[PATCH] eal: introduce division round up macro

[PATCH] eal: introduce division round up macro

[Joshua Washington]

A number of drivers define the DIV_ROUND_UP macro -- add it to
RTE_COMMON so that other code looking to use that functionality doesn't
have to re-define it.

Replace usage of DIV_ROUND_UP in existing places with RTE_DIV_ROUND_UP.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 app/proc-info/main.c                        |  3 +--
 drivers/common/nfp/nfp_platform.h           |  2 --
 drivers/net/bnx2x/bnx2x.c                   |  4 ++--
 drivers/net/bnx2x/bnx2x.h                   | 11 ++++-------
 drivers/net/cxgbe/base/t4_hw.c              |  4 ++--
 drivers/net/cxgbe/base/t4vf_hw.c            | 10 +++++-----
 drivers/net/cxgbe/cxgbe_compat.h            |  3 +--
 drivers/net/cxgbe/cxgbe_filter.c            |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h              |  4 ++--
 drivers/net/cxgbe/sge.c                     | 14 +++++++-------
 drivers/net/ena/base/ena_eth_com.h          |  4 ++--
 drivers/net/ena/base/ena_plat_dpdk.h        |  1 -
 drivers/net/hinic/base/hinic_compat.h       |  5 +----
 drivers/net/hns3/hns3_ethdev.h              |  2 --
 drivers/net/hns3/hns3_fdir.c                |  2 +-
 drivers/net/hns3/hns3_regs.c                |  8 ++++----
 drivers/net/hns3/hns3_rss.c                 |  8 ++++----
 drivers/net/intel/i40e/base/i40e_osdep.h    |  3 +--
 drivers/net/intel/ice/base/ice_osdep.h      |  3 +--
 drivers/net/intel/idpf/base/idpf_osdep.h    |  3 +--
 drivers/net/nfp/nfdk/nfp_nfdk.h             |  5 +++--
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c          |  3 ++-
 drivers/net/nfp/nfpcore/nfp_nsp.c           |  2 +-
 drivers/net/qede/base/bcm_osal.h            |  2 --
 drivers/net/qede/base/ecore_chain.h         |  2 +-
 drivers/net/qede/base/ecore_cxt.c           | 20 +++++++++++---------
 drivers/net/qede/base/ecore_init_fw_funcs.c | 21 ++++++++++++---------
 drivers/net/qede/base/ecore_mcp.c           |  6 +++---
 drivers/net/qede/qede_debug.c               | 16 ++++++++--------
 drivers/net/rnp/base/rnp_osdep.h            |  5 ++---
 drivers/net/rnp/rnp_rxtx.c                  |  2 +-
 drivers/raw/ifpga/base/opae_osdep.h         |  3 +--
 lib/eal/include/rte_common.h                |  5 +++++
 33 files changed, 92 insertions(+), 100 deletions(-)

diff --git a/app/proc-info/main.c b/app/proc-info/main.c
index 5925bbb765..9c22fb8f01 100644
--- a/app/proc-info/main.c
+++ b/app/proc-info/main.c
@@ -48,7 +48,6 @@
 
 #define ETHDEV_FWVERS_LEN 32
 #define RTE_RETA_CONF_GROUP_NUM 32
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define EEPROM_DUMP_CHUNKSIZE 1024
 
 #define STATS_BDR_FMT "========================================"
@@ -1830,7 +1829,7 @@ show_port_rss_reta_info(void)
 			return;
 		}
 
-		num = DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
+		num = RTE_DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
 		memset(reta_conf, 0, sizeof(reta_conf));
 		for (i = 0; i < num; i++)
 			reta_conf[i].mask = ~0ULL;
diff --git a/drivers/common/nfp/nfp_platform.h b/drivers/common/nfp/nfp_platform.h
index 0b02fcf1e8..814f90b5a2 100644
--- a/drivers/common/nfp/nfp_platform.h
+++ b/drivers/common/nfp/nfp_platform.h
@@ -10,8 +10,6 @@
 
 #include <rte_bitops.h>
 
-#define DIV_ROUND_UP(n, d)             (((n) + (d) - 1) / (d))
-
 #define DMA_BIT_MASK(n)    ((1ULL << (n)) - 1)
 
 #define BITS_PER_LONG      (__SIZEOF_LONG__ * 8)
diff --git a/drivers/net/bnx2x/bnx2x.c b/drivers/net/bnx2x/bnx2x.c
index 8790c858d5..ce9dbdf7c6 100644
--- a/drivers/net/bnx2x/bnx2x.c
+++ b/drivers/net/bnx2x/bnx2x.c
@@ -2334,8 +2334,8 @@ static void bnx2x_ilt_set_info(struct bnx2x_softc *sc)
 		ilt_client->start = line;
 
 		/* 4 bytes for each cid */
-		line += DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
-				     QM_ILT_PAGE_SZ);
+		line += RTE_DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
+					 QM_ILT_PAGE_SZ);
 
 		ilt_client->end = (line - 1);
 	}
diff --git a/drivers/net/bnx2x/bnx2x.h b/drivers/net/bnx2x/bnx2x.h
index 35206b4758..0cdd118ad5 100644
--- a/drivers/net/bnx2x/bnx2x.h
+++ b/drivers/net/bnx2x/bnx2x.h
@@ -37,9 +37,6 @@
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(arr) RTE_DIM(arr)
 #endif
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-#endif
 #ifndef roundup
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #endif
@@ -393,7 +390,7 @@ union cdu_context {
 #define CNIC_ISCSI_CID_MAX 256
 #define CNIC_FCOE_CID_MAX  2048
 #define CNIC_CID_MAX       (CNIC_ISCSI_CID_MAX + CNIC_FCOE_CID_MAX)
-#define CNIC_ILT_LINES     DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
+#define CNIC_ILT_LINES     RTE_DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
 
 #define QM_ILT_PAGE_SZ_HW  0
 #define QM_ILT_PAGE_SZ     (4096 << QM_ILT_PAGE_SZ_HW) /* 4K */
@@ -405,7 +402,7 @@ union cdu_context {
 /*#define TM_CONN_NUM        (CNIC_STARTING_CID+CNIC_ISCSI_CXT_MAX) */
 #define TM_CONN_NUM        1024
 #define TM_ILT_SZ          (8 * TM_CONN_NUM)
-#define TM_ILT_LINES       DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
+#define TM_ILT_LINES       RTE_DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
 
 /* SRC (Searcher) host DB constants */
 #define SRC_ILT_PAGE_SZ_HW 0
@@ -414,7 +411,7 @@ union cdu_context {
 #define SRC_CONN_NUM       (1 << SRC_HASH_BITS) /* 1024 */
 #define SRC_ILT_SZ         (sizeof(struct src_ent) * SRC_CONN_NUM)
 #define SRC_T2_SZ          SRC_ILT_SZ
-#define SRC_ILT_LINES      DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
+#define SRC_ILT_LINES      RTE_DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
 
 struct hw_context {
     struct bnx2x_dma    vcxt_dma;
@@ -1258,7 +1255,7 @@ struct bnx2x_softc {
 #define BNX2X_L2_CID_COUNT(sc)                                             \
 	(BNX2X_NUM_ETH_QUEUES(sc) * ECORE_MULTI_TX_COS + 2 * CNIC_SUPPORT(sc))
 #define L2_ILT_LINES(sc)                                \
-	(DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
+	(RTE_DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
 
 	int qm_cid_count;
 
diff --git a/drivers/net/cxgbe/base/t4_hw.c b/drivers/net/cxgbe/base/t4_hw.c
index d02da39b84..944e173da2 100644
--- a/drivers/net/cxgbe/base/t4_hw.c
+++ b/drivers/net/cxgbe/base/t4_hw.c
@@ -2630,8 +2630,8 @@ static int t4_get_exprom_version(struct adapter *adapter, u32 *vers)
 		unsigned char hdr_arr[16];      /* must start with 0x55aa */
 		unsigned char hdr_ver[4];       /* Expansion ROM version */
 	} *hdr;
-	u32 exprom_header_buf[DIV_ROUND_UP(sizeof(struct exprom_header),
-					   sizeof(u32))];
+	u32 exprom_header_buf[RTE_DIV_ROUND_UP(sizeof(struct exprom_header),
+					       sizeof(u32))];
 	int ret;
 
 	ret = t4_read_flash(adapter, FLASH_EXP_ROM_START,
diff --git a/drivers/net/cxgbe/base/t4vf_hw.c b/drivers/net/cxgbe/base/t4vf_hw.c
index 573ccb1b05..561e4a91a4 100644
--- a/drivers/net/cxgbe/base/t4vf_hw.c
+++ b/drivers/net/cxgbe/base/t4vf_hw.c
@@ -362,8 +362,8 @@ int t4vf_query_params(struct adapter *adapter, unsigned int nparams,
 	cmd.op_to_vfn = cpu_to_be32(V_FW_CMD_OP(FW_PARAMS_CMD) |
 				    F_FW_CMD_REQUEST |
 				    F_FW_CMD_READ);
-	len16 = DIV_ROUND_UP(offsetof(struct fw_params_cmd,
-			     param[nparams]), 16);
+	len16 = RTE_DIV_ROUND_UP(offsetof(struct fw_params_cmd,
+					  param[nparams]), 16);
 	cmd.retval_len16 = cpu_to_be32(V_FW_CMD_LEN16(len16));
 	for (i = 0, p = &cmd.param[0]; i < nparams; i++, p++)
 		p->mnem = cpu_to_be32(*params++);
@@ -458,8 +458,8 @@ int t4vf_set_params(struct adapter *adapter, unsigned int nparams,
 	cmd.op_to_vfn = cpu_to_be32(V_FW_CMD_OP(FW_PARAMS_CMD) |
 				    F_FW_CMD_REQUEST |
 				    F_FW_CMD_WRITE);
-	len16 = DIV_ROUND_UP(offsetof(struct fw_params_cmd,
-			     param[nparams]), 16);
+	len16 = RTE_DIV_ROUND_UP(offsetof(struct fw_params_cmd,
+					  param[nparams]), 16);
 	cmd.retval_len16 = cpu_to_be32(V_FW_CMD_LEN16(len16));
 	for (i = 0, p = &cmd.param[0]; i < nparams; i++, p++) {
 		p->mnem = cpu_to_be32(*params++);
@@ -626,7 +626,7 @@ static int t4vf_get_port_stats_fw(struct adapter *adapter, int pidx,
 		struct fw_vi_stats_cmd cmd, rpl;
 		size_t len = (offsetof(struct fw_vi_stats_cmd, u) +
 			      sizeof(struct fw_vi_stats_ctl));
-		size_t len16 = DIV_ROUND_UP(len, 16);
+		size_t len16 = RTE_DIV_ROUND_UP(len, 16);
 		int ret;
 
 		memset(&cmd, 0, sizeof(cmd));
diff --git a/drivers/net/cxgbe/cxgbe_compat.h b/drivers/net/cxgbe/cxgbe_compat.h
index e74d476f6d..14ebe914eb 100644
--- a/drivers/net/cxgbe/cxgbe_compat.h
+++ b/drivers/net/cxgbe/cxgbe_compat.h
@@ -166,11 +166,10 @@ typedef uint64_t  dma_addr_t;
 typedef char *caddr_t;
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 static inline uint8_t hweight32(uint32_t word32)
 {
diff --git a/drivers/net/cxgbe/cxgbe_filter.c b/drivers/net/cxgbe/cxgbe_filter.c
index 5a7efe7a73..83f86a747a 100644
--- a/drivers/net/cxgbe/cxgbe_filter.c
+++ b/drivers/net/cxgbe/cxgbe_filter.c
@@ -219,7 +219,7 @@ static inline void mk_set_tcb_field_ulp(struct filter_entry *f,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*req) - sizeof(struct work_request_hdr));
 	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_SET_TCB_FIELD, f->tid));
@@ -371,7 +371,7 @@ static void mk_abort_req_ulp(struct cpl_abort_req *abort_req,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_req) -
 			      sizeof(struct work_request_hdr));
@@ -395,7 +395,7 @@ static void mk_abort_rpl_ulp(struct cpl_abort_rpl *abort_rpl,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_rpl), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_rpl), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_rpl) -
 			      sizeof(struct work_request_hdr));
diff --git a/drivers/net/cxgbe/cxgbe_ofld.h b/drivers/net/cxgbe/cxgbe_ofld.h
index fd1e7d8002..ef56eef426 100644
--- a/drivers/net/cxgbe/cxgbe_ofld.h
+++ b/drivers/net/cxgbe/cxgbe_ofld.h
@@ -14,7 +14,7 @@
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_TP_WR) | \
 				V_FW_WR_IMMDLEN(sizeof(*w) - sizeof(w->wr))); \
 	(w)->wr.wr_mid = cpu_to_be32( \
-				V_FW_WR_LEN16(DIV_ROUND_UP(sizeof(*w), 16)) | \
+				V_FW_WR_LEN16(RTE_DIV_ROUND_UP(sizeof(*w), 16)) | \
 				V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
@@ -27,7 +27,7 @@
 #define INIT_ULPTX_WR(w, wrlen, atomic, tid) do { \
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_ULPTX_WR) | \
 				    V_FW_WR_ATOMIC(atomic)); \
-	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(DIV_ROUND_UP(wrlen, 16)) | \
+	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(RTE_DIV_ROUND_UP(wrlen, 16)) | \
 				     V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
diff --git a/drivers/net/cxgbe/sge.c b/drivers/net/cxgbe/sge.c
index e9d45f24c4..eb25d80e33 100644
--- a/drivers/net/cxgbe/sge.c
+++ b/drivers/net/cxgbe/sge.c
@@ -459,7 +459,7 @@ static inline unsigned int sgl_len(unsigned int n)
  */
 static inline unsigned int flits_to_desc(unsigned int n)
 {
-	return DIV_ROUND_UP(n, 8);
+	return RTE_DIV_ROUND_UP(n, 8);
 }
 
 /**
@@ -505,7 +505,7 @@ static inline unsigned int calc_tx_flits(const struct rte_mbuf *m,
 
 	hdrlen = is_eth_imm(m);
 	if (hdrlen)
-		return DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
+		return RTE_DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
 
 	/*
 	 * Otherwise, we're going to have to construct a Scatter gather list
@@ -779,7 +779,7 @@ static inline void ship_tx_pkt_coalesce_wr(struct adapter *adap,
 	wr = (void *)&q->desc[q->pidx];
 	vmwr = (void *)&q->desc[q->pidx];
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(q->coalesce.flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(q->coalesce.flits, 2));
 	ndesc = flits_to_desc(q->coalesce.flits);
 	wr->equiq_to_len16 = htonl(wr_mid);
 	wr->plen = cpu_to_be16(q->coalesce.len);
@@ -866,7 +866,7 @@ static inline int should_tx_packet_coalesce(struct sge_eth_txq *txq,
 
 	/* If coalescing is on, the mbuf is added to a pkts WR */
 	if (q->coalesce.idx) {
-		ndesc = DIV_ROUND_UP(q->coalesce.flits + flits, 8);
+		ndesc = RTE_DIV_ROUND_UP(q->coalesce.flits + flits, 8);
 		credits = txq_avail(q) - ndesc;
 
 		if (unlikely(wraps_around(q, ndesc)))
@@ -950,7 +950,7 @@ static inline int tx_do_packet_coalesce(struct sge_eth_txq *txq,
 		mc->cmd_dest = htonl(V_ULPTX_CMD(4) | V_ULP_TXPKT_DEST(0) |
 				     V_ULP_TXPKT_FID(adap->sge.fw_evtq.cntxt_id) |
 				     F_ULP_TXPKT_RO);
-		mc->len = htonl(DIV_ROUND_UP(flits, 2));
+		mc->len = htonl(RTE_DIV_ROUND_UP(flits, 2));
 		sc_imm = (struct ulptx_idata *)(mc + 1);
 		sc_imm->cmd_more = htonl(V_ULPTX_CMD(ULP_TX_SC_IMM) |
 					 F_ULP_TX_SC_MORE);
@@ -1116,7 +1116,7 @@ int t4_eth_xmit(struct sge_eth_txq *txq, struct rte_mbuf *mbuf,
 		goto out_free;
 	}
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(flits, 2));
 	if (Q_IDXDIFF(&txq->q, equeidx)  >= 64) {
 		txq->q.equeidx = txq->q.pidx;
 		wr_mid |= F_FW_WR_EQUEQ;
@@ -1322,7 +1322,7 @@ static int ctrl_xmit(struct sge_ctrl_txq *q, struct rte_mbuf *mbuf)
 	}
 
 	reclaim_completed_tx_imm(&q->q);
-	ndesc = DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
+	ndesc = RTE_DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
 	t4_os_lock(&q->ctrlq_lock);
 
 	q->full = txq_avail(&q->q) < ndesc ? 1 : 0;
diff --git a/drivers/net/ena/base/ena_eth_com.h b/drivers/net/ena/base/ena_eth_com.h
index f773699467..4c82415181 100644
--- a/drivers/net/ena/base/ena_eth_com.h
+++ b/drivers/net/ena/base/ena_eth_com.h
@@ -160,8 +160,8 @@ static inline bool ena_com_is_doorbell_needed(struct ena_com_io_sq *io_sq,
 
 	if (num_descs > llq_info->descs_num_before_header) {
 		descs_after_first_entry = num_descs - llq_info->descs_num_before_header;
-		num_entries_needed += DIV_ROUND_UP(descs_after_first_entry,
-						   llq_info->descs_per_entry);
+		num_entries_needed += RTE_DIV_ROUND_UP(descs_after_first_entry,
+						       llq_info->descs_per_entry);
 	}
 
 	ena_trc_dbg(ena_com_io_sq_to_ena_dev(io_sq),
diff --git a/drivers/net/ena/base/ena_plat_dpdk.h b/drivers/net/ena/base/ena_plat_dpdk.h
index c84420de22..9c8a96c918 100644
--- a/drivers/net/ena/base/ena_plat_dpdk.h
+++ b/drivers/net/ena/base/ena_plat_dpdk.h
@@ -304,7 +304,6 @@ ena_mem_alloc_coherent(struct rte_eth_dev_data *data, size_t size,
 			rte_write64_relaxed(*from, to);			       \
 	} while(0)
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 
 #define ENA_FFS(x) ffs(x)
 
diff --git a/drivers/net/hinic/base/hinic_compat.h b/drivers/net/hinic/base/hinic_compat.h
index 707a3b92b9..f4398a7847 100644
--- a/drivers/net/hinic/base/hinic_compat.h
+++ b/drivers/net/hinic/base/hinic_compat.h
@@ -70,11 +70,8 @@ typedef uint64_t  dma_addr_t;
 #define upper_32_bits(n) ((u32)(((n) >> 16) >> 16))
 #define lower_32_bits(n) ((u32)(n))
 
-/* Returns X / Y, rounding up.  X must be nonnegative to round correctly. */
-#define DIV_ROUND_UP(X, Y) (((X) + ((Y) - 1)) / (Y))
-
 /* Returns X rounded up to the nearest multiple of Y. */
-#define ROUND_UP(X, Y) (DIV_ROUND_UP(X, Y) * (Y))
+#define ROUND_UP(X, Y) (RTE_DIV_ROUND_UP(X, Y) * (Y))
 
 #undef  ALIGN
 #define ALIGN(x, a)  RTE_ALIGN(x, a)
diff --git a/drivers/net/hns3/hns3_ethdev.h b/drivers/net/hns3/hns3_ethdev.h
index 209b042816..b4feff4b94 100644
--- a/drivers/net/hns3/hns3_ethdev.h
+++ b/drivers/net/hns3/hns3_ethdev.h
@@ -978,8 +978,6 @@ static inline struct hns3_vf *HNS3_DEV_HW_TO_VF(struct hns3_hw *hw)
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define rounddown(x, y) ((x) - ((x) % (y)))
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-
 /*
  * Because hardware always access register in little-endian mode based on hns3
  * network engine, so driver should also call rte_cpu_to_le_32 to convert data
diff --git a/drivers/net/hns3/hns3_fdir.c b/drivers/net/hns3/hns3_fdir.c
index 50572ae430..dfcc5c0d99 100644
--- a/drivers/net/hns3/hns3_fdir.c
+++ b/drivers/net/hns3/hns3_fdir.c
@@ -129,7 +129,7 @@ static const struct key_info tuple_key_info[] = {
 #define MAX_KEY_LENGTH		400
 #define MAX_200B_KEY_LENGTH	200
 #define MAX_META_DATA_LENGTH	16
-#define MAX_KEY_DWORDS	DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
+#define MAX_KEY_DWORDS	RTE_DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
 #define MAX_KEY_BYTES	(MAX_KEY_DWORDS * 4)
 
 enum HNS3_FD_PACKET_TYPE {
diff --git a/drivers/net/hns3/hns3_regs.c b/drivers/net/hns3/hns3_regs.c
index 5c74f9ae2e..066b836556 100644
--- a/drivers/net/hns3/hns3_regs.c
+++ b/drivers/net/hns3/hns3_regs.c
@@ -1034,8 +1034,8 @@ hns3_get_32_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
-			       HNS3_32_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
+				   HNS3_32_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-32bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
@@ -1093,8 +1093,8 @@ hns3_get_64_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
-			       HNS3_64_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
+				   HNS3_64_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-64bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
diff --git a/drivers/net/hns3/hns3_rss.c b/drivers/net/hns3/hns3_rss.c
index 508b3e26d0..6cb8d9482d 100644
--- a/drivers/net/hns3/hns3_rss.c
+++ b/drivers/net/hns3/hns3_rss.c
@@ -295,7 +295,7 @@ hns3_rss_set_algo_key(struct hns3_hw *hw, uint8_t hash_algo,
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
 
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  false);
@@ -335,7 +335,7 @@ hns3_rss_get_algo_key(struct hns3_hw *hw,  uint8_t *hash_algo,
 	int ret;
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  true);
@@ -381,7 +381,7 @@ hns3_set_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  false);
@@ -431,7 +431,7 @@ hns3_get_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  true);
diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h b/drivers/net/intel/i40e/base/i40e_osdep.h
index 9c5a94f055..93620f5c50 100644
--- a/drivers/net/intel/i40e/base/i40e_osdep.h
+++ b/drivers/net/intel/i40e/base/i40e_osdep.h
@@ -227,12 +227,11 @@ struct i40e_spinlock {
 #define i40e_memset(a, b, c, d) memset((a), (b), (c))
 #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
 
-#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us_sleep(x)
 #define i40e_usec_delay(x) DELAY(x)
 #define i40e_msec_delay(x) DELAY(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000*(x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #endif /* _I40E_OSDEP_H_ */
diff --git a/drivers/net/intel/ice/base/ice_osdep.h b/drivers/net/intel/ice/base/ice_osdep.h
index 5d92cf5aa4..3ef52eae6d 100644
--- a/drivers/net/intel/ice/base/ice_osdep.h
+++ b/drivers/net/intel/ice/base/ice_osdep.h
@@ -311,13 +311,12 @@ ice_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define ice_usec_delay(x, y) rte_delay_us(x)
 #define ice_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 struct ice_list_entry {
 	LIST_ENTRY(ice_list_entry) next;
diff --git a/drivers/net/intel/idpf/base/idpf_osdep.h b/drivers/net/intel/idpf/base/idpf_osdep.h
index 67bc33f2e6..64bbd00eea 100644
--- a/drivers/net/intel/idpf/base/idpf_osdep.h
+++ b/drivers/net/intel/idpf/base/idpf_osdep.h
@@ -291,13 +291,12 @@ idpf_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define idpf_usec_delay(x) rte_delay_us(x)
 #define idpf_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #ifndef IDPF_DBG_TRACE
 #define IDPF_DBG_TRACE	  BIT_ULL(0)
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk.h b/drivers/net/nfp/nfdk/nfp_nfdk.h
index 64c2b31308..b831e0d69e 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk.h
+++ b/drivers/net/nfp/nfdk/nfp_nfdk.h
@@ -168,8 +168,9 @@ static inline uint16_t
 nfp_net_nfdk_headlen_to_segs(uint16_t headlen)
 {
 	/* First descriptor fits less data, so adjust for that */
-	return DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC - NFDK_TX_MAX_DATA_PER_HEAD,
-			NFDK_TX_MAX_DATA_PER_DESC);
+	return RTE_DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC -
+				NFDK_TX_MAX_DATA_PER_HEAD,
+				NFDK_TX_MAX_DATA_PER_DESC);
 }
 
 /* Set TX CSUM offload flags in TX descriptor of nfdk */
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
index 15867ab62f..86b3c5b64a 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
+++ b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
@@ -103,7 +103,8 @@ nfp_net_nfdk_tx_maybe_close_block(struct nfp_net_txq *txq,
 	n_descs = nfp_net_nfdk_headlen_to_segs(pkt_temp->data_len);
 	while (pkt_temp->next != NULL) {
 		pkt_temp = pkt_temp->next;
-		n_descs += DIV_ROUND_UP(pkt_temp->data_len, NFDK_TX_MAX_DATA_PER_DESC);
+		n_descs += RTE_DIV_ROUND_UP(pkt_temp->data_len,
+					    NFDK_TX_MAX_DATA_PER_DESC);
 	}
 
 	if (unlikely(n_descs > NFDK_TX_DESC_GATHER_MAX))
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp.c b/drivers/net/nfp/nfpcore/nfp_nsp.c
index 43c13baabe..bd02efcc4a 100644
--- a/drivers/net/nfp/nfpcore/nfp_nsp.c
+++ b/drivers/net/nfp/nfpcore/nfp_nsp.c
@@ -542,7 +542,7 @@ nfp_nsp_command_buf_dma_sg(struct nfp_nsp *nsp,
 	struct nfp_nsp_dma_desc *descs;
 
 	chunk_size = RTE_BIT64(chunk_order);
-	nseg = DIV_ROUND_UP(max_size, chunk_size);
+	nseg = RTE_DIV_ROUND_UP(max_size, chunk_size);
 
