Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

cli: program upgrade with offline signing (--sign-only mode) #33860

Merged
merged 8 commits into from
Dec 14, 2023

Conversation

norwnd
Copy link
Contributor

@norwnd norwnd commented Oct 25, 2023

Problem

Resolves #23975.

Summary of Changes

This PR implements offline signing capabilities for solana program upgrade command (solana program-v4 deploy is out of scope of this PR, solana program deploy works as before and doesn't support offline signing) through --sign-only and some helper flags. See rendered docs this PR adds for how it is supposed to work.

Notable things:

  • as you can see in tutorial docs, user needs to pass quite a lot of params between commands involved in offline signing (--sign-only mode), it could potentially be improved by serialising everything into 1 (or maybe 2-3) params - but I think it's better to leave this possible improvement out of this PR (and maybe consider adding to program-v4 later on)
  • durable nonces aren't yet supported (probably better do it in a separate PR)
  • changing program authority once it's been set to offline signer isn't implemented in this PR

@mergify mergify bot added community Community contribution need:merge-assist labels Oct 25, 2023
@mergify mergify bot requested a review from a team October 25, 2023 12:11
@norwnd norwnd changed the title program: --sign-only (offline signing) cli: program deploy with offline signing (--sign-only mode) Oct 25, 2023
cli/src/program.rs Outdated Show resolved Hide resolved
cli/src/program.rs Outdated Show resolved Hide resolved
Comment on lines 69 to 70
pub trait Signer {
pub trait Signer: Debug {
Copy link
Contributor Author

@norwnd norwnd Oct 25, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure what's the best approach in Rust to be able to println!( a Trait at runtime, but this is the simplest one I got working, is it okay to keep Debug constraint here (or should I remove it) ?

The downside - seems like I have to derive Debug for each implementation (including test mocks), didn't find a way to do it in more re-usable manner.

The error compiler complains about looks like this:

error[E0521]: borrowed data escapes outside of function
    --> cli/src/program.rs:1869:23
     |
1858 |     program_signers: Option<&[&dyn Signer]>,
     |     ---------------           - let's call the lifetime of this reference `'1`
     |     |
     |     `program_signers` is a reference that is only valid in the function body
...
1869 |     println!("signers {:?}", program_signers);
     |                       ^^^^
     |                       |
     |                       `program_signers` escapes the function body here
     |                       argument requires that `'1` must outlive `'static`
     |
     = note: this error originates in the macro `$crate::format_args_nl` which comes from the expansion of the macro `println` (in Nightly builds, run with -Z macro-backtrace for more info)

or

error[E0521]: borrowed data escapes outside of function
    --> cli/src/program.rs:1869:23
     |
1858 |     program_signers: Option<&[&dyn Signer]>,
     |     ---------------           - let's call the lifetime of this reference `'1`
     |     |
     |     `program_signers` is a reference that is only valid in the function body
...
1869 |     println!("signers {:?}", &program_signers);
     |                       ^^^^
     |                       |
     |                       `program_signers` escapes the function body here
     |                       argument requires that `'1` must outlive `'static`
     |
     = note: this error originates in the macro `$crate::format_args_nl` which comes from the expansion of the macro `println` (in Nightly builds, run with -Z macro-backtrace for more info)

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's revert this change: Signer has pubkey() in its implementation, which is all that we should ever print anyway, so the previous implementation gives more flexibility.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I used it to debug some things, wasn't sure whether it's better to leave it out or not,

but removed now (see latest PR version)

sdk/src/signer/mod.rs Outdated Show resolved Hide resolved
@norwnd norwnd force-pushed the program-sign-only branch from 423bd08 to 2c57502 Compare October 25, 2023 13:14
sdk/src/signer/mod.rs Outdated Show resolved Hide resolved
@norwnd
Copy link
Contributor Author

norwnd commented Oct 26, 2023

Wrote some tests, and found rough edges ... maybe hold on proper review for now, I'll push another commit soon to address those.

@norwnd norwnd force-pushed the program-sign-only branch 2 times, most recently from 0ac8a3d to fd5af67 Compare October 28, 2023 13:37
@norwnd
Copy link
Contributor Author

norwnd commented Oct 28, 2023

Wrote some tests, and found rough edges ... maybe hold on proper review for now, I'll push another commit soon to address those.

