_t dev_id)
TEST_CASE_NAMED_WITH_DATA("m2d_autofree",
test_dmadev_autofree_setup, NULL,
runtest, ¶m[TEST_M2D]),
+
+
TEST_CASES_END()
}
};
@@ -1317,9 +1337,10 @@ test_dma(void)
return TEST_SKIPPED;
RTE_DMA_FOREACH_DEV(i) {
- if (test_dma_api(i) < 0)
- ERR_RETURN("Error performing API tests\n");
+/* if (test_dma_api(i) < 0)
+ * ERR_RETURN("Error performing API tests\n");
+ */
if (test_dmadev_instance(i) < 0)
ERR_RETURN("Error, test failure for device %d\n", i);
}
diff --git a/drivers/dma/ae4dma/ae4dma_dmadev.c b/drivers/dma/ae4dma/ae4dma_dmadev.c
new file mode 100644
index 0000000000..2dbb7820a9
--- /dev/null
+++ b/drivers/dma/ae4dma/ae4dma_dmadev.c
@@ -0,0 +1,704 @@
+/* SPDX-License-Identifier: BSD-3.0-Clause
+ * Copyright(c) 2021 Advanced Micro Devices, Inc. All rights reserved.
+ */
+
+#include <rte_bus_pci.h>
+#include <bus_pci_driver.h>
+#include <rte_dmadev_pmd.h>
+#include <rte_malloc.h>
+#include <rte_prefetch.h>
+#include <rte_errno.h>
+
+#include "ae4dma_internal.h"
+
+#define MAX_RETRY 10
+int hwq_id = 0;
+
+static struct rte_pci_driver ae4dma_pmd_drv;
+
+RTE_LOG_REGISTER_DEFAULT(ae4dma_pmd_logtype, INFO);
+
+static int ae4dma_dev_dump(const struct rte_dma_dev *dev, FILE *f);
+static int ae4dma_add_queue(struct ae4dma_dmadev *dev, uint8_t qn);
+
+#define DESC_SZ sizeof(struct ae4dma_dma_hw_desc)
+
+#define AE4DMA_PMD_NAME dmadev_ae4dma
+#define AE4DMA_PMD_NAME_STR RTE_STR(AE4DMA_PMD_NAME)
+
+/* AE4DMA operations. */
+enum rte_ae4dma_ops {
+ ae4dma_op_copy = 0, /* Standard DMA Operation */
+ ae4dma_op_fill /* Block Fill */
+};
+
+static const struct rte_memzone *
+ae4dma_queue_dma_zone_reserve(const char *queue_name,
+ uint32_t queue_size, int socket_id)
+{
+ const struct rte_memzone *mz;
+
+ mz = rte_memzone_lookup(queue_name);
+ if (mz != 0) {
+ if (((size_t)queue_size <= mz->len) &&
+ ((socket_id == SOCKET_ID_ANY) ||
+ (socket_id == mz->socket_id))) {
+ AE4DMA_PMD_INFO("re-use memzone already allocated for %s",
+ queue_name);
+ return mz;
+ }
+ AE4DMA_PMD_ERR("Incompatible memzone already allocated %s, size %u, socket %d. Requested size %u, socket %u",
+ queue_name, (uint32_t)mz->len,
+ mz->socket_id, queue_size, socket_id);
+ return NULL;
+ }
+ /* AE4DMA_PMD_INFO("Allocate memzone for %s, size %u on socket %u",
+ queue_name, queue_size, socket_id);
+ */
+ return rte_memzone_reserve_aligned(queue_name, queue_size,
+ socket_id, RTE_MEMZONE_IOVA_CONTIG, queue_size);
+}
+
+
+/* Configure a device. */
+static int
+ae4dma_dev_configure(struct rte_dma_dev *dev __rte_unused, const struct rte_dma_conf *dev_conf,
+ uint32_t conf_sz)
+{
+ if (sizeof(struct rte_dma_conf) != conf_sz)
+ return -EINVAL;
+
+ if (dev_conf->nb_vchans != 1)
+ return -EINVAL;
+
+ return 0;
+}
+
+/* Setup a virtual channel for AE4DMA, only 1 vchan is supported. */
+static int
+ae4dma_vchan_setup(struct rte_dma_dev *dev, uint16_t vchan __rte_unused,
+ const struct rte_dma_vchan_conf *qconf, uint32_t qconf_sz)
+{
+ struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ uint16_t max_desc = qconf->nb_desc;
+
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+
+ if (sizeof(struct rte_dma_vchan_conf) != qconf_sz)
+ return -EINVAL;
+
+ cmd_q->qcfg = *qconf;
+
+ if (!rte_is_power_of_2(max_desc)) {
+ max_desc = rte_align32pow2(max_desc);
+ printf("DMA dev %u using %u descriptors\n", dev->data->dev_id, max_desc);
+ AE4DMA_PMD_DEBUG("DMA dev %u using %u descriptors", dev->data->dev_id, max_desc);
+ cmd_q->qcfg.nb_desc = max_desc;
+ }
+
+ /* Ensure all counters are reset, if reconfiguring/restarting device.*/
+ /* Reset Stats. */
+ memset(&cmd_q->stats, 0, sizeof(cmd_q->stats));
+ return 0;
+}
+
+
+/* Start a configured device. */
+static int
+ae4dma_dev_start(struct rte_dma_dev *dev)
+{
+ struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+
+ if (cmd_q->qcfg.nb_desc == 0)
+ return -EBUSY;
+ return 0;
+}
+
+/* Stop a configured device. */
+static int
+ae4dma_dev_stop(struct rte_dma_dev *dev)
+{
+ struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+
+ if (cmd_q->qcfg.nb_desc == 0)
+ return -EBUSY;
+ return 0;
+}
+
+/* Get device information of a device. */
+static int
+ae4dma_dev_info_get(const struct rte_dma_dev *dev, struct rte_dma_info *info, uint32_t size)
+{
+
+ if (size < sizeof(*info))
+ return -EINVAL;
+ info->dev_name = dev->device->name;
+ info->dev_capa = RTE_DMA_CAPA_MEM_TO_MEM;
+
+ info->max_vchans = 1;
+ info->min_desc = 2;
+ info->max_desc = 32;
+ info->nb_vchans = 1;
+ return 0;
+}
+
+/* Close a configured device. */
+static int
+ae4dma_dev_close(struct rte_dma_dev *dev)
+{
+ RTE_SET_USED(dev);
+ return 0;
+}
+
+/* trigger h/w to process enqued desc:doorbell - by next_write */
+static inline void
+__submit(struct ae4dma_dmadev *ae4dma)
+{
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+ volatile uint16_t write_idx = cmd_q->next_write;
+
+ AE4DMA_WRITE_REG(&cmd_q->hwq_regs->write_idx, write_idx);
+
+ cmd_q->stats.submitted += (uint16_t)(cmd_q->next_write - cmd_q->last_write + AE4DMA_DESCRITPTORS_PER_CMDQ) % AE4DMA_DESCRITPTORS_PER_CMDQ;
+ cmd_q->last_write = cmd_q->next_write;
+}
+
+/* External submit function wrapper. */
+
+static int
+ae4dma_submit(void *dev_private, uint16_t qid __rte_unused)
+{
+
+ struct ae4dma_dmadev *ae4dma = dev_private;
+
+ __submit(ae4dma);
+
+ return 0;
+}
+
+/* Write descriptor for enqueue. */
+
+static inline int
+__write_desc(void *dev_private, uint32_t op, uint64_t src, phys_addr_t dst,
+ unsigned int len, uint64_t flags)
+{
+ struct ae4dma_dmadev *ae4dma = dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+ struct ae4dma_desc *dma_desc;
+ uint16_t ret;
+ const uint16_t mask = cmd_q->qcfg.nb_desc - 1;
+ const uint16_t read = cmd_q->next_read;
+ uint16_t write = cmd_q->next_write;
+ const uint16_t space = mask + read - write;
+
+ if (cmd_q->ring_buff_count >= 28) {
+ AE4DMA_PMD_DEBUG("NO SPACE : ring_buff_count : %d\n", cmd_q->ring_buff_count);
+ return -ENOSPC;
+ }
+ if (op)
+ AE4DMA_PMD_WARN("FILL not supported:performing COPY\n");
+
+ dma_desc = &ae4dma->cmd_q[hwq_id].qbase_desc[write];
+ dma_desc->dw0.byte0 = 0;
+ dma_desc->dw1.status = 0;
+ dma_desc->dw1.err_code = 0;
+ dma_desc->dw1.desc_id = 0;
+ dma_desc->length = len;
+ dma_desc->src_hi = upper_32_bits(src);
+ dma_desc->src_lo = lower_32_bits(src);
+ dma_desc->dst_hi = upper_32_bits(dst);
+ dma_desc->dst_lo = lower_32_bits(dst);
+
+ cmd_q->ring_buff_count++;
+ cmd_q->next_write = (write + 1) % (AE4DMA_DESCRITPTORS_PER_CMDQ);
+ ret = write;
+ if (flags & RTE_DMA_OP_FLAG_SUBMIT)
+ __submit(ae4dma);
+ return ret;
+}
+
+/* Enqueue a fill operation onto the ae4dma device. */
+static int
+ae4dma_enqueue_fill(void *dev_private, uint16_t qid __rte_unused, uint64_t pattern,
+ rte_iova_t dst, unsigned int length, uint64_t flags)
+{
+ return __write_desc(dev_private, ae4dma_op_fill, pattern, dst, length, flags);
+}
+
+/* Enqueue a copy operation onto the ae4dma device. */
+static int
+ae4dma_enqueue_copy(void *dev_private, uint16_t qid __rte_unused, rte_iova_t src,
+ rte_iova_t dst, unsigned int length, uint64_t flags)
+{
+ return __write_desc(dev_private, ae4dma_op_copy, src, dst, length, flags);
+}
+
+/* Dump DMA device info. */
+static int
+ae4dma_dev_dump(const struct rte_dma_dev *dev, FILE *f)
+{
+ struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ struct ae4dma_cmd_queue *cmd_q;
+ void *ae4dma_mmio_base_addr = (uint8_t *) ae4dma->io_regs;
+
+ cmd_q = &ae4dma->cmd_q[hwq_id];
+ fprintf(f, "cmd_q->id = %" PRIx64 "\n", cmd_q->id);
+ fprintf(f, "cmd_q->qidx = %" PRIx64 "\n", cmd_q->qidx);
+ fprintf(f, "cmd_q->qsize = %" PRIx64 "\n", cmd_q->qsize);
+ fprintf(f, "mmio_base_addr = %p\n", ae4dma_mmio_base_addr);
+ fprintf(f, "queues per ae4dma engine = %d\n", AE4DMA_READ_REG_OFFSET(ae4dma_mmio_base_addr, AE4DMA_COMMON_CONFIG_OFFSET));
+ fprintf(f, "== Private Data ==\n");
+ fprintf(f, " Config: { ring_size: %u }\n", cmd_q->qcfg.nb_desc);
+ fprintf(f, " Ring IOVA: %#lx\t%#lx\t%#lx\n", cmd_q->qbase_desc, cmd_q->qbase_addr, cmd_q->qbase_phys_addr);
+ fprintf(f, " Next write: %u\n", cmd_q->next_write);
+ fprintf(f, " Next read: %u\n", cmd_q->next_read);
+ fprintf(f, " current queue depth: %u\n", cmd_q->ring_buff_count);
+ fprintf(f, " }\n");
+ fprintf(f, " Key Stats { submitted: %"PRIu64", comp: %"PRIu64", failed: %"PRIu64" }\n",
+ cmd_q->stats.submitted,
+ cmd_q->stats.completed,
+ cmd_q->stats.errors);
+ return 0;
+}
+
+/* Translates AE4DMA ChanERRs to DMA error codes. */
+static inline enum rte_dma_status_code
+__translate_status_ae4dma_to_dma(enum ae4dma_dma_err status)
+{
+ AE4DMA_PMD_INFO("ae4dma desc status = %d\n", status);
+ /*
+ * to be modified for proper error mapping of ae4dma
+ */
+
+ switch (status) {
+ case AE4DMA_DMA_ERR_NO_ERR:
+ return RTE_DMA_STATUS_SUCCESSFUL;
+ case AE4DMA_DMA_ERR_INV_LEN:
+ return RTE_DMA_STATUS_INVALID_LENGTH;
+ case AE4DMA_DMA_ERR_INV_SRC:
+ return RTE_DMA_STATUS_INVALID_SRC_ADDR;
+ case AE4DMA_DMA_ERR_INV_DST:
+ return RTE_DMA_STATUS_INVALID_DST_ADDR;
+ case AE4DMA_DMA_ERR_INV_ALIGN:
+ return RTE_DMA_STATUS_DATA_POISION;
+ case AE4DMA_DMA_ERR_INV_HEADER:
+ case AE4DMA_DMA_ERR_INV_STATUS:
+ return RTE_DMA_STATUS_ERROR_UNKNOWN;
+ default:
+ return RTE_DMA_STATUS_ERROR_UNKNOWN;
+
+ }
+ return 0;
+}
+
+static inline uint16_t *get_static_ptr()
+{
+ static uint16_t *ptr = NULL;
+
+ ptr = (ptr == NULL) ? (uint16_t *)malloc(sizeof(uint16_t)) : ptr;
+ return ptr;
+}
+/*
+ * icans h/w queues for descriptor processed status returns total processed count of descriptor
+ *@param cmd_q
+ *@param maximum ops expected
+ *the ae4dma h/w queue info struct
+ *@param[out] failed_count
+ * transfer error count
+ * @return
+ * The number of operations that completed - both success and failes
+ */
+
+static inline uint16_t
+ae4dma_scan_hwq(struct ae4dma_cmd_queue *cmd_q, const uint16_t max_ops, uint16_t *failed_count)
+{
+ volatile struct ae4dma_desc *hw_desc;
+ uint32_t events_count = 0, fails = 0;
+ volatile uint32_t tail;
+ volatile uint32_t desc_status;
+ uint32_t retry_count = MAX_RETRY;
+ uint32_t sub_desc_cnt;
+
+ tail = cmd_q->next_read;
+
+ /* process all the submitted descriptors for the HW queue */
+ sub_desc_cnt = cmd_q->ring_buff_count;
+
+ if (max_ops < sub_desc_cnt)
+ sub_desc_cnt = max_ops;
+
+ while (sub_desc_cnt) {
+ desc_status = 0;
+ retry_count = MAX_RETRY;
+ do {
+ hw_desc = &cmd_q->qbase_desc[tail];
+ desc_status = hw_desc->dw1.status;
+
+ if (desc_status) {
+ if (desc_status != AE4DMA_DMA_DESC_COMPLETED) {
+ fails++;
+ AE4DMA_PMD_WARN("WARNING:Desc error code : %d\n", hw_desc->dw1.err_code);
+ }
+
+ if (cmd_q->ring_buff_count)
+ cmd_q->ring_buff_count--;
+
+ cmd_q->status[events_count] = hw_desc->dw1.err_code;
+ events_count++;
+ tail = (tail + 1) % AE4DMA_DESCRITPTORS_PER_CMDQ;
+ sub_desc_cnt--;
+ }
+ } while (!desc_status && retry_count--);
+
+ if (desc_status == 0)
+ break;
+ }
+
+ cmd_q->stats.completed += events_count;
+ cmd_q->stats.errors += fails;
+ cmd_q->next_read = tail;
+
+ *failed_count = fails;
+ return events_count;
+}
+
+/* Returns successfull operations count and sets error flag if any errors. */
+
+static uint16_t
+ae4dma_completed(void *dev_private, uint16_t qid __rte_unused, const uint16_t max_ops,
+ uint16_t *last_idx, bool *has_error)
+{
+
+ struct ae4dma_dmadev *ae4dma = dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+ const uint16_t read = cmd_q->next_read;
+ uint16_t cpl_count, sl_count;
+ *has_error = false;
+ uint16_t err_count = 0;
+
+ cpl_count = ae4dma_scan_hwq(cmd_q, max_ops, &err_count);
+
+ if (cpl_count > max_ops)
+ cpl_count = max_ops;
+ if (cpl_count <= max_ops)
+ *last_idx = (cmd_q->next_read - 1 + AE4DMA_DESCRITPTORS_PER_CMDQ) % AE4DMA_DESCRITPTORS_PER_CMDQ;
+
+ sl_count = cpl_count - err_count;
+
+ if (err_count)
+ *has_error = true;
+
+ return sl_count;
+}
+
+/* Returns detailed status information about operations that have been completed. */
+
+static uint16_t
+ae4dma_completed_status(void *dev_private, uint16_t qid __rte_unused,
+ uint16_t max_ops, uint16_t *last_idx, enum rte_dma_status_code *status)
+
+{
+ struct ae4dma_dmadev *ae4dma = dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+ const uint16_t read = cmd_q->next_read;
+ uint16_t cpl_count;
+ uint16_t i;
+ uint16_t err_count = 0;
+
+ cpl_count = ae4dma_scan_hwq(cmd_q, max_ops, &err_count);
+
+ if (cpl_count > max_ops)
+ cpl_count = max_ops;
+ if (cpl_count <= max_ops)
+ *last_idx = (cmd_q->next_read-1+AE4DMA_DESCRITPTORS_PER_CMDQ) % AE4DMA_DESCRITPTORS_PER_CMDQ;
+
+ if (likely(!err_count)) {
+ for (i = 0; i < cpl_count; i++)
+ status[i] = RTE_DMA_STATUS_SUCCESSFUL;
+ }
+ if (unlikely(err_count >= 1)) {
+ for (i = 0; i < cpl_count; i++)
+ status[i] = __translate_status_ae4dma_to_dma(cmd_q->status[i]);
+ }
+
+ return cpl_count;
+}
+
+/* Get the remaining capacity of the ring. */
+static uint16_t
+ae4dma_burst_capacity(const void *dev_private, uint16_t vchan __rte_unused)
+
+{
+ const struct ae4dma_dmadev *ae4dma = dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+ unsigned short size = cmd_q->qcfg.nb_desc - 1;
+ unsigned short read = cmd_q->next_read;
+ unsigned short write = cmd_q->next_write;
+ unsigned short space = size - (write - read);
+
+ return space;
+}
+
+/* Retrieve the generic stats of a DMA device. */
+static int
+ae4dma_stats_get(const struct rte_dma_dev *dev, uint16_t vchan __rte_unused,
+ struct rte_dma_stats *rte_stats, uint32_t size)
+{
+ const struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+ struct rte_dma_stats *stats = &cmd_q->stats;
+
+ if (size < sizeof(rte_stats))
+ return -EINVAL;
+ if (rte_stats == NULL)
+ return -EINVAL;
+
+ *rte_stats = *stats;
+ return 0;
+}
+
+/* Reset the generic stat counters for the DMA device. */
+static int
+ae4dma_stats_reset(struct rte_dma_dev *dev, uint16_t vchan __rte_unused)
+{
+ struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ struct ae4dma_cmd_queue *cmd_q = &ae4dma->cmd_q[hwq_id];
+
+ memset(&cmd_q->stats, 0, sizeof(cmd_q->stats));
+ return 0;
+}
+
+/* Check if the AE4DMA device is idle. */
+static int
+ae4dma_vchan_status(const struct rte_dma_dev *dev, uint16_t vchan __rte_unused,
+ enum rte_dma_vchan_status *status)
+{
+ struct ae4dma_dmadev *ae4dma = dev->fp_obj->dev_private;
+ struct ae4dma_cmd_queue *cmd_q;
+ uint32_t cmd_q_ctrl;
+
+ cmd_q = &ae4dma->cmd_q[0];
+ /* check here cmd_q is full/empty based on this set status */
+/*
+ * if (*to check*)
+ * *status = RTE_DMA_VCHAN_ACTIVE;
+ * else
+ * *status = RTE_DMA_VCHAN_IDLE;
+ *
+ * return 0;
+ */
+/*
+ * As of now returning -1, as this functionality is not
+ * supported by ae4dma and it's valid also as this status
+ * callback implemetaion by driver is optional.
