sunxi: initial upstreamining effort

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

This has been stripped back for mainlining and supports only sun7i. These changes are not useful by themselves but are split out to make the patch sizes more manageable.

As well as the following signed-off-by the sunxi branch shows commits to these files authored by the following:
Carl van Schaik
Henrik Nordstrom
Stefan Roese
Tom Cubie

Signed-off-by Chen-Yu Tsai <[email protected]>

Signed-off-by Hans de Goede <[email protected]>

Signed-off-by Ma Haijun <[email protected]>

Signed-off-by Oliver Schinagl <[email protected]>

Signed-off-by Ian Campbell <[email protected]>

And as for the rest of the code, lots of magic numbers to #define what/why (why udelay(2) and 22?)

fre 2014-03-14 klockan 10:17 -0400 skrev Tom Rini:

Quoted 2 lines

And as for the rest of the code, lots of magic numbers to #define what/why (why udelay(2) and 22?)

Unfortunately the only documentation we have on the sun7i DRAM controller is Allwinner boot0 + boot1 code, and that code uses.. magic numbers & shifts with no explanations.

Unfortunately the only documentation we have on the sun7i DRAM controller is Allwinner boot0 + boot1 code, and that code uses.. magic numbers & shifts with no explanations.

Quoted 3 lines

Unfortunately the only documentation we have on the sun7i DRAM controller is Allwinner boot0 + boot1 code, and that code uses.. magic numbers & shifts with no explanations.

To avoid having to explain this over-and-over again, the code could include some comment(s) pointing out that these are unexplained magic constants from Allwinner's code.

On Fri, Mar 14, 2014 at 10:33:45AM +0000, Ian Campbell wrote:

[snip]

Quoted 4 lines

+static void mctl_ddr3_reset(void) +{ + struct sunxi_dram_reg
*dram = + (struct sunxi_dram_reg *)SUNXI_DRAMC_BASE; + + { +
clrbits_le32(&dram->mcr, DRAM_MCR_RESET); + udelay(2); +
setbits_le32(&dram->mcr, DRAM_MCR_RESET); + }

That seems like an odd construction, why the extra braces?

And as for the rest of the code, lots of magic numbers to #define what/why (why udelay(2) and 22?)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 10 lines · Tom Rini

On Fri, Mar 14, 2014 at 10:33:45AM +0000, Ian Campbell wrote:

[snip]

Quoted 4 lines

+static void mctl_ddr3_reset(void) +{ + struct sunxi_dram_reg
*dram = + (struct sunxi_dram_reg *)SUNXI_DRAMC_BASE; + + { +
clrbits_le32(&dram->mcr, DRAM_MCR_RESET); + udelay(2); +
setbits_le32(&dram->mcr, DRAM_MCR_RESET); + }

That seems like an odd construction, why the extra braces?

This originally had a conditional depending on the SoC family. (Yeah, they need to reset the ram differently). It seems it wasn't removed properly.

Quoted 3 lines

And as for the rest of the code, lots of magic numbers to #define what/why (why udelay(2) and 22?)

Before going into more detail, remember this is ram initialization code. That's always going to be a pain :(.
There's nothing magic here. It's just a fact of life. Every step is going to need a different delay. No need to bloat the headers by #defining each. It also makes raminit code more unreadable.

We got these numbers from allwinner code dumps. We used to have these as sdelay() numbers, which usually meant units of 2 clock cycles. So we had to convert them to udelay() to at least make the delays independent of CPU clock. The old sdelay() numbers made no sense either.

+/* The sunxi internal brom will try to loader external bootloader
+ * from mmc0, nannd flash, mmc2.

board/sunxi/board.c new file +6 -0

+#ifdef CONFIG_SPL_BUILD
+void sunxi_board_init(void)
+{
+	int power_failed = 0;
+	unsigned long ramsize;
+

Quoted 5 lines · Ian Campbell

+#ifdef CONFIG_SPL_BUILD +void sunxi_board_init(void) +{ + int
power_failed = 0; + unsigned long ramsize; + + printf("DRAM:"); +
ramsize = sunxi_dram_init(); + if (!ramsize) { + printf(" ?"); +
ramsize = sunxi_dram_init(); + } + if (!ramsize) { + printf("
?"); + ramsize = sunxi_dram_init(); + }

We have to try 3 times? What?

Quoted 3 lines

(quoted diff) +1 -0

+	/* +	 * Only clock up the CPU to full speed if we are

reasonably + * assured it's being powered with suitable core
voltage + */ + if (!power_failed) + clock_set_pll1(912000000);

Define that please.

+#define CONFIG_SUN7I /* sun7i SoC generation */

+#define CONFIG_SYS_PROMPT_HUSH_PS2 "> "

+#define CONFIG_SYS_BAUDRATE_TABLE { 9600, 19200, 38400, 57600, 115200 }

include/configs/sunxi-common.h line 81

Quoted 5 lines · Ian Campbell

include/configs/sunxi-common.h new file +5 -0

+/*
+ * Size of malloc() pool
+ * 1MB = 0x100000, 0x100000 = 1024 * 1024
+ */
+#define CONFIG_SYS_MALLOC_LEN		(CONFIG_ENV_SIZE + (1 << 20))

This is really really small, and precludes using UBI on NAND for example.

Quoted 10 lines · Tom Rini

include/configs/sunxi-common.h line 81

Quoted 5 lines · Ian Campbell

include/configs/sunxi-common.h new file +5 -0

+/*
+ * Size of malloc() pool
+ * 1MB = 0x100000, 0x100000 = 1024 * 1024
+ */
+#define CONFIG_SYS_MALLOC_LEN		(CONFIG_ENV_SIZE + (1 << 20))

This is really really small, and precludes using UBI on NAND for example.

What is a suitable value? I see a lots of boards using 128kB or <1MB and only a handful using more. Tegra uses 4MB -- is that a reasonable value?

Quoted 12 lines · Ian Campbell

Quoted 10 lines · Tom Rini

include/configs/sunxi-common.h line 81

Quoted 5 lines · Ian Campbell

include/configs/sunxi-common.h new file +5 -0

+/*
+ * Size of malloc() pool
+ * 1MB = 0x100000, 0x100000 = 1024 * 1024
+ */
+#define CONFIG_SYS_MALLOC_LEN		(CONFIG_ENV_SIZE + (1 << 20))

This is really really small, and precludes using UBI on NAND for example.

Quoted 2 lines

What is a suitable value? I see a lots of boards using 128kB or <1MB and only a handful using more. Tegra uses 4MB -- is that a reasonable value?

4MB is reasonable when you start having to worry about 4KB page NAND and UBI. I might have been a little harsh saying "really really small" and 1MB might be just enough to work with smaller page size NAND.

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

Signed-off-by Oliver Schinagl <[email protected]>

Signed-off-by Ian Campbell <[email protected]>

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

As well as the following signed-off-by the sunxi branch shows commits to these files authored by the following:
Stefan Roese
Tom Cubie
yemao

Signed-off-by Henrik Nordstrom <[email protected]>

Signed-off-by Luke Leighton <[email protected]>

Signed-off-by Oliver Schinagl <[email protected]>

Signed-off-by Wills Wang <[email protected]>

Signed-off-by Ian Campbell <[email protected]>

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

As well as the following signed-off-by the sunxi branch shows commits to these files authored by the following:
Stefan Roese
Tom Cubie
yemao

