Remove P3A protobuf reporting

[rillian]

Description

Once we've confirmed the json-format P3A reports are behaving properly, and we've adjusted the stats backend to use them over the old protobuf format, we should stop sending the old format.

[diracdeltas]

@rillian does this mean that https://github.com/brave/brave-core/blob/920bba00b6efb79aa6a1844c2dcca41eee066442/components/p3a/p3a_message.cc#L22 is no longer being used?

some of that looks unsafe to me but not obviously exploitable... still would be good to remove that function if no longer needed

there is a potential overflow here https://github.com/brave/brave-core/blob/920bba00b6efb79aa6a1844c2dcca41eee066442/components/p3a/p3a_message.cc#L58
like if metastring.size() and metric_value_str.size() are UINT_MAX/2, metastring.size() + metric_value_str.size() is 0 which will bypass that CHECK
but then the memcpy's in https://github.com/brave/brave-core/blob/920bba00b6efb79aa6a1844c2dcca41eee066442/components/p3a/p3a_message.cc#L60-L62 will be out of bounds of data

[rillian]

@diracdeltas Unfortunately the code is still needed to send P2A messages until we've also migrated that backend to json reporting.

AFAIR the overflow you pointed out isn't possible with data as constructed; all the literal values are too small. I can add checked arithmetic to guard against values from corrupted state though?

Thanks for the review!

[diracdeltas]

I can add checked arithmetic to guard against values from corrupted state though?

sgtm!

[GeetaSarvadnya]

QA blocked until #23754 is resolved

[LaurenWags]

Since this is being reverted by #23754, should this issue now be release-notes/exclude?

wdyt @rillian @rebron @GeetaSarvadnya ?

[Uni-verse]

Verified on Samsung GS 21 running version 12 using 1.41.93, Chromium 103.0.5060.114

Went through STR/Test Plan from brave/brave-core#13539 (comment)

PASS - P3A reports are sent on a fresh profile
PASS - Message all go to p3a-json.brave.com and are formatted as readable JSON.
PASS - Traffic is sent to p3a.brave.com

Encountered - #23922

[GeetaSarvadnya]

Since this is being reverted by #23754, should this issue now be release-notes/exclude?

Yes, this should be release note exclude

[GeetaSarvadnya]

Verification PASSED on

Brave | 1.41.93 Chromium: 103.0.5060.114 (Official Build) (64-bit)
-- | --
Revision | a1c2360c5b02a6d4d6ab33796ad8a268a6128226-refs/branch-heads/5060@{#1124}
OS | Windows 10 Version 21H2 (Build 19044.1766)

Basically we ensured P3A and P2A works as it works in 1.40.x due to brave/brave-core#14015

• Confirmed that P3A traffic is sent to the endpoint p3a-json.brave.com in 1.41.x
• Confirmed that all the P3A metrics are sent to the json format to p3a-json.brave.com in 1.41.x
• Confirmed that after reverting P3A protobuf submission Revert stopping P3A protobuf submission for 1.41 brave-core#14015, the traffic is sent to the old p3a.brave.com endpoint in the encoded format in 1.41.x
• Confirmed P2A metrics data sent correctly in both old p2a.brave.com and new json p2a.json.brave.com formats in 1.41.x
• Confirmed brave://local-state file has 6 records for p2a and same is shown in both p2a-json.brave.com and p2a.brave.com endpoints
• Confirmed brave://local-state file has 47 records for p3a and same is shown in both p3a-json.brave.com and p3a.brave.com endpoints

p3a data in brave://local-state	p3a.brave.com	p3a-json.brave.com

p2a data in brave://local-state	p2a.brave.com	p2a-json.brave.com

[rillian]

Since this is being reverted by #23754, should this issue now be release-notes/exclude?

It shouldn't be in the 1.41 release notes, but it's addressed for 1.42 by brave/brave-core#14017 I've updated the milestone but left the release-notes/include tag.