If same allocator (alloc call) allocates multiple dynamic memory blocks,
the additional blocks are added to extended memory block list, and
the allocator list reference count and total sizes are increased.
When statistics are calculated, it is only needed to process the
allocator list, not the whole memory block list.
Added also optional hash function to calculate the (start) index where
memory block information is stored on extended memory block list. Hash
is calculated from memory block address.
It is expected that hash distributes the free index searches evenly
to whole memory area allocated for extended memory block list. So if
area is large enough, in most cases the index pointed by the hash, or
the index few indexes from that, should be empty and available for use.
Similarly, when memory block is freed, the search for the block on
the extended memory block list, starts from the index pointed by the
hash.
If hash is not used, extended memory block list searches start always
from index 0.
Nanostack dynamic memory tracker tracks memory allocations. It has following
interface:
ns_dyn_mem_tracker_lib_alloc: is called on alloc
ns_dyn_mem_tracker_lib_free: is called on free
ns_dyn_mem_tracker_lib_step: updates the lifetime of individual memory
blocks (is used to detected which blocks are permanent and which
have been allocated lately)
ns_dyn_mem_tracker_lib_allocator_lists_update: updates allocator lists,
there are lists for top allocators, permanent allocators and allocators
going to permanent allocators list
ns_dyn_mem_tracker_lib_max_snap_shot_update: can be used to update memory
maximum usage snap shot, this can be called on alloc if snap shot is
needed
BLOCK2 code-branch was missing handling for duplicate packets. As part of the fix, added also
a call to update the duplicate package data via a new function
sn_coap_protocol_update_duplicate_package_data_all.
The new implementation handles all CoAP messages, not just those with COAP_MSG_TYPE_ACKNOWLEDGEMENT.
Add also 4.13 (Request Entity Too Large) responses to duplicate info list.
Add client library configurations for DEFAULT_RESPONSE_TIMEOUT and SN_COAP_DUPLICATION_MAX_TIME_MSGS_STORED.
Increased the default timeouts of DEFAULT_RESPONSE_TIMEOUT and SN_COAP_DUPLICATION_MAX_TIME_MSGS_STORED to 300 seconds.
These two are critical parameters for low-bandwidth high-latency networks. The defaults should be more geared towards such networks that are likely to have issues with transmissions.
The increased defaults can increase the runtime HEAP usage when there is a lot of duplicates or retransmissions.
-Added the mbedtls,crypto,psa,filesystem,fat,littlefs in baremetal.json to resolve compiler issue
-Disable metrics that are not available for bare metal
-Moved the baremetal.json inside TESTS/configs directory
-Added the baremetal.json on Tests directory to enable bare-metal
-Conditional enable of greentea metrics
Note:
Run green tea with bare metal
mbed test -m target -t GCC_ARM -n tests-mbed_platform-transaction --app-config TESTS/baremetal.json
This is related to https://github.com/ARMmbed/mbed-os/issues/11256
When TARGET_PSA is enabled on Cortex-A boards, the SECURE
is redeclared. So to minimize the impact we can redeclare this in
val.h which is used only by Mbed. And `security_t` itself is used in
mbed-os\components\TARGET_PSA\TESTS, but the SECURE string is not used.
Renamed to `caller_security_t` which aligns with
https://github.com/ARM-software/psa-arch-tests
Random initialization sequence is causing start up issues in multiple platform
when done at construction phase.
The right thing is to delay the random initialization to later stage when the
message id is actually required. This provides system to do all necessary allocation
upfront without causing any random race condition at startup phase.
By default CoAP will create a copy of the whole data to be passed to application and it keeps the backward compatibility.
If enabled, application must NOT free the payload when it gets the COAP_STATUS_PARSER_BLOCKWISE_MSG_RECEIVED status.
And application must call sn_coap_protocol_block_remove() instead.
- Reduce heap footprint by storing only single block when receiving a blockwise message.
* User is now responsible of freeing the data by calling sn_coap_protocol_block_remove() and must not free the payload separately.
- Bug fix: Request blockwise transfer if incoming payload length is too large and when it comes without block indication.
We'll use the new ARM-software/psa-arch-tests directly instead for PSA
Crypto API 1.0b3. This commit removes the crypto compliance tests only
for now, leaving attestation and storage tests.
- Store ACK's also into duplicate info list.
- ROM size optimization. Flash size has gone down ~1100 bytes.
**Closed issues:**
- IOTCLT-3592 - Client does not handle Duplicate ACK messages during blockwise registration correctly
- Add the no confidentiality & no replay protection flags
- Add actual size parameter in PS/ITS get APIs
- Change a few size parameters from uint32_t to size_t
Prevent compilation issues when someone has included <stdint.h> before
a header file that needs to include <ns_list.h>.
Some toolchains like ARM C 5 will not provide UINT_FAST8_MAX in C++
unless __STDC_LIMIT_MACROS is defined, and if this was not defined the
first time <stdint.h> was included, it's too late.
We can get the maximum value for our unsigned list offset by casting -1
to it, thanks to modulo arithmetic.
Make sn_coap_protocol_linked_list_duplication_info_remove API to public. User might want to delete some messages from the duplicate list.
Enable support for unified client configuration.
Arto Kinnunen
Mika Leppänen
Teemu Takaluoma
Antti Yli-Tokola
Darryl Green
Risto Huhtala
Martin Kojtal
RAJKUMAR KANAGARAJ
jeromecoutant
Vikas Katariya
int_szyk
Yogesh Pande
Seppo Takalo
Hugues Kamba
Jaeden Amero
David Saada
Kevin Bracey
Oren Cohen