[PATCH 0/7] fix XXX_DEBUG warnings

[PATCH 0/7] fix XXX_DEBUG warnings

[Stephen Hemminger]

I recently modified my build test to automatically enable all
DEBUG flags. This is because developers often add debug code,
forget about it, and it bit rots.

This patch series fixes all the places where #ifdef XXX_DEBUG
code was present but would not build without warnings.

Stephen Hemminger (7):
  bus/fslmc: fix build of qbman debug code
  bus/fslmc: fix signed comparison in buffer acquire
  crypto/caam_jr: fix descriptor dump build
  crypto/ipsec_mb: fix missing header for debug build
  power/amd_uncore: fix debug log build
  net/rnp: fix register debug log format
  net/vmxnet3: remove unused queue dump functions

 drivers/bus/fslmc/qbman/qbman_debug.c   |  3 ++
 drivers/bus/fslmc/qbman/qbman_portal.c  |  4 +-
 drivers/crypto/caam_jr/caam_jr.c        |  2 +-
 drivers/crypto/caam_jr/caam_jr_desc.h   |  6 +--
 drivers/crypto/ipsec_mb/pmd_aesni_gcm.c |  2 +
 drivers/net/rnp/base/rnp_osdep.h        |  4 +-
 drivers/net/vmxnet3/vmxnet3_rxtx.c      | 58 -------------------------
 drivers/power/amd_uncore/amd_uncore.c   |  2 +-
 8 files changed, 14 insertions(+), 67 deletions(-)

-- 
2.53.0

[PATCH 1/7] bus/fslmc: fix build of qbman debug code

[Stephen Hemminger]

QBMAN_RESPONSE_VERB_MASK is a private define in qbman_portal.c, not
in the qbman_portal.h header included by qbman_debug.c.
The debug file has therefore never compiled since it was added;
it fails with seven "undeclared identifier" errors when
RTE_LIBRTE_DPAA2_DEBUG_BUS is defined.

Define the mask locally, next to the other management command codes
that qbman_debug.c already defines for itself.

Fixes: 64f131a82fbe ("bus/fslmc: add qbman debug")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/bus/fslmc/qbman/qbman_debug.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/bus/fslmc/qbman/qbman_debug.c b/drivers/bus/fslmc/qbman/qbman_debug.c
index f13168dce3..1c5c62addb 100644
--- a/drivers/bus/fslmc/qbman/qbman_debug.c
+++ b/drivers/bus/fslmc/qbman/qbman_debug.c
@@ -9,6 +9,9 @@
 
 #include <eal_export.h>
 
+/* Mask used to extract the response verb from a management command result */
+#define QBMAN_RESPONSE_VERB_MASK  0x7f
+
 /* QBMan portal management command code */
 #define QBMAN_BP_QUERY            0x32
 #define QBMAN_FQ_QUERY            0x44
-- 
2.53.0

[PATCH 2/7] bus/fslmc: fix signed comparison in buffer acquire

[Stephen Hemminger]

The acquired buffer count is masked with BMAN_VALID_RSLT_NUM_MASK so
it is always in the range 0..7, but it was held in an int and then
compared against the unsigned num_buffers parameter, giving a
-Wsign-compare warning in both acquire paths.

Make the count unsigned to match num_buffers.  The error paths return
negative errno values directly and are unaffected.

Fixes: a116979a03c6 ("bus/fslmc: improve BMAN buffer acquire")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/bus/fslmc/qbman/qbman_portal.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/bus/fslmc/qbman/qbman_portal.c b/drivers/bus/fslmc/qbman/qbman_portal.c
index c93bec5dd3..7b666d4ef9 100644
--- a/drivers/bus/fslmc/qbman/qbman_portal.c
+++ b/drivers/bus/fslmc/qbman/qbman_portal.c
@@ -2680,7 +2680,7 @@ static int qbman_swp_acquire_direct(struct qbman_swp *s, uint16_t bpid,
 {
 	struct qbman_acquire_desc *p;
 	struct qbman_acquire_rslt *r;
-	int num;
+	unsigned int num;
 
 	if (!num_buffers || (num_buffers > BMAN_VALID_RSLT_NUM_MASK))
 		return -EINVAL;
@@ -2727,7 +2727,7 @@ static int qbman_swp_acquire_cinh_direct(struct qbman_swp *s, uint16_t bpid,
 {
 	struct qbman_acquire_desc *p;
 	struct qbman_acquire_rslt *r;
-	int num;
+	unsigned int num;
 
 	if (!num_buffers || (num_buffers > BMAN_VALID_RSLT_NUM_MASK))
 		return -EINVAL;
-- 
2.53.0

