* RE: [PATCH 0/6] ip_frag: fix reassembly defects and add test
From: Morten Brørup @ 2026-06-19 13:24 UTC (permalink / raw)
To: Stephen Hemminger, dev
In-Reply-To: <20260616210656.464062-1-stephen@networkplumber.org>
> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> Sent: Tuesday, 16 June 2026 23.06
>
> The IP reassembly library tracks only a running byte total and reserved
> slots for the first and last fragments, with no coverage map. As a
> result
> it mishandles duplicate, overlapping, oversized, and misheadered
> fragments, and the IPv4 key is missing a field RFC 791 requires. There
> was also no functional test to catch any of it.
>
> These came out of reviewing a duplicate-fragment report on the list.
>
> Patches 1 and 2 are interdependent: the overlap discard relies on the
> duplicate handling so an exact duplicate is dropped on its own rather
> than discarding the whole datagram. The rest are independent.
>
> Patch 6 adds a functional test modeled on the Linux selftest
> ip_defrag.c.
> It passes on this series; with any single fix reverted the matching
> case
> fails.
With patch 2/6 fixed,
Series-acked-by: Morten Brørup <mb@smartsharesystems.com>
^ permalink raw reply
* RE: [PATCH 2/6] ip_frag: discard datagrams with overlapping fragments
From: Morten Brørup @ 2026-06-19 13:12 UTC (permalink / raw)
To: Stephen Hemminger, dev; +Cc: stable, Konstantin Ananyev
In-Reply-To: <20260616210656.464062-3-stephen@networkplumber.org>
> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> Sent: Tuesday, 16 June 2026 23.06
>
> Existing code does not handle overlapping fragments.
>
> RFC 8200 (IPv6) requires that on overlap all reassembly is abandoned
> andall received fragments are dropped. RFC 791 (IPv4) originally called
> fortrimming and rewriting, but Linux discards for IPv4 as well, since
> overlap has no legitimate use and is a known attack vector.
>
> Depends on the duplicate-tolerance change so that an exact duplicate is
> dropped on its own rather than discarding the whole datagram.
>
> Fixes: cc8f4d020c0b ("examples/ip_reassembly: initial import")
> Cc: stable@dpdk.org
>
> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
> lib/ip_frag/ip_frag_internal.c | 34 ++++++++++++++++++++++++++--------
> 1 file changed, 26 insertions(+), 8 deletions(-)
>
> diff --git a/lib/ip_frag/ip_frag_internal.c
> b/lib/ip_frag/ip_frag_internal.c
> index 9a03ef995a..2505314a29 100644
> --- a/lib/ip_frag/ip_frag_internal.c
> +++ b/lib/ip_frag/ip_frag_internal.c
> @@ -92,16 +92,34 @@ ip_frag_process(struct ip_frag_pkt *fp, struct
> rte_ip_frag_death_row *dr,
> uint32_t i, idx;
>
> /*
> - * Discard an exact duplicate fragment. If a previously stored
> fragment
> - * already covers the same offset and length, this fragment
> carries no
> - * new data. Reassembly is tolerant of duplicates (RFC 791), so
> drop
> - * only this mbuf and keep the reassembly entry intact rather
> than
> - * treating it as an error. Fragments overlapping an existing one
> with
> - * different bounds are not handled here.
> + * Scan the fragments already collected for this datagram before
> + * storing the new one. The stored set is kept free of duplicates
> and
> + * overlaps, so a single pass is sufficient.
> */
> for (i = 0; i != fp->last_idx; i++) {
> - if (fp->frags[i].mb != NULL && fp->frags[i].ofs == ofs &&
> - fp->frags[i].len == len) {
> + if (fp->frags[i].mb == NULL)
> + continue;
> +
> + /*
> + * Exact duplicate: carries no new data. Reassembly
> tolerates
> + * duplicates (RFC 791), so drop only this mbuf and keep
> the
> + * entry.
> + */
> + if (fp->frags[i].ofs == ofs && fp->frags[i].len == len) {
> + IP_FRAG_MBUF2DR(dr, mb);
> + return NULL;
> + }
> +
> + /*
> + * Overlap with an existing fragment. Per RFC 8200 section
> 4.5
> + * (and RFC 5722) the datagram must be discarded; the same
> is
> + * applied to IPv4. Free all collected fragments, drop this
> one,
> + * and invalidate the entry.
> + */
> + if (ofs < fp->frags[i].ofs + fp->frags[i].len &&
> + fp->frags[i].ofs < ofs + len) {
This only catches fragments that are smaller than existing fragments, i.e. fit within one of the existing fragments.
It should be:
if ((ofs >= fp->frags[i].ofs &&
ofs < fp->frags[i].ofs + fp->frags[i].len) ||
(ofs + len >= fp->frags[i].ofs &&
ofs + len < fp->frags[i].ofs + fp->frags[i].len)) {
> + ip_frag_free(fp, dr);
> + ip_frag_key_invalidate(&fp->key);
> IP_FRAG_MBUF2DR(dr, mb);
> return NULL;
> }
> --
> 2.53.0
^ permalink raw reply
* [PATCH 18/18] dma/dpaa: add SG data validation and ERR050757 fix
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Gagandeep Singh
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Gagandeep Singh <g.singh@nxp.com>
Add scatter-gather (SG) support to the QDMA driver, enabled by default
via the s_sg_enable flag. Add optional data validation mode controlled
by the s_data_validation flag for debugging transfer correctness.
Add a workaround for hardware errata ERR050757: when
RTE_DMA_DPAA_ERRATA_ERR050757 is defined, configure the source frame
descriptor with stride settings (sss/ssd = FSL_QDMA_CMD_SS_ERR050757_LEN)
to force PCI read transactions to stay within the errata-safe length
limit, preventing data corruption on affected silicon.
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>
---
drivers/dma/dpaa/dpaa_qdma.c | 99 +++++++++++++++++++++++++++---------
1 file changed, 75 insertions(+), 24 deletions(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/dma/dpaa/dpaa_qdma.c
index bf2a373d70..ca615f8d8b 100644
--- a/drivers/dma/dpaa/dpaa_qdma.c
+++ b/drivers/dma/dpaa/dpaa_qdma.c
@@ -9,9 +9,14 @@
#include "dpaa_qdma.h"
#include "dpaa_qdma_logs.h"
+static int s_data_validation;
+static int s_hw_err_check;
+static int s_sg_enable = 1;
static uint32_t s_sg_max_entry_sz = 2000;
-static bool s_hw_err_check;
+#ifdef RTE_DMA_DPAA_ERRATA_ERR050757
+static int s_pci_read = 1;
+#endif
#define DPAA_DMA_ERROR_CHECK "dpaa_dma_err_check"
static inline void
@@ -112,7 +117,8 @@ dma_pool_alloc(char *nm, int size, int aligned, dma_addr_t *phy_addr)
if (!virt_addr)
return NULL;
- *phy_addr = rte_mem_virt2iova(virt_addr);
+ if (phy_addr)
+ *phy_addr = rte_mem_virt2iova(virt_addr);
return virt_addr;
}
@@ -392,6 +398,8 @@ fsl_qdma_data_validation(struct fsl_qdma_desc *desc[],
char err_msg[512];
int offset;
+ if (likely(!s_data_validation))
+ return;
offset = sprintf(err_msg, "Fatal TC%d/queue%d: ",
fsl_queue->block_id,
@@ -716,19 +724,21 @@ fsl_qdma_enqueue_desc_single(struct fsl_qdma_queue *fsl_queue,
ft = fsl_queue->ft[fsl_queue->ci];
#ifdef RTE_DMA_DPAA_ERRATA_ERR050757
- sdf = &ft->df.sdf;
- sdf->srttype = FSL_QDMA_CMD_RWTTYPE;
+ if (s_pci_read) {
+ sdf = &ft->df.sdf;
+ sdf->srttype = FSL_QDMA_CMD_RWTTYPE;
#ifdef RTE_DMA_DPAA_ERRATA_ERR050265
- sdf->prefetch = 1;
+ sdf->prefetch = 1;
#endif
- if (len > FSL_QDMA_CMD_SS_ERR050757_LEN) {
- sdf->ssen = 1;
- sdf->sss = FSL_QDMA_CMD_SS_ERR050757_LEN;
- sdf->ssd = FSL_QDMA_CMD_SS_ERR050757_LEN;
- } else {
- sdf->ssen = 0;
- sdf->sss = 0;
- sdf->ssd = 0;
+ if (len > FSL_QDMA_CMD_SS_ERR050757_LEN) {
+ sdf->ssen = 1;
+ sdf->sss = FSL_QDMA_CMD_SS_ERR050757_LEN;
+ sdf->ssd = FSL_QDMA_CMD_SS_ERR050757_LEN;
+ } else {
+ sdf->ssen = 0;
+ sdf->sss = 0;
+ sdf->ssd = 0;
+ }
}
#endif
csgf_src = &ft->desc_sbuf;
@@ -837,19 +847,21 @@ fsl_qdma_enqueue_desc_sg(struct fsl_qdma_queue *fsl_queue)
csgf_src->length = total_len;
csgf_dest->length = total_len;
#ifdef RTE_DMA_DPAA_ERRATA_ERR050757
- sdf = &ft->df.sdf;
- sdf->srttype = FSL_QDMA_CMD_RWTTYPE;
+ if (s_pci_read) {
+ sdf = &ft->df.sdf;
+ sdf->srttype = FSL_QDMA_CMD_RWTTYPE;
#ifdef RTE_DMA_DPAA_ERRATA_ERR050265
- sdf->prefetch = 1;
+ sdf->prefetch = 1;
#endif
- if (total_len > FSL_QDMA_CMD_SS_ERR050757_LEN) {
- sdf->ssen = 1;
- sdf->sss = FSL_QDMA_CMD_SS_ERR050757_LEN;
- sdf->ssd = FSL_QDMA_CMD_SS_ERR050757_LEN;
- } else {
- sdf->ssen = 0;
- sdf->sss = 0;
- sdf->ssd = 0;
+ if (total_len > FSL_QDMA_CMD_SS_ERR050757_LEN) {
+ sdf->ssen = 1;
+ sdf->sss = FSL_QDMA_CMD_SS_ERR050757_LEN;
+ sdf->ssd = FSL_QDMA_CMD_SS_ERR050757_LEN;
+ } else {
+ sdf->ssen = 0;
+ sdf->sss = 0;
+ sdf->ssd = 0;
+ }
}
#endif
ret = fsl_qdma_enqueue_desc_to_ring(fsl_queue, num);
@@ -888,6 +900,25 @@ fsl_qdma_enqueue_desc(struct fsl_qdma_queue *fsl_queue)
fsl_queue->pending_num = 0;
}
return ret;
+ } else if (!s_sg_enable) {
+ while (fsl_queue->pending_num > 0) {
+ ret = fsl_qdma_enqueue_desc_single(fsl_queue,
+ fsl_queue->pending_desc[start].dst,
+ fsl_queue->pending_desc[start].src,
+ fsl_queue->pending_desc[start].len);
+ if (!ret) {
+ start = (start + 1) &
+ (fsl_queue->pending_max - 1);
+ fsl_queue->pending_start = start;
+ fsl_queue->pending_num--;
+ } else {
+ DPAA_QDMA_ERR("Eq pending desc failed(%d)",
+ ret);
+ return -EIO;
+ }
+ }
+
+ return 0;
}
return fsl_qdma_enqueue_desc_sg(fsl_queue);
@@ -1344,6 +1375,26 @@ dpaa_qdma_init(struct rte_dma_dev *dmadev)
DPAA_QDMA_INFO("Enable DMA error checks");
}
+ if (getenv("DPAA_QDMA_DATA_VALIDATION"))
+ s_data_validation = 1;
+
+ if (getenv("DPAA_QDMA_HW_ERR_CHECK"))
+ s_hw_err_check = 1;
+
+ penv = getenv("DPAA_QDMA_SG_ENABLE");
+ if (penv)
+ s_sg_enable = atoi(penv);
+
+ penv = getenv("DPAA_QDMA_SG_MAX_ENTRY_SIZE");
+ if (penv)
+ s_sg_max_entry_sz = atoi(penv);
+
+#ifdef RTE_DMA_DPAA_ERRATA_ERR050757
+ penv = getenv("DPAA_QDMA_PCI_READ");
+ if (penv)
+ s_pci_read = atoi(penv);
+#endif
+
fsl_qdma->n_queues = QDMA_QUEUES * QDMA_BLOCKS;
fsl_qdma->num_blocks = QDMA_BLOCKS;
fsl_qdma->block_offset = QDMA_BLOCK_OFFSET;
--
2.25.1
^ permalink raw reply related
* [PATCH 17/18] mempool/dpaa: release bpid on driver destructor
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Track allocated BPIDs in a static per-BPID flag table and register a
driver destructor that releases any BPIDs still marked as in use at
process exit. This prevents BPID leaks when an application exits without
calling rte_mempool_free(). Also tune the per-lcore mempool cache flush
threshold to match the hardware bulk release size (DPAA_MBUF_MAX_ACQ_REL)
so that buffers are returned to HW in optimal burst sizes.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/bman.c | 8 +++
drivers/bus/dpaa/dpaa_bus_base_symbols.c | 1 +
drivers/bus/dpaa/include/fsl_bman.h | 3 ++
drivers/mempool/dpaa/dpaa_mempool.c | 67 ++++++++++++++++++++++--
drivers/mempool/dpaa/dpaa_mempool.h | 3 +-
5 files changed, 78 insertions(+), 4 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/bman.c b/drivers/bus/dpaa/base/qbman/bman.c
index 889d657052..d09266fa5b 100644
--- a/drivers/bus/dpaa/base/qbman/bman.c
+++ b/drivers/bus/dpaa/base/qbman/bman.c
@@ -237,6 +237,14 @@ void bman_free_pool(struct bman_pool *pool)
kfree(pool);
}
+void bman_free_bpid(u8 bpid, u32 flags)
+{
+ if (flags & BMAN_POOL_FLAG_THRESH)
+ bm_pool_set(bpid, zero_thresholds);
+ if (flags & BMAN_POOL_FLAG_DYNAMIC_BPID)
+ bman_release_bpid(bpid);
+}
+
const struct bman_pool_params *bman_get_params(const struct bman_pool *pool)
{
return &pool->params;
diff --git a/drivers/bus/dpaa/dpaa_bus_base_symbols.c b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
index 514ab7b1f1..8bd1a9bc6e 100644
--- a/drivers/bus/dpaa/dpaa_bus_base_symbols.c
+++ b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
@@ -46,6 +46,7 @@ RTE_EXPORT_INTERNAL_SYMBOL(netcfg_acquire)
RTE_EXPORT_INTERNAL_SYMBOL(netcfg_release)
RTE_EXPORT_INTERNAL_SYMBOL(bman_new_pool)
RTE_EXPORT_INTERNAL_SYMBOL(bman_free_pool)
+RTE_EXPORT_INTERNAL_SYMBOL(bman_free_bpid)
RTE_EXPORT_INTERNAL_SYMBOL(bman_get_params)
RTE_EXPORT_INTERNAL_SYMBOL(bman_release)
RTE_EXPORT_INTERNAL_SYMBOL(bman_acquire)
diff --git a/drivers/bus/dpaa/include/fsl_bman.h b/drivers/bus/dpaa/include/fsl_bman.h
index 67a7a09618..6079eedff5 100644
--- a/drivers/bus/dpaa/include/fsl_bman.h
+++ b/drivers/bus/dpaa/include/fsl_bman.h
@@ -317,6 +317,9 @@ struct bman_pool *bman_new_pool(const struct bman_pool_params *params);
__rte_internal
void bman_free_pool(struct bman_pool *pool);
+__rte_internal
+void bman_free_bpid(u8 bpid, u32 flags);
+
/**
* bman_get_params - Returns a pool object's parameters.