Signed-off-by Henrik Nordstrom <[email protected]>

Signed-off-by Luke Leighton <[email protected]>

Signed-off-by Oliver Schinagl <[email protected]>

Signed-off-by Wills Wang <[email protected]>

Signed-off-by Ian Campbell <[email protected]>

---

arch/arm/include/asm/arch-sunxi/mmc.h |  66 ++++
board/sunxi/board.c                   |  13 +
drivers/mmc/Makefile                  |   1 +
drivers/mmc/sunxi_mmc.c               | 650 ++++++++++++++++++++++++++++++++++
include/configs/sunxi-common.h        |  11 +
5 files changed, 741 insertions(+)

create mode 100644 arch/arm/include/asm/arch-sunxi/mmc.h create mode 100755 drivers/mmc/sunxi_mmc.c

arch/arm/include/asm/arch-sunxi/mmc.h new file +66 -0

new file mode 100644
index 0000000..639a7fc
@@ -0,0 +1,66 @@
+/*
+ * (C) Copyright 2007-2011
+ * Allwinner Technology Co., Ltd. <www.allwinnertech.com>
+ * Aaron <[email protected]>
+ *
+ * MMC register definition for allwinner sunxi platform.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA
+ */
+
+#ifndef _SUNXI_MMC_H
+#define _SUNXI_MMC_H
+
+#include <linux/types.h>
+
+struct sunxi_mmc {
+	u32 gctrl;		/* (0x00) SMC Global Control Register */
+	u32 clkcr;		/* (0x04) SMC Clock Control Register */
+	u32 timeout;		/* (0x08) SMC Time Out Register */
+	u32 width;		/* (0x0c) SMC Bus Width Register */
+	u32 blksz;		/* (0x10) SMC Block Size Register */
+	u32 bytecnt;		/* (0x14) SMC Byte Count Register */
+	u32 cmd;		/* (0x18) SMC Command Register */
+	u32 arg;		/* (0x1c) SMC Argument Register */
+	u32 resp0;		/* (0x20) SMC Response Register 0 */
+	u32 resp1;		/* (0x24) SMC Response Register 1 */
+	u32 resp2;		/* (0x28) SMC Response Register 2 */
+	u32 resp3;		/* (0x2c) SMC Response Register 3 */
+	u32 imask;		/* (0x30) SMC Interrupt Mask Register */
+	u32 mint;		/* (0x34) SMC Masked Interrupt Status Reg */
+	u32 rint;		/* (0x38) SMC Raw Interrupt Status Register */
+	u32 status;		/* (0x3c) SMC Status Register */
+	u32 ftrglevel;		/* (0x40) SMC FIFO Threshold Watermark Reg */
+	u32 funcsel;		/* (0x44) SMC Function Select Register */
+	u32 cbcr;		/* (0x48) SMC CIU Byte Count Register */
+	u32 bbcr;		/* (0x4c) SMC BIU Byte Count Register */
+	u32 dbgc;		/* (0x50) SMC Debug Enable Register */
+	u32 res0[11];		/* (0x54~0x7c) */
+	u32 dmac;		/* (0x80) SMC IDMAC Control Register */
+	u32 dlba;		/* (0x84) SMC IDMAC Descr List Base Addr Reg */
+	u32 idst;		/* (0x88) SMC IDMAC Status Register */
+	u32 idie;		/* (0x8c) SMC IDMAC Interrupt Enable Register */
+	u32 chda;		/* (0x90) */
+	u32 cbda;		/* (0x94) */
+	u32 res1[26];		/* (0x98~0xff) */
+	u32 fifo;		/* (0x100) SMC FIFO Access Address */
+};
+
+int sunxi_mmc_init(int sdc_no);
+#endif /* _SUNXI_MMC_H */

board/sunxi/board.c +13 -0

index bffef4f..5dbcf2a 100644
@@ -30,6 +30,7 @@
 #include <common.h>
 #include <asm/arch/clock.h>
 #include <asm/arch/dram.h>
+#include <asm/arch/mmc.h>
 
 DECLARE_GLOBAL_DATA_PTR;
 

@@ -59,6 +60,18 @@ int dram_init(void)
 	return 0;
 }
 
+#ifdef CONFIG_GENERIC_MMC
+int board_mmc_init(bd_t *bis)
+{
+	sunxi_mmc_init(CONFIG_MMC_SUNXI_SLOT);
+#if !defined (CONFIG_SPL_BUILD) && defined (CONFIG_MMC_SUNXI_SLOT_EXTRA)
+	sunxi_mmc_init(CONFIG_MMC_SUNXI_SLOT_EXTRA);
+#endif
+
+	return 0;
+}
+#endif
+
 #ifdef CONFIG_SPL_BUILD
 void sunxi_board_init(void)
 {

drivers/mmc/Makefile +0 -0

index e793ed9..c695841 100644
@@ -27,6 +27,7 @@ obj-$(CONFIG_SPEAR_SDHCI) += spear_sdhci.o

obj-$(CONFIG_TEGRA_MMC) += tegra_mmc.o
obj-$(CONFIG_DWMMC) += dw_mmc.o
obj-$(CONFIG_EXYNOS_DWMMC) += exynos_dw_mmc.o
+obj-$(CONFIG_MMC_SUNXI) += sunxi_mmc.o
obj-$(CONFIG_ZYNQ_SDHCI) += zynq_sdhci.o
obj-$(CONFIG_SOCFPGA_DWMMC) += socfpga_dw_mmc.o
ifdef CONFIG_SPL_BUILD

drivers/mmc/sunxi_mmc.c new file +64 -0

new file mode 100755
index 0000000..13eba76
@@ -0,0 +1,650 @@
+/*
+ * (C) Copyright 2007-2011
+ * Allwinner Technology Co., Ltd. <www.allwinnertech.com>
+ * Aaron <[email protected]>
+ *
+ * MMC driver for allwinner sunxi platform.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA
+ */
+
+#include <common.h>
+#include <malloc.h>
+#include <mmc.h>
+#include <asm/io.h>
+#include <asm/arch/clock.h>
+#include <asm/arch/cpu.h>
+#include <asm/arch/gpio.h>
+#include <asm/arch/mmc.h>
+
+static void dumphex32(char *name, char *base, int len)
+{
+	__u32 i;
+
+	debug("dump %s registers:", name);
+	for (i = 0; i < len; i += 4) {
+		if (!(i & 0xf))
+			debug("\n0x%p : ", base + i);
+		debug("0x%08x ", readl(base + i));
+	}
+	debug("\n");
+}
+
+static void dumpmmcreg(struct sunxi_mmc *reg)
+{
+	debug("dump mmc registers:\n");
+	debug("gctrl     0x%08x\n", reg->gctrl);
+	debug("clkcr     0x%08x\n", reg->clkcr);
+	debug("timeout   0x%08x\n", reg->timeout);
+	debug("width     0x%08x\n", reg->width);
+	debug("blksz     0x%08x\n", reg->blksz);
+	debug("bytecnt   0x%08x\n", reg->bytecnt);
+	debug("cmd       0x%08x\n", reg->cmd);
+	debug("arg       0x%08x\n", reg->arg);
+	debug("resp0     0x%08x\n", reg->resp0);
+	debug("resp1     0x%08x\n", reg->resp1);
+	debug("resp2     0x%08x\n", reg->resp2);
+	debug("resp3     0x%08x\n", reg->resp3);
+	debug("imask     0x%08x\n", reg->imask);

drivers/mmc/sunxi_mmc.c new file +26 -0

+
+struct sunxi_mmc_des {
+	u32 reserved1_1:1;
+	u32 dic:1;		/* disable interrupt on completion */
+	u32 last_des:1;		/* 1-this data buffer is the last buffer */
+	u32 first_des:1;		/* 1-data buffer is the first buffer,
+				   0-data buffer contained in the next
+				   descriptor is 1st buffer */
+	u32 des_chain:1;	/* 1-the 2nd address in the descriptor is the
+				   next descriptor address */
+	u32 end_of_ring:1;	/* 1-last descriptor flag when using dual
+				   data buffer in descriptor */
+	u32 reserved1_2:24;
+	u32 card_err_sum:1;	/* transfer error flag */
+	u32 own:1;		/* des owner:1-idma owns it, 0-host owns it */
+#define SDXC_DES_NUM_SHIFT 16
+#define SDXC_DES_BUFFER_MAX_LEN	(1 << SDXC_DES_NUM_SHIFT)
+	u32 data_buf1_sz:16;
+	u32 data_buf2_sz:16;
+	u32 buf_addr_ptr1;
+	u32 buf_addr_ptr2;
+};
+
+struct sunxi_mmc_host {
+	unsigned mmc_no;
+	uint32_t *mclkreg;

