feat: nostd, core+alloc support #22

roblabla · 2019-08-18T16:53:25Z

Missing: docs, tests and a way to avoid having base64 break our builds to rust-lang/cargo#4866.

dignifiedquire · 2019-08-18T17:06:57Z

Cargo.toml

@@ -50,7 +51,8 @@ name = "key"
 # debug = true

 [features]
-default = []
+default = ["std"]
+std = []


shouldn’t this enable the std features on the crates that got it disabled above?

newpavlov · 2019-08-18T18:49:38Z

I wonder if we can write a custom library for stack allocated bigints bounded by a prime constant. it would be also useful for the SRP implementation. Do you know any prior art libraries which have done it?

dignifiedquire · 2019-08-18T20:27:10Z

doesnt smallvec already do this, given it is the backing structure of the bigints?

…

On 18. Aug 2019, 20:49 +0200, Artyom Pavlov ***@***.***>, wrote: I wonder if we can write a custom library for stack allocating bigints bounded by a prime constant. it would be also useful for SRP implementation. Do you know any prior art libraries which have done it? — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

roblabla · 2019-08-18T20:58:37Z

That's exactly what smallvec does. A SmallVec<[u8; 4096]> will store up to 4096 bytes on the stack, and then switch to heap-allocated vector. BigUint uses a SmallVec<[u64; 4]> (or u32; 8 if u64_digit feature is not used), so digits of up to 256 bit will be stored on the stack.

newpavlov · 2019-08-18T21:08:39Z

I meant bigints without the heap fallback. Since most operations are done modulo some value with a known length, it should be doable? Plus this hypothetical crate can be significantly smaller, since we don't need general purpose bigints.

roblabla · 2019-08-18T21:28:14Z

Well for statically allocated vecs, arrayvec exists and is pretty good. But it would require changing the bigint library to be generic over the storage backend somehow, since otherwise we'd have to copy paste a fairly significant portion of the bigint algorithms over.

dignifiedquire · 2019-10-13T14:32:30Z

@roblabla I have updated num-bigint-dig, want to update this PR so we can ship it?

roblabla · 2019-10-13T15:13:22Z

I'll try to squeeze some time during the week to get this up and running.

roblabla · 2019-10-23T14:44:01Z

I did some fixes here and there, but I'm stuck against a wall again. base64 is enabling the std feature of byteorder, which breaks no_std build. Now, base64 is a dev dependency, but because cargo is, frustratingly, still susceptible to unifying dev-dependencies and dependencies features.

A PR was merged in August allowing nostd usage of base64, but there hasn't been a new release of base64 since. So we're waiting for upstream to release a new version.

tarcieri · 2019-10-23T14:47:57Z

@roblabla you can use subtle_encoding::base64 instead of base64, which is fully no_std friendly, has no dependencies on crates with activated std features, and provides constant-time encoding and decoding if you do end up using it for private keys: https://github.com/iqlusioninc/crates/tree/develop/subtle-encoding

roblabla · 2019-10-25T15:37:35Z

The new base64 dependency breaks Rust 1.31 builds 😭. Subtle-encoding also has an MSRV of 1.36. May I bump the MSRV @dignifiedquire ?

dignifiedquire · 2019-10-25T16:06:14Z

there is always sth :D feel free to bump the version

dignifiedquire · 2019-12-11T00:12:24Z

src/algorithms.rs

@@ -2,6 +2,7 @@ use num_bigint::traits::ModInverse;
 use num_bigint::{BigUint, RandPrime};
 use num_traits::{FromPrimitive, One, Zero};
 use rand::Rng;
+use libm::F64Ext;


Could you update to [email protected], I did that in num-bigint-dig, unfortunately that doesn't have the extension anymore, so you need to do some manual wrapping of things.

dignifiedquire · 2019-12-11T00:12:57Z

.travis.yml

@@ -2,14 +2,16 @@ language: rust

 matrix:
  include:
-    - rust: 1.31.0
+    - rust: 1.34.0


I bumped the msr to 1.36, because of [email protected] in #33

dignifiedquire · 2020-06-11T12:11:47Z

@roblabla merged everything else on master, any chance you want to pick this up again and bring it over the finishline?