* @pool: the pool object
diff --git a/drivers/mempool/dpaa/dpaa_mempool.c b/drivers/mempool/dpaa/dpaa_mempool.c
index 3fdbcba646..210ea3bcf9 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.c
+++ b/drivers/mempool/dpaa/dpaa_mempool.c
@@ -25,10 +25,22 @@
#include <rte_eal.h>
#include <rte_malloc.h>
#include <rte_ring.h>
+#include <rte_common.h>
#include <dpaa_mempool.h>
#include <dpaax_iova_table.h>
+struct dpaa_bpid_flag {
+ uint32_t flags;
+ int used;
+};
+
+/** Be referenced in destructor to release bpid allocated.
+ * Destructor can't access bman_pool from eal mem,
+ * we release ID with flag directly.
+ */
+static struct dpaa_bpid_flag s_dpaa_bpid_allocated_flag[DPAA_MAX_BPOOLS];
+
#define FMAN_ERRATA_BOUNDARY ((uint64_t)4096)
#define FMAN_ERRATA_BOUNDARY_MASK (~(FMAN_ERRATA_BOUNDARY - 1))
@@ -58,6 +70,8 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
struct bman_pool_params params = {
.flags = BMAN_POOL_FLAG_DYNAMIC_BPID
};
+ unsigned int lcore_id;
+ struct rte_mempool_cache *cache;
MEMPOOL_INIT_FUNC_TRACE();
@@ -115,7 +129,7 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
rte_dpaa_bpid_info[bpid].ptov_off = 0;
rte_dpaa_bpid_info[bpid].flags = 0;
- bp_info = rte_malloc(NULL,
+ bp_info = rte_zmalloc(NULL,
sizeof(struct dpaa_bp_info),
RTE_CACHE_LINE_SIZE);
if (!bp_info) {
@@ -127,6 +141,20 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
rte_memcpy(bp_info, (void *)&rte_dpaa_bpid_info[bpid],
sizeof(struct dpaa_bp_info));
mp->pool_data = (void *)bp_info;
+ s_dpaa_bpid_allocated_flag[bpid].flags = params.flags;
+ s_dpaa_bpid_allocated_flag[bpid].used = true;
+ /* Update per core mempool cache threshold to optimal value which is
+ * number of buffers that can be released to HW buffer pool in
+ * a single API call.
+ */
+ for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
+ cache = &mp->local_cache[lcore_id];
+ DPAA_MEMPOOL_DEBUG("lCore %d: cache->flushthresh %d -> %d",
+ lcore_id, cache->flushthresh,
+ (uint32_t)(cache->size + DPAA_MBUF_MAX_ACQ_REL));
+ if (cache->flushthresh)
+ cache->flushthresh = cache->size + DPAA_MBUF_MAX_ACQ_REL;
+ }
DPAA_MEMPOOL_INFO("BMAN pool created for bpid =%d", bpid);
return 0;
@@ -136,6 +164,7 @@ static void
dpaa_mbuf_free_pool(struct rte_mempool *mp)
{
struct dpaa_bp_info *bp_info = DPAA_MEMPOOL_TO_POOL_INFO(mp);
+ uint16_t i;
MEMPOOL_INIT_FUNC_TRACE();
@@ -143,10 +172,25 @@ dpaa_mbuf_free_pool(struct rte_mempool *mp)
bman_free_pool(bp_info->bp);
DPAA_MEMPOOL_INFO("BMAN pool freed for bpid =%d",
bp_info->bpid);
- rte_free(mp->pool_data);
- bp_info->bp = NULL;
+ rte_dpaa_bpid_info[bp_info->bpid].mp = NULL;
+ rte_dpaa_bpid_info[bp_info->bpid].bp = NULL;
+ s_dpaa_bpid_allocated_flag[bp_info->bpid].used = false;
+ rte_free(bp_info);
mp->pool_data = NULL;
}
+
+ if (!rte_dpaa_bpid_info)
+ return;
+
+ for (i = 0; i < DPAA_MAX_BPOOLS; i++) {
+ if (rte_dpaa_bpid_info[i].mp)
+ break;
+ }
+
+ if (i == DPAA_MAX_BPOOLS) {
+ rte_free(rte_dpaa_bpid_info);
+ rte_dpaa_bpid_info = NULL;
+ }
}
static int
@@ -481,4 +525,21 @@ static const struct rte_mempool_ops dpaa_mpool_ops = {
.populate = dpaa_populate,
};
+#define RTE_PRIORITY_104 104
+
+RTE_FINI_PRIO(dpaa_mpool_finish, 104)
+{
+ uint16_t bpid;
+
+ for (bpid = 0; bpid < DPAA_MAX_BPOOLS; bpid++) {
+ if (s_dpaa_bpid_allocated_flag[bpid].used) {
+ bman_free_bpid(bpid, s_dpaa_bpid_allocated_flag[bpid].flags);
+ s_dpaa_bpid_allocated_flag[bpid].used = false;
+ }
+ }
+ /** The rte_dpaa_bpid_info and bman_pool from EAL mem have been released
+ * with EAL mem pool being destroyed.
+ */
+}
+
RTE_MEMPOOL_REGISTER_OPS(dpaa_mpool_ops);
diff --git a/drivers/mempool/dpaa/dpaa_mempool.h b/drivers/mempool/dpaa/dpaa_mempool.h
index 865b533b8f..d7ee49b557 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.h
+++ b/drivers/mempool/dpaa/dpaa_mempool.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
- * Copyright 2017,2019,2024 -2025 NXP
+ * Copyright 2017,2019,2024 -2026 NXP
*
*/
#ifndef __DPAA_MEMPOOL_H__
@@ -24,6 +24,7 @@
/* total number of bpools on SoC */
#define DPAA_MAX_BPOOLS 256
+#define DPAA_INVALID_BPID DPAA_MAX_BPOOLS
/* Maximum release/acquire from BMAN */
#define DPAA_MBUF_MAX_ACQ_REL FSL_BM_BURST_MAX
--
2.25.1
^ permalink raw reply related
* [PATCH 17/18] drivers: release DPAA bpid on driver destructor
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Track allocated BPIDs in a static per-BPID flag table and register a
driver destructor that releases any BPIDs still marked as in use at
process exit. This prevents BPID leaks when an application exits without
calling rte_mempool_free(). Also tune the per-lcore mempool cache flush
threshold to match the hardware bulk release size (DPAA_MBUF_MAX_ACQ_REL)
so that buffers are returned to HW in optimal burst sizes.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/bman.c | 8 +++
drivers/bus/dpaa/dpaa_bus_base_symbols.c | 1 +
drivers/bus/dpaa/include/fsl_bman.h | 3 ++
drivers/mempool/dpaa/dpaa_mempool.c | 67 ++++++++++++++++++++++--
drivers/mempool/dpaa/dpaa_mempool.h | 3 +-
5 files changed, 78 insertions(+), 4 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/bman.c b/drivers/bus/dpaa/base/qbman/bman.c
index 889d657052..d09266fa5b 100644
--- a/drivers/bus/dpaa/base/qbman/bman.c
+++ b/drivers/bus/dpaa/base/qbman/bman.c
@@ -237,6 +237,14 @@ void bman_free_pool(struct bman_pool *pool)
kfree(pool);
}
+void bman_free_bpid(u8 bpid, u32 flags)
+{
+ if (flags & BMAN_POOL_FLAG_THRESH)
+ bm_pool_set(bpid, zero_thresholds);
+ if (flags & BMAN_POOL_FLAG_DYNAMIC_BPID)
+ bman_release_bpid(bpid);
+}
+
const struct bman_pool_params *bman_get_params(const struct bman_pool *pool)
{
return &pool->params;
diff --git a/drivers/bus/dpaa/dpaa_bus_base_symbols.c b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
index 514ab7b1f1..8bd1a9bc6e 100644
--- a/drivers/bus/dpaa/dpaa_bus_base_symbols.c
+++ b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
@@ -46,6 +46,7 @@ RTE_EXPORT_INTERNAL_SYMBOL(netcfg_acquire)
RTE_EXPORT_INTERNAL_SYMBOL(netcfg_release)
RTE_EXPORT_INTERNAL_SYMBOL(bman_new_pool)
RTE_EXPORT_INTERNAL_SYMBOL(bman_free_pool)
+RTE_EXPORT_INTERNAL_SYMBOL(bman_free_bpid)
RTE_EXPORT_INTERNAL_SYMBOL(bman_get_params)
RTE_EXPORT_INTERNAL_SYMBOL(bman_release)
RTE_EXPORT_INTERNAL_SYMBOL(bman_acquire)
diff --git a/drivers/bus/dpaa/include/fsl_bman.h b/drivers/bus/dpaa/include/fsl_bman.h
index 67a7a09618..6079eedff5 100644
--- a/drivers/bus/dpaa/include/fsl_bman.h
+++ b/drivers/bus/dpaa/include/fsl_bman.h
@@ -317,6 +317,9 @@ struct bman_pool *bman_new_pool(const struct bman_pool_params *params);
__rte_internal
void bman_free_pool(struct bman_pool *pool);
+__rte_internal
+void bman_free_bpid(u8 bpid, u32 flags);
+
/**
* bman_get_params - Returns a pool object's parameters.
* @pool: the pool object
diff --git a/drivers/mempool/dpaa/dpaa_mempool.c b/drivers/mempool/dpaa/dpaa_mempool.c
index 3fdbcba646..210ea3bcf9 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.c
+++ b/drivers/mempool/dpaa/dpaa_mempool.c
@@ -25,10 +25,22 @@
#include <rte_eal.h>
#include <rte_malloc.h>
#include <rte_ring.h>
+#include <rte_common.h>
#include <dpaa_mempool.h>
#include <dpaax_iova_table.h>
+struct dpaa_bpid_flag {
+ uint32_t flags;
+ int used;
+};
+
+/** Be referenced in destructor to release bpid allocated.
+ * Destructor can't access bman_pool from eal mem,
+ * we release ID with flag directly.
+ */
+static struct dpaa_bpid_flag s_dpaa_bpid_allocated_flag[DPAA_MAX_BPOOLS];
+
#define FMAN_ERRATA_BOUNDARY ((uint64_t)4096)
#define FMAN_ERRATA_BOUNDARY_MASK (~(FMAN_ERRATA_BOUNDARY - 1))
@@ -58,6 +70,8 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
struct bman_pool_params params = {
.flags = BMAN_POOL_FLAG_DYNAMIC_BPID
};
+ unsigned int lcore_id;
+ struct rte_mempool_cache *cache;
MEMPOOL_INIT_FUNC_TRACE();
@@ -115,7 +129,7 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
rte_dpaa_bpid_info[bpid].ptov_off = 0;
rte_dpaa_bpid_info[bpid].flags = 0;
- bp_info = rte_malloc(NULL,
+ bp_info = rte_zmalloc(NULL,
sizeof(struct dpaa_bp_info),
RTE_CACHE_LINE_SIZE);
if (!bp_info) {
@@ -127,6 +141,20 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
rte_memcpy(bp_info, (void *)&rte_dpaa_bpid_info[bpid],
sizeof(struct dpaa_bp_info));
mp->pool_data = (void *)bp_info;
+ s_dpaa_bpid_allocated_flag[bpid].flags = params.flags;
+ s_dpaa_bpid_allocated_flag[bpid].used = true;
+ /* Update per core mempool cache threshold to optimal value which is
+ * number of buffers that can be released to HW buffer pool in
+ * a single API call.
+ */
+ for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
+ cache = &mp->local_cache[lcore_id];
+ DPAA_MEMPOOL_DEBUG("lCore %d: cache->flushthresh %d -> %d",
+ lcore_id, cache->flushthresh,
+ (uint32_t)(cache->size + DPAA_MBUF_MAX_ACQ_REL));
+ if (cache->flushthresh)
+ cache->flushthresh = cache->size + DPAA_MBUF_MAX_ACQ_REL;
+ }
DPAA_MEMPOOL_INFO("BMAN pool created for bpid =%d", bpid);
return 0;
@@ -136,6 +164,7 @@ static void
dpaa_mbuf_free_pool(struct rte_mempool *mp)
{
struct dpaa_bp_info *bp_info = DPAA_MEMPOOL_TO_POOL_INFO(mp);
+ uint16_t i;
MEMPOOL_INIT_FUNC_TRACE();
@@ -143,10 +172,25 @@ dpaa_mbuf_free_pool(struct rte_mempool *mp)
bman_free_pool(bp_info->bp);
DPAA_MEMPOOL_INFO("BMAN pool freed for bpid =%d",
bp_info->bpid);
- rte_free(mp->pool_data);
- bp_info->bp = NULL;
+ rte_dpaa_bpid_info[bp_info->bpid].mp = NULL;
+ rte_dpaa_bpid_info[bp_info->bpid].bp = NULL;
+ s_dpaa_bpid_allocated_flag[bp_info->bpid].used = false;
+ rte_free(bp_info);
mp->pool_data = NULL;
}
+
+ if (!rte_dpaa_bpid_info)
+ return;
+
+ for (i = 0; i < DPAA_MAX_BPOOLS; i++) {
+ if (rte_dpaa_bpid_info[i].mp)
+ break;
+ }
+
+ if (i == DPAA_MAX_BPOOLS) {
+ rte_free(rte_dpaa_bpid_info);
+ rte_dpaa_bpid_info = NULL;
+ }
}
static int
@@ -481,4 +525,21 @@ static const struct rte_mempool_ops dpaa_mpool_ops = {
.populate = dpaa_populate,
};
+#define RTE_PRIORITY_104 104
+
+RTE_FINI_PRIO(dpaa_mpool_finish, 104)
+{
+ uint16_t bpid;
+
+ for (bpid = 0; bpid < DPAA_MAX_BPOOLS; bpid++) {
+ if (s_dpaa_bpid_allocated_flag[bpid].used) {
+ bman_free_bpid(bpid, s_dpaa_bpid_allocated_flag[bpid].flags);
+ s_dpaa_bpid_allocated_flag[bpid].used = false;
+ }
+ }
+ /** The rte_dpaa_bpid_info and bman_pool from EAL mem have been released
+ * with EAL mem pool being destroyed.
+ */
+}
+
RTE_MEMPOOL_REGISTER_OPS(dpaa_mpool_ops);
diff --git a/drivers/mempool/dpaa/dpaa_mempool.h b/drivers/mempool/dpaa/dpaa_mempool.h
index 865b533b8f..d7ee49b557 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.h
+++ b/drivers/mempool/dpaa/dpaa_mempool.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
- * Copyright 2017,2019,2024 -2025 NXP
+ * Copyright 2017,2019,2024 -2026 NXP
*
*/
#ifndef __DPAA_MEMPOOL_H__
@@ -24,6 +24,7 @@
/* total number of bpools on SoC */
#define DPAA_MAX_BPOOLS 256
+#define DPAA_INVALID_BPID DPAA_MAX_BPOOLS
/* Maximum release/acquire from BMAN */
#define DPAA_MBUF_MAX_ACQ_REL FSL_BM_BURST_MAX
--
2.25.1
^ permalink raw reply related
* [PATCH 16/18] mempool/dpaa: optimize multi-entry buffer pool operations
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Replace the hardcoded buffer acquire count of 8 with the FSL_BM_BURST_MAX
constant when acquiring buffers from the buffer pool. Use a single
bm_hw_buf_desc structure for HW initialization of the first entry and copy
it to remaining entries, ensuring consistent HW descriptor state across
all entries in the pool.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/bman.c | 49 ++++++-----------------------
drivers/bus/dpaa/include/fsl_bman.h | 46 ++++++++++++++++++++++-----
drivers/mempool/dpaa/dpaa_mempool.c | 8 ++---
3 files changed, 52 insertions(+), 51 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/bman.c b/drivers/bus/dpaa/base/qbman/bman.c
index ee4232d0a0..889d657052 100644
--- a/drivers/bus/dpaa/base/qbman/bman.c
+++ b/drivers/bus/dpaa/base/qbman/bman.c
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2016 Freescale Semiconductor Inc.
