Strange problem : SDMMC and USB msc not working with DMA - and SWV "wrong"

[AScha.3]

Hi,

so i have a strange problem with SDMMC and USB msc :

- custom board with H743ZI , 2x sd-card , 1x USB msc , for USB stick data;

- software using HAL, FatFS etc; made on IDE 1.10 , migrate ...13, 15, 17, 18, now 1.18.1 (actual)

- program working fine since > 1 year , but without DMA on sdmmc and usb/host ;

- for debug/messages with SWO, SWV working , as it should;

Now i tried using the DMA setting for sd and usb - no, nothing working.

SD stays on waiting for first data transfer, USB hangs in a loop: connecting-disconnected-connecting....

So i made a dummy test project, just debug + SDMMC1 + USB host, msc :

surprisingly works "out of the box", can mount sd and usb and read, ok; but no SWV at all - nothing!

Then i checked with CubeProgrammer (has SWV also) and it shows : something is sent, but hangs him up,

last thing CubeProg showing is this : <unknown packet>  (then need to kill the prog, not any reaction)

And now the really strange thing: on same hardware, same 200M core clock, same settings for used pins:

- the old program cannot work with sd or usb + DMA , but sending SWO correct.

- the dummy program working fine with sd1+usb DMA enabled, but no SWO messages at all.

- the dummy program not working with sdmmc2+ DMA enabled, but no SWO messages at all.

- both programs working with sdmmc1+2+USB , but no DMA , and no SWO messages at dummy prog.

 

Then i compared the used/included :

- libs from HAL etc : same

- startup code : same

- looked in ioc files : "same" (i could not find any important diff, besides the many more pins + periph used in the program, dummy uses only the needed pins for debug, sd , usb)

 

So i have no idea, why this all is so different in errors , on same IDE, compiler, libs, settings , and same hardware .

Do i anything wrong ?

Would be great, if someone has a good idea...

+ update:

made a new project,dummiii,  same hardware, at 200M, but only : debug + swv , nothing else.

and this .... can send swv, no problem !

so i copy the send "program" to first dummy-prog, but now : no debug, just :

Info : STLINK V2J46S7 (API v2) VID:PID 0483:3748
Info : Target voltage: 3.267200
Info : clock speed 4000 kHz
Info : stlink_dap_op_connect(connect)
Info : SWD DPIDR 0x6ba02477
Info : [STM32H743ZITx.ap2] Examination succeed
Error: Failed to read memory at 0x5c001000
Error: [STM32H743ZITx.cm7] Examination failed
Warn : target STM32H743ZITx.cm7 examination failed
Info : gdb port disabled
Info : starting gdb server for STM32H743ZITx.cm7 on 3333
Info : Listening on port 3333 for gdb connections
Info : accepting 'gdb' connection on tcp/3333
Error: Target not examined yet
Error: auto_probe failed

.. debug dont remember the H743 ??? or what ?

+

then i set on dummiii optimizer to -O2 (was -O0 , default)

now, also no debug:

File download complete
Time elapsed during download operation: 00:00:01.001



Verifying ...




Download verified successfully 


Shutting down...
Exit.

so flash ok, but then...exit.

perfect.

So checking, what was flashed, with CubeProgrammer, it shows:

Fine, dummiii flashed, and swo working. But IDE cannot debug any more , what it just flashed ??

After 3 attempts, unplug st-link, try again...suddenly it works again :

(Better, i dont write, what i think now.)

Then tried dummy , with optimizer set to -O3 , no debug , it tells:

Info : flash size probed value 1024k
Info : STM32H7 flash has dual banks
Info : Bank (0) size is 512 kb, base address is 0x08000000
Info : Device: STM32H74x/75x
Info : flash size probed value 1024k
Info : STM32H7 flash has dual banks
Info : Bank (1) size is 512 kb, base address is 0x08100000
Info : New GDB Connection: 1, Target STM32H743ZITx.cm7, state: running
Warn : GDB connection 1 on target STM32H743ZITx.cm7 not halted
undefined debug reason 8 (UNDEFINED) - target needs reset
Error: Failed to read memory at 0xe00e3fc8
Error executing event gdb-detach on target STM32H743ZITx.cm7:
read_memory: failed to read memory
Info : dropped 'gdb' connection
shutdown command invoked

 

hmmm - so check with connect on CubeProgrammer :  hey, it has new info for me !!!! see:

btw I never played around with trustZone or chip protection levels.

