u-boot/drivers/spi/mpc8xxx_spi.c

299 lines
7.1 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0+
/*
* Copyright (c) 2006 Ben Warren, Qstreams Networks Inc.
* With help from the common/soft_spi and arch/powerpc/cpu/mpc8260 drivers
*/
#include <common.h>
#include <clk.h>
#include <dm.h>
#include <errno.h>
#include <log.h>
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
#include <malloc.h>
#include <spi.h>
#include <asm/mpc8xxx_spi.h>
#include <asm-generic/gpio.h>
#include <dm/device_compat.h>
#include <linux/bitops.h>
#include <linux/delay.h>
#include <asm/arch/soc.h>
enum {
SPI_EV_NE = BIT(31 - 22), /* Receiver Not Empty */
SPI_EV_NF = BIT(31 - 23), /* Transmitter Not Full */
};
enum {
SPI_MODE_LOOP = BIT(31 - 1), /* Loopback mode */
SPI_MODE_CI = BIT(31 - 2), /* Clock invert */
SPI_MODE_CP = BIT(31 - 3), /* Clock phase */
SPI_MODE_DIV16 = BIT(31 - 4), /* Divide clock source by 16 */
SPI_MODE_REV = BIT(31 - 5), /* Reverse mode - MSB first */
SPI_MODE_MS = BIT(31 - 6), /* Always master */
SPI_MODE_EN = BIT(31 - 7), /* Enable interface */
SPI_MODE_OP = BIT(31 - 17), /* CPU Mode, QE otherwise */
SPI_MODE_LEN_MASK = 0xf00000,
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
SPI_MODE_LEN_SHIFT = 20,
SPI_MODE_PM_SHIFT = 16,
SPI_MODE_PM_MASK = 0xf0000,
SPI_COM_LST = BIT(31 - 9),
};
struct mpc8xxx_priv {
spi8xxx_t *spi;
struct gpio_desc gpios[16];
int cs_count;
ulong clk_rate;
};
#define SPI_TIMEOUT 1000
static int mpc8xxx_spi_of_to_plat(struct udevice *dev)
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
{
struct mpc8xxx_priv *priv = dev_get_priv(dev);
struct clk clk;
int ret;
priv->spi = dev_read_addr_ptr(dev);
ret = gpio_request_list_by_name(dev, "gpios", priv->gpios,
ARRAY_SIZE(priv->gpios), GPIOD_IS_OUT | GPIOD_ACTIVE_LOW);
if (ret < 0)
return -EINVAL;
priv->cs_count = ret;
ret = clk_get_by_index(dev, 0, &clk);
if (ret) {
dev_err(dev, "%s: clock not defined\n", __func__);
return ret;
}
priv->clk_rate = clk_get_rate(&clk);
if (!priv->clk_rate) {
dev_err(dev, "%s: failed to get clock rate\n", __func__);
return -EINVAL;
}
return 0;
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
}
static int mpc8xxx_spi_probe(struct udevice *dev)
{
struct mpc8xxx_priv *priv = dev_get_priv(dev);
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
spi8xxx_t *spi = priv->spi;
/*
* SPI pins on the MPC83xx are not muxed, so all we do is initialize
* some registers
*/
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
out_be32(&priv->spi->mode, SPI_MODE_REV | SPI_MODE_MS);
if (dev_get_driver_data(dev) == SOC_MPC832X)
setbits_be32(&priv->spi->mode, SPI_MODE_OP);
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
/* set len to 8 bits */
setbits_be32(&spi->mode, (8 - 1) << SPI_MODE_LEN_SHIFT);
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
setbits_be32(&spi->mode, SPI_MODE_EN);
/* Clear all SPI events */
setbits_be32(&priv->spi->event, 0xffffffff);