- * Copyright 2017, 2024 NXP
+ * Copyright 2017, 2024-2026 NXP
*
*/
#include <rte_memcpy.h>
@@ -17,20 +17,6 @@
#define IRQNAME "BMan portal %d"
#define MAX_IRQNAME 16 /* big enough for "BMan portal %d" */
-
-#define MAX_U16 UINT16_MAX
-#define MAX_U32 UINT32_MAX
-#ifndef BIT_SIZE
-#define BIT_SIZE(t) (sizeof(t) * 8)
-#endif
-#define MAX_U48 \
- ((((uint64_t)MAX_U16) << BIT_SIZE(uint32_t)) | MAX_U32)
-#define HI16_OF_U48(x) \
- (((x) >> BIT_SIZE(rte_be32_t)) & MAX_U16)
-#define LO32_OF_U48(x) ((x) & MAX_U32)
-#define U48_BY_HI16_LO32(hi, lo) \
- (((hi) << BIT_SIZE(uint32_t)) | (lo))
-
struct bman_portal {
struct bm_portal p;
/* 2-element array. pools[0] is mask, pools[1] is snapshot. */
@@ -273,7 +259,7 @@ bman_release_fast(struct bman_pool *pool, const uint64_t *bufs,
struct bm_rcr_entry *r;
uint8_t i, avail;
uint64_t bpid = pool->params.bpid;
- struct bm_hw_buf_desc bm_bufs[FSL_BM_BURST_MAX];
+ struct bm_buffer bm_bufs[FSL_BM_BURST_MAX];
#ifdef RTE_LIBRTE_DPAA_HWDEBUG
if (!num || (num > FSL_BM_BURST_MAX))
@@ -290,19 +276,15 @@ bman_release_fast(struct bman_pool *pool, const uint64_t *bufs,
if (unlikely(!r))
return -EBUSY;
+ bm_bufs[0].be_desc.bpid = bpid;
+ for (i = 0; i < num; i++)
+ bm_buffer_set64_to_be(&bm_bufs[i], bufs[i]);
/*
* we can copy all but the first entry, as this can trigger badness
* with the valid-bit
*/
- bm_bufs[0].bpid = bpid;
- bm_bufs[0].hi_addr = cpu_to_be16(HI16_OF_U48(bufs[0]));
- bm_bufs[0].lo_addr = cpu_to_be32(LO32_OF_U48(bufs[0]));
- for (i = 1; i < num; i++) {
- bm_bufs[i].hi_addr = cpu_to_be16(HI16_OF_U48(bufs[i]));
- bm_bufs[i].lo_addr = cpu_to_be32(LO32_OF_U48(bufs[i]));
- }
-
- memcpy(r->bufs, bm_bufs, sizeof(struct bm_buffer) * num);
+ r->bufs[0].opaque = bm_bufs[0].opaque;
+ rte_memcpy(&r->bufs[1], &bm_bufs[1], sizeof(struct bm_buffer) * (num - 1));
bm_rcr_pvb_commit(&p->p, BM_RCR_VERB_CMD_BPID_SINGLE |
(num & BM_RCR_VERB_BUFCOUNT_MASK));
@@ -360,16 +342,6 @@ __rte_unused bman_extract_addr(struct bm_buffer *buf)
return buf->addr;
}
-static inline uint64_t
-bman_hw_extract_addr(struct bm_hw_buf_desc *buf)
-{
- uint64_t hi, lo;
-
- hi = be16_to_cpu(buf->hi_addr);
- lo = be32_to_cpu(buf->lo_addr);
- return U48_BY_HI16_LO32(hi, lo);
-}
-
RTE_EXPORT_INTERNAL_SYMBOL(bman_acquire_fast)
int
bman_acquire_fast(struct bman_pool *pool, uint64_t *bufs, uint8_t num)
@@ -378,7 +350,7 @@ bman_acquire_fast(struct bman_pool *pool, uint64_t *bufs, uint8_t num)
struct bm_mc_command *mcc;
struct bm_mc_result *mcr;
uint8_t i, rst;
- struct bm_hw_buf_desc bm_bufs[FSL_BM_BURST_MAX];
+ struct bm_buffer bm_bufs[FSL_BM_BURST_MAX];
#ifdef RTE_LIBRTE_DPAA_HWDEBUG
if (!num || (num > FSL_BM_BURST_MAX))
@@ -397,11 +369,10 @@ bman_acquire_fast(struct bman_pool *pool, uint64_t *bufs, uint8_t num)
if (unlikely(rst < 1 || rst > FSL_BM_BURST_MAX))
return -EINVAL;
- rte_memcpy(bm_bufs, mcr->acquire.bufs,
- sizeof(struct bm_buffer) * rst);
+ rte_memcpy(bm_bufs, mcr->acquire.bufs, sizeof(struct bm_buffer) * rst);
for (i = 0; i < rst; i++)
- bufs[i] = bman_hw_extract_addr(&bm_bufs[i]);
+ bufs[i] = bm_buffer_get64_from_be(&bm_bufs[i]);
return rst;
}
diff --git a/drivers/bus/dpaa/include/fsl_bman.h b/drivers/bus/dpaa/include/fsl_bman.h
index 2d24b89889..67a7a09618 100644
--- a/drivers/bus/dpaa/include/fsl_bman.h
+++ b/drivers/bus/dpaa/include/fsl_bman.h
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2012 Freescale Semiconductor, Inc.
- * Copyright 2024 NXP
+ * Copyright 2024-2026 NXP
*
*/
@@ -42,6 +42,13 @@ struct bm_mc_result; /* MC result */
* pool id specific to this buffer is needed (BM_RCR_VERB_CMD_BPID_MULTI,
* BM_MCC_VERB_ACQUIRE), the 'bpid' field is used.
*/
+struct __rte_packed_begin bm_hw_buf_desc {
+ uint8_t rsv;
+ uint8_t bpid;
+ rte_be16_t hi; /* High 16-bits of 48-bit address */
+ rte_be32_t lo; /* Low 32-bits of 48-bit address */
+} __rte_packed_end;
+
struct __rte_aligned(8) bm_buffer {
union {
struct {
@@ -66,17 +73,11 @@ struct __rte_aligned(8) bm_buffer {
u64 __notaddress:16;
#endif
};
+ struct bm_hw_buf_desc be_desc;
u64 opaque;
};
};
-struct __rte_packed_begin bm_hw_buf_desc {
- uint8_t rsv;
- uint8_t bpid;
- rte_be16_t hi_addr; /* High 16-bits of 48-bit address */
- rte_be32_t lo_addr; /* Low 32-bits of 48-bit address */
-} __rte_packed_end;
-
static inline u64 bm_buffer_get64(const struct bm_buffer *buf)
{
return buf->addr;
@@ -87,6 +88,17 @@ static inline dma_addr_t bm_buf_addr(const struct bm_buffer *buf)
return (dma_addr_t)buf->addr;
}
+#ifndef BIT_SIZE
+#define BIT_SIZE(t) (sizeof(t) * 8)
+#endif
+#define MAX_U48 \
+ ((((uint64_t)UINT16_MAX) << BIT_SIZE(uint32_t)) | UINT32_MAX)
+#define HI16_OF_U48(x) \
+ (((x) >> BIT_SIZE(uint32_t)) & UINT16_MAX)
+#define LO32_OF_U48(x) ((x) & UINT32_MAX)
+#define U48_BY_HI16_LO32(hi, lo) \
+ (((hi) << BIT_SIZE(uint32_t)) | (lo))
+
#define bm_buffer_set64(buf, v) \
do { \
struct bm_buffer *__buf931 = (buf); \
@@ -94,6 +106,24 @@ static inline dma_addr_t bm_buf_addr(const struct bm_buffer *buf)
__buf931->lo = lower_32_bits(v); \
} while (0)
+#define bm_buffer_set64_to_be(buf, v) \
+ do { \
+ struct bm_buffer *__buf931 = (buf); \
+ \
+ __buf931->be_desc.hi = cpu_to_be16(HI16_OF_U48(v)); \
+ __buf931->be_desc.lo = cpu_to_be32(LO32_OF_U48(v)); \
+ } while (0)
+
+#define bm_buffer_get64_from_be(buf) \
+ ({ \
+ uint64_t hi, lo; \
+ struct bm_buffer *__buf931 = (buf); \
+ \
+ hi = be16_to_cpu(__buf931->be_desc.hi); \
+ lo = be32_to_cpu(__buf931->be_desc.lo); \
+ U48_BY_HI16_LO32(hi, lo); \
+ })
+
#define FSL_BM_BURST_MAX 8
/* See 1.5.3.5.4: "Release Command" */
diff --git a/drivers/mempool/dpaa/dpaa_mempool.c b/drivers/mempool/dpaa/dpaa_mempool.c
index 2f8555a026..3fdbcba646 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.c
+++ b/drivers/mempool/dpaa/dpaa_mempool.c
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
- * Copyright 2017,2019,2023-2025 NXP
+ * Copyright 2017,2019,2023-2026 NXP
*
*/
@@ -50,7 +50,7 @@ static int
dpaa_mbuf_create_pool(struct rte_mempool *mp)
{
struct bman_pool *bp;
- struct bm_buffer bufs[8];
+ struct bm_buffer bufs[FSL_BM_BURST_MAX];
struct dpaa_bp_info *bp_info;
uint8_t bpid;
int num_bufs = 0, ret = 0;
@@ -83,8 +83,8 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
* then in 1s for the remainder.
*/
if (ret != 1)
- ret = bman_acquire(bp, bufs, 8, 0);
- if (ret < 8)
+ ret = bman_acquire(bp, bufs, FSL_BM_BURST_MAX, 0);
+ if (ret < FSL_BM_BURST_MAX)
ret = bman_acquire(bp, bufs, 1, 0);
if (ret > 0)
num_bufs += ret;
--
2.25.1
^ permalink raw reply related
* [PATCH 16/18] drivers: optimize DPAA multi-entry buffer pool operations
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Replace the hardcoded buffer acquire count of 8 with the FSL_BM_BURST_MAX
constant when acquiring buffers from the buffer pool. Use a single
bm_hw_buf_desc structure for HW initialization of the first entry and copy
it to remaining entries, ensuring consistent HW descriptor state across
all entries in the pool.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/bman.c | 49 ++++++-----------------------
drivers/bus/dpaa/include/fsl_bman.h | 46 ++++++++++++++++++++++-----
drivers/mempool/dpaa/dpaa_mempool.c | 8 ++---
3 files changed, 52 insertions(+), 51 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/bman.c b/drivers/bus/dpaa/base/qbman/bman.c
index ee4232d0a0..889d657052 100644
--- a/drivers/bus/dpaa/base/qbman/bman.c
+++ b/drivers/bus/dpaa/base/qbman/bman.c
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2016 Freescale Semiconductor Inc.
- * Copyright 2017, 2024 NXP
+ * Copyright 2017, 2024-2026 NXP
*
*/
#include <rte_memcpy.h>
@@ -17,20 +17,6 @@
#define IRQNAME "BMan portal %d"
#define MAX_IRQNAME 16 /* big enough for "BMan portal %d" */
-
-#define MAX_U16 UINT16_MAX
-#define MAX_U32 UINT32_MAX
-#ifndef BIT_SIZE
-#define BIT_SIZE(t) (sizeof(t) * 8)
-#endif
-#define MAX_U48 \
- ((((uint64_t)MAX_U16) << BIT_SIZE(uint32_t)) | MAX_U32)
-#define HI16_OF_U48(x) \
- (((x) >> BIT_SIZE(rte_be32_t)) & MAX_U16)
-#define LO32_OF_U48(x) ((x) & MAX_U32)
-#define U48_BY_HI16_LO32(hi, lo) \
- (((hi) << BIT_SIZE(uint32_t)) | (lo))
-
struct bman_portal {
struct bm_portal p;
/* 2-element array. pools[0] is mask, pools[1] is snapshot. */
@@ -273,7 +259,7 @@ bman_release_fast(struct bman_pool *pool, const uint64_t *bufs,
struct bm_rcr_entry *r;
uint8_t i, avail;
uint64_t bpid = pool->params.bpid;
- struct bm_hw_buf_desc bm_bufs[FSL_BM_BURST_MAX];
+ struct bm_buffer bm_bufs[FSL_BM_BURST_MAX];
#ifdef RTE_LIBRTE_DPAA_HWDEBUG
if (!num || (num > FSL_BM_BURST_MAX))
@@ -290,19 +276,15 @@ bman_release_fast(struct bman_pool *pool, const uint64_t *bufs,
if (unlikely(!r))
return -EBUSY;
+ bm_bufs[0].be_desc.bpid = bpid;
+ for (i = 0; i < num; i++)
+ bm_buffer_set64_to_be(&bm_bufs[i], bufs[i]);
/*
* we can copy all but the first entry, as this can trigger badness
* with the valid-bit
*/
- bm_bufs[0].bpid = bpid;
- bm_bufs[0].hi_addr = cpu_to_be16(HI16_OF_U48(bufs[0]));
- bm_bufs[0].lo_addr = cpu_to_be32(LO32_OF_U48(bufs[0]));
- for (i = 1; i < num; i++) {
- bm_bufs[i].hi_addr = cpu_to_be16(HI16_OF_U48(bufs[i]));
- bm_bufs[i].lo_addr = cpu_to_be32(LO32_OF_U48(bufs[i]));
- }
-
- memcpy(r->bufs, bm_bufs, sizeof(struct bm_buffer) * num);
+ r->bufs[0].opaque = bm_bufs[0].opaque;
+ rte_memcpy(&r->bufs[1], &bm_bufs[1], sizeof(struct bm_buffer) * (num - 1));
bm_rcr_pvb_commit(&p->p, BM_RCR_VERB_CMD_BPID_SINGLE |
(num & BM_RCR_VERB_BUFCOUNT_MASK));
@@ -360,16 +342,6 @@ __rte_unused bman_extract_addr(struct bm_buffer *buf)
return buf->addr;
}
-static inline uint64_t
-bman_hw_extract_addr(struct bm_hw_buf_desc *buf)
-{
- uint64_t hi, lo;
-
- hi = be16_to_cpu(buf->hi_addr);
- lo = be32_to_cpu(buf->lo_addr);
- return U48_BY_HI16_LO32(hi, lo);
-}
-
RTE_EXPORT_INTERNAL_SYMBOL(bman_acquire_fast)
int
bman_acquire_fast(struct bman_pool *pool, uint64_t *bufs, uint8_t num)
@@ -378,7 +350,7 @@ bman_acquire_fast(struct bman_pool *pool, uint64_t *bufs, uint8_t num)
struct bm_mc_command *mcc;
struct bm_mc_result *mcr;
uint8_t i, rst;
- struct bm_hw_buf_desc bm_bufs[FSL_BM_BURST_MAX];
+ struct bm_buffer bm_bufs[FSL_BM_BURST_MAX];
#ifdef RTE_LIBRTE_DPAA_HWDEBUG
if (!num || (num > FSL_BM_BURST_MAX))
@@ -397,11 +369,10 @@ bman_acquire_fast(struct bman_pool *pool, uint64_t *bufs, uint8_t num)
if (unlikely(rst < 1 || rst > FSL_BM_BURST_MAX))
return -EINVAL;
- rte_memcpy(bm_bufs, mcr->acquire.bufs,
- sizeof(struct bm_buffer) * rst);
+ rte_memcpy(bm_bufs, mcr->acquire.bufs, sizeof(struct bm_buffer) * rst);
for (i = 0; i < rst; i++)
- bufs[i] = bman_hw_extract_addr(&bm_bufs[i]);
+ bufs[i] = bm_buffer_get64_from_be(&bm_bufs[i]);
return rst;
}
diff --git a/drivers/bus/dpaa/include/fsl_bman.h b/drivers/bus/dpaa/include/fsl_bman.h
index 2d24b89889..67a7a09618 100644
--- a/drivers/bus/dpaa/include/fsl_bman.h
+++ b/drivers/bus/dpaa/include/fsl_bman.h
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2012 Freescale Semiconductor, Inc.
