Ensure symlinked realpaths are not loaded twice #3189

rwstauner · 2023-07-27T17:00:08Z

Use a hash to cache realpath lookups
and another to cache realpath -> feature.

This is based off of what I was reading in the CRuby code:
https://github.com/ruby/ruby/blob/499eb3990faeaac2603787f2a41b2d9625e180dc/load.c#L358-L402
https://github.com/ruby/ruby/blob/499eb3990faeaac2603787f2a41b2d9625e180dc/load.c#L1195-L1289
but I could easily be wrong about something here.

Let me know if you see problems with the approach or a better way to do something here.

I did have a few questions remaining:
Should the body of the load_unless_realpath_loaded method be synchronized? Does it make sense to wrap it in with_synchronized_features?

There was a commit in CRuby:
ruby/ruby@972f274#:~:text=To%20avoid%20concurrency%20issues%20when%20rebuilding%20the%20loaded%20features%0Aindex%2C%20the%20building%20of%20the%20index%20itself%20is%20left%20alone%2C%20and%0Aafterwards%2C%20a%20separate%20loop%20is%20done%20on%20a%20copy%20of%20the%20loaded%20feature%0Asnapshot%20in%20order%20to%20rebuild%20the%20realpaths%20hash.

To avoid concurrency issues when rebuilding the loaded features
index, the building of the index itself is left alone, and
afterwards, a separate loop is done on a copy of the loaded feature
snapshot in order to rebuild the realpaths hash.

Do we not need to worry about that here because it's already being called inside with_synchronized_features? Or would that still be a concern?

closes oracle#3138 Use a hash to cache realpath lookups and another to cache realpath -> feature similar to what was done in CRuby: https://github.com/ruby/ruby/blob/499eb3990faeaac2603787f2a41b2d9625e180dc/load.c#L358-L402 https://github.com/ruby/ruby/blob/499eb3990faeaac2603787f2a41b2d9625e180dc/load.c#L1195-L1289

eregon · 2023-07-28T16:39:57Z

Design-wise that change looks pretty hacky to me.
The simplest for TruffleRuby would be to realpath when adding to $LOADED_FEATURES.
But that may not be compatible enough, unsure.

eregon · 2023-07-28T16:40:54Z

src/main/ruby/truffleruby/core/truffle/feature_loader.rb

@@ -140,6 +143,16 @@ def self.find_feature_or_file(feature, use_feature_provided = true)
      end
    end

+    def self.load_unless_realpath_loaded(feature, path)
+      # TODO: does this need to be synchronized?
+      realpath = (@features_realpath_cache[path] ||= File.realpath(path) || path)


AFAIK File.realpath never returns false|nil, always a String, or an exception if some part does not exist.

That did seem to be the case for me but I wasn't sure if it was true on all platforms.
I think I was looking at this C code and it made me want to program defensively (the semantics are probably different, though).
https://github.com/ruby/ruby/blob/7ce4b716bdb5bcfc8b30ffcd034ce7aded1f72b9/load.c#L87-L88

eregon · 2023-07-28T16:43:09Z

src/main/ruby/truffleruby/core/truffle/feature_loader.rb

@@ -140,6 +143,16 @@ def self.find_feature_or_file(feature, use_feature_provided = true)
      end
    end

+    def self.load_unless_realpath_loaded(feature, path)
+      # TODO: does this need to be synchronized?


Yes of course, you could use a TruffleRuby::ConcurrentMap.
But I think this code would need to move to Java anyway, because we cannot have an extra backtrace entry for require, everything must happen in Primitive.load_feature (or another Primitive reusing it).

i was wondering if it should be under with_synchronized_features.

That's a good point, though, might be able to do the whole change in java 🤔

eregon · 2023-07-28T16:44:42Z

src/main/ruby/truffleruby/core/truffle/feature_loader.rb

@@ -35,6 +35,9 @@ module FeatureLoader
    # A snapshot of $LOADED_FEATURES, to check if the @loaded_features_index cache is up to date.
    @loaded_features_version = -1

+    @features_realpath_cache = {}


I don't think this cache is correct, for instance it seems broken if CWD changes or $LOAD_PATH changes.
Probably that part shouldn't be cached.

