spi: mtk_snor: avoid alloc in mtk_snor_cmd_program()

Rework mtk_snor_cmd_program() to avoid allocating a temporary buffer
for tx data. This improves performance a bit by avoiding the need to
allocate memory and copy data an extra time.

Signed-off-by: Noah.Shen <noah.shen@mediatek.com>
Reviewed-by: Julien Stephan <jstephan@baylibre.com>
Link: https://patch.msgid.link/20260406-mtk-spi-nor-improvements-v1-2-66f675cbbd3e@baylibre.com
Signed-off-by: David Lechner <dlechner@baylibre.com>

diff --git a/drivers/spi/mtk_snor.c b/drivers/spi/mtk_snor.c
index eb36c9dd5e8c6f2e9112f8c37f2a05812eadbdc0..0ff4a8b23d51e24afbda82f50cd1e96cec8275aa 100644 (file)
--- a/drivers/spi/mtk_snor.c
+++ b/drivers/spi/mtk_snor.c
@@ -383,50 +383,63 @@ static int mtk_snor_pp_unbuffered(struct mtk_snor_priv *priv,
 static int mtk_snor_cmd_program(struct mtk_snor_priv *priv,
                                const struct spi_mem_op *op)
 {
-       u32 tx_len = 0;
-       u32 trx_len = 0;
+       int rx_len = 0;
        int reg_offset = MTK_NOR_REG_PRGDATA_MAX;
+       int tx_len, prg_len;
+       int i;
        void __iomem *reg;
-       u8 *txbuf;
-       int tx_cnt = 0;
-       u8 *rxbuf = op->data.buf.in;
-       int i = 0;
+       u8 val;
 
-       tx_len = 1 + op->addr.nbytes + op->dummy.nbytes;
-       trx_len = tx_len + op->data.nbytes;
-       if (op->data.dir == SPI_MEM_DATA_OUT)
-               tx_len += op->data.nbytes;
-
-       txbuf = kmalloc_array(tx_len, sizeof(u8), GFP_KERNEL);
-       memset(txbuf, 0x0, tx_len * sizeof(u8));
-
-       /* Join all bytes to be transferred */
-       txbuf[tx_cnt] = op->cmd.opcode;
-       tx_cnt++;
-       for (i = op->addr.nbytes; i > 0; i--, tx_cnt++)
-               txbuf[tx_cnt] = ((u8 *)&op->addr.val)[i - 1];
-       for (i = op->dummy.nbytes; i > 0; i--, tx_cnt++)
-               txbuf[tx_cnt] = 0x0;
+       tx_len = op->cmd.nbytes + op->addr.nbytes;
+
+       /* count dummy bytes only if we need to write data after it */
        if (op->data.dir == SPI_MEM_DATA_OUT)
-               for (i = op->data.nbytes; i > 0; i--, tx_cnt++)
-                       txbuf[tx_cnt] = ((u8 *)op->data.buf.out)[i - 1];
+               tx_len += op->dummy.nbytes + op->data.nbytes;
+       else if (op->data.dir == SPI_MEM_DATA_IN)
+               rx_len = op->data.nbytes;
 
-       for (i = MTK_NOR_REG_PRGDATA_MAX; i >= 0; i--)
-               writeb(0, priv->base + MTK_NOR_REG_PRGDATA(i));
+       prg_len = op->cmd.nbytes + op->addr.nbytes + op->dummy.nbytes +
+                 op->data.nbytes;
 
-       for (i = 0; i < tx_len; i++, reg_offset--)
-               writeb(txbuf[i], priv->base + MTK_NOR_REG_PRGDATA(reg_offset));
+       /* fill tx data */
 
-       kfree(txbuf);
+       for (i = op->cmd.nbytes; i > 0; i--, reg_offset--) {
+               reg = priv->base + MTK_NOR_REG_PRGDATA(reg_offset);
+               val = (op->cmd.opcode >> ((i - 1) * BITS_PER_BYTE)) & 0xff;
+               writeb(val, reg);
+       }
 
-       writel(trx_len * BITS_PER_BYTE, priv->base + MTK_NOR_REG_PRG_CNT);
+       for (i = op->addr.nbytes; i > 0; i--, reg_offset--) {
+               reg = priv->base + MTK_NOR_REG_PRGDATA(reg_offset);
+               val = (op->addr.val >> ((i - 1) * BITS_PER_BYTE)) & 0xff;
+               writeb(val, reg);
+       }
 
-       mtk_snor_cmd_exec(priv, MTK_NOR_CMD_PROGRAM, trx_len * BITS_PER_BYTE);
+       for (i = 0; i < op->dummy.nbytes; i++, reg_offset--) {
+               reg = priv->base + MTK_NOR_REG_PRGDATA(reg_offset);
+               writeb(0, reg);
+       }
 
-       reg_offset = op->data.nbytes - 1;
        for (i = 0; i < op->data.nbytes; i++, reg_offset--) {
+               reg = priv->base + MTK_NOR_REG_PRGDATA(reg_offset);
+               writeb(((const u8 *)(op->data.buf.out))[i], reg);
+       }
+
+       for (; reg_offset >= 0; reg_offset--) {
+               reg = priv->base + MTK_NOR_REG_PRGDATA(reg_offset);
+               writeb(0, reg);
+       }
+
+       /* trigger op */
+       writel(prg_len * BITS_PER_BYTE, priv->base + MTK_NOR_REG_PRG_CNT);
+
+       mtk_snor_cmd_exec(priv, MTK_NOR_CMD_PROGRAM, prg_len * BITS_PER_BYTE);
+
+       /* fetch read data */
+       reg_offset = 0;
+       for (i = op->data.nbytes - 1; i >= 0; i--, reg_offset++) {
                reg = priv->base + MTK_NOR_REG_SHIFT(reg_offset);
-               rxbuf[i] = readb(reg);
+               ((u8 *)(op->data.buf.in))[i] = readb(reg);
        }
 
        return 0;