FreeRTOS+FAT Debugging New SD Card Media Driver: data fading away?!

carlk3 · February 23, 2020, 4:43am

I am working on a new +FAT port. This is a media driver for SD Card over SPI.

At first I tested it with something similar to the FATFileSystem example at https://os.mbed.com/docs/mbed-os/v5.15/apis/fatfilesystem.html. That seemed to work well. Next, I tried FreeRTOSv10.2.1_191129\FreeRTOS-Labs\Demo\Common\FreeRTOS_Plus_FAT_Demos\CreateAndVerifyExampleFiles.c, and I am going crazy trying to debug this. It gets through prvCreateDemoFilesUsing_ff_fwrite OK, but, when it gets to prvVerifyDemoFileUsing_ff_fread, the files are gone!

Creating file /fs/root001.txt in /fs
Creating file /fs/root002.txt in /fs
Creating file /fs/root003.txt in /fs
Creating file /fs/root004.txt in /fs
Creating file /fs/root005.txt in /fs
Reading file root001.txt from /fs
FF_fopen(root001.txt): No such file or directory (-2)
assertion "pxFile" failed: file "CreateAndVerifyExampleFiles.c", line 236, function: prvVerifyDemoFileUsing_ff_fread

Sure enough, if I yank the card and put it into the slot on my PC, I don’t see the root???.txt files.

I added some more printf()s to the FATFileSystem example, and I added a directory listing in the middle:

ls();
err = ff_chdir("fs");
if (err < 0) {
	FF_PRINTF("chdir error: %s (%d)\n", strerror(stdioGET_ERRNO()),
			-stdioGET_ERRNO());
}
ls();

Sure enough, the numbers.txt file is there:

FF_Mount:
FF_FS_Add:
Reading FAT and calculating Free Space
Partition Nr          0
Type                 12 (FAT32)
VolLabel       'NO NAME    '
TotalSectors   15515648
SecsPerCluster       64
Size            7753728 KB
FreeSize        7753536 KB ( 100 perc free )
Opening "/fs/numbers.txt"... Fail :(
No file found, creating a new file... OK
Writing numbers (10/10)... OK
Seeking file... OK

Incrementing numbers (0/10)...
Incrementing numbers (1/10)...
Incrementing numbers (2/10)...
Incrementing numbers (3/10)...
Incrementing numbers (4/10)...
Incrementing numbers (5/10)...
Incrementing numbers (6/10)...
Incrementing numbers (7/10)...
Incrementing numbers (8/10)...
Incrementing numbers (10/10)... OK
Closing "/fs/numbers.txt"... OK
Directory Listing:
fs [directory] [size=1024]
. [directory] [size=1024]
Directory Listing:
System Volume Information [directory] [size=0]
numbers.txt [writable file] [size=60]
.. [directory] [size=1024]
. [directory] [size=1024]
Opening "/fs/numbers.txt"... OK
numbers:
    5
    6
    7
    8
    9
    10
    11
    12
    13

Closing "/fs/numbers.txt"... OK

When I pull the card and mount it in Windows, sometimes I see numbers.txt, and sometimes I don’t. Similary, at the start of the test, sometimes the old file is found, but usually I get:

Opening "/fs/numbers.txt"... Fail :(
No file found, creating a new file... OK

What could be happening here? Am I missing some step needed to commit changes to the card? Is my hardware no good? (I did try a couple different microSD cards). What should I be looking at next?

CreateAndVerifyExampleFiles.c (16.0 KB)
x.c (5.9 KB)

htibosch · February 23, 2020, 9:08am

Hello carlk3, I recommend to test your driver first with a simple self-made test.
The best way to do this is to use the interface in ff_stdio.h.
Create a file, write to it, and close it. Then open, read and close it. Then try it on a PC. Something like that.

I am not sure what /fs stands for in the example. It it the pseudo directory on which the SD-card is mounted?

What parameter did you pass to FF_SDDiskInit()? Was it "/" of "/fs"?

You wrote the driver your self. Q : does it keep sectors in cache? It should not do, because FreeRTOS+TCP already uses cache memory when necessary.

