Hi Keguang, > > > + > > > + nfc->data->parse_address(chip, addrs, naddrs, cmd); > > > + /* set NAND address */ > > > + nand_writel(nfc, NAND_ADDR1, nfc->addr1_reg); > > > + nand_writel(nfc, NAND_ADDR2, nfc->addr2_reg); > > > + break; > > > + case NAND_OP_DATA_IN_INSTR: > > > + case NAND_OP_DATA_OUT_INSTR: > > > + offset = nand_subop_get_data_start_off(subop, op_id); > > > + nfc->len = nand_subop_get_data_len(subop, op_id); > > > + if (instr->type == NAND_OP_DATA_IN_INSTR) > > > + nfc->buf = > > > + (void *)instr->ctx.data.buf.in + offset; > > > + else if (instr->type == NAND_OP_DATA_OUT_INSTR) > > > + nfc->buf = > > > + (void *)instr->ctx.data.buf.out + offset; > > > > The buf pointer feels clunky. You don't know for how long the buffer > > you point to will be valid, please don't do that. > > > The buf pointer is used for DMA transfer afterwards. > I referred to other drivers, and they used the same approach. > https://elixir.bootlin.com/linux/v6.8.9/source/drivers/mtd/nand/raw/arasan-nand-controller.c#L647 nfc_op is a local variable, it doesn't live past the function call. > https://elixir.bootlin.com/linux/v6.8.9/source/drivers/mtd/nand/raw/rockchip-nand-controller.c#L366 Thanks, Miquèl