You could cache absolute_path->realpath though.

good point, we probably want absolute path as the key.

eregon · 2023-07-28T16:45:02Z

src/main/ruby/truffleruby/core/truffle/feature_loader.rb

@@ -292,13 +305,24 @@ def self.get_loaded_features_index
      unless @loaded_features_version == $LOADED_FEATURES.version
        raise '$LOADED_FEATURES is frozen; cannot append feature' if $LOADED_FEATURES.frozen?
        @loaded_features_index.clear
+        previous_realpaths = @features_realpath_cache.dup
+        @features_realpath_cache.clear
+        @loaded_features_realpaths.clear


Maybe this cache never needs to be cleared.

I think it was necessary to satisfy the specs, and I saw the CRuby code doing something simliar:
https://github.com/ruby/ruby/blob/7ce4b716bdb5bcfc8b30ffcd034ce7aded1f72b9/load.c#L363-L364
https://github.com/ruby/ruby/blob/7ce4b716bdb5bcfc8b30ffcd034ce7aded1f72b9/load.c#L377-L384

eregon · 2023-07-28T16:47:20Z

If #3138 is not really an issue in practice I would probably delay this.
It seems pretty messy to do correctly.
I'd rather we convince CRuby to realpath all LOADED_FEATURES eagerly or do that and accept the small semantics difference (it's much simpler and likely more efficient).

rwstauner · 2023-08-07T17:11:35Z

The simplest for TruffleRuby would be to realpath when adding to $LOADED_FEATURES.
But that may not be compatible enough, unsure.

In my testing it seemed that the first time the realpath was stored it was saved to LOADED_FEATURES as whatever was actually requested.
Then if that realpath was requested again, it would be ignored.

I tried to follow what the ruby C code was doing, but it's probably more complicated than what I have so far here.

I'd rather we convince CRuby to realpath all LOADED_FEATURES eagerly or do that and accept the small semantics difference (it's much simpler and likely more efficient).

Might be worth a discussion 👍

eregon · 2023-08-16T12:12:06Z

Action item for me here is I need to file a CRuby issue to suggest using realpath for everything, it would be so much simpler and consistent, and have that discussed in a dev meeting.
In parallel we could try to implement the semantics of "everything realpath" in TruffleRuby and see if that's compatible enough.
In either case I think this is too risky to integrate for the 23.1 release.

Also would be useful to get an answer to #3138 (comment), cc @nirvdrum

oracle-contributor-agreement bot added the OCA Verified All contributors have signed the Oracle Contributor Agreement. label Jul 27, 2023

rwstauner added the shopify label Jul 27, 2023

rwstauner force-pushed the rwstauner/loaded-features-realpaths branch from 1df367c to 89e47ef Compare July 27, 2023 20:03

eregon reviewed Jul 28, 2023

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Ensure symlinked realpaths are not loaded twice #3189

Ensure symlinked realpaths are not loaded twice #3189

rwstauner commented Jul 27, 2023

eregon commented Jul 28, 2023

eregon Jul 28, 2023

rwstauner Aug 7, 2023

eregon Jul 28, 2023

rwstauner Aug 7, 2023

eregon Jul 28, 2023

rwstauner Aug 7, 2023

eregon Jul 28, 2023

rwstauner Aug 7, 2023

eregon commented Jul 28, 2023

rwstauner commented Aug 7, 2023

eregon commented Aug 16, 2023 •

edited

Loading

Ensure symlinked realpaths are not loaded twice #3189

Are you sure you want to change the base?

Ensure symlinked realpaths are not loaded twice #3189

Conversation

rwstauner commented Jul 27, 2023

eregon commented Jul 28, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eregon commented Jul 28, 2023

rwstauner commented Aug 7, 2023

eregon commented Aug 16, 2023 • edited Loading

eregon commented Aug 16, 2023 •

edited

Loading