 	/* Malloc memory */
 	total_len = sizeof(struct nfp_nsp_dma_buf) +
diff --git a/drivers/net/qede/base/bcm_osal.h b/drivers/net/qede/base/bcm_osal.h
index 357981f63d..a2f3bc838b 100644
--- a/drivers/net/qede/base/bcm_osal.h
+++ b/drivers/net/qede/base/bcm_osal.h
@@ -396,8 +396,6 @@ int qede_save_fw_dump(uint16_t port_id);
 
 /* Utility functions */
 
-#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-#define DIV_ROUND_UP(size, to_what) RTE_DIV_ROUND_UP(size, to_what)
 #define RTE_ROUNDUP(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define ROUNDUP(value, to_what) RTE_ROUNDUP((value), (to_what))
 
diff --git a/drivers/net/qede/base/ecore_chain.h b/drivers/net/qede/base/ecore_chain.h
index c69920be54..50b3196f02 100644
--- a/drivers/net/qede/base/ecore_chain.h
+++ b/drivers/net/qede/base/ecore_chain.h
@@ -157,7 +157,7 @@ struct ecore_chain {
 	UNUSABLE_ELEMS_PER_PAGE(elem_size, mode)))
 
 #define ECORE_CHAIN_PAGE_CNT(elem_cnt, elem_size, mode)		\
-	DIV_ROUND_UP(elem_cnt, USABLE_ELEMS_PER_PAGE(elem_size, mode))
+	RTE_DIV_ROUND_UP(elem_cnt, USABLE_ELEMS_PER_PAGE(elem_size, mode))
 
 #define is_chain_u16(p)	((p)->cnt_type == ECORE_CHAIN_CNT_TYPE_U16)
 #define is_chain_u32(p)	((p)->cnt_type == ECORE_CHAIN_CNT_TYPE_U32)
diff --git a/drivers/net/qede/base/ecore_cxt.c b/drivers/net/qede/base/ecore_cxt.c
index d3025724b6..f877d31ecc 100644
--- a/drivers/net/qede/base/ecore_cxt.c
+++ b/drivers/net/qede/base/ecore_cxt.c
@@ -341,7 +341,8 @@ static void ecore_ilt_cli_adv_line(struct ecore_hwfn *p_hwfn,
 		p_cli->first.val = *p_line;
 
 	p_cli->active = true;
-	*p_line += DIV_ROUND_UP(p_blk->total_size, p_blk->real_size_in_page);
+	*p_line += RTE_DIV_ROUND_UP(p_blk->total_size,
+				    p_blk->real_size_in_page);
 	p_cli->last.val = *p_line - 1;
 
 	DP_VERBOSE(p_hwfn, ECORE_MSG_ILT,
@@ -761,7 +762,7 @@ static enum _ecore_status_t ecore_cxt_src_t2_alloc(struct ecore_hwfn *p_hwfn)
 	/* use the same page size as the SRC ILT client */
 	psz = ILT_PAGE_IN_BYTES(p_src->p_size.val);
 	p_t2 = &p_mngr->src_t2;
-	p_t2->num_pages = DIV_ROUND_UP(total_size, psz);
+	p_t2->num_pages = RTE_DIV_ROUND_UP(total_size, psz);
 
 	/* allocate t2 */
 	p_t2->dma_mem = OSAL_ZALLOC(p_hwfn->p_dev, GFP_KERNEL,
@@ -877,7 +878,8 @@ ecore_ilt_blk_alloc(struct ecore_hwfn *p_hwfn,
 
 	sz_left = p_blk->total_size;
 	lines_to_skip = p_blk->dynamic_line_cnt;
-	lines = DIV_ROUND_UP(sz_left, p_blk->real_size_in_page) - lines_to_skip;
+	lines = RTE_DIV_ROUND_UP(sz_left, p_blk->real_size_in_page) -
+		lines_to_skip;
 	line = p_blk->start_line + start_line_offset -
 	       p_hwfn->p_cxt_mngr->pf_start_line;
 	first_skipped_line = line + p_blk->dynamic_line_offset;
@@ -1000,7 +1002,7 @@ __ecore_cid_map_alloc_single(struct ecore_hwfn *p_hwfn, u32 type,
 	if (!cid_count)
 		return ECORE_SUCCESS;
 
-	size = MAP_WORD_SIZE * DIV_ROUND_UP(cid_count, BITS_PER_MAP_WORD);
+	size = MAP_WORD_SIZE * RTE_DIV_ROUND_UP(cid_count, BITS_PER_MAP_WORD);
 	p_map->cid_map = OSAL_ZALLOC(p_hwfn->p_dev, GFP_KERNEL, size);
 	if (p_map->cid_map == OSAL_NULL)
 		return ECORE_NOMEM;
@@ -1217,8 +1219,8 @@ void ecore_cxt_mngr_setup(struct ecore_hwfn *p_hwfn)
 		p_cfg = &p_mngr->conn_cfg[type];
 		if (p_cfg->cid_count) {
 			p_map = &p_mngr->acquired[type];
-			len = DIV_ROUND_UP(p_map->max_count,
-					   BITS_PER_MAP_WORD) *
+			len = RTE_DIV_ROUND_UP(p_map->max_count,
+					       BITS_PER_MAP_WORD) *
 			      MAP_WORD_SIZE;
 			OSAL_MEM_ZERO(p_map->cid_map, len);
 		}
@@ -1228,8 +1230,8 @@ void ecore_cxt_mngr_setup(struct ecore_hwfn *p_hwfn)
 
 		for (vf = 0; vf < max_num_vfs; vf++) {
 			p_map = &p_mngr->acquired_vf[type][vf];
-			len = DIV_ROUND_UP(p_map->max_count,
-					   BITS_PER_MAP_WORD) *
+			len = RTE_DIV_ROUND_UP(p_map->max_count,
+					       BITS_PER_MAP_WORD) *
 			      MAP_WORD_SIZE;
 			OSAL_MEM_ZERO(p_map->cid_map, len);
 		}
@@ -2184,7 +2186,7 @@ static u16 ecore_blk_calculate_pages(struct ecore_ilt_cli_blk *p_blk)
 	if (p_blk->real_size_in_page == 0)
 		return 0;
 
-	return DIV_ROUND_UP(p_blk->total_size, p_blk->real_size_in_page);
+	return RTE_DIV_ROUND_UP(p_blk->total_size, p_blk->real_size_in_page);
 }
 
 u16 ecore_get_cdut_num_pf_init_pages(struct ecore_hwfn *p_hwfn)
diff --git a/drivers/net/qede/base/ecore_init_fw_funcs.c b/drivers/net/qede/base/ecore_init_fw_funcs.c
index 4e4d1dc374..e3a39d0958 100644
--- a/drivers/net/qede/base/ecore_init_fw_funcs.c
+++ b/drivers/net/qede/base/ecore_init_fw_funcs.c
@@ -24,9 +24,11 @@ static u16 task_region_offsets[1][NUM_OF_CONNECTION_TYPES] = {
 
 /* General constants */
 #define QM_PQ_MEM_4KB(pq_size) \
-	(pq_size ? DIV_ROUND_UP((pq_size + 1) * QM_PQ_ELEMENT_SIZE, 0x1000) : 0)
+	(pq_size \
+		? RTE_DIV_ROUND_UP((pq_size + 1) * QM_PQ_ELEMENT_SIZE, 0x1000) \
+		: 0)
 #define QM_PQ_SIZE_256B(pq_size) \
-	(pq_size ? DIV_ROUND_UP(pq_size, 0x100) - 1 : 0)
+	(pq_size ? RTE_DIV_ROUND_UP(pq_size, 0x100) - 1 : 0)
 #define QM_INVALID_PQ_ID		0xffff
 
 /* Max link speed (in Mbps) */
@@ -1313,10 +1315,10 @@ void ecore_init_brb_ram(struct ecore_hwfn *p_hwfn,
 	u32 active_port_blocks, reg_offset = 0;
 	u8 port, active_ports = 0;
 
-	tc_headroom_blocks = (u32)DIV_ROUND_UP(req->headroom_per_tc,
-					       BRB_BLOCK_SIZE);
-	min_pkt_size_blocks = (u32)DIV_ROUND_UP(req->min_pkt_size,
-						BRB_BLOCK_SIZE);
+	tc_headroom_blocks = (u32)RTE_DIV_ROUND_UP(req->headroom_per_tc,
+						   BRB_BLOCK_SIZE);
+	min_pkt_size_blocks = (u32)RTE_DIV_ROUND_UP(req->min_pkt_size,
+						    BRB_BLOCK_SIZE);
 	total_blocks = ECORE_IS_K2(p_hwfn->p_dev) ? BRB_TOTAL_RAM_BLOCKS_K2 :
 						    BRB_TOTAL_RAM_BLOCKS_BB;
 
@@ -1334,8 +1336,9 @@ void ecore_init_brb_ram(struct ecore_hwfn *p_hwfn,
 		u8 tc;
 
 		/* Calculate per-port sizes */
-		tc_guaranteed_blocks = (u32)DIV_ROUND_UP(req->guranteed_per_tc,
-							 BRB_BLOCK_SIZE);
+		tc_guaranteed_blocks =
+			(u32)RTE_DIV_ROUND_UP(req->guranteed_per_tc,
+					      BRB_BLOCK_SIZE);
 		port_blocks = req->num_active_tcs[port] ? active_port_blocks :
 							  0;
 		port_guaranteed_blocks = req->num_active_tcs[port] *
@@ -2059,7 +2062,7 @@ void ecore_enable_context_validation(struct ecore_hwfn *p_hwfn,
 	ecore_wr(p_hwfn, p_ptt, CDU_REG_TCFC_CTX_VALID0, ctx_validation);
 }
 
-#define PHYS_ADDR_DWORDS        DIV_ROUND_UP(sizeof(dma_addr_t), 4)
+#define PHYS_ADDR_DWORDS        RTE_DIV_ROUND_UP(sizeof(dma_addr_t), 4)
 #define OVERLAY_HDR_SIZE_DWORDS (sizeof(struct fw_overlay_buf_hdr) / 4)
 
 static u32 ecore_get_overlay_addr_ram_addr(struct ecore_hwfn *p_hwfn,
diff --git a/drivers/net/qede/base/ecore_mcp.c b/drivers/net/qede/base/ecore_mcp.c
index ec83b244e6..07ba41e1a8 100644
--- a/drivers/net/qede/base/ecore_mcp.c
+++ b/drivers/net/qede/base/ecore_mcp.c
@@ -713,7 +713,7 @@ ecore_mcp_cmd_and_union(struct ecore_hwfn *p_hwfn,
 	}
 #endif
 	if (ECORE_MB_FLAGS_IS_SET(p_mb_params, CAN_SLEEP)) {
-		max_retries = DIV_ROUND_UP(max_retries, 1000);
+		max_retries = RTE_DIV_ROUND_UP(max_retries, 1000);
 		usecs *= 1000;
 	}
 
@@ -4243,8 +4243,8 @@ ecore_mcp_resc_lock(struct ecore_hwfn *p_hwfn, struct ecore_ptt *p_ptt,
 		if (retry_cnt) {
 			if (p_params->sleep_b4_retry) {
 				u16 retry_interval_in_ms =
-					DIV_ROUND_UP(p_params->retry_interval,
-						     1000);
+					RTE_DIV_ROUND_UP(p_params->retry_interval,
+							 1000);
 
 				OSAL_MSLEEP(retry_interval_in_ms);
 			} else {
diff --git a/drivers/net/qede/qede_debug.c b/drivers/net/qede/qede_debug.c
index 1d3147b724..c4cf40b15d 100644
--- a/drivers/net/qede/qede_debug.c
+++ b/drivers/net/qede/qede_debug.c
@@ -332,7 +332,7 @@ struct split_type_defs {
 
 #define BYTES_IN_DWORD			sizeof(u32)
 /* In the macros below, size and offset are specified in bits */
-#define CEIL_DWORDS(size)		DIV_ROUND_UP(size, 32)
+#define CEIL_DWORDS(size)		RTE_DIV_ROUND_UP(size, 32)
 #define FIELD_BIT_OFFSET(type, field)	type ## _ ## field ## _ ## OFFSET
 #define FIELD_BIT_SIZE(type, field)	type ## _ ## field ## _ ## SIZE
 #define FIELD_DWORD_OFFSET(type, field) \
@@ -3060,7 +3060,7 @@ static u32 qed_grc_dump_big_ram(struct ecore_hwfn *p_hwfn,
 		return offset + ram_size;
 
 	/* Dump Big RAM */
-	for (i = 0; i < DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
+	for (i = 0; i < RTE_DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
 	     i++) {
 		u32 addr, len;
 
@@ -4161,8 +4161,8 @@ static enum dbg_status qed_mcp_trace_dump(struct ecore_hwfn *p_hwfn,
 
 	/* Find trace data size */
 	trace_data_size_dwords =
-	    DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
-			 BYTES_IN_DWORD);
+	    RTE_DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
+			     BYTES_IN_DWORD);
 
 	/* Dump trace data section header and param */
 	offset += qed_dump_section_hdr(dump_buf + offset,
@@ -4817,10 +4817,10 @@ static u32 qed_ilt_dump(struct ecore_hwfn *p_hwfn,
 		offset += num_pages * PAGE_MEM_DESC_SIZE_DWORDS;
 	}
 
-	valid_conn_pf_pages = DIV_ROUND_UP(valid_conn_pf_cids,
-					   num_cids_per_page);
-	valid_conn_vf_pages = DIV_ROUND_UP(valid_conn_vf_cids,
-					   num_cids_per_page);
+	valid_conn_pf_pages = RTE_DIV_ROUND_UP(valid_conn_pf_cids,
+					       num_cids_per_page);
+	valid_conn_vf_pages = RTE_DIV_ROUND_UP(valid_conn_vf_cids,
+					       num_cids_per_page);
 
 	/* Dump ILT pages IDs */
 	offset += qed_ilt_dump_pages_section(p_hwfn,
diff --git a/drivers/net/rnp/base/rnp_osdep.h b/drivers/net/rnp/base/rnp_osdep.h
index 58dcb40f90..b0e7906fab 100644
--- a/drivers/net/rnp/base/rnp_osdep.h
+++ b/drivers/net/rnp/base/rnp_osdep.h
@@ -53,10 +53,9 @@ typedef uint64_t u64;
 #define cpu_to_le32(v)	rte_cpu_to_le_32((u32)(v))
 #endif
 
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d)      (((n) + (d) - 1) / (d))
+#ifndef BITS_TO_LONGS
 #define BITS_PER_BYTE           (8)
-#define BITS_TO_LONGS(nr)       DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
+#define BITS_TO_LONGS(nr)       RTE_DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
 #endif
 
 #define fls(n)	rte_fls_u32(n)
diff --git a/drivers/net/rnp/rnp_rxtx.c b/drivers/net/rnp/rnp_rxtx.c
index 81e8c6ba44..143cd11ffe 100644
--- a/drivers/net/rnp/rnp_rxtx.c
+++ b/drivers/net/rnp/rnp_rxtx.c
@@ -1165,7 +1165,7 @@ rnp_calc_pkt_desc(struct rte_mbuf *tx_pkt)
 	uint16_t count = 0;
 
 	while (txd != NULL) {
-		count += DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
+		count += RTE_DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
 		txd = txd->next;
 	}
 
diff --git a/drivers/raw/ifpga/base/opae_osdep.h b/drivers/raw/ifpga/base/opae_osdep.h
index e35a21c80e..b0045a36a0 100644
--- a/drivers/raw/ifpga/base/opae_osdep.h
+++ b/drivers/raw/ifpga/base/opae_osdep.h
@@ -71,10 +71,9 @@ struct uuid {
 } while (0)
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define udelay(x) opae_udelay(x)
 #define msleep(x) opae_udelay(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #define time_after(a, b)	((long)((b) - (a)) < 0)
 #define time_before(a, b)	time_after(b, a)
diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
index 79d2a0ab93..168b0fe592 100644
--- a/lib/eal/include/rte_common.h
+++ b/lib/eal/include/rte_common.h
@@ -717,6 +717,11 @@ static void __attribute__((destructor(RTE_PRIO(prio)), used)) func(void)
 		(ceil - (v)) > ((v) - floor) ? floor : ceil;	\
 	})
 
+/**
+ * Macro to divide dividend n by divisor d, rounding up if d does not divide n.
+ */
+#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+
 /**
  * Checks if a pointer is aligned to a given power-of-two value
  *
-- 
2.55.0.691.gc56d675ccc-goog

Re: [PATCH] eal: introduce division round up macro

[Stephen Hemminger]

On Wed, 12 Aug 2026 15:30:17 -0700
Joshua Washington <joshwash@google.com> wrote:

> A number of drivers define the DIV_ROUND_UP macro -- add it to
> RTE_COMMON so that other code looking to use that functionality doesn't
> have to re-define it.
> 
> Replace usage of DIV_ROUND_UP in existing places with RTE_DIV_ROUND_UP.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
> ---

Makes sense, but this is the kind of change that is best done as a patch series
so that each driver vendor looks at their code; and if something is broken
only one change needs to be reverted. Make blast radius smaller

[PATCH v2 00/14] Introduce RTE_DIV_ROUND_UP

[Joshua Washington]

Numerous drivers independently define the DIV_ROUND_UP macro, using the
same definition. Attempt to commonize by introducing RTE_DIV_ROUND_UP to
rte_common.h and replacing usages of DIV_ROUND_UP with RTE_DIV_ROUND_UP.

Changes in v2:
  - Split single patch into series.

Joshua Washington (14):
  eal: introduce division round up macro
  net/i40e: use common division round up macro
  net/idpf: use common division round up macro
  net/ice: use common division round up macro
  net/qede: use common division round up macro
  net/rnp: use common division round up macro
  net/hns3: use common division round up macro
  net/hinic: use common division round up macro
  net/bnx2x: use common division round up macro
  net/cxgbe: use common division round up macro
  net/ena: use common division round up macro
  raw/ifpga/base: use common division round up macro
  drivers: use common division round up macro
  app/procinfo: use common division round up macro

 app/proc-info/main.c                        |  3 +--
 drivers/common/nfp/nfp_platform.h           |  2 --
 drivers/net/bnx2x/bnx2x.c                   |  4 ++--
 drivers/net/bnx2x/bnx2x.h                   | 11 ++++-------
 drivers/net/cxgbe/base/t4_hw.c              |  4 ++--
 drivers/net/cxgbe/base/t4vf_hw.c            | 10 +++++-----
 drivers/net/cxgbe/cxgbe_compat.h            |  3 +--
 drivers/net/cxgbe/cxgbe_filter.c            |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h              |  4 ++--
 drivers/net/cxgbe/sge.c                     | 14 +++++++-------
 drivers/net/ena/base/ena_eth_com.h          |  4 ++--
 drivers/net/ena/base/ena_plat_dpdk.h        |  1 -
 drivers/net/hinic/base/hinic_compat.h       |  5 +----
 drivers/net/hns3/hns3_ethdev.h              |  2 --
 drivers/net/hns3/hns3_fdir.c                |  2 +-
 drivers/net/hns3/hns3_regs.c                |  8 ++++----
 drivers/net/hns3/hns3_rss.c                 |  8 ++++----
 drivers/net/intel/i40e/base/i40e_osdep.h    |  3 +--
 drivers/net/intel/ice/base/ice_osdep.h      |  3 +--
 drivers/net/intel/idpf/base/idpf_osdep.h    |  3 +--
 drivers/net/nfp/nfdk/nfp_nfdk.h             |  5 +++--
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c          |  3 ++-
 drivers/net/nfp/nfpcore/nfp_nsp.c           |  2 +-
 drivers/net/qede/base/bcm_osal.h            |  2 --
 drivers/net/qede/base/ecore_chain.h         |  2 +-
 drivers/net/qede/base/ecore_cxt.c           | 20 +++++++++++---------
 drivers/net/qede/base/ecore_init_fw_funcs.c | 21 ++++++++++++---------
 drivers/net/qede/base/ecore_mcp.c           |  6 +++---
 drivers/net/qede/qede_debug.c               | 16 ++++++++--------
 drivers/net/rnp/base/rnp_osdep.h            |  5 ++---
 drivers/net/rnp/rnp_rxtx.c                  |  2 +-
 drivers/raw/ifpga/base/opae_osdep.h         |  3 +--
 lib/eal/include/rte_common.h                |  5 +++++
 33 files changed, 92 insertions(+), 100 deletions(-)

-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 01/14] eal: introduce division round up macro

[Joshua Washington]

A number of drivers define the DIV_ROUND_UP macro -- add it to
rte_common.h so that other code looking to use that functionality
doesn't have to re-define it.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v2:
  - Split out patches replacing usages
---
 lib/eal/include/rte_common.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
index 79d2a0ab93..168b0fe592 100644
--- a/lib/eal/include/rte_common.h
+++ b/lib/eal/include/rte_common.h
@@ -717,6 +717,11 @@ static void __attribute__((destructor(RTE_PRIO(prio)), used)) func(void)
 		(ceil - (v)) > ((v) - floor) ? floor : ceil;	\
 	})
 
+/**
+ * Macro to divide dividend n by divisor d, rounding up if d does not divide n.
+ */
+#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+
 /**
  * Checks if a pointer is aligned to a given power-of-two value
  *
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 02/14] net/i40e: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/intel/i40e/base/i40e_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h b/drivers/net/intel/i40e/base/i40e_osdep.h
index 9c5a94f055..93620f5c50 100644
--- a/drivers/net/intel/i40e/base/i40e_osdep.h
+++ b/drivers/net/intel/i40e/base/i40e_osdep.h
@@ -227,12 +227,11 @@ struct i40e_spinlock {
 #define i40e_memset(a, b, c, d) memset((a), (b), (c))
 #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
 
-#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us_sleep(x)
 #define i40e_usec_delay(x) DELAY(x)
 #define i40e_msec_delay(x) DELAY(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000*(x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #endif /* _I40E_OSDEP_H_ */
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 03/14] net/idpf: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/intel/idpf/base/idpf_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/idpf/base/idpf_osdep.h b/drivers/net/intel/idpf/base/idpf_osdep.h
index 67bc33f2e6..64bbd00eea 100644
--- a/drivers/net/intel/idpf/base/idpf_osdep.h
+++ b/drivers/net/intel/idpf/base/idpf_osdep.h
@@ -291,13 +291,12 @@ idpf_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define idpf_usec_delay(x) rte_delay_us(x)
 #define idpf_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #ifndef IDPF_DBG_TRACE
 #define IDPF_DBG_TRACE	  BIT_ULL(0)
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 04/14] net/ice: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/intel/ice/base/ice_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/ice/base/ice_osdep.h b/drivers/net/intel/ice/base/ice_osdep.h
index 5d92cf5aa4..3ef52eae6d 100644
--- a/drivers/net/intel/ice/base/ice_osdep.h
+++ b/drivers/net/intel/ice/base/ice_osdep.h
@@ -311,13 +311,12 @@ ice_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define ice_usec_delay(x, y) rte_delay_us(x)
 #define ice_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 struct ice_list_entry {
 	LIST_ENTRY(ice_list_entry) next;
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 05/14] net/qede: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP and
RTE_DIV_ROUND_UP definitions redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/qede/base/bcm_osal.h            |  2 --
 drivers/net/qede/base/ecore_chain.h         |  2 +-
 drivers/net/qede/base/ecore_cxt.c           | 20 +++++++++++---------
 drivers/net/qede/base/ecore_init_fw_funcs.c | 21 ++++++++++++---------
 drivers/net/qede/base/ecore_mcp.c           |  6 +++---
 drivers/net/qede/qede_debug.c               | 16 ++++++++--------
 6 files changed, 35 insertions(+), 32 deletions(-)

diff --git a/drivers/net/qede/base/bcm_osal.h b/drivers/net/qede/base/bcm_osal.h
index 357981f63d..a2f3bc838b 100644
--- a/drivers/net/qede/base/bcm_osal.h
+++ b/drivers/net/qede/base/bcm_osal.h
@@ -396,8 +396,6 @@ int qede_save_fw_dump(uint16_t port_id);
 
 /* Utility functions */
 
-#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-#define DIV_ROUND_UP(size, to_what) RTE_DIV_ROUND_UP(size, to_what)
 #define RTE_ROUNDUP(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define ROUNDUP(value, to_what) RTE_ROUNDUP((value), (to_what))
 
diff --git a/drivers/net/qede/base/ecore_chain.h b/drivers/net/qede/base/ecore_chain.h
index c69920be54..50b3196f02 100644
--- a/drivers/net/qede/base/ecore_chain.h
+++ b/drivers/net/qede/base/ecore_chain.h
@@ -157,7 +157,7 @@ struct ecore_chain {
 	UNUSABLE_ELEMS_PER_PAGE(elem_size, mode)))
 
 #define ECORE_CHAIN_PAGE_CNT(elem_cnt, elem_size, mode)		\
-	DIV_ROUND_UP(elem_cnt, USABLE_ELEMS_PER_PAGE(elem_size, mode))
+	RTE_DIV_ROUND_UP(elem_cnt, USABLE_ELEMS_PER_PAGE(elem_size, mode))
 
 #define is_chain_u16(p)	((p)->cnt_type == ECORE_CHAIN_CNT_TYPE_U16)
 #define is_chain_u32(p)	((p)->cnt_type == ECORE_CHAIN_CNT_TYPE_U32)
diff --git a/drivers/net/qede/base/ecore_cxt.c b/drivers/net/qede/base/ecore_cxt.c
index d3025724b6..f877d31ecc 100644
--- a/drivers/net/qede/base/ecore_cxt.c
+++ b/drivers/net/qede/base/ecore_cxt.c
@@ -341,7 +341,8 @@ static void ecore_ilt_cli_adv_line(struct ecore_hwfn *p_hwfn,
 		p_cli->first.val = *p_line;
 
 	p_cli->active = true;
-	*p_line += DIV_ROUND_UP(p_blk->total_size, p_blk->real_size_in_page);
+	*p_line += RTE_DIV_ROUND_UP(p_blk->total_size,
+				    p_blk->real_size_in_page);
 	p_cli->last.val = *p_line - 1;
 
