lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5693B47A.9050705@atmel.com>
Date:	Mon, 11 Jan 2016 14:56:10 +0100
From:	Cyrille Pitchen <cyrille.pitchen@...el.com>
To:	Boris Brezillon <boris.brezillon@...e-electrons.com>
CC:	<computersforpeace@...il.com>, <linux-mtd@...ts.infradead.org>,
	<nicolas.ferre@...el.com>, <marex@...x.de>, <vigneshr@...com>,
	<linux-kernel@...r.kernel.org>,
	<linux-arm-kernel@...ts.infradead.org>,
	<devicetree@...r.kernel.org>, <robh+dt@...nel.org>,
	<pawel.moll@....com>, <mark.rutland@....com>,
	<ijc+devicetree@...lion.org.uk>, <galak@...eaurora.org>
Subject: Re: [PATCH linux-next v2 02/14] mtd: spi-nor: properly detect the
 memory when it boots in Quad or Dual mode

Hi Boris,

Le 11/01/2016 11:08, Boris Brezillon a écrit :
> Hi Cyrille,
> 
> Sorry for this pretty useless review, but I think you know more about
> SPI-NOR than I do. So, consider my comments as nitpicks, which shouldn't
> prevent your series from being applied.
> 
> On Fri, 8 Jan 2016 17:02:14 +0100
> Cyrille Pitchen <cyrille.pitchen@...el.com> wrote:
> 
>> The quad (or dual) mode of a spi-nor memory may be enabled at boot time by
>> non-volatile bits in some setting register. Also such a mode may have
>> already been enabled at early stage by some boot loader.
>>
>> Hence, we should not guess the spi-nor memory is always configured for the
>> regular SPI 1-1-1 protocol.
>>
>> Micron and Macronix memories, once their Quad (or dual for Micron) mode
>> enabled, no longer process the regular JEDEC Read ID (0x9f) command but
>> instead reply to a new command: JEDEC Read ID Multiple I/O (0xaf).
>> Besides, in Quad mode both memory manufacturers expect ALL commands to
>> use the SPI 4-4-4 protocol. For Micron memories, enabling their Dual mode
>> implies to use the SPI 2-2-2 protocol for ALL commands.
>>
>> Winbond memories, once their Quad mode enabled, expect ALL commands to use
>> the SPI 4-4-4 protocol. Unlike Micron and Macronix memories, they still
>> reply to the regular JEDEC Read ID (0x9f) command.
>>
>> Signed-off-by: Cyrille Pitchen <cyrille.pitchen@...el.com>
>> ---
>>  drivers/mtd/spi-nor/spi-nor.c | 83 ++++++++++++++++++++++++++++++++++++++++---
>>  include/linux/mtd/spi-nor.h   | 50 ++++++++++++++++++++++++--
>>  2 files changed, 127 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/mtd/spi-nor/spi-nor.c b/drivers/mtd/spi-nor/spi-nor.c
>> index 3028c06547c1..8967319ea7da 100644
>> --- a/drivers/mtd/spi-nor/spi-nor.c
>> +++ b/drivers/mtd/spi-nor/spi-nor.c
