evidence: check evidence hasn't already been stored #4632

cmwaters · 2020-04-02T18:56:06Z

Closes: #4614

Description

Uses the Has() function from the DB store interface to first confirm that a piece of received evidence is not already stored before commencing verification

For contributor use:

Wrote tests
Updated CHANGELOG_PENDING.md
Linked to Github issue with discussion and accepted design OR link to spec that describes this work.
Updated relevant documentation (docs/) and code comments
Re-reviewed Files changed in the Github PR explorer

melekes

cool!

don't forget to add CHANGELOG_PENDING.md entry and maybe write a test

evidence/reactor.go

melekes · 2020-04-03T07:41:34Z

evidence/errors.go

+	"github.com/tendermint/tendermint/types"
+)
+
+type ErrInvalidEvidence struct {


why expose?

Okay, perhaps I need a better understanding of when errors should be exposed or not because almost all the errors in all other packages are exposed, so I have just been copying them

I think in general it's just good practice. Unless, however, you know with 100% certainty, the error will never be used/needed by the outside (package) world.

If we'll be creating a typed error every time we return an error in public function, we will end up with pile of extra code, which is not needed in reality

evidence/errors.go

evidence/pool.go

cmwaters · 2020-04-03T08:16:08Z

Previously the evidence reactor would StopPeerForError when any error was returned from AddEvidence. Now, I've changed it such that only when the validation fails that it disconnects the peer. It doesn't seem necessary to disconnect because it has already received the evidence. Is this okay? Do you think we should further punish the peer by banning them?

Also is it okay that AddNewEvidence ignores any error that the DB might return?

evidence/errors.go

evidence/pool.go

alexanderbez

LGTM 👍

alexanderbez · 2020-04-06T14:31:54Z

evidence/errors.go

+	"github.com/tendermint/tendermint/types"
+)
+
+type ErrInvalidEvidence struct {


I think in general it's just good practice. Unless, however, you know with 100% certainty, the error will never be used/needed by the outside (package) world.

evidence/store.go

codecov-io · 2020-04-07T05:58:04Z

Codecov Report

Merging #4632 into master will increase coverage by 0.00%.
The diff coverage is 46.15%.

@@           Coverage Diff           @@
##           master    #4632   +/-   ##
=======================================
  Coverage   66.00%   66.01%           
=======================================
  Files         220      221    +1     
  Lines       19766    19800   +34     
=======================================
+ Hits        13047    13070   +23     
- Misses       5676     5686   +10     
- Partials     1043     1044    +1

Impacted Files	Coverage Δ
evidence/errors.go	`0.00% <0.00%> (ø)`
rpc/core/evidence.go	`0.00% <0.00%> (ø)`
evidence/reactor.go	`61.01% <40.00%> (-0.81%)`	⬇️
evidence/store.go	`82.60% <62.50%> (-5.63%)`	⬇️
evidence/pool.go	`76.00% <66.66%> (-2.09%)`	⬇️
blockchain/v0/reactor.go	`73.91% <0.00%> (-1.74%)`	⬇️
blockchain/v0/pool.go	`78.98% <0.00%> (+0.31%)`	⬆️
blockchain/v1/peer.go	`95.91% <0.00%> (+0.62%)`	⬆️
consensus/replay.go	`73.64% <0.00%> (+0.77%)`	⬆️
... and 3 more

evidence/errors.go

evidence/reactor.go

melekes · 2020-04-07T08:18:38Z

evidence/errors.go

+type ErrEvidenceAlreadyStored struct{}
+
+func (e ErrEvidenceAlreadyStored) Error() string {
+	return fmt.Sprint("evidence is already stored")


Suggested change

return fmt.Sprint("evidence is already stored")

return "evidence is already stored"

Add Has function, create better handling of errors when adding evidence, usage of error types.

melekes · 2020-04-14T06:45:48Z

Previously the evidence reactor would StopPeerForError when any error was returned from AddEvidence. Now, I've changed it such that only when the validation fails that it disconnects the peer. It doesn't seem necessary to disconnect because it has already received the evidence. Is this okay? Do you think we should further punish the peer by banning them?

