FDS Write: Wrong Written Data

Question

Good Afternoon, 
 
 I'm writing some config data on an empty record the first time I turn on my board and updating that same record every next time I turn on the board. 
 While the writing of the first data is correctly written with fds_record_write() - when I check the data with fds_record_open() -, the data is totally wrongly written when I update the record with new config data, with fds_record_update(). 
 When the board initialises I initialise FDS and call flash_write_record_setup(&DEVICE_struct_w_config2write) after initialising the softdevice. 
 By looking into the code below, do you have any cue why the data is not correctly written when I update the record? 
 
 Thank you, 
 João Oliveira 
 
 /* Declarations in flash.h */

/* File ID and Key used for the configuration record. */

#define CONFIG_FILE (0xF010)
#define CONFIG_REC_KEY (0x7010)

typedef struct
{
	uint8_t enable;
	uint8_t time_adv;
}ADVERTISE;

typedef struct
{
	uint8_t state;
	ADVERTISE adv2;
}DEVICE;

/* flash_write_record_setup function in flash.c */

void flash_write_record_setup(DEVICE * config2write)
{
 ret_code_t rc;

 fds_record_desc_t desc = {0};
 fds_find_token_t tok = {0};

 //CONFIG_FILE and CONFIG_REC_KEY defined in flash.h
 rc = fds_record_find(CONFIG_FILE, CONFIG_REC_KEY, &desc, &tok);

 if (rc == FDS_SUCCESS) //record exists => update it
 {
 /* A config file is in flash. Let's update it. */
 fds_flash_record_t config = {0};

 /* Open the record and read its contents. */
 rc = fds_record_open(&desc, &config);
 APP_ERROR_CHECK(rc);

 /* Close the record when done reading. */
 rc = fds_record_close(&desc);
 APP_ERROR_CHECK(rc);
 
 //compare existing config in flash with new one
 uint8_t ret = memcmp(config2write, config.p_data, sizeof(DEVICE));

 if(ret != 0) //if different update config in flash
 {
 /* Write the updated record to flash. */
 fds_record_t const rec =
 {
 .file_id = CONFIG_FILE,
 .key = CONFIG_REC_KEY,
 .data.p_data = config2write,
 .data.length_words = (sizeof(DEVICE) + 3) / sizeof(uint32_t)
 };

 rc = fds_record_update(&desc, &rec);
 APP_ERROR_CHECK(rc);

 m_write_finished = false;
 }
 else
 {
 /* Same config data as previous, nothing to be done */
 }
 }
 else /* System config not found; write a new one. */
 {
 fds_record_t const rec =
 {
 .file_id = CONFIG_FILE,
 .key = CONFIG_REC_KEY,
 .data.p_data = config2write,
 .data.length_words = (sizeof(DEVICE) + 3) / sizeof(uint32_t)
 };

 rc = fds_record_write(&desc, &rec);
 APP_ERROR_CHECK(rc);

 m_write_finished = false;
 }
}

Rune Holmgren · Accepted Answer

Sorry, I should have realized this when first looking at your post. The problem is not the flash, but rather that you are trying to print strings which are not null terminated. 
 The reason you are successfully able to write and later print strings that are not a length divisible by 4 is that the flash memory will be padded with '0's. When you write "tes" to flash, the actual flash will be padded with a 0 byte at the end. In flash the following values will be present [116, 101, 115, 0]. To C this is the string "tes" with the final byte telling sprintf that the string has ended. When you write the string "test" to flash, the actual content of flash will be [116, 101, 115, 116] with no null terminator. When you send a reference to this flash address to the logging framework, nrf_log will attempt to print all the bytes it can find until a '0' is found. This may be an incredibly long string and the buffers overflow and cause the device to hang. 
 To resolve the issue, simply ensure that the data you are writing to flash is either null terminated, or that you add a null terminator after reading it out. 
 
 In this image, I have changed the printing to print specifically the first 4 characters present in flash. As you can see the word "test" is present, but there is no null terminator. You can't send this string as is to sprintf without adding a null terminator. 
 Best regards, Rune Holmgren

FDS Write: Wrong Written Data

Top Replies