>> @@ -73,6 +73,12 @@ struct flash_info {
>>  
>>  #define JEDEC_MFR(info)	((info)->id[0])
>>  
>> +struct read_id_config {
>> +	enum read_mode		mode;
>> +	enum spi_nor_protocol	proto;
>> +	u8			opcode;
>> +};
>> +
>>  static const struct flash_info *spi_nor_match_id(const char *name);
>>  
>>  /*
>> @@ -879,11 +885,22 @@ static const struct flash_info spi_nor_ids[] = {
>>  	{ },
>>  };
>>  
>> -static const struct flash_info *spi_nor_read_id(struct spi_nor *nor)
>> +static const struct flash_info *spi_nor_read_id(struct spi_nor *nor,
>> +						enum read_mode mode)
>>  {
>> -	int			tmp;
>> +	int			i, tmp;
>>  	u8			id[SPI_NOR_MAX_ID_LEN];
>>  	const struct flash_info	*info;
>> +	static const struct read_id_config configs[] = {
>> +		/* Winbond QPI mode */
>> +		{SPI_NOR_QUAD, SNOR_PROTO_4_4_4, SPINOR_OP_RDID},
>> +
>> +		/* Micron Quad mode & Macronix QPI mode */
>> +		{SPI_NOR_QUAD, SNOR_PROTO_4_4_4, SPINOR_OP_MIO_RDID},
>> +
>> +		/* Micron Dual mode */
>> +		{SPI_NOR_DUAL, SNOR_PROTO_2_2_2, SPINOR_OP_MIO_RDID}
>> +	};
>>  
>>  	tmp = nor->read_reg(nor, SPINOR_OP_RDID, id, SPI_NOR_MAX_ID_LEN);
>>  	if (tmp < 0) {
>> @@ -891,6 +908,58 @@ static const struct flash_info *spi_nor_read_id(struct spi_nor *nor)
>>  		return ERR_PTR(tmp);
>>  	}
>>  
>> +	/*
>> +	 * Check whether the SPI NOR memory has already been configured (at
>> +	 * reset or by some bootloader) to use a protocol other than SPI 1-1-1.
>> +	 */
>> +	for (i = 0; i < ARRAY_SIZE(configs); ++i) {
>> +		int len = SPI_NOR_MAX_ID_LEN;
>> +		bool is_multi = false;
>> +
>> +		/*
>> +		 * Check the latest read Manufacturer ID + Device ID (3 bytes):
>> +		 * if they are different from both 0x000000 and 0xffffff, we
>> +		 * assume that we succeeded in reading a valid JEDEC ID so we
>> +		 * don't need to try other SPI protocols.
>> +		 * Indeed when either the protocol or the op code are not valid,
>> +		 * the SPI NOR memory should not reply to the command. Hence the
>> +		 * SPI I/O lines remain in their default state: 1 when connected
>> +		 * to pull-up resistors or 0 with pull-down.
>> +		 */
>> +		if (!((id[0] == 0xff && id[1] == 0xff && id[2] == 0xff) ||
>> +		      (id[0] == 0x00 && id[1] == 0x00 && id[2] == 0x00)))
>> +			break;
> 
> Hopefully, you'll have all the pins configured with pull-up or
> pull-down resistors, but what if one of them is configured differently?
> 
> Shouldn't we try to find a valid chip definition by iterating over the
> spi_nor_ids[] table after each mode change?
> 