 	DP_VERBOSE(p_hwfn, ECORE_MSG_ILT,
@@ -761,7 +762,7 @@ static enum _ecore_status_t ecore_cxt_src_t2_alloc(struct ecore_hwfn *p_hwfn)
 	/* use the same page size as the SRC ILT client */
 	psz = ILT_PAGE_IN_BYTES(p_src->p_size.val);
 	p_t2 = &p_mngr->src_t2;
-	p_t2->num_pages = DIV_ROUND_UP(total_size, psz);
+	p_t2->num_pages = RTE_DIV_ROUND_UP(total_size, psz);
 
 	/* allocate t2 */
 	p_t2->dma_mem = OSAL_ZALLOC(p_hwfn->p_dev, GFP_KERNEL,
@@ -877,7 +878,8 @@ ecore_ilt_blk_alloc(struct ecore_hwfn *p_hwfn,
 
 	sz_left = p_blk->total_size;
 	lines_to_skip = p_blk->dynamic_line_cnt;
-	lines = DIV_ROUND_UP(sz_left, p_blk->real_size_in_page) - lines_to_skip;
+	lines = RTE_DIV_ROUND_UP(sz_left, p_blk->real_size_in_page) -
+		lines_to_skip;
 	line = p_blk->start_line + start_line_offset -
 	       p_hwfn->p_cxt_mngr->pf_start_line;
 	first_skipped_line = line + p_blk->dynamic_line_offset;
@@ -1000,7 +1002,7 @@ __ecore_cid_map_alloc_single(struct ecore_hwfn *p_hwfn, u32 type,
 	if (!cid_count)
 		return ECORE_SUCCESS;
 
-	size = MAP_WORD_SIZE * DIV_ROUND_UP(cid_count, BITS_PER_MAP_WORD);
+	size = MAP_WORD_SIZE * RTE_DIV_ROUND_UP(cid_count, BITS_PER_MAP_WORD);
 	p_map->cid_map = OSAL_ZALLOC(p_hwfn->p_dev, GFP_KERNEL, size);
 	if (p_map->cid_map == OSAL_NULL)
 		return ECORE_NOMEM;
@@ -1217,8 +1219,8 @@ void ecore_cxt_mngr_setup(struct ecore_hwfn *p_hwfn)
 		p_cfg = &p_mngr->conn_cfg[type];
 		if (p_cfg->cid_count) {
 			p_map = &p_mngr->acquired[type];
-			len = DIV_ROUND_UP(p_map->max_count,
-					   BITS_PER_MAP_WORD) *
+			len = RTE_DIV_ROUND_UP(p_map->max_count,
+					       BITS_PER_MAP_WORD) *
 			      MAP_WORD_SIZE;
 			OSAL_MEM_ZERO(p_map->cid_map, len);
 		}
@@ -1228,8 +1230,8 @@ void ecore_cxt_mngr_setup(struct ecore_hwfn *p_hwfn)
 
 		for (vf = 0; vf < max_num_vfs; vf++) {
 			p_map = &p_mngr->acquired_vf[type][vf];
-			len = DIV_ROUND_UP(p_map->max_count,
-					   BITS_PER_MAP_WORD) *
+			len = RTE_DIV_ROUND_UP(p_map->max_count,
+					       BITS_PER_MAP_WORD) *
 			      MAP_WORD_SIZE;
 			OSAL_MEM_ZERO(p_map->cid_map, len);
 		}
@@ -2184,7 +2186,7 @@ static u16 ecore_blk_calculate_pages(struct ecore_ilt_cli_blk *p_blk)
 	if (p_blk->real_size_in_page == 0)
 		return 0;
 
-	return DIV_ROUND_UP(p_blk->total_size, p_blk->real_size_in_page);
+	return RTE_DIV_ROUND_UP(p_blk->total_size, p_blk->real_size_in_page);
 }
 
 u16 ecore_get_cdut_num_pf_init_pages(struct ecore_hwfn *p_hwfn)
diff --git a/drivers/net/qede/base/ecore_init_fw_funcs.c b/drivers/net/qede/base/ecore_init_fw_funcs.c
index 4e4d1dc374..e3a39d0958 100644
--- a/drivers/net/qede/base/ecore_init_fw_funcs.c
+++ b/drivers/net/qede/base/ecore_init_fw_funcs.c
@@ -24,9 +24,11 @@ static u16 task_region_offsets[1][NUM_OF_CONNECTION_TYPES] = {
 
 /* General constants */
 #define QM_PQ_MEM_4KB(pq_size) \
-	(pq_size ? DIV_ROUND_UP((pq_size + 1) * QM_PQ_ELEMENT_SIZE, 0x1000) : 0)
+	(pq_size \
+		? RTE_DIV_ROUND_UP((pq_size + 1) * QM_PQ_ELEMENT_SIZE, 0x1000) \
+		: 0)
 #define QM_PQ_SIZE_256B(pq_size) \
-	(pq_size ? DIV_ROUND_UP(pq_size, 0x100) - 1 : 0)
+	(pq_size ? RTE_DIV_ROUND_UP(pq_size, 0x100) - 1 : 0)
 #define QM_INVALID_PQ_ID		0xffff
 
 /* Max link speed (in Mbps) */
@@ -1313,10 +1315,10 @@ void ecore_init_brb_ram(struct ecore_hwfn *p_hwfn,
 	u32 active_port_blocks, reg_offset = 0;
 	u8 port, active_ports = 0;
 
-	tc_headroom_blocks = (u32)DIV_ROUND_UP(req->headroom_per_tc,
-					       BRB_BLOCK_SIZE);
-	min_pkt_size_blocks = (u32)DIV_ROUND_UP(req->min_pkt_size,
-						BRB_BLOCK_SIZE);
+	tc_headroom_blocks = (u32)RTE_DIV_ROUND_UP(req->headroom_per_tc,
+						   BRB_BLOCK_SIZE);
+	min_pkt_size_blocks = (u32)RTE_DIV_ROUND_UP(req->min_pkt_size,
+						    BRB_BLOCK_SIZE);
 	total_blocks = ECORE_IS_K2(p_hwfn->p_dev) ? BRB_TOTAL_RAM_BLOCKS_K2 :
 						    BRB_TOTAL_RAM_BLOCKS_BB;
 
@@ -1334,8 +1336,9 @@ void ecore_init_brb_ram(struct ecore_hwfn *p_hwfn,
 		u8 tc;
 
 		/* Calculate per-port sizes */
-		tc_guaranteed_blocks = (u32)DIV_ROUND_UP(req->guranteed_per_tc,
-							 BRB_BLOCK_SIZE);
+		tc_guaranteed_blocks =
+			(u32)RTE_DIV_ROUND_UP(req->guranteed_per_tc,
+					      BRB_BLOCK_SIZE);
 		port_blocks = req->num_active_tcs[port] ? active_port_blocks :
 							  0;
 		port_guaranteed_blocks = req->num_active_tcs[port] *
@@ -2059,7 +2062,7 @@ void ecore_enable_context_validation(struct ecore_hwfn *p_hwfn,
 	ecore_wr(p_hwfn, p_ptt, CDU_REG_TCFC_CTX_VALID0, ctx_validation);
 }
 
-#define PHYS_ADDR_DWORDS        DIV_ROUND_UP(sizeof(dma_addr_t), 4)
+#define PHYS_ADDR_DWORDS        RTE_DIV_ROUND_UP(sizeof(dma_addr_t), 4)
 #define OVERLAY_HDR_SIZE_DWORDS (sizeof(struct fw_overlay_buf_hdr) / 4)
 
 static u32 ecore_get_overlay_addr_ram_addr(struct ecore_hwfn *p_hwfn,
diff --git a/drivers/net/qede/base/ecore_mcp.c b/drivers/net/qede/base/ecore_mcp.c
index ec83b244e6..07ba41e1a8 100644
--- a/drivers/net/qede/base/ecore_mcp.c
+++ b/drivers/net/qede/base/ecore_mcp.c
@@ -713,7 +713,7 @@ ecore_mcp_cmd_and_union(struct ecore_hwfn *p_hwfn,
 	}
 #endif
 	if (ECORE_MB_FLAGS_IS_SET(p_mb_params, CAN_SLEEP)) {
-		max_retries = DIV_ROUND_UP(max_retries, 1000);
+		max_retries = RTE_DIV_ROUND_UP(max_retries, 1000);
 		usecs *= 1000;
 	}
 
@@ -4243,8 +4243,8 @@ ecore_mcp_resc_lock(struct ecore_hwfn *p_hwfn, struct ecore_ptt *p_ptt,
 		if (retry_cnt) {
 			if (p_params->sleep_b4_retry) {
 				u16 retry_interval_in_ms =
-					DIV_ROUND_UP(p_params->retry_interval,
-						     1000);
+					RTE_DIV_ROUND_UP(p_params->retry_interval,
+							 1000);
 
 				OSAL_MSLEEP(retry_interval_in_ms);
 			} else {
diff --git a/drivers/net/qede/qede_debug.c b/drivers/net/qede/qede_debug.c
index 1d3147b724..c4cf40b15d 100644
--- a/drivers/net/qede/qede_debug.c
+++ b/drivers/net/qede/qede_debug.c
@@ -332,7 +332,7 @@ struct split_type_defs {
 
 #define BYTES_IN_DWORD			sizeof(u32)
 /* In the macros below, size and offset are specified in bits */
-#define CEIL_DWORDS(size)		DIV_ROUND_UP(size, 32)
+#define CEIL_DWORDS(size)		RTE_DIV_ROUND_UP(size, 32)
 #define FIELD_BIT_OFFSET(type, field)	type ## _ ## field ## _ ## OFFSET
 #define FIELD_BIT_SIZE(type, field)	type ## _ ## field ## _ ## SIZE
 #define FIELD_DWORD_OFFSET(type, field) \
@@ -3060,7 +3060,7 @@ static u32 qed_grc_dump_big_ram(struct ecore_hwfn *p_hwfn,
 		return offset + ram_size;
 
 	/* Dump Big RAM */
-	for (i = 0; i < DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
+	for (i = 0; i < RTE_DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
 	     i++) {
 		u32 addr, len;
 
@@ -4161,8 +4161,8 @@ static enum dbg_status qed_mcp_trace_dump(struct ecore_hwfn *p_hwfn,
 
 	/* Find trace data size */
 	trace_data_size_dwords =
-	    DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
-			 BYTES_IN_DWORD);
+	    RTE_DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
+			     BYTES_IN_DWORD);
 
 	/* Dump trace data section header and param */
 	offset += qed_dump_section_hdr(dump_buf + offset,
@@ -4817,10 +4817,10 @@ static u32 qed_ilt_dump(struct ecore_hwfn *p_hwfn,
 		offset += num_pages * PAGE_MEM_DESC_SIZE_DWORDS;
 	}
 
-	valid_conn_pf_pages = DIV_ROUND_UP(valid_conn_pf_cids,
-					   num_cids_per_page);
-	valid_conn_vf_pages = DIV_ROUND_UP(valid_conn_vf_cids,
-					   num_cids_per_page);
+	valid_conn_pf_pages = RTE_DIV_ROUND_UP(valid_conn_pf_cids,
+					       num_cids_per_page);
+	valid_conn_vf_pages = RTE_DIV_ROUND_UP(valid_conn_vf_cids,
+					       num_cids_per_page);
 
 	/* Dump ILT pages IDs */
 	offset += qed_ilt_dump_pages_section(p_hwfn,
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 06/14] net/rnp: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Base BITS_PER_BYTE and BITS_TO_LONGS definitions on whether
BITS_TO_LONGS is defined instead.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/rnp/base/rnp_osdep.h | 5 ++---
 drivers/net/rnp/rnp_rxtx.c       | 2 +-
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/net/rnp/base/rnp_osdep.h b/drivers/net/rnp/base/rnp_osdep.h
index 58dcb40f90..b0e7906fab 100644
--- a/drivers/net/rnp/base/rnp_osdep.h
+++ b/drivers/net/rnp/base/rnp_osdep.h
@@ -53,10 +53,9 @@ typedef uint64_t u64;
 #define cpu_to_le32(v)	rte_cpu_to_le_32((u32)(v))
 #endif
 
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d)      (((n) + (d) - 1) / (d))
+#ifndef BITS_TO_LONGS
 #define BITS_PER_BYTE           (8)
-#define BITS_TO_LONGS(nr)       DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
+#define BITS_TO_LONGS(nr)       RTE_DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
 #endif
 
 #define fls(n)	rte_fls_u32(n)
diff --git a/drivers/net/rnp/rnp_rxtx.c b/drivers/net/rnp/rnp_rxtx.c
index 81e8c6ba44..143cd11ffe 100644
--- a/drivers/net/rnp/rnp_rxtx.c
+++ b/drivers/net/rnp/rnp_rxtx.c
@@ -1165,7 +1165,7 @@ rnp_calc_pkt_desc(struct rte_mbuf *tx_pkt)
 	uint16_t count = 0;
 
 	while (txd != NULL) {
-		count += DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
+		count += RTE_DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
 		txd = txd->next;
 	}
 
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 07/14] net/hns3: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/hns3/hns3_ethdev.h | 2 --
 drivers/net/hns3/hns3_fdir.c   | 2 +-
 drivers/net/hns3/hns3_regs.c   | 8 ++++----
 drivers/net/hns3/hns3_rss.c    | 8 ++++----
 4 files changed, 9 insertions(+), 11 deletions(-)

diff --git a/drivers/net/hns3/hns3_ethdev.h b/drivers/net/hns3/hns3_ethdev.h
index 209b042816..b4feff4b94 100644
--- a/drivers/net/hns3/hns3_ethdev.h
+++ b/drivers/net/hns3/hns3_ethdev.h
@@ -978,8 +978,6 @@ static inline struct hns3_vf *HNS3_DEV_HW_TO_VF(struct hns3_hw *hw)
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define rounddown(x, y) ((x) - ((x) % (y)))
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-
 /*
  * Because hardware always access register in little-endian mode based on hns3
  * network engine, so driver should also call rte_cpu_to_le_32 to convert data
diff --git a/drivers/net/hns3/hns3_fdir.c b/drivers/net/hns3/hns3_fdir.c
index 50572ae430..dfcc5c0d99 100644
--- a/drivers/net/hns3/hns3_fdir.c
+++ b/drivers/net/hns3/hns3_fdir.c
@@ -129,7 +129,7 @@ static const struct key_info tuple_key_info[] = {
 #define MAX_KEY_LENGTH		400
 #define MAX_200B_KEY_LENGTH	200
 #define MAX_META_DATA_LENGTH	16
-#define MAX_KEY_DWORDS	DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
+#define MAX_KEY_DWORDS	RTE_DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
 #define MAX_KEY_BYTES	(MAX_KEY_DWORDS * 4)
 
 enum HNS3_FD_PACKET_TYPE {
diff --git a/drivers/net/hns3/hns3_regs.c b/drivers/net/hns3/hns3_regs.c
index 5c74f9ae2e..066b836556 100644
--- a/drivers/net/hns3/hns3_regs.c
+++ b/drivers/net/hns3/hns3_regs.c
@@ -1034,8 +1034,8 @@ hns3_get_32_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
-			       HNS3_32_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
+				   HNS3_32_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-32bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
@@ -1093,8 +1093,8 @@ hns3_get_64_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
-			       HNS3_64_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
+				   HNS3_64_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-64bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
diff --git a/drivers/net/hns3/hns3_rss.c b/drivers/net/hns3/hns3_rss.c
index 508b3e26d0..6cb8d9482d 100644
--- a/drivers/net/hns3/hns3_rss.c
+++ b/drivers/net/hns3/hns3_rss.c
@@ -295,7 +295,7 @@ hns3_rss_set_algo_key(struct hns3_hw *hw, uint8_t hash_algo,
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
 
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  false);
@@ -335,7 +335,7 @@ hns3_rss_get_algo_key(struct hns3_hw *hw,  uint8_t *hash_algo,
 	int ret;
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  true);
@@ -381,7 +381,7 @@ hns3_set_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  false);
@@ -431,7 +431,7 @@ hns3_get_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  true);
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 08/14] net/hinic: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/hinic/base/hinic_compat.h | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/net/hinic/base/hinic_compat.h b/drivers/net/hinic/base/hinic_compat.h
index 707a3b92b9..f4398a7847 100644
--- a/drivers/net/hinic/base/hinic_compat.h
+++ b/drivers/net/hinic/base/hinic_compat.h
@@ -70,11 +70,8 @@ typedef uint64_t  dma_addr_t;
 #define upper_32_bits(n) ((u32)(((n) >> 16) >> 16))
 #define lower_32_bits(n) ((u32)(n))
 
-/* Returns X / Y, rounding up.  X must be nonnegative to round correctly. */
-#define DIV_ROUND_UP(X, Y) (((X) + ((Y) - 1)) / (Y))
-
 /* Returns X rounded up to the nearest multiple of Y. */
-#define ROUND_UP(X, Y) (DIV_ROUND_UP(X, Y) * (Y))
+#define ROUND_UP(X, Y) (RTE_DIV_ROUND_UP(X, Y) * (Y))
 
 #undef  ALIGN
 #define ALIGN(x, a)  RTE_ALIGN(x, a)
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 09/14] net/bnx2x: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/bnx2x/bnx2x.c |  4 ++--
 drivers/net/bnx2x/bnx2x.h | 11 ++++-------
 2 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/drivers/net/bnx2x/bnx2x.c b/drivers/net/bnx2x/bnx2x.c
index 8790c858d5..ce9dbdf7c6 100644
--- a/drivers/net/bnx2x/bnx2x.c
+++ b/drivers/net/bnx2x/bnx2x.c
@@ -2334,8 +2334,8 @@ static void bnx2x_ilt_set_info(struct bnx2x_softc *sc)
 		ilt_client->start = line;
 
 		/* 4 bytes for each cid */
-		line += DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
-				     QM_ILT_PAGE_SZ);
+		line += RTE_DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
+					 QM_ILT_PAGE_SZ);
 
 		ilt_client->end = (line - 1);
 	}
diff --git a/drivers/net/bnx2x/bnx2x.h b/drivers/net/bnx2x/bnx2x.h
index 35206b4758..0cdd118ad5 100644
--- a/drivers/net/bnx2x/bnx2x.h
+++ b/drivers/net/bnx2x/bnx2x.h
@@ -37,9 +37,6 @@
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(arr) RTE_DIM(arr)
 #endif
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-#endif
 #ifndef roundup
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #endif
@@ -393,7 +390,7 @@ union cdu_context {
 #define CNIC_ISCSI_CID_MAX 256
 #define CNIC_FCOE_CID_MAX  2048
 #define CNIC_CID_MAX       (CNIC_ISCSI_CID_MAX + CNIC_FCOE_CID_MAX)
-#define CNIC_ILT_LINES     DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
+#define CNIC_ILT_LINES     RTE_DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
 
 #define QM_ILT_PAGE_SZ_HW  0
 #define QM_ILT_PAGE_SZ     (4096 << QM_ILT_PAGE_SZ_HW) /* 4K */
@@ -405,7 +402,7 @@ union cdu_context {
 /*#define TM_CONN_NUM        (CNIC_STARTING_CID+CNIC_ISCSI_CXT_MAX) */
 #define TM_CONN_NUM        1024
 #define TM_ILT_SZ          (8 * TM_CONN_NUM)
-#define TM_ILT_LINES       DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
+#define TM_ILT_LINES       RTE_DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
 
 /* SRC (Searcher) host DB constants */
 #define SRC_ILT_PAGE_SZ_HW 0
@@ -414,7 +411,7 @@ union cdu_context {
 #define SRC_CONN_NUM       (1 << SRC_HASH_BITS) /* 1024 */
 #define SRC_ILT_SZ         (sizeof(struct src_ent) * SRC_CONN_NUM)
 #define SRC_T2_SZ          SRC_ILT_SZ
-#define SRC_ILT_LINES      DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
+#define SRC_ILT_LINES      RTE_DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
 
 struct hw_context {
     struct bnx2x_dma    vcxt_dma;
@@ -1258,7 +1255,7 @@ struct bnx2x_softc {
 #define BNX2X_L2_CID_COUNT(sc)                                             \
 	(BNX2X_NUM_ETH_QUEUES(sc) * ECORE_MULTI_TX_COS + 2 * CNIC_SUPPORT(sc))
 #define L2_ILT_LINES(sc)                                \
-	(DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
+	(RTE_DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
 
 	int qm_cid_count;
 
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 10/14] net/cxgbe: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/cxgbe/base/t4_hw.c   |  4 ++--
 drivers/net/cxgbe/base/t4vf_hw.c | 10 +++++-----
 drivers/net/cxgbe/cxgbe_compat.h |  3 +--
 drivers/net/cxgbe/cxgbe_filter.c |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h   |  4 ++--
 drivers/net/cxgbe/sge.c          | 14 +++++++-------
 6 files changed, 20 insertions(+), 21 deletions(-)

diff --git a/drivers/net/cxgbe/base/t4_hw.c b/drivers/net/cxgbe/base/t4_hw.c
index d02da39b84..944e173da2 100644
--- a/drivers/net/cxgbe/base/t4_hw.c
+++ b/drivers/net/cxgbe/base/t4_hw.c
@@ -2630,8 +2630,8 @@ static int t4_get_exprom_version(struct adapter *adapter, u32 *vers)
 		unsigned char hdr_arr[16];      /* must start with 0x55aa */
 		unsigned char hdr_ver[4];       /* Expansion ROM version */
 	} *hdr;
-	u32 exprom_header_buf[DIV_ROUND_UP(sizeof(struct exprom_header),
-					   sizeof(u32))];
+	u32 exprom_header_buf[RTE_DIV_ROUND_UP(sizeof(struct exprom_header),
+					       sizeof(u32))];
 	int ret;
 
 	ret = t4_read_flash(adapter, FLASH_EXP_ROM_START,
diff --git a/drivers/net/cxgbe/base/t4vf_hw.c b/drivers/net/cxgbe/base/t4vf_hw.c
index 573ccb1b05..561e4a91a4 100644
--- a/drivers/net/cxgbe/base/t4vf_hw.c
+++ b/drivers/net/cxgbe/base/t4vf_hw.c
@@ -362,8 +362,8 @@ int t4vf_query_params(struct adapter *adapter, unsigned int nparams,
 	cmd.op_to_vfn = cpu_to_be32(V_FW_CMD_OP(FW_PARAMS_CMD) |
 				    F_FW_CMD_REQUEST |
 				    F_FW_CMD_READ);
-	len16 = DIV_ROUND_UP(offsetof(struct fw_params_cmd,
-			     param[nparams]), 16);
+	len16 = RTE_DIV_ROUND_UP(offsetof(struct fw_params_cmd,
+					  param[nparams]), 16);
 	cmd.retval_len16 = cpu_to_be32(V_FW_CMD_LEN16(len16));
 	for (i = 0, p = &cmd.param[0]; i < nparams; i++, p++)
 		p->mnem = cpu_to_be32(*params++);
@@ -458,8 +458,8 @@ int t4vf_set_params(struct adapter *adapter, unsigned int nparams,
 	cmd.op_to_vfn = cpu_to_be32(V_FW_CMD_OP(FW_PARAMS_CMD) |
 				    F_FW_CMD_REQUEST |
 				    F_FW_CMD_WRITE);
-	len16 = DIV_ROUND_UP(offsetof(struct fw_params_cmd,
-			     param[nparams]), 16);
+	len16 = RTE_DIV_ROUND_UP(offsetof(struct fw_params_cmd,
+					  param[nparams]), 16);
 	cmd.retval_len16 = cpu_to_be32(V_FW_CMD_LEN16(len16));
 	for (i = 0, p = &cmd.param[0]; i < nparams; i++, p++) {
 		p->mnem = cpu_to_be32(*params++);
@@ -626,7 +626,7 @@ static int t4vf_get_port_stats_fw(struct adapter *adapter, int pidx,
 		struct fw_vi_stats_cmd cmd, rpl;
 		size_t len = (offsetof(struct fw_vi_stats_cmd, u) +
 			      sizeof(struct fw_vi_stats_ctl));
-		size_t len16 = DIV_ROUND_UP(len, 16);
+		size_t len16 = RTE_DIV_ROUND_UP(len, 16);
 		int ret;
 
 		memset(&cmd, 0, sizeof(cmd));
diff --git a/drivers/net/cxgbe/cxgbe_compat.h b/drivers/net/cxgbe/cxgbe_compat.h
index e74d476f6d..14ebe914eb 100644
--- a/drivers/net/cxgbe/cxgbe_compat.h
+++ b/drivers/net/cxgbe/cxgbe_compat.h
@@ -166,11 +166,10 @@ typedef uint64_t  dma_addr_t;
 typedef char *caddr_t;
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 static inline uint8_t hweight32(uint32_t word32)
 {
diff --git a/drivers/net/cxgbe/cxgbe_filter.c b/drivers/net/cxgbe/cxgbe_filter.c
index 5a7efe7a73..83f86a747a 100644
--- a/drivers/net/cxgbe/cxgbe_filter.c
+++ b/drivers/net/cxgbe/cxgbe_filter.c
@@ -219,7 +219,7 @@ static inline void mk_set_tcb_field_ulp(struct filter_entry *f,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*req) - sizeof(struct work_request_hdr));
 	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_SET_TCB_FIELD, f->tid));
@@ -371,7 +371,7 @@ static void mk_abort_req_ulp(struct cpl_abort_req *abort_req,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_req) -
 			      sizeof(struct work_request_hdr));
@@ -395,7 +395,7 @@ static void mk_abort_rpl_ulp(struct cpl_abort_rpl *abort_rpl,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_rpl), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_rpl), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_rpl) -
 			      sizeof(struct work_request_hdr));
diff --git a/drivers/net/cxgbe/cxgbe_ofld.h b/drivers/net/cxgbe/cxgbe_ofld.h
index fd1e7d8002..ef56eef426 100644
--- a/drivers/net/cxgbe/cxgbe_ofld.h
+++ b/drivers/net/cxgbe/cxgbe_ofld.h
@@ -14,7 +14,7 @@
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_TP_WR) | \
 				V_FW_WR_IMMDLEN(sizeof(*w) - sizeof(w->wr))); \
 	(w)->wr.wr_mid = cpu_to_be32( \
-				V_FW_WR_LEN16(DIV_ROUND_UP(sizeof(*w), 16)) | \
+				V_FW_WR_LEN16(RTE_DIV_ROUND_UP(sizeof(*w), 16)) | \
 				V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
@@ -27,7 +27,7 @@
 #define INIT_ULPTX_WR(w, wrlen, atomic, tid) do { \
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_ULPTX_WR) | \
 				    V_FW_WR_ATOMIC(atomic)); \
-	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(DIV_ROUND_UP(wrlen, 16)) | \
+	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(RTE_DIV_ROUND_UP(wrlen, 16)) | \
 				     V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
diff --git a/drivers/net/cxgbe/sge.c b/drivers/net/cxgbe/sge.c
index e9d45f24c4..eb25d80e33 100644
--- a/drivers/net/cxgbe/sge.c
+++ b/drivers/net/cxgbe/sge.c
@@ -459,7 +459,7 @@ static inline unsigned int sgl_len(unsigned int n)
  */
 static inline unsigned int flits_to_desc(unsigned int n)
 {
-	return DIV_ROUND_UP(n, 8);
+	return RTE_DIV_ROUND_UP(n, 8);
 }
 