/* Mask all SPI interrupts */
clrbits_be32(&priv->spi->mask, 0xffffffff);
/* LST bit doesn't do anything, so disregard */
out_be32(&priv->spi->com, 0);
return 0;
}
static void mpc8xxx_spi_cs_activate(struct udevice *dev)
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
{
struct mpc8xxx_priv *priv = dev_get_priv(dev->parent);
struct dm_spi_slave_plat *plat = dev_get_parent_plat(dev);
dm_gpio_set_value(&priv->gpios[plat->cs], 1);
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
}
static void mpc8xxx_spi_cs_deactivate(struct udevice *dev)
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
{
struct mpc8xxx_priv *priv = dev_get_priv(dev->parent);
struct dm_spi_slave_plat *plat = dev_get_parent_plat(dev);
dm_gpio_set_value(&priv->gpios[plat->cs], 0);
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
}
static int mpc8xxx_spi_xfer(struct udevice *dev, uint bitlen,
const void *dout, void *din, ulong flags)
{
struct udevice *bus = dev->parent;
struct mpc8xxx_priv *priv = dev_get_priv(bus);
spi8xxx_t *spi = priv->spi;
struct dm_spi_slave_plat *plat = dev_get_parent_plat(dev);
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
u32 tmpdin = 0, tmpdout = 0, n;
const u8 *cout = dout;
u8 *cin = din;
ulong type = dev_get_driver_data(bus);
debug("%s: slave %s:%u dout %08X din %08X bitlen %u\n", __func__,
bus->name, plat->cs, (uint)dout, (uint)din, bitlen);
if (plat->cs >= priv->cs_count) {
dev_err(dev, "chip select index %d too large (cs_count=%d)\n",
plat->cs, priv->cs_count);
return -EINVAL;
}
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
if (bitlen % 8) {
printf("*** spi_xfer: bitlen must be multiple of 8\n");
return -ENOTSUPP;
}
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
if (flags & SPI_XFER_BEGIN)
mpc8xxx_spi_cs_activate(dev);
/* Clear all SPI events */
setbits_be32(&spi->event, 0xffffffff);
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
n = bitlen / 8;
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
/* Handle data in 8-bit chunks */
while (n--) {
ulong start;
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
if (cout)
tmpdout = *cout++;
if (type == SOC_MPC832X)
tmpdout <<= 24;
/* Write the data out */
out_be32(&spi->tx, tmpdout);
debug("*** %s: ... %08x written\n", __func__, tmpdout);
/*
* Wait for SPI transmit to get out
* or time out (1 second = 1000 ms)
* The NE event must be read and cleared first
*/
start = get_timer(0);
do {
u32 event = in_be32(&spi->event);
bool have_ne = event & SPI_EV_NE;
bool have_nf = event & SPI_EV_NF;
if (!have_ne)
continue;
tmpdin = in_be32(&spi->rx);
setbits_be32(&spi->event, SPI_EV_NE);
if (type == SOC_MPC832X)
tmpdin >>= 16;
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
if (cin)
*cin++ = tmpdin;
/*
* Only bail when we've had both NE and NF events.
* This will cause timeouts on RO devices, so maybe
* in the future put an arbitrary delay after writing
* the device. Arbitrary delays suck, though...