- * Copyright 2024 NXP
+ * Copyright 2024-2026 NXP
*
*/
@@ -42,6 +42,13 @@ struct bm_mc_result; /* MC result */
* pool id specific to this buffer is needed (BM_RCR_VERB_CMD_BPID_MULTI,
* BM_MCC_VERB_ACQUIRE), the 'bpid' field is used.
*/
+struct __rte_packed_begin bm_hw_buf_desc {
+ uint8_t rsv;
+ uint8_t bpid;
+ rte_be16_t hi; /* High 16-bits of 48-bit address */
+ rte_be32_t lo; /* Low 32-bits of 48-bit address */
+} __rte_packed_end;
+
struct __rte_aligned(8) bm_buffer {
union {
struct {
@@ -66,17 +73,11 @@ struct __rte_aligned(8) bm_buffer {
u64 __notaddress:16;
#endif
};
+ struct bm_hw_buf_desc be_desc;
u64 opaque;
};
};
-struct __rte_packed_begin bm_hw_buf_desc {
- uint8_t rsv;
- uint8_t bpid;
- rte_be16_t hi_addr; /* High 16-bits of 48-bit address */
- rte_be32_t lo_addr; /* Low 32-bits of 48-bit address */
-} __rte_packed_end;
-
static inline u64 bm_buffer_get64(const struct bm_buffer *buf)
{
return buf->addr;
@@ -87,6 +88,17 @@ static inline dma_addr_t bm_buf_addr(const struct bm_buffer *buf)
return (dma_addr_t)buf->addr;
}
+#ifndef BIT_SIZE
+#define BIT_SIZE(t) (sizeof(t) * 8)
+#endif
+#define MAX_U48 \
+ ((((uint64_t)UINT16_MAX) << BIT_SIZE(uint32_t)) | UINT32_MAX)
+#define HI16_OF_U48(x) \
+ (((x) >> BIT_SIZE(uint32_t)) & UINT16_MAX)
+#define LO32_OF_U48(x) ((x) & UINT32_MAX)
+#define U48_BY_HI16_LO32(hi, lo) \
+ (((hi) << BIT_SIZE(uint32_t)) | (lo))
+
#define bm_buffer_set64(buf, v) \
do { \
struct bm_buffer *__buf931 = (buf); \
@@ -94,6 +106,24 @@ static inline dma_addr_t bm_buf_addr(const struct bm_buffer *buf)
__buf931->lo = lower_32_bits(v); \
} while (0)
+#define bm_buffer_set64_to_be(buf, v) \
+ do { \
+ struct bm_buffer *__buf931 = (buf); \
+ \
+ __buf931->be_desc.hi = cpu_to_be16(HI16_OF_U48(v)); \
+ __buf931->be_desc.lo = cpu_to_be32(LO32_OF_U48(v)); \
+ } while (0)
+
+#define bm_buffer_get64_from_be(buf) \
+ ({ \
+ uint64_t hi, lo; \
+ struct bm_buffer *__buf931 = (buf); \
+ \
+ hi = be16_to_cpu(__buf931->be_desc.hi); \
+ lo = be32_to_cpu(__buf931->be_desc.lo); \
+ U48_BY_HI16_LO32(hi, lo); \
+ })
+
#define FSL_BM_BURST_MAX 8
/* See 1.5.3.5.4: "Release Command" */
diff --git a/drivers/mempool/dpaa/dpaa_mempool.c b/drivers/mempool/dpaa/dpaa_mempool.c
index 2f8555a026..3fdbcba646 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.c
+++ b/drivers/mempool/dpaa/dpaa_mempool.c
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
- * Copyright 2017,2019,2023-2025 NXP
+ * Copyright 2017,2019,2023-2026 NXP
*
*/
@@ -50,7 +50,7 @@ static int
dpaa_mbuf_create_pool(struct rte_mempool *mp)
{
struct bman_pool *bp;
- struct bm_buffer bufs[8];
+ struct bm_buffer bufs[FSL_BM_BURST_MAX];
struct dpaa_bp_info *bp_info;
uint8_t bpid;
int num_bufs = 0, ret = 0;
@@ -83,8 +83,8 @@ dpaa_mbuf_create_pool(struct rte_mempool *mp)
* then in 1s for the remainder.
*/
if (ret != 1)
- ret = bman_acquire(bp, bufs, 8, 0);
- if (ret < 8)
+ ret = bman_acquire(bp, bufs, FSL_BM_BURST_MAX, 0);
+ if (ret < FSL_BM_BURST_MAX)
ret = bman_acquire(bp, bufs, 1, 0);
if (ret > 0)
num_bufs += ret;
--
2.25.1
^ permalink raw reply related
* [PATCH 15/18] net/dpaa: report error on using deferred start
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
This patch add support to report on error
for rx and tx deferred start config
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index a223703421..c884d7e5e4 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -1171,6 +1171,12 @@ int dpaa_eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
rxq->nb_desc = UINT16_MAX;
rxq->offloads = rx_conf->offloads;
+ /* Rx deferred start is not supported */
+ if (rx_conf->rx_deferred_start) {
+ DPAA_PMD_ERR("%p:Rx deferred start not supported", (void *)dev);
+ return -EINVAL;
+ }
+
DPAA_PMD_INFO("Rx queue setup for queue index: %d fq_id (0x%x)",
queue_idx, rxq->fqid);
@@ -1477,6 +1483,12 @@ int dpaa_eth_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
txq->nb_desc = UINT16_MAX;
txq->offloads = tx_conf->offloads;
+ /* Tx deferred start is not supported */
+ if (tx_conf->tx_deferred_start) {
+ DPAA_PMD_ERR("%p:Tx deferred start not supported", (void *)dev);
+ return -EINVAL;
+ }
+
if (queue_idx >= dev->data->nb_tx_queues) {
rte_errno = EOVERFLOW;
DPAA_PMD_ERR("%p: queue index out of range (%u >= %u)",
--
2.25.1
^ permalink raw reply related
* [PATCH 14/18] net/dpaa: optimize FMC MAC type parsing
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
For ls104xa, MAC9 and MAC10's type could be either of 10G/2.5G/1G
up to serdes configuration, MAC index should be identified by
port name instead of parsing MAC type and port number.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/net/dpaa/dpaa_fmc.c | 73 ++++++++++++++++++++++---------------
1 file changed, 44 insertions(+), 29 deletions(-)
diff --git a/drivers/net/dpaa/dpaa_fmc.c b/drivers/net/dpaa/dpaa_fmc.c
index 7dc42f6e23..3034f534a5 100644
--- a/drivers/net/dpaa/dpaa_fmc.c
+++ b/drivers/net/dpaa/dpaa_fmc.c
@@ -1,5 +1,5 @@
/* SPDX-License-Identifier: BSD-3-Clause
- * Copyright 2017-2023 NXP
+ * Copyright 2017-2026 NXP
*/
/* System headers */
@@ -204,6 +204,36 @@ struct fmc_model_t {
struct fmc_model_t *g_fmc_model;
+static int
+dpaa_port_fmc_get_idx_from_name(const char *name)
+{
+ const char *found;
+ int idx_str_start = -1, idx;
+
+#define FMC_PORT_NAME_MAC "MAC/"
+#define FMC_PORT_NAME_OFFLINE "OFFLINE/"
+
+ found = strstr(name, FMC_PORT_NAME_MAC);
+ if (!found) {
+ found = strstr(name, FMC_PORT_NAME_OFFLINE);
+ if (found)
+ idx_str_start = strlen(FMC_PORT_NAME_OFFLINE);
+ } else {
+ idx_str_start = strlen(FMC_PORT_NAME_MAC);
+ }
+
+ if (!found) {
+ DPAA_PMD_ERR("Invalid fmc port name: %s", name);
+ return -EINVAL;
+ }
+
+ idx = atoi(&found[idx_str_start]);
+
+ DPAA_PMD_INFO("MAC index of %s is %d", name, idx);
+
+ return idx;
+}
+
static int
dpaa_port_fmc_port_parse(struct fman_if *fif,
const struct fmc_model_t *fmc_model,
@@ -211,7 +241,10 @@ dpaa_port_fmc_port_parse(struct fman_if *fif,
{
int current_port = fmc_model->apply_order[apply_idx].index;
const fmc_port *pport = &fmc_model->port[current_port];
- uint32_t num;
+ int num = dpaa_port_fmc_get_idx_from_name(pport->name);
+
+ if (num < 0)
+ return num;
if (pport->type == e_FM_PORT_TYPE_OH_OFFLINE_PARSING &&
pport->number == fif->mac_idx &&
@@ -219,40 +252,22 @@ dpaa_port_fmc_port_parse(struct fman_if *fif,
fif->mac_type == fman_onic))
return current_port;
- if (fif->mac_type == fman_mac_1g) {
- if (pport->type != e_FM_PORT_TYPE_RX)
- return -ENODEV;
- num = pport->number + DPAA_1G_MAC_START_IDX;
- if (fif->mac_idx == num)
- return current_port;
-
+ if (fif->mac_type == fman_mac_1g &&
+ pport->type != e_FM_PORT_TYPE_RX)
return -ENODEV;
- }
-
- if (fif->mac_type == fman_mac_2_5g) {
- if (pport->type != e_FM_PORT_TYPE_RX_2_5G)
- return -ENODEV;
- num = pport->number + DPAA_2_5G_MAC_START_IDX;
- if (fif->mac_idx == num)
- return current_port;
+ if (fif->mac_type == fman_mac_2_5g &&
+ pport->type != e_FM_PORT_TYPE_RX_2_5G)
return -ENODEV;
- }
-
- if (fif->mac_type == fman_mac_10g) {
- if (pport->type != e_FM_PORT_TYPE_RX_10G)
- return -ENODEV;
- num = pport->number + DPAA_10G_MAC_START_IDX;
- if (fif->mac_idx == num)
- return current_port;
+ if (fif->mac_type == fman_mac_10g &&
+ pport->type != e_FM_PORT_TYPE_RX_10G)
return -ENODEV;
- }
- DPAA_PMD_ERR("Invalid MAC(mac_idx=%d) type(%d)",
- fif->mac_idx, fif->mac_type);
+ if (fif->mac_idx == num)
+ return current_port;
- return -EINVAL;
+ return -ENODEV;
}
static int
--
2.25.1
^ permalink raw reply related
* [PATCH 13/18] bus/dpaa: improve log macro and fix bus detection
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Replace DPAA_BUS_LOG(LEVEL, ...) calls with shorthand macros
(DPAA_BUS_INFO, DPAA_BUS_ERR, DPAA_BUS_WARN, DPAA_BUS_DEBUG) for
consistency across the driver.
Move bus detection (sysfs path check), portal key creation and
dpaa_bus.detected guard into dpaa_bus_dev_compare() so that bus
probe is properly gated on hardware presence.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/bus/dpaa/base/fman/fman.c | 9 ++++----
drivers/bus/dpaa/dpaa_bus.c | 35 ++++++++++++++++++++++++-------
2 files changed, 32 insertions(+), 12 deletions(-)
diff --git a/drivers/bus/dpaa/base/fman/fman.c b/drivers/bus/dpaa/base/fman/fman.c
index 55f466d751..67f77265ca 100644
--- a/drivers/bus/dpaa/base/fman/fman.c
+++ b/drivers/bus/dpaa/base/fman/fman.c
@@ -119,7 +119,7 @@ _fman_init(const struct device_node *fman_node, int fd)
ip_rev_1 = in_be32((uint8_t *)fman->ccsr_vir + FMAN_IP_REV_1);
fman->ip_rev = ip_rev_1 >> FMAN_IP_REV_1_MAJOR_SHIFT;
fman->ip_rev &= FMAN_IP_REV_1_MAJOR_MASK;
- DPAA_BUS_LOG(NOTICE, "FMan version is 0x%02x", fman->ip_rev);
+ DPAA_BUS_INFO("FMan version is 0x%02x", fman->ip_rev);
if (fman->ip_rev >= FMAN_V3) {
/*
@@ -795,8 +795,7 @@ fman_if_init(const struct device_node *dpa_node, int fd)
fman_if_vsp_init(__if);
/* Parsing of the network interface is complete, add it to the list */
- DPAA_BUS_LOG(DEBUG, "Found %s, Tx Channel = %x, FMAN = %x,"
- "Port ID = %x",
+ DPAA_BUS_DEBUG("Found %s, Tx Channel = %x, FMAN = %x, Port ID = %x",
dname, __if->__if.tx_channel_id, __if->__if.fman->idx,
__if->__if.mac_idx);
@@ -1109,14 +1108,14 @@ fman_init(void)
fd = open(FMAN_DEVICE_PATH, O_RDWR);
if (unlikely(fd < 0)) {
- DPAA_BUS_LOG(ERR, "Unable to open %s: %s", FMAN_DEVICE_PATH, strerror(errno));
+ DPAA_BUS_ERR("Unable to open %s: %s", FMAN_DEVICE_PATH, strerror(errno));
return fd;
}
fman_ccsr_map_fd = fd;
parent_node = of_find_compatible_node(NULL, NULL, "fsl,dpaa");
if (!parent_node) {
- DPAA_BUS_LOG(ERR, "Unable to find fsl,dpaa node");
+ DPAA_BUS_ERR("Unable to find fsl,dpaa node");
return -ENODEV;
}
diff --git a/drivers/bus/dpaa/dpaa_bus.c b/drivers/bus/dpaa/dpaa_bus.c
index 02a8c5882e..79d0b6311a 100644
--- a/drivers/bus/dpaa/dpaa_bus.c
+++ b/drivers/bus/dpaa/dpaa_bus.c
@@ -560,13 +560,37 @@ rte_dpaa_bus_parse(const char *name, void *out)
static int
dpaa_bus_dev_compare(const char *name1, const char *name2)
{
+ int ret = 0;
char devname1[32], devname2[32];
if (rte_dpaa_bus_parse(name1, devname1) != 0 ||
rte_dpaa_bus_parse(name2, devname2) != 0)
return 1;
- return strncmp(devname1, devname2, sizeof(devname1));
+#define DPAA_DEV_PATH1 "/sys/devices/platform/soc/soc:fsl,dpaa"
+#define DPAA_DEV_PATH2 "/sys/devices/platform/fsl,dpaa"
+ if ((access(DPAA_DEV_PATH1, F_OK) != 0) &&
+ (access(DPAA_DEV_PATH2, F_OK) != 0)) {
+ DPAA_BUS_DEBUG("DPAA Bus not present. Skipping.");
+ return 0;
+ }
+
+ if (dpaa_bus.detected)
+ return 0;
+
+ dpaa_bus.detected = 1;
+
+ /* create the key, supplying a function that'll be invoked
+ * when a portal affined thread will be deleted.