Finished tests, applied adjustments as needed and squashed everything into single commit, keeping PR in draft until all of these are added,

but expecting to get feedback on the overall approach first,

cc @joncinque

Comment on lines 201 to 207
/// loaded. If multiple equivalent (same pub key) signers are provided - only
/// one of those will be returned in the result, such that NullSigner(s)
/// always get lower priority.
///
/// The returned value includes all of the `bulk_signers` that were not
/// `None`, and maybe the default signer, if it was loaded.
/// `None`, and maybe the default signer (if it was loaded). There is no
/// guarantees on resulting signers ordering.
Copy link
Contributor Author

@norwnd norwnd Oct 28, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is needed to allow specifying same signer seamlessly multiple times (with some of those "duplicates" being NullSigner(s)). One case I found this to be helpful in is doing something like this:

solana program deploy --sign-only --fee-payer <OFFLINE_SIGNER_PUB_KEY> --program <PROGRAM_SIGNER_PUB_KEY> --upgrade-authority <OFFLINE_SIGNER> --buffer <BUFFER_PUB_KEY> --blockhash <VALUE> --max-len <VALUE> --min-rent-balance <VALUE>

where <OFFLINE_SIGNER> can have same pubkey as <OFFLINE_SIGNER_PUB_KEY> (and even perhaps <PROGRAM_SIGNER_PUB_KEY>).

Also, I'm guessing ALL the changes I've introduced in clap-utils must be copied 1-to-1 into clap-v3-utils ?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry, I'm not understanding the motivation for this. Why would someone specify the same signer different ways?

Unless I'm missing something, I think these changes should be in a separate PR so we can discuss there, since they don't seem to be needed for offline signing of program upgrades.

Copy link
Contributor Author

@norwnd norwnd Dec 2, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Essentially, this is a prerequisite work for implementing this "refactoring": #33860 (comment)

One reason (closely-related to this PR) is the use-case of specifying both --upgrade-authority <OFFLINE_SIGNER_PUB_KEY> and --signer <OFFLINE_SIGNER_SIGNATURE> when doing solana program upgrade:

solana program upgrade --fee-payer /Users/norwnd/.config/solana/id.json --program-id BgY6WizXe2Hfrq71TjjvV2ntjiMX3tjHKAARvg47SJUC --upgrade-authority GuKy3G22F4avoPKGtpqBEuoQYLEWaCRmncGGc5Q7JoPL --buffer C1o2XLFAU6EqZuMPVfW9C2gFEtBxMuXaNkEYLyNtsbiU --blockhash qyof8GDHGrVwUqtiDAFfit9MY62Xub2i1cEAnVXpQ62 --signer GuKy3G22F4avoPKGtpqBEuoQYLEWaCRmncGGc5Q7JoPL=Qonons7tm1oRxR2akT4cVu9HyXQjah2DxAzS8Ht7RwU3Gx1Nk2SUTv9X9LMa8mYFrPSRZbpYDsNVPG5BNNLGsh7

since OFFLINE_SIGNER_PUB_KEY will convert to NullSigner we need to differentiate between these 2 signers (preferring the most "complete" of these 2). Which is why I didn't create a separate PR for it - but can potentially move these changes into "a separate PR with preliminary changes" this PR depends on, if that helps with anything.

To sum up, I'm using the power of NullSigner to avoid unnecessary ifs and duplicate code:

  • I'm using signer_of_or_null_signer to treat pubkey arguments as any other signer (through NullSigner)
  • generate_unique_signers will choose the best signer for me (preferring real signers over NullSigner(s))
  • fn process_program_deploy, fn process_program_upgrade will take only signer and work with it (corresponding pubkey, instead of being a 2nd "duplicate" argument, is abstracted away behind Signer interface)

Alternative way to address this (along the lines of what's done in current master) would be to try parse --upgrade-authority as both signer and pubkey in ("upgrade", Some(matches)) => { and doing some if-checks to decide which should be used, plus we'll need more if-checks in fn process_program_deploy and fn process_program_upgrade (which I removed as mentioned here). Same for --fee-payer.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In the CLI, this is all done for you with signer_of. If you follow the code to signer_from_path_with_config, you'll see that if a pubkey is provided, the function will then check the --signer args to see if a signature is provided, and use that signature for any transaction.

For example, see how process_transfer is using the different signers:

pub fn process_transfer(

So the code is all in there for you already!