+ */
+ return -1;
+}
+
+int
+ae4dma_add_queue(struct ae4dma_dmadev *dev, uint8_t qn)
+{
+ uint32_t dma_addr_lo, dma_addr_hi;
+ uint32_t q_per_eng = 0;
+ struct ae4dma_cmd_queue *cmd_q;
+ const struct rte_memzone *q_mz;
+ void *ae4dma_mmio_base_addr;
+ int i;
+ static int dev_id;
+
+ if (dev == NULL)
+ return -1;
+ dev->qidx = 0;
+ q_per_eng = AE4DMA_MAX_HW_QUEUES;
+ dev->io_regs = (void *)(dev->pci.mem_resource[AE4DMA_PCIE_BAR].addr);
+ ae4dma_mmio_base_addr = (uint8_t *) dev->io_regs;
+ /* Set the number of HW queues for this AE4DMA engine */
+ AE4DMA_WRITE_REG_OFFSET(ae4dma_mmio_base_addr, AE4DMA_COMMON_CONFIG_OFFSET, q_per_eng);
+ q_per_eng = AE4DMA_READ_REG_OFFSET(ae4dma_mmio_base_addr, AE4DMA_COMMON_CONFIG_OFFSET);
+ AE4DMA_PMD_INFO("AE4DMA queues per engine = %d\n", q_per_eng);
+
+ dev->id = dev_id++;
+ dev->cmd_q_count = 0;
+ i = qn;
+ /* Find available queues */
+
+ cmd_q = &dev->cmd_q[dev->cmd_q_count++];
+ cmd_q->id = i;
+ cmd_q->qidx = 0;
+ /* Queue_size: 32*sizeof(struct ae4dmadma_desc) */
+ cmd_q->qsize = AE4DMA_QUEUE_SIZE(AE4DMA_QUEUE_DESC_SIZE);
+
+ cmd_q->hwq_regs = (volatile struct ae4dma_hwq_regs *)dev->io_regs + (i + 1);
+ /* AE4DMA queue memory */
+ snprintf(cmd_q->memz_name, sizeof(cmd_q->memz_name),
+ "%s_%d_%s_%d_%s",
+ "ae4dma_dev",
+ (int)dev->id, "queue",
+ (int)cmd_q->id, "mem");
+ q_mz = ae4dma_queue_dma_zone_reserve(cmd_q->memz_name,
+ cmd_q->qsize, rte_socket_id());
+ cmd_q->qbase_addr = (void *)q_mz->addr;
+ cmd_q->qbase_desc = (void *)q_mz->addr;
+ cmd_q->qbase_phys_addr = q_mz->iova;
+
+ /* Max Index (cmd queue length) */
+ AE4DMA_WRITE_REG(&cmd_q->hwq_regs->max_idx, AE4DMA_DESCRITPTORS_PER_CMDQ);
+
+ /* Queue Enable */
+ AE4DMA_WRITE_REG(&cmd_q->hwq_regs->control_reg.control_raw, AE4DMA_CMD_QUEUE_ENABLE);
+
+ /* Disabling the interrupt */
+ AE4DMA_WRITE_REG(&cmd_q->hwq_regs->intr_status_reg.intr_status_raw, AE4DMA_DISABLE_INTR);
+
+ /* AE4DMA_WRITE_REG(&cmd_q->hwq_regs->write_idx,0); */
+ cmd_q->next_write = AE4DMA_READ_REG(&cmd_q->hwq_regs->write_idx);
+ /* while(AE4DMA_READ_REG(&cmd_q->hwq_regs->read_idx));*/
+
+ cmd_q->next_read = AE4DMA_READ_REG(&cmd_q->hwq_regs->read_idx);
+ cmd_q->ring_buff_count = 0;
+ /* Update the device registers with queue addresses. */
+ dma_addr_lo = low32_value(cmd_q->qbase_phys_addr);
+ AE4DMA_WRITE_REG(&cmd_q->hwq_regs->qbase_lo,
+ (uint32_t)dma_addr_lo);
+ dma_addr_hi = high32_value(cmd_q->qbase_phys_addr);
+ AE4DMA_WRITE_REG(&cmd_q->hwq_regs->qbase_hi,
+ (uint32_t)dma_addr_hi);
+ if (dev->cmd_q_count == 0) {
+ AE4DMA_PMD_ERR("Error in enabling HW queues.No HW queues available\n");
+ return -1;
+ }
+
+ return 0;
+}
+
+/* Create a dmadev(dpdk DMA device) */
+static int
+ae4dma_dmadev_create(const char *name, struct rte_pci_device *dev, uint8_t qn)
+{
+ static const struct rte_dma_dev_ops ae4dma_dmadev_ops = {
+ .dev_close = ae4dma_dev_close,
+ .dev_configure = ae4dma_dev_configure,
+ .dev_dump = ae4dma_dev_dump,
+ .dev_info_get = ae4dma_dev_info_get,
+ .dev_start = ae4dma_dev_start,
+ .dev_stop = ae4dma_dev_stop,
+ .stats_get = ae4dma_stats_get,
+ .stats_reset = ae4dma_stats_reset,
+ .vchan_status = ae4dma_vchan_status,
+ .vchan_setup = ae4dma_vchan_setup,
+ };
+
+ struct rte_dma_dev *dmadev = NULL;
+ struct ae4dma_dmadev *ae4dma = NULL;
+ char hwq_dev_name[RTE_DEV_NAME_MAX_LEN];
+
+ if (!name) {
+ AE4DMA_PMD_ERR("Invalid name of the device!");
+ return -EINVAL;
+ }
+ memset(hwq_dev_name, 0, sizeof(hwq_dev_name));
+ (void)snprintf(hwq_dev_name, sizeof(hwq_dev_name), "%s-ch%u", name, qn);
+
+ /* Allocate device structure. */
+ dmadev = rte_dma_pmd_allocate(hwq_dev_name, dev->device.numa_node, sizeof(struct ae4dma_dmadev));
+ if (dmadev == NULL) {
+ AE4DMA_PMD_ERR("Unable to allocate dma device");
+ return -ENOMEM;
+ }
+ dmadev->device = &dev->device;
+ dmadev->fp_obj->dev_private = dmadev->data->dev_private;
+ dmadev->dev_ops = &ae4dma_dmadev_ops;
+
+ dmadev->fp_obj->burst_capacity = ae4dma_burst_capacity;
+ dmadev->fp_obj->completed = ae4dma_completed;
+ dmadev->fp_obj->completed_status = ae4dma_completed_status;
+ dmadev->fp_obj->copy = ae4dma_enqueue_copy;
+ dmadev->fp_obj->fill = ae4dma_enqueue_fill;
+ dmadev->fp_obj->submit = ae4dma_submit;
+
+ ae4dma = dmadev->data->dev_private;
+ ae4dma->dmadev = dmadev;
+ /* ae4dma->qcfg.nb_desc = 0; */
+ ae4dma->pci = *dev;
+ /* ae4dma->io_regs = (void *)(dev->mem_resource[AE4DMA_PCIE_BAR].addr); */
+ /* device is valid, add queue details */
+ if (ae4dma_add_queue(ae4dma, qn))
+ goto init_error;
+ return 0;
+
+init_error:
+ AE4DMA_PMD_ERR("driver %s(): failed", __func__);
+ return -EFAULT;
+}
+
+/* Destroy a DMA device. */
+static int
+ae4dma_dmadev_destroy(const char *name)
+{
+ int ret;
+
+ if (!name) {
+ AE4DMA_PMD_ERR("Invalid device name");
+ return -EINVAL;
+ }
+
+ ret = rte_dma_pmd_release(name);
+ if (ret)
+ AE4DMA_PMD_DEBUG("Device cleanup failed");
+
+ return 0;
+}
+
+/* Probe DMA device. */
+static int
+ae4dma_dmadev_probe(struct rte_pci_driver *drv, struct rte_pci_device *dev)
+{
+ char name[32];
+ int ret;
+
+ rte_pci_device_name(&dev->addr, name, sizeof(name));
+ AE4DMA_PMD_INFO("Init %s on NUMA node %d", name, dev->device.numa_node);
+ dev->device.driver = &drv->driver;
+ for (uint8_t i = 0; i < AE4DMA_MAX_HW_QUEUES; i++) {
+ ret = ae4dma_dmadev_create(name, dev, i);
+ if (ret) {
+ AE4DMA_PMD_ERR("%s create dmadev %u failed!",
+ name, i);
+ break;
+ }
+ }
+
+ return ret;
+}
+
+/* Remove DMA device. */
+static int
+ae4dma_dmadev_remove(struct rte_pci_device *dev)
+{
+ char name[32];
+
+ rte_pci_device_name(&dev->addr, name, sizeof(name));
+
+ AE4DMA_PMD_INFO("Closing %s on NUMA node %d",
+ name, dev->device.numa_node);
+
+ return ae4dma_dmadev_destroy(name);
+}
+
+static const struct rte_pci_id pci_id_ae4dma_map[] = {
+ { RTE_PCI_DEVICE(AMD_VENDOR_ID, AE4DMA_DEVICE_ID) },
+ { .vendor_id = 0, /* sentinel */ },
+};
+
+static struct rte_pci_driver ae4dma_pmd_drv = {
+ .id_table = pci_id_ae4dma_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+ .probe = ae4dma_dmadev_probe,
+ .remove = ae4dma_dmadev_remove,
+};
+
+RTE_PMD_REGISTER_PCI(AE4DMA_PMD_NAME, ae4dma_pmd_drv);
+RTE_PMD_REGISTER_PCI_TABLE(AE4DMA_PMD_NAME, pci_id_ae4dma_map);
+RTE_PMD_REGISTER_KMOD_DEP(AE4DMA_PMD_NAME, "* igb_uio | uio_pci_generic | vfio-pci");
diff --git a/drivers/dma/ae4dma/ae4dma_hw_defs.h b/drivers/dma/ae4dma/ae4dma_hw_defs.h
new file mode 100644
index 0000000000..814723a737
--- /dev/null
+++ b/drivers/dma/ae4dma/ae4dma_hw_defs.h
@@ -0,0 +1,223 @@
+/* SPDX-License-Identifier: BSD-3.0-Clause
+ * Copyright(c) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ */
+
+#ifndef __AE4DMA_HW_DEFS_H__
+#define __AE4DMA_HW_DEFS_H__
+
+#include <rte_bus_pci.h>
+#include <rte_byteorder.h>
+#include <rte_io.h>
+#include <rte_pci.h>
+#include <rte_spinlock.h>
+#include <rte_memzone.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * utility macros for bit setting and genmask
+ */
+
+#define BIT(nr) (1 << (nr))
+
+#define BITS_PER_LONG (__SIZEOF_LONG__ * 8)
+#define GENMASK(h, l) (((~0UL) << (l)) & (~0UL >> (BITS_PER_LONG - 1 - (h))))
+
+/* ae4dma device details */
+#define AMD_VENDOR_ID 0x1022
+#define AE4DMA_DEVICE_ID 0x149b
+#define AE4DMA_PCIE_BAR 0
+
+/*
+ * An AE4DMA engine has 16 DMA queues. Each queue supports 32 descriptors
+ */
+
+#define AE4DMA_MAX_HW_QUEUES 2
+#define AE4DMA_QUEUE_START_INDEX 0
+#define AE4DMA_CMD_QUEUE_ENABLE 0x1
+
+/* Common to all queues */
+#define AE4DMA_COMMON_CONFIG_OFFSET 0x00
+
+#define AE4DMA_DISABLE_INTR 0x01
+
+
+/* temp defs added, need to remove if not required - start*/
+
+
+/* Address offset for virtual queue registers */
+#define CMD_Q_STATUS_INCR 0x1000
+
+/* Bit masks */
+
+#define CMD_Q_LEN 32
+#define CMD_Q_RUN BIT(0)
+#define CMD_Q_HALT BIT(1)
+#define CMD_Q_MEM_LOCATION BIT(2)
+#define CMD_Q_STATUS GENMASK(9, 7)
+#define CMD_Q_SIZE GENMASK(4, 0)
+#define CMD_Q_SHIFT GENMASK(1, 0)
+#define COMMANDS_PER_QUEUE 8192
+
+
+#define QUEUE_SIZE_VAL ((ffs(COMMANDS_PER_QUEUE) - 2) & \
+ CMD_Q_SIZE)
+#define Q_PTR_MASK (2 << (QUEUE_SIZE_VAL + 5) - 1)
+#define Q_DESC_SIZE sizeof(struct ae4dma_desc)
+#define Q_SIZE(n) (COMMANDS_PER_QUEUE * (n))
+
+#define INT_COMPLETION BIT(0)
+#define INT_ERROR BIT(1)
+#define INT_QUEUE_STOPPED BIT(2)
+#define INT_EMPTY_QUEUE BIT(3)
+#define SUPPORTED_INTERRUPTS (INT_COMPLETION | INT_ERROR)
+#define ALL_INTERRUPTS (INT_COMPLETION | INT_ERROR | \
+ INT_QUEUE_STOPPED)
+
+/* bitmap */
+enum {
+ BITS_PER_WORD = sizeof(unsigned long) * CHAR_BIT
+};
+
+#define WORD_OFFSET(b) ((b) / BITS_PER_WORD)
+#define BIT_OFFSET(b) ((b) % BITS_PER_WORD)
+
+#define AE4DMA_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#define AE4DMA_BITMAP_SIZE(nr) \
+ AE4DMA_DIV_ROUND_UP(nr, CHAR_BIT * sizeof(unsigned long))
+
+#define AE4DMA_BITMAP_FIRST_WORD_MASK(start) \
+ (~0UL << ((start) & (BITS_PER_WORD - 1)))
+#define AE4DMA_BITMAP_LAST_WORD_MASK(nbits) \
+ (~0UL >> (-(nbits) & (BITS_PER_WORD - 1)))
+
+#define __ae4dma_round_mask(x, y) ((typeof(x))((y)-1))
+#define ae4dma_round_down(x, y) ((x) & ~__ae4dma_round_mask(x, y))
+
+/* temp defs added, need to remove if not required - end*/
+
+/* Descriptor status */
+enum ae4dma_dma_status {
+ AE4DMA_DMA_DESC_SUBMITTED = 0,
+ AE4DMA_DMA_DESC_VALIDATED = 1,
+ AE4DMA_DMA_DESC_PROCESSED = 2,
+ AE4DMA_DMA_DESC_COMPLETED = 3,
+ AE4DMA_DMA_DESC_ERROR = 4,
+};
+
+/* Descriptor error-code */
+enum ae4dma_dma_err {
+ AE4DMA_DMA_ERR_NO_ERR = 0,
+ AE4DMA_DMA_ERR_INV_HEADER = 1,
+ AE4DMA_DMA_ERR_INV_STATUS = 2,
+ AE4DMA_DMA_ERR_INV_LEN = 3,
+ AE4DMA_DMA_ERR_INV_SRC = 4,
+ AE4DMA_DMA_ERR_INV_DST = 5,
+ AE4DMA_DMA_ERR_INV_ALIGN = 6,
+ AE4DMA_DMA_ERR_UNKNOWN = 7,
+};
+
+/* HW Queue status */
+enum ae4dma_hwqueue_status {
+ AE4DMA_HWQUEUE_EMPTY = 0,
+ AE4DMA_HWQUEUE_FULL = 1,
+ AE4DMA_HWQUEUE_NOT_EMPTY = 4
+};
+/*
+ * descriptor for AE4DMA commands
+ * 8 32-bit words:
+ * word 0: source memory type; destination memory type ; control bits
+ * word 1: desc_id; error code; status
+ * word 2: length
+ * word 3: reserved
+ * word 4: upper 32 bits of source pointer
+ * word 5: low 32 bits of source pointer
+ * word 6: upper 32 bits of destination pointer
+ * word 7: low 32 bits of destination pointer
+ */
+
+/* AE4DMA Descriptor - DWORD0 - Controls bits: Reserved for future use */
+#define AE4DMA_DWORD0_STOP_ON_COMPLETION BIT(0)
+#define AE4DMA_DWORD0_INTERRUPT_ON_COMPLETION BIT(1)
+#define AE4DMA_DWORD0_START_OF_MESSAGE BIT(3)
+#define AE4DMA_DWORD0_END_OF_MESSAGE BIT(4)
+#define AE4DMA_DWORD0_DESTINATION_MEMORY_TYPE GENMASK(5, 4)
+#define AE4DMA_DWORD0_SOURCE_MEMEORY_TYPE GENMASK(7, 6)
+
+#define AE4DMA_DWORD0_DESTINATION_MEMORY_TYPE_MEMORY 0x0
+#define AE4DMA_DWORD0_DESTINATION_MEMORY_TYPE_IOMEMORY (1<<4)
+#define AE4DMA_DWORD0_SOURCE_MEMEORY_TYPE_MEMORY 0x0
+#define AE4DMA_DWORD0_SOURCE_MEMEORY_TYPE_IOMEMORY (1<<6)
+
+struct ae4dma_desc_dword0 {
+ uint8_t byte0;
+ uint8_t byte1;
+ uint16_t timestamp;
+};
+
+struct ae4dma_desc_dword1 {
+ uint8_t status;
+ uint8_t err_code;
+ uint16_t desc_id;
+};
+
+struct ae4dma_desc {
+ struct ae4dma_desc_dword0 dw0;
+ struct ae4dma_desc_dword1 dw1;
+ uint32_t length;
+ uint32_t reserved;
+ uint32_t src_lo;
+ uint32_t src_hi;
+ uint32_t dst_lo;
+ uint32_t dst_hi;
+};
+
+/*