 /**
@@ -505,7 +505,7 @@ static inline unsigned int calc_tx_flits(const struct rte_mbuf *m,
 
 	hdrlen = is_eth_imm(m);
 	if (hdrlen)
-		return DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
+		return RTE_DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
 
 	/*
 	 * Otherwise, we're going to have to construct a Scatter gather list
@@ -779,7 +779,7 @@ static inline void ship_tx_pkt_coalesce_wr(struct adapter *adap,
 	wr = (void *)&q->desc[q->pidx];
 	vmwr = (void *)&q->desc[q->pidx];
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(q->coalesce.flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(q->coalesce.flits, 2));
 	ndesc = flits_to_desc(q->coalesce.flits);
 	wr->equiq_to_len16 = htonl(wr_mid);
 	wr->plen = cpu_to_be16(q->coalesce.len);
@@ -866,7 +866,7 @@ static inline int should_tx_packet_coalesce(struct sge_eth_txq *txq,
 
 	/* If coalescing is on, the mbuf is added to a pkts WR */
 	if (q->coalesce.idx) {
-		ndesc = DIV_ROUND_UP(q->coalesce.flits + flits, 8);
+		ndesc = RTE_DIV_ROUND_UP(q->coalesce.flits + flits, 8);
 		credits = txq_avail(q) - ndesc;
 
 		if (unlikely(wraps_around(q, ndesc)))
@@ -950,7 +950,7 @@ static inline int tx_do_packet_coalesce(struct sge_eth_txq *txq,
 		mc->cmd_dest = htonl(V_ULPTX_CMD(4) | V_ULP_TXPKT_DEST(0) |
 				     V_ULP_TXPKT_FID(adap->sge.fw_evtq.cntxt_id) |
 				     F_ULP_TXPKT_RO);
-		mc->len = htonl(DIV_ROUND_UP(flits, 2));
+		mc->len = htonl(RTE_DIV_ROUND_UP(flits, 2));
 		sc_imm = (struct ulptx_idata *)(mc + 1);
 		sc_imm->cmd_more = htonl(V_ULPTX_CMD(ULP_TX_SC_IMM) |
 					 F_ULP_TX_SC_MORE);
@@ -1116,7 +1116,7 @@ int t4_eth_xmit(struct sge_eth_txq *txq, struct rte_mbuf *mbuf,
 		goto out_free;
 	}
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(flits, 2));
 	if (Q_IDXDIFF(&txq->q, equeidx)  >= 64) {
 		txq->q.equeidx = txq->q.pidx;
 		wr_mid |= F_FW_WR_EQUEQ;
@@ -1322,7 +1322,7 @@ static int ctrl_xmit(struct sge_ctrl_txq *q, struct rte_mbuf *mbuf)
 	}
 
 	reclaim_completed_tx_imm(&q->q);
-	ndesc = DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
+	ndesc = RTE_DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
 	t4_os_lock(&q->ctrlq_lock);
 
 	q->full = txq_avail(&q->q) < ndesc ? 1 : 0;
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 11/14] net/ena: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/net/ena/base/ena_eth_com.h   | 4 ++--
 drivers/net/ena/base/ena_plat_dpdk.h | 1 -
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ena/base/ena_eth_com.h b/drivers/net/ena/base/ena_eth_com.h
index f773699467..4c82415181 100644
--- a/drivers/net/ena/base/ena_eth_com.h
+++ b/drivers/net/ena/base/ena_eth_com.h
@@ -160,8 +160,8 @@ static inline bool ena_com_is_doorbell_needed(struct ena_com_io_sq *io_sq,
 
 	if (num_descs > llq_info->descs_num_before_header) {
 		descs_after_first_entry = num_descs - llq_info->descs_num_before_header;
-		num_entries_needed += DIV_ROUND_UP(descs_after_first_entry,
-						   llq_info->descs_per_entry);
+		num_entries_needed += RTE_DIV_ROUND_UP(descs_after_first_entry,
+						       llq_info->descs_per_entry);
 	}
 
 	ena_trc_dbg(ena_com_io_sq_to_ena_dev(io_sq),
diff --git a/drivers/net/ena/base/ena_plat_dpdk.h b/drivers/net/ena/base/ena_plat_dpdk.h
index c84420de22..9c8a96c918 100644
--- a/drivers/net/ena/base/ena_plat_dpdk.h
+++ b/drivers/net/ena/base/ena_plat_dpdk.h
@@ -304,7 +304,6 @@ ena_mem_alloc_coherent(struct rte_eth_dev_data *data, size_t size,
 			rte_write64_relaxed(*from, to);			       \
 	} while(0)
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 
 #define ENA_FFS(x) ffs(x)
 
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 12/14] raw/ifpga/base: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/raw/ifpga/base/opae_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/raw/ifpga/base/opae_osdep.h b/drivers/raw/ifpga/base/opae_osdep.h
index e35a21c80e..b0045a36a0 100644
--- a/drivers/raw/ifpga/base/opae_osdep.h
+++ b/drivers/raw/ifpga/base/opae_osdep.h
@@ -71,10 +71,9 @@ struct uuid {
 } while (0)
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define udelay(x) opae_udelay(x)
 #define msleep(x) opae_udelay(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #define time_after(a, b)	((long)((b) - (a)) < 0)
 #define time_before(a, b)	time_after(b, a)
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 13/14] drivers: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 drivers/common/nfp/nfp_platform.h  | 2 --
 drivers/net/nfp/nfdk/nfp_nfdk.h    | 5 +++--
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c | 3 ++-
 drivers/net/nfp/nfpcore/nfp_nsp.c  | 2 +-
 4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/common/nfp/nfp_platform.h b/drivers/common/nfp/nfp_platform.h
index 0b02fcf1e8..814f90b5a2 100644
--- a/drivers/common/nfp/nfp_platform.h
+++ b/drivers/common/nfp/nfp_platform.h
@@ -10,8 +10,6 @@
 
 #include <rte_bitops.h>
 
-#define DIV_ROUND_UP(n, d)             (((n) + (d) - 1) / (d))
-
 #define DMA_BIT_MASK(n)    ((1ULL << (n)) - 1)
 
 #define BITS_PER_LONG      (__SIZEOF_LONG__ * 8)
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk.h b/drivers/net/nfp/nfdk/nfp_nfdk.h
index 64c2b31308..b831e0d69e 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk.h
+++ b/drivers/net/nfp/nfdk/nfp_nfdk.h
@@ -168,8 +168,9 @@ static inline uint16_t
 nfp_net_nfdk_headlen_to_segs(uint16_t headlen)
 {
 	/* First descriptor fits less data, so adjust for that */
-	return DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC - NFDK_TX_MAX_DATA_PER_HEAD,
-			NFDK_TX_MAX_DATA_PER_DESC);
+	return RTE_DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC -
+				NFDK_TX_MAX_DATA_PER_HEAD,
+				NFDK_TX_MAX_DATA_PER_DESC);
 }
 
 /* Set TX CSUM offload flags in TX descriptor of nfdk */
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
index 15867ab62f..86b3c5b64a 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
+++ b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
@@ -103,7 +103,8 @@ nfp_net_nfdk_tx_maybe_close_block(struct nfp_net_txq *txq,
 	n_descs = nfp_net_nfdk_headlen_to_segs(pkt_temp->data_len);
 	while (pkt_temp->next != NULL) {
 		pkt_temp = pkt_temp->next;
-		n_descs += DIV_ROUND_UP(pkt_temp->data_len, NFDK_TX_MAX_DATA_PER_DESC);
+		n_descs += RTE_DIV_ROUND_UP(pkt_temp->data_len,
+					    NFDK_TX_MAX_DATA_PER_DESC);
 	}
 
 	if (unlikely(n_descs > NFDK_TX_DESC_GATHER_MAX))
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp.c b/drivers/net/nfp/nfpcore/nfp_nsp.c
index 43c13baabe..bd02efcc4a 100644
--- a/drivers/net/nfp/nfpcore/nfp_nsp.c
+++ b/drivers/net/nfp/nfpcore/nfp_nsp.c
@@ -542,7 +542,7 @@ nfp_nsp_command_buf_dma_sg(struct nfp_nsp *nsp,
 	struct nfp_nsp_dma_desc *descs;
 
 	chunk_size = RTE_BIT64(chunk_order);
-	nseg = DIV_ROUND_UP(max_size, chunk_size);
+	nseg = RTE_DIV_ROUND_UP(max_size, chunk_size);
 
 	/* Malloc memory */
 	total_len = sizeof(struct nfp_nsp_dma_buf) +
-- 
2.55.0.691.gc56d675ccc-goog

[PATCH v2 14/14] app/procinfo: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
 app/proc-info/main.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/app/proc-info/main.c b/app/proc-info/main.c
index 5925bbb765..9c22fb8f01 100644
--- a/app/proc-info/main.c
+++ b/app/proc-info/main.c
@@ -48,7 +48,6 @@
 
 #define ETHDEV_FWVERS_LEN 32
 #define RTE_RETA_CONF_GROUP_NUM 32
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define EEPROM_DUMP_CHUNKSIZE 1024
 
 #define STATS_BDR_FMT "========================================"
@@ -1830,7 +1829,7 @@ show_port_rss_reta_info(void)
 			return;
 		}
 
-		num = DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
+		num = RTE_DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
 		memset(reta_conf, 0, sizeof(reta_conf));
 		for (i = 0; i < num; i++)
 			reta_conf[i].mask = ~0ULL;
-- 
2.55.0.691.gc56d675ccc-goog

Re: [PATCH v2 00/14] Introduce RTE_DIV_ROUND_UP

[Stephen Hemminger]

On Thu, 13 Aug 2026 10:22:25 -0700
Joshua Washington <joshwash@google.com> wrote:

> Numerous drivers independently define the DIV_ROUND_UP macro, using the
> same definition. Attempt to commonize by introducing RTE_DIV_ROUND_UP to
> rte_common.h and replacing usages of DIV_ROUND_UP with RTE_DIV_ROUND_UP.
> 
> Changes in v2:
>   - Split single patch into series.
> 
> Joshua Washington (14):
>   eal: introduce division round up macro
>   net/i40e: use common division round up macro
>   net/idpf: use common division round up macro
>   net/ice: use common division round up macro
>   net/qede: use common division round up macro
>   net/rnp: use common division round up macro
>   net/hns3: use common division round up macro
>   net/hinic: use common division round up macro
>   net/bnx2x: use common division round up macro
>   net/cxgbe: use common division round up macro
>   net/ena: use common division round up macro
>   raw/ifpga/base: use common division round up macro
>   drivers: use common division round up macro
>   app/procinfo: use common division round up macro
> 
>  app/proc-info/main.c                        |  3 +--
>  drivers/common/nfp/nfp_platform.h           |  2 --
>  drivers/net/bnx2x/bnx2x.c                   |  4 ++--
>  drivers/net/bnx2x/bnx2x.h                   | 11 ++++-------
>  drivers/net/cxgbe/base/t4_hw.c              |  4 ++--
>  drivers/net/cxgbe/base/t4vf_hw.c            | 10 +++++-----
>  drivers/net/cxgbe/cxgbe_compat.h            |  3 +--
>  drivers/net/cxgbe/cxgbe_filter.c            |  6 +++---
>  drivers/net/cxgbe/cxgbe_ofld.h              |  4 ++--
>  drivers/net/cxgbe/sge.c                     | 14 +++++++-------
>  drivers/net/ena/base/ena_eth_com.h          |  4 ++--
>  drivers/net/ena/base/ena_plat_dpdk.h        |  1 -
>  drivers/net/hinic/base/hinic_compat.h       |  5 +----
>  drivers/net/hns3/hns3_ethdev.h              |  2 --
>  drivers/net/hns3/hns3_fdir.c                |  2 +-
>  drivers/net/hns3/hns3_regs.c                |  8 ++++----
>  drivers/net/hns3/hns3_rss.c                 |  8 ++++----
>  drivers/net/intel/i40e/base/i40e_osdep.h    |  3 +--
>  drivers/net/intel/ice/base/ice_osdep.h      |  3 +--
>  drivers/net/intel/idpf/base/idpf_osdep.h    |  3 +--
>  drivers/net/nfp/nfdk/nfp_nfdk.h             |  5 +++--
>  drivers/net/nfp/nfdk/nfp_nfdk_dp.c          |  3 ++-
>  drivers/net/nfp/nfpcore/nfp_nsp.c           |  2 +-
>  drivers/net/qede/base/bcm_osal.h            |  2 --
>  drivers/net/qede/base/ecore_chain.h         |  2 +-
>  drivers/net/qede/base/ecore_cxt.c           | 20 +++++++++++---------
>  drivers/net/qede/base/ecore_init_fw_funcs.c | 21 ++++++++++++---------
>  drivers/net/qede/base/ecore_mcp.c           |  6 +++---
>  drivers/net/qede/qede_debug.c               | 16 ++++++++--------
>  drivers/net/rnp/base/rnp_osdep.h            |  5 ++---
>  drivers/net/rnp/rnp_rxtx.c                  |  2 +-
>  drivers/raw/ifpga/base/opae_osdep.h         |  3 +--
>  lib/eal/include/rte_common.h                |  5 +++++
>  33 files changed, 92 insertions(+), 100 deletions(-)
> 

Looks good, the only feedback from AI runs were noise
about blank lines etc.

Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH v2 00/14] Introduce RTE_DIV_ROUND_UP

[Stephen Hemminger]

On Thu, 13 Aug 2026 10:22:25 -0700
Joshua Washington <joshwash@google.com> wrote:

> Numerous drivers independently define the DIV_ROUND_UP macro, using the
> same definition. Attempt to commonize by introducing RTE_DIV_ROUND_UP to
> rte_common.h and replacing usages of DIV_ROUND_UP with RTE_DIV_ROUND_UP.


Looks ok, predates use of AI review, so ran it manually.

The biggest issue reported is tat changing base/ code here
is problematic. You can do it a base/ wrapper header but
not directly at call sites.

Tested: applies cleanly to main, builds with -Werror on x86/gcc for all
touched drivers, and no DIV_ROUND_UP users remain.

Patch 1:

    Needs a test in app/test/test_common.c (exact, remainder, n=0, d=1).
    Needs a release notes entry.
    Doxygen should document the params and the caveats: d evaluated twice,
    unsigned only, and n + d - 1 can wrap (RTE_DIV_ROUND_UP(UINT32_MAX, 2)
    on uint32_t is 0). Consider ((n) / (d) + ((n) % (d) != 0)) instead.
    DIV_CEIL in app/test/test_compressdev.c and
    app/test-compress-perf/comp_perf_test_common.c could be converted too.

Base code (qede, cxgbe, ena):
Please don't edit call sites under base/. That code is synced from
vendor trees (ena-com is shared with Linux and FreeBSD) and this will
conflict on the next update. Do what i40e/ice/idpf did: keep
#define DIV_ROUND_UP(n, d) RTE_DIV_ROUND_UP(n, d)
in the osdep/platform header and leave base/ alone.

Minor:

    qede, nfp: drop the rewraps; the original lines fit in 100 columns.
    hinic: the "X must be nonnegative" comment is lost; move it to the
    rte_common.h doxygen.
    ena: leaves a double blank line in ena_plat_dpdk.h.
    13/14: subject should be "net/nfp:", not "drivers:".

The osdep-only patches (i40e, idpf, ice, hns3, bnx2x, ifpga, procinfo)
look fine.

[PATCH v3 00/16] Introduce RTE_DIV_ROUND_UP

[Joshua Washington]

Numerous components independently define the DIV_ROUND_UP macro, using
the same definition. Attempt to commonize by introducing
RTE_DIV_ROUND_UP to rte_common.h and replacing usages of DIV_ROUND_UP
with RTE_DIV_ROUND_UP. Similarly, replace DIV_CEIL, which is
functionally identical, to RTE_DIV_ROUND_UP.

Changes in v3:
  - eal: Add unit tests in app/test/test_common.c.
  - eal: Add release notes entry for RTE_DIV_ROUND_UP.
  - eal: Expand Doxygen documentation with parameters and caveats
    (divisor evaluation, non-negative inputs, wraparound).
  - base code (qede, cxgbe, ena): Keep DIV_ROUND_UP macro in osdep/platform
    headers and avoid modifying call sites under base/.
  - nfp: Change commit subject prefix from "drivers:" to "net/nfp:".
  - qede, nfp: Drop unnecessary line rewraps.
  - ena: Remove extraneous blank line in ena_plat_dpdk.h.
  - app: Convert DIV_CEIL in app/test and app/test-compress-perf to
    RTE_DIV_ROUND_UP.

Changes in v2:
  - Split single patch into series.

Joshua Washington (16):
  eal: introduce division round up macro
  net/i40e: use common division round up macro
  net/idpf: use common division round up macro
  net/ice: use common division round up macro
  net/qede: use common division round up macro
  net/rnp: use common division round up macro
  net/hns3: use common division round up macro
  net/hinic: use common division round up macro
  net/bnx2x: use common division round up macro
  net/cxgbe: use common division round up macro
  net/ena: use common division round up macro
  raw/ifpga/base: use common division round up macro
  net/nfp: use common division round up macro
  app/procinfo: use common division round up macro
  app/test: use common division round up macro
  app/test-compress-perf: use common division round up macro

 app/proc-info/main.c                           |  3 +--
 app/test-compress-perf/comp_perf_test_common.c | 13 +++++--------
 app/test/test_common.c                         | 10 ++++++++++
 app/test/test_compressdev.c                    |  4 +---
 doc/guides/rel_notes/release_26_11.rst         |  5 +++++
 drivers/common/nfp/nfp_platform.h              |  2 --
 drivers/net/bnx2x/bnx2x.c                      |  4 ++--
 drivers/net/bnx2x/bnx2x.h                      | 11 ++++-------
 drivers/net/cxgbe/cxgbe_compat.h               |  2 +-
 drivers/net/cxgbe/cxgbe_filter.c               |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h                 |  4 ++--
 drivers/net/cxgbe/sge.c                        | 14 +++++++-------
 drivers/net/ena/base/ena_plat_dpdk.h           |  2 +-
 drivers/net/hinic/base/hinic_compat.h          |  5 +----
 drivers/net/hns3/hns3_ethdev.h                 |  2 --
 drivers/net/hns3/hns3_fdir.c                   |  2 +-
 drivers/net/hns3/hns3_regs.c                   |  8 ++++----
 drivers/net/hns3/hns3_rss.c                    |  8 ++++----
 drivers/net/intel/i40e/base/i40e_osdep.h       |  3 +--
 drivers/net/intel/ice/base/ice_osdep.h         |  3 +--
 drivers/net/intel/idpf/base/idpf_osdep.h       |  3 +--
 drivers/net/nfp/nfdk/nfp_nfdk.h                |  2 +-
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c             |  2 +-
 drivers/net/nfp/nfpcore/nfp_nsp.c              |  2 +-
 drivers/net/qede/base/bcm_osal.h               |  1 -
 drivers/net/qede/qede_debug.c                  | 16 ++++++++--------
 drivers/net/rnp/base/rnp_osdep.h               |  5 ++---
 drivers/net/rnp/rnp_rxtx.c                     |  2 +-
 drivers/raw/ifpga/base/opae_osdep.h            |  3 +--
 lib/eal/include/rte_common.h                   | 18 ++++++++++++++++++
 30 files changed, 88 insertions(+), 77 deletions(-)

-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 01/16] eal: introduce division round up macro

[Joshua Washington]

A number of drivers define the DIV_ROUND_UP macro -- add it to
rte_common.h so that other code looking to use that functionality
doesn't have to re-define it.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - Add unit tests in app/test/test_common.c.
  - Add release notes entry.
  - Enhance Doxygen documentation to document parameters and caveats
    (divisor evaluation, non-negative inputs, wraparound).
Changes in v2:
  - Split out patches replacing usages
---
 app/test/test_common.c                 | 10 ++++++++++
 doc/guides/rel_notes/release_26_11.rst |  5 +++++
 lib/eal/include/rte_common.h           | 18 ++++++++++++++++++
 3 files changed, 33 insertions(+)

diff --git a/app/test/test_common.c b/app/test/test_common.c
index 3e1c7df0c1..bce2776ea6 100644
--- a/app/test/test_common.c
+++ b/app/test/test_common.c
@@ -47,6 +47,16 @@ test_macros(int __rte_unused unused_parm)
 		"RTE_MAX");
 	RTE_TEST_ASSERT_EQUAL(RTE_MIN(SMALLER, BIGGER), SMALLER,
 		"RTE_MIN");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(10, 5), 2,
+		"RTE_DIV_ROUND_UP exact");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(11, 5), 3,
+		"RTE_DIV_ROUND_UP remainder");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(0, 5), 0,
+		"RTE_DIV_ROUND_UP n=0");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(0, 1), 0,
+		"RTE_DIV_ROUND_UP n=0, d=1");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(5, 1), 5,
+		"RTE_DIV_ROUND_UP d=1");
 
 	RTE_TEST_ASSERT_EQUAL(RTE_PTR_ADD(arr + 1, sizeof(arr[0])), &arr[2],
 		"RTE_PTR_ADD(expr, x)");
diff --git a/doc/guides/rel_notes/release_26_11.rst b/doc/guides/rel_notes/release_26_11.rst
index c8cc86295d..7720fd9f96 100644
--- a/doc/guides/rel_notes/release_26_11.rst
+++ b/doc/guides/rel_notes/release_26_11.rst
@@ -55,6 +55,11 @@ New Features
      Also, make sure to start the actual text at the margin.
      =======================================================
 
+* **Added division round up macro.**
+
+  Added the ``RTE_DIV_ROUND_UP`` macro in ``rte_common.h`` to divide dividend
+  by divisor, rounding up to the nearest integer.
+
 
 Removed Items
 -------------
diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
index 79d2a0ab93..9707f7e255 100644
--- a/lib/eal/include/rte_common.h
+++ b/lib/eal/include/rte_common.h
@@ -717,6 +717,24 @@ static void __attribute__((destructor(RTE_PRIO(prio)), used)) func(void)
 		(ceil - (v)) > ((v) - floor) ? floor : ceil;	\
 	})
 
+/**
+ * Macro to divide dividend @p n by divisor @p d, rounding up if @p d does not divide @p n.
+ *
+ * @param n
+ *   Dividend. Must be a non-negative integer to round correctly.
+ * @param d
+ *   Divisor. Must be a positive integer. Evaluated twice.
+ * @return
+ *   Result of dividing @p n by @p d, rounded up.
+ *
+ * @note The divisor @p d is evaluated twice. Arguments must not have side effects.
+ * @note Both @p n and @p d must be non-negative to round correctly.
+ * @note The sum @c (n) + (d) - 1 can wrap around if @p n is close to the maximum
+ *       representable value of its type (e.g. RTE_DIV_ROUND_UP(UINT32_MAX, 2)
+ *       on uint32_t wraps to 0).
+ */
+#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+
 /**
  * Checks if a pointer is aligned to a given power-of-two value
  *
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 02/16] net/i40e: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/intel/i40e/base/i40e_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h b/drivers/net/intel/i40e/base/i40e_osdep.h
index 9c5a94f055..93620f5c50 100644
--- a/drivers/net/intel/i40e/base/i40e_osdep.h
+++ b/drivers/net/intel/i40e/base/i40e_osdep.h
@@ -227,12 +227,11 @@ struct i40e_spinlock {
 #define i40e_memset(a, b, c, d) memset((a), (b), (c))
 #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
 