roblabla · 2020-06-11T13:06:59Z

I'll take a look!

roblabla · 2020-06-11T13:50:21Z

Some quick thoughts:

RSA now relies on thiserror. This crate is not no-std compatible. I'll probably have to rip it out and implement the traits by hand.
Regarding the possibility of a future no-std no-alloc version, the RSA crate now makes some fairly heavy use of the DynDigest trait, which relies on liballoc. In theory it could be swapped with a trait that takes a buffer instead of returning a boxed slice, and panics if the buffer is the wrong size.
The parse functionality will have to be gated on libstd for now, since simple_asn1 is not no_std-compatible. I don't think people working in no_std environments are likely to need PEM parsing though, so that shouldn't be a big issue.

tarcieri · 2020-06-11T14:09:40Z

You can also use displaydoc if you'd like a proc macro for impl'ing Display for the error type which supports no_std.

roblabla · 2020-06-11T14:19:34Z

I ended up using cargo expand and copy pasting the output of thiserror lol. I'll look into displaydoc. I pushed a first draft of the no-std support. It's WIP and untested, but at least it builds.

Things to do:

I added an "alloc" feature, the idea being that down the line, we might want a no-std no-alloc build to be possible. In order to support this use-case without introducing a breaking change, I figured the best idea would be to add an alloc feature now, and force the build to fail if neither std nor alloc features are given. Does that sound good?
I'll try to get qemu setup to run the tests on ARM on CI.
Fix the tests. I had to play with the dev-dependencies, which broke them :^).

Things that will happen in a follow-up PR:

ASN1 parsing in core+alloc environment.
Find a way to ditch lazy_static. That dependency is a pain in the neck, it uses a feature to enable libcore support (instead of the convention of adding a feature to add libstd support), which means we force lazy_static to spinlock even in std environments.

tarcieri · 2020-06-11T14:20:54Z

+1 to no_std + alloc. Many of the other RustCrypto crates support features which use this combination.

Also +1 to ditching lazy_static. I have similar pains regarding its (mis)use of subtractive cargo features.

You might check out conquer-once as a no_std-friendly replacement (I haven't used it personally, but I've heard good things):

https://github.com/oliver-giersch/conquer-once

~~Edit: hrmm, it looks like conquer-once doesn't provide lazy init on no_std. Alas.~~ Never mind!

roblabla · 2020-06-11T14:28:32Z

Ooh, I hadn't heard of conquer-once. And it does seem to provide lazy init on no_std: https://docs.rs/conquer-once/0.2.0/conquer_once/spin/type.Lazy.html is available in no-std, and is exactly what we want. We should be able to swap between this and the top-level Lazy depending on whether we enabled the std feature!

I'll try it out in a follow-up PR.

tarcieri · 2020-07-17T03:34:12Z

Is this ready to merge? Looks good to me at first glance

.travis.yml

dignifiedquire · 2020-07-29T10:42:58Z

@roblabla almost there one more rebase please and can you check out those clippy warnings please? Sorry for the long turnaround times

roblabla · 2020-08-07T16:11:29Z

Should be good to go I hope!

dignifiedquire

awesome, thanks for getting this through!

dignifiedquire reviewed Aug 18, 2019

View reviewed changes

roblabla force-pushed the nostd branch 3 times, most recently from b9fb1fe to 49505e8 Compare October 23, 2019 14:26

dignifiedquire reviewed Dec 11, 2019

View reviewed changes

roblabla force-pushed the nostd branch from 398c6a1 to 9d171a6 Compare June 11, 2020 14:12

roblabla changed the title ~~feat: nostd support~~ feat: nostd, core+alloc support Jun 11, 2020

roblabla mentioned this pull request Jun 11, 2020

core-only "heapless" support #51

Open

roblabla force-pushed the nostd branch from 24a4b1b to ec31591 Compare June 11, 2020 14:50

dignifiedquire reviewed Jul 17, 2020

View reviewed changes

.travis.yml Outdated Show resolved Hide resolved

roblabla force-pushed the nostd branch 3 times, most recently from 6b45d82 to 9b935e5 Compare July 25, 2020 14:25