+ * Registers for each queue :4 bytes length
+ * Effective address : offset + reg
+ */
+
+struct ae4dma_hwq_regs {
+ union {
+ uint32_t control_raw;
+ struct {
+ uint32_t queue_enable: 1;
+ uint32_t reserved_internal: 31;
+ } control;
+ } control_reg;
+
+ union {
+ uint32_t status_raw;
+ struct {
+ uint32_t reserved0: 1;
+ uint32_t queue_status: 2; /* 0–empty, 1–full, 2–stopped, 3–error , 4–Not Empty */
+ uint32_t reserved1: 21;
+ uint32_t interrupt_type: 4;
+ uint32_t reserved2: 4;
+ } status;
+ } status_reg;
+
+ uint32_t max_idx;
+ uint32_t read_idx;
+ uint32_t write_idx;
+
+ union {
+ uint32_t intr_status_raw;
+ struct {
+ uint32_t intr_status: 1;
+ uint32_t reserved: 31;
+ } intr_status;
+ } intr_status_reg;
+
+ uint32_t qbase_lo;
+ uint32_t qbase_hi;
+
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* AE4DMA_HW_DEFS_H */
diff --git a/drivers/dma/ae4dma/ae4dma_internal.h b/drivers/dma/ae4dma/ae4dma_internal.h
new file mode 100644
index 0000000000..3a95dc5b91
--- /dev/null
+++ b/drivers/dma/ae4dma/ae4dma_internal.h
@@ -0,0 +1,138 @@
+/* SPDX-License-Identifier: BSD-3.0-Clause
+ * Copyright(c) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ */
+
+#ifndef _AE4DMA_INTERNAL_H_
+#define _AE4DMA_INTERNAL_H_
+
+#include "ae4dma_hw_defs.h"
+
+//#ifdef STRUCT_DEFINED_REGS
+ #define NO_OFFSET 0
+//#endif
+
+#define ENABLE_DEBUG_LOG 0
+/**
+ * upper_32_bits - return bits 32-63 of a number
+ * @n: the number we're accessing
+ */
+#define upper_32_bits(n) ((uint32_t)(((n) >> 16) >> 16))
+
+/**
+ * lower_32_bits - return bits 0-31 of a number
+ * @n: the number we're accessing
+ */
+#define lower_32_bits(n) ((uint32_t)((n) & 0xffffffff))
+
+#define AE4DMA_DESCRITPTORS_PER_CMDQ 32
+#define AE4DMA_QUEUE_DESC_SIZE sizeof(struct ae4dma_desc)
+#define AE4DMA_QUEUE_SIZE(n) (AE4DMA_DESCRITPTORS_PER_CMDQ * (n))
+
+/** AE4DMA registers Write/Read */
+static inline void ae4dma_pci_reg_write(void *base, int offset,
+ uint32_t value)
+{
+ volatile void *reg_addr = ((uint8_t *)base + offset);
+
+ rte_write32((rte_cpu_to_le_32(value)), reg_addr);
+}
+
+static inline uint32_t ae4dma_pci_reg_read(void *base, int offset)
+{
+ volatile void *reg_addr = ((uint8_t *)base + offset);
+
+ return rte_le_to_cpu_32(rte_read32(reg_addr));
+}
+
+#define AE4DMA_READ_REG_OFFSET(hw_addr, reg_offset) \
+ ae4dma_pci_reg_read(hw_addr, reg_offset)
+
+#define AE4DMA_WRITE_REG_OFFSET(hw_addr, reg_offset, value) \
+ ae4dma_pci_reg_write(hw_addr, reg_offset, value)
+
+
+#define AE4DMA_READ_REG(hw_addr) \
+ ae4dma_pci_reg_read(hw_addr, 0)
+
+#define AE4DMA_WRITE_REG(hw_addr, value) \
+ ae4dma_pci_reg_write(hw_addr, 0, value)
+
+static inline uint32_t
+low32_value(unsigned long addr)
+{
+ return ((uint64_t)addr) & 0x0ffffffff;
+}
+
+static inline uint32_t
+high32_value(unsigned long addr)
+{
+ return ((uint64_t)addr >> 32) & 0x00000ffff;
+}
+
+/**
+ * A structure describing a AE4DMA command queue.
+ */
+struct ae4dma_cmd_queue {
+ char *wr_src;
+ phys_addr_t wr_src_phy;
+ char *wr_dst;
+ phys_addr_t wr_dst_phy;
+
+ char memz_name[RTE_MEMZONE_NAMESIZE];
+ volatile struct ae4dma_hwq_regs *hwq_regs;
+
+ struct rte_dma_vchan_conf qcfg;
+ struct rte_dma_stats stats;
+
+ /* Queue address */
+ struct ae4dma_desc *qbase_desc;
+ void *qbase_addr;
+ phys_addr_t qbase_phys_addr;
+ enum ae4dma_dma_err status[AE4DMA_DESCRITPTORS_PER_CMDQ];
+
+ /* Queue identifier */
+ uint64_t id; /**< queue id */
+ uint64_t qidx; /**< queue index */
+ uint64_t qsize; /**< queue size */
+
+ /* Queue Statistics */
+ uint64_t tail;
+ uint32_t ring_buff_count;
+
+ unsigned short next_read;
+ unsigned short next_write;
+ unsigned short last_write; /* Used to compute submitted count. */
+
+ /* queue-page registers addr */
+ void *reg_base;
+
+} __rte_cache_aligned;
+
+struct ae4dma_dmadev {
+ struct rte_dma_dev *dmadev;
+// struct rte_dma_vchan_conf qcfg;
+
+ phys_addr_t status_addr;
+ phys_addr_t ring_addr;
+ void *io_regs;
+
+ int id; /**< ae4dma dev id on platform */
+// struct ae4dma_cmd_queue cmd_q[AE4DMA_MAX_HW_QUEUES]; /**< ae4dma queue */
+ struct ae4dma_cmd_queue cmd_q[1]; /**< ae4dma queue */
+ int cmd_q_count; /**< no. of ae4dma Queues */
+ struct rte_pci_device pci; /**< ae4dma pci identifier */
+ int qidx;
+};
+
+
+extern int ae4dma_pmd_logtype;
+
+#define AE4DMA_PMD_LOG(level, fmt, args...) rte_log(RTE_LOG_ ## level, \
+ ae4dma_pmd_logtype, "AE4DMA: %s(): " fmt "\n", __func__, ##args)
+
+#define AE4DMA_PMD_DEBUG(fmt, args...) AE4DMA_PMD_LOG(DEBUG, fmt, ## args)
+#define AE4DMA_PMD_INFO(fmt, args...) AE4DMA_PMD_LOG(INFO, fmt, ## args)
+#define AE4DMA_PMD_ERR(fmt, args...) AE4DMA_PMD_LOG(ERR, fmt, ## args)
+#define AE4DMA_PMD_WARN(fmt, args...) AE4DMA_PMD_LOG(WARNING, fmt, ## args)
+
+#endif /* _AE4DMA_INTERNAL_H_ */
diff --git a/drivers/dma/ae4dma/meson.build b/drivers/dma/ae4dma/meson.build
new file mode 100644
index 0000000000..e48ab0d561
--- /dev/null
+++ b/drivers/dma/ae4dma/meson.build
@@ -0,0 +1,7 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright 2024 Advanced Micro Devices, Inc. All rights reserved.
+
+build = dpdk_conf.has('RTE_ARCH_X86')
+reason = 'only supported on x86'
+sources = files('ae4dma_dmadev.c')
+deps += ['bus_pci', 'dmadev']
diff --git a/drivers/dma/meson.build b/drivers/dma/meson.build
index 358132759a..6039f12cab 100644
--- a/drivers/dma/meson.build
+++ b/drivers/dma/meson.build
@@ -9,6 +9,7 @@ drivers = [
'idxd',
'ioat',
'odm',
+ 'ae4dma',
'skeleton',
]
std_deps = ['dmadev']
diff --git a/drivers/raw/ntb/meson.build b/drivers/raw/ntb/meson.build
index 9096f2b25a..f07ddd4c5d 100644
--- a/drivers/raw/ntb/meson.build
+++ b/drivers/raw/ntb/meson.build
@@ -3,5 +3,6 @@
deps += ['rawdev', 'mbuf', 'mempool', 'pci', 'bus_pci']
sources = files('ntb.c',
+ 'ntb_hw_amd.c',
'ntb_hw_intel.c')
headers = files('rte_pmd_ntb.h')
diff --git a/drivers/raw/ntb/ntb.c b/drivers/raw/ntb/ntb.c
index 0ed4c14592..685a5b8c86 100644
--- a/drivers/raw/ntb/ntb.c
+++ b/drivers/raw/ntb/ntb.c
@@ -26,6 +26,8 @@
static const struct rte_pci_id pci_id_ntb_map[] = {
{ RTE_PCI_DEVICE(NTB_INTEL_VENDOR_ID, NTB_INTEL_DEV_ID_B2B_SKX) },
{ RTE_PCI_DEVICE(NTB_INTEL_VENDOR_ID, NTB_INTEL_DEV_ID_B2B_ICX) },
+ { RTE_PCI_DEVICE(NTB_AMD_VENDOR_ID, NTB_AMD_DEV_ID_PRI) },
+ { RTE_PCI_DEVICE(NTB_AMD_VENDOR_ID, NTB_AMD_DEV_ID_SEC) },
{ .vendor_id = 0, /* sentinel */ },
};
@@ -84,46 +86,98 @@ ntb_handshake_work(const struct rte_rawdev *dev)
ret = (*hw->ntb_ops->spad_write)(dev, SPAD_NUM_MWS, 1, hw->mw_cnt);
if (ret < 0)
return ret;
- for (i = 0; i < hw->mw_cnt; i++) {
- NTB_LOG(INFO, "Local %u mw size: 0x%"PRIx64"", i,
- hw->mw_size[i]);
- val = hw->mw_size[i] >> 32;
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_SZ_H + 2 * i,
- 1, val);
+
+ if ((hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_PRI) || (hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_SEC)) {
+ for (i = 0; i < hw->mw_cnt; i++) {
+ val = hw->mw_size[i] >> 32;
+ NTB_LOG(INFO, "SPAD_MW0_SZ_H Local %u mw size: 0x%"PRIx64" val 0x%x ", i,
+ hw->mw_size[i], val);
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_SZ_H + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ val = hw->mw_size[i];
+ NTB_LOG(INFO, "SPAD_MW0_SZ_L Local %u mw size: 0x%"PRIx64" val 0x%x ", i,
+ hw->mw_size[i], val);
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_SZ_L + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ }
+
+ NTB_LOG(INFO, "writing hw->queue_size %d", hw->queue_size);
+
+ /* Tell peer about the queue info and map memory to the peer. */
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_Q_SZ, 1, hw->queue_size);
if (ret < 0)
return ret;
- val = hw->mw_size[i];
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_SZ_L + 2 * i,
- 1, val);
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_NUM_QPS, 1,
+ hw->queue_pairs);
if (ret < 0)
return ret;
- }
- /* Tell peer about the queue info and map memory to the peer. */
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_Q_SZ, 1, hw->queue_size);
- if (ret < 0)
- return ret;
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_NUM_QPS, 1,
- hw->queue_pairs);
- if (ret < 0)
- return ret;
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_USED_MWS, 1,
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_USED_MWS, 1,
hw->used_mw_num);
- if (ret < 0)
- return ret;
- for (i = 0; i < hw->used_mw_num; i++) {
- val = (uint64_t)(size_t)(hw->mz[i]->addr) >> 32;
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_BA_H + 2 * i,
- 1, val);
+ NTB_LOG(INFO, "after spad_write SPAD_USED_MWS hw->used_mw_num %d", hw->used_mw_num);
+ if (ret < 0)
+ return ret;
+
+ for (i = 0; i < hw->used_mw_num; i++) {
+ val = (uint64_t)(size_t)(hw->mz[i]->addr) >> 32;
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_BA_H + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ NTB_LOG(INFO, "after spad_write SPAD_MW0_BA_H SPAD_MW0_BA_H + 2 * i 0x%x, val 0x%x", SPAD_MW0_BA_H + 2 * i, val);
+ val = (uint64_t)(size_t)(hw->mz[i]->addr);
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_BA_L + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ NTB_LOG(INFO, "after spad_write SPAD_MW0_BA_L SPAD_MW0_BA_L + 2 * i 0x%x, val 0x%x", SPAD_MW0_BA_L + 2 * i, val);
+ }
+ } else {
+ NTB_LOG(INFO, " in else part");
+ for (i = 0; i < hw->mw_cnt; i++) {
+ NTB_LOG(INFO, "Local %u mw size: 0x%"PRIx64"", i,
+ hw->mw_size[i]);
+ val = hw->mw_size[i] >> 32;
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_SZ_H + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ val = hw->mw_size[i];
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_SZ_L + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ }
+
+ /* Tell peer about the queue info and map memory to the peer. */
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_Q_SZ, 1, hw->queue_size);
if (ret < 0)
return ret;
- val = (uint64_t)(size_t)(hw->mz[i]->addr);
- ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_BA_L + 2 * i,
- 1, val);
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_NUM_QPS, 1,
+ hw->queue_pairs);
if (ret < 0)
return ret;
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_USED_MWS, 1,
+ hw->used_mw_num);
+ if (ret < 0)
+ return ret;
+ for (i = 0; i < hw->used_mw_num; i++) {
+ val = (uint64_t)(size_t)(hw->mz[i]->addr) >> 32;
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_BA_H + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ val = (uint64_t)(size_t)(hw->mz[i]->addr);
+ ret = (*hw->ntb_ops->spad_write)(dev, SPAD_MW0_BA_L + 2 * i,
+ 1, val);
+ if (ret < 0)
+ return ret;
+ }
}
-
for (i = 0; i < hw->used_mw_num; i++) {
ret = (*hw->ntb_ops->mw_set_trans)(dev, i, hw->mz[i]->iova,
hw->mz[i]->len);
@@ -150,55 +204,9 @@ ntb_dev_intr_handler(void *param)
uint8_t peer_mw_cnt;
int i = 0;
- if (hw->ntb_ops->db_read == NULL ||
- hw->ntb_ops->db_clear == NULL ||
- hw->ntb_ops->peer_db_set == NULL) {
- NTB_LOG(ERR, "Doorbell is not supported.");
- return;
- }
-
- db_bits = (*hw->ntb_ops->db_read)(dev);
- if (!db_bits)
- NTB_LOG(ERR, "No doorbells");
-
- /* Doorbell 0 is for peer device ready. */
- if (db_bits & 1) {
- NTB_LOG(INFO, "DB0: Peer device is up.");
- /* Clear received doorbell. */
- (*hw->ntb_ops->db_clear)(dev, 1);
- /**
- * Peer dev is already up. All mw settings are already done.