+ */
+ ret = pthread_key_create(&dpaa_portal_key, dpaa_portal_finish);
+ if (ret) {
+ DPAA_BUS_DEBUG("Unable to create pthread key. (%d)", ret);
+ dpaa_clean_device_list();
+ return ret;
+ }
+
+ return 0;
}
/* register a dpaa bus based dpaa driver */
@@ -667,8 +691,6 @@ static int rte_dpaa_setup_intr(struct rte_intr_handle *intr_handle)
return 0;
}
-#define DPAA_DEV_PATH1 "/sys/devices/platform/soc/soc:fsl,dpaa"
-#define DPAA_DEV_PATH2 "/sys/devices/platform/fsl,dpaa"
static int
rte_dpaa_bus_scan(void)
@@ -715,12 +737,11 @@ rte_dpaa_bus_scan(void)
dpaa_bus.svr_ver = 0;
}
if (dpaa_bus.svr_ver == SVR_LS1046A_FAMILY) {
- DPAA_BUS_LOG(INFO, "This is LS1046A family SoC.");
+ DPAA_BUS_INFO("This is LS1046A family SoC.");
} else if (dpaa_bus.svr_ver == SVR_LS1043A_FAMILY) {
- DPAA_BUS_LOG(INFO, "This is LS1043A family SoC.");
+ DPAA_BUS_INFO("This is LS1043A family SoC.");
} else {
- DPAA_BUS_LOG(WARNING,
- "This is Unknown(%08x) DPAA1 family SoC.",
+ DPAA_BUS_WARN("This is Unknown(%08x) DPAA1 family SoC.",
dpaa_bus.svr_ver);
}
--
2.25.1
^ permalink raw reply related
* [PATCH 12/18] net/dpaa: optimize FM deconfig
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Consolidate FM deconfiguration to avoid duplicate calls.
Move the fm_deconfig call to a single location and remove
redundant checks in the device close path.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 37 +++++++++++++++++++++++++---------
drivers/net/dpaa/dpaa_flow.c | 9 +++++----
2 files changed, 32 insertions(+), 14 deletions(-)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 3b66fbe18a..a223703421 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -526,10 +526,12 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
/* DPAA FM deconfig */
if (!(default_q || fmc_q)) {
- ret = dpaa_fm_deconfig(dpaa_intf, dev->process_private);
- if (ret) {
- DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
- dev->data->name, ret);
+ if (dpaa_intf->port_handle) {
+ ret = dpaa_fm_deconfig(dpaa_intf, dev->process_private);
+ if (ret) {
+ DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
+ dev->data->name, ret);
+ }
}
}
@@ -575,6 +577,23 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
rte_free(dpaa_intf->fc_conf);
dpaa_intf->fc_conf = NULL;
+ /** For FMCLESS mode of share MAC, deconfig FM to direct
+ * ingress traffic to kernel before fq shutdown.
+ */
+ if (!(default_q || fmc_q)) {
+ ret = dpaa_fm_deconfig(dpaa_intf, dev->process_private);
+ if (ret) {
+ DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
+ dev->data->name, ret);
+ }
+ }
+ if (fif->num_profiles) {
+ ret = dpaa_port_vsp_cleanup(dpaa_intf, fif);
+ if (ret) {
+ DPAA_PMD_WARN("%s: cleanup VSP failed(%d)",
+ dev->data->name, ret);
+ }
+ }
/** Release congestion Groups after releasing FQIDs*/
/* Release RX congestion Groups */
if (dpaa_intf->cgr_rx) {
@@ -647,12 +666,10 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
rte_free(dpaa_intf->tx_conf_queues);
dpaa_intf->tx_conf_queues = NULL;
- if (dpaa_intf->port_handle) {
- ret = dpaa_fm_deconfig(dpaa_intf, fif);
- if (ret) {
- DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
- dev->data->name, ret);
- }
+ ret = dpaa_fm_deconfig(dpaa_intf, fif);
+ if (ret) {
+ DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
+ dev->data->name, ret);
}
if (fif->num_profiles) {
ret = dpaa_port_vsp_cleanup(dpaa_intf, fif);
diff --git a/drivers/net/dpaa/dpaa_flow.c b/drivers/net/dpaa/dpaa_flow.c
index f21950f64d..bfe294d21d 100644
--- a/drivers/net/dpaa/dpaa_flow.c
+++ b/drivers/net/dpaa/dpaa_flow.c
@@ -728,6 +728,9 @@ int dpaa_fm_deconfig(struct dpaa_if *dpaa_intf,
PMD_INIT_FUNC_TRACE();
+ if (!dpaa_intf->port_handle)
+ return 0;
+
/* FM PORT Disable */
ret = fm_port_disable(dpaa_intf->port_handle);
if (ret != E_OK) {
@@ -787,10 +790,8 @@ int dpaa_fm_config(struct rte_eth_dev *dev, uint64_t req_dist_set)
unsigned int i = 0;
PMD_INIT_FUNC_TRACE();
- if (dpaa_intf->port_handle) {
- if (dpaa_fm_deconfig(dpaa_intf, fif))
- DPAA_PMD_ERR("DPAA FM deconfig failed");
- }
+ if (dpaa_fm_deconfig(dpaa_intf, fif))
+ DPAA_PMD_ERR("DPAA FM deconfig failed");
if (!dev->data->nb_rx_queues)
return 0;
--
2.25.1
^ permalink raw reply related
* [PATCH 12/18] net/dpaa: optimize fm_deconfig
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Consolidate FM deconfiguration to avoid duplicate calls.
Move the fm_deconfig call to a single location and remove
redundant checks in the device close path.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 37 +++++++++++++++++++++++++---------
drivers/net/dpaa/dpaa_flow.c | 9 +++++----
2 files changed, 32 insertions(+), 14 deletions(-)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 3b66fbe18a..a223703421 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -526,10 +526,12 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
/* DPAA FM deconfig */
if (!(default_q || fmc_q)) {
- ret = dpaa_fm_deconfig(dpaa_intf, dev->process_private);
- if (ret) {
- DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
- dev->data->name, ret);
+ if (dpaa_intf->port_handle) {
+ ret = dpaa_fm_deconfig(dpaa_intf, dev->process_private);
+ if (ret) {
+ DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
+ dev->data->name, ret);
+ }
}
}
@@ -575,6 +577,23 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
rte_free(dpaa_intf->fc_conf);
dpaa_intf->fc_conf = NULL;
+ /** For FMCLESS mode of share MAC, deconfig FM to direct
+ * ingress traffic to kernel before fq shutdown.
+ */
+ if (!(default_q || fmc_q)) {
+ ret = dpaa_fm_deconfig(dpaa_intf, dev->process_private);
+ if (ret) {
+ DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
+ dev->data->name, ret);
+ }
+ }
+ if (fif->num_profiles) {
+ ret = dpaa_port_vsp_cleanup(dpaa_intf, fif);
+ if (ret) {
+ DPAA_PMD_WARN("%s: cleanup VSP failed(%d)",
+ dev->data->name, ret);
+ }
+ }
/** Release congestion Groups after releasing FQIDs*/
/* Release RX congestion Groups */
if (dpaa_intf->cgr_rx) {
@@ -647,12 +666,10 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
rte_free(dpaa_intf->tx_conf_queues);
dpaa_intf->tx_conf_queues = NULL;
- if (dpaa_intf->port_handle) {
- ret = dpaa_fm_deconfig(dpaa_intf, fif);
- if (ret) {
- DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
- dev->data->name, ret);
- }
+ ret = dpaa_fm_deconfig(dpaa_intf, fif);
+ if (ret) {
+ DPAA_PMD_WARN("%s: FM deconfig failed(%d)",
+ dev->data->name, ret);
}
if (fif->num_profiles) {
ret = dpaa_port_vsp_cleanup(dpaa_intf, fif);
diff --git a/drivers/net/dpaa/dpaa_flow.c b/drivers/net/dpaa/dpaa_flow.c
index f21950f64d..bfe294d21d 100644
--- a/drivers/net/dpaa/dpaa_flow.c
+++ b/drivers/net/dpaa/dpaa_flow.c
@@ -728,6 +728,9 @@ int dpaa_fm_deconfig(struct dpaa_if *dpaa_intf,
PMD_INIT_FUNC_TRACE();
+ if (!dpaa_intf->port_handle)
+ return 0;
+
/* FM PORT Disable */
ret = fm_port_disable(dpaa_intf->port_handle);
if (ret != E_OK) {
@@ -787,10 +790,8 @@ int dpaa_fm_config(struct rte_eth_dev *dev, uint64_t req_dist_set)
unsigned int i = 0;
PMD_INIT_FUNC_TRACE();
- if (dpaa_intf->port_handle) {
- if (dpaa_fm_deconfig(dpaa_intf, fif))
- DPAA_PMD_ERR("DPAA FM deconfig failed");
- }
+ if (dpaa_fm_deconfig(dpaa_intf, fif))
+ DPAA_PMD_ERR("DPAA FM deconfig failed");
if (!dev->data->nb_rx_queues)
return 0;
--
2.25.1
^ permalink raw reply related
* [PATCH 11/18] net/dpaa: remove redundant FQ shutdown from Rx queue setup
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Remove the redundant qman_shutdown_fq() call from
dpaa_eth_rx_queue_setup(). The FQ is shut down during device stop,
so calling it again at queue setup time is unnecessary and may
interfere with a clean queue initialization.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index b427b1df13..3b66fbe18a 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -1157,9 +1157,6 @@ int dpaa_eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
DPAA_PMD_INFO("Rx queue setup for queue index: %d fq_id (0x%x)",
queue_idx, rxq->fqid);
- /* Shutdown FQ before configure */
- qman_shutdown_fq(rxq->fqid);
-
if (!fif->num_profiles) {
if (dpaa_intf->bp_info && dpaa_intf->bp_info->bp &&
dpaa_intf->bp_info->mp != mp) {
--
2.25.1
^ permalink raw reply related
* [PATCH 11/18] net/dpaa: remove redundant FQ shutdown from rx_queue_setup
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Remove the redundant qman_shutdown_fq() call from dpaa_eth_rx_queue_setup().
The FQ is shut down during device stop, so calling it again at queue setup
time is unnecessary and may interfere with a clean queue initialization.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index b427b1df13..3b66fbe18a 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -1157,9 +1157,6 @@ int dpaa_eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
DPAA_PMD_INFO("Rx queue setup for queue index: %d fq_id (0x%x)",
queue_idx, rxq->fqid);
- /* Shutdown FQ before configure */
- qman_shutdown_fq(rxq->fqid);
-
if (!fif->num_profiles) {
if (dpaa_intf->bp_info && dpaa_intf->bp_info->bp &&
dpaa_intf->bp_info->mp != mp) {
--
2.25.1
^ permalink raw reply related
* [PATCH 10/18] net/dpaa: clean Tx confirmation FQ on device stop
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Ensure the Tx confirmation FQ is also cleaned up during device
stop, preventing stale FQ state on subsequent device restarts.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index e1856580b9..b427b1df13 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -573,6 +573,7 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
/* release configuration memory */
rte_free(dpaa_intf->fc_conf);
+ dpaa_intf->fc_conf = NULL;
/** Release congestion Groups after releasing FQIDs*/
/* Release RX congestion Groups */
@@ -642,6 +643,10 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
rte_free(dpaa_intf->tx_queues);
dpaa_intf->tx_queues = NULL;
+
+ rte_free(dpaa_intf->tx_conf_queues);
+ dpaa_intf->tx_conf_queues = NULL;
+
if (dpaa_intf->port_handle) {
ret = dpaa_fm_deconfig(dpaa_intf, fif);
if (ret) {
@@ -2537,6 +2542,8 @@ dpaa_dev_init(struct rte_eth_dev *eth_dev)
return 0;
free_tx:
+ rte_free(dpaa_intf->tx_conf_queues);
+ dpaa_intf->tx_conf_queues= NULL;
rte_free(dpaa_intf->tx_queues);
dpaa_intf->tx_queues = NULL;
dpaa_intf->nb_tx_queues = 0;
--
2.25.1
^ permalink raw reply related
* [PATCH 10/18] net/dpaa: clean tx_conf_fq on device stop
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
Ensure the Tx confirmation FQ is also cleaned up during device
stop, preventing stale FQ state on subsequent device restarts.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/net/dpaa/dpaa_ethdev.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index e1856580b9..b427b1df13 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -573,6 +573,7 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
/* release configuration memory */
rte_free(dpaa_intf->fc_conf);
+ dpaa_intf->fc_conf = NULL;
/** Release congestion Groups after releasing FQIDs*/
/* Release RX congestion Groups */
@@ -642,6 +643,10 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
rte_free(dpaa_intf->tx_queues);
dpaa_intf->tx_queues = NULL;
+
+ rte_free(dpaa_intf->tx_conf_queues);
+ dpaa_intf->tx_conf_queues = NULL;
+
if (dpaa_intf->port_handle) {
ret = dpaa_fm_deconfig(dpaa_intf, fif);
if (ret) {
@@ -2537,6 +2542,8 @@ dpaa_dev_init(struct rte_eth_dev *eth_dev)
return 0;
free_tx:
+ rte_free(dpaa_intf->tx_conf_queues);
+ dpaa_intf->tx_conf_queues= NULL;
rte_free(dpaa_intf->tx_queues);
dpaa_intf->tx_queues = NULL;
dpaa_intf->nb_tx_queues = 0;
--
2.25.1
^ permalink raw reply related
* [PATCH 09/18] drivers: add DPAA cgrid cleanup support
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Add qman_find_fq_by_cgid() to find frame queues associated with
a given CGID. This allows the driver to verify that all FQs
using a CGR are shut down before releasing the CGR ID, preventing
use-after-free of CGR resources.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/bus/dpaa/dpaa_bus_base_symbols.c | 1 +
drivers/bus/dpaa/include/fsl_qman.h | 3 +++
drivers/net/dpaa/dpaa_ethdev.c | 29 ++++++++++++++++++++++--
3 files changed, 31 insertions(+), 2 deletions(-)
diff --git a/drivers/bus/dpaa/dpaa_bus_base_symbols.c b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
index 52abec2b4c..514ab7b1f1 100644
--- a/drivers/bus/dpaa/dpaa_bus_base_symbols.c
+++ b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
@@ -56,6 +56,7 @@ RTE_EXPORT_INTERNAL_SYMBOL(qman_reserve_fqid_range)
RTE_EXPORT_INTERNAL_SYMBOL(qman_alloc_pool_range)
RTE_EXPORT_INTERNAL_SYMBOL(qman_alloc_cgrid_range)
RTE_EXPORT_INTERNAL_SYMBOL(qman_release_cgrid_range)
+RTE_EXPORT_INTERNAL_SYMBOL(qman_find_fq_by_cgrid)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_intr_enable)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_intr_disable)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_get_ioctl_version_number)
diff --git a/drivers/bus/dpaa/include/fsl_qman.h b/drivers/bus/dpaa/include/fsl_qman.h
index bd46207232..20321ed355 100644
--- a/drivers/bus/dpaa/include/fsl_qman.h