drivers/mmc/sunxi_mmc.c new file +44 -0

+static int mmc_resource_init(int sdc_no)
+{
+	struct sunxi_mmc_host *mmchost = &mmc_host[sdc_no];
+	struct sunxi_ccm_reg *ccm = (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
+
+	debug("init mmc %d resource\n", sdc_no);
+
+	switch (sdc_no) {
+	case 0:
+		mmchost->reg = (struct sunxi_mmc *)SUNXI_MMC0_BASE;
+		mmchost->mclkreg = &ccm->sd0_clk_cfg;
+		break;
+	case 1:
+		mmchost->reg = (struct sunxi_mmc *)SUNXI_MMC1_BASE;
+		mmchost->mclkreg = &ccm->sd1_clk_cfg;
+		break;
+	case 2:
+		mmchost->reg = (struct sunxi_mmc *)SUNXI_MMC2_BASE;
+		mmchost->mclkreg = &ccm->sd2_clk_cfg;
+		break;
+	case 3:
+		mmchost->reg = (struct sunxi_mmc *)SUNXI_MMC3_BASE;
+		mmchost->mclkreg = &ccm->sd3_clk_cfg;
+		break;
+	default:
+		printf("Wrong mmc number %d\n", sdc_no);
+		return -1;
+	}
+	mmchost->database = (unsigned int)mmchost->reg + 0x100;
+	mmchost->mmc_no = sdc_no;
+
+	return 0;
+}
+
+static int mmc_clk_io_on(int sdc_no)
+{
+	unsigned int rval;
+	unsigned int pll5_clk;
+	unsigned int divider;
+	struct sunxi_mmc_host *mmchost = &mmc_host[sdc_no];
+	static struct sunxi_gpio *gpio_c =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_C];
+	static struct sunxi_gpio *gpio_f =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_F];

drivers/mmc/sunxi_mmc.c new file +36 -0

+	debug("init mmc %d clock and io\n", sdc_no);
+
+	/* config gpio */
+	switch (sdc_no) {
+	case 0:
+		/* D1-PF0, D0-PF1, CLK-PF2, CMD-PF3, D3-PF4, D4-PF5 */
+		writel(0x222222, &gpio_f->cfg[0]);
+		writel(0x555, &gpio_f->pull[0]);
+		writel(0xaaa, &gpio_f->drv[0]);
+		break;
+
+	case 1:
+#if CONFIG_MMC1_PG
+		/* PG0-CMD, PG1-CLK, PG2~5-D0~3 : 4 */
+		writel(0x444444, &gpio_g->cfg[0]);
+		writel(0x555, &gpio_g->pull[0]);
+		writel(0xaaa, &gpio_g->drv[0]);
+#else
+		/* PH22-CMD, PH23-CLK, PH24~27-D0~D3 : 5 */
+		writel(0x55 << 24, &gpio_h->cfg[2]);
+		writel(0x5555, &gpio_h->cfg[3]);
+		writel(0x555 << 12, &gpio_h->pull[1]);
+		writel(0xaaa << 12, &gpio_h->drv[1]);
+#endif
+		break;
+
+	case 2:
+		/* CMD-PC6, CLK-PC7, D0-PC8, D1-PC9, D2-PC10, D3-PC11 */
+		writel(0x33 << 24, &gpio_c->cfg[0]);
+		writel(0x3333, &gpio_c->cfg[1]);
+		writel(0x555 << 12, &gpio_c->pull[0]);
+		writel(0xaaa << 12, &gpio_c->drv[0]);
+		break;
+
+	case 3:
+		/* PI4-CMD, PI5-CLK, PI6~9-D0~D3 : 2 */

drivers/mmc/sunxi_mmc.c new file +24 -0

+	/* config ahb clock */
+	rval = readl(&ccm->ahb_gate0);
+	rval |= (1 << (8 + sdc_no));
+	writel(rval, &ccm->ahb_gate0);
+
+	/* config mod clock */
+	pll5_clk = clock_get_pll5();
+	if (pll5_clk > 400000000)
+		divider = 4;
+	else
+		divider = 3;
+	writel((0x1 << 31) | (0x2 << 24) | divider, mmchost->mclkreg);
+	mmchost->mod_clk = pll5_clk / (divider + 1);
+
+	dumphex32("ccmu", (char *)SUNXI_CCM_BASE, 0x100);
+	dumphex32("gpio", (char *)SUNXI_PIO_BASE, 0x100);
+	dumphex32("mmc", (char *)mmchost->reg, 0x100);
+	dumpmmcreg(mmchost->reg);
+
+	return 0;
+}
+
+static int mmc_update_clk(struct mmc *mmc)
+{

drivers/mmc/sunxi_mmc.c new file +17 -0

+	return 0;
+}
+
+static int mmc_config_clock(struct mmc *mmc, unsigned div)
+{
+	struct sunxi_mmc_host *mmchost = (struct sunxi_mmc_host *)mmc->priv;
+	unsigned rval = readl(&mmchost->reg->clkcr);
+
+	/*
+	 * CLKCREG[7:0]: divider
+	 * CLKCREG[16]:  on/off
+	 * CLKCREG[17]:  power save
+	 */
+	/* Disable Clock */
+	rval &= ~(0x1 << 16);
+	writel(rval, &mmchost->reg->clkcr);
+	if (mmc_update_clk(mmc))

drivers/mmc/sunxi_mmc.c new file +14 -0

+	writel(rval, &mmchost->reg->clkcr);
+
+	if (mmc_update_clk(mmc))
+		return -1;
+
+	return 0;
+}
+
+static void mmc_set_ios(struct mmc *mmc)
+{
+	struct sunxi_mmc_host *mmchost = (struct sunxi_mmc_host *)mmc->priv;
+	unsigned int clkdiv = 0;
+
+	debug("set ios: bus_width: %x, clock: %d, mod_clk: %d\n",

drivers/mmc/sunxi_mmc.c new file +6 -0

+	/* Change bus width */
+	if (mmc->bus_width == 8)
+		writel(0x2, &mmchost->reg->width);
+	else if (mmc->bus_width == 4)
+		writel(0x1, &mmchost->reg->width);
+	else

drivers/mmc/sunxi_mmc.c new file +23 -0

+	return 0;
+}
+
+static int mmc_trans_data_by_cpu(struct mmc *mmc, struct mmc_data *data)
+{
+	struct sunxi_mmc_host *mmchost = (struct sunxi_mmc_host *)mmc->priv;
+	unsigned i;
+	unsigned byte_cnt = data->blocksize * data->blocks;
+	unsigned *buff;
+	unsigned timeout = 0xfffff;
+
+	if (data->flags & MMC_DATA_READ) {
+		buff = (unsigned int *)data->dest;
+		for (i = 0; i < (byte_cnt >> 2); i++) {
+			while (--timeout &&
+				 (readl(&mmchost->reg->status) & (0x1 << 2)));
+			if (timeout <= 0)
+				goto out;
+			buff[i] = readl(mmchost->database);
+			timeout = 0xfffff;
+		}
+	} else {
+		buff = (unsigned int *)data->src;