- * Skip them.
- */
- if (hw->peer_dev_up)
- return;
-
- if (hw->ntb_ops->spad_read == NULL) {
- NTB_LOG(ERR, "Scratchpad read is not supported.");
- return;
- }
-
- /* Check if mw setting on the peer is the same as local. */
- peer_mw_cnt = (*hw->ntb_ops->spad_read)(dev, SPAD_NUM_MWS, 0);
- if (peer_mw_cnt != hw->mw_cnt) {
- NTB_LOG(ERR, "Both mw cnt must be the same.");
- return;
- }
-
- for (i = 0; i < hw->mw_cnt; i++) {
- val_h = (*hw->ntb_ops->spad_read)
- (dev, SPAD_MW0_SZ_H + 2 * i, 0);
- val_l = (*hw->ntb_ops->spad_read)
- (dev, SPAD_MW0_SZ_L + 2 * i, 0);
- peer_mw_size = ((uint64_t)val_h << 32) | val_l;
- NTB_LOG(DEBUG, "Peer %u mw size: 0x%"PRIx64"", i,
- peer_mw_size);
- if (peer_mw_size != hw->mw_size[i]) {
- NTB_LOG(ERR, "Mw config must be the same.");
- return;
- }
- }
+ if ((hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_PRI) || (hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_SEC)) {
+ (*hw->ntb_ops->interrupt_handler)(param);
hw->peer_dev_up = 1;
@@ -212,42 +220,117 @@ ntb_dev_intr_handler(void *param)
NTB_LOG(ERR, "Handshake work failed.");
return;
}
-
/* To get the link info. */
if (hw->ntb_ops->get_link_status == NULL) {
NTB_LOG(ERR, "Not supported to get link status.");
return;
}
+
(*hw->ntb_ops->get_link_status)(dev);
- NTB_LOG(INFO, "Link is up. Link speed: %u. Link width: %u",
- hw->link_speed, hw->link_width);
return;
- }
- if (db_bits & (1 << 1)) {
- NTB_LOG(INFO, "DB1: Peer device is down.");
- /* Clear received doorbell. */
- (*hw->ntb_ops->db_clear)(dev, 2);
+ } else {
+ if (hw->ntb_ops->db_read == NULL ||
+ hw->ntb_ops->db_clear == NULL ||
+ hw->ntb_ops->peer_db_set == NULL) {
+ NTB_LOG(ERR, "Doorbell is not supported.");
+ return;
+ }
- /* Peer device will be down, So clean local side too. */
- ntb_link_cleanup(dev);
+ db_bits = (*hw->ntb_ops->db_read)(dev);
+ if (!db_bits)
+ NTB_LOG(ERR, "No doorbells");
- hw->peer_dev_up = 0;
- /* Response peer's dev_stop request. */
- (*hw->ntb_ops->peer_db_set)(dev, 2);
- return;
- }
+ /* Doorbell 0 is for peer device ready. */
+ if (db_bits & 1) {
+ /* Clear received doorbell. */
+ (*hw->ntb_ops->db_clear)(dev, 1);
- if (db_bits & (1 << 2)) {
- NTB_LOG(INFO, "DB2: Peer device agrees dev to be down.");
- /* Clear received doorbell. */
- (*hw->ntb_ops->db_clear)(dev, (1 << 2));
- hw->peer_dev_up = 0;
- return;
- }
+ /**
+ * * * * Peer dev is already up. All mw settings are already done.
+ * * * * Skip them.
+ * * * */
+ if (hw->peer_dev_up)
+ return;
+
+ if (hw->ntb_ops->spad_read == NULL) {
+ NTB_LOG(ERR, "Scratchpad read is not supported.");
+ return;
+ }
+
+ /* Check if mw setting on the peer is the same as local. */
+ peer_mw_cnt = (*hw->ntb_ops->spad_read)(dev, SPAD_NUM_MWS, 0);
+ if (peer_mw_cnt != hw->mw_cnt) {
+ NTB_LOG(ERR, "Both mw cnt must be the same. peer_mw_cnt %d hw->mw_cnt %d ", peer_mw_cnt, hw->mw_cnt);
+ return;
+ }
+
+ for (i = 0; i < hw->mw_cnt; i++) {
+ val_h = (*hw->ntb_ops->spad_read)
+ (dev, SPAD_MW0_SZ_H + 2 * i, 0);
+ val_l = (*hw->ntb_ops->spad_read)
+ (dev, SPAD_MW0_SZ_L + 2 * i, 0);
+ peer_mw_size = ((uint64_t)val_h << 32) | val_l;
+ NTB_LOG(DEBUG, "Peer %u mw size: 0x%"PRIx64"", i,
+ peer_mw_size);
+ if (peer_mw_size != hw->mw_size[i]) {
+ NTB_LOG(ERR, "Mw config must be the same.");
+ return;
+ }
+ }
+
+ hw->peer_dev_up = 1;
+
+ /**
+ * Handshake with peer. Spad_write & mw_set_trans only works
+ * when both devices are up. So write spad again when db is
+ * received. And set db again for the later device who may miss
+ * the 1st db.
+ */
+ if (ntb_handshake_work(dev) < 0) {
+ NTB_LOG(ERR, "Handshake work failed.");
+ return;
+ }
+
+ /* To get the link info. */
+ if (hw->ntb_ops->get_link_status == NULL) {
+ NTB_LOG(ERR, "Not supported to get link status.");
+ return;
+ }
- /* Clear other received doorbells. */
- (*hw->ntb_ops->db_clear)(dev, db_bits);
+ (*hw->ntb_ops->get_link_status)(dev);
+ NTB_LOG(INFO, "Link is up. Link speed: %u. Link width: %u",
+ hw->link_speed, hw->link_width);
+ return;
+ }
+
+ if (db_bits & (1 << 1)) {
+
+ NTB_LOG(INFO, "DB1: Peer device is down.");
+ /* Clear received doorbell. */
+ (*hw->ntb_ops->db_clear)(dev, 2);
+
+ /* Peer device will be down, So clean local side too. */
+ ntb_link_cleanup(dev);
+
+ hw->peer_dev_up = 0;
+ /* Response peer's dev_stop request. */
+ (*hw->ntb_ops->peer_db_set)(dev, 2);
+ return;
+ }
+
+ if (db_bits & (1 << 2)) {
+
+ NTB_LOG(INFO, "DB2: Peer device agrees dev to be down.");
+ /* Clear received doorbell. */
+ (*hw->ntb_ops->db_clear)(dev, (1 << 2));
+ hw->peer_dev_up = 0;
+ return;
+ }
+
+ /* Clear other received doorbells. */
+ (*hw->ntb_ops->db_clear)(dev, db_bits);
+ }
}
static int
@@ -320,8 +403,7 @@ ntb_rxq_setup(struct rte_rawdev *dev,
RTE_CACHE_LINE_SIZE,
dev->socket_id);
if (!rxq) {
- NTB_LOG(ERR, "Failed to allocate memory for "
- "rx queue data structure.");
+ NTB_LOG(ERR, "Failed to allocate memory for rx queue data structure.");
return -ENOMEM;
}
@@ -406,8 +488,7 @@ ntb_txq_setup(struct rte_rawdev *dev,
RTE_CACHE_LINE_SIZE,
dev->socket_id);
if (!txq) {
- NTB_LOG(ERR, "Failed to allocate memory for "
- "tx queue structure");
+ NTB_LOG(ERR, "Failed to allocate memory for tx queue structure");
return -ENOMEM;
}
@@ -438,14 +519,16 @@ ntb_txq_setup(struct rte_rawdev *dev,
prev = i;
}
- txq->tx_free_thresh = txq_conf->tx_free_thresh ?
- txq_conf->tx_free_thresh :
- NTB_DFLT_TX_FREE_THRESH;
- if (txq->tx_free_thresh >= txq->nb_tx_desc - 3) {
- NTB_LOG(ERR, "tx_free_thresh must be less than nb_desc - 3. "
- "(tx_free_thresh=%u qp_id=%u)", txq->tx_free_thresh,
- qp_id);
- return -EINVAL;
+ if (!((hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_PRI) || (hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_SEC))) {
+ txq->tx_free_thresh = txq_conf->tx_free_thresh ?
+ txq_conf->tx_free_thresh :
+ NTB_DFLT_TX_FREE_THRESH;
+ if (txq->tx_free_thresh >= txq->nb_tx_desc - 3) {
+ NTB_LOG(ERR, "tx_free_thresh must be less than nb_desc - 3. (tx_free_thresh=%u qp_id=%u)",
+ txq->tx_free_thresh,
+ qp_id);
+ return -EINVAL;
+ }
}
hw->tx_queues[qp_id] = txq;
@@ -495,6 +578,7 @@ static uint16_t
ntb_queue_count(struct rte_rawdev *dev)
{
struct ntb_hw *hw = dev->dev_private;
+
return hw->queue_pairs;
}
@@ -523,6 +607,7 @@ ntb_queue_init(struct rte_rawdev *dev, uint16_t qp_id)
bar_addr = (*hw->ntb_ops->get_peer_mw_addr)(dev, 0);
if (bar_addr == NULL)
return -EINVAL;
+
remote_hdr = (struct ntb_header *)
((size_t)bar_addr + hdr_offset);
@@ -536,6 +621,7 @@ ntb_queue_init(struct rte_rawdev *dev, uint16_t qp_id)
for (i = 0; i < rxq->nb_rx_desc - 1; i++) {
struct rte_mbuf *mbuf = rte_mbuf_raw_alloc(rxq->mpool);
+
if (unlikely(!mbuf)) {
NTB_LOG(ERR, "Failed to allocate mbuf for RX");
return -ENOMEM;
@@ -651,8 +737,7 @@ ntb_enqueue_bufs(struct rte_rawdev *dev,
goto end_of_tx;
}
if (txm->data_len > tx_item->len) {
- NTB_LOG(ERR, "Data length exceeds buf length."
- " Only %u data would be transmitted.",
+ NTB_LOG(ERR, "Data length exceeds buf length. Only %u data would be transmitted.",
tx_item->len);
txm->data_len = tx_item->len;
}
@@ -689,6 +774,7 @@ ntb_enqueue_bufs(struct rte_rawdev *dev,
end_of_tx:
if (nb_tx) {
uint16_t nb1, nb2;
+
if (nb_mbufs > txq->nb_tx_desc - last_used) {
nb1 = txq->nb_tx_desc - last_used;
nb2 = nb_mbufs - txq->nb_tx_desc + last_used;
@@ -795,6 +881,7 @@ ntb_dequeue_bufs(struct rte_rawdev *dev,
end_of_rx:
if (nb_rx) {
uint16_t nb1, nb2;
+
if (nb_mbufs > rxq->nb_rx_desc - last_avail) {
nb1 = rxq->nb_rx_desc - last_avail;
nb2 = nb_mbufs - rxq->nb_rx_desc + last_avail;
@@ -903,6 +990,7 @@ ntb_dev_start(struct rte_rawdev *dev)
uint64_t peer_base_h;
uint32_t i;
int ret;
+ void *bar_addr;
if (!hw->link_status || !hw->peer_dev_up)
return -EINVAL;
@@ -934,20 +1022,22 @@ ntb_dev_start(struct rte_rawdev *dev)
goto err_up;
}
- peer_val = (*hw->ntb_ops->spad_read)(dev, SPAD_Q_SZ, 0);
- if (peer_val != hw->queue_size) {
- NTB_LOG(ERR, "Inconsistent queue size! (local: %u peer: %u)",
- hw->queue_size, peer_val);
- ret = -EINVAL;
- goto err_up;
- }
+ if (!((hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_PRI) || (hw->pci_dev->id.device_id == NTB_AMD_DEV_ID_SEC))) {
+ peer_val = (*hw->ntb_ops->spad_read)(dev, SPAD_Q_SZ, 0);
+ if (peer_val != hw->queue_size) {
+ NTB_LOG(ERR, "Inconsistent queue size! (local: %u peer: %u)",
+ hw->queue_size, peer_val);
+ ret = -EINVAL;
+ goto err_up;
+ }
- peer_val = (*hw->ntb_ops->spad_read)(dev, SPAD_NUM_QPS, 0);
- if (peer_val != hw->queue_pairs) {
- NTB_LOG(ERR, "Inconsistent number of queues! (local: %u peer:"
- " %u)", hw->queue_pairs, peer_val);
- ret = -EINVAL;
- goto err_up;
+ peer_val = (*hw->ntb_ops->spad_read)(dev, SPAD_NUM_QPS, 0);
+ if (peer_val != hw->queue_pairs) {
+ NTB_LOG(ERR, "Inconsistent number of queues! (local: %u peer: %u)",
+ hw->queue_pairs, peer_val);
+ ret = -EINVAL;
+ goto err_up;
+ }
}
hw->peer_used_mws = (*hw->ntb_ops->spad_read)(dev, SPAD_USED_MWS, 0);
@@ -960,6 +1050,12 @@ ntb_dev_start(struct rte_rawdev *dev)
hw->peer_mw_base[i] = (peer_base_h << 32) + peer_base_l;
}
+ bar_addr = (*hw->ntb_ops->get_peer_mw_addr)(dev, 0);
+ if (bar_addr == NULL)
+ return -EINVAL;
+
+ hw->base_addr = bar_addr;
+ hw->base_addr_phys = (*hw->ntb_ops->get_peer_mw_addr_phys)(dev, 0);
dev->started = 1;
return 0;
@@ -1138,6 +1234,20 @@ ntb_attr_get(struct rte_rawdev *dev, const char *attr_name,
return 0;
}
+ if (!strncmp(attr_name, NTB_BASE_ADDR, NTB_ATTR_NAME_LEN)) {
+ *attr_value = hw->base_addr;
+ NTB_LOG(DEBUG, "Attribute (%s) Value (%" PRIu64 ")",
+ attr_name, *attr_value);
+ return 0;
+ }
+
+ if (!strncmp(attr_name, NTB_BASE_ADDR_PHYS, NTB_ATTR_NAME_LEN)) {
+ *attr_value = hw->base_addr_phys;
+ NTB_LOG(DEBUG, "Attribute (%s) Value (%" PRIu64 ")",
+ attr_name, *attr_value);
+ return 0;
+ }
+
if (!strncmp(attr_name, NTB_LINK_STATUS_NAME, NTB_ATTR_NAME_LEN)) {
/* hw->link_status only indicates hw link status. */
*attr_value = hw->link_status && hw->peer_dev_up;
@@ -1382,6 +1492,11 @@ ntb_init_hw(struct rte_rawdev *dev, struct rte_pci_device *pci_dev)
case NTB_INTEL_DEV_ID_B2B_ICX:
hw->ntb_ops = &intel_ntb_ops;
break;
+
+ case NTB_AMD_DEV_ID_PRI:
+ case NTB_AMD_DEV_ID_SEC:
+ hw->ntb_ops = &amd_ntb_ops;
+ break;
default:
NTB_LOG(ERR, "Not supported device.");
return -EINVAL;
diff --git a/drivers/raw/ntb/ntb.h b/drivers/raw/ntb/ntb.h
index 8c7a2230f9..70b3d7ca5b 100644
--- a/drivers/raw/ntb/ntb.h
+++ b/drivers/raw/ntb/ntb.h
@@ -15,11 +15,14 @@ extern int ntb_logtype;
/* Vendor ID */
#define NTB_INTEL_VENDOR_ID 0x8086
+#define NTB_AMD_VENDOR_ID 0x1022
/* Device IDs */
#define NTB_INTEL_DEV_ID_B2B_SKX 0x201C
#define NTB_INTEL_DEV_ID_B2B_ICX 0x347E
#define NTB_INTEL_DEV_ID_B2B_SPR 0x347E
+#define NTB_AMD_DEV_ID_PRI 0x14C0
+#define NTB_AMD_DEV_ID_SEC 0x14C3
/* Reserved to app to use. */
#define NTB_SPAD_USER "spad_user_"
@@ -40,6 +43,8 @@ enum ntb_xstats_idx {
enum ntb_topo {
NTB_TOPO_NONE = 0,
+ NTB_TOPO_PRI,
+ NTB_TOPO_SEC,
NTB_TOPO_B2B_USD,
NTB_TOPO_B2B_DSD,
};
@@ -88,6 +93,7 @@ enum ntb_spad_idx {
* NTB device operations
* @ntb_dev_init: Init ntb dev.