-#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us_sleep(x)
 #define i40e_usec_delay(x) DELAY(x)
 #define i40e_msec_delay(x) DELAY(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000*(x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #endif /* _I40E_OSDEP_H_ */
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 03/16] net/idpf: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/intel/idpf/base/idpf_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/idpf/base/idpf_osdep.h b/drivers/net/intel/idpf/base/idpf_osdep.h
index 67bc33f2e6..64bbd00eea 100644
--- a/drivers/net/intel/idpf/base/idpf_osdep.h
+++ b/drivers/net/intel/idpf/base/idpf_osdep.h
@@ -291,13 +291,12 @@ idpf_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define idpf_usec_delay(x) rte_delay_us(x)
 #define idpf_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #ifndef IDPF_DBG_TRACE
 #define IDPF_DBG_TRACE	  BIT_ULL(0)
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 04/16] net/ice: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/intel/ice/base/ice_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/ice/base/ice_osdep.h b/drivers/net/intel/ice/base/ice_osdep.h
index 5d92cf5aa4..3ef52eae6d 100644
--- a/drivers/net/intel/ice/base/ice_osdep.h
+++ b/drivers/net/intel/ice/base/ice_osdep.h
@@ -311,13 +311,12 @@ ice_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define ice_usec_delay(x, y) rte_delay_us(x)
 #define ice_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 struct ice_list_entry {
 	LIST_ENTRY(ice_list_entry) next;
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 05/16] net/qede: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP and
RTE_DIV_ROUND_UP definitions redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - Leave call sites in base/ alone and map DIV_ROUND_UP to
    RTE_DIV_ROUND_UP in bcm_osal.h.
  - Drop line rewraps.
---
 drivers/net/qede/base/bcm_osal.h |  1 -
 drivers/net/qede/qede_debug.c    | 16 ++++++++--------
 2 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/drivers/net/qede/base/bcm_osal.h b/drivers/net/qede/base/bcm_osal.h
index 357981f63d..7c47a8f614 100644
--- a/drivers/net/qede/base/bcm_osal.h
+++ b/drivers/net/qede/base/bcm_osal.h
@@ -396,7 +396,6 @@ int qede_save_fw_dump(uint16_t port_id);
 
 /* Utility functions */
 
-#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DIV_ROUND_UP(size, to_what) RTE_DIV_ROUND_UP(size, to_what)
 #define RTE_ROUNDUP(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define ROUNDUP(value, to_what) RTE_ROUNDUP((value), (to_what))
diff --git a/drivers/net/qede/qede_debug.c b/drivers/net/qede/qede_debug.c
index 1d3147b724..c4cf40b15d 100644
--- a/drivers/net/qede/qede_debug.c
+++ b/drivers/net/qede/qede_debug.c
@@ -332,7 +332,7 @@ struct split_type_defs {
 
 #define BYTES_IN_DWORD			sizeof(u32)
 /* In the macros below, size and offset are specified in bits */
-#define CEIL_DWORDS(size)		DIV_ROUND_UP(size, 32)
+#define CEIL_DWORDS(size)		RTE_DIV_ROUND_UP(size, 32)
 #define FIELD_BIT_OFFSET(type, field)	type ## _ ## field ## _ ## OFFSET
 #define FIELD_BIT_SIZE(type, field)	type ## _ ## field ## _ ## SIZE
 #define FIELD_DWORD_OFFSET(type, field) \
@@ -3060,7 +3060,7 @@ static u32 qed_grc_dump_big_ram(struct ecore_hwfn *p_hwfn,
 		return offset + ram_size;
 
 	/* Dump Big RAM */
-	for (i = 0; i < DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
+	for (i = 0; i < RTE_DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
 	     i++) {
 		u32 addr, len;
 
@@ -4161,8 +4161,8 @@ static enum dbg_status qed_mcp_trace_dump(struct ecore_hwfn *p_hwfn,
 
 	/* Find trace data size */
 	trace_data_size_dwords =
-	    DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
-			 BYTES_IN_DWORD);
+	    RTE_DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
+			     BYTES_IN_DWORD);
 
 	/* Dump trace data section header and param */
 	offset += qed_dump_section_hdr(dump_buf + offset,
@@ -4817,10 +4817,10 @@ static u32 qed_ilt_dump(struct ecore_hwfn *p_hwfn,
 		offset += num_pages * PAGE_MEM_DESC_SIZE_DWORDS;
 	}
 
-	valid_conn_pf_pages = DIV_ROUND_UP(valid_conn_pf_cids,
-					   num_cids_per_page);
-	valid_conn_vf_pages = DIV_ROUND_UP(valid_conn_vf_cids,
-					   num_cids_per_page);
+	valid_conn_pf_pages = RTE_DIV_ROUND_UP(valid_conn_pf_cids,
+					       num_cids_per_page);
+	valid_conn_vf_pages = RTE_DIV_ROUND_UP(valid_conn_vf_cids,
+					       num_cids_per_page);
 
 	/* Dump ILT pages IDs */
 	offset += qed_ilt_dump_pages_section(p_hwfn,
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 06/16] net/rnp: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Base BITS_PER_BYTE and BITS_TO_LONGS definitions on whether
BITS_TO_LONGS is defined instead.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/rnp/base/rnp_osdep.h | 5 ++---
 drivers/net/rnp/rnp_rxtx.c       | 2 +-
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/net/rnp/base/rnp_osdep.h b/drivers/net/rnp/base/rnp_osdep.h
index 58dcb40f90..b0e7906fab 100644
--- a/drivers/net/rnp/base/rnp_osdep.h
+++ b/drivers/net/rnp/base/rnp_osdep.h
@@ -53,10 +53,9 @@ typedef uint64_t u64;
 #define cpu_to_le32(v)	rte_cpu_to_le_32((u32)(v))
 #endif
 
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d)      (((n) + (d) - 1) / (d))
+#ifndef BITS_TO_LONGS
 #define BITS_PER_BYTE           (8)
-#define BITS_TO_LONGS(nr)       DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
+#define BITS_TO_LONGS(nr)       RTE_DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
 #endif
 
 #define fls(n)	rte_fls_u32(n)
diff --git a/drivers/net/rnp/rnp_rxtx.c b/drivers/net/rnp/rnp_rxtx.c
index 81e8c6ba44..143cd11ffe 100644
--- a/drivers/net/rnp/rnp_rxtx.c
+++ b/drivers/net/rnp/rnp_rxtx.c
@@ -1165,7 +1165,7 @@ rnp_calc_pkt_desc(struct rte_mbuf *tx_pkt)
 	uint16_t count = 0;
 
 	while (txd != NULL) {
-		count += DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
+		count += RTE_DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
 		txd = txd->next;
 	}
 
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 07/16] net/hns3: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/hns3/hns3_ethdev.h | 2 --
 drivers/net/hns3/hns3_fdir.c   | 2 +-
 drivers/net/hns3/hns3_regs.c   | 8 ++++----
 drivers/net/hns3/hns3_rss.c    | 8 ++++----
 4 files changed, 9 insertions(+), 11 deletions(-)

diff --git a/drivers/net/hns3/hns3_ethdev.h b/drivers/net/hns3/hns3_ethdev.h
index 209b042816..b4feff4b94 100644
--- a/drivers/net/hns3/hns3_ethdev.h
+++ b/drivers/net/hns3/hns3_ethdev.h
@@ -978,8 +978,6 @@ static inline struct hns3_vf *HNS3_DEV_HW_TO_VF(struct hns3_hw *hw)
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define rounddown(x, y) ((x) - ((x) % (y)))
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-
 /*
  * Because hardware always access register in little-endian mode based on hns3
  * network engine, so driver should also call rte_cpu_to_le_32 to convert data
diff --git a/drivers/net/hns3/hns3_fdir.c b/drivers/net/hns3/hns3_fdir.c
index 50572ae430..dfcc5c0d99 100644
--- a/drivers/net/hns3/hns3_fdir.c
+++ b/drivers/net/hns3/hns3_fdir.c
@@ -129,7 +129,7 @@ static const struct key_info tuple_key_info[] = {
 #define MAX_KEY_LENGTH		400
 #define MAX_200B_KEY_LENGTH	200
 #define MAX_META_DATA_LENGTH	16
-#define MAX_KEY_DWORDS	DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
+#define MAX_KEY_DWORDS	RTE_DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
 #define MAX_KEY_BYTES	(MAX_KEY_DWORDS * 4)
 
 enum HNS3_FD_PACKET_TYPE {
diff --git a/drivers/net/hns3/hns3_regs.c b/drivers/net/hns3/hns3_regs.c
index 5c74f9ae2e..066b836556 100644
--- a/drivers/net/hns3/hns3_regs.c
+++ b/drivers/net/hns3/hns3_regs.c
@@ -1034,8 +1034,8 @@ hns3_get_32_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
-			       HNS3_32_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
+				   HNS3_32_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-32bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
@@ -1093,8 +1093,8 @@ hns3_get_64_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
-			       HNS3_64_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
+				   HNS3_64_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-64bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
diff --git a/drivers/net/hns3/hns3_rss.c b/drivers/net/hns3/hns3_rss.c
index 508b3e26d0..6cb8d9482d 100644
--- a/drivers/net/hns3/hns3_rss.c
+++ b/drivers/net/hns3/hns3_rss.c
@@ -295,7 +295,7 @@ hns3_rss_set_algo_key(struct hns3_hw *hw, uint8_t hash_algo,
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
 
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  false);
@@ -335,7 +335,7 @@ hns3_rss_get_algo_key(struct hns3_hw *hw,  uint8_t *hash_algo,
 	int ret;
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  true);
@@ -381,7 +381,7 @@ hns3_set_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  false);
@@ -431,7 +431,7 @@ hns3_get_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  true);
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 08/16] net/hinic: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/hinic/base/hinic_compat.h | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/net/hinic/base/hinic_compat.h b/drivers/net/hinic/base/hinic_compat.h
index 707a3b92b9..f4398a7847 100644
--- a/drivers/net/hinic/base/hinic_compat.h
+++ b/drivers/net/hinic/base/hinic_compat.h
@@ -70,11 +70,8 @@ typedef uint64_t  dma_addr_t;
 #define upper_32_bits(n) ((u32)(((n) >> 16) >> 16))
 #define lower_32_bits(n) ((u32)(n))
 
-/* Returns X / Y, rounding up.  X must be nonnegative to round correctly. */
-#define DIV_ROUND_UP(X, Y) (((X) + ((Y) - 1)) / (Y))
-
 /* Returns X rounded up to the nearest multiple of Y. */
-#define ROUND_UP(X, Y) (DIV_ROUND_UP(X, Y) * (Y))
+#define ROUND_UP(X, Y) (RTE_DIV_ROUND_UP(X, Y) * (Y))
 
 #undef  ALIGN
 #define ALIGN(x, a)  RTE_ALIGN(x, a)
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 09/16] net/bnx2x: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/bnx2x/bnx2x.c |  4 ++--
 drivers/net/bnx2x/bnx2x.h | 11 ++++-------
 2 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/drivers/net/bnx2x/bnx2x.c b/drivers/net/bnx2x/bnx2x.c
index 8790c858d5..ce9dbdf7c6 100644
--- a/drivers/net/bnx2x/bnx2x.c
+++ b/drivers/net/bnx2x/bnx2x.c
@@ -2334,8 +2334,8 @@ static void bnx2x_ilt_set_info(struct bnx2x_softc *sc)
 		ilt_client->start = line;
 
 		/* 4 bytes for each cid */
-		line += DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
-				     QM_ILT_PAGE_SZ);
+		line += RTE_DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
+					 QM_ILT_PAGE_SZ);
 
 		ilt_client->end = (line - 1);
 	}
diff --git a/drivers/net/bnx2x/bnx2x.h b/drivers/net/bnx2x/bnx2x.h
index 35206b4758..0cdd118ad5 100644
--- a/drivers/net/bnx2x/bnx2x.h
+++ b/drivers/net/bnx2x/bnx2x.h
@@ -37,9 +37,6 @@
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(arr) RTE_DIM(arr)
 #endif
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-#endif
 #ifndef roundup
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #endif
@@ -393,7 +390,7 @@ union cdu_context {
 #define CNIC_ISCSI_CID_MAX 256
 #define CNIC_FCOE_CID_MAX  2048
 #define CNIC_CID_MAX       (CNIC_ISCSI_CID_MAX + CNIC_FCOE_CID_MAX)
-#define CNIC_ILT_LINES     DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
+#define CNIC_ILT_LINES     RTE_DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
 
 #define QM_ILT_PAGE_SZ_HW  0
 #define QM_ILT_PAGE_SZ     (4096 << QM_ILT_PAGE_SZ_HW) /* 4K */
@@ -405,7 +402,7 @@ union cdu_context {
 /*#define TM_CONN_NUM        (CNIC_STARTING_CID+CNIC_ISCSI_CXT_MAX) */
 #define TM_CONN_NUM        1024
 #define TM_ILT_SZ          (8 * TM_CONN_NUM)
-#define TM_ILT_LINES       DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
+#define TM_ILT_LINES       RTE_DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
 
 /* SRC (Searcher) host DB constants */
 #define SRC_ILT_PAGE_SZ_HW 0
@@ -414,7 +411,7 @@ union cdu_context {
 #define SRC_CONN_NUM       (1 << SRC_HASH_BITS) /* 1024 */
 #define SRC_ILT_SZ         (sizeof(struct src_ent) * SRC_CONN_NUM)
 #define SRC_T2_SZ          SRC_ILT_SZ
-#define SRC_ILT_LINES      DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
+#define SRC_ILT_LINES      RTE_DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
 
 struct hw_context {
     struct bnx2x_dma    vcxt_dma;
@@ -1258,7 +1255,7 @@ struct bnx2x_softc {
 #define BNX2X_L2_CID_COUNT(sc)                                             \
 	(BNX2X_NUM_ETH_QUEUES(sc) * ECORE_MULTI_TX_COS + 2 * CNIC_SUPPORT(sc))
 #define L2_ILT_LINES(sc)                                \
-	(DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
+	(RTE_DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
 
 	int qm_cid_count;
 
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 10/16] net/cxgbe: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - Leave call sites in base/ alone and map DIV_ROUND_UP to
    RTE_DIV_ROUND_UP in cxgbe_compat.h.
---
 drivers/net/cxgbe/cxgbe_compat.h |  2 +-
 drivers/net/cxgbe/cxgbe_filter.c |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h   |  4 ++--
 drivers/net/cxgbe/sge.c          | 14 +++++++-------
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/net/cxgbe/cxgbe_compat.h b/drivers/net/cxgbe/cxgbe_compat.h
index e74d476f6d..8c03a07fde 100644
--- a/drivers/net/cxgbe/cxgbe_compat.h
+++ b/drivers/net/cxgbe/cxgbe_compat.h
@@ -166,7 +166,7 @@ typedef uint64_t  dma_addr_t;
 typedef char *caddr_t;
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#define DIV_ROUND_UP(n, d) RTE_DIV_ROUND_UP(n, d)
 #define DELAY(x) rte_delay_us(x)
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
diff --git a/drivers/net/cxgbe/cxgbe_filter.c b/drivers/net/cxgbe/cxgbe_filter.c
index 5a7efe7a73..83f86a747a 100644
--- a/drivers/net/cxgbe/cxgbe_filter.c
+++ b/drivers/net/cxgbe/cxgbe_filter.c
@@ -219,7 +219,7 @@ static inline void mk_set_tcb_field_ulp(struct filter_entry *f,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*req) - sizeof(struct work_request_hdr));
 	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_SET_TCB_FIELD, f->tid));
@@ -371,7 +371,7 @@ static void mk_abort_req_ulp(struct cpl_abort_req *abort_req,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_req) -
 			      sizeof(struct work_request_hdr));
@@ -395,7 +395,7 @@ static void mk_abort_rpl_ulp(struct cpl_abort_rpl *abort_rpl,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_rpl), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_rpl), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_rpl) -
 			      sizeof(struct work_request_hdr));
diff --git a/drivers/net/cxgbe/cxgbe_ofld.h b/drivers/net/cxgbe/cxgbe_ofld.h
index fd1e7d8002..ef56eef426 100644
--- a/drivers/net/cxgbe/cxgbe_ofld.h
+++ b/drivers/net/cxgbe/cxgbe_ofld.h
@@ -14,7 +14,7 @@
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_TP_WR) | \
 				V_FW_WR_IMMDLEN(sizeof(*w) - sizeof(w->wr))); \
 	(w)->wr.wr_mid = cpu_to_be32( \
-				V_FW_WR_LEN16(DIV_ROUND_UP(sizeof(*w), 16)) | \
+				V_FW_WR_LEN16(RTE_DIV_ROUND_UP(sizeof(*w), 16)) | \
 				V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
@@ -27,7 +27,7 @@
 #define INIT_ULPTX_WR(w, wrlen, atomic, tid) do { \
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_ULPTX_WR) | \
 				    V_FW_WR_ATOMIC(atomic)); \
-	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(DIV_ROUND_UP(wrlen, 16)) | \
+	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(RTE_DIV_ROUND_UP(wrlen, 16)) | \
 				     V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
diff --git a/drivers/net/cxgbe/sge.c b/drivers/net/cxgbe/sge.c
index e9d45f24c4..eb25d80e33 100644
--- a/drivers/net/cxgbe/sge.c
+++ b/drivers/net/cxgbe/sge.c
@@ -459,7 +459,7 @@ static inline unsigned int sgl_len(unsigned int n)
  */
 static inline unsigned int flits_to_desc(unsigned int n)
 {
-	return DIV_ROUND_UP(n, 8);
+	return RTE_DIV_ROUND_UP(n, 8);
 }
 
 /**
@@ -505,7 +505,7 @@ static inline unsigned int calc_tx_flits(const struct rte_mbuf *m,
 
 	hdrlen = is_eth_imm(m);
 	if (hdrlen)
-		return DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
+		return RTE_DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
 
 	/*
 	 * Otherwise, we're going to have to construct a Scatter gather list
@@ -779,7 +779,7 @@ static inline void ship_tx_pkt_coalesce_wr(struct adapter *adap,
 	wr = (void *)&q->desc[q->pidx];
 	vmwr = (void *)&q->desc[q->pidx];
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(q->coalesce.flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(q->coalesce.flits, 2));
 	ndesc = flits_to_desc(q->coalesce.flits);
 	wr->equiq_to_len16 = htonl(wr_mid);
 	wr->plen = cpu_to_be16(q->coalesce.len);
@@ -866,7 +866,7 @@ static inline int should_tx_packet_coalesce(struct sge_eth_txq *txq,
 
 	/* If coalescing is on, the mbuf is added to a pkts WR */
 	if (q->coalesce.idx) {
-		ndesc = DIV_ROUND_UP(q->coalesce.flits + flits, 8);
+		ndesc = RTE_DIV_ROUND_UP(q->coalesce.flits + flits, 8);
 		credits = txq_avail(q) - ndesc;
 
 		if (unlikely(wraps_around(q, ndesc)))
@@ -950,7 +950,7 @@ static inline int tx_do_packet_coalesce(struct sge_eth_txq *txq,
 		mc->cmd_dest = htonl(V_ULPTX_CMD(4) | V_ULP_TXPKT_DEST(0) |
 				     V_ULP_TXPKT_FID(adap->sge.fw_evtq.cntxt_id) |
 				     F_ULP_TXPKT_RO);
-		mc->len = htonl(DIV_ROUND_UP(flits, 2));
+		mc->len = htonl(RTE_DIV_ROUND_UP(flits, 2));
 		sc_imm = (struct ulptx_idata *)(mc + 1);
 		sc_imm->cmd_more = htonl(V_ULPTX_CMD(ULP_TX_SC_IMM) |
 					 F_ULP_TX_SC_MORE);
@@ -1116,7 +1116,7 @@ int t4_eth_xmit(struct sge_eth_txq *txq, struct rte_mbuf *mbuf,
 		goto out_free;
 	}
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(flits, 2));
 	if (Q_IDXDIFF(&txq->q, equeidx)  >= 64) {
 		txq->q.equeidx = txq->q.pidx;
 		wr_mid |= F_FW_WR_EQUEQ;
@@ -1322,7 +1322,7 @@ static int ctrl_xmit(struct sge_ctrl_txq *q, struct rte_mbuf *mbuf)
 	}
 
 	reclaim_completed_tx_imm(&q->q);
-	ndesc = DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
+	ndesc = RTE_DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
 	t4_os_lock(&q->ctrlq_lock);
 
 	q->full = txq_avail(&q->q) < ndesc ? 1 : 0;
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 11/16] net/ena: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - Leave call sites in base/ alone and map DIV_ROUND_UP to
    RTE_DIV_ROUND_UP in ena_plat_dpdk.h.
  - Remove extraneous blank line in ena_plat_dpdk.h.
---
 drivers/net/ena/base/ena_plat_dpdk.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ena/base/ena_plat_dpdk.h b/drivers/net/ena/base/ena_plat_dpdk.h
index c84420de22..ec47c885b1 100644
--- a/drivers/net/ena/base/ena_plat_dpdk.h
+++ b/drivers/net/ena/base/ena_plat_dpdk.h
@@ -304,7 +304,7 @@ ena_mem_alloc_coherent(struct rte_eth_dev_data *data, size_t size,
 			rte_write64_relaxed(*from, to);			       \
 	} while(0)
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#define DIV_ROUND_UP(n, d) RTE_DIV_ROUND_UP(n, d)
 
 #define ENA_FFS(x) ffs(x)
 
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 12/16] raw/ifpga/base: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/raw/ifpga/base/opae_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/raw/ifpga/base/opae_osdep.h b/drivers/raw/ifpga/base/opae_osdep.h
index e35a21c80e..b0045a36a0 100644
--- a/drivers/raw/ifpga/base/opae_osdep.h
+++ b/drivers/raw/ifpga/base/opae_osdep.h
@@ -71,10 +71,9 @@ struct uuid {
 } while (0)
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define udelay(x) opae_udelay(x)
 #define msleep(x) opae_udelay(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #define time_after(a, b)	((long)((b) - (a)) < 0)
 #define time_before(a, b)	time_after(b, a)
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 13/16] net/nfp: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - Change commit subject prefix from "drivers:" to "net/nfp:".
  - Drop line rewraps in nfp_nfdk.h and nfp_nfdk_dp.c.
---
 drivers/common/nfp/nfp_platform.h  | 2 --
 drivers/net/nfp/nfdk/nfp_nfdk.h    | 2 +-
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c | 2 +-
 drivers/net/nfp/nfpcore/nfp_nsp.c  | 2 +-
 4 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/drivers/common/nfp/nfp_platform.h b/drivers/common/nfp/nfp_platform.h
index 0b02fcf1e8..814f90b5a2 100644
--- a/drivers/common/nfp/nfp_platform.h
+++ b/drivers/common/nfp/nfp_platform.h
@@ -10,8 +10,6 @@
 
 #include <rte_bitops.h>
 
-#define DIV_ROUND_UP(n, d)             (((n) + (d) - 1) / (d))
-
 #define DMA_BIT_MASK(n)    ((1ULL << (n)) - 1)
 
 #define BITS_PER_LONG      (__SIZEOF_LONG__ * 8)
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk.h b/drivers/net/nfp/nfdk/nfp_nfdk.h
index 64c2b31308..df4d1eb8d9 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk.h
+++ b/drivers/net/nfp/nfdk/nfp_nfdk.h
@@ -168,7 +168,7 @@ static inline uint16_t
 nfp_net_nfdk_headlen_to_segs(uint16_t headlen)
 {
 	/* First descriptor fits less data, so adjust for that */
-	return DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC - NFDK_TX_MAX_DATA_PER_HEAD,
+	return RTE_DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC - NFDK_TX_MAX_DATA_PER_HEAD,
 			NFDK_TX_MAX_DATA_PER_DESC);
 }
 
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
index 15867ab62f..777550b6df 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
+++ b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
@@ -103,7 +103,7 @@ nfp_net_nfdk_tx_maybe_close_block(struct nfp_net_txq *txq,
 	n_descs = nfp_net_nfdk_headlen_to_segs(pkt_temp->data_len);
 	while (pkt_temp->next != NULL) {
 		pkt_temp = pkt_temp->next;
-		n_descs += DIV_ROUND_UP(pkt_temp->data_len, NFDK_TX_MAX_DATA_PER_DESC);
+		n_descs += RTE_DIV_ROUND_UP(pkt_temp->data_len, NFDK_TX_MAX_DATA_PER_DESC);
 	}
 
 	if (unlikely(n_descs > NFDK_TX_DESC_GATHER_MAX))
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp.c b/drivers/net/nfp/nfpcore/nfp_nsp.c
index 43c13baabe..bd02efcc4a 100644
--- a/drivers/net/nfp/nfpcore/nfp_nsp.c
+++ b/drivers/net/nfp/nfpcore/nfp_nsp.c
@@ -542,7 +542,7 @@ nfp_nsp_command_buf_dma_sg(struct nfp_nsp *nsp,
 	struct nfp_nsp_dma_desc *descs;
 
 	chunk_size = RTE_BIT64(chunk_order);
-	nseg = DIV_ROUND_UP(max_size, chunk_size);
+	nseg = RTE_DIV_ROUND_UP(max_size, chunk_size);
 
 	/* Malloc memory */
 	total_len = sizeof(struct nfp_nsp_dma_buf) +
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 14/16] app/procinfo: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 app/proc-info/main.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/app/proc-info/main.c b/app/proc-info/main.c
index 5925bbb765..9c22fb8f01 100644
--- a/app/proc-info/main.c
+++ b/app/proc-info/main.c
@@ -48,7 +48,6 @@
 
 #define ETHDEV_FWVERS_LEN 32
 #define RTE_RETA_CONF_GROUP_NUM 32
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define EEPROM_DUMP_CHUNKSIZE 1024
 
 #define STATS_BDR_FMT "========================================"
@@ -1830,7 +1829,7 @@ show_port_rss_reta_info(void)
 			return;
 		}
 
