From: Jonas Malaco <jonas@xxxxxxxxxxxx> commit c0689e46be23160d925dca95dfc411f1a0462708 upstream. Commit effa453168a7 ("i2c: i801: Don't silently correct invalid transfer size") revealed that ee1004_eeprom_read() did not properly limit how many bytes to read at once. In particular, i2c_smbus_read_i2c_block_data_or_emulated() takes the length to read as an u8. If count == 256 after taking into account the offset and page boundary, the cast to u8 overflows. And this is common when user space tries to read the entire EEPROM at once. To fix it, limit each read to I2C_SMBUS_BLOCK_MAX (32) bytes, already the maximum length i2c_smbus_read_i2c_block_data_or_emulated() allows. Fixes: effa453168a7 ("i2c: i801: Don't silently correct invalid transfer size") Cc: stable@xxxxxxxxxxxxxxx Reviewed-by: Heiner Kallweit <hkallweit1@xxxxxxxxx> Signed-off-by: Jonas Malaco <jonas@xxxxxxxxxxxx> Link: https://lore.kernel.org/r/20220203165024.47767-1-jonas@xxxxxxxxxxxx Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- drivers/misc/eeprom/ee1004.c | 3 +++ 1 file changed, 3 insertions(+) --- a/drivers/misc/eeprom/ee1004.c +++ b/drivers/misc/eeprom/ee1004.c @@ -82,6 +82,9 @@ static ssize_t ee1004_eeprom_read(struct if (unlikely(offset + count > EE1004_PAGE_SIZE)) count = EE1004_PAGE_SIZE - offset; + if (count > I2C_SMBUS_BLOCK_MAX) + count = I2C_SMBUS_BLOCK_MAX; + status = i2c_smbus_read_i2c_block_data_or_emulated(client, offset, count, buf); dev_dbg(&client->dev, "read %zu@%d --> %d\n", count, offset, status); Patches currently in stable-queue which might be from jonas@xxxxxxxxxxxx are queue-5.4/eeprom-ee1004-limit-i2c-reads-to-i2c_smbus_block_max.patch