+++ b/drivers/bus/dpaa/include/fsl_qman.h
@@ -1907,6 +1907,9 @@ static inline int qman_shutdown_fq_by_fqid(u32 fqid)
return qman_shutdown_fq(&fq);
}
+__rte_internal
+int qman_find_fq_by_cgrid(u32 cgrid, u32 *fqid);
+
/**
* qman_reserve_fqid_range - Reserve the specified range of frame queue IDs
* @fqid: the base FQID of the range to deallocate
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 1615cf3892..e1856580b9 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -511,7 +511,7 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
struct rte_eth_link *link = &dev->data->dev_link;
struct dpaa_if *dpaa_intf = dev->data->dev_private;
struct qman_fq *fq;
- int loop;
+ uint32_t fqid, loop;
int ret;
PMD_INIT_FUNC_TRACE();
@@ -574,28 +574,53 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
/* release configuration memory */
rte_free(dpaa_intf->fc_conf);
+ /** Release congestion Groups after releasing FQIDs*/
/* Release RX congestion Groups */
if (dpaa_intf->cgr_rx) {
for (loop = 0; loop < dpaa_intf->nb_rx_queues; loop++) {
+ ret = qman_find_fq_by_cgrid(dpaa_intf->cgr_rx[loop].cgrid, &fqid);
+ if (!ret) {
+ /** Should be FQ not cleaned in previous program.*/
+ DPAA_PMD_DEBUG("FQ(fqid=0x%x) with rx cgid=%d is still alive?",
+ fqid, dpaa_intf->cgr_rx[loop].cgrid);
+ ret = qman_shutdown_fq_by_fqid(fqid);
+ if (ret) {
+ DPAA_PMD_WARN("Failed(%d) to shutdown fq(fqid=0x%x)",
+ ret, fqid);
+ }
+ }
ret = qman_delete_cgr(&dpaa_intf->cgr_rx[loop]);
if (ret) {
DPAA_PMD_WARN("%s: delete rxq%d's cgr err(%d)",
dev->data->name, loop, ret);
}
}
+ qman_release_cgrid_range(dpaa_intf->cgr_rx[0].cgrid, dpaa_intf->nb_rx_queues);
rte_free(dpaa_intf->cgr_rx);
dpaa_intf->cgr_rx = NULL;
}
/* Release TX congestion Groups */
if (dpaa_intf->cgr_tx) {
- for (loop = 0; loop < MAX_DPAA_CORES; loop++) {
+ for (loop = 0; loop < dpaa_intf->nb_tx_queues; loop++) {
+ ret = qman_find_fq_by_cgrid(dpaa_intf->cgr_tx[loop].cgrid, &fqid);
+ if (!ret) {
+ /** Should be FQ not cleaned in previous program.*/
+ DPAA_PMD_DEBUG("FQ(fqid=0x%x) with tx cgid=%d is still alive?",
+ fqid, dpaa_intf->cgr_tx[loop].cgrid);
+ ret = qman_shutdown_fq_by_fqid(fqid);
+ if (ret) {
+ DPAA_PMD_WARN("Failed(%d) to shutdown fq(fqid=0x%x)",
+ ret, fqid);
+ }
+ }
ret = qman_delete_cgr(&dpaa_intf->cgr_tx[loop]);
if (ret) {
DPAA_PMD_WARN("%s: delete txq%d's cgr err(%d)",
dev->data->name, loop, ret);
}
}
+ qman_release_cgrid_range(dpaa_intf->cgr_tx[0].cgrid, dpaa_intf->nb_tx_queues);
rte_free(dpaa_intf->cgr_tx);
dpaa_intf->cgr_tx = NULL;
}
--
2.25.1
^ permalink raw reply related
* [PATCH 09/18] bus/dpaa: add cgrid cleanup support
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Add qman_find_fq_by_cgid() to find frame queues associated with
a given CGID. This allows the driver to verify that all FQs
using a CGR are shut down before releasing the CGR ID, preventing
use-after-free of CGR resources.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
---
drivers/bus/dpaa/dpaa_bus_base_symbols.c | 1 +
drivers/bus/dpaa/include/fsl_qman.h | 3 +++
drivers/net/dpaa/dpaa_ethdev.c | 29 ++++++++++++++++++++++--
3 files changed, 31 insertions(+), 2 deletions(-)
diff --git a/drivers/bus/dpaa/dpaa_bus_base_symbols.c b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
index 52abec2b4c..514ab7b1f1 100644
--- a/drivers/bus/dpaa/dpaa_bus_base_symbols.c
+++ b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
@@ -56,6 +56,7 @@ RTE_EXPORT_INTERNAL_SYMBOL(qman_reserve_fqid_range)
RTE_EXPORT_INTERNAL_SYMBOL(qman_alloc_pool_range)
RTE_EXPORT_INTERNAL_SYMBOL(qman_alloc_cgrid_range)
RTE_EXPORT_INTERNAL_SYMBOL(qman_release_cgrid_range)
+RTE_EXPORT_INTERNAL_SYMBOL(qman_find_fq_by_cgrid)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_intr_enable)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_intr_disable)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_get_ioctl_version_number)
diff --git a/drivers/bus/dpaa/include/fsl_qman.h b/drivers/bus/dpaa/include/fsl_qman.h
index bd46207232..20321ed355 100644
--- a/drivers/bus/dpaa/include/fsl_qman.h
+++ b/drivers/bus/dpaa/include/fsl_qman.h
@@ -1907,6 +1907,9 @@ static inline int qman_shutdown_fq_by_fqid(u32 fqid)
return qman_shutdown_fq(&fq);
}
+__rte_internal
+int qman_find_fq_by_cgrid(u32 cgrid, u32 *fqid);
+
/**
* qman_reserve_fqid_range - Reserve the specified range of frame queue IDs
* @fqid: the base FQID of the range to deallocate
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 1615cf3892..e1856580b9 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -511,7 +511,7 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
struct rte_eth_link *link = &dev->data->dev_link;
struct dpaa_if *dpaa_intf = dev->data->dev_private;
struct qman_fq *fq;
- int loop;
+ uint32_t fqid, loop;
int ret;
PMD_INIT_FUNC_TRACE();
@@ -574,28 +574,53 @@ static int dpaa_eth_dev_close(struct rte_eth_dev *dev)
/* release configuration memory */
rte_free(dpaa_intf->fc_conf);
+ /** Release congestion Groups after releasing FQIDs*/
/* Release RX congestion Groups */
if (dpaa_intf->cgr_rx) {
for (loop = 0; loop < dpaa_intf->nb_rx_queues; loop++) {
+ ret = qman_find_fq_by_cgrid(dpaa_intf->cgr_rx[loop].cgrid, &fqid);
+ if (!ret) {
+ /** Should be FQ not cleaned in previous program.*/
+ DPAA_PMD_DEBUG("FQ(fqid=0x%x) with rx cgid=%d is still alive?",
+ fqid, dpaa_intf->cgr_rx[loop].cgrid);
+ ret = qman_shutdown_fq_by_fqid(fqid);
+ if (ret) {
+ DPAA_PMD_WARN("Failed(%d) to shutdown fq(fqid=0x%x)",
+ ret, fqid);
+ }
+ }
ret = qman_delete_cgr(&dpaa_intf->cgr_rx[loop]);
if (ret) {
DPAA_PMD_WARN("%s: delete rxq%d's cgr err(%d)",
dev->data->name, loop, ret);
}
}
+ qman_release_cgrid_range(dpaa_intf->cgr_rx[0].cgrid, dpaa_intf->nb_rx_queues);
rte_free(dpaa_intf->cgr_rx);
dpaa_intf->cgr_rx = NULL;
}
/* Release TX congestion Groups */
if (dpaa_intf->cgr_tx) {
- for (loop = 0; loop < MAX_DPAA_CORES; loop++) {
+ for (loop = 0; loop < dpaa_intf->nb_tx_queues; loop++) {
+ ret = qman_find_fq_by_cgrid(dpaa_intf->cgr_tx[loop].cgrid, &fqid);
+ if (!ret) {
+ /** Should be FQ not cleaned in previous program.*/
+ DPAA_PMD_DEBUG("FQ(fqid=0x%x) with tx cgid=%d is still alive?",
+ fqid, dpaa_intf->cgr_tx[loop].cgrid);
+ ret = qman_shutdown_fq_by_fqid(fqid);
+ if (ret) {
+ DPAA_PMD_WARN("Failed(%d) to shutdown fq(fqid=0x%x)",
+ ret, fqid);
+ }
+ }
ret = qman_delete_cgr(&dpaa_intf->cgr_tx[loop]);
if (ret) {
DPAA_PMD_WARN("%s: delete txq%d's cgr err(%d)",
dev->data->name, loop, ret);
}
}
+ qman_release_cgrid_range(dpaa_intf->cgr_tx[0].cgrid, dpaa_intf->nb_tx_queues);
rte_free(dpaa_intf->cgr_tx);
dpaa_intf->cgr_tx = NULL;
}
--
2.25.1
^ permalink raw reply related
* [PATCH 08/18] bus/dpaa: enhance DPAA FQ shutdown
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Gagandeep Singh
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Gagandeep Singh <g.singh@nxp.com>
Improve the FQ shutdown sequence to handle edge cases more
robustly, including better handling of ORL (Order Restoration
List) presence and improved error recovery paths.
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>
---
drivers/bus/dpaa/base/qbman/qman.c | 79 ++++++++++++++++++++++--------
1 file changed, 59 insertions(+), 20 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/qman.c b/drivers/bus/dpaa/base/qbman/qman.c
index 9d2ca25f0e..9832c0be24 100644
--- a/drivers/bus/dpaa/base/qbman/qman.c
+++ b/drivers/bus/dpaa/base/qbman/qman.c
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2016 Freescale Semiconductor Inc.
- * Copyright 2017,2019-2025 NXP
+ * Copyright 2017,2019-2026 NXP
*
*/
@@ -2843,9 +2843,10 @@ qman_shutdown_fq(struct qman_fq *fq)
}
res = mcr->result; /* Make a copy as we reuse MCR below */
- if (res == QM_MCR_RESULT_OK) {
+ if (res == QM_MCR_RESULT_OK)
drain_mr_fqrni(&p->p);
- } else if (res == QM_MCR_RESULT_PENDING) {
+
+ if (res == QM_MCR_RESULT_PENDING) {
/*
* Need to wait for the FQRN in the message ring, which
* will only occur once the FQ has been drained. In
@@ -2853,27 +2854,29 @@ qman_shutdown_fq(struct qman_fq *fq)
* to dequeue from the channel the FQ is scheduled on
*/
int found_fqrn = 0;
- const u16 pool_ch_start = dpaa_get_qm_channel_pool();
- const u16 pool_ch_end = pool_ch_start + dpaa_get_qm_channel_pool_num();
- u32 sdqcr = p->sdqcr;
/* Flag that we need to drain FQ */
drain = 1;
+ const u16 pool_ch_start = dpaa_get_qm_channel_pool();
+ const u16 pool_ch_end = pool_ch_start +
+ dpaa_get_qm_channel_pool_num();
if (channel >= pool_ch_start && channel < pool_ch_end) {
- /* Pool channel, enable the bit in the portal */
+ /* Pool channel - must use affine portal */
if (p->config->channel != channel) {
- DPAA_BUS_ERR("Portal affine channel(0x%04x) != wq channel(0x%04x)",
+ DPAA_BUS_ERR("Portal ch(0x%04x) != FQ ch(0x%04x)",
p->config->channel, channel);
ret = -EINVAL;
goto out;
}
} else if (channel < pool_ch_start) {
/* Dedicated channel */
- sdqcr = QM_SDQCR_TYPE_ACTIVE | QM_SDQCR_CHANNELS_DEDICATED;
- qm_dqrr_sdqcr_set(&p->p, sdqcr);
+ qm_dqrr_sdqcr_set(&p->p,
+ QM_SDQCR_TYPE_ACTIVE |
+ QM_SDQCR_CHANNELS_DEDICATED);
} else {
- DPAA_BUS_ERR("Can't recover FQ 0x%x, Invalid channel: 0x%x", fqid, channel);
+ DPAA_BUS_ERR("Invalid channel 0x%x for FQ 0x%x",
+ channel, fqid);
ret = -EBUSY;
goto out;
}
@@ -2881,15 +2884,16 @@ qman_shutdown_fq(struct qman_fq *fq)
/* Keep draining DQRR while checking the MR*/
qm_dqrr_drain_nomatch(&p->p);
/* Process message ring too */
- found_fqrn = qm_mr_drain(&p->p,
- FQRN);
+ found_fqrn = qm_mr_drain(&p->p, FQRN);
cpu_relax();
} while (!found_fqrn);
- /* Restore SDQCR */
- if (sdqcr != p->sdqcr)
- qm_dqrr_sdqcr_set(&p->p, p->sdqcr);
- } else {
- DPAA_BUS_ERR("retire_fq failed: FQ 0x%x, res=0x%x", fqid, res);
+ qm_dqrr_sdqcr_set(&p->p, p->sdqcr);
+
+ }
+ if (res != QM_MCR_RESULT_OK &&
+ res != QM_MCR_RESULT_PENDING) {
+ DPAA_BUS_ERR("retire_fq failed: FQ 0x%x, res=0x%x",
+ fqid, res);
ret = -EIO;
goto out;
}
@@ -2932,7 +2936,7 @@ qman_shutdown_fq(struct qman_fq *fq)
if (mcr->result != QM_MCR_RESULT_OK) {
DPAA_BUS_ERR("OOS after drain fail: FQ 0x%x (0x%x)",
- fqid, mcr->result);
+ fqid, mcr->result);
ret = -EIO;
goto out;
}
@@ -2951,7 +2955,7 @@ qman_shutdown_fq(struct qman_fq *fq)
if (mcr->result != QM_MCR_RESULT_OK) {
DPAA_BUS_ERR("OOS fail: FQ 0x%x (0x%x)",
- fqid, mcr->result);
+ fqid, mcr->result);
ret = -EIO;
goto out;
}
@@ -2968,3 +2972,38 @@ qman_shutdown_fq(struct qman_fq *fq)
out:
return ret;
}
+
+int qman_find_fq_by_cgrid(u32 cgrid, u32 *fqid)
+{
+ struct qman_fq fq = {
+ .fqid = 1
+ };
+ struct qm_mcr_queryfq_np np;
+ struct qm_fqd fqd;
+ int err;
+
+ do {
+ err = qman_query_fq_np(&fq, &np);
+ if (err == -ERANGE) {
+ DPAA_BUS_INFO("No FQ found with cgrid(0x%x)", cgrid);
+ return err;
+ } else if (err) {
+ DPAA_BUS_WARN("Failed(%d) to Query np FQ(fqid=0x%x)", err, fq.fqid);
+ return err;
+ }
+ if ((np.state & QM_MCR_NP_STATE_MASK) != QM_MCR_NP_STATE_OOS) {
+ err = qman_query_fq(&fq, &fqd);
+ if (err) {
+ DPAA_BUS_WARN("Failed(%d) to Query FQ(fqid=0x%x)", err, fq.fqid);
+ } else if ((fqd.fq_ctrl & QM_FQCTRL_CGE) && fqd.cgid == cgrid) {
+ if (fqid)
+ *fqid = fq.fqid;
+ return 0;
+ }
+ }
+ /* Move to the next FQID */
+ fq.fqid++;
+ } while (1);
+
+ return -ENODEV;
+}
--
2.25.1
^ permalink raw reply related
* [PATCH 07/18] bus/dpaa: improve FQ shutdown with channel validation
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Fix hardcoded channel range check by using DTS-derived pool
channel start/end values. Add validation that the portal's
affine channel matches the FQ's channel for pool-channel FQs,
and only restore SDQCR when it was actually changed.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/qman.c | 55 ++++++++++-------------
drivers/bus/dpaa/base/qbman/qman_driver.c | 29 ++++++++++--
drivers/bus/dpaa/dpaa_bus_base_symbols.c | 2 +
drivers/bus/dpaa/include/fsl_qman.h | 8 ++--
4 files changed, 54 insertions(+), 40 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/qman.c b/drivers/bus/dpaa/base/qbman/qman.c
index 2315b81065..9d2ca25f0e 100644
--- a/drivers/bus/dpaa/base/qbman/qman.c
+++ b/drivers/bus/dpaa/base/qbman/qman.c
@@ -2792,7 +2792,7 @@ qman_shutdown_fq(struct qman_fq *fq)
int orl_empty, drain = 0, ret = 0;
u32 res, fqid = fq->fqid;
u8 state;
- u32 channel, wq;
+ u16 channel;
DPAA_BUS_DEBUG("In shutdown for queue = %x", fqid);
if (!p)
@@ -2806,9 +2806,10 @@ qman_shutdown_fq(struct qman_fq *fq)
ret = -ETIMEDOUT;
goto out;
}
+
state = mcr->queryfq_np.state & QM_MCR_NP_STATE_MASK;
if (state == QM_MCR_NP_STATE_OOS) {
- DPAA_BUS_ERR("Already in OOS");
+ DPAA_BUS_DEBUG("fqid(0x%x) Already in OOS", fqid);
goto out; /* Already OOS, no need to do anymore checks */
}
@@ -2824,7 +2825,6 @@ qman_shutdown_fq(struct qman_fq *fq)
/* Need to store these since the MCR gets reused */
channel = qm_fqd_get_chan(&mcr->queryfq.fqd);
- wq = qm_fqd_get_wq(&mcr->queryfq.fqd);
switch (state) {
case QM_MCR_NP_STATE_TEN_SCHED:
@@ -2843,10 +2843,9 @@ qman_shutdown_fq(struct qman_fq *fq)
}
res = mcr->result; /* Make a copy as we reuse MCR below */
- if (res == QM_MCR_RESULT_OK)