* @get_peer_mw_addr: To get the addr of peer mw[mw_idx].
+ * @get_peer_mw_addr_phys: To get the physical addr of peer mw[mw_idx].
* @mw_set_trans: Set translation of internal memory that remote can access.
* @ioremap: Translate the remote host address to bar address.
* @get_link_status: get link status, link speed and link width.
@@ -104,6 +110,7 @@ enum ntb_spad_idx {
struct ntb_dev_ops {
int (*ntb_dev_init)(const struct rte_rawdev *dev);
void *(*get_peer_mw_addr)(const struct rte_rawdev *dev, int mw_idx);
+ uint64_t (*get_peer_mw_addr_phys)(const struct rte_rawdev *dev, int mw_idx);
int (*mw_set_trans)(const struct rte_rawdev *dev, int mw_idx,
uint64_t addr, uint64_t size);
void *(*ioremap)(const struct rte_rawdev *dev, uint64_t addr);
@@ -119,6 +126,7 @@ struct ntb_dev_ops {
int (*peer_db_set)(const struct rte_rawdev *dev, uint8_t db_bit);
int (*vector_bind)(const struct rte_rawdev *dev, uint8_t intr,
uint8_t msix);
+ int (*interrupt_handler)(void *param);
};
struct ntb_desc {
@@ -209,6 +217,8 @@ struct ntb_hw {
const struct ntb_dev_ops *ntb_ops;
struct rte_pci_device *pci_dev;
+ /**< PMD-specific private data */
+ rte_rawdev_obj_t pmd_private;
char *hw_addr;
uint8_t peer_dev_up;
@@ -216,6 +226,9 @@ struct ntb_hw {
/* remote mem base addr */
uint64_t *peer_mw_base;
+ void *base_addr;
+ uint64_t base_addr_phys;
+
uint16_t queue_pairs;
uint16_t queue_size;
uint32_t hdr_size_per_queue;
@@ -236,4 +249,6 @@ struct ntb_hw {
int spad_user_list[NTB_SPAD_USER_MAX_NUM];
};
+extern const struct ntb_dev_ops amd_ntb_ops;
+
#endif /* _NTB_H_ */
diff --git a/drivers/raw/ntb/ntb_hw_amd.c b/drivers/raw/ntb/ntb_hw_amd.c
new file mode 100644
index 0000000000..b616551e81
--- /dev/null
+++ b/drivers/raw/ntb/ntb_hw_amd.c
@@ -0,0 +1,738 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2025 Advanced Micro Devices, Inc. All Rights Reserved.
+ */
+
+#include <stdint.h>
+#include <stdio.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <rte_io.h>
+#include <rte_eal.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
+#include <bus_pci_driver.h>
+#include <rte_rawdev.h>
+#include <rte_rawdev_pmd.h>
+
+#include "ntb.h"
+#include "ntb_hw_amd.h"
+#include "rte_bus_pci.h"
+
+
+static int
+amd_ntb_dev_init(const struct rte_rawdev *dev)
+{
+ uint32_t ntb_topo;
+ int i, j, bar;
+
+ struct ntb_hw *ntb = dev->dev_private;
+
+ struct amd_ntb_hw *amd_hw = NULL;
+
+ ntb->pmd_private = rte_zmalloc("amd_hw", sizeof(struct amd_ntb_hw), 0);
+ if (ntb->pmd_private == NULL) {
+ NTB_LOG(ERR, "Failed to allocate memory for amd_hw.");
+ return -EINVAL;
+ }
+
+ amd_hw = ntb->pmd_private;
+ if (amd_hw == NULL) {
+ NTB_LOG(ERR, "Invalid device.");
+ return -EINVAL;
+ }
+
+ ntb->hw_addr = (char *)ntb->pci_dev->mem_resource[0].addr;
+
+ amd_hw->self_mmio = ntb->hw_addr;
+ amd_hw->peer_mmio = amd_hw->self_mmio + AMD_PEER_OFFSET;
+ amd_hw->int_mask = AMD_EVENT_INTMASK;
+
+ // Bit 0 of link status register
+ ntb_topo = rte_read32(ntb->hw_addr + AMD_SIDEINFO_OFFSET);
+
+ if (ntb_topo & AMD_SIDE_MASK)
+ ntb->topo = NTB_TOPO_SEC;
+ else
+ ntb->topo = NTB_TOPO_PRI;
+
+ NTB_LOG(INFO, "dev_init() - ntb_topo 0x%x amd_hw->ntb->topo %d", ntb_topo, ntb->topo);
+ NTB_LOG(INFO, "dev_init() - amd_hw-> self_mmio 0x%x amd_hw->peermmio 0x%x", amd_hw->self_mmio, amd_hw->peer_mmio);
+
+ // memory window, door bell, scratcpad count
+ ntb->mw_cnt = AMD_MW_COUNT;
+ ntb->db_cnt = AMD_DB_COUNT;
+ ntb->spad_cnt = AMD_SPADS_COUNT;
+
+ ntb->mw_size = rte_zmalloc("ntb_mw_size",
+ ntb->mw_cnt * sizeof(uint64_t), 0);
+ if (ntb->mw_size == NULL) {
+ NTB_LOG(ERR, "Cannot allocate memory for mw size.");
+ return -ENOMEM;
+ }
+ for (i = 0; i < ntb->mw_cnt; i++) {
+ bar = amd_ntb_bar[i];
+ ntb->mw_size[i] = ntb->pci_dev->mem_resource[bar].len;
+ NTB_LOG(INFO, "dev_init() - bar %d ntb->mw_size[%d] - 0x%x ", bar, i, ntb->mw_size[i]);
+ }
+
+ // splitting the scratchpad registers to half and assign it to each side
+
+// ntb->spad_cnt >>= 1; //// commenting for now.
+ if (ntb->topo == NTB_TOPO_PRI) {
+ amd_hw->self_spad = 0;
+ amd_hw->peer_spad = 0x20;
+ } else {
+ amd_hw->self_spad = 0x20;
+ amd_hw->peer_spad = 0;
+ }
+
+ // Reserve the last 2 spad registers for users.
+ for (j = 0; j < NTB_SPAD_USER_MAX_NUM; j++)
+ ntb->spad_user_list[j] = ntb->spad_cnt;
+
+ ntb->spad_user_list[0] = ntb->spad_cnt - 2;
+ ntb->spad_user_list[1] = ntb->spad_cnt - 1;
+
+ return 0;
+}
+
+// Memory window address - particular bar
+static void *
+amd_ntb_get_peer_mw_addr(const struct rte_rawdev *dev, int mw_idx)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ void *bar_addr;
+ uint64_t phys_bar_addr;
+
+ if (mw_idx < 0 || mw_idx > ntb->mw_cnt)
+ return -EINVAL;
+
+ bar_addr = ntb->pci_dev->mem_resource[AMD_MW_IDX << mw_idx].addr;
+ phys_bar_addr = ntb->pci_dev->mem_resource[AMD_MW_IDX << mw_idx].phys_addr;
+ NTB_LOG(INFO, "get_peer_mw_addr - bar virtual addr 0x%llx", bar_addr);
+ NTB_LOG(INFO, "get_peer_mw_addr - bar phys addr 0x%llx", phys_bar_addr);
+
+ return bar_addr;
+}
+
+static uint64_t
+amd_ntb_get_peer_mw_addr_phys(const struct rte_rawdev *dev, int mw_idx)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ uint64_t phys_addr;
+
+ if (mw_idx < 0 || mw_idx > ntb->mw_cnt)
+ return -EINVAL;
+
+ phys_addr = ntb->pci_dev->mem_resource[AMD_MW_IDX << mw_idx].phys_addr;
+ NTB_LOG(INFO, "get_peer_mw_addr - bar phys addr 0x%llx", phys_addr);
+
+ return phys_addr;
+}
+// Setting the registers with memory window details (start addresses and sizes)
+static int
+amd_ntb_mw_set_trans(const struct rte_rawdev *dev, int mw_idx,
+ uint64_t addr, uint64_t size)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ void *mmio, *peer_mmio;
+ uint32_t bar, mw_size;
+ uint64_t reg_val, limit, base_addr;
+
+ if (mw_idx < 0 || mw_idx > ntb->mw_cnt)
+ return -EINVAL;
+
+ mmio = amd_hw->self_mmio;
+ peer_mmio = amd_hw->peer_mmio;
+
+ /* Set the limit if supported */
+ limit = size;
+
+ NTB_LOG(INFO, " mw_idx %d, size 0x%x ", mw_idx, size);
+
+ bar = AMD_MW_IDX << mw_idx;
+
+ mw_size = ntb->pci_dev->mem_resource[bar].len;
+ if (size > mw_size)
+ return -EINVAL;
+
+ base_addr = ntb->pci_dev->mem_resource[bar].addr;
+
+ NTB_LOG(INFO, "mmio 0x%x peer_mmio 0x%x, limit 0x%x, bar %d mw_size 0x%x, base addr 0x%x ",
+ mmio, peer_mmio, limit, bar, mw_size, base_addr);
+
+ /* bar = 2 */
+ if (bar == amd_ntb_bar[0]) {
+ rte_write64(addr, peer_mmio + AMD_BAR23_XLAT_OFFSET);
+ reg_val = rte_read64(peer_mmio + AMD_BAR23_XLAT_OFFSET);
+ if (reg_val != addr) {
+ rte_write64(0, peer_mmio + AMD_BAR23_XLAT_OFFSET);
+ return -EIO;
+ }
+
+ rte_write64(limit, peer_mmio + AMD_BAR23_LIMIT_OFFSET);
+ reg_val = rte_read64(peer_mmio + AMD_BAR23_LIMIT_OFFSET);
+ if (reg_val != limit) {
+ rte_write64(base_addr, mmio + AMD_BAR23_LIMIT_OFFSET);
+ rte_write64(0, peer_mmio + AMD_BAR23_LIMIT_OFFSET);
+ return -EIO;
+ }
+ } /* bar = 4 */
+ else if (bar == amd_ntb_bar[1]) {
+ rte_write64(addr, peer_mmio + AMD_BAR45_XLAT_OFFSET);
+ reg_val = rte_read64(peer_mmio + AMD_BAR45_XLAT_OFFSET);
+ if (reg_val != addr) {
+ rte_write64(0, peer_mmio + AMD_BAR45_XLAT_OFFSET);
+ return -EIO;
+ }
+
+ rte_write64(limit, peer_mmio + AMD_BAR45_LIMIT_OFFSET);
+ reg_val = rte_read64(peer_mmio + AMD_BAR45_LIMIT_OFFSET);
+ if (reg_val != limit) {
+ rte_write64(base_addr, mmio + AMD_BAR45_LIMIT_OFFSET);
+ rte_write64(0, peer_mmio + AMD_BAR45_LIMIT_OFFSET);
+ return -EIO;
+ }
+ }
+
+ return 0;
+}
+
+/* Updating the memory addresses */
+static void *
+amd_ntb_ioremap(const struct rte_rawdev *dev, uint64_t addr)
+{
+ void *mapped = NULL;
+ void *base;
+ uint32_t i;
+
+ struct ntb_hw *ntb = dev->dev_private;
+
+ if (ntb == NULL) {
+ NTB_LOG(ERR, "Invalid NTB device.");
+ return NULL;
+ }
+
+ for (i = 0; i < ntb->peer_used_mws; i++) {
+ if (addr >= ntb->peer_mw_base[i] &&
+ addr <= ntb->peer_mw_base[i] + ntb->mw_size[i]) {
+ base = amd_ntb_get_peer_mw_addr(dev, i);
+ mapped = (void *)(size_t)(addr - ntb->peer_mw_base[i] + (size_t)base);
+ break;
+ }
+ }
+ return mapped;
+}
+
+static void
+amd_clear_side_info_reg(struct amd_ntb_hw *amd_hw, bool peer)
+{
+ void *mmio = NULL;
+ uint32_t reg;
+
+ if (peer)
+ mmio = amd_hw->peer_mmio;
+ else
+ mmio = amd_hw->self_mmio;
+
+ reg = rte_read32(mmio + AMD_SIDEINFO_OFFSET);
+ if (reg & AMD_SIDE_READY) {
+ reg &= ~AMD_SIDE_READY;
+ rte_write32(reg, mmio + AMD_SIDEINFO_OFFSET);
+ rte_read32(mmio + AMD_SIDEINFO_OFFSET);
+ }
+}
+
+
+struct rte_pci_device *pci_upstream_bridge(struct rte_pci_device *pdev)
+{
+
+ struct rte_pci_device *dev;
+ struct rte_pci_bus *pci_bus = NULL;
+
+ if (!pdev) {
+ printf("Invalid PCI device\n");
+ return NULL;
+ }
+ pci_bus = (struct rte_pci_bus *) rte_bus_find_by_name("pci");
+ if (!pci_bus) {
+ printf("failed to get pci bus\n");
+ return NULL;
+ }
+
+
+ TAILQ_FOREACH(dev, &pci_bus->device_list, next) {
+ if (dev->id.class_id == PCI_CLASS_BRIDGE_PCI) {
+ if (dev->addr.bus == (pdev->addr.bus - 1)) {
+ printf("upstream bridge found for device %04x:%02x:%02x.%x\n",
+ pdev->addr.domain, pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+ return dev;
+ }
+ }
+ }
+ printf("No upstream bridge found for device %04x:%02x:%02x.%x\n",
+ pdev->addr.domain, pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+ return NULL;
+}
+
+int read_config(const char *bdf, int offset, void *buf, int size)
+{
+ char path[128];
+ int fd;
+
+ snprintf(path, sizeof(path), "/sys/bus/pci/devices/%s/config", bdf);
+ fd = open(path, O_RDONLY);
+ if (fd < 0) {
+ perror("open");
+ return -1;
+ }
+
+ if (lseek(fd, offset, SEEK_SET) != offset) {
+ perror("lseek");
+ close(fd);
+ return -1;
+ }
+
+ if (read(fd, buf, size) != size) {
+ perror("read");
+ close(fd);
+ return -1;
+ }
+
+ close(fd);
+ return 0;
+}
+
+// Find PCIe capability offset
+int find_pcie_capability(const char *bdf, uint8_t *cap_offset)
+{
+ uint8_t status;
+ uint8_t pos;
+
+ if (read_config(bdf, 0x06, &status, 1) != 0)
+ return -1;
+
+ if (!(status & 0x10)) {
+ fprintf(stderr, "Capabilities list not supported\n");
+ return -1;
+ }
+
+ if (read_config(bdf, 0x34, &pos, 1) != 0)
+ return -1;
+
+ while (pos) {
+ uint8_t cap_id, next;