-		num = DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
+		num = RTE_DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
 		memset(reta_conf, 0, sizeof(reta_conf));
 		for (i = 0; i < num; i++)
 			reta_conf[i].mask = ~0ULL;
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 15/16] app/test: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - New patch converting DIV_CEIL in test_compressdev.c.
---
 app/test/test_compressdev.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/app/test/test_compressdev.c b/app/test/test_compressdev.c
index 293d1d757f..2891b98d56 100644
--- a/app/test/test_compressdev.c
+++ b/app/test/test_compressdev.c
@@ -18,8 +18,6 @@
 #include "test_compressdev_test_buffer.h"
 #include "test.h"
 
-#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
-
 #define DEFAULT_WINDOW_SIZE 15
 #define DEFAULT_MEM_LEVEL 8
 #define MAX_DEQD_RETRIES 10
@@ -688,7 +686,7 @@ prepare_sgl_bufs(const char *test_buf, struct rte_mbuf *head_buf,
 		uint16_t seg_size)
 {
 	uint32_t remaining_data = total_data_size;
-	uint16_t num_remaining_segs = DIV_CEIL(remaining_data, seg_size);
+	uint16_t num_remaining_segs = RTE_DIV_ROUND_UP(remaining_data, seg_size);
 	struct rte_mempool *pool;
 	struct rte_mbuf *next_seg;
 	uint32_t data_size;
-- 
2.55.0.1082.g2b9226bbc0-goog

[PATCH v3 16/16] app/test-compress-perf: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v3:
  - New patch converting DIV_CEIL in comp_perf_test_common.c.
---
 app/test-compress-perf/comp_perf_test_common.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)

diff --git a/app/test-compress-perf/comp_perf_test_common.c b/app/test-compress-perf/comp_perf_test_common.c
index 78487196ad..3f70c222c4 100644
--- a/app/test-compress-perf/comp_perf_test_common.c
+++ b/app/test-compress-perf/comp_perf_test_common.c
@@ -14,9 +14,6 @@
 #include "comp_perf_test_common.h"
 #include "comp_perf_test_verify.h"
 
-
-#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
-
 struct cperf_buffer_info {
 	uint16_t total_segments;
 	uint16_t segment_sz;
@@ -238,8 +235,8 @@ comp_perf_allocate_memory(struct comp_test_data *test_data,
 		 * Number of segments for input and output
 		 * (compression and decompression)
 		 */
-		test_data->total_segs = DIV_CEIL(test_data->input_data_sz,
-						 test_data->seg_sz);
+		test_data->total_segs = RTE_DIV_ROUND_UP(test_data->input_data_sz,
+							 test_data->seg_sz);
 	} else {
 		/*
 		 * When application does decompression only, input data is
@@ -292,8 +289,8 @@ comp_perf_allocate_memory(struct comp_test_data *test_data,
 		return -1;
 	}
 
-	mem->total_bufs = DIV_CEIL(test_data->total_segs,
-				   test_data->max_sgl_segs);
+	mem->total_bufs = RTE_DIV_ROUND_UP(test_data->total_segs,
+					   test_data->max_sgl_segs);
 
 	snprintf(pool_name, sizeof(pool_name), "op_pool_%u_qp_%u",
 			mem->dev_id, mem->qp_id);
@@ -557,7 +554,7 @@ prepare_bufs(struct comp_test_data *test_data, struct cperf_mem_resources *mem)
 void
 print_test_dynamics(const struct comp_test_data *test_data)
 {
-	uint32_t opt_total_segs = DIV_CEIL(buffer_info.input_data_sz,
+	uint32_t opt_total_segs = RTE_DIV_ROUND_UP(buffer_info.input_data_sz,
 			MAX_SEG_SIZE);
 
 	if (buffer_info.total_buffs > 1) {
-- 
2.55.0.1082.g2b9226bbc0-goog

Re: [PATCH v3 01/16] eal: introduce division round up macro

[Bruce Richardson]

On Tue, Sep 22, 2026 at 06:07:24PM -0700, Joshua Washington wrote:
> A number of drivers define the DIV_ROUND_UP macro -- add it to
> rte_common.h so that other code looking to use that functionality
> doesn't have to re-define it.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

Re: [PATCH v3 02/16] net/i40e: use common division round up macro

[Bruce Richardson]

On Tue, Sep 22, 2026 at 06:07:25PM -0700, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
>  drivers/net/intel/i40e/base/i40e_osdep.h | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h b/drivers/net/intel/i40e/base/i40e_osdep.h
> index 9c5a94f055..93620f5c50 100644
> --- a/drivers/net/intel/i40e/base/i40e_osdep.h
> +++ b/drivers/net/intel/i40e/base/i40e_osdep.h
> @@ -227,12 +227,11 @@ struct i40e_spinlock {
>  #define i40e_memset(a, b, c, d) memset((a), (b), (c))
>  #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
>  
> -#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
>  #define DELAY(x) rte_delay_us_sleep(x)
>  #define i40e_usec_delay(x) DELAY(x)
>  #define i40e_msec_delay(x) DELAY(1000 * (x))
>  #define udelay(x) DELAY(x)
>  #define msleep(x) DELAY(1000*(x))
> -#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
> +#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
>  
>  #endif /* _I40E_OSDEP_H_ */
> -- 
> 2.55.0.1082.g2b9226bbc0-goog
> 

Re: [PATCH v3 03/16] net/idpf: use common division round up macro

[Bruce Richardson]

On Tue, Sep 22, 2026 at 06:07:26PM -0700, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

>  drivers/net/intel/idpf/base/idpf_osdep.h | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/net/intel/idpf/base/idpf_osdep.h b/drivers/net/intel/idpf/base/idpf_osdep.h
> index 67bc33f2e6..64bbd00eea 100644
> --- a/drivers/net/intel/idpf/base/idpf_osdep.h
> +++ b/drivers/net/intel/idpf/base/idpf_osdep.h
> @@ -291,13 +291,12 @@ idpf_hweight32(u32 num)
>  	return bits;
>  }
>  
> -#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
>  #define DELAY(x) rte_delay_us(x)
>  #define idpf_usec_delay(x) rte_delay_us(x)
>  #define idpf_msec_delay(x, y) rte_delay_us(1000 * (x))
>  #define udelay(x) DELAY(x)
>  #define msleep(x) DELAY(1000 * (x))
> -#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
> +#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
>  
>  #ifndef IDPF_DBG_TRACE
>  #define IDPF_DBG_TRACE	  BIT_ULL(0)
> -- 
> 2.55.0.1082.g2b9226bbc0-goog
> 

Re: [PATCH v3 04/16] net/ice: use common division round up macro

[Bruce Richardson]

On Tue, Sep 22, 2026 at 06:07:27PM -0700, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
>  drivers/net/intel/ice/base/ice_osdep.h | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/net/intel/ice/base/ice_osdep.h b/drivers/net/intel/ice/base/ice_osdep.h
> index 5d92cf5aa4..3ef52eae6d 100644
> --- a/drivers/net/intel/ice/base/ice_osdep.h
> +++ b/drivers/net/intel/ice/base/ice_osdep.h
> @@ -311,13 +311,12 @@ ice_hweight32(u32 num)
>  	return bits;
>  }
>  
> -#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
>  #define DELAY(x) rte_delay_us(x)
>  #define ice_usec_delay(x, y) rte_delay_us(x)
>  #define ice_msec_delay(x, y) rte_delay_us(1000 * (x))
>  #define udelay(x) DELAY(x)
>  #define msleep(x) DELAY(1000 * (x))
> -#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
> +#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
>  
>  struct ice_list_entry {
>  	LIST_ENTRY(ice_list_entry) next;
> -- 
> 2.55.0.1082.g2b9226bbc0-goog
> 

Re: [PATCH v3 02/16] net/i40e: use common division round up macro

[Bruce Richardson]

On Wed, Sep 23, 2026 at 08:30:18AM +0100, Bruce Richardson wrote:
> On Tue, Sep 22, 2026 at 06:07:25PM -0700, Joshua Washington wrote:
> > The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> > 
> > Signed-off-by: Joshua Washington <joshwash@google.com>
> > Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> > ---
> >  drivers/net/intel/i40e/base/i40e_osdep.h | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> > 

FYI: Just a note for i40e and the other intel drivers: the osdep.h files
have moved in the repo in next-net-intel tree, so this patch will conflict
once those changes are merged to main. Therefore, I'd suggest not merging
this till after the pull, and then adjusting the paths in the patch on
apply for simplicity of merge.

Thanks,
/Bruce

> > diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h b/drivers/net/intel/i40e/base/i40e_osdep.h
> > index 9c5a94f055..93620f5c50 100644
> > --- a/drivers/net/intel/i40e/base/i40e_osdep.h
> > +++ b/drivers/net/intel/i40e/base/i40e_osdep.h
> > @@ -227,12 +227,11 @@ struct i40e_spinlock {
> >  #define i40e_memset(a, b, c, d) memset((a), (b), (c))
> >  #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
> >  
> > -#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
> >  #define DELAY(x) rte_delay_us_sleep(x)
> >  #define i40e_usec_delay(x) DELAY(x)
> >  #define i40e_msec_delay(x) DELAY(1000 * (x))
> >  #define udelay(x) DELAY(x)
> >  #define msleep(x) DELAY(1000*(x))
> > -#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
> > +#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
> >  
> >  #endif /* _I40E_OSDEP_H_ */
> > -- 
> > 2.55.0.1082.g2b9226bbc0-goog
> > 

Re: [PATCH v3 15/16] app/test: use common division round up macro

[Bruce Richardson]

On Tue, Sep 22, 2026 at 06:07:38PM -0700, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

> ---
> Changes in v3:
>   - New patch converting DIV_CEIL in test_compressdev.c.
> ---
>  app/test/test_compressdev.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/app/test/test_compressdev.c b/app/test/test_compressdev.c
> index 293d1d757f..2891b98d56 100644
> --- a/app/test/test_compressdev.c
> +++ b/app/test/test_compressdev.c
> @@ -18,8 +18,6 @@
>  #include "test_compressdev_test_buffer.h"
>  #include "test.h"
>  
> -#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
> -
>  #define DEFAULT_WINDOW_SIZE 15
>  #define DEFAULT_MEM_LEVEL 8
>  #define MAX_DEQD_RETRIES 10
> @@ -688,7 +686,7 @@ prepare_sgl_bufs(const char *test_buf, struct rte_mbuf *head_buf,
>  		uint16_t seg_size)
>  {
>  	uint32_t remaining_data = total_data_size;
> -	uint16_t num_remaining_segs = DIV_CEIL(remaining_data, seg_size);
> +	uint16_t num_remaining_segs = RTE_DIV_ROUND_UP(remaining_data, seg_size);
>  	struct rte_mempool *pool;
>  	struct rte_mbuf *next_seg;
>  	uint32_t data_size;
> -- 
> 2.55.0.1082.g2b9226bbc0-goog
> 

Re: [PATCH v3 16/16] app/test-compress-perf: use common division round up macro

[Bruce Richardson]

On Tue, Sep 22, 2026 at 06:07:39PM -0700, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

> ---
> Changes in v3:
>   - New patch converting DIV_CEIL in comp_perf_test_common.c.
> ---
>  app/test-compress-perf/comp_perf_test_common.c | 13 +++++--------
>  1 file changed, 5 insertions(+), 8 deletions(-)
> 
> diff --git a/app/test-compress-perf/comp_perf_test_common.c b/app/test-compress-perf/comp_perf_test_common.c
> index 78487196ad..3f70c222c4 100644
> --- a/app/test-compress-perf/comp_perf_test_common.c
> +++ b/app/test-compress-perf/comp_perf_test_common.c
> @@ -14,9 +14,6 @@
>  #include "comp_perf_test_common.h"
>  #include "comp_perf_test_verify.h"
>  
> -
> -#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
> -
>  struct cperf_buffer_info {
>  	uint16_t total_segments;
>  	uint16_t segment_sz;
> @@ -238,8 +235,8 @@ comp_perf_allocate_memory(struct comp_test_data *test_data,
>  		 * Number of segments for input and output
>  		 * (compression and decompression)
>  		 */
> -		test_data->total_segs = DIV_CEIL(test_data->input_data_sz,
> -						 test_data->seg_sz);
> +		test_data->total_segs = RTE_DIV_ROUND_UP(test_data->input_data_sz,
> +							 test_data->seg_sz);
>  	} else {
>  		/*
>  		 * When application does decompression only, input data is
> @@ -292,8 +289,8 @@ comp_perf_allocate_memory(struct comp_test_data *test_data,
>  		return -1;
>  	}
>  
> -	mem->total_bufs = DIV_CEIL(test_data->total_segs,
> -				   test_data->max_sgl_segs);
> +	mem->total_bufs = RTE_DIV_ROUND_UP(test_data->total_segs,
> +					   test_data->max_sgl_segs);
>  
>  	snprintf(pool_name, sizeof(pool_name), "op_pool_%u_qp_%u",
>  			mem->dev_id, mem->qp_id);
> @@ -557,7 +554,7 @@ prepare_bufs(struct comp_test_data *test_data, struct cperf_mem_resources *mem)
>  void
>  print_test_dynamics(const struct comp_test_data *test_data)
>  {
> -	uint32_t opt_total_segs = DIV_CEIL(buffer_info.input_data_sz,
> +	uint32_t opt_total_segs = RTE_DIV_ROUND_UP(buffer_info.input_data_sz,
>  			MAX_SEG_SIZE);
>  
>  	if (buffer_info.total_buffs > 1) {
> -- 
> 2.55.0.1082.g2b9226bbc0-goog
> 

Re: [PATCH v3 00/16] Introduce RTE_DIV_ROUND_UP

[Stephen Hemminger]

On Tue, 22 Sep 2026 18:07:23 -0700
Joshua Washington <joshwash@google.com> wrote:

> Numerous components independently define the DIV_ROUND_UP macro, using
> the same definition. Attempt to commonize by introducing
> RTE_DIV_ROUND_UP to rte_common.h and replacing usages of DIV_ROUND_UP
> with RTE_DIV_ROUND_UP. Similarly, replace DIV_CEIL, which is
> functionally identical, to RTE_DIV_ROUND_UP.
> 
> Changes in v3:
>   - eal: Add unit tests in app/test/test_common.c.
>   - eal: Add release notes entry for RTE_DIV_ROUND_UP.
>   - eal: Expand Doxygen documentation with parameters and caveats
>     (divisor evaluation, non-negative inputs, wraparound).
>   - base code (qede, cxgbe, ena): Keep DIV_ROUND_UP macro in osdep/platform
>     headers and avoid modifying call sites under base/.
>   - nfp: Change commit subject prefix from "drivers:" to "net/nfp:".
>   - qede, nfp: Drop unnecessary line rewraps.
>   - ena: Remove extraneous blank line in ena_plat_dpdk.h.
>   - app: Convert DIV_CEIL in app/test and app/test-compress-perf to
>     RTE_DIV_ROUND_UP.
> 
> Changes in v2:
>   - Split single patch into series.
> 
> Joshua Washington (16):
>   eal: introduce division round up macro
>   net/i40e: use common division round up macro
>   net/idpf: use common division round up macro
>   net/ice: use common division round up macro
>   net/qede: use common division round up macro
>   net/rnp: use common division round up macro
>   net/hns3: use common division round up macro
>   net/hinic: use common division round up macro
>   net/bnx2x: use common division round up macro
>   net/cxgbe: use common division round up macro
>   net/ena: use common division round up macro
>   raw/ifpga/base: use common division round up macro
>   net/nfp: use common division round up macro
>   app/procinfo: use common division round up macro
>   app/test: use common division round up macro
>   app/test-compress-perf: use common division round up macro
> 
>  app/proc-info/main.c                           |  3 +--
>  app/test-compress-perf/comp_perf_test_common.c | 13 +++++--------
>  app/test/test_common.c                         | 10 ++++++++++
>  app/test/test_compressdev.c                    |  4 +---
>  doc/guides/rel_notes/release_26_11.rst         |  5 +++++
>  drivers/common/nfp/nfp_platform.h              |  2 --
>  drivers/net/bnx2x/bnx2x.c                      |  4 ++--
>  drivers/net/bnx2x/bnx2x.h                      | 11 ++++-------
>  drivers/net/cxgbe/cxgbe_compat.h               |  2 +-
>  drivers/net/cxgbe/cxgbe_filter.c               |  6 +++---
>  drivers/net/cxgbe/cxgbe_ofld.h                 |  4 ++--
>  drivers/net/cxgbe/sge.c                        | 14 +++++++-------
>  drivers/net/ena/base/ena_plat_dpdk.h           |  2 +-
>  drivers/net/hinic/base/hinic_compat.h          |  5 +----
>  drivers/net/hns3/hns3_ethdev.h                 |  2 --
>  drivers/net/hns3/hns3_fdir.c                   |  2 +-
>  drivers/net/hns3/hns3_regs.c                   |  8 ++++----
>  drivers/net/hns3/hns3_rss.c                    |  8 ++++----
>  drivers/net/intel/i40e/base/i40e_osdep.h       |  3 +--
>  drivers/net/intel/ice/base/ice_osdep.h         |  3 +--
>  drivers/net/intel/idpf/base/idpf_osdep.h       |  3 +--
>  drivers/net/nfp/nfdk/nfp_nfdk.h                |  2 +-
>  drivers/net/nfp/nfdk/nfp_nfdk_dp.c             |  2 +-
>  drivers/net/nfp/nfpcore/nfp_nsp.c              |  2 +-
>  drivers/net/qede/base/bcm_osal.h               |  1 -
>  drivers/net/qede/qede_debug.c                  | 16 ++++++++--------
>  drivers/net/rnp/base/rnp_osdep.h               |  5 ++---
>  drivers/net/rnp/rnp_rxtx.c                     |  2 +-
>  drivers/raw/ifpga/base/opae_osdep.h            |  3 +--
>  lib/eal/include/rte_common.h                   | 18 ++++++++++++++++++
>  30 files changed, 88 insertions(+), 77 deletions(-)
> 

Since the next-net-intel tree is now merged in main this will need
a rebase. Detailed AI review also found other issues:

Review: [PATCH v3 01/16..16/16] common division round up macro

Series
------

Error

Patches 02-04 do not apply to main. The osdep headers moved out of
base/ in:

  968a7e52ef50 ("net/i40e: move editable files out of base folder")
  2eb2950a8e17 ("net/ice: move editable files out of base folder")
  7b8be7b1c614 ("net/idpf: move editable files out of base folder")

Patch 01 also conflicts in release_26_11.rst. Needs a rebase; the
changes carry over as is to drivers/net/intel/{i40e,ice,idpf}/.
With paths adjusted, every commit builds with -Dwerror=true for all
touched drivers and apps, and common_autotest passes.

Warning

Patches 05, 10 and 11 say DIV_ROUND_UP is now redundant, but v3
keeps it in bcm_osal.h, cxgbe_compat.h and ena_plat_dpdk.h as an
alias for base/ code. Update the commit bodies to match, e.g.:

  Map DIV_ROUND_UP to RTE_DIV_ROUND_UP for base code and use
  RTE_DIV_ROUND_UP directly elsewhere in the driver.

For 05, what actually goes away is the local RTE_DIV_ROUND_UP.

Info

Private copies not covered by the series, for a follow-up:

  drivers/crypto/ccp/ccp_dev.h              CCP_DIV_ROUND_UP
  drivers/common/qat/dev/qat_dev_gen_lce.c  LCE_DIV_ROUND_UP (unused)
  drivers/common/sxe2/sxe2_osal.h           SXE2_DIV_ROUND_UP
  examples/ip_fragmentation/main.c          ROUNDUP_DIV (unused)


Patch 01/16 eal: introduce division round up macro
--------------------------------------------------

Warning

The Doxygen block contradicts itself on the divisor:

  + *   Divisor. Must be a positive integer. Evaluated twice.
  + * @note Both @p n and @p d must be non-negative to round correctly.

d == 0 is non-negative. The first two @note lines only repeat the
@param text; drop them. Also @c formats a single word, so
"@c (n) + (d) - 1" renders only "(n)" as code. Suggest:

  /**
   * Divide @p n by @p d, rounding up.
   *
   * @param n
   *   Dividend. Must be non-negative.
   * @param d
   *   Divisor. Must be positive. Evaluated twice.
   * @return
   *   @p n divided by @p d, rounded up.
   *
   * @note (n) + (d) - 1 wraps if @p n is near the maximum of its
   *   type, e.g. RTE_DIV_ROUND_UP(UINT32_MAX, 2) is 0.
   */


Patch 03/16 net/idpf: use common division round up macro
--------------------------------------------------------

Info

Pre-existing, not introduced by this patch. After the rebase
idpf_osdep.h still carries a second copy, with no users in the tree:

  #define DIVIDE_AND_ROUND_UP(a, b) (((a) + (b) - 1) / (b))

Drop it or alias it to RTE_DIV_ROUND_UP.


Patch 04/16 net/ice: use common division round up macro
-------------------------------------------------------

Info

Pre-existing, not introduced by this patch. ice_osdep.h has the same
DIVIDE_AND_ROUND_UP copy, used by base code (ice_acl_ctrl.c,
ice_flow.c, ice_sched.c). Now that the file is outside base/, map it
to RTE_DIV_ROUND_UP as done for cxgbe and ena.

Re: [PATCH v3 01/16] eal: introduce division round up macro

[fengchengwen]

Acked-by: Chengwen Feng <fengchengwen@huawei.com>

On 9/23/2026 9:07 AM, Joshua Washington wrote:
> A number of drivers define the DIV_ROUND_UP macro -- add it to
> rte_common.h so that other code looking to use that functionality
> doesn't have to re-define it.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>

Re: [PATCH v3 07/16] net/hns3: use common division round up macro

[fengchengwen]

Acked-by: Chengwen Feng <fengchengwen@huawei.com>

On 9/23/2026 9:07 AM, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
>  drivers/net/hns3/hns3_ethdev.h | 2 --
>  drivers/net/hns3/hns3_fdir.c   | 2 +-
>  drivers/net/hns3/hns3_regs.c   | 8 ++++----
>  drivers/net/hns3/hns3_rss.c    | 8 ++++----
>  4 files changed, 9 insertions(+), 11 deletions(-)

Re: [PATCH v3 14/16] app/procinfo: use common division round up macro

[fengchengwen]

Acked-by: Chengwen Feng <fengchengwen@huawei.com>

On 9/23/2026 9:07 AM, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>
> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH v3 15/16] app/test: use common division round up macro

[fengchengwen]

Acked-by: Chengwen Feng <fengchengwen@huawei.com>

On 9/23/2026 9:07 AM, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>

Re: [PATCH v3 16/16] app/test-compress-perf: use common division round up macro

[fengchengwen]

Acked-by: Chengwen Feng <fengchengwen@huawei.com>

On 9/23/2026 9:07 AM, Joshua Washington wrote:
> The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.
> 
> Signed-off-by: Joshua Washington <joshwash@google.com>

Re: [PATCH v3 02/16] net/i40e: use common division round up macro

[Joshua Washington]

[-- Attachment #1: Type: text/plain, Size: 2033 bytes --]

On Wed, Sep 23, 2026, 00:35 Bruce Richardson <bruce.richardson@intel.com>
wrote:

> On Wed, Sep 23, 2026 at 08:30:18AM +0100, Bruce Richardson wrote:
> > On Tue, Sep 22, 2026 at 06:07:25PM -0700, Joshua Washington wrote:
> > > The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.
> > >
> > > Signed-off-by: Joshua Washington <joshwash@google.com>
> > > Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
> > Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> > > ---
> > >  drivers/net/intel/i40e/base/i40e_osdep.h | 3 +--
> > >  1 file changed, 1 insertion(+), 2 deletions(-)
> > >
>
> FYI: Just a note for i40e and the other intel drivers: the osdep.h files
> have moved in the repo in next-net-intel tree, so this patch will conflict
> once those changes are merged to main. Therefore, I'd suggest not merging
> this till after the pull, and then adjusting the paths in the patch on
> apply for simplicity of merge.
>
> Thanks,
> /Bruce
>
> > > diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h
> b/drivers/net/intel/i40e/base/i40e_osdep.h
> > > index 9c5a94f055..93620f5c50 100644
> > > --- a/drivers/net/intel/i40e/base/i40e_osdep.h
> > > +++ b/drivers/net/intel/i40e/base/i40e_osdep.h
> > > @@ -227,12 +227,11 @@ struct i40e_spinlock {
> > >  #define i40e_memset(a, b, c, d) memset((a), (b), (c))
> > >  #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
> > >
> > > -#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
> > >  #define DELAY(x) rte_delay_us_sleep(x)
> > >  #define i40e_usec_delay(x) DELAY(x)
> > >  #define i40e_msec_delay(x) DELAY(1000 * (x))
> > >  #define udelay(x) DELAY(x)
> > >  #define msleep(x) DELAY(1000*(x))
> > > -#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
> > > +#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
> > >
> > >  #endif /* _I40E_OSDEP_H_ */
> > > --
> > > 2.55.0.1082.g2b9226bbc0-goog
> > >


Ah, I missed this. Thanks for the update. I will hold off on a v3 if it is
necessary until the main branch is updated.