drivers/mmc/sunxi_mmc.c new file +83 -0

+out:
+	if (timeout <= 0)
+		return -1;
+
+	return 0;
+}
+
+static int mmc_trans_data_by_dma(struct mmc *mmc, struct mmc_data *data)
+{
+	struct sunxi_mmc_host *mmchost = (struct sunxi_mmc_host *)mmc->priv;
+	unsigned byte_cnt = data->blocksize * data->blocks;
+	unsigned char *buff;
+	unsigned des_idx = 0;
+	unsigned buff_frag_num =
+		(byte_cnt + SDXC_DES_BUFFER_MAX_LEN - 1) >> SDXC_DES_NUM_SHIFT;
+	unsigned remain;
+	unsigned i, rval;
+	ALLOC_CACHE_ALIGN_BUFFER(struct sunxi_mmc_des, pdes, buff_frag_num);
+
+	buff = data->flags & MMC_DATA_READ ?
+	    (unsigned char *)data->dest : (unsigned char *)data->src;
+	remain = byte_cnt & (SDXC_DES_BUFFER_MAX_LEN - 1);
+	if (!remain)
+		remain = SDXC_DES_BUFFER_MAX_LEN;
+
+	flush_cache((unsigned long)buff, (unsigned long)byte_cnt);
+	for (i = 0; i < buff_frag_num; i++, des_idx++) {
+		memset((void *)&pdes[des_idx], 0, sizeof(struct sunxi_mmc_des));
+		pdes[des_idx].des_chain = 1;
+		pdes[des_idx].own = 1;
+		pdes[des_idx].dic = 1;
+		if (buff_frag_num > 1 && i != buff_frag_num - 1)
+			pdes[des_idx].data_buf1_sz =
+			    (SDXC_DES_BUFFER_MAX_LEN -
+			     1) & SDXC_DES_BUFFER_MAX_LEN;
+		else
+			pdes[des_idx].data_buf1_sz = remain;
+
+		pdes[des_idx].buf_addr_ptr1 =
+		    (u32) buff + i * SDXC_DES_BUFFER_MAX_LEN;
+		if (i == 0)
+			pdes[des_idx].first_des = 1;
+
+		if (i == buff_frag_num - 1) {
+			pdes[des_idx].dic = 0;
+			pdes[des_idx].last_des = 1;
+			pdes[des_idx].end_of_ring = 1;
+			pdes[des_idx].buf_addr_ptr2 = 0;
+		} else {
+			pdes[des_idx].buf_addr_ptr2 = (u32)&pdes[des_idx + 1];
+		}
+		debug("frag %d, remain %d, des[%d](%08x): ",
+		      i, remain, des_idx, (u32)&pdes[des_idx]);
+		debug("[0] = %08x, [1] = %08x, [2] = %08x, [3] = %08x\n",
+		      (u32)((u32 *)&pdes[des_idx])[0],
+		      (u32)((u32 *)&pdes[des_idx])[1],
+		      (u32)((u32 *)&pdes[des_idx])[2],
+		      (u32)((u32 *)&pdes[des_idx])[3]);
+	}
+	flush_cache((unsigned long)pdes,
+		    sizeof(struct sunxi_mmc_des) * (des_idx + 1));
+
+	/*
+	 * GCTRLREG
+	 * GCTRL[2]     : DMA reset
+	 * GCTRL[5]     : DMA enable
+	 *
+	 * IDMACREG
+	 * IDMAC[0]     : IDMA soft reset
+	 * IDMAC[1]     : IDMA fix burst flag
+	 * IDMAC[7]     : IDMA on
+	 *
+	 * IDIECREG
+	 * IDIE[0]      : IDMA transmit interrupt flag
+	 * IDIE[1]      : IDMA receive interrupt flag
+	 */
+	rval = readl(&mmchost->reg->gctrl);
+	/* Enable DMA */
+	writel(rval | (0x1 << 5) | (0x1 << 2), &mmchost->reg->gctrl);
+	/* Reset iDMA */
+	writel((0x1 << 0), &mmchost->reg->dmac);
+	/* Enable iDMA */
+	writel((0x1 << 1) | (1 << 7), &mmchost->reg->dmac);

drivers/mmc/sunxi_mmc.c new file +100 -0

+	return 0;
+}
+
+static int mmc_send_cmd(struct mmc *mmc, struct mmc_cmd *cmd,
+			struct mmc_data *data)
+{
+	struct sunxi_mmc_host *mmchost = (struct sunxi_mmc_host *)mmc->priv;
+	unsigned int cmdval = 0x80000000;
+	signed int timeout = 0;
+	int error = 0;
+	unsigned int status = 0;
+	unsigned int usedma = 0;
+	unsigned int bytecnt = 0;
+
+	if (mmchost->fatal_err)
+		return -1;
+	if (cmd->resp_type & MMC_RSP_BUSY)
+		debug("mmc cmd %d check rsp busy\n", cmd->cmdidx);
+	if (cmd->cmdidx == 12)
+		return 0;
+
+	/*
+	 * CMDREG
+	 * CMD[5:0]     : Command index
+	 * CMD[6]       : Has response
+	 * CMD[7]       : Long response
+	 * CMD[8]       : Check response CRC
+	 * CMD[9]       : Has data
+	 * CMD[10]      : Write
+	 * CMD[11]      : Steam mode
+	 * CMD[12]      : Auto stop
+	 * CMD[13]      : Wait previous over
+	 * CMD[14]      : About cmd
+	 * CMD[15]      : Send initialization
+	 * CMD[21]      : Update clock
+	 * CMD[31]      : Load cmd
+	 */
+	if (!cmd->cmdidx)
+		cmdval |= (0x1 << 15);
+	if (cmd->resp_type & MMC_RSP_PRESENT)
+		cmdval |= (0x1 << 6);
+	if (cmd->resp_type & MMC_RSP_136)
+		cmdval |= (0x1 << 7);
+	if (cmd->resp_type & MMC_RSP_CRC)
+		cmdval |= (0x1 << 8);
+
+	if (data) {
+		if ((u32) data->dest & 0x3) {
+			error = -1;
+			goto out;
+		}
+
+		cmdval |= (0x1 << 9) | (0x1 << 13);
+		if (data->flags & MMC_DATA_WRITE)
+			cmdval |= (0x1 << 10);
+		if (data->blocks > 1)
+			cmdval |= (0x1 << 12);
+		writel(data->blocksize, &mmchost->reg->blksz);
+		writel(data->blocks * data->blocksize, &mmchost->reg->bytecnt);
+	}
+
+	debug("mmc %d, cmd %d(0x%08x), arg 0x%08x\n", mmchost->mmc_no,
+	      cmd->cmdidx, cmdval | cmd->cmdidx, cmd->cmdarg);
+	writel(cmd->cmdarg, &mmchost->reg->arg);
+
+	if (!data)
+		writel(cmdval | cmd->cmdidx, &mmchost->reg->cmd);
+
+	/*
+	 * transfer data and check status
+	 * STATREG[2] : FIFO empty
+	 * STATREG[3] : FIFO full
+	 */
+	if (data) {
+		int ret = 0;
+
+		bytecnt = data->blocksize * data->blocks;
+		debug("trans data %d bytes\n", bytecnt);
+#if defined(CONFIG_MMC_SUNXI_USE_DMA) && !defined(CONFIG_SPL_BUILD)
+		if (bytecnt > 64) {
+#else
+		if (0) {
+#endif
+			usedma = 1;
+			writel(readl(&mmchost->reg->gctrl) & ~(0x1 << 31),
+			       &mmchost->reg->gctrl);
+			ret = mmc_trans_data_by_dma(mmc, data);
+			writel(cmdval | cmd->cmdidx, &mmchost->reg->cmd);
+		} else {
+			writel(readl(&mmchost->reg->gctrl) | 0x1 << 31,
+			       &mmchost->reg->gctrl);
+			writel(cmdval | cmd->cmdidx, &mmchost->reg->cmd);
+			ret = mmc_trans_data_by_cpu(mmc, data);
+		}
+		if (ret) {
+			error = readl(&mmchost->reg->rint) & 0xbfc2;
+			error = TIMEOUT;
+			goto out;
+		}
+	}

drivers/mmc/sunxi_mmc.c new file +8 -0

+	} while (!(status & 0x4));
+
+	if (data) {
+		unsigned done = 0;
+		timeout = usedma ? 0xffff * bytecnt : 0xffff;
+		debug("cacl timeout %x\n", timeout);
+		do {
+			status = readl(&mmchost->reg->rint);