+
+ if (read_config(bdf, pos, &cap_id, 1) != 0 ||
+ read_config(bdf, pos + 1, &next, 1) != 0)
+ return -1;
+ if (cap_id == RTE_PCI_CAP_ID_EXP) {
+ *cap_offset = pos;
+ printf("PCIe capability FOUND\n");
+ return 0;
+ }
+ pos = next;
+ }
+
+ fprintf(stderr, "PCIe capability not found\n");
+ return -1;
+}
+
+int amd_read_link_status(struct rte_pci_device *pci_swus, uint16_t *lnk_word)
+{
+ const char bdf[32];
+ uint8_t pcie_cap;
+ uint16_t linkctl, linksta;
+
+ rte_pci_device_name(&pci_swus->addr, bdf, sizeof(bdf));
+ printf("bdf = %s\n", bdf);
+
+ if (find_pcie_capability(bdf, &pcie_cap) != 0)
+ return EXIT_FAILURE;
+
+ if (read_config(bdf, pcie_cap + 0x10, &linkctl, 2) != 0 ||
+ read_config(bdf, pcie_cap + 0x12, &linksta, 2) != 0) {
+ return EXIT_FAILURE;
+ }
+
+ *lnk_word = linksta;
+
+ if (linkctl & 0x03)
+ printf("ASPM Enabled\n");
+ else
+ printf("ASPM Disabled\n");
+
+ printf("Link Speed : Gen%d Link Width : x%d\n", linksta & 0xF, (linksta >> 4) & 0x3F);
+ return 0;
+}
+
+static int
+amd_link_status(struct ntb_hw *ntb)
+{
+ struct rte_pci_device *pdev = ntb->pci_dev;
+ struct pci_dev *pci_swus = NULL;
+ struct pci_dev *pci_swds = NULL;
+ off_t pcie_cap_offset;
+ uint16_t lnk_word;
+ off_t off;
+
+ if (ntb == NULL) {
+ NTB_LOG(ERR, "Invalid device.");
+ return -EINVAL;
+ }
+ if (ntb->topo == NTB_TOPO_PRI) {
+ pcie_cap_offset = amd_read_link_status(pdev, &lnk_word);
+
+ if (pcie_cap_offset) {
+ NTB_LOG(ERR, "[%s()] failed to find the pcie capability", __func__);
+ return pcie_cap_offset;
+ }
+ } else if (ntb->topo == NTB_TOPO_SEC) {
+ pci_swds = pci_upstream_bridge(pdev);
+
+ if (pci_swds)
+ pci_swus = pci_upstream_bridge(pci_swds);
+
+ if (pci_swus)
+ pcie_cap_offset = amd_read_link_status(pci_swus, &lnk_word);
+
+ if (pcie_cap_offset) {
+ NTB_LOG(ERR, "[%s()] failed to find the pcie capability", __func__);
+ return pcie_cap_offset;
+ }
+
+ }
+
+ NTB_LOG(INFO, "&lnk_word %x", lnk_word);
+ ntb->link_status = lnk_word;
+ return 0;
+}
+
+static int
+amd_poll_link(struct ntb_hw *ntb)
+{
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ void *mmio = amd_hw->peer_mmio;
+ uint32_t reg;
+
+ reg = rte_read32(mmio + AMD_SIDEINFO_OFFSET);
+ reg &= AMD_SIDE_READY;
+
+ amd_hw->ctl_status = reg;
+
+ amd_link_status(ntb);
+
+ return amd_hw->ctl_status;
+}
+
+static int
+amd_link_is_up(struct ntb_hw *ntb)
+{
+ int ret;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+
+ ret = amd_poll_link(ntb);
+ if (ret) {
+ if (ntb->topo == NTB_TOPO_PRI) {
+ if ((amd_hw->peer_status & AMD_LINK_UP_EVENT) || (amd_hw->peer_status == 0))
+ return ret;
+ else if (amd_hw->peer_status & AMD_LINK_DOWN_EVENT) {
+ amd_clear_side_info_reg(amd_hw, true);
+ return 0;
+ }
+ } else /* NTB_TOPO_SEC */
+ return ret;
+ }
+ return 0;
+}
+
+/* Get link status */
+static int
+amd_ntb_get_link_status(const struct rte_rawdev *dev)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ int ret;
+
+ ret = amd_link_is_up(ntb);
+ if (ret) {
+ if (ntb->link_status) {
+ ntb->link_speed = AMD_NTB_LNK_STA_SPEED(ntb->link_status);
+ ntb->link_width = AMD_NTB_LNK_STA_WIDTH(ntb->link_status);
+ } else {
+ ntb->link_speed = NTB_SPEED_NONE;
+ ntb->link_width = NTB_WIDTH_NONE;
+ }
+ }
+
+ return 0;
+}
+
+static int
+amd_ntb_set_link(const struct rte_rawdev *dev, bool up)
+{
+ uint32_t ntb_ctl, ntb_side_ready;
+
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+
+ amd_hw->self_mmio = ntb->hw_addr;
+
+ void *mmio = amd_hw->self_mmio;
+
+ ntb_side_ready = rte_read32(mmio + AMD_SIDEINFO_OFFSET);
+ if (!(ntb_side_ready & AMD_SIDE_READY)) {
+ // AMD_SIDE_READY is a read-only register which tells topology (prim\secondary)
+ ntb_side_ready |= AMD_SIDE_READY;
+ rte_write32(ntb_side_ready, mmio + AMD_SIDEINFO_OFFSET);
+ }
+
+ ntb_ctl = rte_read32(mmio + AMD_CNTL_OFFSET);
+ ntb_ctl |= (PMM_REG_CTL | SMM_REG_CTL);
+ rte_write32(ntb_ctl, mmio + AMD_CNTL_OFFSET);
+
+ NTB_LOG(INFO, "ntb_ctl_reg 0x%x ", rte_read32(mmio + AMD_CNTL_OFFSET));
+
+ return 0;
+}
+
+// scratch pad registed read
+static uint32_t
+amd_ntb_spad_read(const struct rte_rawdev *dev, int spad, bool peer)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ uint32_t offset;
+ void *reg_addr;
+
+ if (spad < 0 || spad >= ntb->spad_cnt) {
+ NTB_LOG(ERR, "Invalid scratchpad count.");
+ return -EINVAL;
+ }
+
+ if (peer) {
+ offset = amd_hw->peer_spad;
+ NTB_LOG(INFO, "peer side");
+ } else
+ offset = amd_hw->self_spad;
+
+ return rte_read32(ntb->hw_addr + AMD_SPAD_OFFSET + offset + (spad << 2));
+}
+
+//scratchpad register write
+static int
+amd_ntb_spad_write(const struct rte_rawdev *dev, int spad,
+ bool peer, uint32_t spad_v)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ uint32_t offset;
+ void *reg_addr;
+
+ if (spad < 0 || spad >= ntb->spad_cnt) {
+ NTB_LOG(ERR, "Invalid scratchpad count.");
+ return -EINVAL;
+ }
+
+ if (peer)
+ offset = amd_hw->peer_spad + (spad << 2);
+ else
+ offset = amd_hw->self_spad + (spad << 2);
+
+ rte_write32(spad_v, ntb->hw_addr + AMD_SPAD_OFFSET + offset);
+
+ return 0;
+}
+
+static uint64_t
+amd_ntb_db_read(const struct rte_rawdev *dev)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ uint64_t db_bits;
+
+ if (amd_hw == NULL)
+ NTB_LOG(ERR, "Invalid device.");
+
+ db_bits = (uint64_t)rte_read16(amd_hw->self_mmio + AMD_DBSTAT_OFFSET);
+
+ return db_bits;
+}
+
+static int
+amd_ntb_db_clear(const struct rte_rawdev *dev, uint64_t db_bits)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+
+ if (amd_hw == NULL)
+ NTB_LOG(ERR, "Invalid device.");
+
+ rte_write16((uint16_t)db_bits, amd_hw->self_mmio + AMD_DBSTAT_OFFSET);
+ return 0;
+}
+
+static int
+amd_ntb_db_set_mask(const struct rte_rawdev *dev, uint64_t db_mask)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+
+ if (db_mask & (~ntb->db_valid_mask))
+ return -EINVAL;
+ ntb->db_mask |= db_mask;
+
+ rte_write16((uint16_t)ntb->db_mask, amd_hw->self_mmio + AMD_DBMASK_OFFSET);
+ return 0;
+}
+
+static int
+amd_ntb_peer_db_set(const struct rte_rawdev *dev, uint8_t db_idx)
+{
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+
+ if (((uint64_t)1 << db_idx) & ~ntb->db_valid_mask) {
+ NTB_LOG(ERR, "Invalid doorbell.");
+ return -EINVAL;
+ }
+
+ rte_write16((uint16_t)1 << db_idx, ntb->hw_addr + AMD_DBREQ_OFFSET);
+ return 0;
+}
+
+static void
+amd_ntb_dev_interrupt_handler(void *param)
+{
+ struct rte_rawdev *dev = (struct rte_rawdev *)param;
+ struct ntb_hw *ntb = dev->dev_private;
+ struct amd_ntb_hw *amd_hw = ntb->pmd_private;
+ uint64_t db_bits = 0;
+ uint32_t event_value, event_set, peer_mw_cnt;
+
+ db_bits = amd_ntb_db_read(dev);
+ if (!db_bits)
+ NTB_LOG(ERR, "No doorbells");
+
+ /* Doorbell 0 is for peer device ready. */
+ if (db_bits & 1) {
+ /* Clear received doorbell. */
+ amd_ntb_db_clear(dev, 1);
+
+ /**
+ * * Peer dev is already up. All mw settings are already done.
+ */
+ if (ntb->peer_dev_up)
+ return;
+
+ /* Check if mw setting on the peer is the same as local. */
+ peer_mw_cnt = amd_ntb_spad_read(dev, SPAD_NUM_MWS, 1);
+
+ if (peer_mw_cnt != ntb->mw_cnt) {
+ NTB_LOG(ERR, "Both mw cnt must be the same peer_mw_cnt %d, ntb->mw_cnt %d .", peer_mw_cnt, ntb->mw_cnt);
+ return;
+ }
+
+ ntb->peer_dev_up = 1;
+
+ /**
+ * * Handshake with peer. Spad_write & mw_set_trans only works
+ * * when both devices are up. So write spad again when db is
+ * * received. And set db again for the later device who may miss
+ * * the 1st db.
+ */
+
+ return;
+ }
+
+ if (db_bits & (1 << 1)) {
+ NTB_LOG(INFO, "DB1: Peer device is down.");
+
+ /* Clear received doorbell. */
+ amd_ntb_db_clear(dev, 2);
+
+ /* Peer device will be down, So clean local side too. */
+ ntb->peer_dev_up = 0;
+
+ /* Response peer's dev_stop request. */
+ amd_ntb_peer_db_set(dev, 2);
+ return;
+ }
+
+ if (db_bits & (1 << 2)) {
+ NTB_LOG(INFO, "DB2: Peer device agrees dev to be down.");
+
+ /* Clear received doorbell. */
+ amd_ntb_db_clear(dev, (1 << 2));
+ ntb->peer_dev_up = 0;
+
+ return;
+ }
+
+ /* Clear other received doorbells. */
+ amd_ntb_db_clear(dev, db_bits);
+
+ event_value = rte_read32(amd_hw->self_mmio + AMD_INTSTAT_OFFSET);
+
+ if (event_value != 0) {
+ event_value &= AMD_EVENT_INTMASK;
+
+ switch (event_value) {
+
+ case AMD_PEER_FLUSH_EVENT:
+ NTB_LOG(INFO, "Peer Flush Event occurred.");
+ break;
+ case AMD_PEER_RESET_EVENT:
+ case AMD_LINK_DOWN_EVENT:
+ case AMD_PEER_D3_EVENT:
+ case AMD_PEER_PMETO_EVENT:
+ case AMD_LINK_UP_EVENT:
+ event_set = rte_read32(amd_hw->self_mmio + AMD_SMUACK_OFFSET);
+ event_set |= event_value;
+ rte_write32(event_set, amd_hw->self_mmio + AMD_SMUACK_OFFSET);
+ break;
+
+ case AMD_PEER_D0_EVENT:
+ event_set = rte_read32(amd_hw->self_mmio + AMD_PMESTAT_OFFSET);
+ if (event_set & 0x0001)
+ NTB_LOG(INFO, "D0 Wakeup completed for NTB");
+ event_set = rte_read32(amd_hw->self_mmio + AMD_SMUACK_OFFSET);
+ event_set |= event_value;
+ rte_write32(event_set, amd_hw->self_mmio + AMD_SMUACK_OFFSET);
+ break;
+ default:
+ NTB_LOG(ERR, "Invalid Interrupt event.");
+ break;
+ }
+ }
+}
+
+static int
+amd_ntb_vector_bind(const struct rte_rawdev *dev, uint8_t intr, uint8_t msix)
+{
+ return 0;
+}
+
+
+
+/* operations for primary side of local ntb */
+const struct ntb_dev_ops amd_ntb_ops = {
+ .ntb_dev_init = amd_ntb_dev_init,
+ .get_peer_mw_addr = amd_ntb_get_peer_mw_addr,
+ .get_peer_mw_addr_phys = amd_ntb_get_peer_mw_addr_phys,
+ .mw_set_trans = amd_ntb_mw_set_trans,
+ .ioremap = amd_ntb_ioremap,
+ .get_link_status = amd_ntb_get_link_status,
+ .set_link = amd_ntb_set_link,
+ .spad_read = amd_ntb_spad_read,
+ .spad_write = amd_ntb_spad_write,
+ .db_read = amd_ntb_db_read,
+ .db_clear = amd_ntb_db_clear,
+ .db_set_mask = amd_ntb_db_set_mask,
+ .peer_db_set = amd_ntb_peer_db_set,
+ .vector_bind = amd_ntb_vector_bind,
+ .interrupt_handler = amd_ntb_dev_interrupt_handler,
+};
+
diff --git a/drivers/raw/ntb/ntb_hw_amd.h b/drivers/raw/ntb/ntb_hw_amd.h
new file mode 100644
index 0000000000..1346c6942c
--- /dev/null
+++ b/drivers/raw/ntb/ntb_hw_amd.h
@@ -0,0 +1,123 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2025 Advanced Micro Devices, Inc. All Rights Reserved.