Honestly I don't know and I didn't test what would happen if no resistors were
used. So maybe you're right, it would be safer to look in the ID table up after
every read attempt.
>> +
>> +		/* Only try protocols supported by the user. */
>> +		if (configs[i].mode != mode)
>> +			continue;
>> +
>> +		/* Set this protocol for all commands. */
>> +		nor->reg_proto = configs[i].proto;
>> +		nor->read_proto = configs[i].proto;
>> +		nor->write_proto = configs[i].proto;
>> +		nor->erase_proto = configs[i].proto;
>> +
>> +		/*
>> +		 * Multiple I/O Read ID only returns the Manufacturer ID
>> +		 * (1 byte) and the Device ID (2 bytes). So we reset the
>> +		 * remaining bytes.
>> +		 */
>> +		if (configs[i].opcode == SPINOR_OP_MIO_RDID) {
>> +			is_multi = true;
>> +			len = 3;
>> +			memset(id + len, 0, sizeof(id) - len);
>> +		}
>> +
>> +		tmp = nor->read_reg(nor, configs[i].opcode, id, len);
>> +		if (tmp < 0) {
>> +			dev_dbg(nor->dev,
>> +				"error %d reading JEDEC ID%s\n",
>> +				tmp, (is_multi ? " Multi I/O" : ""));
>> +			return ERR_PTR(tmp);
>> +		}
>> +	}
>> +
>>  	for (tmp = 0; tmp < ARRAY_SIZE(spi_nor_ids) - 1; tmp++) {
>>  		info = &spi_nor_ids[tmp];
>>  		if (info->id_len) {
>> @@ -1148,11 +1217,17 @@ int spi_nor_scan(struct spi_nor *nor, const char *name, enum read_mode mode)
>>  	if (ret)
>>  		return ret;
>>  
>> +	/* Reset SPI protocol for all commands */
>> +	nor->erase_proto = SNOR_PROTO_1_1_1;
>> +	nor->read_proto = SNOR_PROTO_1_1_1;
>> +	nor->write_proto = SNOR_PROTO_1_1_1;
>> +	nor->reg_proto = SNOR_PROTO_1_1_1;
>> +
>>  	if (name)
>>  		info = spi_nor_match_id(name);
>>  	/* Try to auto-detect if chip name wasn't specified or not found */
>>  	if (!info)
>> -		info = spi_nor_read_id(nor);
>> +		info = spi_nor_read_id(nor, mode);
>>  	if (IS_ERR_OR_NULL(info))
>>  		return -ENOENT;
>>  
>> @@ -1163,7 +1238,7 @@ int spi_nor_scan(struct spi_nor *nor, const char *name, enum read_mode mode)
>>  	if (name && info->id_len) {
>>  		const struct flash_info *jinfo;
>>  
>> -		jinfo = spi_nor_read_id(nor);
>> +		jinfo = spi_nor_read_id(nor, mode);
>>  		if (IS_ERR(jinfo)) {
>>  			return PTR_ERR(jinfo);
>>  		} else if (jinfo != info) {
>> diff --git a/include/linux/mtd/spi-nor.h b/include/linux/mtd/spi-nor.h
>> index 62356d50815b..53932c87bcf2 100644
>> --- a/include/linux/mtd/spi-nor.h
>> +++ b/include/linux/mtd/spi-nor.h
>> @@ -75,8 +75,9 @@
>>  #define SPINOR_OP_BRWR		0x17	/* Bank register write */
>>  
>>  /* Used for Micron flashes only. */
>> -#define SPINOR_OP_RD_EVCR      0x65    /* Read EVCR register */
>> -#define SPINOR_OP_WD_EVCR      0x61    /* Write EVCR register */
>> +#define SPINOR_OP_MIO_RDID	0xaf	/* Multiple I/O Read JEDEC ID */
>> +#define SPINOR_OP_RD_EVCR	0x65    /* Read EVCR register */
>> +#define SPINOR_OP_WD_EVCR	0x61    /* Write EVCR register */
>>  
>>  /* Status Register bits. */
>>  #define SR_WIP			BIT(0)	/* Write in progress */
>> @@ -105,6 +106,43 @@ enum read_mode {
>>  	SPI_NOR_QUAD,
>>  };
>>  
>> +
> 
> Extra blank line here ^
> 
>> +#define SNOR_PROTO_CMD_OFF	8
>> +#define SNOR_PROTO_CMD_MASK	GENMASK(11, 8)
>> +#define SNOR_PROTO_CMD_TO_PROTO(cmd) \
>> +	(((cmd) << SNOR_PROTO_CMD_OFF) & SNOR_PROTO_CMD_MASK)
>> +#define SNOR_PROTO_CMD_FROM_PROTO(proto) \
>> +	((((u32)(proto)) & SNOR_PROTO_CMD_MASK) >> SNOR_PROTO_CMD_OFF)
>> +
>> +#define SNOR_PROTO_ADDR_OFF	4
>> +#define SNOR_PROTO_ADDR_MASK	GENMASK(7, 4)
>> +#define SNOR_PROTO_ADDR_TO_PROTO(addr) \
>> +	(((addr) << SNOR_PROTO_ADDR_OFF) & SNOR_PROTO_ADDR_MASK)
>> +#define SNOR_PROTO_ADDR_FROM_PROTO(proto) \
>> +	((((u32)(proto)) & SNOR_PROTO_ADDR_MASK) >> SNOR_PROTO_ADDR_OFF)
>> +
>> +#define SNOR_PROTO_DATA_OFF	0
>> +#define SNOR_PROTO_DATA_MASK	GENMASK(3, 0)
> 
> Why not directly reserving 8 bits per cycle type. Your end result is
> stored on a 32 bit integer anyway, and since you'll always have only 3
> cycle types (ADDR, CMD and DATA). This way, you'll be ready for
> SPI-8-8-8 :-).
> 