drivers/mmc/sunxi_mmc.c new file +58 -0

+	}
+
+	if (cmd->resp_type & MMC_RSP_136) {
+		cmd->response[0] = readl(&mmchost->reg->resp3);
+		cmd->response[1] = readl(&mmchost->reg->resp2);
+		cmd->response[2] = readl(&mmchost->reg->resp1);
+		cmd->response[3] = readl(&mmchost->reg->resp0);
+		debug("mmc resp 0x%08x 0x%08x 0x%08x 0x%08x\n",
+		      cmd->response[3], cmd->response[2],
+		      cmd->response[1], cmd->response[0]);
+	} else {
+		cmd->response[0] = readl(&mmchost->reg->resp0);
+		debug("mmc resp 0x%08x\n", cmd->response[0]);
+	}
+out:
+	if (data && usedma) {
+		/* IDMASTAREG
+		 * IDST[0] : idma tx int
+		 * IDST[1] : idma rx int
+		 * IDST[2] : idma fatal bus error
+		 * IDST[4] : idma descriptor invalid
+		 * IDST[5] : idma error summary
+		 * IDST[8] : idma normal interrupt sumary
+		 * IDST[9] : idma abnormal interrupt sumary
+		 */
+		status = readl(&mmchost->reg->idst);
+		writel(status, &mmchost->reg->idst);
+		writel(0, &mmchost->reg->idie);
+		writel(0, &mmchost->reg->dmac);
+		writel(readl(&mmchost->reg->gctrl) & ~(0x1 << 5),
+		       &mmchost->reg->gctrl);
+	}
+	if (error < 0) {
+		writel(0x7, &mmchost->reg->gctrl);
+		mmc_update_clk(mmc);
+	}
+	writel(0xffffffff, &mmchost->reg->rint);
+	writel(readl(&mmchost->reg->gctrl) | (1 << 1), &mmchost->reg->gctrl);
+
+	return error;
+}
+
+int sunxi_mmc_init(int sdc_no)
+{
+	struct mmc *mmc;
+
+	memset(&mmc_dev[sdc_no], 0, sizeof(struct mmc));
+	memset(&mmc_host[sdc_no], 0, sizeof(struct sunxi_mmc_host));
+	mmc = &mmc_dev[sdc_no];
+
+	sprintf(mmc->name, "SUNXI SD/MMC");
+	mmc->priv = &mmc_host[sdc_no];
+	mmc->send_cmd = mmc_send_cmd;
+	mmc->set_ios = mmc_set_ios;
+	mmc->init = mmc_core_init;
+
+	mmc->voltages = MMC_VDD_32_33 | MMC_VDD_33_34;
+	mmc->host_caps = MMC_MODE_4BIT;

drivers/mmc/sunxi_mmc.c new file +2 -0

+	return 0;
+}

include/configs/sunxi-common.h +11 -0

index bdfc169..1afdeb2 100644
@@ -74,6 +74,16 @@
 #define CONFIG_INITRD_TAG
 #define CONFIG_CMDLINE_EDITING
 
+/* mmc config */
+#define CONFIG_MMC
+#define CONFIG_GENERIC_MMC
+#define CONFIG_CMD_MMC
+#define CONFIG_MMC_SUNXI
+#define CONFIG_MMC_SUNXI_SLOT		0
+#define CONFIG_MMC_SUNXI_USE_DMA
+#define CONFIG_ENV_IS_IN_MMC
+#define CONFIG_SYS_MMC_ENV_DEV		0	/* first detected MMC controller */
+
 /*
 * Size of malloc() pool
 * 1MB = 0x100000, 0x100000 = 1024 * 1024

@@ -118,6 +128,7 @@
 #define CONFIG_SYS_MONITOR_LEN		(512 << 10)	/* 512 KiB */
 #define CONFIG_IDENT_STRING		" Allwinner Technology"
 
+#define CONFIG_ENV_OFFSET		(544 << 10) /* (8 + 24 + 512) KiB */
 #define CONFIG_ENV_SIZE			(128 << 10)	/* 128 KiB */
 
 #define CONFIG_BOOTDELAY	3

Signature · 6 lines

1.8.5.3

_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot

[...]

+

drivers/mmc/sunxi_mmc.c line 56

Quoted 8 lines

drivers/mmc/sunxi_mmc.c new file +8 -0

+static void dumpmmcreg(struct sunxi_mmc *reg)
+{
+	debug("dump mmc registers:\n");
+	debug("gctrl     0x%08x\n", reg->gctrl);
+	debug("clkcr     0x%08x\n", reg->clkcr);
+	debug("timeout   0x%08x\n", reg->timeout);
+	debug("width     0x%08x\n", reg->width);
+	debug("blksz     0x%08x\n", reg->blksz);

[...] lots more debug(foo)

Quoted 1 line

+}

^^^ #ifdef DEBUG here?

[...]

drivers/mmc/sunxi_mmc.c line 110

Quoted 4 lines

drivers/mmc/sunxi_mmc.c new file +4 -0

+/* support 4 mmc hosts */
+struct mmc mmc_dev[4];
+struct sunxi_mmc_host mmc_host[4];
+

^ hosts & mmc structs can be allocated even for SPL now

Quoted 9 lines

drivers/mmc/sunxi_mmc.c new file +9 -0

+	struct sunxi_mmc_host *mmchost = &mmc_host[sdc_no];
+	static struct sunxi_gpio *gpio_c =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_C];
+	static struct sunxi_gpio *gpio_f =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_F];
+#if CONFIG_MMC1_PG
+	static struct sunxi_gpio *gpio_g =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_G];
+#endif

drivers/mmc/sunxi_mmc.c line 164

Quoted 6 lines

drivers/mmc/sunxi_mmc.c new file +6 -0

+	static struct sunxi_gpio *gpio_h =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_H];
+	static struct sunxi_gpio *gpio_i =
+	    &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_I];
+	struct sunxi_ccm_reg *ccm = (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
+

^ Castings are ugly; rework with a temporary variable.

[...]

drivers/mmc/sunxi_mmc.c line 210

Quoted 12 lines

drivers/mmc/sunxi_mmc.c new file +12 -0

+	case 3:
+		/* PI4-CMD, PI5-CLK, PI6~9-D0~D3 : 2 */
+		writel(0x2222 << 16, &gpio_i->cfg[0]);
+		writel(0x22, &gpio_i->cfg[1]);
+		writel(0x555 << 8, &gpio_i->pull[0]);
+		writel(0x555 << 8, &gpio_i->drv[0]);
+		break;
+
+	default:
+		return -1;
+	}
+

Lots of magic constants. I have no idea what's going on here. Use a few defines.

[...]> + cmd = (0x1 << 31) | (0x1 << 21) | (0x1 << 13);

drivers/mmc/sunxi_mmc.c line 244

Quoted 5 lines

drivers/mmc/sunxi_mmc.c new file +5 -0

+	writel(cmd, &mmchost->reg->cmd);
+	while ((readl(&mmchost->reg->cmd) & (0x1 << 31)) && timeout--);
+	if (!timeout)
+		return -1;
+

^ Eeek! Use udelay and a time based timeout.

drivers/mmc/sunxi_mmc.c line 246

Quoted 2 lines

drivers/mmc/sunxi_mmc.c new file +2 -0

+	writel(readl(&mmchost->reg->rint), &mmchost->reg->rint);
+

^ Same here - Not even a timeout?

drivers/mmc/sunxi_mmc.c line 273

Quoted 1 line

drivers/mmc/sunxi_mmc.c new file +1 -0

+	rval |= (0x1 << 16);

#define ?

Ambiguous formatting. Use { }

[...]

drivers/mmc/sunxi_mmc.c line 313

Quoted 3 lines

drivers/mmc/sunxi_mmc.c new file +3 -0

+	/* Reset controller */
+	writel(0x7, &mmchost->reg->gctrl);
+

Magic value again.

Quoted 6 lines

drivers/mmc/sunxi_mmc.c new file +6 -0

+	} else {
+		buff = (unsigned int *)data->src;
+		for (i = 0; i < (byte_cnt >> 2); i++) {
+			while (--timeout &&
+				 (readl(&mmchost->reg->status) & (0x1 << 3)));
+			if (timeout <= 0)

drivers/mmc/sunxi_mmc.c line 346

Quoted 6 lines

drivers/mmc/sunxi_mmc.c new file +6 -0

+				goto out;
+			writel(buff[i], mmchost->database);
+			timeout = 0xfffff;
+		}
+	}
+

^ Timeouts using time values? udelay? See above.