+ if (res == QM_MCR_RESULT_OK) {
drain_mr_fqrni(&p->p);
-
- if (res == QM_MCR_RESULT_PENDING) {
+ } else if (res == QM_MCR_RESULT_PENDING) {
/*
* Need to wait for the FQRN in the message ring, which
* will only occur once the FQ has been drained. In
@@ -2854,35 +2853,30 @@ qman_shutdown_fq(struct qman_fq *fq)
* to dequeue from the channel the FQ is scheduled on
*/
int found_fqrn = 0;
+ const u16 pool_ch_start = dpaa_get_qm_channel_pool();
+ const u16 pool_ch_end = pool_ch_start + dpaa_get_qm_channel_pool_num();
+ u32 sdqcr = p->sdqcr;
/* Flag that we need to drain FQ */
drain = 1;
- __maybe_unused u16 dequeue_wq = 0;
- if (channel >= qm_channel_pool1 &&
- channel < (u16)(qm_channel_pool1 + 15)) {
+ if (channel >= pool_ch_start && channel < pool_ch_end) {
/* Pool channel, enable the bit in the portal */
- dequeue_wq = (channel -
- qm_channel_pool1 + 1) << 4 | wq;
- } else if (channel < qm_channel_pool1) {
+ if (p->config->channel != channel) {
+ DPAA_BUS_ERR("Portal affine channel(0x%04x) != wq channel(0x%04x)",
+ p->config->channel, channel);
+ ret = -EINVAL;
+ goto out;
+ }
+ } else if (channel < pool_ch_start) {
/* Dedicated channel */
- dequeue_wq = wq;
+ sdqcr = QM_SDQCR_TYPE_ACTIVE | QM_SDQCR_CHANNELS_DEDICATED;
+ qm_dqrr_sdqcr_set(&p->p, sdqcr);
} else {
- DPAA_BUS_ERR("Can't recover FQ 0x%x, ch: 0x%x",
- fqid, channel);
+ DPAA_BUS_ERR("Can't recover FQ 0x%x, Invalid channel: 0x%x", fqid, channel);
ret = -EBUSY;
goto out;
}
- /* Set the sdqcr to drain this channel */
- if (channel < qm_channel_pool1)
- qm_dqrr_sdqcr_set(&p->p,
- QM_SDQCR_TYPE_ACTIVE |
- QM_SDQCR_CHANNELS_DEDICATED);
- else
- qm_dqrr_sdqcr_set(&p->p,
- QM_SDQCR_TYPE_ACTIVE |
- QM_SDQCR_CHANNELS_POOL_CONV
- (channel));
do {
/* Keep draining DQRR while checking the MR*/
qm_dqrr_drain_nomatch(&p->p);
@@ -2892,13 +2886,10 @@ qman_shutdown_fq(struct qman_fq *fq)
cpu_relax();
} while (!found_fqrn);
/* Restore SDQCR */
- qm_dqrr_sdqcr_set(&p->p,
- p->sdqcr);
- }
- if (res != QM_MCR_RESULT_OK &&
- res != QM_MCR_RESULT_PENDING) {
- DPAA_BUS_ERR("retire_fq failed: FQ 0x%x, res=0x%x",
- fqid, res);
+ if (sdqcr != p->sdqcr)
+ qm_dqrr_sdqcr_set(&p->p, p->sdqcr);
+ } else {
+ DPAA_BUS_ERR("retire_fq failed: FQ 0x%x, res=0x%x", fqid, res);
ret = -EIO;
goto out;
}
diff --git a/drivers/bus/dpaa/base/qbman/qman_driver.c b/drivers/bus/dpaa/base/qbman/qman_driver.c
index 45b094e0c6..51a360d36e 100644
--- a/drivers/bus/dpaa/base/qbman/qman_driver.c
+++ b/drivers/bus/dpaa/base/qbman/qman_driver.c
@@ -17,9 +17,10 @@
* where CCSR isn't available).
*/
u16 qman_ip_rev;
-u16 qm_channel_pool1 = QMAN_CHANNEL_POOL1;
-u16 qm_channel_caam = QMAN_CHANNEL_CAAM;
-u16 qm_channel_pme = QMAN_CHANNEL_PME;
+static u16 qm_channel_pool1 = QMAN_CHANNEL_POOL1;
+static u16 qm_channel_caam = QMAN_CHANNEL_CAAM;
+static u16 qm_channel_pme = QMAN_CHANNEL_PME;
+static u16 qm_channel_pool_num;
/* Ccsr map address to access ccsrbased register */
static void *qman_ccsr_map;
@@ -65,6 +66,11 @@ u16 dpaa_get_qm_channel_pool(void)
return qm_channel_pool1;
}
+u16 dpaa_get_qm_channel_pool_num(void)
+{
+ return qm_channel_pool_num;
+}
+
static int fsl_qman_portal_init(uint32_t index, int is_shared)
{
struct qman_portal *portal;
@@ -275,7 +281,7 @@ int qman_global_init(void)
uint64_t phys_addr;
uint64_t regs_size;
const u32 *clk;
-
+ u16 pool_channel;
static int done;
if (done)
@@ -336,6 +342,21 @@ int qman_global_init(void)
return -EINVAL;
}
+ if (lenp != sizeof(rte_be32_t) * 2) {
+ pr_err("pool-channel-range should have 2 items.\n");
+ return -EINVAL;
+ }
+ pool_channel = rte_be_to_cpu_32(chanid[0]);
+ qm_channel_pool_num = rte_be_to_cpu_32(chanid[1]);
+
+ if (pool_channel != qm_channel_pool1) {
+ pr_warn("Pool channel(%04x) configured != default(0x%04x)\n",
+ pool_channel, qm_channel_pool1);
+ }
+ qm_channel_pool1 = pool_channel;
+ pr_debug("Pool channel starts from 0x%04x, number=%d, lenp:%ld\n",
+ qm_channel_pool1, qm_channel_pool_num, lenp);
+
/* get ccsr base */
dt_node = of_find_compatible_node(NULL, NULL, "fsl,qman");
if (!dt_node) {
diff --git a/drivers/bus/dpaa/dpaa_bus_base_symbols.c b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
index 522cdca27e..52abec2b4c 100644
--- a/drivers/bus/dpaa/dpaa_bus_base_symbols.c
+++ b/drivers/bus/dpaa/dpaa_bus_base_symbols.c
@@ -1,5 +1,6 @@
/* SPDX-License-Identifier: BSD-3-Clause
* Copyright (c) 2025 Red Hat, Inc.
+ * Copyright 2026 NXP
*/
#include <eal_export.h>
@@ -94,6 +95,7 @@ RTE_EXPORT_INTERNAL_SYMBOL(qman_create_cgr)
RTE_EXPORT_INTERNAL_SYMBOL(qman_delete_cgr)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_get_qm_channel_caam)
RTE_EXPORT_INTERNAL_SYMBOL(dpaa_get_qm_channel_pool)
+RTE_EXPORT_INTERNAL_SYMBOL(dpaa_get_qm_channel_pool_num)
RTE_EXPORT_INTERNAL_SYMBOL(qman_thread_fd)
RTE_EXPORT_INTERNAL_SYMBOL(qman_thread_irq)
RTE_EXPORT_INTERNAL_SYMBOL(qman_fq_portal_thread_irq)
diff --git a/drivers/bus/dpaa/include/fsl_qman.h b/drivers/bus/dpaa/include/fsl_qman.h
index 673859ed2e..bd46207232 100644
--- a/drivers/bus/dpaa/include/fsl_qman.h
+++ b/drivers/bus/dpaa/include/fsl_qman.h
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2012 Freescale Semiconductor, Inc.
- * Copyright 2019-2022 NXP
+ * Copyright 2019-2022, 2026 NXP
*
*/
@@ -35,9 +35,6 @@ extern "C" {
#define QMAN_CHANNEL_POOL1_REV3 0x401
#define QMAN_CHANNEL_CAAM_REV3 0x840
#define QMAN_CHANNEL_PME_REV3 0x860
-extern u16 qm_channel_pool1;
-extern u16 qm_channel_caam;
-extern u16 qm_channel_pme;
enum qm_dc_portal {
qm_dc_portal_fman0 = 0,
qm_dc_portal_fman1 = 1,
@@ -51,6 +48,9 @@ u16 dpaa_get_qm_channel_caam(void);
__rte_internal
u16 dpaa_get_qm_channel_pool(void);
+__rte_internal
+u16 dpaa_get_qm_channel_pool_num(void);
+
/* Portal processing (interrupt) sources */
#define QM_PIRQ_CCSCI 0x00200000 /* CEETM Congestion State Change */
#define QM_PIRQ_CSCI 0x00100000 /* Congestion State Change */
--
2.25.1
^ permalink raw reply related
* [PATCH 06/18] drivers: shutdown DPAA FQ by fq descriptor
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Pass the full FQ descriptor to qman_shutdown_fq() instead of
just the fqid, so that channel-affine portals can be correctly
accessed when shutting down push-mode Rx queues.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/qman.c | 9 +++++----
drivers/bus/dpaa/include/fsl_qman.h | 11 ++++++++++-
drivers/net/dpaa/dpaa_ethdev.c | 7 +++++++
3 files changed, 22 insertions(+), 5 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/qman.c b/drivers/bus/dpaa/base/qbman/qman.c
index 931d487de4..2315b81065 100644
--- a/drivers/bus/dpaa/base/qbman/qman.c
+++ b/drivers/bus/dpaa/base/qbman/qman.c
@@ -2784,18 +2784,19 @@ qm_mc_result_timeout(struct qm_portal *portal,
RTE_EXPORT_INTERNAL_SYMBOL(qman_shutdown_fq)
int
-qman_shutdown_fq(u32 fqid)
+qman_shutdown_fq(struct qman_fq *fq)
{
- struct qman_portal *p;
+ struct qman_portal *p = fq->qp;
struct qm_mc_command *mcc;
struct qm_mc_result *mcr;
int orl_empty, drain = 0, ret = 0;
- u32 res;
+ u32 res, fqid = fq->fqid;
u8 state;
u32 channel, wq;
DPAA_BUS_DEBUG("In shutdown for queue = %x", fqid);
- p = get_affine_portal();
+ if (!p)
+ p = get_affine_portal();
/* Determine the state of the FQID */
mcc = qm_mc_start(&p->p);
mcc->queryfq_np.fqid = cpu_to_be32(fqid);
diff --git a/drivers/bus/dpaa/include/fsl_qman.h b/drivers/bus/dpaa/include/fsl_qman.h
index 82269cdf99..673859ed2e 100644
--- a/drivers/bus/dpaa/include/fsl_qman.h
+++ b/drivers/bus/dpaa/include/fsl_qman.h
@@ -1896,7 +1896,16 @@ static inline void qman_release_fqid(u32 fqid)
void qman_seed_fqid_range(u32 fqid, unsigned int count);
__rte_internal
-int qman_shutdown_fq(u32 fqid);
+int qman_shutdown_fq(struct qman_fq *fq);
+
+static inline int qman_shutdown_fq_by_fqid(u32 fqid)
+{
+ struct qman_fq fq;
+
+ memset(&fq, 0, sizeof(struct qman_fq));
+ fq.fqid = fqid;
+ return qman_shutdown_fq(&fq);
+}
/**
* qman_reserve_fqid_range - Reserve the specified range of frame queue IDs
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 70d1e9acc3..1615cf3892 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -2363,6 +2363,13 @@ dpaa_dev_init(struct rte_eth_dev *eth_dev)
vsp_id = dev_vspids[loop];
+ /* Shutdown FQ before configure to clean the queue */
+ ret = qman_shutdown_fq_by_fqid(fqid);
+ if (ret < 0) {
+ DPAA_PMD_ERR("Failed shutdown %s:rxq-%d-fqid = 0x%08x",
+ dpaa_intf->name, loop, fqid);
+ }
+
if (dpaa_intf->cgr_rx)
dpaa_intf->cgr_rx[loop].cgrid = cgrid[loop];
--
2.25.1
^ permalink raw reply related
* [PATCH 06/18] bus/dpaa: shutdown FQ by fq descriptor
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Pass the full FQ descriptor to qman_shutdown_fq() instead of
just the fqid, so that channel-affine portals can be correctly
accessed when shutting down push-mode Rx queues.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/qman.c | 9 +++++----
drivers/bus/dpaa/include/fsl_qman.h | 11 ++++++++++-
drivers/net/dpaa/dpaa_ethdev.c | 7 +++++++
3 files changed, 22 insertions(+), 5 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/qman.c b/drivers/bus/dpaa/base/qbman/qman.c
index 931d487de4..2315b81065 100644
--- a/drivers/bus/dpaa/base/qbman/qman.c
+++ b/drivers/bus/dpaa/base/qbman/qman.c
@@ -2784,18 +2784,19 @@ qm_mc_result_timeout(struct qm_portal *portal,
RTE_EXPORT_INTERNAL_SYMBOL(qman_shutdown_fq)
int
-qman_shutdown_fq(u32 fqid)
+qman_shutdown_fq(struct qman_fq *fq)
{
- struct qman_portal *p;
+ struct qman_portal *p = fq->qp;
struct qm_mc_command *mcc;
struct qm_mc_result *mcr;
int orl_empty, drain = 0, ret = 0;
- u32 res;
+ u32 res, fqid = fq->fqid;
u8 state;
u32 channel, wq;
DPAA_BUS_DEBUG("In shutdown for queue = %x", fqid);
- p = get_affine_portal();
+ if (!p)
+ p = get_affine_portal();
/* Determine the state of the FQID */
mcc = qm_mc_start(&p->p);
mcc->queryfq_np.fqid = cpu_to_be32(fqid);
diff --git a/drivers/bus/dpaa/include/fsl_qman.h b/drivers/bus/dpaa/include/fsl_qman.h
index 82269cdf99..673859ed2e 100644
--- a/drivers/bus/dpaa/include/fsl_qman.h
+++ b/drivers/bus/dpaa/include/fsl_qman.h
@@ -1896,7 +1896,16 @@ static inline void qman_release_fqid(u32 fqid)
void qman_seed_fqid_range(u32 fqid, unsigned int count);
__rte_internal
-int qman_shutdown_fq(u32 fqid);
+int qman_shutdown_fq(struct qman_fq *fq);
+
+static inline int qman_shutdown_fq_by_fqid(u32 fqid)
+{
+ struct qman_fq fq;
+
+ memset(&fq, 0, sizeof(struct qman_fq));
+ fq.fqid = fqid;
+ return qman_shutdown_fq(&fq);
+}
/**
* qman_reserve_fqid_range - Reserve the specified range of frame queue IDs
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 70d1e9acc3..1615cf3892 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -2363,6 +2363,13 @@ dpaa_dev_init(struct rte_eth_dev *eth_dev)
vsp_id = dev_vspids[loop];
+ /* Shutdown FQ before configure to clean the queue */
+ ret = qman_shutdown_fq_by_fqid(fqid);
+ if (ret < 0) {
+ DPAA_PMD_ERR("Failed shutdown %s:rxq-%d-fqid = 0x%08x",
+ dpaa_intf->name, loop, fqid);
+ }
+
if (dpaa_intf->cgr_rx)
dpaa_intf->cgr_rx[loop].cgrid = cgrid[loop];
--
2.25.1
^ permalink raw reply related
* [PATCH 05/18] bus/dpaa: define helpers for qman channel and wq
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Add inline helper functions to extract channel and work queue
from a frame queue descriptor, replacing open-coded bit
manipulation throughout the driver.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/qbman/qman.c | 14 ++------------
drivers/bus/dpaa/base/qbman/qman.h | 23 ++++++++++++++++++++++-
2 files changed, 24 insertions(+), 13 deletions(-)
diff --git a/drivers/bus/dpaa/base/qbman/qman.c b/drivers/bus/dpaa/base/qbman/qman.c
index d289df2d33..931d487de4 100644
--- a/drivers/bus/dpaa/base/qbman/qman.c
+++ b/drivers/bus/dpaa/base/qbman/qman.c
@@ -2707,14 +2707,6 @@ int qman_delete_cgr(struct qman_cgr *cgr)
return ret;
}
-#define GENMASK(h, l) \
- (((~0U) >> (sizeof(unsigned int) * 8 - ((h) - (l) + 1))) << (l))
-
-/* 'fqid' is a 24-bit field in every h/w descriptor */
-#define QM_FQID_MASK GENMASK(23, 0)
-#define qm_fqid_set(p, v) ((p)->fqid = cpu_to_be32((v) & QM_FQID_MASK))
-#define qm_fqid_get(p) (be32_to_cpu((p)->fqid) & QM_FQID_MASK)
-
static int
_qm_mr_consume_and_match_verb(struct qm_portal *p, int v)
{
@@ -2801,7 +2793,6 @@ qman_shutdown_fq(u32 fqid)
u32 res;
u8 state;
u32 channel, wq;
- u16 dest_wq;
DPAA_BUS_DEBUG("In shutdown for queue = %x", fqid);
p = get_affine_portal();
@@ -2831,9 +2822,8 @@ qman_shutdown_fq(u32 fqid)
}
/* Need to store these since the MCR gets reused */
- dest_wq = be16_to_cpu(mcr->queryfq.fqd.dest_wq);
- channel = dest_wq & 0x7;
- wq = dest_wq >> 3;
+ channel = qm_fqd_get_chan(&mcr->queryfq.fqd);
+ wq = qm_fqd_get_wq(&mcr->queryfq.fqd);
switch (state) {
case QM_MCR_NP_STATE_TEN_SCHED:
diff --git a/drivers/bus/dpaa/base/qbman/qman.h b/drivers/bus/dpaa/base/qbman/qman.h
index 43a16d1e3b..bd97689a91 100644
--- a/drivers/bus/dpaa/base/qbman/qman.h
+++ b/drivers/bus/dpaa/base/qbman/qman.h
@@ -1,12 +1,15 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
*
* Copyright 2008-2016 Freescale Semiconductor Inc.