[PATCH 3/7] crypto/caam_jr: fix descriptor dump build

[Stephen Hemminger]

SEC_DUMP_DESC() grew a FILE * argument when the NXP drivers were
converted to fprintf(), but the only caller was never updated, so
building with RTE_LIBRTE_PMD_CAAM_JR_DEBUG failed with a macro
arity error.

Compiling the macro body for the first time then exposed two
further problems in it: the log format string ended in a newline,
which RTE_LOG_LINE() rejects, and descriptor pointers were cast to uint32_t,
truncating them on 64-bit builds.

Fixes: a8794e398279 ("drivers: use fprintf for debug dumps in NXP drivers")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/crypto/caam_jr/caam_jr.c      | 2 +-
 drivers/crypto/caam_jr/caam_jr_desc.h | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/crypto/caam_jr/caam_jr.c b/drivers/crypto/caam_jr/caam_jr.c
index a57dc56b80..104ce52d11 100644
--- a/drivers/crypto/caam_jr/caam_jr.c
+++ b/drivers/crypto/caam_jr/caam_jr.c
@@ -462,7 +462,7 @@ caam_jr_prep_cdb(struct caam_jr_session *ses)
 	}
 
 #if CAAM_JR_DBG
-	SEC_DUMP_DESC(cdb->sh_desc);
+	SEC_DUMP_DESC(cdb->sh_desc, stdout);
 #endif
 
 	cdb->sh_hdr.hi.field.idlen = shared_desc_len;
diff --git a/drivers/crypto/caam_jr/caam_jr_desc.h b/drivers/crypto/caam_jr/caam_jr_desc.h
index eb4e68578a..845facec26 100644
--- a/drivers/crypto/caam_jr/caam_jr_desc.h
+++ b/drivers/crypto/caam_jr/caam_jr_desc.h
@@ -112,12 +112,12 @@
 /* Helper macro for dumping the hex representation of a descriptor */
 #define SEC_DUMP_DESC(descriptor, f) {					\
 	int __i;							\
-	CAAM_JR_INFO("Des@ 0x%08x\n", (uint32_t)((uint32_t *)(descriptor)));\
+	CAAM_JR_INFO("Des@ %p", (const void *)(descriptor));		\
 	for (__i = 0;						\
 		__i < SEC_GET_DESC_LEN(descriptor);			\
 		__i++) {						\
-		fprintf(f, "0x%08x: 0x%08x\n",			\
-			(uint32_t)(((uint32_t *)(descriptor)) + __i),	\
+		fprintf(f, "%p: 0x%08x\n",			\
+			(const void *)(((uint32_t *)(descriptor)) + __i), \
 			*(((uint32_t *)(descriptor)) + __i));		\
 	}								\
 }
-- 
2.53.0

[PATCH 4/7] crypto/ipsec_mb: fix missing header for debug build

[Stephen Hemminger]

Building with RTE_LIBRTE_PMD_AESNI_GCM_DEBUG fails with an implicit
declaration of rte_hexdump()

Fixes: 746825e5c0ea ("crypto/ipsec_mb: move aesni_gcm PMD")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/crypto/ipsec_mb/pmd_aesni_gcm.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/crypto/ipsec_mb/pmd_aesni_gcm.c b/drivers/crypto/ipsec_mb/pmd_aesni_gcm.c
index 8c35820ef7..3c266cc47e 100644
--- a/drivers/crypto/ipsec_mb/pmd_aesni_gcm.c
+++ b/drivers/crypto/ipsec_mb/pmd_aesni_gcm.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2016-2021 Intel Corporation
  */
 
+#include <rte_hexdump.h>
+
 #include "pmd_aesni_gcm_priv.h"
 
 static void
-- 
2.53.0

[PATCH 5/7] power/amd_uncore: fix debug log build

[Stephen Hemminger]

The debug log referenced ui->num_uncore_freqs, which is the field name
used by the intel_uncore driver.  The AMD uncore_power_info struct
calls it nb_freqs, so building with RTE_LIBRTE_POWER_DEBUG failed.

