FDS Example - so what does "freeable words" actually mean?

[awneil]

Even immediately after issuing a gc (garbage collect) command, "freeable words" is still 1022:

fds example:~$ stat

total pages:    2
total records:  3
valid records:  1
dirty records:  2
largest contig: 1024
freeable words: 1031 (4124 bytes)
reserved words: 0
used     words: 2

fds example:~$ gc
<info> app: Event: FDS_EVT_GC received (FDS_SUCCESS)
fds example:~$ stat
total pages:    2
total records:  2
valid records:  1
dirty records:  1
largest contig: 1024
freeable words: 1022 (4088 bytes)
reserved words: 0
used     words: 2

fds example:~$

Why?

What does this number actually mean?

The documentation just says,

The largest number of words that can be reclaimed by garbage collection.

The actual amount of space freed by garbage collection might be less than this value if records are open while garbage collection is run.

But, even after doing delete all, and gc, it is still 1022:

fds example:~$ delete_all
<info> app: Deleting next record.
<info> app: Event: FDS_EVT_DEL_RECORD received (FDS_SUCCESS)
<info> app: Record ID:  0x0015
<info> app: File ID:    0xF010
<info> app: Record key: 0x7010
<info> app: Deleting next record.
<info> app: No records left to delete.
fds example:~$ gc
<info> app: Event: FDS_EVT_GC received (FDS_SUCCESS)
fds example:~$ stat
total pages:    2
total records:  1
valid records:  0
dirty records:  1
largest contig: 1024
freeable words: 1022 (4088 bytes)
reserved words: 0
used     words: 2
fds example:~$

Why?

If they are "freeable" - why does gc not free them??!

[Edvin]

Hello,

I was not able to reproduce the freeable words with the FDS example from SDK15. What SDK do you use? And do you use the project with or without the softdevice (is the path pca10040\s132 or pca10040\blank) ?

 

Below is the log from the unmodified fds example from SDK15\examples\peripheral\flash_fds\pca10040\s132\...

 

What operations did you do to get the large freeable words?

It seems like freeable is suppose to say how many words that will be freed with a "gc".

 

Best regards,

Edvin

[awneil]

Yes, it's SDK15 - and with SoftDevice S132.

Edvin said:

What operations did you do to get the large freeable words?

None - just built the project in SES, downloaded, and ran it.

Edvin said:

It seems like freeable is suppose to say how many words that will be freed with a "gc".

Well, yes:  that is what I thought it should mean - hence my puzzlement that it does not seem to mean that at all!

[Edvin]

Did you write anything in putty before delete_all in the log that you posted in your initial post? If not, what have you changed in the project? Does it happen if you program an unmodified version of the project? You can try with a fresh unzipped version of the SDK as well, in case you have modified something outside the project folder. If it still happens, can you send me a .zip of your project folder, so that I can try to replicate it? 

 

It doesn't happen when I flash the example, so I am curious why you get the large number on freeable words. 

 

Best regards,

Edvin

[awneil]

Where/how would I send it?

I think it would be rather large to email...

[awneil]

If I zip just the flash_fds folder from the SDK, it's 8MB.

If I try to attach that here, it says "Error"

No, it doesn't give anything more informative than just "Error" (in red)

[Edvin]

Yes It does that sometimes. Devzone is (since the update) a bit flaky. You can remove the _build or output (depending on what IDE you are using) folder inside the project. This will reduce the size significantly. Then try again.

 

Can you please describe what commands you wrote in the UART terminal? Was it only the part that is included in the log output in your initial post?

 

BR,

Edvin

[awneil]

Just as shown, I think.

In my real app (on a different board), the initialisation was failing with FD_ERROR_NO_PAGES.

I've just tried doing a complete chip erase, then reprogramming the SD and app - and that error goes away!

So it looks like the FDS is (or can be) sensitive to what's been in flash before?

I'll have to try the full erase & reprogram with the example ...

[Edvin]

From fds.c:

/**@brief The largest number of words that can be reclaimed by garbage collection.
*
* The actual amount of space freed by garbage collection might be less than this value if
* records are open while garbage collection is run.
*/
uint16_t freeable_words;

 

So freeable_words is the total number of words in your dirty_records. I have not looked to much in the details of the example, but it might be that the record was open when you ran gc. If you have modified the example, that is possible, but I don't think the example should leave any records open when you run the cli commands.

 

BR,

Edvin

[awneil]

and the number - 1022 - is far larger than the size of a record!

[tesc]

awneil said:

So it looks like the FDS is (or can be) sensitive to what's been in flash before?

Yes. The application data area of flash is separate from that of the application itself, and so the data is left unchanged when you re-download the application only. I.e. whatever application data is stored from previously is still there, unless explicitly cleared. Doing a full erase solves this issue, as does erasing the application data flash pages (e.g. with `nrfjprog --erasepage'.)

[awneil]

This was the first time FDS had been used on that board - but I had used Flash Storage on it before.

So it looks like the documentation needs updating to mention this explicitly.

EDIT

That would be both the FDS documentation itself, and the example.

[Edvin]

If there is something stored via FDS or fstorage before you uploaded the example, this might have been what caused the weird numbers. E.g. the peer manager stores peer data in flash, which may have been what you saw.

 

Best regards,

Edvin