*/
if (have_nf)
break;
mdelay(1);
} while (get_timer(start) < SPI_TIMEOUT);
if (get_timer(start) >= SPI_TIMEOUT) {
debug("*** %s: Time out during SPI transfer\n",
__func__);
return -ETIMEDOUT;
}
debug("*** %s: transfer ended. Value=%08x\n", __func__, tmpdin);
}
SPI API improvements This patch gets rid of the spi_chipsel table and adds a handful of new functions that makes the SPI layer cleaner and more flexible. Instead of the spi_chipsel table, each board that wants to use SPI gets to implement three hooks: * spi_cs_activate(): Activates the chipselect for a given slave * spi_cs_deactivate(): Deactivates the chipselect for a given slave * spi_cs_is_valid(): Determines if the given bus/chipselect combination can be activated. Not all drivers may need those extra functions however. If that's the case, the board code may just leave them out (assuming they know what the driver needs) or rely on the linker to strip them out (assuming --gc-sections is being used.) To set up communication parameters for a given slave, the driver needs to call spi_setup_slave(). This returns a pointer to an opaque spi_slave struct which must be passed as a parameter to subsequent SPI calls. This struct can be freed by calling spi_free_slave(), but most driver probably don't want to do this. Before starting one or more SPI transfers, the driver must call spi_claim_bus() to gain exclusive access to the SPI bus and initialize the hardware. When all transfers are done, the driver must call spi_release_bus() to make the bus available to others, and possibly shut down the SPI controller hardware. spi_xfer() behaves mostly the same as before, but it now takes a spi_slave parameter instead of a spi_chipsel function pointer. It also got a new parameter, flags, which is used to specify chip select behaviour. This may be extended with other flags in the future. This patch has been build-tested on all powerpc and arm boards involved. I have not tested NIOS since I don't have a toolchain for it installed, so I expect some breakage there even though I've tried fixing up everything I could find by visual inspection. I have run-time tested this on AVR32 ATNGW100 using the atmel_spi and DataFlash drivers posted as a follow-up. I'd like some help testing other boards that use the existing SPI API. But most of all, I'd like some comments on the new API. Is this stuff usable for everyone? If not, why? Changed in v4: - Build fixes for various boards, drivers and commands - Provide common struct spi_slave definition that can be extended by drivers - Pass a struct spi_slave * to spi_cs_activate and spi_cs_deactivate - Make default bus and mode build-time configurable - Override default SPI bus ID and mode on mx32ads and imx31_litekit. Changed in v3: - Add opaque struct spi_slave for controller-specific data associated with a slave. - Add spi_claim_bus() and spi_release_bus() - Add spi_free_slave() - spi_setup() is now called spi_setup_slave() and returns a struct spi_slave - soft_spi now supports four SPI modes (CPOL|CPHA) - Add bus parameter to spi_setup_slave() - Convert the new i.MX32 SPI driver - Convert the new MC13783 RTC driver Changed in v2: - Convert the mpc8xxx_spi driver and the mpc8349emds board to the new API. Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Tested-by: Guennadi Liakhovetski <lg@denx.de>
2008-05-16 09:10:31 +00:00
if (flags & SPI_XFER_END)
mpc8xxx_spi_cs_deactivate(dev);
return 0;
}
static int mpc8xxx_spi_set_speed(struct udevice *dev, uint speed)
{
struct mpc8xxx_priv *priv = dev_get_priv(dev);
spi8xxx_t *spi = priv->spi;
u32 bits, mask, div16, pm;
u32 mode;
ulong clk;
clk = priv->clk_rate;
if (clk / 64 > speed) {
div16 = SPI_MODE_DIV16;
clk /= 16;
} else {
div16 = 0;
}
pm = (clk - 1)/(4*speed) + 1;
if (pm > 16) {
dev_err(dev, "requested speed %u too small\n", speed);
return -EINVAL;
}
pm--;
bits = div16 | (pm << SPI_MODE_PM_SHIFT);
mask = SPI_MODE_DIV16 | SPI_MODE_PM_MASK;
mode = in_be32(&spi->mode);
if ((mode & mask) != bits) {
/* Must clear mode[EN] while changing speed. */
mode &= ~(mask | SPI_MODE_EN);
out_be32(&spi->mode, mode);