PS. when a file was written to, the caches will be flushed as soon as ff_close() is called.

carlk3 · February 23, 2020, 2:50pm

Hi Hein,

I recommend to test your driver first with a simple self-made test.
The best way to do this is to use the interface in ff_stdio.h .
Create a file, write to it, and close it. Then open, read and close it. Then try it on a PC. Something like that.

That is what I’m trying to do in the snippet I attached as “x.c”. For very simple tests, it seems to work. But, somehow, later activity seems to wipe out files created earlier.

Now, I’m thinking of ~~stealing~~ porting this “Low level disk I/O module function checker” code http://elm-chan.org/fsw/ff/res/app4.c to test my media driver at a lower level.

I am not sure what /fs stands for in the example. It it the pseudo directory on which the SD-card is mounted?

/fs is supposed to be the, umm, “mount point”:

/* Where the RAM disk is mounted. */
#define mainSD_DISK_NAME			"/fs"
// ...

		/*
		 * Add a file system
		 * The path must be absolute, e.g. start with a slash
		 * The second argument is the FF_Disk_t structure that is handling the driver
		 */
		int iReturn = FF_FS_Add(mainSD_DISK_NAME, pxDisk);
    	FF_PRINTF("FF_FS_Add: %s\n", (const char *) FF_GetErrMessage(xError));
		configASSERT(iReturn);
	}

What parameter did you pass to FF_SDDiskInit() ? Was it "/" of "/fs" ?

I took the FF_Mount and FF_FS_Add out of FF_SDDiskInit, so I am not using a parameter there anymore. You can see the calls in the snippet I attached as “x.c”.

Q : does it keep sectors in cache?

Not intentionally! It simply passes everything through to SPI commands.

I am pretty unsure how to configure the +FAT cache. This is what I have now:

	const BaseType_t SECTOR_SIZE = 512;
	const uint32_t xIOManagerCacheSize = 2 * SECTOR_SIZE;
// ...

		memset(&xParameters, 0, sizeof xParameters);
		xParameters.pucCacheMemory = NULL;
		xParameters.ulMemorySize = xIOManagerCacheSize;
		xParameters.ulSectorSize = SECTOR_SIZE;
		xParameters.fnWriteBlocks = FFWrite;
// ...
		pxDisk->pxIOManager = FF_CreateIOManger(&xParameters, &xError);

Thanks for your help!

       Carl

FreeRTOSFATConfig.h (14.0 KB)
ff_sddisk.c (68.4 KB)

htibosch · February 24, 2020, 3:46pm

#define mainSD_DISK_NAME "/fs"

I think it is less confusing when you define this as "/", and create a physical directory "/fs".

const uint32_t xIOManagerCacheSize = 2 * SECTOR_SIZE;

That is very small. Not sure if it gives problems somewhere.

Yes it would be a good idea to test the low-level SPI read and write functions.

carlk3 · February 24, 2020, 10:20pm

#define mainSD_DISK_NAME "/fs"

I think it is less confusing when you define this as "/" , and create a physical directory "/fs" .

Well, I am confusing I’m still unclear on this whole subject. I plan to eventually have 3 SD cards mounted in the system. I presume I’ll have to do something like this?

FF_FS_Add("/sd0", pxDisk0);
FF_FS_Add("/sd1", pxDisk1);
FF_FS_Add("/sd2", pxDisk2);

const uint32_t xIOManagerCacheSize = 2 * SECTOR_SIZE;

That is very small. Not sure if it gives problems somewhere.

OK, I will try 16 * SECTOR_SIZE and see how that works. I was thinking it would only affect performance (which is important, but I’m still trying to debug all this).

Yes it would be a good idea to test the low-level SPI read and write functions.

I have now done that, and it was very informative. I found two logic errors in my driver. Dumb mistakes. To quote Michael Bolton in Office Space: “Okay! I must have put a decimal point in the wrong place or something. Sh*t! I always do that. I always mess up some mundane detail.”

Update: I now have FreeRTOS+FAT+CLI running on PSoC® 6 BLE Pioneer Kit (CY8CKIT-062-BLE) with an Adafruit Data Logger Shield.

app4 - IO module function checker.c (12.5 KB)