+ */
+#ifndef _NTB_HW_AMD_H_
+#define _NTB_HW_AMD_H_
+
+#include <rte_bus_pci.h>
+#include <bus_driver.h>
+#include <bus_pci_driver.h>
+#include <rte_bus_pci.h>
+#include <rte_byteorder.h>
+#include <rte_io.h>
+#include <rte_pci.h>
+#include <rte_spinlock.h>
+#include <rte_memzone.h>
+
+
+/**
+ * Structure describing the PCI bus
+ */
+#if 1
+struct rte_pci_bus {
+ struct rte_bus bus; /**< Inherit the generic class */
+ RTE_TAILQ_HEAD(, rte_pci_device) device_list; /**< List of PCI devices */
+ RTE_TAILQ_HEAD(, rte_pci_driver) driver_list; /**< List of PCI drivers */
+};
+#endif
+
+extern struct rte_pci_bus rte_pci_bus;
+
+/* PCI Bus iterators */
+#define FOREACH_DEVICE_ON_PCIBUS(p) \
+ RTE_TAILQ_FOREACH(p, &(rte_pci_bus.device_list), next)
+
+#define FOREACH_DRIVER_ON_PCIBUS(p) \
+ RTE_TAILQ_FOREACH(p, &(rte_pci_bus.driver_list), next)
+
+
+#define BIT(nr) (1 << (nr))
+
+/* AMD NTB Regsiters */
+#define AMD_DB_COUNT 16
+#define AMD_MSIX_VECTOR_COUNT 24
+#define AMD_SPADS_COUNT 16
+#define AMD_MW_COUNT 2
+#define AMD_MW_IDX 2
+
+#define AMD_PEER_OFFSET 0x400
+#define AMD_SIDEINFO_OFFSET 0x408
+#define AMD_DBFM_OFFSET 0x450
+#define AMD_DBREQ_OFFSET 0x454
+#define AMD_DBMASK_OFFSET 0x45C
+#define AMD_DBSTAT_OFFSET 0x460
+#define AMD_INTMASK_OFFSET 0x470
+#define AMD_INTSTAT_OFFSET 0x474
+#define AMD_PMESTAT_OFFSET 0x480
+
+/* event type */
+#define AMD_PEER_FLUSH_EVENT BIT(0)
+#define AMD_PEER_RESET_EVENT BIT(1)
+#define AMD_PEER_D3_EVENT BIT(2)
+#define AMD_PEER_PMETO_EVENT BIT(3)
+#define AMD_PEER_D0_EVENT BIT(4)
+#define AMD_LINK_UP_EVENT BIT(5)
+#define AMD_LINK_DOWN_EVENT BIT(6)
+#define AMD_EVENT_INTMASK (AMD_PEER_FLUSH_EVENT | AMD_PEER_RESET_EVENT | \
+ AMD_PEER_D3_EVENT | AMD_PEER_PMETO_EVENT | AMD_PEER_D0_EVENT | \
+ AMD_LINK_UP_EVENT | AMD_LINK_DOWN_EVENT)
+
+
+/* Bar Registers */
+#define AMD_BAR23_XLAT_OFFSET 0x438
+#define AMD_BAR23_LIMIT_OFFSET 0x418
+#define AMD_BAR45_XLAT_OFFSET 0x440
+#define AMD_BAR45_LIMIT_OFFSET 0x420
+
+#define AMD_SIDE_MASK BIT(0)
+#define AMD_SIDE_READY BIT(1)
+
+#define AMD_SPAD_OFFSET 0x210
+#define AMD_CNTL_OFFSET 0x200
+
+#define PMM_REG_CTL BIT(21)
+#define SMM_REG_CTL BIT(20)
+
+/* SMU registers */
+#define AMD_SMUACK_OFFSET 0x4A0
+
+#define AMD_NTB_LNK_STA_ACTIVE_BIT 0x2000
+#define AMD_NTB_LNK_STA_SPEED_MASK 0x000f
+#define AMD_NTB_LNK_STA_WIDTH_MASK 0x03f0
+#define AMD_NTB_LNK_STA_ACTIVE(x) (!!((x) & AMD_NTB_LNK_STA_ACTIVE_BIT))
+#define AMD_NTB_LNK_STA_SPEED(x) ((x) & AMD_NTB_LNK_STA_SPEED_MASK)
+#define AMD_NTB_LNK_STA_WIDTH(x) (((x) & AMD_NTB_LNK_STA_WIDTH_MASK) >> 4)
+
+#define PCI_CLASS_BRIDGE_PCI 0x060400
+
+enum ntb_bar {
+ NTB_BAR23 = 2,
+ NTB_BAR45 = 4,
+};
+
+static enum ntb_bar amd_ntb_bar[] = {
+ NTB_BAR23,
+ NTB_BAR45,
+};
+
+struct amd_ntb_hw {
+
+ void *self_mmio;
+ void *peer_mmio;
+ unsigned int self_spad;
+ unsigned int peer_spad;
+
+ uint32_t peer_status;
+ uint32_t ctl_status;
+
+ uint32_t int_mask;
+
+
+};
+
+#endif
diff --git a/drivers/raw/ntb/ntb_hw_intel.c b/drivers/raw/ntb/ntb_hw_intel.c
index 9b4465176a..d9c45d6891 100644
--- a/drivers/raw/ntb/ntb_hw_intel.c
+++ b/drivers/raw/ntb/ntb_hw_intel.c
@@ -613,6 +613,26 @@ intel_ntb_vector_bind(const struct rte_rawdev *dev, uint8_t intr, uint8_t msix)
return 0;
}
+static void
+intel_ntb_dev_interrupt_handler(void *param)
+{
+ struct rte_rawdev *dev = (struct rte_rawdev *)param;
+ struct ntb_hw *hw = dev->dev_private;
+ uint32_t val_h, val_l;
+ uint64_t peer_mw_size;
+ uint64_t db_bits = 0;
+ uint8_t peer_mw_cnt;
+ int i = 0;
+
+ if (hw->ntb_ops->db_read == NULL ||
+ hw->ntb_ops->db_clear == NULL ||
+ hw->ntb_ops->peer_db_set == NULL) {
+ NTB_LOG(ERR, "Doorbell is not supported.");
+ return;
+ }
+}
+
+
/* operations for primary side of local ntb */
const struct ntb_dev_ops intel_ntb_ops = {
.ntb_dev_init = intel_ntb_dev_init,
@@ -628,4 +648,5 @@ const struct ntb_dev_ops intel_ntb_ops = {
.db_set_mask = intel_ntb_db_set_mask,
.peer_db_set = intel_ntb_peer_db_set,
.vector_bind = intel_ntb_vector_bind,
+ .interrupt_handler = intel_ntb_dev_interrupt_handler,
};
diff --git a/drivers/raw/ntb/rte_pmd_ntb.h b/drivers/raw/ntb/rte_pmd_ntb.h
index 6591ce7931..d9e5da2533 100644
--- a/drivers/raw/ntb/rte_pmd_ntb.h
+++ b/drivers/raw/ntb/rte_pmd_ntb.h
@@ -15,6 +15,8 @@
#define NTB_MW_CNT_NAME "mw_count"
#define NTB_DB_CNT_NAME "db_count"
#define NTB_SPAD_CNT_NAME "spad_count"
+#define NTB_BASE_ADDR "base_addr"
+#define NTB_BASE_ADDR_PHYS "base_addr_phy"
#define NTB_MAX_DESC_SIZE 1024
#define NTB_MIN_DESC_SIZE 64
diff --git a/examples/ntb/commands.list b/examples/ntb/commands.list
index a26b8acfa3..04fff5f66d 100644
--- a/examples/ntb/commands.list
+++ b/examples/ntb/commands.list
@@ -8,4 +8,4 @@ start # start pkt fwd between ntb and ethdev
stop # stop packet forwarding
show port stats # show statistics for all ports
clear port stats # clear all port statistics
-set fwd <(file-trans,iofwd,txonly,rxonly)>mode # set forwarding mode as file-trans|rxonly|txonly|iofwd
+set fwd <(file-trans,ntbperf,iofwd,txonly,rxonly)>mode # set forwarding mode as file-trans|ntbperf|rxonly|txonly|iofwd
diff --git a/examples/ntb/ntb_fwd.c b/examples/ntb/ntb_fwd.c
index 37d60208e3..ba1a400852 100644
--- a/examples/ntb/ntb_fwd.c
+++ b/examples/ntb/ntb_fwd.c
@@ -1,6 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause
* Copyright(c) 2019 Intel Corporation
*/
+
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
@@ -17,12 +18,41 @@
#include <rte_rawdev.h>
#include <rte_ethdev.h>
#include <rte_malloc.h>
+#include <rte_memory.h>
#include <rte_lcore.h>
#include <rte_cycles.h>
#include <rte_pmd_ntb.h>
#include <rte_mbuf_pool_ops.h>
+
+#include "lib/dmadev/rte_dmadev.h"
#include "commands.h"
+#define ALIGN_4K 4096
+#define TOTAL_ORDER (1048576)
+#define CHUNK_ORDER (1024)
+#define PBAR23 0x8df00000000
+#define PBAR23_XLAT 0x17b330000
+
+#define PBAR45 0x8dec0000000
+#define MAX_DMA_DEVS 64
+
+void *out_buf;
+uint64_t out_buf_phys;
+static uint8_t dmacopy;
+
+static void
+__rte_format_printf(3, 4)
+print_err(const char *func, int lineno, const char *format, ...)
+{
+ va_list ap;
+
+ fprintf(stderr, "In %s:%d - ", func, lineno);
+ va_start(ap, format);
+ vfprintf(stderr, format, ap);
+}
+#define ERR_RETURN(...) do { print_err(__func__, __LINE__, __VA_ARGS__); return -1; } while (0)
+
+
/* Per-port statistics struct */
struct __rte_cache_aligned ntb_port_statistics {
uint64_t tx;
@@ -44,8 +74,19 @@ struct ntb_fwd_lcore_conf {
uint8_t stopped;
};
+struct ntb_perf_conf {
+ void *outbuf;
+ uint64_t outbuf_phys;
+ uint16_t use_dma;
+ uint16_t dma_dev_id;
+};
+
+static uint16_t dma_dev_count;
+uint16_t dma_dev_ids[MAX_DMA_DEVS];
+
enum ntb_fwd_mode {
FILE_TRANS = 0,
+ NTBPERF,
RXONLY,
TXONLY,
IOFWD,
@@ -53,6 +94,7 @@ enum ntb_fwd_mode {
};
static const char *const fwd_mode_s[] = {
"file-trans",
+ "ntbperf",
"rxonly",
"txonly",
"iofwd",
@@ -85,7 +127,12 @@ static uint16_t tx_free_thresh;
#define NTB_MAX_PKT_BURST 32
#define NTB_DFLT_PKT_BURST 32
+#define NUM_BUFFERS_PER_LCORE (TOTAL_ORDER / CHUNK_ORDER)
+
static uint16_t pkt_burst = NTB_DFLT_PKT_BURST;
+static uint64_t total_order = TOTAL_ORDER;
+static uint64_t chunk_order = CHUNK_ORDER;
+
#define BURST_TX_RETRIES 64
@@ -114,19 +161,13 @@ cmd_help_parsed(__rte_unused void *parsed_result,
"\n"
"The following commands are currently available:\n\n"
"Control:\n"
- " quit :"
- " Quit the application.\n"
+ " quit : Quit the application.\n"
"\nTransmission:\n"
- " send [path] :"
- " Send [path] file. Only take effect in file-trans mode\n"
- " start :"
- " Start transmissions.\n"
- " stop :"
- " Stop transmissions.\n"
- " clear/show port stats :"
- " Clear/show port stats.\n"
- " set fwd file-trans/rxonly/txonly/iofwd :"
- " Set packet forwarding mode.\n"
+ " send [path] : Send [path] file. Only take effect in file-trans mode\n"
+ " start : Start transmissions.\n"
+ " stop : Stop transmissions.\n"
+ " clear/show port stats : Clear/show port stats.\n"
+ " set fwd file-trans/rxonly/txonly/iofwd/ntbperf : Set packet forwarding mode.\n"
);
}
@@ -356,6 +397,107 @@ start_polling_recv_file(void *param)
return 0;
}
+static int
+start_perf_data_trans(void *param)
+{
+ int32_t i, j, ret;
+ uint16_t dma_dev_id;
+ uint8_t *dst;
+ uint64_t start, end;
+ double cpu_time_seconds, cpu_time_ms;
+ double bandwidth_MBps;
+ int32_t total_buffers;
+ uint32_t current_dst_offset = 0;
+ uint32_t lcore_id = rte_lcore_id();
+
+ char mempool_name[RTE_MEMPOOL_NAMESIZE];
+ uint8_t *src_buffers[NUM_BUFFERS_PER_LCORE];
+ struct rte_mempool *src_mp = {NULL};
+ struct ntb_perf_conf *perf_conf = param;
+
+ total_buffers = NUM_BUFFERS_PER_LCORE;
+
+
+ snprintf(mempool_name, sizeof(mempool_name), "src_mempool_lcore%u", lcore_id);
+
+ src_mp = rte_mempool_create(mempool_name, total_buffers,
+ chunk_order, MEMPOOL_CACHE_SIZE, 0,
+ NULL, NULL, NULL, NULL,
+ rte_socket_id(), 0);
+ if (src_mp == NULL)
+ rte_exit(EXIT_FAILURE, "Failed to create mempool 'src_mempool' for socket %d: %s\n",
+ rte_socket_id(), rte_strerror(rte_errno));
+
+ for (i = 0; i < total_buffers; i++) {
+
+ ret = rte_mempool_get(src_mp, (void *)&src_buffers[i]);
+ if (ret < 0)
+ rte_exit(EXIT_FAILURE, "Error getting buffer from mempool %u: %s\n",
+ i, rte_strerror(rte_errno));
+ for (j = 0; j < chunk_order; j++)
+ src_buffers[i][j] = (uint8_t)(rand() % 256);
+ }
+
+ if (total_order == 0 || chunk_order == 0) {
+ printf("Error: Invalid parameters for copy.\n");
+ return -EINVAL;
+ }
+
+
+ start = rte_get_tsc_cycles();
+
+ if (perf_conf->use_dma == 0) {
+ printf(" its cpu_copy\n");
+ dst = perf_conf->outbuf;
+ printf("dst cpu = %lx\n", dst);
+
+ for (i = 0; i < total_buffers; i++) {
+ if (current_dst_offset + chunk_order > total_order) {
+ rte_memcpy(dst + current_dst_offset, src_buffers[i], total_order - current_dst_offset);
+ break;
+ }
+ rte_memcpy(dst + current_dst_offset, src_buffers[i], chunk_order);
+ current_dst_offset += chunk_order;
+ }
+ } else {
+ printf(" its dma_copy\n");
+ dst = perf_conf->outbuf_phys;
+ dma_dev_id = perf_conf->dma_dev_id;
+ for (i = 0; i < total_buffers; i++) {
+ if (current_dst_offset + chunk_order > total_order) {
+ rte_dma_copy(dma_dev_id, 0, rte_mempool_virt2iova(src_buffers[i]), dst + current_dst_offset,
+ total_order-current_dst_offset, RTE_DMA_OP_FLAG_SUBMIT);
+ rte_dma_completed(dma_dev_id, 0, 1, NULL, NULL);
+ break;
+ }
+
+ rte_dma_copy(dma_dev_id, 0, rte_mempool_virt2iova(src_buffers[i]), dst + current_dst_offset,
+ chunk_order, RTE_DMA_OP_FLAG_SUBMIT);
+ rte_dma_completed(dma_dev_id, 0, 1, NULL, NULL);
+ current_dst_offset += chunk_order;
+ }
+
+ }
+ end = rte_get_tsc_cycles();
+
+ cpu_time_seconds = (double)(end - start) / rte_get_tsc_hz();
+ cpu_time_ms = cpu_time_seconds * 1000.0;
+
+ bandwidth_MBps = (double)total_order / (cpu_time_seconds * 1024 * 1024);
+
+ if (perf_conf->use_dma)
+ printf("Data copied: %u bytes in %.3f ms. Bandwidth: %.2f MBytes/s on thread %u with dmadev %u\n", total_order, cpu_time_ms, bandwidth_MBps, lcore_id, dma_dev_id);
+ else
+ printf("Data copied: %u bytes in %.3f ms. Bandwidth: %.2f MBytes/s on thread %u\n", total_order, cpu_time_ms, bandwidth_MBps, lcore_id);
+
+ for (i = 0; i < total_buffers; i++)
+ rte_mempool_put(src_mp, src_buffers[i]);
+
+ rte_mempool_free(src_mp);
+ return 0;
+}
+
+
static int
start_iofwd_per_lcore(void *param)
{
@@ -538,10 +680,49 @@ start_txonly_per_lcore(void *param)
return 0;
}
+static int
+ntb_perf_config_dmadevs(int dev_id)
+{
+ char *dma_name;
+ int ret;
+ uint16_t vchan = 0;
+ struct rte_dma_info info;
+ const struct rte_dma_conf conf = { .nb_vchans = 1};
+ const struct rte_dma_vchan_conf qconf = {
+ .direction = RTE_DMA_DIR_MEM_TO_MEM,
+ .nb_desc = 32,
+ };
+
+ if (dev_id < 0) {
+ fprintf(stderr, "Error: Fail to find DMA %s.\n", dma_name);
+ return -1;
+ }
+
+ if (rte_dma_configure(dev_id, &conf) != 0)
+ ERR_RETURN("Error with rte_dma_configure()\n");
+
+ if (rte_dma_vchan_setup(dev_id, vchan, &qconf) < 0)
+ ERR_RETURN("Error with queue configuration\n");
+
+ ret = rte_dma_info_get(dev_id, &info);
+ if (ret != 0 || info.nb_vchans != 1)
+ ERR_RETURN("Error, no configured queues reported on device id %u\n", dev_id);