I've reserved 4 bits per cycle type so the possible values range into [0 - 15].
I thought it would be enough to anticipate future evolution like SPI 8-8-8.
Also it leaves more bits available for future usage.
Finally, I've chosen 4 bits and this order so when debugging and/or printing
the value in hex format, you can easily decode the corresponding protocol:

SPI 1-1-4 -> 0x00000114
SPI 1-2-2 -> 0x00000122
SPI 4-4-4 -> 0x00000444

Anyway I don't mind using 8 bits instead if it is preferred.

>> +#define SNOR_PROTO_DATA_TO_PROTO(data) \
>> +	(((data) << SNOR_PROTO_DATA_OFF) & SNOR_PROTO_DATA_MASK)
>> +#define SNOR_PROTO_DATA_FROM_PROTO(proto) \
>> +	((((u32)(proto)) & SNOR_PROTO_DATA_MASK) >> SNOR_PROTO_DATA_OFF)
>> +
>> +#define SNOR_PROTO(cmd, addr, data)	  \
>> +	(SNOR_PROTO_CMD_TO_PROTO(cmd) |   \
>> +	 SNOR_PROTO_ADDR_TO_PROTO(addr) | \
>> +	 SNOR_PROTO_DATA_TO_PROTO(data))
>> +
>> +enum spi_nor_protocol {
>> +	SNOR_PROTO_1_1_1 = SNOR_PROTO(1, 1, 1),	/* SPI */
>> +	SNOR_PROTO_1_1_2 = SNOR_PROTO(1, 1, 2),	/* Dual Output */
>> +	SNOR_PROTO_1_1_4 = SNOR_PROTO(1, 1, 4),	/* Quad Output */
>> +	SNOR_PROTO_1_2_2 = SNOR_PROTO(1, 2, 2),	/* Dual IO */
>> +	SNOR_PROTO_1_4_4 = SNOR_PROTO(1, 4, 4),	/* Quad IO */
>> +	SNOR_PROTO_2_2_2 = SNOR_PROTO(2, 2, 2),	/* Dual Command */
>> +	SNOR_PROTO_4_4_4 = SNOR_PROTO(4, 4, 4),	/* Quad Command */
>> +};
> 
> Do you really need this enum definition? I mean, directly using
> SNOR_PROTO(X, Y, Z) is just as easy as using the SNOR_PROTO_X_Y_Z enum
> definitions, and you won't have to define new ones if spi-nor vendors
> decide to support new combinations.
> 

This enum was created only for stronger type checking; Except for that, there
is no strong reason to keep the enum. So I don't mind removing it if other
people agree with you.


>> +
>>  #define SPI_NOR_MAX_CMD_SIZE	8
>>  enum spi_nor_ops {
>>  	SPI_NOR_OPS_READ = 0,
>> @@ -132,6 +170,10 @@ enum spi_nor_option_flags {
>>   * @flash_read:		the mode of the read
>>   * @sst_write_second:	used by the SST write operation
>>   * @flags:		flag options for the current SPI-NOR (SNOR_F_*)
>> + * @erase_proto:	the SPI protocol used by erase operations
>> + * @read_proto:		the SPI protocol used by read operations
>> + * @write_proto:	the SPI protocol used by write operations
>> + * @reg_proto		the SPI protocol used by read_reg/write_reg operations
>>   * @cmd_buf:		used by the write_reg
>>   * @prepare:		[OPTIONAL] do some preparations for the
>>   *			read/write/erase/lock/unlock operations
>> @@ -160,6 +202,10 @@ struct spi_nor {
>>  	u8			read_opcode;
>>  	u8			read_dummy;
>>  	u8			program_opcode;
>> +	enum spi_nor_protocol	erase_proto;
>> +	enum spi_nor_protocol	read_proto;
>> +	enum spi_nor_protocol	write_proto;
>> +	enum spi_nor_protocol	reg_proto;
> 
> Should be u32 fields if you decide to drop the enum definition.
> 
>>  	enum read_mode		flash_read;
>>  	bool			sst_write_second;
>>  	u32			flags;
> 
> 
> 

Thanks for your review :)

Best regards,

Cyrille

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