linux-kernel - Re: [PATCH v4 1/3] mtd: spi-nor: always use bounce buffer for register read/writes

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]

Date:   Mon, 5 Aug 2019 11:24:14 +0000
From:   <Tudor.Ambarus@...rochip.com>
To:     <vigneshr@...com>, <miquel.raynal@...tlin.com>, <richard@....at>
CC:     <marek.vasut@...il.com>, <bbrezillon@...nel.org>,
        <linux-kernel@...r.kernel.org>, <linux-mtd@...ts.infradead.org>,
        <tmaimon77@...il.com>
Subject: Re: [PATCH v4 1/3] mtd: spi-nor: always use bounce buffer for
 register read/writes



On 08/05/2019 01:38 PM, Vignesh Raghavendra wrote:
> External E-Mail
> 
> 
> 
> On 05/08/19 2:36 PM, Tudor.Ambarus@...rochip.com wrote:
>>
>>
>> On 08/01/2019 07:22 PM, Vignesh Raghavendra wrote:
>>> External E-Mail
>>>
>>>
>>> spi-mem layer expects all buffers passed to it to be DMA'able. But
>>> spi-nor layer mostly allocates buffers on stack for reading/writing to
>>> registers and therefore are not DMA'able. Introduce bounce buffer to be
>>> used to read/write to registers. This ensures that buffer passed to
>>> spi-mem layer during register read/writes is DMA'able. With this change
>>> nor->cmd-buf is no longer used, so drop it.
>>>
>>> Reviewed-by: Boris Brezillon <boris.brezillon@...labora.com>
>>> Signed-off-by: Vignesh Raghavendra <vigneshr@...com>
>>> ---
>>>
>>> v4:
>>> Avoid memcpy during READID
>>>
>>> v3: new patch
>>>
>>>  drivers/mtd/spi-nor/spi-nor.c | 70 ++++++++++++++++++++---------------
>>>  include/linux/mtd/spi-nor.h   |  7 +++-
>>>  2 files changed, 45 insertions(+), 32 deletions(-)
>>>
>>> diff --git a/drivers/mtd/spi-nor/spi-nor.c b/drivers/mtd/spi-nor/spi-nor.c
>>> index 03cc788511d5..e02376e1127b 100644
>>> --- a/drivers/mtd/spi-nor/spi-nor.c
>>> +++ b/drivers/mtd/spi-nor/spi-nor.c
>>
>> cut
>>
>>>  /**
>>> @@ -1404,9 +1401,11 @@ static int write_sr_cr(struct spi_nor *nor, u8 *sr_cr)
>>>  {
>>>  	int ret;
>>>  
>>> +	memcpy(nor->bouncebuf, sr_cr, 2);
>>
>> I'm thinking out loud. This can be avoided by forcing all the callers to use
>> nor->bouncebuf. That would result in a:
>>
> 
> I can make this change and make all callers use nor->bouncebuf in next
> version.
> 

should be ok, it will not change the flow logic.

>> static int write_sr(struct spi_nor *nor, size_t len)
>>
>> write_sr_cr() can be removed. Memcopying 2 bytes is a small price to pay, we can
>> keep things as they are, to not be too invasive. But if you think that this idea
>> is worth it, tell.
>>
> 
> Sounds good to me. But replacing write_sr_cr() with above defintion of
> write_sr() should be a patch IMO>

I'll do a patch after we finish to integrate the big fat changes.

Thanks,
ta