+
+ fprintf(stdout, "iinfo.devname= %s;info.nb_vchn = %d;info.min_desc = %d;info.max_desc = %d\n",
+ info.dev_name, info.nb_vchans, info.min_desc, info.max_desc);
+
+ if (rte_dma_start(dev_id) != 0)
+ ERR_RETURN("Error with rte_dma_start()\n");
+
+ return dev_id;
+}
+
+
static int
ntb_fwd_config_setup(void)
{
uint16_t i;
+ int dma_dev_id;
/* Make sure iofwd has valid ethdev. */
if (fwd_mode == IOFWD && eth_port_id >= RTE_MAX_ETHPORTS) {
@@ -584,6 +765,33 @@ ntb_fwd_config_setup(void)
return 0;
}
+ if (fwd_mode == NTBPERF) {
+ if (dmacopy) {
+ dma_dev_count = rte_dma_count_avail();
+ printf("rte_dma_count_avail() = %d\n", dma_dev_count);
+ i = 0;
+ RTE_DMA_FOREACH_DEV(dma_dev_id) {
+ dma_dev_id = ntb_perf_config_dmadevs(dma_dev_id);
+ printf(" dma_dev_id => %d\n", dma_dev_id);
+ dma_dev_ids[i] = dma_dev_id;
+ i++;
+ }
+ }
+ num_queues = rte_lcore_count();
+ printf("num_queus = %u\n", num_queues);
+
+ fwd_streams = rte_zmalloc("ntb_fwd: fwd_streams",
+ sizeof(struct ntb_fwd_stream) * num_queues,
+ RTE_CACHE_LINE_SIZE);
+ for (i = 0; i < num_queues; i++) {
+ fwd_streams[i].qp_id = i;
+ fwd_streams[i].tx_port = RTE_MAX_ETHPORTS;
+ fwd_streams[i].rx_port = dev_id;
+ fwd_streams[i].tx_ntb = 1;
+ }
+ return 0;
+ }
+
if (fwd_mode == TXONLY) {
fwd_streams = rte_zmalloc("ntb_fwd: fwd_streams",
sizeof(struct ntb_fwd_stream) * num_queues,
@@ -647,8 +855,8 @@ assign_stream_to_lcores(void)
for (i = 0; i < conf->nb_stream; i++) {
fs = &fwd_streams[conf->stream_id + i];
if (fwd_mode == IOFWD)
- printf(" + Stream %u : %s%u RX -> %s%u TX,"
- " Q=%u\n", conf->stream_id + i,
+ printf(" + Stream %u : %s%u RX -> %s%u TX, Q=%u\n",
+ conf->stream_id + i,
fs->tx_ntb ? "Eth" : "NTB", fs->rx_port,
fs->tx_ntb ? "NTB" : "Eth", fs->tx_port,
fs->qp_id);
@@ -666,6 +874,7 @@ static void
start_pkt_fwd(void)
{
struct ntb_fwd_lcore_conf *conf;
+ struct ntb_perf_conf *perf_conf;
struct rte_eth_link eth_link;
uint32_t lcore_id;
int ret, i;
@@ -705,10 +914,12 @@ start_pkt_fwd(void)
assign_stream_to_lcores();
in_test = 1;
+ i = 0;
RTE_LCORE_FOREACH_WORKER(lcore_id) {
conf = &fwd_lcore_conf[lcore_id];
+
if (!conf->nb_stream)
continue;
@@ -716,7 +927,27 @@ start_pkt_fwd(void)
if (fwd_mode == FILE_TRANS)
rte_eal_remote_launch(start_polling_recv_file,
conf, lcore_id);
- else if (fwd_mode == IOFWD)
+ else if (fwd_mode == NTBPERF) {
+ perf_conf = (struct ntb_perf_conf *)
+ malloc(sizeof(struct ntb_perf_conf));
+ if (dmacopy == 0) {
+ perf_conf->use_dma = 0;
+ perf_conf->dma_dev_id = -1;
+ perf_conf->outbuf = out_buf;
+ } else if (dmacopy == 1) {
+ perf_conf->use_dma = 1;
+ perf_conf->outbuf_phys = out_buf_phys;
+ perf_conf->dma_dev_id = dma_dev_ids[i++];
+ printf("perf_conf.dma_dev_id = %d\n", perf_conf->dma_dev_id);
+ if (i >= dma_dev_count) {
+ printf("dma_dev_count crossed -> i = %d\n", i);
+ i = 0;
+ }
+
+ }
+ rte_eal_remote_launch(start_perf_data_trans,
+ (void *)perf_conf, lcore_id);
+ } else if (fwd_mode == IOFWD)
rte_eal_remote_launch(start_iofwd_per_lcore,
conf, lcore_id);
else if (fwd_mode == RXONLY)
@@ -947,6 +1178,9 @@ signal_handler(int signum)
#define OPT_TXFREET "txfreet"
#define OPT_BURST "burst"
#define OPT_QP "qp"
+#define OPT_USE_DMA "use_dma"
+#define OPT_TOTAL_ORDER "total_order"
+#define OPT_CHUNK_ORDER "chunk_order"
enum {
/* long options mapped to a short option */
@@ -957,6 +1191,9 @@ enum {
OPT_TXFREET_NUM,
OPT_BURST_NUM,
OPT_QP_NUM,
+ OPT_USE_DMA_NUM,
+ OPT_TOTAL_ORDER_NUM,
+ OPT_CHUNK_ORDER_NUM,
};
static const char short_options[] =
@@ -970,6 +1207,9 @@ static const struct option lgopts[] = {
{OPT_TXFREET, 1, NULL, OPT_TXFREET_NUM },
{OPT_BURST, 1, NULL, OPT_BURST_NUM },
{OPT_QP, 1, NULL, OPT_QP_NUM },
+ {OPT_USE_DMA, 1, NULL, OPT_USE_DMA_NUM },
+ {OPT_TOTAL_ORDER, 1, NULL, OPT_TOTAL_ORDER_NUM },
+ {OPT_CHUNK_ORDER, 1, NULL, OPT_CHUNK_ORDER_NUM },
{0, 0, NULL, 0 }
};
@@ -979,14 +1219,12 @@ ntb_usage(const char *prgname)
printf("%s [EAL options] -- [options]\n"
"-i: run in interactive mode.\n"
"-qp=N: set number of queues as N (N > 0, default: 1).\n"
- "--fwd-mode=N: set fwd mode (N: file-trans | rxonly | "
- "txonly | iofwd, default: file-trans)\n"
- "--buf-size=N: set mbuf dataroom size as N (0 < N < 65535,"
- " default: 2048).\n"
- "--nb-desc=N: set number of descriptors as N (%u <= N <= %u,"
- " default: 1024).\n"
+ "--fwd-mode=N: set fwd mode (N: file-trans | rxonly | txonly | ntbperf | iofwd, default: file-trans)\n"
+ "--buf-size=N: set mbuf dataroom size as N (0 < N < 65535, default: 2048).\n"
+ "--nb-desc=N: set number of descriptors as N (%u <= N <= %u, default: 1024).\n"
"--txfreet=N: set tx free thresh for NTB driver as N. (N >= 0)\n"
"--burst=N: set pkt burst as N (0 < N <= %u default: 32).\n",
+ "--perfdma=N: set for ntb_perf to use DMA(1) or CPU (0) copy.\n",
prgname, NTB_MIN_DESC_SIZE, NTB_MAX_DESC_SIZE,
NTB_MAX_PKT_BURST);
}
@@ -1011,13 +1249,35 @@ ntb_parse_args(int argc, char **argv)
else
rte_exit(EXIT_FAILURE, "q must be > 0.\n");
break;
+ case OPT_USE_DMA_NUM:
+ n = atoi(optarg);
+ if (n == 0)
+ dmacopy = 0;
+ else if (n == 1)
+ dmacopy = 1;
+ else
+ rte_exit(EXIT_FAILURE, "pls select cpu copy or dma copy.\n");
+ break;
+ case OPT_TOTAL_ORDER_NUM:
+ n = atoi(optarg);
+ if (n > 0)
+ total_order = n;
+ else
+ rte_exit(EXIT_FAILURE, "pls enter a valid total order %d.\n", n);
+ break;
+ case OPT_CHUNK_ORDER_NUM:
+ n = atoi(optarg);
+ if (n > 0)
+ chunk_order = n;
+ else
+ rte_exit(EXIT_FAILURE, "pls enter a valid chunk order.\n");
+ break;
case OPT_BUF_SIZE_NUM:
n = atoi(optarg);
if (n > RTE_PKTMBUF_HEADROOM && n <= 0xFFFF)
ntb_buf_size = n;
else
- rte_exit(EXIT_FAILURE, "buf-size must be > "
- "%u and < 65536.\n",
+ rte_exit(EXIT_FAILURE, "buf-size must be > %u and < 65536.\n",
RTE_PKTMBUF_HEADROOM);
break;
case OPT_FWD_MODE_NUM:
@@ -1028,34 +1288,31 @@ ntb_parse_args(int argc, char **argv)
}
}
if (i == MAX_FWD_MODE)
- rte_exit(EXIT_FAILURE, "Unsupported mode. "
- "(Should be: file-trans | rxonly | txonly "
- "| iofwd)\n");
+ rte_exit(EXIT_FAILURE, "Unsupported mode. (Should be: file-trans \
+ | rxonly | txonly | iofwd | ntbperf)\n");
break;
case OPT_NB_DESC_NUM:
n = atoi(optarg);
if (n >= NTB_MIN_DESC_SIZE && n <= NTB_MAX_DESC_SIZE)
nb_desc = n;
else
- rte_exit(EXIT_FAILURE, "nb-desc must be within"
- " [%u, %u].\n", NTB_MIN_DESC_SIZE,
- NTB_MAX_DESC_SIZE);
+ rte_exit(EXIT_FAILURE, "nb-desc must be within [%u, %u].\n",
+ NTB_MIN_DESC_SIZE, NTB_MAX_DESC_SIZE);
break;
case OPT_TXFREET_NUM:
n = atoi(optarg);
if (n >= 0)
tx_free_thresh = n;
else
- rte_exit(EXIT_FAILURE, "txfreet must be"
- " >= 0\n");
+ rte_exit(EXIT_FAILURE, "txfreet must be >= 0\n");
break;
case OPT_BURST_NUM:
n = atoi(optarg);
if (n > 0 && n <= NTB_MAX_PKT_BURST)
pkt_burst = n;
else
- rte_exit(EXIT_FAILURE, "burst must be within "
- "(0, %u].\n", NTB_MAX_PKT_BURST);
+ rte_exit(EXIT_FAILURE, "burst must be within (0, %u].\n",
+ NTB_MAX_PKT_BURST);
break;
default:
@@ -1072,6 +1329,7 @@ ntb_mempool_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
void *opaque)
{
const struct rte_memzone *mz = opaque;
+
rte_memzone_free(mz);
}
@@ -1118,9 +1376,8 @@ ntb_mbuf_pool_create(uint16_t mbuf_seg_size, uint32_t nb_mbuf,
/* Put ntb header on mw0. */
if (ntb_info.mw_size[0] < ntb_info.ntb_hdr_size) {
- printf("mw0 (size: %" PRIu64 ") is not enough for ntb hdr"
- " (size: %u)\n", ntb_info.mw_size[0],
- ntb_info.ntb_hdr_size);
+ printf("mw0 (size: %" PRIu64 ") is not enough for ntb hdr (size: %u)\n",
+ ntb_info.mw_size[0], ntb_info.ntb_hdr_size);
goto fail;
}
@@ -1141,10 +1398,10 @@ ntb_mbuf_pool_create(uint16_t mbuf_seg_size, uint32_t nb_mbuf,
if (!mz_len)
continue;
mz = rte_memzone_reserve_aligned(mz_name, mz_len, socket_id,
- RTE_MEMZONE_IOVA_CONTIG, align);
+ RTE_MEMZONE_IOVA_CONTIG, ALIGN_4K);
if (mz == NULL) {
- printf("Cannot allocate %" PRIu64 " aligned memzone"
- " %u\n", align, mz_id);
+ printf("Cannot allocate %" PRIu64 " aligned memzone %u\n",
+ align, mz_id);
goto fail;
}
left_sz -= mz_len;
@@ -1246,8 +1503,8 @@ main(int argc, char **argv)
ntb_rawdev_conf.dev_private = (rte_rawdev_obj_t)(&ntb_conf);
ret = rte_rawdev_configure(dev_id, &ntb_rawdev_conf, sizeof(ntb_conf));
if (ret)
- rte_exit(EXIT_FAILURE, "Can't config ntb dev: err=%d, "
- "port=%u\n", ret, dev_id);
+ rte_exit(EXIT_FAILURE, "Can't config ntb dev: err=%d, port=%u\n",
+ ret, dev_id);
ntb_q_conf.tx_free_thresh = tx_free_thresh;
ntb_q_conf.nb_desc = nb_desc;
@@ -1284,6 +1541,16 @@ main(int argc, char **argv)
/* Find 1st ethdev */
eth_port_id = rte_eth_find_next(0);
+ rte_rawdev_get_attr(dev_id, NTB_BASE_ADDR, (void *) &out_buf);
+ if (out_buf == 0)
+ printf("outbuf is NULL: ERROR\n");
+
+ rte_rawdev_get_attr(dev_id, NTB_BASE_ADDR_PHYS, (uint64_t)&out_buf_phys);
+ if (out_buf_phys == 0)
+ printf("outbuf_phys is NULL: ERROR\n");
+ else
+ printf("outbuf_phys is = 0x%llx\n", out_buf_phys);
+
if (eth_port_id < RTE_MAX_ETHPORTS) {
ret = rte_eth_dev_info_get(eth_port_id, ðdev_info);
if (ret)
@@ -1294,8 +1561,8 @@ main(int argc, char **argv)
ret = rte_eth_dev_configure(eth_port_id, num_queues,
num_queues, ð_pconf);
if (ret)
- rte_exit(EXIT_FAILURE, "Can't config ethdev: err=%d, "
- "port=%u\n", ret, eth_port_id);
+ rte_exit(EXIT_FAILURE, "Can't config ethdev: err=%d, port=%u\n",
+ ret, eth_port_id);
eth_rx_conf = ethdev_info.default_rxconf;
eth_rx_conf.offloads = eth_pconf.rxmode.offloads;
eth_tx_conf = ethdev_info.default_txconf;
@@ -1319,8 +1586,8 @@ main(int argc, char **argv)
ret = rte_eth_dev_start(eth_port_id);
if (ret < 0)
- rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, "
- "port=%u\n", ret, eth_port_id);
+ rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%u\n",
+ ret, eth_port_id);
}
/* initialize port stats */
@@ -1331,6 +1598,11 @@ main(int argc, char **argv)
printf("Set default fwd mode as iofwd.\n");
fwd_mode = IOFWD;
}
+ if (fwd_mode == NTBPERF) {
+ printf(" Set default fwd mode as data-tran (ntb_perf).\n");
+ fwd_mode = NTBPERF;
+ }
+
if (fwd_mode == MAX_FWD_MODE) {
printf("Set default fwd mode as file-trans.\n");
fwd_mode = FILE_TRANS;
diff --git a/usertools/dpdk-devbind.py b/usertools/dpdk-devbind.py
index 283707fc16..0ddea56323 100755
--- a/usertools/dpdk-devbind.py
+++ b/usertools/dpdk-devbind.py
@@ -66,6 +66,12 @@
'SVendor': None, 'SDevice': None}
intel_ntb_icx = {'Class': '06', 'Vendor': '8086', 'Device': '347e',
'SVendor': None, 'SDevice': None}
+amd_ntb_pri = {'Class': '06', 'Vendor': '1022', 'Device': '14c0',
+ 'SVendor': None, 'SDevice': None}
+
+amd_ntb_sec = {'Class': '06', 'Vendor': '1022', 'Device': '14c3',
+ 'SVendor': None, 'SDevice': None}
+
cnxk_sso = {'Class': '08', 'Vendor': '177d', 'Device': 'a0f9,a0fa',
'SVendor': None, 'SDevice': None}
@@ -74,6 +80,9 @@
cn9k_ree = {'Class': '08', 'Vendor': '177d', 'Device': 'a0f4',
'SVendor': None, 'SDevice': None}
+amd_ae4dma = {'Class': '08', 'Vendor': '1022', 'Device': '149b',
+ 'SVendor': None, 'SDevice': None}
+
virtio_blk = {'Class': '01', 'Vendor': "1af4", 'Device': '1001,1042',
'SVendor': None, 'SDevice': None}
@@ -83,7 +92,7 @@
network_devices = [network_class, cavium_pkx, avp_vnic, ifpga_class]
baseband_devices = [acceleration_class]
crypto_devices = [encryption_class, intel_processor_class]
-dma_devices = [cnxk_dma, hisilicon_dma,
+dma_devices = [amd_ae4dma, cnxk_dma, hisilicon_dma,
intel_idxd_spr, intel_ioat_bdw, intel_ioat_icx, intel_ioat_skx,
odm_dma]
eventdev_devices = [cavium_sso, cavium_tim, intel_dlb, cnxk_sso]
@@ -92,8 +101,8 @@
regex_devices = [cn9k_ree]
ml_devices = [cnxk_ml]
misc_devices = [cnxk_bphy, cnxk_bphy_cgx, cnxk_inl_dev,
- intel_ntb_skx, intel_ntb_icx,
- virtio_blk]
+ intel_ntb_skx, intel_ntb_icx, amd_ntb_pri,
+ amd_ntb_sec, virtio_blk]
# global dict ethernet devices present. Dictionary indexed by PCI address.
# Each device within this is itself a dictionary of device properties
--
2.34.1
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2025-07-05 14:10 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-07-05 14:10 [PATCH] ntb/ae4dma: added Amd userspace DMA and NTB drivers Bhagyada Modali
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).