Copy link
Contributor Author

@norwnd norwnd Dec 6, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Right, what you describe here solves it specifically for --signer argument,

I've described here why we might want to solve it for other parameters too, might be a bit of a stretch on my part, wdyt ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Changed to get rid of this NullSigner-based approach I've suggested above (see latest PR version).

@github-actions github-actions bot added the stale [bot only] Added to stale content; results in auto-close after a week. label Nov 16, 2023
@github-actions github-actions bot closed this Nov 24, 2023
@CriesofCarrots
Copy link
Contributor

@norwnd , are you still working on this? Happy to re-open if so.

@norwnd
Copy link
Contributor Author

norwnd commented Nov 26, 2023

Hi @CriesofCarrots! I'm happy to proceed if you or @joncinque can provide some feedback on this PR,

it's in Draft because there are couple of minor things I still need to add, but it's ready for review otherwise (I'd rather wait for general-approach review first though).

@joncinque joncinque reopened this Nov 28, 2023
@joncinque
Copy link
Contributor

Hi @norwnd -- I sincerely apologize for taking such a long time to get to this. I really appreciate the effort that you've put into this.

Looking at how many changes this requires, I'm leaning towards a slightly different way to implement it, through a totally new command called upgrade which only does the upgrade instruction.

Currently, solana program deploy does way too much work to cover many different situations. As you found out, since it works for initial deployments and upgrades, it has many possible branches to figure out. This forced you to take into account too many situations.

So rather than mess with solana program deploy, let's keep that functionality as it exists currently, and add solana program upgrade, which only sends or signs the UpgradeableLoaderInstruction::Upgrade instruction. What do you think?

@norwnd
Copy link
Contributor Author

norwnd commented Nov 28, 2023

Hey @joncinque, no worries,

So rather than mess with solana program deploy, let's keep that functionality as it exists currently, and add solana program upgrade, which only sends or signs the UpgradeableLoaderInstruction::Upgrade instruction. What do you think?

I think we can move the upgrade functionality into a separate command to get cleaner result - I'll get to it in a day or two to try it out. Just want to note, --sign-only feature still complicates solana program deploy either way, so pls let me know if there is anything else we can/should do about it.

Update:

I just sketched out a "very rough diff" for introducing solana program upgrade would look like (assuming, solana program deploy still preserves an ability to upgrade program - for backward compatibility) - 72ad73f - and as it turns out, the improvement is rather marginal. Considering long-term code maintenance, I'd prefer to add it (since it does somewhat simplify complex some conditions I had to add in process_program_deploy func) even at the cost of some code duplication (which we can maybe further de-dup if we want).

It doesn't quite address the overall complexity of --sign-only though. I did consider splitting solana program deploy in two (current version, and --sign-only version) - but that's seems quite arbitrary and splitting by deploy/upgrade is more fitting imo (if we prefer splitting it at all).

@norwnd norwnd force-pushed the program-sign-only branch 2 times, most recently from d6eee19 to 72ad73f Compare November 28, 2023 16:16
@joncinque
Copy link
Contributor

It doesn't quite address the overall complexity of --sign-only though. I did consider splitting solana program deploy in two (current version, and --sign-only version) - but that's seems quite arbitrary and splitting by deploy/upgrade is more fitting imo (if we prefer splitting it at all).

Ah sorry, there was a misunderstanding. For solana program upgrade, it should only send one transaction, which contains the UpgradeableLoaderInstruction::Upgrade, and that's it. It assumes that the buffer is already prepared for the upgrade.

The interface would be something like:

solana program upgrade <PROGRAM_ID> <BUFFER_ADDRESS> [--upgrade-authority <SIGNER>] [--fee-payer <SIGNER>] [--sign-only]  [--blockhash <BLOCKHASH>] [--nonce <NONCE_PUBKEY>] [--nonce-authority <SIGNER>] [--signer <PUBKEY>=<SIGNATURE> ...]

Or, in the enum, this would give:

    Upgrade {
        program_pubkey: Pubkey,
        buffer_pubkey: Pubkey,
        upgrade_authority_signer_index: SignerIndex,
        fee_payer: SignerIndex,
        // ... plus all the sign-only stuff ...
        sign_only: bool,
        dump_transaction_message: bool,
        blockhash_query: BlockhashQuery,
        nonce_account: Option<Pubkey>,
        nonce_authority: SignerIndex,
    },

Does that make more sense?

