linux-kernel - Re: [PATCH] mtd: spi-nor: Return error when nor->addr

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: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <96397ec0-14de-f09a-a18d-c67396e33fba@microchip.com>
Date:   Thu, 15 Nov 2018 10:54:39 +0000
From:   <Tudor.Ambarus@...rochip.com>
To:     <boris.brezillon@...tlin.com>, <liu.xiang6@....com.cn>,
        <cyrille.pitchen@...ev4u.fr>
CC:     <linux-mtd@...ts.infradead.org>, <linux-kernel@...r.kernel.org>,
        <marek.vasut@...il.com>, <dwmw2@...radead.org>,
        <computersforpeace@...il.com>, <richard@....at>,
        <liuxiang_1999@....com>
Subject: Re: [PATCH] mtd: spi-nor: Return error when nor->addr_width not match
 the device size

Hi, Liu, Boris, Cyrille,

On 11/14/2018 03:51 PM, Boris Brezillon wrote:
> On Wed, 14 Nov 2018 20:56:05 +0800
> Liu Xiang <liu.xiang6@....com.cn> wrote:
> 
>> In is25lp256, the DWORD1 of JEDEC Basic Flash Parameter Header
>> is 0xfff920e5. So the DWORD1[18:17] Address Bytes bits are 0b00,

Liu, can you point us to a datasheet that has the JEDEC BFPT tables described? I
couldn't find one ...

>> means that 3-Byte only addressing.
> 
> According to your other patch this NOR supports 4B opcode, which means
> the SFDP table is wrong.
> 
>> But the device size is larger
>> than 16MB, nor->addr_width must be 4 to access the whole address.
>> An error should be returned when nor->addr_width not match
> 
> 						   ^does not
> 
>> the device size in spi_nor_parse_sfdp().
>>
>> Suggested-by: Boris Brezillon <boris.brezillon@...tlin.com>
>> Signed-off-by: Liu Xiang <liu.xiang6@....com.cn>
>> ---
>>  drivers/mtd/spi-nor/spi-nor.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
>> diff --git a/drivers/mtd/spi-nor/spi-nor.c b/drivers/mtd/spi-nor/spi-nor.c
>> index 3eba13a..77eaf22 100644
>> --- a/drivers/mtd/spi-nor/spi-nor.c
>> +++ b/drivers/mtd/spi-nor/spi-nor.c
>> @@ -2669,6 +2669,10 @@ static int spi_nor_parse_bfpt(struct spi_nor *nor,
>>  	}
>>  	params->size >>= 3; /* Convert to bytes. */
>>  
>> +	/*if the device exceeds 16MiB, addr_width must be 4*/
> 
> Please add a white space after '/*' and before '*/':
> 
> 	/* If the device exceeds 16MiB, ->addr_width must be 4. */
> 
>> +	if ((params->size > 0x1000000) && (nor->addr_width == 3))
> 
> Parens are not needed around sub-conditions:
> 
> 	if (params->size > 0x1000000 && nor->addr_width == 3)
> 
>> +		return -EINVAL;
>> +
> 
> I'm not sure this is correct. Looks like some NORs only support 3B
> opcodes but have a "4-byte addressing" mode (see set_4byte() [1]).
> Don't know what's reported by the BFPT section in this case though
> (BFPT_DWORD1_ADDRESS_BYTES_3_ONLY or BFPT_DWORD1_ADDRESS_BYTES_3_OR_4).

Boris, this is in close relation with your second patch: [PATCH v3 2/2] mtd:
spi-nor: Use 4B opcodes when the NOR advertises both 3B and 4B.

When looking again at this, I would say that for the flashes that have a "4-byte
addressing" mode, but just 3B opcodes, I would expect the DWORD1[18:17] to be of
value BFPT_DWORD1_ADDRESS_BYTES_3_OR_4 (enters 4-Byte mode on command - uses 3B
opcodes).

If BFPT_DWORD1_ADDRESS_BYTES_3_OR_4 and 4B opcodes, then we can query BFPT
DWORD16[31:24]: it should have value xx1x_xxxxb to indicate that 4B opcodes are
supported. But which 4B opcodes are supported? Do all 3B opcodes have a 4B
opcode correspondent if SFDP 4-byte table is not available? This might be a good
assumption, but I can't see it anywhere in jesd216c.

Cyrille, when looking at ba3ae6a1d4c ("mtd: spi-nor: add a stateless method to
support memory size above 128Mib") I see that all flashes that declare
SPI_NOR_4B_OPCODES will have the 3B opcodes converted to 4B opcodes, assuming
that all 3B opcodes have a 4B correspondent. Is this assumption a general rule,
or it's just for the Spansion flashes?

Cheers,
ta

> 
> Anyway, I think this check should be moved here [2] to cover the
> non-SFDP case.
> 
>>  	/* Fast Read settings. */
>>  	for (i = 0; i < ARRAY_SIZE(sfdp_bfpt_reads); i++) {
>>  		const struct sfdp_bfpt_read *rd = &sfdp_bfpt_reads[i];
> 
> [1]https://elixir.bootlin.com/linux/v4.20-rc2/source/drivers/mtd/spi-nor/spi-nor.c#L278
> [2]https://elixir.bootlin.com/linux/v4.20-rc2/source/drivers/mtd/spi-nor/spi-nor.c#L3758
>