[....]

drivers/mmc/sunxi_mmc.c line 439

Quoted 5 lines

drivers/mmc/sunxi_mmc.c new file +5 -0

+	writel(rval, &mmchost->reg->idie);
+	writel((u32) pdes, &mmchost->reg->dlba);
+	writel((0x2 << 28) | (0x7 << 16) | (0x01 << 3),
+	       &mmchost->reg->ftrglevel);
+

^ #define again?

[...]

drivers/mmc/sunxi_mmc.c line 549

Quoted 9 lines

drivers/mmc/sunxi_mmc.c new file +9 -0

+	timeout = 0xfffff;
+	do {
+		status = readl(&mmchost->reg->rint);
+		if (!timeout-- || (status & 0xbfc2)) {
+			error = status & 0xbfc2;
+			debug("cmd timeout %x\n", error);
+			error = TIMEOUT;
+			goto out;
+		}

^ Again timeouts without using time values.

[...]

Quoted 20 lines

drivers/mmc/sunxi_mmc.c new file +20 -0

+int sunxi_mmc_init(int sdc_no)
+{
+	struct mmc *mmc;
+
+	memset(&mmc_dev[sdc_no], 0, sizeof(struct mmc));
+	memset(&mmc_host[sdc_no], 0, sizeof(struct sunxi_mmc_host));
+	mmc = &mmc_dev[sdc_no];
+
+	sprintf(mmc->name, "SUNXI SD/MMC");
+	mmc->priv = &mmc_host[sdc_no];
+	mmc->send_cmd = mmc_send_cmd;
+	mmc->set_ios = mmc_set_ios;
+	mmc->init = mmc_core_init;
+
+	mmc->voltages = MMC_VDD_32_33 | MMC_VDD_33_34;
+	mmc->host_caps = MMC_MODE_4BIT;
+	mmc->host_caps |= MMC_MODE_HS_52MHz | MMC_MODE_HS;
+
+	mmc->f_min = 400000;
+	mmc->f_max = 52000000;

drivers/mmc/sunxi_mmc.c line 648

Quoted 6 lines

drivers/mmc/sunxi_mmc.c new file +6 -0

+
+	mmc_resource_init(sdc_no);
+	mmc_clk_io_on(sdc_no);
+
+	mmc_register(mmc);
+

^ The mmc_registeration sequence has changed, but not landed at master yet.

Quoted 1 line · Pantelis Antoniou

[...]

Thanks for your review. It seems there are still quite a few issues dating back to the original allwinner dumps here.

@linux-sunxi: if anyone wants to volunteer to help cleanup this particular driver I'd be very happy -- there's a lot of it!

Quoted 10 lines

+

drivers/mmc/sunxi_mmc.c line 56

Quoted 8 lines

drivers/mmc/sunxi_mmc.c new file +8 -0

+static void dumpmmcreg(struct sunxi_mmc *reg)
+{
+   debug("dump mmc registers:\n");
+   debug("gctrl     0x%08x\n", reg->gctrl);
+   debug("clkcr     0x%08x\n", reg->clkcr);
+   debug("timeout   0x%08x\n", reg->timeout);
+   debug("width     0x%08x\n", reg->width);
+   debug("blksz     0x%08x\n", reg->blksz);

[...] lots more debug(foo)

Quoted 3 lines

Quoted 1 line

+}

^^^ #ifdef DEBUG here?

I can if you prefer but debug() itself effectively includes the same ifdef so the end result is already the same.

Quoted 8 lines

[...]

drivers/mmc/sunxi_mmc.c line 110

Quoted 4 lines

drivers/mmc/sunxi_mmc.c new file +4 -0

+/* support 4 mmc hosts */
+struct mmc mmc_dev[4];
+struct sunxi_mmc_host mmc_host[4];
+

^ hosts & mmc structs can be allocated even for SPL now

Can be or must be?

Quoted 9 lines

Quoted 9 lines

drivers/mmc/sunxi_mmc.c new file +9 -0

+   struct sunxi_mmc_host *mmchost = &mmc_host[sdc_no];
+   static struct sunxi_gpio *gpio_c =
+       &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_C];
+   static struct sunxi_gpio *gpio_f =
+       &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_F];
+#if CONFIG_MMC1_PG
+   static struct sunxi_gpio *gpio_g =
+       &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_G];
+#endif

Quoted 8 lines

drivers/mmc/sunxi_mmc.c line 164

Quoted 6 lines

drivers/mmc/sunxi_mmc.c new file +6 -0

+   static struct sunxi_gpio *gpio_h =
+       &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_H];
+   static struct sunxi_gpio *gpio_i =
+       &((struct sunxi_gpio_reg *)SUNXI_PIO_BASE)->gpio_bank[SUNXI_GPIO_I];
+   struct sunxi_ccm_reg *ccm = (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
+

^ Castings are ugly; rework with a temporary variable.

Ack.

The static's here are odd too and date back to the original alwinner code dumps. I'll get rid of them too.

Quoted 16 lines

[...]

drivers/mmc/sunxi_mmc.c line 210

Quoted 12 lines

drivers/mmc/sunxi_mmc.c new file +12 -0

+   case 3:
+           /* PI4-CMD, PI5-CLK, PI6~9-D0~D3 : 2 */
+           writel(0x2222 << 16, &gpio_i->cfg[0]);
+           writel(0x22, &gpio_i->cfg[1]);
+           writel(0x555 << 8, &gpio_i->pull[0]);
+           writel(0x555 << 8, &gpio_i->drv[0]);
+           break;
+
+   default:
+           return -1;
+   }
+

Lots of magic constants. I have no idea what's going on here. Use a few defines.

Right. These came from the original allwinner dumps so I was worried that they might be undocumented magic, but actually since the are gpio frobbing I reckon I can figure them out.

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

Quoted 15 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Quoted 3 lines

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

Quoted 20 lines · Tom Rini

Quoted 15 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Quoted 3 lines

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

Quoted 6 lines

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

So, I'm confused about what to do here ;-)

AFAICS none of the existing platforms in u-boot.git are setting bootm_low in their default environment.

README suggests that bootm_low is the lowest address allowed for use by bootm/z rather than the limit, from the docs it seems that CONFIG_SYS_BOOTMAPSZ (overridden by env bootm_mapsize) is the limit

bootm_low defaults to CONFIG_SYS_SDRAM_BASE, which sunxi sets, and this seems logical to me since kernel's lowmem mapping starts at offset 0 into RAM.

I think we probably do want to set BOOTMAPSZ to something like 256MB, which seems to be the highest in tree (although the vast majority use 8MB...). But I'm not sure that explains why fdt_high is needed today. I'll have a play.

Quoted 28 lines · Ian Campbell

Quoted 26 lines · Tom Rini

Quoted 15 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Quoted 3 lines

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

Quoted 4 lines

So, I'm confused about what to do here ;-)

AFAICS none of the existing platforms in u-boot.git are setting bootm_low in their default environment.

Quoted 4 lines

README suggests that bootm_low is the lowest address allowed for use by bootm/z rather than the limit, from the docs it seems that CONFIG_SYS_BOOTMAPSZ (overridden by env bootm_mapsize) is the limit

Quoted 8 lines

bootm_low defaults to CONFIG_SYS_SDRAM_BASE, which sunxi sets, and this seems logical to me since kernel's lowmem mapping starts at offset 0 into RAM.

I think we probably do want to set BOOTMAPSZ to something like 256MB, which seems to be the highest in tree (although the vast majority use 8MB...). But I'm not sure that explains why fdt_high is needed today. I'll have a play.

So setting BOOTMAPSZ to e.g. 256MB limits the fdt load address as you would expect.