Fixes: da4d64d0e803 ("power/amd_uncore: add uncore for AMD EPYC processors")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/power/amd_uncore/amd_uncore.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/power/amd_uncore/amd_uncore.c b/drivers/power/amd_uncore/amd_uncore.c
index c3e95cdc08..b341cfef75 100644
--- a/drivers/power/amd_uncore/amd_uncore.c
+++ b/drivers/power/amd_uncore/amd_uncore.c
@@ -100,7 +100,7 @@ power_get_available_uncore_freqs(struct uncore_power_info *ui)
 	}
 
 	POWER_DEBUG_LOG("%d frequency(s) of pkg %02u die %02u are available",
-			ui->num_uncore_freqs, ui->pkg, ui->die);
+			ui->nb_freqs, ui->pkg, ui->die);
 
 	return 0;
 }
-- 
2.53.0

[PATCH 6/7] net/rnp: fix register debug log format

[Stephen Hemminger]

The register offset is a size_t but was printed with %p, producing a
-Wformat warning at every call site (22 in a full build) when
RTE_LIBRTE_RNP_REG_DEBUG is enabled.  Print it with %zx instead.

Fixes: 18d555f74fcf ("net/rnp: add mailbox basic operations")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/rnp/base/rnp_osdep.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/rnp/base/rnp_osdep.h b/drivers/net/rnp/base/rnp_osdep.h
index 58dcb40f90..a95a37f2fe 100644
--- a/drivers/net/rnp/base/rnp_osdep.h
+++ b/drivers/net/rnp/base/rnp_osdep.h
@@ -90,14 +90,14 @@ rnp_reg_read32(const void *base, size_t offset)
 {
 	u32 v = rte_read32(((const u8 *)base + offset));
 
-	RNP_PMD_REG_LOG(DEBUG, "offset=%p val=%#"PRIx32"", offset, v);
+	RNP_PMD_REG_LOG(DEBUG, "offset=%#zx val=%#"PRIx32"", offset, v);
 	return v;
 }
 
 static inline void
 rnp_reg_write32(volatile void *base, size_t offset, u32 val)
 {
-	RNP_PMD_REG_LOG(DEBUG, "offset=%p val=%#"PRIx32"", offset, val);
+	RNP_PMD_REG_LOG(DEBUG, "offset=%#zx val=%#"PRIx32"", offset, val);
 	rte_write32(val, ((volatile u8 *)base + offset));
 }
 
-- 
2.53.0

[PATCH 7/7] net/vmxnet3: remove unused queue dump functions

[Stephen Hemminger]

vmxnet3_rxq_dump() and vmxnet3_txq_dump() are guarded by
RTE_LIBRTE_VMXNET3_DEBUG_DRIVER_NOT_USED, a macro that is defined
nowhere and whose name records that the callers were removed long
ago. Building with the macro forced on only produces -Wunused-function warnings.

Remove the dead code.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/vmxnet3/vmxnet3_rxtx.c | 58 ------------------------------
 1 file changed, 58 deletions(-)

diff --git a/drivers/net/vmxnet3/vmxnet3_rxtx.c b/drivers/net/vmxnet3/vmxnet3_rxtx.c
index 552b8a9823..aaebfc8675 100644
--- a/drivers/net/vmxnet3/vmxnet3_rxtx.c
+++ b/drivers/net/vmxnet3/vmxnet3_rxtx.c
@@ -59,64 +59,6 @@
 