Soo... ?

[AScha.3]

un-re plugging st-link , CubeProgrammer :

So still dummiii prog running, and swo works.

-------------------------------------------

Then set debug in IDE to GDB, un-re-plug st-link, debug ends:

 STM32CubeProgrammer v2.19.0 
 -------------------------------------------------------------------



Log output file: /tmp/STM32CubeProgrammer_2zXUa4.log
ST-LINK SN : 56FF68064972495625570387
ST-LINK FW : V2J46S7
Board : --
Voltage : 3,26V
SWD freq : 4000 KHz
Connect mode: Under Reset
Reset mode : Hardware reset
Device ID : 0x450
Revision ID : Rev V
Device name : STM32H7xx
Flash size : 1 MBytes
Device type : MCU
Device CPU : Cortex-M7
BL Version : 0x91

Opening and parsing file: ST-LINK_GDB_server_JLABD4.srec


Memory Programming ...
 File : ST-LINK_GDB_server_JLABD4.srec
 Size : 64.39 KB 
 Address : 0x08000000 



Erasing memory corresponding to sector 0:
Erasing internal memory sector 0
Download in Progress:


File download complete
Time elapsed during download operation: 00:00:01.445



Verifying ...




Download verified successfully 


Shutting down...
Exit.

--- and CubeProgrammer ..?

obviously dummy flashed, but still no correct swo:

So set debug to OCD, debug hangs in startup and exit. After third attempt, suddenly strt debug:

--- but still no swo , correct swo.

Finish for today.

[Nemui Trinomius]

I have no idea what to say because you didn't suggest any "custom-board" schematic

and any sourcecode of the project....

 

....BUT, To solve your problem for you, I'll enforce Dark Force!....uMMMM...!!!

I found...!!!
Reason of "SDMMC works but not with DMA" is cache coherency problem!
Do proper cache control before/after DMA!
In addition,you should consider memory address alignment on DMA.

Second...!Reason for "strange behaviour" of CubeProgrammer is...
Program Runaway!
You should connect STlink/V2 with "connect under reset" ! and do "mass-erase" then re-power for now!

And to divide your problem,check simple SDMMC+DMA(eliminate ANY USB routine!).
If comfirmed,then check SDMMC with DMA and USB(WITHOUT DMA).
If get no problem,finally test SDMMC with DMA + USB with DMA.

...uhh ...today I'm so tired of using Dark Force that I can't give you any more advice.
But I believe you can solve your problem by yourself!

 

Best regards,
Nemui.

[AScha.3]

Thx , @Nemui Trinomius  , for thinking about - and even using your Dark Force!

Right, cache might be a problem, so i made all tests with d-cache OFF. (forgot to tell)

The hardware i didnt show, because : its working fine and has nothing special, just cpu , 3v3 linear supply, sd and usb ; and DACs and analog circuits have nothing to do with these problems; on my test board anyway not fitted, only the digital part + TFT connection.

 

But other problem i found : (my error) can the DMA reach the memory at all? 

So looking at H7  bus matrix   :

This might explain, why dma on sdmmc2 working, but not on sdmmc1, depending on the RAM, its using. (or i gave it...) So this is more or less clear now.

But still two problems i could not explain:

- on my old project , working fine without sd/usb DMA , no way to get it working with DMA (even with cache OFF and all data in main RAM); and/but  the new test dummy progs running fine with DMA . But might be still a problem with the RAM, because on old prog 95% RAM used, on test progs only about 51% used. But no error message, the old prog with DMA just waiting for data - until timeout.

- SWO working fine there on old prog, but not on new "dummy" program;

and on new "dummiii" prog , same settings, SWO working ok. 

The not working program SWO sending something , "unknown packet", but i have no idea, why this is always "wrong", with same settings in clock tree , same hardware, all optimizer settings tried - no change.

(And all on same hardware always.)