But it seems to not make any difference for the initramfs which still gets loaded to the top of RAM. This is consistent with README which says that unless initrd_high is set the ramdisk will be loaded as high as possible, without reference to BOOTMAPSZ. Supposedly setting initrd_high to "no", "off" or "0" should cause initrd relocation to obey BOOTMAPSZ -- but at least in my experiments it does not seem to do so.

boot_ramdisk_high() doesn't seem to make any reference to bootm_{low,size,etc} like I would expect and with initrd_high==0 it calls lmb_alloc which allocates anywhere. This is in contrast with boot_relocate_fdt which uses getenv_bootm_mapsize() and getenv_bootm_low().

It looks to me like the initrd reloc logic is wrong -- but it's been that way for such a long time I think I must be mistaken.

Quoted 29 lines · Ian Campbell

Quoted 28 lines · Ian Campbell

Quoted 26 lines · Tom Rini

Quoted 15 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

Review

Quoted 3 lines

(quoted diff) +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Quoted 3 lines

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

Quoted 4 lines

Quoted 4 lines

So, I'm confused about what to do here ;-)

AFAICS none of the existing platforms in u-boot.git are setting bootm_low in their default environment.

Yes. http://patchwork.ozlabs.org/patch/329210/ gets things right for some subset of TI platforms and I'll be picking that up soon. Or maybe a v2.

Quoted 29 lines

Quoted 4 lines

README suggests that bootm_low is the lowest address allowed for use by bootm/z rather than the limit, from the docs it seems that CONFIG_SYS_BOOTMAPSZ (overridden by env bootm_mapsize) is the limit

Quoted 8 lines

bootm_low defaults to CONFIG_SYS_SDRAM_BASE, which sunxi sets, and this seems logical to me since kernel's lowmem mapping starts at offset 0 into RAM.

I think we probably do want to set BOOTMAPSZ to something like 256MB, which seems to be the highest in tree (although the vast majority use 8MB...). But I'm not sure that explains why fdt_high is needed today. I'll have a play.

So setting BOOTMAPSZ to e.g. 256MB limits the fdt load address as you would expect.

But it seems to not make any difference for the initramfs which still gets loaded to the top of RAM. This is consistent with README which says that unless initrd_high is set the ramdisk will be loaded as high as possible, without reference to BOOTMAPSZ. Supposedly setting initrd_high to "no", "off" or "0" should cause initrd relocation to obey BOOTMAPSZ -- but at least in my experiments it does not seem to do so.

boot_ramdisk_high() doesn't seem to make any reference to bootm_{low,size,etc} like I would expect and with initrd_high==0 it calls lmb_alloc which allocates anywhere. This is in contrast with boot_relocate_fdt which uses getenv_bootm_mapsize() and getenv_bootm_low().

Quoted 2 lines

It looks to me like the initrd reloc logic is wrong -- but it's been that way for such a long time I think I must be mistaken.

I think after reading what you're saying, things are very unclear, unfortunately.

What you can do, but is very odd after thinking about it
is if you set bootm_low only and not any of the size things, it works as a constraint on the upper bounds so initrd nor fdt are put above that. But that's really seeming like a side effect rather than an intent.

I'm going to whack at this more..

Quoted 35 lines · Tom Rini

Quoted 33 lines · Ian Campbell

Quoted 32 lines · Ian Campbell

Quoted 20 lines · Tom Rini

Quoted 15 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

Review

Quoted 3 lines

(quoted diff) +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Quoted 3 lines

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

Quoted 6 lines

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

So, I'm confused about what to do here ;-)

AFAICS none of the existing platforms in u-boot.git are setting bootm_low in their default environment.

Quoted 3 lines

Yes. http://patchwork.ozlabs.org/patch/329210/ gets things right for some subset of TI platforms and I'll be picking that up soon. Or maybe a v2.

You are dropping fdt_high=0xffffffff here, is that deliberate?

I suppose this is offset by the addition of bootm_size which keeps things working? (or does it, I'm very confused now...)

Quoted 29 lines

Quoted 29 lines

Quoted 4 lines

README suggests that bootm_low is the lowest address allowed for use by bootm/z rather than the limit, from the docs it seems that CONFIG_SYS_BOOTMAPSZ (overridden by env bootm_mapsize) is the limit

Quoted 8 lines

bootm_low defaults to CONFIG_SYS_SDRAM_BASE, which sunxi sets, and this seems logical to me since kernel's lowmem mapping starts at offset 0 into RAM.

I think we probably do want to set BOOTMAPSZ to something like 256MB, which seems to be the highest in tree (although the vast majority use 8MB...). But I'm not sure that explains why fdt_high is needed today. I'll have a play.

So setting BOOTMAPSZ to e.g. 256MB limits the fdt load address as you would expect.

But it seems to not make any difference for the initramfs which still gets loaded to the top of RAM. This is consistent with README which says that unless initrd_high is set the ramdisk will be loaded as high as possible, without reference to BOOTMAPSZ. Supposedly setting initrd_high to "no", "off" or "0" should cause initrd relocation to obey BOOTMAPSZ -- but at least in my experiments it does not seem to do so.

boot_ramdisk_high() doesn't seem to make any reference to bootm_{low,size,etc} like I would expect and with initrd_high==0 it calls lmb_alloc which allocates anywhere. This is in contrast with boot_relocate_fdt which uses getenv_bootm_mapsize() and getenv_bootm_low().

Quoted 5 lines

Quoted 2 lines

It looks to me like the initrd reloc logic is wrong -- but it's been that way for such a long time I think I must be mistaken.

I think after reading what you're saying, things are very unclear, unfortunately.

I'm glad it's not just me!

Quoted 4 lines

What you can do, but is very odd after thinking about it
is if you set bootm_low only and not any of the size things, it works as a constraint on the upper bounds so initrd nor fdt are put above that. But that's really seeming like a side effect rather than an intent.

Right, I don't think I want to add any weirdness, unless all the cool kids are going to be doing it of course ;-)

Quoted 1 line

I'm going to whack at this more..

Thanks. I think for now with this series I can just set BOOTMAPSZ to 256MB without making this problem any better or worse and we can fixup things later in whatever is determined to be the right solution. Is that OK with you?

Quoted 20 lines · Tom Rini

Quoted 18 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

Quoted 6 lines

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

So, I'm confused about what to do here ;-)

Quoted 21 lines · Ian Campbell

Quoted 20 lines · Tom Rini

Quoted 18 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

include/configs/sunxi-common.h line 103

Quoted 3 lines

include/configs/sunxi-common.h +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

Quoted 8 lines

Quoted 6 lines

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

So, I'm confused about what to do here ;-)

And I've not made things clearer with a mis-recollection of things. I don't know why I keep saying "bootm_low" when I mean "bootm_size" like I've done in later patches (and thankfully, when poking people on G+).

There's three ways to say "Please ensure that the FDT and if passed initrd do not relocate above a certain location".
1) In the environment, set bootm_size to kernel lowmem. This means that boot_start_lmb restricts the pool used by both fdt and initrd to that value at the top.
2) In the environment, set fdt_high (and if using initrd, initrd_high) to the top of lowmem. This means that we'll make sure they don't get relocated above that value.
3) In the environment set initrd_high to top of lowmem and set bootm_mapsize to lowmem.

In all of the above, lowmem can be replaced with any valid size that's also smaller than lowmem, such as 256MB. Another option, in the environment, is to set initrd_high and fdt_high to 0xffffffff and then relocation is disabled. I would _not_ recommend this in the general case as one of the points of relocation is to ensure we don't get overwritten by the kernel BSS.

Quoted 46 lines · Tom Rini

Quoted 21 lines · Ian Campbell

Quoted 20 lines · Tom Rini

Quoted 18 lines · Hans de Goede

Hi,

Quoted 11 lines · Tom Rini

Quoted 2 lines · Ian Campbell

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

[snip]

Review

Quoted 3 lines

(quoted diff) +3 -0

+/* Flat Device Tree (FDT/DT) support */
+#define CONFIG_OF_LIBFDT
+#define CONFIG_SYS_BOOTMAPSZ		(16 << 20)

This seems pretty small. This is to keep things from being relocated into highmem right?