 static int vmxnet3_post_rx_bufs(vmxnet3_rx_queue_t*, uint8_t);
 static void vmxnet3_tq_tx_complete(vmxnet3_tx_queue_t *);
-#ifdef RTE_LIBRTE_VMXNET3_DEBUG_DRIVER_NOT_USED
-static void vmxnet3_rxq_dump(struct vmxnet3_rx_queue *);
-static void vmxnet3_txq_dump(struct vmxnet3_tx_queue *);
-#endif
-
-#ifdef RTE_LIBRTE_VMXNET3_DEBUG_DRIVER_NOT_USED
-static void
-vmxnet3_rxq_dump(struct vmxnet3_rx_queue *rxq)
-{
-	uint32_t avail = 0;
-
-	if (rxq == NULL)
-		return;
-
-	PMD_RX_LOG(DEBUG,
-		   "RXQ: cmd0 base : %p cmd1 base : %p comp ring base : %p.",
-		   rxq->cmd_ring[0].base, rxq->cmd_ring[1].base, rxq->comp_ring.base);
-	PMD_RX_LOG(DEBUG,
-		   "RXQ: cmd0 basePA : 0x%lx cmd1 basePA : 0x%lx comp ring basePA : 0x%lx.",
-		   (unsigned long)rxq->cmd_ring[0].basePA,
-		   (unsigned long)rxq->cmd_ring[1].basePA,
-		   (unsigned long)rxq->comp_ring.basePA);
-
-	avail = vmxnet3_cmd_ring_desc_avail(&rxq->cmd_ring[0]);
-	PMD_RX_LOG(DEBUG,
-		   "RXQ:cmd0: size=%u; free=%u; next2proc=%u; queued=%u",
-		   (uint32_t)rxq->cmd_ring[0].size, avail,
-		   rxq->comp_ring.next2proc,
-		   rxq->cmd_ring[0].size - avail);
-
-	avail = vmxnet3_cmd_ring_desc_avail(&rxq->cmd_ring[1]);
-	PMD_RX_LOG(DEBUG, "RXQ:cmd1 size=%u; free=%u; next2proc=%u; queued=%u",
-		   (uint32_t)rxq->cmd_ring[1].size, avail, rxq->comp_ring.next2proc,
-		   rxq->cmd_ring[1].size - avail);
-
-}
-
-static void
-vmxnet3_txq_dump(struct vmxnet3_tx_queue *txq)
-{
-	uint32_t avail = 0;
-
-	if (txq == NULL)
-		return;
-
-	PMD_TX_LOG(DEBUG, "TXQ: cmd base : %p comp ring base : %p data ring base : %p.",
-		   txq->cmd_ring.base, txq->comp_ring.base, txq->data_ring.base);
-	PMD_TX_LOG(DEBUG, "TXQ: cmd basePA : 0x%lx comp ring basePA : 0x%lx data ring basePA : 0x%lx.",
-		   (unsigned long)txq->cmd_ring.basePA,
-		   (unsigned long)txq->comp_ring.basePA,
-		   (unsigned long)txq->data_ring.basePA);
-
-	avail = vmxnet3_cmd_ring_desc_avail(&txq->cmd_ring);
-	PMD_TX_LOG(DEBUG, "TXQ: size=%u; free=%u; next2proc=%u; queued=%u",
-		   (uint32_t)txq->cmd_ring.size, avail,
-		   txq->comp_ring.next2proc, txq->cmd_ring.size - avail);
-}
-#endif
 
 static void
 vmxnet3_tx_cmd_ring_release_mbufs(vmxnet3_cmd_ring_t *ring)
-- 
2.53.0

Re: [PATCH 1/7] bus/fslmc: fix build of qbman debug code

[hemant.agrawal]

On 05/09/2026 00:42, Stephen Hemminger wrote:
> QBMAN_RESPONSE_VERB_MASK is a private define in qbman_portal.c, not
> in the qbman_portal.h header included by qbman_debug.c.
> The debug file has therefore never compiled since it was added;
> it fails with seven "undeclared identifier" errors when
> RTE_LIBRTE_DPAA2_DEBUG_BUS is defined.
>
> Define the mask locally, next to the other management command codes
> that qbman_debug.c already defines for itself.
>
> Fixes: 64f131a82fbe ("bus/fslmc: add qbman debug")
> Cc: stable@dpdk.org
>
> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

> ---
>   drivers/bus/fslmc/qbman/qbman_debug.c | 3 +++
>   1 file changed, 3 insertions(+)
>
> diff --git a/drivers/bus/fslmc/qbman/qbman_debug.c b/drivers/bus/fslmc/qbman/qbman_debug.c
> index f13168dce3..1c5c62addb 100644
> --- a/drivers/bus/fslmc/qbman/qbman_debug.c
> +++ b/drivers/bus/fslmc/qbman/qbman_debug.c
> @@ -9,6 +9,9 @@
>   
>   #include <eal_export.h>
>   
> +/* Mask used to extract the response verb from a management command result */
> +#define QBMAN_RESPONSE_VERB_MASK  0x7f
> +
>   /* QBMan portal management command code */
>   #define QBMAN_BP_QUERY            0x32
>   #define QBMAN_FQ_QUERY            0x44