[-- Attachment #2: Type: text/html, Size: 3157 bytes --]

Re: [PATCH v3 02/16] net/i40e: use common division round up macro

[Bruce Richardson]

On Tue, Sep 29, 2026 at 03:44:02PM -0700, Joshua Washington wrote:
> 
>    On Wed, Sep 23, 2026, 00:35 Bruce Richardson
>    <[1]bruce.richardson@intel.com> wrote:
> 
>      On Wed, Sep 23, 2026 at 08:30:18AM +0100, Bruce Richardson wrote:
>      > On Tue, Sep 22, 2026 at 06:07:25PM -0700, Joshua Washington wrote:
>      > > The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP
>      redundant.
>      > >
>      > > Signed-off-by: Joshua Washington <[2]joshwash@google.com>
>      > > Reviewed-by: Stephen Hemminger <[3]stephen@networkplumber.org>
>      > Acked-by: Bruce Richardson <[4]bruce.richardson@intel.com>
>      > > ---
>      > >  drivers/net/intel/i40e/base/i40e_osdep.h | 3 +--
>      > >  1 file changed, 1 insertion(+), 2 deletions(-)
>      > >
>      FYI: Just a note for i40e and the other intel drivers: the osdep.h
>      files
>      have moved in the repo in next-net-intel tree, so this patch will
>      conflict
>      once those changes are merged to main. Therefore, I'd suggest not
>      merging
>      this till after the pull, and then adjusting the paths in the patch
>      on
>      apply for simplicity of merge.
>      Thanks,
>      /Bruce
>      > > diff --git a/drivers/net/intel/i40e/base/i40e_osdep.h
>      b/drivers/net/intel/i40e/base/i40e_osdep.h
>      > > index 9c5a94f055..93620f5c50 100644
>      > > --- a/drivers/net/intel/i40e/base/i40e_osdep.h
>      > > +++ b/drivers/net/intel/i40e/base/i40e_osdep.h
>      > > @@ -227,12 +227,11 @@ struct i40e_spinlock {
>      > >  #define i40e_memset(a, b, c, d) memset((a), (b), (c))
>      > >  #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
>      > >
>      > > -#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
>      > >  #define DELAY(x) rte_delay_us_sleep(x)
>      > >  #define i40e_usec_delay(x) DELAY(x)
>      > >  #define i40e_msec_delay(x) DELAY(1000 * (x))
>      > >  #define udelay(x) DELAY(x)
>      > >  #define msleep(x) DELAY(1000*(x))
>      > > -#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
>      > > +#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min,
>      1000))
>      > >
>      > >  #endif /* _I40E_OSDEP_H_ */
>      > > --
>      > > 2.55.0.1082.g2b9226bbc0-goog
>      > >
> 
>    Ah, I missed this. Thanks for the update. I will hold off on a v3 if it
>    is necessary until the main branch is updated.
> 
The next-net-intel tree has been pulled, so the move of the osdep files is
completed now. Maybe respin this patch so it applies cleanly again?

[PATCH v4 00/15] Introduce RTE_DIV_ROUND_UP

[Joshua Washington]

Numerous components independently define the DIV_ROUND_UP macro, using
the same definition. Attempt to commonize by introducing
RTE_DIV_ROUND_UP to rte_common.h and replacing usages of DIV_ROUND_UP
with RTE_DIV_ROUND_UP. Similarly, replace DIV_CEIL, which is
functionally identical, to RTE_DIV_ROUND_UP.

Changes in v4:
  - eal: Fix merge conflict in release notes after rebase.
  - various commits: add rte_common.h includes where missing to fix
    compilation issues.
  - opae: drop RTE_DIV_ROUND_UP change due to shered osdep file for RTE
    and raw drivers.

Changes in v3:
  - eal: Add unit tests in app/test/test_common.c.
  - eal: Add release notes entry for RTE_DIV_ROUND_UP.
  - eal: Expand Doxygen documentation with parameters and caveats
    (divisor evaluation, non-negative inputs, wraparound).
  - base code (qede, cxgbe, ena): Keep DIV_ROUND_UP macro in osdep/platform
    headers and avoid modifying call sites under base/.
  - nfp: Change commit subject prefix from "drivers:" to "net/nfp:".
  - qede, nfp: Drop unnecessary line rewraps.
  - ena: Remove extraneous blank line in ena_plat_dpdk.h.
  - app: Convert DIV_CEIL in app/test and app/test-compress-perf to
    RTE_DIV_ROUND_UP.

Changes in v2:
  - Split single patch into series.

Joshua Washington (15):
  eal: introduce division round up macro
  net/i40e: use common division round up macro
  net/idpf: use common division round up macro
  net/ice: use common division round up macro
  net/qede: use common division round up macro
  net/rnp: use common division round up macro
  net/hns3: use common division round up macro
  net/hinic: use common division round up macro
  net/bnx2x: use common division round up macro
  net/cxgbe: use common division round up macro
  net/ena: use common division round up macro
  net/nfp: use common division round up macro
  app/procinfo: use common division round up macro
  app/test: use common division round up macro
  app/test-compress-perf: use common division round up macro

 app/proc-info/main.c                           |  3 +--
 app/test-compress-perf/comp_perf_test_common.c | 14 ++++++--------
 app/test/test_common.c                         | 10 ++++++++++
 app/test/test_compressdev.c                    |  5 ++---
 doc/guides/rel_notes/release_26_11.rst         |  5 +++++
 drivers/common/nfp/nfp_platform.h              |  2 --
 drivers/net/bnx2x/bnx2x.c                      |  4 ++--
 drivers/net/bnx2x/bnx2x.h                      | 12 +++++-------
 drivers/net/cxgbe/cxgbe_compat.h               |  2 +-
 drivers/net/cxgbe/cxgbe_filter.c               |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h                 |  4 ++--
 drivers/net/cxgbe/sge.c                        | 14 +++++++-------
 drivers/net/ena/base/ena_plat_dpdk.h           |  3 ++-
 drivers/net/hinic/base/hinic_compat.h          |  5 +----
 drivers/net/hns3/hns3_ethdev.h                 |  2 --
 drivers/net/hns3/hns3_fdir.c                   |  3 ++-
 drivers/net/hns3/hns3_regs.c                   |  9 +++++----
 drivers/net/hns3/hns3_rss.c                    |  9 +++++----
 drivers/net/intel/i40e/i40e_osdep.h            |  3 +--
 drivers/net/intel/ice/ice_osdep.h              |  3 +--
 drivers/net/intel/idpf/idpf_osdep.h            |  3 +--
 drivers/net/nfp/nfdk/nfp_nfdk.h                |  3 ++-
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c             |  2 +-
 drivers/net/nfp/nfpcore/nfp_nsp.c              |  2 +-
 drivers/net/qede/base/bcm_osal.h               |  2 +-
 drivers/net/qede/qede_debug.c                  | 16 ++++++++--------
 drivers/net/rnp/base/rnp_osdep.h               |  5 ++---
 drivers/net/rnp/rnp_rxtx.c                     |  2 +-
 lib/eal/include/rte_common.h                   | 18 ++++++++++++++++++
 29 files changed, 96 insertions(+), 75 deletions(-)

-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 01/15] eal: introduce division round up macro

[Joshua Washington]

A number of drivers define the DIV_ROUND_UP macro -- add it to
rte_common.h so that other code looking to use that functionality
doesn't have to re-define it.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>
---
Changes in v4:
  - Rebased changes
Changes in v3:
  - Add unit tests in app/test/test_common.c.
  - Add release notes entry.
  - Enhance Doxygen documentation to document parameters and caveats
    (divisor evaluation, non-negative inputs, wraparound).
Changes in v2:
  - Split out patches replacing usages
---
 app/test/test_common.c                 | 10 ++++++++++
 doc/guides/rel_notes/release_26_11.rst |  5 +++++
 lib/eal/include/rte_common.h           | 18 ++++++++++++++++++
 3 files changed, 33 insertions(+)

diff --git a/app/test/test_common.c b/app/test/test_common.c
index 3e1c7df0c1..bce2776ea6 100644
--- a/app/test/test_common.c
+++ b/app/test/test_common.c
@@ -47,6 +47,16 @@ test_macros(int __rte_unused unused_parm)
 		"RTE_MAX");
 	RTE_TEST_ASSERT_EQUAL(RTE_MIN(SMALLER, BIGGER), SMALLER,
 		"RTE_MIN");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(10, 5), 2,
+		"RTE_DIV_ROUND_UP exact");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(11, 5), 3,
+		"RTE_DIV_ROUND_UP remainder");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(0, 5), 0,
+		"RTE_DIV_ROUND_UP n=0");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(0, 1), 0,
+		"RTE_DIV_ROUND_UP n=0, d=1");
+	RTE_TEST_ASSERT_EQUAL(RTE_DIV_ROUND_UP(5, 1), 5,
+		"RTE_DIV_ROUND_UP d=1");
 
 	RTE_TEST_ASSERT_EQUAL(RTE_PTR_ADD(arr + 1, sizeof(arr[0])), &arr[2],
 		"RTE_PTR_ADD(expr, x)");
diff --git a/doc/guides/rel_notes/release_26_11.rst b/doc/guides/rel_notes/release_26_11.rst
index 030bd84cea..430e5ab1c1 100644
--- a/doc/guides/rel_notes/release_26_11.rst
+++ b/doc/guides/rel_notes/release_26_11.rst
@@ -138,6 +138,11 @@ New Features
   Added ``rte_bbdev_queue_stats_get()`` function to retrieve statistics
   for a specific queue, complementing the existing device-level statistics API.
 
+* **Added division round up macro.**
+
+  Added the ``RTE_DIV_ROUND_UP`` macro in ``rte_common.h`` to divide dividend
+  by divisor, rounding up to the nearest integer.
+
 
 Removed Items
 -------------
diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
index f872d3eabb..f3ff632713 100644
--- a/lib/eal/include/rte_common.h
+++ b/lib/eal/include/rte_common.h
@@ -704,6 +704,24 @@ static void __attribute__((destructor(RTE_PRIO(prio)), used)) func(void)
 		(ceil - (v)) > ((v) - floor) ? floor : ceil;	\
 	})
 
+/**
+ * Macro to divide dividend @p n by divisor @p d, rounding up if @p d does not divide @p n.
+ *
+ * @param n
+ *   Dividend. Must be a non-negative integer to round correctly.
+ * @param d
+ *   Divisor. Must be a positive integer. Evaluated twice.
+ * @return
+ *   Result of dividing @p n by @p d, rounded up.
+ *
+ * @note The divisor @p d is evaluated twice. Arguments must not have side effects.
+ * @note Both @p n and @p d must be non-negative to round correctly.
+ * @note The sum @c (n) + (d) - 1 can wrap around if @p n is close to the maximum
+ *       representable value of its type (e.g. RTE_DIV_ROUND_UP(UINT32_MAX, 2)
+ *       on uint32_t wraps to 0).
+ */
+#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+
 /**
  * Checks if a pointer is aligned to a given power-of-two value
  *
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 02/15] net/i40e: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/intel/i40e/i40e_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/i40e/i40e_osdep.h b/drivers/net/intel/i40e/i40e_osdep.h
index 1139c9c1fb..e499ae1d11 100644
--- a/drivers/net/intel/i40e/i40e_osdep.h
+++ b/drivers/net/intel/i40e/i40e_osdep.h
@@ -229,12 +229,11 @@ struct i40e_spinlock {
 #define i40e_memset(a, b, c, d) memset((a), (b), (c))
 #define i40e_memcpy(a, b, c, d) memcpy((a), (b), (c))
 
-#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us_sleep(x)
 #define i40e_usec_delay(x) DELAY(x)
 #define i40e_msec_delay(x) DELAY(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000*(x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #endif /* _I40E_OSDEP_H_ */
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 03/15] net/idpf: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/intel/idpf/idpf_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/idpf/idpf_osdep.h b/drivers/net/intel/idpf/idpf_osdep.h
index 9c758fb3c0..7d99a23cf8 100644
--- a/drivers/net/intel/idpf/idpf_osdep.h
+++ b/drivers/net/intel/idpf/idpf_osdep.h
@@ -293,13 +293,12 @@ idpf_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define idpf_usec_delay(x) rte_delay_us(x)
 #define idpf_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 #ifndef IDPF_DBG_TRACE
 #define IDPF_DBG_TRACE	  BIT_ULL(0)
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 04/15] net/ice: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/intel/ice/ice_osdep.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/intel/ice/ice_osdep.h b/drivers/net/intel/ice/ice_osdep.h
index 599bd4086b..f4ec8b8994 100644
--- a/drivers/net/intel/ice/ice_osdep.h
+++ b/drivers/net/intel/ice/ice_osdep.h
@@ -313,13 +313,12 @@ ice_hweight32(u32 num)
 	return bits;
 }
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DELAY(x) rte_delay_us(x)
 #define ice_usec_delay(x, y) rte_delay_us(x)
 #define ice_msec_delay(x, y) rte_delay_us(1000 * (x))
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
-#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+#define usleep_range(min, max) msleep(RTE_DIV_ROUND_UP(min, 1000))
 
 struct ice_list_entry {
 	LIST_ENTRY(ice_list_entry) next;
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 05/15] net/qede: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP and
RTE_DIV_ROUND_UP definitions redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v4:
  - Add rte_common.h include.
Changes in v3:
  - Leave call sites in base/ alone and map DIV_ROUND_UP to
    RTE_DIV_ROUND_UP in bcm_osal.h.
  - Drop line rewraps.
---
 drivers/net/qede/base/bcm_osal.h |  2 +-
 drivers/net/qede/qede_debug.c    | 16 ++++++++--------
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/net/qede/base/bcm_osal.h b/drivers/net/qede/base/bcm_osal.h
index 357981f63d..7bcd2d1e32 100644
--- a/drivers/net/qede/base/bcm_osal.h
+++ b/drivers/net/qede/base/bcm_osal.h
@@ -11,6 +11,7 @@
 #include <pthread.h>
 #include <time.h>
 
+#include <rte_common.h>
 #include <rte_bitops.h>
 #include <rte_branch_prediction.h>
 #include <rte_byteorder.h>
@@ -396,7 +397,6 @@ int qede_save_fw_dump(uint16_t port_id);
 
 /* Utility functions */
 
-#define RTE_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define DIV_ROUND_UP(size, to_what) RTE_DIV_ROUND_UP(size, to_what)
 #define RTE_ROUNDUP(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define ROUNDUP(value, to_what) RTE_ROUNDUP((value), (to_what))
diff --git a/drivers/net/qede/qede_debug.c b/drivers/net/qede/qede_debug.c
index d14ef93a3c..9e1f19ad96 100644
--- a/drivers/net/qede/qede_debug.c
+++ b/drivers/net/qede/qede_debug.c
@@ -332,7 +332,7 @@ struct split_type_defs {
 
 #define BYTES_IN_DWORD			sizeof(u32)
 /* In the macros below, size and offset are specified in bits */
-#define CEIL_DWORDS(size)		DIV_ROUND_UP(size, 32)
+#define CEIL_DWORDS(size)		RTE_DIV_ROUND_UP(size, 32)
 #define FIELD_BIT_OFFSET(type, field)	type ## _ ## field ## _ ## OFFSET
 #define FIELD_BIT_SIZE(type, field)	type ## _ ## field ## _ ## SIZE
 #define FIELD_DWORD_OFFSET(type, field) \
@@ -3053,7 +3053,7 @@ static u32 qed_grc_dump_big_ram(struct ecore_hwfn *p_hwfn,
 		return offset + ram_size;
 
 	/* Dump Big RAM */
-	for (i = 0; i < DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
+	for (i = 0; i < RTE_DIV_ROUND_UP(ram_size, BRB_REG_BIG_RAM_DATA_SIZE);
 	     i++) {
 		u32 addr, len;
 
@@ -4154,8 +4154,8 @@ static enum dbg_status qed_mcp_trace_dump(struct ecore_hwfn *p_hwfn,
 
 	/* Find trace data size */
 	trace_data_size_dwords =
-	    DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
-			 BYTES_IN_DWORD);
+	    RTE_DIV_ROUND_UP(trace_data_size_bytes + sizeof(struct mcp_trace),
+			     BYTES_IN_DWORD);
 
 	/* Dump trace data section header and param */
 	offset += qed_dump_section_hdr(dump_buf + offset,
@@ -4810,10 +4810,10 @@ static u32 qed_ilt_dump(struct ecore_hwfn *p_hwfn,
 		offset += num_pages * PAGE_MEM_DESC_SIZE_DWORDS;
 	}
 
-	valid_conn_pf_pages = DIV_ROUND_UP(valid_conn_pf_cids,
-					   num_cids_per_page);
-	valid_conn_vf_pages = DIV_ROUND_UP(valid_conn_vf_cids,
-					   num_cids_per_page);
+	valid_conn_pf_pages = RTE_DIV_ROUND_UP(valid_conn_pf_cids,
+					       num_cids_per_page);
+	valid_conn_vf_pages = RTE_DIV_ROUND_UP(valid_conn_vf_cids,
+					       num_cids_per_page);
 
 	/* Dump ILT pages IDs */
 	offset += qed_ilt_dump_pages_section(p_hwfn,
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 06/15] net/rnp: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Base BITS_PER_BYTE and BITS_TO_LONGS definitions on whether
BITS_TO_LONGS is defined instead.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/rnp/base/rnp_osdep.h | 5 ++---
 drivers/net/rnp/rnp_rxtx.c       | 2 +-
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/net/rnp/base/rnp_osdep.h b/drivers/net/rnp/base/rnp_osdep.h
index a95a37f2fe..c03de79e7e 100644
--- a/drivers/net/rnp/base/rnp_osdep.h
+++ b/drivers/net/rnp/base/rnp_osdep.h
@@ -53,10 +53,9 @@ typedef uint64_t u64;
 #define cpu_to_le32(v)	rte_cpu_to_le_32((u32)(v))
 #endif
 
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d)      (((n) + (d) - 1) / (d))
+#ifndef BITS_TO_LONGS
 #define BITS_PER_BYTE           (8)
-#define BITS_TO_LONGS(nr)       DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
+#define BITS_TO_LONGS(nr)       RTE_DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
 #endif
 
 #define fls(n)	rte_fls_u32(n)
diff --git a/drivers/net/rnp/rnp_rxtx.c b/drivers/net/rnp/rnp_rxtx.c
index 81e8c6ba44..143cd11ffe 100644
--- a/drivers/net/rnp/rnp_rxtx.c
+++ b/drivers/net/rnp/rnp_rxtx.c
@@ -1165,7 +1165,7 @@ rnp_calc_pkt_desc(struct rte_mbuf *tx_pkt)
 	uint16_t count = 0;
 
 	while (txd != NULL) {
-		count += DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
+		count += RTE_DIV_ROUND_UP(txd->data_len, RNP_MAX_TSO_SEG_LEN);
 		txd = txd->next;
 	}
 
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 07/15] net/hns3: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
Changes in v4:
  - Add rte_common.h includes.
---
 drivers/net/hns3/hns3_ethdev.h | 2 --
 drivers/net/hns3/hns3_fdir.c   | 3 ++-
 drivers/net/hns3/hns3_regs.c   | 9 +++++----
 drivers/net/hns3/hns3_rss.c    | 9 +++++----
 4 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/drivers/net/hns3/hns3_ethdev.h b/drivers/net/hns3/hns3_ethdev.h
index 209b042816..b4feff4b94 100644
--- a/drivers/net/hns3/hns3_ethdev.h
+++ b/drivers/net/hns3/hns3_ethdev.h
@@ -978,8 +978,6 @@ static inline struct hns3_vf *HNS3_DEV_HW_TO_VF(struct hns3_hw *hw)
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #define rounddown(x, y) ((x) - ((x) % (y)))
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-
 /*
  * Because hardware always access register in little-endian mode based on hns3
  * network engine, so driver should also call rte_cpu_to_le_32 to convert data
diff --git a/drivers/net/hns3/hns3_fdir.c b/drivers/net/hns3/hns3_fdir.c
index 50572ae430..0096aa21e9 100644
--- a/drivers/net/hns3/hns3_fdir.c
+++ b/drivers/net/hns3/hns3_fdir.c
@@ -7,6 +7,7 @@
 #include <rte_hash_crc.h>
 #include <rte_io.h>
 #include <rte_malloc.h>
+#include <rte_common.h>
 
 #include "hns3_ethdev.h"
 #include "hns3_logs.h"
@@ -129,7 +130,7 @@ static const struct key_info tuple_key_info[] = {
 #define MAX_KEY_LENGTH		400
 #define MAX_200B_KEY_LENGTH	200
 #define MAX_META_DATA_LENGTH	16
-#define MAX_KEY_DWORDS	DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
+#define MAX_KEY_DWORDS	RTE_DIV_ROUND_UP(MAX_KEY_LENGTH / HNS3_BITS_PER_BYTE, 4)
 #define MAX_KEY_BYTES	(MAX_KEY_DWORDS * 4)
 
 enum HNS3_FD_PACKET_TYPE {
diff --git a/drivers/net/hns3/hns3_regs.c b/drivers/net/hns3/hns3_regs.c
index 5c74f9ae2e..2bf0baa5c1 100644
--- a/drivers/net/hns3/hns3_regs.c
+++ b/drivers/net/hns3/hns3_regs.c
@@ -4,6 +4,7 @@
 
 #include <ethdev_pci.h>
 #include <rte_io.h>
+#include <rte_common.h>
 
 #include "hns3_ethdev.h"
 #include "hns3_logs.h"
@@ -1034,8 +1035,8 @@ hns3_get_32_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
-			       HNS3_32_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_32_BIT_DESC_NODATA_LEN,
+				   HNS3_32_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-32bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
@@ -1093,8 +1094,8 @@ hns3_get_64_bit_regs(struct hns3_hw *hw, uint32_t regs_num, struct rte_dev_reg_i
 	if (regs_num == 0)
 		return 0;
 
-	cmd_num = DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
-			       HNS3_64_BIT_REG_RTN_DATANUM);
+	cmd_num = RTE_DIV_ROUND_UP(regs_num + HNS3_64_BIT_DESC_NODATA_LEN,
+				   HNS3_64_BIT_REG_RTN_DATANUM);
 	desc = rte_zmalloc("hns3-64bit-regs",
 			   sizeof(struct hns3_cmd_desc) * cmd_num, 0);
 	if (desc == NULL) {
diff --git a/drivers/net/hns3/hns3_rss.c b/drivers/net/hns3/hns3_rss.c
index 508b3e26d0..080800d60a 100644
--- a/drivers/net/hns3/hns3_rss.c
+++ b/drivers/net/hns3/hns3_rss.c
@@ -5,6 +5,7 @@
 #include <rte_ethdev.h>
 #include <rte_io.h>
 #include <rte_malloc.h>
+#include <rte_common.h>
 
 #include "hns3_ethdev.h"
 #include "hns3_logs.h"
@@ -295,7 +296,7 @@ hns3_rss_set_algo_key(struct hns3_hw *hw, uint8_t hash_algo,
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
 
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  false);
@@ -335,7 +336,7 @@ hns3_rss_get_algo_key(struct hns3_hw *hw,  uint8_t *hash_algo,
 	int ret;
 
 	req = (struct hns3_rss_generic_config_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
+	max_bd_num = RTE_DIV_ROUND_UP(key_len, HNS3_RSS_HASH_KEY_NUM);
 	for (idx = 0; idx < max_bd_num; idx++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_GENERIC_CONFIG,
 					  true);
@@ -381,7 +382,7 @@ hns3_set_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  false);
@@ -431,7 +432,7 @@ hns3_get_rss_indir_table(struct hns3_hw *hw, uint16_t *indir, uint16_t size)
 	int ret;
 
 	req = (struct hns3_rss_indirection_table_cmd *)desc.data;
-	max_bd_num = DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
+	max_bd_num = RTE_DIV_ROUND_UP(size, HNS3_RSS_CFG_TBL_SIZE);
 	for (i = 0; i < max_bd_num; i++) {
 		hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_RSS_INDIR_TABLE,
 					  true);
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 08/15] net/hinic: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/hinic/base/hinic_compat.h | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/net/hinic/base/hinic_compat.h b/drivers/net/hinic/base/hinic_compat.h
index c53b88b96d..c37d970f59 100644
--- a/drivers/net/hinic/base/hinic_compat.h
+++ b/drivers/net/hinic/base/hinic_compat.h
@@ -70,11 +70,8 @@ typedef uint64_t  dma_addr_t;
 #define upper_32_bits(n) ((u32)(((n) >> 16) >> 16))
 #define lower_32_bits(n) ((u32)(n))
 
-/* Returns X / Y, rounding up.  X must be nonnegative to round correctly. */
-#define DIV_ROUND_UP(X, Y) (((X) + ((Y) - 1)) / (Y))
-
 /* Returns X rounded up to the nearest multiple of Y. */
-#define ROUND_UP(X, Y) (DIV_ROUND_UP(X, Y) * (Y))
+#define ROUND_UP(X, Y) (RTE_DIV_ROUND_UP(X, Y) * (Y))
 
 #undef  ALIGN
 #define ALIGN(x, a)  RTE_ALIGN(x, a)
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 09/15] net/bnx2x: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
Changes in v4:
  - Add rte_common.h include.
---
 drivers/net/bnx2x/bnx2x.c |  4 ++--
 drivers/net/bnx2x/bnx2x.h | 12 +++++-------
 2 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/drivers/net/bnx2x/bnx2x.c b/drivers/net/bnx2x/bnx2x.c
index 13b451bae9..2924fed66b 100644
--- a/drivers/net/bnx2x/bnx2x.c
+++ b/drivers/net/bnx2x/bnx2x.c
@@ -2334,8 +2334,8 @@ static void bnx2x_ilt_set_info(struct bnx2x_softc *sc)
 		ilt_client->start = line;
 
 		/* 4 bytes for each cid */
-		line += DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
-				     QM_ILT_PAGE_SZ);
+		line += RTE_DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
+					 QM_ILT_PAGE_SZ);
 
 		ilt_client->end = (line - 1);
 	}
diff --git a/drivers/net/bnx2x/bnx2x.h b/drivers/net/bnx2x/bnx2x.h
index c5de4b71aa..fe375219dd 100644
--- a/drivers/net/bnx2x/bnx2x.h
+++ b/drivers/net/bnx2x/bnx2x.h
@@ -18,6 +18,7 @@
 #include <rte_spinlock.h>
 #include <bus_pci_driver.h>
 #include <rte_io.h>
+#include <rte_common.h>
 