roblabla added 7 commits August 5, 2020 16:45

No-std support

4f49813

Fix tests

af5cae2

Cleanly error out when building without the alloc feature

0281322

Run no-std tests on arm-linux-gnu target

780b263

Fix nostd tests

5014c52

Attempt 2 at fixing nostd tests

4b91c82

Fix warnings when running tests in nostd mode

914f12a

roblabla force-pushed the nostd branch from 1903437 to 914f12a Compare August 5, 2020 16:49

fixup! No-std support

62ece78

dignifiedquire approved these changes Aug 7, 2020

View reviewed changes

dignifiedquire merged commit e815294 into RustCrypto:master Aug 7, 2020

alt3r-3go mentioned this pull request Aug 30, 2021

no_std-targeted build fails, getrandom dependency uses std #112

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: nostd, core+alloc support #22

feat: nostd, core+alloc support #22

roblabla commented Aug 18, 2019 •

edited

Loading

dignifiedquire Aug 18, 2019

newpavlov commented Aug 18, 2019 •

edited

Loading

dignifiedquire commented Aug 18, 2019 via email

roblabla commented Aug 18, 2019 •

edited

Loading

newpavlov commented Aug 18, 2019 •

edited

Loading

roblabla commented Aug 18, 2019 •

edited

Loading

dignifiedquire commented Oct 13, 2019

roblabla commented Oct 13, 2019

roblabla commented Oct 23, 2019

tarcieri commented Oct 23, 2019

roblabla commented Oct 25, 2019 •

edited

Loading

dignifiedquire commented Oct 25, 2019

dignifiedquire Dec 11, 2019

dignifiedquire Dec 11, 2019

dignifiedquire commented Jun 11, 2020

roblabla commented Jun 11, 2020

roblabla commented Jun 11, 2020

tarcieri commented Jun 11, 2020

roblabla commented Jun 11, 2020 •

edited

Loading

tarcieri commented Jun 11, 2020 •

edited

Loading

roblabla commented Jun 11, 2020 •

edited

Loading

tarcieri commented Jul 17, 2020

dignifiedquire commented Jul 29, 2020

roblabla commented Aug 7, 2020

dignifiedquire left a comment

feat: nostd, core+alloc support #22

feat: nostd, core+alloc support #22

Conversation

roblabla commented Aug 18, 2019 • edited Loading

dignifiedquire Aug 18, 2019

Choose a reason for hiding this comment

newpavlov commented Aug 18, 2019 • edited Loading

dignifiedquire commented Aug 18, 2019 via email

roblabla commented Aug 18, 2019 • edited Loading

newpavlov commented Aug 18, 2019 • edited Loading

roblabla commented Aug 18, 2019 • edited Loading

dignifiedquire commented Oct 13, 2019

roblabla commented Oct 13, 2019

roblabla commented Oct 23, 2019

tarcieri commented Oct 23, 2019

roblabla commented Oct 25, 2019 • edited Loading

dignifiedquire commented Oct 25, 2019

dignifiedquire Dec 11, 2019

Choose a reason for hiding this comment

dignifiedquire Dec 11, 2019

Choose a reason for hiding this comment

dignifiedquire commented Jun 11, 2020

roblabla commented Jun 11, 2020

roblabla commented Jun 11, 2020

tarcieri commented Jun 11, 2020

roblabla commented Jun 11, 2020 • edited Loading

tarcieri commented Jun 11, 2020 • edited Loading

roblabla commented Jun 11, 2020 • edited Loading

tarcieri commented Jul 17, 2020

dignifiedquire commented Jul 29, 2020

roblabla commented Aug 7, 2020

dignifiedquire left a comment

Choose a reason for hiding this comment

roblabla commented Aug 18, 2019 •

edited

Loading

newpavlov commented Aug 18, 2019 •

edited

Loading

roblabla commented Aug 18, 2019 •

edited

Loading

newpavlov commented Aug 18, 2019 •

edited

Loading

roblabla commented Aug 18, 2019 •

edited

Loading

roblabla commented Oct 25, 2019 •

edited

Loading

roblabla commented Jun 11, 2020 •

edited

Loading

tarcieri commented Jun 11, 2020 •

edited

Loading

roblabla commented Jun 11, 2020 •

edited

Loading