Hmm, this reminds me that we currently need to do a "env set fdt_high ffffffff" in our boot scripts to get ftd to work at all. Would be nice to fix this for upstream. I'm afraid I'm clueless as to why we (sunxi) need it, but we do.

Quoted 8 lines

Quoted 6 lines

You want to be setting bootm_low (even for bootz, it's about the underlying boot mechanics that bootz and bootelf and ... hook into) to the amount of lowmem the kernel will have. We do this because we do want to make sure that the device tree isn't overwritten by the kernel BSS or similar. Everyone with more DDR than kernel lowmem needs to be doing something along these lines.

So, I'm confused about what to do here ;-)

And I've not made things clearer with a mis-recollection of things. I don't know why I keep saying "bootm_low" when I mean "bootm_size" like I've done in later patches (and thankfully, when poking people on G+).

There's three ways to say "Please ensure that the FDT and if passed initrd do not relocate above a certain location".
1) In the environment, set bootm_size to kernel lowmem. This means that boot_start_lmb restricts the pool used by both fdt and initrd to that value at the top.
2) In the environment, set fdt_high (and if using initrd, initrd_high) to the top of lowmem. This means that we'll make sure they don't get relocated above that value.
3) In the environment set initrd_high to top of lowmem and set bootm_mapsize to lowmem.

Quoted 6 lines

In all of the above, lowmem can be replaced with any valid size that's also smaller than lowmem, such as 256MB. Another option, in the environment, is to set initrd_high and fdt_high to 0xffffffff and then relocation is disabled. I would _not_ recommend this in the general case as one of the points of relocation is to ensure we don't get overwritten by the kernel BSS.

IIRC, we added it not to make sure that it was in lowmem, but because we had issues with large enough kernels (multi_v7_defconfig + embedded initramfs) that would overwrite the relocated DTB whenever it was decompressing itself.

Quoted 1 line

So, I'm confused about what to do here ;-)

And I've not made things clearer with a mis-recollection of things. I don't know why I keep saying "bootm_low" when I mean "bootm_size" like I've done in later patches (and thankfully, when poking people on G+).

There's three ways to say "Please ensure that the FDT and if passed initrd do not relocate above a certain location".
1) In the environment, set bootm_size to kernel lowmem. This means that boot_start_lmb restricts the pool used by both fdt and initrd to that value at the top.
2) In the environment, set fdt_high (and if using initrd, initrd_high) to the top of lowmem. This means that we'll make sure they don't get relocated above that value.
3) In the environment set initrd_high to top of lowmem and set bootm_mapsize to lowmem.

In all of the above, lowmem can be replaced with any valid size that's also smaller than lowmem, such as 256MB.

Another option, in the
environment, is to set initrd_high and fdt_high to 0xffffffff and then relocation is disabled. I would _not_ recommend this in the general case as one of the points of relocation is to ensure we don't get overwritten by the kernel BSS.

Based linux-sunxi#sunxi commit d854c4de2f57 "arm: Handle .gnu.hash section in ldscripts" vs v2014.01.

Signed-off-by Chen-Yu Tsai <[email protected]>

Signed-off-by Jens Kuske <[email protected]>

Signed-off-by Ian Campbell <[email protected]>

---
 arch/arm/cpu/armv7/sunxi/board.c | 15 +++++++++
 boards.cfg                       |  4 +--
 drivers/net/Makefile             |  1 +
 drivers/net/sunxi_gmac.c         | 34 ++++++++++++++++++++
 include/configs/sunxi-common.h   | 68 ++++++++++++++++++++++++++++++++++++++++
 5 files changed, 120 insertions(+), 2 deletions(-)
 create mode 100644 drivers/net/sunxi_gmac.c

drivers/net/sunxi_gmac.c new file +21 -0

new file mode 100644
index 0000000..432d7b2
@@ -0,0 +1,34 @@
+#include <common.h>
+#include <netdev.h>
+#include <miiphy.h>
+#include <asm/gpio.h>
+#include <asm/io.h>
+#include <asm/arch/clock.h>
+#include <asm/arch/gpio.h>
+
+int sunxi_gmac_initialize(bd_t *bis)
+{
+       int pin;
+       struct sunxi_ccm_reg *const ccm =
+               (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
+
+       /* Set up clock gating */
+       setbits_le32(&ccm->ahb_gate1, 0x1 << AHB_GATE_OFFSET_GMAC);
+
+       /* Set MII clock */
+       setbits_le32(&ccm->gmac_clk_cfg, CCM_GMAC_CTRL_TX_CLK_SRC_INT_RGMII |
+               CCM_GMAC_CTRL_GPIT_RGMII);
+

drivers/net/sunxi_gmac.c new file +3 -0

+
+       return 0;
+}

Thanks for working on this!

I see you left out all the CONFIG_RGMII ifdefs from this file. Not sure if it's because you're aiming to support only the Cubietruck first.

I think you should keep them in the same patch. You can then just add the appropriate config options when support of other A20 boards roll in.

There are 3 changes to the designware driver since u-boot-sunxi's last merge (2014.1-rc1):

50b0df8 net/designware: make driver compatible with data cache 92a190a net/designware - switch driver to phylib usage 74cb708 net/designware: add explicit reset of {tx|rx}_currdescnum

The move to phylib will likely break the current code. Maybe we could merge 2014.4-rc1 in to fix it first.

Also, GMAC support depends on

5a51af3 net/designware: Reduce DMA burst length

by Jens Kuske (jemk). Neither of us bothered to send this upstream yet.

On Fri, 2014-03-14 at 19:11 +0800, Chen-Yu Tsai wrote:

Quoted 3 lines

I think you should keep them in the same patch. You can then just add the appropriate config options when support of other A20 boards roll in.

I don't have any boards which require this, so I cannot test it. I'd prefer to leave it to whoever adds support for the first such board to put this stuff back in, it's a pretty trivial git diff invocation to spot the needed bits.

Quoted 1 line · Ian Campbell

On Fri, 2014-03-14 at 19:11 +0800, Chen-Yu Tsai wrote:

[snip]

Quoted 4 lines

Quoted 3 lines

I think you should keep them in the same patch. You can then just add the appropriate config options when support of other A20 boards roll in.

Quoted 4 lines

I don't have any boards which require this, so I cannot test it. I'd prefer to leave it to whoever adds support for the first such board to put this stuff back in, it's a pretty trivial git diff invocation to spot the needed bits.

I agree with this as well.

Quoted 12 lines · Tom Rini

Quoted 1 line · Ian Campbell

On Fri, 2014-03-14 at 19:11 +0800, Chen-Yu Tsai wrote:

[snip]

Quoted 8 lines

Quoted 3 lines

I think you should keep them in the same patch. You can then just add the appropriate config options when support of other A20 boards roll in.

I don't have any boards which require this, so I cannot test it. I'd prefer to leave it to whoever adds support for the first such board to put this stuff back in, it's a pretty trivial git diff invocation to spot the needed bits.

Quoted 1 line

I agree with this as well.

With the request to add RGMII here or my reasons for not doing so? (sorry if I'm being dense)

include/configs/sunxi-common.h +15 -0

+
+#ifdef CONFIG_CMD_NET
+#define CONFIG_CMD_PING
+#define CONFIG_CMD_DHCP
+#define CONFIG_CMD_NFS
+#define CONFIG_CMD_SNTP
+#define CONFIG_TIMESTAMP		/* Needed by SNTP */
+#define CONFIG_CMD_DNS
+#define CONFIG_NETCONSOLE
+#define CONFIG_BOOTP_SUBNETMASK
+#define CONFIG_BOOTP_GATEWAY
+#define CONFIG_BOOTP_HOSTNAME
+#define CONFIG_BOOTP_NISDOMAIN
+#define CONFIG_BOOTP_BOOTPATH
+#define CONFIG_BOOTP_BOOTFILESIZE