 #include "bnx2x_osal.h"
 #include "bnx2x_ethdev.h"
@@ -37,9 +38,6 @@
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(arr) RTE_DIM(arr)
 #endif
-#ifndef DIV_ROUND_UP
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
-#endif
 #ifndef roundup
 #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
 #endif
@@ -393,7 +391,7 @@ union cdu_context {
 #define CNIC_ISCSI_CID_MAX 256
 #define CNIC_FCOE_CID_MAX  2048
 #define CNIC_CID_MAX       (CNIC_ISCSI_CID_MAX + CNIC_FCOE_CID_MAX)
-#define CNIC_ILT_LINES     DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
+#define CNIC_ILT_LINES     RTE_DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
 
 #define QM_ILT_PAGE_SZ_HW  0
 #define QM_ILT_PAGE_SZ     (4096 << QM_ILT_PAGE_SZ_HW) /* 4K */
@@ -405,7 +403,7 @@ union cdu_context {
 /*#define TM_CONN_NUM        (CNIC_STARTING_CID+CNIC_ISCSI_CXT_MAX) */
 #define TM_CONN_NUM        1024
 #define TM_ILT_SZ          (8 * TM_CONN_NUM)
-#define TM_ILT_LINES       DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
+#define TM_ILT_LINES       RTE_DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
 
 /* SRC (Searcher) host DB constants */
 #define SRC_ILT_PAGE_SZ_HW 0
@@ -414,7 +412,7 @@ union cdu_context {
 #define SRC_CONN_NUM       (1 << SRC_HASH_BITS) /* 1024 */
 #define SRC_ILT_SZ         (sizeof(struct src_ent) * SRC_CONN_NUM)
 #define SRC_T2_SZ          SRC_ILT_SZ
-#define SRC_ILT_LINES      DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
+#define SRC_ILT_LINES      RTE_DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
 
 struct hw_context {
     struct bnx2x_dma    vcxt_dma;
@@ -1258,7 +1256,7 @@ struct bnx2x_softc {
 #define BNX2X_L2_CID_COUNT(sc)                                             \
 	(BNX2X_NUM_ETH_QUEUES(sc) * ECORE_MULTI_TX_COS + 2 * CNIC_SUPPORT(sc))
 #define L2_ILT_LINES(sc)                                \
-	(DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
+	(RTE_DIV_ROUND_UP(BNX2X_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
 
 	int qm_cid_count;
 
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 10/15] net/cxgbe: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Potnuri Bharat Teja <bharat@chelsio.com>
---
Changes in v3:
  - Leave call sites in base/ alone and map DIV_ROUND_UP to
    RTE_DIV_ROUND_UP in cxgbe_compat.h.
---
 drivers/net/cxgbe/cxgbe_compat.h |  2 +-
 drivers/net/cxgbe/cxgbe_filter.c |  6 +++---
 drivers/net/cxgbe/cxgbe_ofld.h   |  4 ++--
 drivers/net/cxgbe/sge.c          | 14 +++++++-------
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/net/cxgbe/cxgbe_compat.h b/drivers/net/cxgbe/cxgbe_compat.h
index e74d476f6d..8c03a07fde 100644
--- a/drivers/net/cxgbe/cxgbe_compat.h
+++ b/drivers/net/cxgbe/cxgbe_compat.h
@@ -166,7 +166,7 @@ typedef uint64_t  dma_addr_t;
 typedef char *caddr_t;
 #endif
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#define DIV_ROUND_UP(n, d) RTE_DIV_ROUND_UP(n, d)
 #define DELAY(x) rte_delay_us(x)
 #define udelay(x) DELAY(x)
 #define msleep(x) DELAY(1000 * (x))
diff --git a/drivers/net/cxgbe/cxgbe_filter.c b/drivers/net/cxgbe/cxgbe_filter.c
index 5a7efe7a73..83f86a747a 100644
--- a/drivers/net/cxgbe/cxgbe_filter.c
+++ b/drivers/net/cxgbe/cxgbe_filter.c
@@ -219,7 +219,7 @@ static inline void mk_set_tcb_field_ulp(struct filter_entry *f,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*req) - sizeof(struct work_request_hdr));
 	OPCODE_TID(req) = cpu_to_be32(MK_OPCODE_TID(CPL_SET_TCB_FIELD, f->tid));
@@ -371,7 +371,7 @@ static void mk_abort_req_ulp(struct cpl_abort_req *abort_req,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_req), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_req), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_req) -
 			      sizeof(struct work_request_hdr));
@@ -395,7 +395,7 @@ static void mk_abort_rpl_ulp(struct cpl_abort_rpl *abort_rpl,
 
 	txpkt->cmd_dest = cpu_to_be32(V_ULPTX_CMD(ULP_TX_PKT) |
 				      V_ULP_TXPKT_DEST(0));
-	txpkt->len = cpu_to_be32(DIV_ROUND_UP(sizeof(*abort_rpl), 16));
+	txpkt->len = cpu_to_be32(RTE_DIV_ROUND_UP(sizeof(*abort_rpl), 16));
 	sc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
 	sc->len = cpu_to_be32(sizeof(*abort_rpl) -
 			      sizeof(struct work_request_hdr));
diff --git a/drivers/net/cxgbe/cxgbe_ofld.h b/drivers/net/cxgbe/cxgbe_ofld.h
index fd1e7d8002..ef56eef426 100644
--- a/drivers/net/cxgbe/cxgbe_ofld.h
+++ b/drivers/net/cxgbe/cxgbe_ofld.h
@@ -14,7 +14,7 @@
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_TP_WR) | \
 				V_FW_WR_IMMDLEN(sizeof(*w) - sizeof(w->wr))); \
 	(w)->wr.wr_mid = cpu_to_be32( \
-				V_FW_WR_LEN16(DIV_ROUND_UP(sizeof(*w), 16)) | \
+				V_FW_WR_LEN16(RTE_DIV_ROUND_UP(sizeof(*w), 16)) | \
 				V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
@@ -27,7 +27,7 @@
 #define INIT_ULPTX_WR(w, wrlen, atomic, tid) do { \
 	(w)->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_ULPTX_WR) | \
 				    V_FW_WR_ATOMIC(atomic)); \
-	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(DIV_ROUND_UP(wrlen, 16)) | \
+	(w)->wr.wr_mid = cpu_to_be32(V_FW_WR_LEN16(RTE_DIV_ROUND_UP(wrlen, 16)) | \
 				     V_FW_WR_FLOWID(tid)); \
 	(w)->wr.wr_lo = cpu_to_be64(0); \
 } while (0)
diff --git a/drivers/net/cxgbe/sge.c b/drivers/net/cxgbe/sge.c
index e9d45f24c4..eb25d80e33 100644
--- a/drivers/net/cxgbe/sge.c
+++ b/drivers/net/cxgbe/sge.c
@@ -459,7 +459,7 @@ static inline unsigned int sgl_len(unsigned int n)
  */
 static inline unsigned int flits_to_desc(unsigned int n)
 {
-	return DIV_ROUND_UP(n, 8);
+	return RTE_DIV_ROUND_UP(n, 8);
 }
 
 /**
@@ -505,7 +505,7 @@ static inline unsigned int calc_tx_flits(const struct rte_mbuf *m,
 
 	hdrlen = is_eth_imm(m);
 	if (hdrlen)
-		return DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
+		return RTE_DIV_ROUND_UP(m->pkt_len + hdrlen, sizeof(__be64));
 
 	/*
 	 * Otherwise, we're going to have to construct a Scatter gather list
@@ -779,7 +779,7 @@ static inline void ship_tx_pkt_coalesce_wr(struct adapter *adap,
 	wr = (void *)&q->desc[q->pidx];
 	vmwr = (void *)&q->desc[q->pidx];
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(q->coalesce.flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(q->coalesce.flits, 2));
 	ndesc = flits_to_desc(q->coalesce.flits);
 	wr->equiq_to_len16 = htonl(wr_mid);
 	wr->plen = cpu_to_be16(q->coalesce.len);
@@ -866,7 +866,7 @@ static inline int should_tx_packet_coalesce(struct sge_eth_txq *txq,
 
 	/* If coalescing is on, the mbuf is added to a pkts WR */
 	if (q->coalesce.idx) {
-		ndesc = DIV_ROUND_UP(q->coalesce.flits + flits, 8);
+		ndesc = RTE_DIV_ROUND_UP(q->coalesce.flits + flits, 8);
 		credits = txq_avail(q) - ndesc;
 
 		if (unlikely(wraps_around(q, ndesc)))
@@ -950,7 +950,7 @@ static inline int tx_do_packet_coalesce(struct sge_eth_txq *txq,
 		mc->cmd_dest = htonl(V_ULPTX_CMD(4) | V_ULP_TXPKT_DEST(0) |
 				     V_ULP_TXPKT_FID(adap->sge.fw_evtq.cntxt_id) |
 				     F_ULP_TXPKT_RO);
-		mc->len = htonl(DIV_ROUND_UP(flits, 2));
+		mc->len = htonl(RTE_DIV_ROUND_UP(flits, 2));
 		sc_imm = (struct ulptx_idata *)(mc + 1);
 		sc_imm->cmd_more = htonl(V_ULPTX_CMD(ULP_TX_SC_IMM) |
 					 F_ULP_TX_SC_MORE);
@@ -1116,7 +1116,7 @@ int t4_eth_xmit(struct sge_eth_txq *txq, struct rte_mbuf *mbuf,
 		goto out_free;
 	}
 
-	wr_mid = V_FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	wr_mid = V_FW_WR_LEN16(RTE_DIV_ROUND_UP(flits, 2));
 	if (Q_IDXDIFF(&txq->q, equeidx)  >= 64) {
 		txq->q.equeidx = txq->q.pidx;
 		wr_mid |= F_FW_WR_EQUEQ;
@@ -1322,7 +1322,7 @@ static int ctrl_xmit(struct sge_ctrl_txq *q, struct rte_mbuf *mbuf)
 	}
 
 	reclaim_completed_tx_imm(&q->q);
-	ndesc = DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
+	ndesc = RTE_DIV_ROUND_UP(mbuf->pkt_len, sizeof(struct tx_desc));
 	t4_os_lock(&q->ctrlq_lock);
 
 	q->full = txq_avail(&q->q) < ndesc ? 1 : 0;
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 11/15] net/ena: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v4:
  - Add rte_common.h include.
Changes in v3:
  - Leave call sites in base/ alone and map DIV_ROUND_UP to
    RTE_DIV_ROUND_UP in ena_plat_dpdk.h.
  - Remove extraneous blank line in ena_plat_dpdk.h.
---
 drivers/net/ena/base/ena_plat_dpdk.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ena/base/ena_plat_dpdk.h b/drivers/net/ena/base/ena_plat_dpdk.h
index 83b354d9da..4a6d53d0a2 100644
--- a/drivers/net/ena/base/ena_plat_dpdk.h
+++ b/drivers/net/ena/base/ena_plat_dpdk.h
@@ -17,6 +17,7 @@
 #include <ethdev_driver.h>
 #include <rte_atomic.h>
 #include <rte_branch_prediction.h>
+#include <rte_common.h>
 #include <rte_cycles.h>
 #include <rte_io.h>
 #include <rte_log.h>
@@ -308,7 +309,7 @@ ena_mem_alloc_coherent(struct rte_eth_dev_data *data, size_t size,
 			rte_write64_relaxed(*from, to);			       \
 	} while(0)
 
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#define DIV_ROUND_UP(n, d) RTE_DIV_ROUND_UP(n, d)
 
 #define ENA_FFS(x) ffs(x)
 
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 12/15] net/nfp: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
---
Changes in v4:
  - Add rte_common.h include.
Changes in v3:
  - Change commit subject prefix from "drivers:" to "net/nfp:".
  - Drop line rewraps in nfp_nfdk.h and nfp_nfdk_dp.c.
---
 drivers/common/nfp/nfp_platform.h  | 2 --
 drivers/net/nfp/nfdk/nfp_nfdk.h    | 3 ++-
 drivers/net/nfp/nfdk/nfp_nfdk_dp.c | 2 +-
 drivers/net/nfp/nfpcore/nfp_nsp.c  | 2 +-
 4 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/drivers/common/nfp/nfp_platform.h b/drivers/common/nfp/nfp_platform.h
index 0b02fcf1e8..814f90b5a2 100644
--- a/drivers/common/nfp/nfp_platform.h
+++ b/drivers/common/nfp/nfp_platform.h
@@ -10,8 +10,6 @@
 
 #include <rte_bitops.h>
 
-#define DIV_ROUND_UP(n, d)             (((n) + (d) - 1) / (d))
-
 #define DMA_BIT_MASK(n)    ((1ULL << (n)) - 1)
 
 #define BITS_PER_LONG      (__SIZEOF_LONG__ * 8)
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk.h b/drivers/net/nfp/nfdk/nfp_nfdk.h
index 64c2b31308..d843162be0 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk.h
+++ b/drivers/net/nfp/nfdk/nfp_nfdk.h
@@ -7,6 +7,7 @@
 #define __NFP_NFDK_H__
 
 #include <nfp_platform.h>
+#include <rte_common.h>
 
 #include "../nfp_rxtx.h"
 #include "nfp_net_common.h"
@@ -168,7 +169,7 @@ static inline uint16_t
 nfp_net_nfdk_headlen_to_segs(uint16_t headlen)
 {
 	/* First descriptor fits less data, so adjust for that */
-	return DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC - NFDK_TX_MAX_DATA_PER_HEAD,
+	return RTE_DIV_ROUND_UP(headlen + NFDK_TX_MAX_DATA_PER_DESC - NFDK_TX_MAX_DATA_PER_HEAD,
 			NFDK_TX_MAX_DATA_PER_DESC);
 }
 
diff --git a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
index 15867ab62f..777550b6df 100644
--- a/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
+++ b/drivers/net/nfp/nfdk/nfp_nfdk_dp.c
@@ -103,7 +103,7 @@ nfp_net_nfdk_tx_maybe_close_block(struct nfp_net_txq *txq,
 	n_descs = nfp_net_nfdk_headlen_to_segs(pkt_temp->data_len);
 	while (pkt_temp->next != NULL) {
 		pkt_temp = pkt_temp->next;
-		n_descs += DIV_ROUND_UP(pkt_temp->data_len, NFDK_TX_MAX_DATA_PER_DESC);
+		n_descs += RTE_DIV_ROUND_UP(pkt_temp->data_len, NFDK_TX_MAX_DATA_PER_DESC);
 	}
 
 	if (unlikely(n_descs > NFDK_TX_DESC_GATHER_MAX))
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp.c b/drivers/net/nfp/nfpcore/nfp_nsp.c
index 43c13baabe..bd02efcc4a 100644
--- a/drivers/net/nfp/nfpcore/nfp_nsp.c
+++ b/drivers/net/nfp/nfpcore/nfp_nsp.c
@@ -542,7 +542,7 @@ nfp_nsp_command_buf_dma_sg(struct nfp_nsp *nsp,
 	struct nfp_nsp_dma_desc *descs;
 
 	chunk_size = RTE_BIT64(chunk_order);
-	nseg = DIV_ROUND_UP(max_size, chunk_size);
+	nseg = RTE_DIV_ROUND_UP(max_size, chunk_size);
 
 	/* Malloc memory */
 	total_len = sizeof(struct nfp_nsp_dma_buf) +
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 13/15] app/procinfo: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_ROUND_UP redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
---
 app/proc-info/main.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/app/proc-info/main.c b/app/proc-info/main.c
index 5925bbb765..9c22fb8f01 100644
--- a/app/proc-info/main.c
+++ b/app/proc-info/main.c
@@ -48,7 +48,6 @@
 
 #define ETHDEV_FWVERS_LEN 32
 #define RTE_RETA_CONF_GROUP_NUM 32
-#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 #define EEPROM_DUMP_CHUNKSIZE 1024
 
 #define STATS_BDR_FMT "========================================"
@@ -1830,7 +1829,7 @@ show_port_rss_reta_info(void)
 			return;
 		}
 
-		num = DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
+		num = RTE_DIV_ROUND_UP(dev_info.reta_size, RTE_ETH_RETA_GROUP_SIZE);
 		memset(reta_conf, 0, sizeof(reta_conf));
 		for (i = 0; i < num; i++)
 			reta_conf[i].mask = ~0ULL;
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 14/15] app/test: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>
---
Changes in v4:
  - Add rte_common.h include.
Changes in v3:
  - New patch converting DIV_CEIL in test_compressdev.c.
---
 app/test/test_compressdev.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/app/test/test_compressdev.c b/app/test/test_compressdev.c
index 293d1d757f..ed0352a909 100644
--- a/app/test/test_compressdev.c
+++ b/app/test/test_compressdev.c
@@ -14,12 +14,11 @@
 #include <rte_mbuf.h>
 #include <rte_compressdev.h>
 #include <rte_string_fns.h>
+#include <rte_common.h>
 
 #include "test_compressdev_test_buffer.h"
 #include "test.h"
 
-#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
-
 #define DEFAULT_WINDOW_SIZE 15
 #define DEFAULT_MEM_LEVEL 8
 #define MAX_DEQD_RETRIES 10
@@ -688,7 +687,7 @@ prepare_sgl_bufs(const char *test_buf, struct rte_mbuf *head_buf,
 		uint16_t seg_size)
 {
 	uint32_t remaining_data = total_data_size;
-	uint16_t num_remaining_segs = DIV_CEIL(remaining_data, seg_size);
+	uint16_t num_remaining_segs = RTE_DIV_ROUND_UP(remaining_data, seg_size);
 	struct rte_mempool *pool;
 	struct rte_mbuf *next_seg;
 	uint32_t data_size;
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

[PATCH v4 15/15] app/test-compress-perf: use common division round up macro

[Joshua Washington]

The RTE_DIV_ROUND_UP in rte_common.h makes DIV_CEIL redundant.

Signed-off-by: Joshua Washington <joshwash@google.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>
---
Changes in v4:
  - Add rte_common.h include.
Changes in v3:
  - New patch converting DIV_CEIL in comp_perf_test_common.c.
---
 app/test-compress-perf/comp_perf_test_common.c | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/app/test-compress-perf/comp_perf_test_common.c b/app/test-compress-perf/comp_perf_test_common.c
index 78487196ad..53685cd57d 100644
--- a/app/test-compress-perf/comp_perf_test_common.c
+++ b/app/test-compress-perf/comp_perf_test_common.c
@@ -6,6 +6,7 @@
 #include <rte_eal.h>
 #include <rte_log.h>
 #include <rte_compressdev.h>
+#include <rte_common.h>
 
 #include "comp_perf.h"
 #include "comp_perf_options.h"
@@ -14,9 +15,6 @@
 #include "comp_perf_test_common.h"
 #include "comp_perf_test_verify.h"
 
-
-#define DIV_CEIL(a, b)  ((a) / (b) + ((a) % (b) != 0))
-
 struct cperf_buffer_info {
 	uint16_t total_segments;
 	uint16_t segment_sz;
@@ -238,8 +236,8 @@ comp_perf_allocate_memory(struct comp_test_data *test_data,
 		 * Number of segments for input and output
 		 * (compression and decompression)
 		 */
-		test_data->total_segs = DIV_CEIL(test_data->input_data_sz,
-						 test_data->seg_sz);
+		test_data->total_segs = RTE_DIV_ROUND_UP(test_data->input_data_sz,
+							 test_data->seg_sz);
 	} else {
 		/*
 		 * When application does decompression only, input data is
@@ -292,8 +290,8 @@ comp_perf_allocate_memory(struct comp_test_data *test_data,
 		return -1;
 	}
 
-	mem->total_bufs = DIV_CEIL(test_data->total_segs,
-				   test_data->max_sgl_segs);
+	mem->total_bufs = RTE_DIV_ROUND_UP(test_data->total_segs,
+					   test_data->max_sgl_segs);
 
 	snprintf(pool_name, sizeof(pool_name), "op_pool_%u_qp_%u",
 			mem->dev_id, mem->qp_id);
@@ -557,7 +555,7 @@ prepare_bufs(struct comp_test_data *test_data, struct cperf_mem_resources *mem)
 void
 print_test_dynamics(const struct comp_test_data *test_data)
 {
-	uint32_t opt_total_segs = DIV_CEIL(buffer_info.input_data_sz,
+	uint32_t opt_total_segs = RTE_DIV_ROUND_UP(buffer_info.input_data_sz,
 			MAX_SEG_SIZE);
 
 	if (buffer_info.total_buffs > 1) {
-- 
2.56.0.rc1.315.gc6ed9934b7-goog

Re: [PATCH v4 00/15] Introduce RTE_DIV_ROUND_UP

[Stephen Hemminger]

On Mon,  5 Oct 2026 19:30:36 -0700
Joshua Washington <joshwash@google.com> wrote:

> Numerous components independently define the DIV_ROUND_UP macro, using
> the same definition. Attempt to commonize by introducing
> RTE_DIV_ROUND_UP to rte_common.h and replacing usages of DIV_ROUND_UP
> with RTE_DIV_ROUND_UP. Similarly, replace DIV_CEIL, which is
> functionally identical, to RTE_DIV_ROUND_UP.
> 
> Changes in v4:
>   - eal: Fix merge conflict in release notes after rebase.
>   - various commits: add rte_common.h includes where missing to fix
>     compilation issues.
>   - opae: drop RTE_DIV_ROUND_UP change due to shered osdep file for RTE
>     and raw drivers.

Looks good, some minor AI notes on commit messages and doc

Review: [PATCH v4 00/15] division round up consolidation

Series applies cleanly to main. Each patch touches a disjoint set of
files, so with 01/15 first the series is bisect safe. Build of all
affected drivers, dpdk-proc-info, dpdk-test and dpdk-test-compress-perf
with -Dwerror=true is clean.

No errors or warnings.

Series

Info: remaining private copies not covered by this series, candidates
for a follow-up:
  drivers/raw/ifpga/base/opae_osdep.h   DIV_ROUND_UP (map like cxgbe)
  drivers/crypto/ccp/ccp_dev.h          CCP_DIV_ROUND_UP
  drivers/common/qat/dev/qat_dev_gen_lce.c  LCE_DIV_ROUND_UP
  drivers/net/sfc/sfc_dp.h              SFC_DIV_ROUND_UP
  drivers/common/sxe2/sxe2_osal.h       SXE2_DIV_ROUND_UP
  drivers/common/cnxk/roc_platform.h    PLT_DIV_CEIL
sfc_efx base (EFX_DIV_ROUND_UP) is shared code and should stay.

Patch 01/15 eal: introduce division round up macro

Info: the Doxygen repeats itself. "Evaluated twice" is in @param d and
again in the first @note; "non-negative" is in @param n and again in
the second @note. Keep the @param text and drop the duplicate notes.

The plain expression form (not a statement expression with typeof) is
correct here: several converted users need an integer constant
expression (array bound via BITS_TO_LONGS in rnp.h, MAX_KEY_DWORDS in
hns3_fdir.c, CNIC_ILT_LINES in bnx2x.h).

Patch 05/15 net/qede, 10/15 net/cxgbe, 11/15 net/ena

Info: commit message says RTE_DIV_ROUND_UP "makes DIV_ROUND_UP
redundant", but these patches keep DIV_ROUND_UP as an alias for base/
code. Say that instead, e.g.:

  Map DIV_ROUND_UP to RTE_DIV_ROUND_UP for the shared base code
  and use RTE_DIV_ROUND_UP directly elsewhere in the driver.

Patch 05/15 net/qede

Info: bcm_osal.h still defines RTE_ROUNDUP in the driver, squatting on
the RTE_ namespace the same way the removed RTE_DIV_ROUND_UP did.
RTE_ALIGN_MUL_CEIL() is the equivalent. Follow-up, not this series.

Patch 06/15 net/rnp

Info: nothing in DPDK defines BITS_TO_LONGS, so the new #ifndef guard
does nothing; drop it.

Re: [PATCH v4 00/15] Introduce RTE_DIV_ROUND_UP

[Thomas Monjalon]

06/10/2026 18:38, Stephen Hemminger:
> On Mon,  5 Oct 2026 19:30:36 -0700
> Joshua Washington <joshwash@google.com> wrote:
> 
> > Numerous components independently define the DIV_ROUND_UP macro, using
> > the same definition. Attempt to commonize by introducing
> > RTE_DIV_ROUND_UP to rte_common.h and replacing usages of DIV_ROUND_UP
> > with RTE_DIV_ROUND_UP. Similarly, replace DIV_CEIL, which is
> > functionally identical, to RTE_DIV_ROUND_UP.
> > 
> > Changes in v4:
> >   - eal: Fix merge conflict in release notes after rebase.
> >   - various commits: add rte_common.h includes where missing to fix
> >     compilation issues.
> >   - opae: drop RTE_DIV_ROUND_UP change due to shered osdep file for RTE
> >     and raw drivers.
> 
> Looks good, some minor AI notes on commit messages and doc

I consider the minor notes as candidates for follow-up or not relevant.

Applied (with a bit of squashing), thanks.


> Info: remaining private copies not covered by this series, candidates
> for a follow-up:
>   drivers/raw/ifpga/base/opae_osdep.h   DIV_ROUND_UP (map like cxgbe)
>   drivers/crypto/ccp/ccp_dev.h          CCP_DIV_ROUND_UP
>   drivers/common/qat/dev/qat_dev_gen_lce.c  LCE_DIV_ROUND_UP
>   drivers/net/sfc/sfc_dp.h              SFC_DIV_ROUND_UP
>   drivers/common/sxe2/sxe2_osal.h       SXE2_DIV_ROUND_UP
>   drivers/common/cnxk/roc_platform.h    PLT_DIV_CEIL

Would be nice to address in a new series, yes.


> Info: the Doxygen repeats itself. "Evaluated twice" is in @param d and
> again in the first @note; "non-negative" is in @param n and again in
> the second @note. Keep the @param text and drop the duplicate notes.

I think this repeat is fine.


> Info: nothing in DPDK defines BITS_TO_LONGS, so the new #ifndef guard
> does nothing; drop it.

It is in rnp_osdep.h, probably to allow including in a different environment.