[PATCH 1/2] mtd: core: fix use of uninitialized struct member

[Ahmad Fatoum]

On 11.06.21 12:23, Sascha Hauer wrote:
> Hi Antony,
> 
> On Fri, Jun 11, 2021 at 09:39:13AM +0300, Antony Pavlov wrote:
>> E.g. mtd_read() calls mtd_read_oob(), mtd_read_oob()
>> can check uninitialized ops->oobbuf. As a result
>> mtd_read_oob() can return -EOPNOTSUPP.
>>
>> Found on a MIPS board during /dev/m25p0 read, e.g.
>>
>>   barebox:/ md -s /dev/m25p0
>>   read: error 95
>>
>> Signed-off-by: Antony Pavlov <antonynpavlov at gmail.com>
>> ---
>>  drivers/mtd/core.c | 18 ++++++++++--------
>>  1 file changed, 10 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/mtd/core.c b/drivers/mtd/core.c
>> index 37fccda6be..98820dfb4f 100644
>> --- a/drivers/mtd/core.c
>> +++ b/drivers/mtd/core.c
>> @@ -379,12 +379,13 @@ int mtd_block_markgood(struct mtd_info *mtd, loff_t ofs)
>>  int mtd_read(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen,
>>               u_char *buf)
>>  {
>> -	struct mtd_oob_ops ops = {
>> -		.len = len,
>> -		.datbuf = buf,
>> -	};
>> +	struct mtd_oob_ops ops;
>>  	int ret;
>>  
>> +	memset(&ops, 0, sizeof(ops));
>> +	ops.len = len;
>> +	ops.datbuf = buf;
> 
> If this fixes something for you then you have severe problems with your
> compiler. Omitted fields are implicitly initialized to zero, often you
> can even find the corresponding call to memset in the objdump.

Perhaps elsewhere something is interpreting the padding, which may
not be zeroed when using an initializer list. This warrants further investigation
for sure.

Cheers,
Ahmad


-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |