TDBStore: Perform garbage collection on failed writes

[davidsaada]

Description

In TDBStore, partial writes may turn the storage unusable, as records are written sequentially. This change adds garbage collection following each write failure, in order to prevent such a scenario.

Pull request type

[x] Fix
[ ] Refactor
[ ] Target update
[ ] Functionality change
[ ] Docs update
[ ] Test update
[ ] Breaking change

[ciarmcom]

@davidsaada, thank you for your changes.
@ARMmbed/mbed-os-storage @ARMmbed/mbed-os-maintainers please review.

[dannybenor]

@davidsaada Very good solution. We need to enhance the triggering of GC to the case a failure was not detected by the software and you will not get a return error. I think we should check the checksum at the end of set_finalize

[davidsaada]

Added read verification following @dannybenor's comment.

[cmonr]

CI started

[mbed-ci]

Test run: FAILED

Summary: 1 of 11 test jobs failed
Build number : 1
Build artifacts

Failed test jobs:

• jenkins-ci/mbed-os-ci_greentea-test

[0xc0170]

CI restarted

[cmonr]

@davidsaada Please take a look at the greentea test failures: http://mbed-os-ci.s3-website-eu-west-1.amazonaws.com/?prefix=jenkins-ci/artifacts/9200/1/greentea-test/685/FAIL/

The K64F looks related to the PR, don't think the NRF52 is.

[davidsaada]

@davidsaada Please take a look at the greentea test failures: http://mbed-os-ci.s3-website-eu-west-1.amazonaws.com/?prefix=jenkins-ci/artifacts/9200/1/greentea-test/685/FAIL/

The K64F looks related to the PR, don't think the NRF52 is.

The NRF52 failure is indeed not related to the PR.
As for the K64F one: The test sets a value, gets it and then compares the read data to the written one - with the failed test producing different results. Now, this test was missing the return code check of the get API, meaning that the call could have failed (maybe due to media error), but we didn't know it. Test was fixed now to include the error code checking (along with a few others with the same problem). Regardless, tried to reproduce the failures on several K64F boards (local & remote), but without success.
Bottom line - please rerun the test. We may either see a real problem or a media (SD) one, according to the results.

[0xc0170]

CI restarated

[mbed-ci]

Test run: FAILED

Summary: 1 of 11 test jobs failed
Build number : 2
Build artifacts

Failed test jobs:

• jenkins-ci/mbed-os-ci_exporter

[0xc0170]

Exporters restarted (pipeline closed error)

[0xc0170]

Bottom line - please rerun the test. We may either see a real problem or a media (SD) one, according to the results.

Tests passed, what shall we do now?

[davidsaada]

Tests passed, what shall we do now?

I think this should merge now. Only thing I can think for that failure was a one time glitch in the storage, which was covered by the missing error code checking.

[dannybenor]

I think this PR should be merged