@github-actions github-actions bot removed the stale [bot only] Added to stale content; results in auto-close after a week. label Nov 29, 2023
@norwnd norwnd changed the title cli: program deploy with offline signing (--sign-only mode) cli: program upgrade with offline signing (--sign-only mode) Dec 11, 2023
@norwnd
Copy link
Contributor Author

norwnd commented Dec 11, 2023

Squashed and rebased, proceeding to resolve other review-comments still left from previous reviews (I'll ping when it's ready for another look).

@norwnd
Copy link
Contributor Author

norwnd commented Dec 11, 2023

Hey @joncinque, I think I've addressed all the pending comments, pending your feedback now. Appreciate thorough reviews!

Copy link
Contributor

@joncinque joncinque left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks really great! We're down to just tiny little changes, then this will be ready to merge. Thanks for all of your patience and edits

cli/src/program.rs Outdated Show resolved Hide resolved
cli/src/program.rs Show resolved Hide resolved
cli/src/program.rs Outdated Show resolved Hide resolved
cli/src/program.rs Outdated Show resolved Hide resolved
cli/src/program.rs Outdated Show resolved Hide resolved
docs/src/cli/deploy-a-program.md Outdated Show resolved Hide resolved
docs/src/cli/deploy-a-program.md Outdated Show resolved Hide resolved
docs/src/cli/deploy-a-program.md Outdated Show resolved Hide resolved
docs/src/cli/deploy-a-program.md Outdated Show resolved Hide resolved
docs/src/cli/deploy-a-program.md Outdated Show resolved Hide resolved
@joncinque
Copy link
Contributor

joncinque commented Dec 12, 2023

Oh also, it looks like the docs just got refactored, so you'll have to make the doc change in docs/src/cli/examples/deploy-a-program.md instead.

@norwnd
Copy link
Contributor Author

norwnd commented Dec 13, 2023

Oh also, it looks like the docs just got refactored, so you'll have to make the doc change in docs/src/cli/examples/deploy-a-program.md instead.

Thanks for the heads-up! This last push ^ was a rebase, it moved my stuff into appropriate file, but seems like there are path-dependent links that broke (I'll push a follow-up commit(s) to fix that).

@joncinque joncinque added the CI Pull Request is ready to enter CI label Dec 13, 2023
@solana-grimes solana-grimes removed the CI Pull Request is ready to enter CI label Dec 13, 2023
@joncinque
Copy link
Contributor

joncinque commented Dec 13, 2023

All the changes look great to me! There's a failure in the sanity checks in CI due to some trailing whitespace https://buildkite.com/solana-labs/solana/builds/105554#018c63f9-526c-4080-bafc-d27ec92899d5, so once that's sorted out and CI passes, we'll merge this in!

@joncinque joncinque added the CI Pull Request is ready to enter CI label Dec 13, 2023
@solana-grimes solana-grimes removed the CI Pull Request is ready to enter CI label Dec 13, 2023
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you revert the changes here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hm, that's creeped from a sloppy rebase, fixed in f3e466f

Copy link

codecov bot commented Dec 13, 2023

Codecov Report

Merging #33860 (f3e466f) into master (501458a) will decrease coverage by 0.1%.
Report is 7 commits behind head on master.
The diff coverage is 14.7%.

Additional details and impacted files
@@            Coverage Diff            @@
##           master   #33860     +/-   ##
=========================================
- Coverage    81.9%    81.8%   -0.1%     
=========================================
  Files         819      819             
  Lines      220931   221019     +88     
=========================================
- Hits       180954   180950      -4     
- Misses      39977    40069     +92     

@joncinque joncinque added the CI Pull Request is ready to enter CI label Dec 14, 2023
@solana-grimes solana-grimes removed the CI Pull Request is ready to enter CI label Dec 14, 2023
Copy link
Contributor

@joncinque joncinque left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is finally ready to go, thanks so much for your time, effort, and understanding!

cheer

@joncinque joncinque merged commit 0be3b6e into solana-labs:master Dec 14, 2023
47 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
community Community contribution need:merge-assist
Projects
None yet
Development

Successfully merging this pull request may close these issues.

RFC: Add --sign-only flag to solana program deploy
4 participants