- * Copyright 2017 NXP
+ * Copyright 2017,2026 NXP
*
*/
#include "qman_priv.h"
+#define GENMASK(h, l) \
+ (((~0U) >> (sizeof(u32) * 8 - ((h) - (l) + 1))) << (l))
+
/***************************/
/* Portal register assists */
/***************************/
@@ -42,6 +45,14 @@
#define QM_CL_RR0 0x3900
#define QM_CL_RR1 0x3940
+#define QM_FQD_CHAN_OFF 3
+#define QM_FQD_WQ_MASK GENMASK(2, 0)
+/* 'fqid' is a 24-bit field in every h/w descriptor */
+#define QM_FQID_MASK GENMASK(23, 0)
+
+#define qm_fqid_set(p, v) ((p)->fqid = cpu_to_be32((v) & QM_FQID_MASK))
+#define qm_fqid_get(p) (be32_to_cpu((p)->fqid) & QM_FQID_MASK)
+
/* BTW, the drivers (and h/w programming model) already obtain the required
* synchronisation for portal accesses via lwsync(), hwsync(), and
* data-dependencies. Use of barrier()s or other order-preserving primitives
@@ -911,3 +922,13 @@ static inline void __qm_isr_write(struct qm_portal *portal, enum qm_isr_reg n,
__qm_out(&portal->addr, QM_REG_ISR + (n << 2), val);
#endif
}
+
+static inline int qm_fqd_get_chan(const struct qm_fqd *fqd)
+{
+ return be16_to_cpu(fqd->dest_wq) >> QM_FQD_CHAN_OFF;
+}
+
+static inline int qm_fqd_get_wq(const struct qm_fqd *fqd)
+{
+ return be16_to_cpu(fqd->dest_wq) & QM_FQD_WQ_MASK;
+}
--
2.25.1
^ permalink raw reply related
* [PATCH 04/18] dpaa: add process-type guards to prevent segfaults in secondary
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Prashant Gupta
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Prashant Gupta <prashant.gupta_3@nxp.com>
Add RTE_PROC_PRIMARY checks in device initialization paths for
net/dpaa, crypto/dpaa_sec and dma/dpaa drivers. Secondary
processes should skip hardware initialization to prevent
segfaults when accessing hardware registers that are only
mapped in the primary process.
Signed-off-by: Prashant Gupta <prashant.gupta_3@nxp.com>
---
drivers/crypto/dpaa_sec/dpaa_sec.c | 3 ---
drivers/dma/dpaa/dpaa_qdma.c | 4 ++++
drivers/net/dpaa/dpaa_ethdev.c | 3 +++
3 files changed, 7 insertions(+), 3 deletions(-)
diff --git a/drivers/crypto/dpaa_sec/dpaa_sec.c b/drivers/crypto/dpaa_sec/dpaa_sec.c
index 65bbd38b17..36f5819b0e 100644
--- a/drivers/crypto/dpaa_sec/dpaa_sec.c
+++ b/drivers/crypto/dpaa_sec/dpaa_sec.c
@@ -3783,9 +3783,6 @@ cryptodev_dpaa_sec_probe(struct rte_dpaa_driver *dpaa_drv __rte_unused,
RTE_DPAA_MAX_NB_SEC_QPS,
};
- if (rte_eal_process_type() != RTE_PROC_PRIMARY)
- return 0;
-
snprintf(cryptodev_name, sizeof(cryptodev_name), "%s", dpaa_dev->name);
cryptodev = rte_cryptodev_pmd_create(cryptodev_name, &dpaa_dev->device, &init_params);
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/dma/dpaa/dpaa_qdma.c
index b20ff24ab6..bf2a373d70 100644
--- a/drivers/dma/dpaa/dpaa_qdma.c
+++ b/drivers/dma/dpaa/dpaa_qdma.c
@@ -1334,6 +1334,10 @@ dpaa_qdma_init(struct rte_dma_dev *dmadev)
int regs_size;
int ret;
uint32_t i, j, k;
+ char *penv;
+
+ if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+ return -ENOTSUP;
if (dpaa_get_devargs(dmadev->device->devargs, DPAA_DMA_ERROR_CHECK)) {
s_hw_err_check = true;
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index ebd50d340d..70d1e9acc3 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -2685,6 +2685,9 @@ rte_dpaa_remove(struct rte_dpaa_device *dpaa_dev)
PMD_INIT_FUNC_TRACE();
+ if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+ return 0;
+
eth_dev = dpaa_dev->eth_dev;
if (eth_dev->state != RTE_ETH_DEV_UNUSED) {
dpaa_eth_dev_close(eth_dev);
--
2.25.1
^ permalink raw reply related
* [PATCH 03/18] net/dpaa: add BMI Tx statistics
From: Hemant Agrawal @ 2026-06-19 12:29 UTC (permalink / raw)
To: stephen, david.marchand, dev; +Cc: Jun Yang
In-Reply-To: <20260619122922.3774666-1-hemant.agrawal@nxp.com>
From: Jun Yang <jun.yang@nxp.com>
Add support for BMI (Buffer Manager Interface) Tx statistics
counters. Extend fman_hw to read Tx BMI registers and expose
them through the xstats interface.
Signed-off-by: Jun Yang <jun.yang@nxp.com>
---
drivers/bus/dpaa/base/fman/fman_hw.c | 2 --
drivers/bus/dpaa/include/fman.h | 24 ++++++++++++++++++++++++
drivers/net/dpaa/dpaa_ethdev.c | 10 +++++++++-
drivers/net/dpaa/dpaa_ethdev.h | 11 +++++++++--
4 files changed, 42 insertions(+), 5 deletions(-)
diff --git a/drivers/bus/dpaa/base/fman/fman_hw.c b/drivers/bus/dpaa/base/fman/fman_hw.c
index ce68581555..aab04bf76a 100644
--- a/drivers/bus/dpaa/base/fman/fman_hw.c
+++ b/drivers/bus/dpaa/base/fman/fman_hw.c
@@ -301,7 +301,6 @@ fman_if_bmi_stats_enable(struct fman_if *p)
uint32_t tmp;
tmp = in_be32(®s->fmbm_rstc);
-
tmp |= FMAN_BMI_COUNTERS_EN;
out_be32(®s->fmbm_rstc, tmp);
@@ -315,7 +314,6 @@ fman_if_bmi_stats_disable(struct fman_if *p)
uint32_t tmp;
tmp = in_be32(®s->fmbm_rstc);
-
tmp &= ~FMAN_BMI_COUNTERS_EN;
out_be32(®s->fmbm_rstc, tmp);
diff --git a/drivers/bus/dpaa/include/fman.h b/drivers/bus/dpaa/include/fman.h
index 6e35af57d7..3e17e0e0b7 100644
--- a/drivers/bus/dpaa/include/fman.h
+++ b/drivers/bus/dpaa/include/fman.h
@@ -306,6 +306,21 @@ struct tx_bmi_regs {
uint32_t fmbm_tfene; /**< Tx Frame Enqueue Next Engine*/
uint32_t fmbm_trlmts; /**< Tx Rate Limiter Scale*/
uint32_t fmbm_trlmt; /**< Tx Rate Limiter*/
+ uint32_t reserved0034[0x73];/**< (0x0034 0x01FF) */
+ uint32_t fmbm_tstc; /**< Tx Statistics Counters*/
+ uint32_t fmbm_tfrc; /**< Tx Frame Counter*/
+ uint32_t fmbm_tfdc; /**< Tx Frames Discard Counter*/
+ uint32_t fmbm_tfledc; /**< Tx Frames Length Error Discard*/
+ uint32_t fmbm_tfufdc; /**< Tx Frames Unsupported Format*/
+ uint32_t fmbm_tbdc; /**< Tx Buffers Deallocate Counter */
+ uint32_t reserved0218[0x1a];/**< (0x0218 0x027F) */
+ uint32_t fmbm_tpc; /**< Tx Performance Counters*/
+ uint32_t fmbm_tpcp; /**< Tx Performance Count Parameters */
+ uint32_t fmbm_tccn; /**< Tx Cycle Counter*/
+ uint32_t fmbm_ttuc; /**< Tx Tasks Utilization Counter */
+ uint32_t fmbm_ttcquc; /**< Tx Transmit Confirm Queue Utilization Counter*/
+ uint32_t fmbm_tduc; /**< Tx DMA Utilization Counter */
+ uint32_t fmbm_tfuc; /**< Tx FIFO Utilization Counter */
};
/* Description FM RTC timer alarm */
@@ -468,6 +483,15 @@ struct __fman_if {
void *qmi_map;
};
+#define MEMMAC_REG_OFFSET(reg) offsetof(struct memac_regs, reg)
+#define BMI_RX_REG_OFFSET(reg) offsetof(struct rx_bmi_regs, reg)
+#define BMI_TX_REG_OFFSET(reg) offsetof(struct tx_bmi_regs, reg)
+
+#define FMAN_IF_BMI_RX_STAT_OFFSET_START BMI_RX_REG_OFFSET(fmbm_rfrc)
+#define FMAN_IF_BMI_RX_STAT_OFFSET_END BMI_RX_REG_OFFSET(fmbm_rbdc)
+#define FMAN_IF_BMI_TX_STAT_OFFSET_START BMI_TX_REG_OFFSET(fmbm_tfrc)
+#define FMAN_IF_BMI_TX_STAT_OFFSET_END BMI_TX_REG_OFFSET(fmbm_tbdc)
+
/* And this is the base list node that the interfaces are added to. (See
* fman_if_enable_all_rx() below for an example of its use.)
*/
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 3ca3455ab7..ebd50d340d 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright 2016 Freescale Semiconductor, Inc. All rights reserved.
- * Copyright 2017-2020,2022-2025 NXP
+ * Copyright 2017-2020,2022-2026 NXP
*
*/
/* System headers */
@@ -143,6 +143,14 @@ static const struct rte_dpaa_xstats_name_off dpaa_xstats_strings[] = {
offsetof(struct dpaa_if_rx_bmi_stats, fmbm_rodc)},
{"rx_buf_deallocate",
offsetof(struct dpaa_if_rx_bmi_stats, fmbm_rbdc)},
+ {"tx_bad_frames_count",
+ offsetof(struct dpaa_if_tx_bmi_stats, fmbm_tfdc)},
+ {"tx_frame_length_discard",
+ offsetof(struct dpaa_if_tx_bmi_stats, fmbm_tfledc)},
+ {"tx_frames_unsupported_format",
+ offsetof(struct dpaa_if_tx_bmi_stats, fmbm_tfufdc)},
+ {"tx_buf_diallocate",
+ offsetof(struct dpaa_if_tx_bmi_stats, fmbm_tbdc)},
};
static struct rte_dpaa_driver rte_dpaa_pmd;
diff --git a/drivers/net/dpaa/dpaa_ethdev.h b/drivers/net/dpaa/dpaa_ethdev.h
index f400030a5c..d342d98f23 100644
--- a/drivers/net/dpaa/dpaa_ethdev.h
+++ b/drivers/net/dpaa/dpaa_ethdev.h
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 2014-2016 Freescale Semiconductor, Inc. All rights reserved.
- * Copyright 2017-2024 NXP
+ * Copyright 2017-2026 NXP
*
*/
#ifndef __DPAA_ETHDEV_H__
@@ -234,7 +234,6 @@ dpaa_rx_cb_atomic(void *event,
void **bufs);
struct dpaa_if_rx_bmi_stats {
- uint32_t fmbm_rstc; /**< Rx Statistics Counters*/
uint32_t fmbm_rfrc; /**< Rx Frame Counter*/
uint32_t fmbm_rfbc; /**< Rx Bad Frames Counter*/
uint32_t fmbm_rlfc; /**< Rx Large Frames Counter*/
@@ -245,6 +244,14 @@ struct dpaa_if_rx_bmi_stats {
uint32_t fmbm_rbdc; /**< Rx Buffers Deallocate Counter*/
};
+struct dpaa_if_tx_bmi_stats {
+ uint32_t fmbm_tfrc; /**< Tx Frame Counter*/
+ uint32_t fmbm_tfdc; /**< Tx Frames Discard Counter*/
+ uint32_t fmbm_tfledc; /**< Tx Frames Length Error Discard*/
+ uint32_t fmbm_tfufdc; /**< Tx Frames Unsupported Format*/
+ uint32_t fmbm_tbdc; /**< Tx Buffers Deallocate Counter */
+};
+
int
dpaa_tx_conf_queue_init(struct qman_fq *fq);
--
2.25.1
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox