Skip to contentSkip to conversation

media: i2c: Copy mt9t112 soc_camera sensor driver

bug report Dan Carpenter 2 participants 3 messages Oct 5, 2021 – Oct 9, 2021 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

Hello Media devs,

The patch 7641b0442195: "media: i2c: Copy mt9t112 soc_camera sensor driver" from Mar 12, 2018, leads to the following Smatch static checker warning:

	drivers/media/i2c/mt9t112.c:176 __mt9t112_reg_read()
	warn: not copying enough bytes for '&ret' (4 vs 2 bytes)

drivers/media/i2c/mt9t112.c

    150 static int __mt9t112_reg_read(const struct i2c_client *client, u16 command)
    151 {
    152         struct i2c_msg msg[2];
    153         u8 buf[2];
    154         int ret;
    155 
    156         command = swab16(command);
                          ^^^^^^^^^^^^^^^

This driver won't work on big endian systems

    157 
    158         msg[0].addr  = client->addr;
    159         msg[0].flags = 0;
    160         msg[0].len   = 2;
    161         msg[0].buf   = (u8 *)&command;
    162 
    163         msg[1].addr  = client->addr;
    164         msg[1].flags = I2C_M_RD;
    165         msg[1].len   = 2;
    166         msg[1].buf   = buf;
    167 
    168         /*
    169          * If return value of this function is < 0, it means error, else,
    170          * below 16bit is valid data.
    171          */
    172         ret = i2c_transfer(client->adapter, msg, 2);
    173         if (ret < 0)
    174                 return ret;
    175 
--> 176         memcpy(&ret, buf, 2);
                       ^^^^

And this is ugly as all heck. I would have fixed it but there were so many other endian bugs and I can't test it.

    177 
    178         return swab16(ret);
    179 }

regards,
dan carpenter

#1 lore raw

Hi Dan,

Quoted 5 lines · Dan Carpenter

Hello Media devs,

The patch 7641b0442195: "media: i2c: Copy mt9t112 soc_camera sensor driver" from Mar 12, 2018, leads to the following Smatch static checker warning:

That commit is from me, but I just copied the code as (ugly) as it was I didn't even have the hw to test it so I preferred not to touch it

Quoted 36 lines
	drivers/media/i2c/mt9t112.c:176 __mt9t112_reg_read()
	warn: not copying enough bytes for '&ret' (4 vs 2 bytes)

drivers/media/i2c/mt9t112.c

    150 static int __mt9t112_reg_read(const struct i2c_client *client, u16 command)
    151 {
    152         struct i2c_msg msg[2];
    153         u8 buf[2];
    154         int ret;
    155
    156         command = swab16(command);
                          ^^^^^^^^^^^^^^^

This driver won't work on big endian systems

    157
    158         msg[0].addr  = client->addr;
    159         msg[0].flags = 0;
    160         msg[0].len   = 2;
    161         msg[0].buf   = (u8 *)&command;
    162
    163         msg[1].addr  = client->addr;
    164         msg[1].flags = I2C_M_RD;
    165         msg[1].len   = 2;
    166         msg[1].buf   = buf;
    167
    168         /*
    169          * If return value of this function is < 0, it means error, else,
    170          * below 16bit is valid data.
    171          */
    172         ret = i2c_transfer(client->adapter, msg, 2);
    173         if (ret < 0)
    174                 return ret;
    175
--> 176         memcpy(&ret, buf, 2);
                       ^^^^
Quoted 2 lines

And this is ugly as all heck. I would have fixed it but there were so many other endian bugs and I can't test it.

Is this an endianess issue or just a complaint about the difference in size between the number of copied bytes and the destination ?

Thanks
j

Quoted 7 lines
    177
    178         return swab16(ret);
    179 }

regards,
dan carpenter

#2 lore raw
Quoted 15 lines · Jacopo Mondi
Quoted 12 lines
    166         msg[1].buf   = buf;
    167
    168         /*
    169          * If return value of this function is < 0, it means error, else,
    170          * below 16bit is valid data.
    171          */
    172         ret = i2c_transfer(client->adapter, msg, 2);
    173         if (ret < 0)
    174                 return ret;
    175
--> 176         memcpy(&ret, buf, 2);
                       ^^^^
Quoted 2 lines

And this is ugly as all heck. I would have fixed it but there were so many other endian bugs and I can't test it.

Quoted 2 lines

Is this an endianess issue or just a complaint about the difference in size between the number of copied bytes and the destination ?

On big endian systems this would translate the value from buf into a very large value but on little endian systems it should work okay. (I think). But all the swap16 stuff are endian bugs as well.

regards,
dan carpenter

Quoted 11 lines

Thanks
j

Quoted 7 lines
    177
    178         return swab16(ret);
    179 }

regards,
dan carpenter