toggle quoted messageShow quoted text
That's some good analysis you've done. Note that this is for "frequent
sender" mesh nodes. For "requent receiver" nodes it's the Replay
Protection List (RPL) that will need to be updated frequently. What's
worse for the RPL is that you can't make similar jumps of 8192 messages,
rather you have to update the RPL for every message that the node
processes (note that relaying does not count as "processing" in this
The implications of failing to keep an updated Seq vs failing to keep an
updated RPL are similar, but not quite the same: in the case of the Seq,
you end up in a situation where other nodes ignore packets from you,
since they think someone is doing a replay attack with old seq values.
In the case of not updating the RPL it's the local node that gets
exposed to potential replay attacks.
On Thu, Mar 22, 2018, Laczen JMS wrote:
As the author of NVS (PR6391
https://github.com/zephyrproject-rtos/zephyr/pull/6391) I am probably
a bit biased. NVS as is can be used directly for managing persistent
data. I am looking into adding NVS as a backend to the configuration
system, but the configuration system comes with a major drawback:
I did a little calculation to compare the configuration system with
FCB as backend to direct use of NVS. Lets make the following
assumptions: Device is nrf51822, flash erase size is 1k, flash erase
cycles is 20000. Flash available for storage is 2k (2 pages). Data
without the sequence number takes up 512 bytes so 512 bytes is
available to store the sequence number.
Now lets see what happens when storing the sequence number, we will
only store it after 8192 messages (this means we only need to store a
11 bit number). This number can be increased but will reduce the
number of reboots that are allowed (to keep the IV index from changing
to fast). The worst case scenario is 24.3 messages/s (see nordicsemi
Calculation for the config system: each store of the sequence number
is a string looking like:
"bt/btmesh/seq=XXXX" which takes up 18 bytes. FCB overhead takes
another 8 bytes, so in total 26 bytes for a write of the sequence
number. After 19 (512/26) writes a sector is full and it will be
erased (using a 1K scratch sector). So with the worst case of 24.3
messages/s: after 19 * 8192 /24.3 s a sector will be erased. Or every
1.78 h. With the erase cycles of 20000 flash will be worn out after
about 4 years.
Calculation for NVS: the sequence number is given an id for storage.
This id is used to differentiate between other data that is stored
(and is part of the NVS overhead). The sequence number takes up only 2
bytes. NVS overhead takes up 8 bytes, so in total 10 bytes for a
write. After 51 (512/10) messages the first page (1k) is full and the
second page is taken into use. After 51 messages the second sector is
full and the first page is erased. So after 102 messages the same page
is erased. With the erase cycle of 20000 flash will be worn out after
about 21.5 years.
Modifying the above assumptions should allow you to get an estimate of
the expected life of your specific application.
To summarize: the configuration system comes at a price, the use of
human readable strings as storage items takes up more space, and
reduces the life of the device.
I am currently working on a solution for the configuration system that
would use NVS as a backend but would also allow some data to be
directly written to NVS. This would allow combination of human
readability for some data with reduced flash wear for other. But the
human readability will always take up more space.
Date: Wed, 21 Mar 2018 14:02:39 -0700
From: Michael Scott <michael@...>
To: Vikrant More <vikrant8051@...>
Subject: Re: [Zephyr-devel] Firmware over the air (FOTA) and FCB
support in 1.11.0
Content-Type: text/plain; charset="utf-8"; Format="flowed"
On 03/21/2018 10:09 AM, Vikrant More wrote:
Hi,IIRC, the idea is to settle on 1 solution which would be appropriate for
base layer for other higher level implementations to manage persistent
Regarding FCB: the initial implementation is done in v1.11,
but the APIs are fairly complex and IMHO it's meant to be used as a
data such as device configuration, system logs and other uses.
higher level services.? (The first PR will probably be re-written to
There are several pull requests in-progress which aim to add these
use FCB as it's base layer):
Are these PRs, suitable to save #BluetoothMesh's Sequence Numbers on
flash which updates frequently ?
most use-cases (including Mesh Sequence Numbers).?? You might want to
have 2 separate storage locations if you have data that only updates
occasionally *and* data that updates frequently.
The guys from Nordic (and possibly Johan) are looking at this issue as well.
On Wed, Mar 21, 2018 at 9:59 PM, Michael Scott
On 03/20/2018 09:14 PM, ashish.shukla@...
Hi all, Hi Ashish,
I've been waiting for FOTA and FCB support in zephyr and now when
it is supported, I cannot see any samples available or proper
documentation to use these features in my project.
Your question is a bit open-ended, and might be difficult to
answer without some details regarding your paricular use-case (BLE
update, IP-based update, Mesh, etc)
For instance, the LwM2M subsystem provides a mechanism for
receiving a firmware update in the LwM2M client, but the
implementation of where to store the incoming binary data is up to
for a callback example that is triggered on each incoming block of
data.? Documentation for the sample itself doesn't discuss the
firmware update mechanism, but it's here for reference:
Then, there is a robust DFU (Device Firmware Update) subsystem to
help implement the image writing portion of a firmware update as
well as integrate with mcuboot (an MCU bootloader) which would
check an image for validity and then move it into the bootable
application slot. ? See:
Regarding FCB: the initial implementation is done in v1.11, but
the APIs are fairly complex and IMHO it's meant to be used as a
base layer for other higher level implementations to manage
persistent data such as device configuration, system logs and
other uses.? There are several pull requests in-progress which aim
to add these higher level services.? (The first PR will probably
be re-written to use FCB as it's base layer):
Hopefully that helps get you started,
Any help regarding the same would be of great help.
Jr. Embedded Engineer
Research & Development
Please consider the environment before printing this e-mail or
Disclaimer: The information contained herein (including any
accompanying documents) is confidential and is intended solely
for the addressee(s). If you have erroneously received this
message, please immediately delete it and notify the sender.
Also, if you are not the intended recipient, you are hereby
notified that any disclosure, copying, distribution or taking any
action in reliance on the contents of this message or any
accompanying document is strictly prohibited and is unlawful. The
organization is not responsible for any damage caused by a virus
or alteration of the e-mail by a third party or otherwise. The
contents of this message may not necessarily represent the views
or policies of Corvi
Zephyr-devel mailing list
Zephyr-devel mailing list
Zephyr-devel mailing list