mode |= bits;
out_be32(&spi->mode, mode);
mode |= SPI_MODE_EN;
out_be32(&spi->mode, mode);
}
debug("requested speed %u, set speed to %lu/(%s4*%u) == %lu\n",
speed, priv->clk_rate, div16 ? "16*" : "", pm + 1,
clk/(4*(pm + 1)));
mpc8xxx_spi: always use 8-bit characters, don't read or write garbage There are a few problems with the current driver. First, it unconditionally reads from dout/writes to din whether or not those pointers are NULL. So for example a simple "sf probe" ends up writing four bytes at address 0: => md.l 0x0 8 00000000: 45454545 45454545 05050505 05050505 EEEEEEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ => sf probe 0 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53618 din 00000000 bitlen 8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 0FB536B8 bitlen 48 SF: Detected s25sl032p with page size 256 Bytes, erase size 64 KiB, total 4 MiB => md.l 0x0 8 00000000: ff000000 45454545 05050505 05050505 ....EEEE........ 00000010: 00000000 00000000 07070707 07070707 ................ (here I've change the first debug statement to a printf, and made it print the din/dout pointers rather than the uints they point at). Second, as we can also see above, it always writes a full 32 bits, even if a smaller amount was requested. So for example => mw.l $loadaddr 0xaabbccdd 8 => md.l $loadaddr 8 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ 02000010: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x400 6 device 0 offset 0x400, size 0x6 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 48 SF: 6 bytes @ 0x400 Read: OK => sf read 0x02000010 0x400 8 device 0 offset 0x400, size 0x8 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB53848 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000010 bitlen 64 SF: 8 bytes @ 0x400 Read: OK => md.l $loadaddr 8 02000000: 45454545 45450000 aabbccdd aabbccdd EEEEEE.......... 02000010: 45454545 45454545 aabbccdd aabbccdd EEEEEEEE........ Finally, when the bitlen is 24 mod 32 (e.g. requesting to read 3 or 7 bytes), the last three bytes and up being the wrong ones, since the driver does a full 32 bit read and then shifts the wrong byte out: => mw.l $loadaddr 0xaabbccdd 4 => md.l $loadaddr 4 02000000: aabbccdd aabbccdd aabbccdd aabbccdd ................ => sf read $loadaddr 0x444 10 device 0 offset 0x444, size 0x10 mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 128 SF: 16 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 3031392e 30342d30 U-Boot 2019.04-0 => mw.l $loadaddr 0xaabbccdd 4 => sf read $loadaddr 0x444 0xb device 0 offset 0x444, size 0xb mpc8xxx_spi_xfer: slave spi@7000:0 dout 0FB536E8 din 00000000 bitlen 40 mpc8xxx_spi_xfer: slave spi@7000:0 dout 00000000 din 02000000 bitlen 88 SF: 11 bytes @ 0x444 Read: OK => md.l $loadaddr 4 02000000: 552d426f 6f742032 31392e00 aabbccdd U-Boot 219...... Fix all of that by always using a character size of 8, and reject transfers that are not a whole number of bytes. While it ends being more work for the CPU, we're mostly bounded by the speed of the SPI bus, and we avoid writing to the mode register in every loop. Based on work by Klaus H. Sørensen. Cc: Klaus H. Sorensen <khso@prevas.dk> Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2020-02-11 15:20:25 +00:00
return 0;
}
static int mpc8xxx_spi_set_mode(struct udevice *dev, uint mode)
{
/* TODO(mario.six@gdsys.cc): Using SPI_CPHA (for clock phase) and
* SPI_CPOL (for clock polarity) should work
*/
return 0;
}
static const struct dm_spi_ops mpc8xxx_spi_ops = {
.xfer = mpc8xxx_spi_xfer,
.set_speed = mpc8xxx_spi_set_speed,
.set_mode = mpc8xxx_spi_set_mode,
/*
* cs_info is not needed, since we require all chip selects to be
* in the device tree explicitly
*/
};
static const struct udevice_id mpc8xxx_spi_ids[] = {
{ .compatible = "fsl,spi" },
{ .compatible = "fsl,mpc832x-spi", .data = SOC_MPC832X },
{ }
};
U_BOOT_DRIVER(mpc8xxx_spi) = {
.name = "mpc8xxx_spi",
.id = UCLASS_SPI,
.of_match = mpc8xxx_spi_ids,
.ops = &mpc8xxx_spi_ops,
.of_to_plat = mpc8xxx_spi_of_to_plat,
.probe = mpc8xxx_spi_probe,
.priv_auto = sizeof(struct mpc8xxx_priv),
};