It's not necessary to disconnect, agree. we should ban peers who send us invalid evidence.

Also is it okay that AddNewEvidence ignores any error that the DB might return?

probably not

cmwaters requested review from melekes and tessr as code owners April 2, 2020 18:56

cmwaters self-assigned this Apr 2, 2020

cmwaters added the C:evidence Component: Evidence label Apr 2, 2020

melekes reviewed Apr 3, 2020

View reviewed changes

cmwaters requested a review from melekes April 3, 2020 14:17

melekes reviewed Apr 3, 2020

View reviewed changes

evidence/errors.go Show resolved Hide resolved

alexanderbez reviewed Apr 3, 2020

View reviewed changes

evidence/errors.go Outdated Show resolved Hide resolved

evidence/pool.go Show resolved Hide resolved

cmwaters requested review from alexanderbez and melekes April 6, 2020 13:37

alexanderbez approved these changes Apr 6, 2020

View reviewed changes

melekes reviewed Apr 7, 2020

View reviewed changes

evidence/errors.go Show resolved Hide resolved

evidence/errors.go Show resolved Hide resolved

evidence/reactor.go Show resolved Hide resolved

melekes reviewed Apr 7, 2020

View reviewed changes

evidence/reactor.go Outdated Show resolved Hide resolved

melekes reviewed Apr 7, 2020

View reviewed changes

melekes approved these changes Apr 7, 2020

View reviewed changes

check evidence hasn't already been stored

c69b1f7

Add Has function, create better handling of errors when adding evidence, usage of error types.

cmwaters force-pushed the callum/check-inbound-evidence branch from 3f517ae to c69b1f7 Compare April 7, 2020 08:31

cmwaters added the S:automerge Automatically merge PR when requirements pass label Apr 7, 2020

Merge branch 'master' into callum/check-inbound-evidence

fa810eb

mergify bot merged commit 88d7007 into master Apr 7, 2020

mergify bot deleted the callum/check-inbound-evidence branch April 7, 2020 08:42

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

evidence: check evidence hasn't already been stored #4632

evidence: check evidence hasn't already been stored #4632

cmwaters commented Apr 2, 2020 •

edited

Loading

melekes left a comment

melekes Apr 3, 2020

cmwaters Apr 3, 2020

alexanderbez Apr 6, 2020

melekes Apr 6, 2020

cmwaters commented Apr 3, 2020

alexanderbez left a comment

alexanderbez Apr 6, 2020

codecov-io commented Apr 7, 2020 •

edited

Loading

melekes Apr 7, 2020

melekes commented Apr 14, 2020

	return fmt.Sprint("evidence is already stored")
	return "evidence is already stored"

evidence: check evidence hasn't already been stored #4632

evidence: check evidence hasn't already been stored #4632

Conversation

cmwaters commented Apr 2, 2020 • edited Loading

Description

melekes left a comment

Choose a reason for hiding this comment

melekes Apr 3, 2020

Choose a reason for hiding this comment

cmwaters Apr 3, 2020

Choose a reason for hiding this comment

alexanderbez Apr 6, 2020

Choose a reason for hiding this comment

melekes Apr 6, 2020

Choose a reason for hiding this comment

cmwaters commented Apr 3, 2020

alexanderbez left a comment

Choose a reason for hiding this comment

alexanderbez Apr 6, 2020

Choose a reason for hiding this comment

codecov-io commented Apr 7, 2020 • edited Loading

Codecov Report

melekes Apr 7, 2020

Choose a reason for hiding this comment

melekes commented Apr 14, 2020

cmwaters commented Apr 2, 2020 •

edited

Loading

codecov-io commented Apr 7, 2020 •

edited

Loading