Closed adfoster-r7 closed 1 year ago
The difference is that on_file_autoloaded
is only invoked if a require
on a managed file returned true
(here), and Ruby guarantees that if multiple threads are loading the same file, only one of them will succeed.
Do you have a way to reproduce or any other insights? Is this a multi-threaded application?
I'm afraid as it's only happened once I don't have any additional datapoints; here's all the information I have:
operating-system: macOS-11.7.7
ruby-version: 3.0.2
bundler: 2.2.33
zeitwerk: 2.6.8
The application does make use of threads, and user interaction via a readline
prompt i.e. reacting to user input, which then attempted to load additional ruby files.
From the CI output I can see that the series of events were:
Msf::Ui::Console::CommandDispatcher::Payload
which zeitwerk crashed onMsf::Ui::Console::CommandDispatcher::Payload
doesn't load again - or at least it seems like zeitwerk doesn't try again, based on the same crash not happeningI based the third point on seeing future calls to the same failure line in the CI output giving a normal uninitialized constant
exception when attempting to load the same constant
when Msf::MODULE_PAYLOAD
dispatcher = Msf::Ui::Console::CommandDispatcher::Payload
Error:
[read] [-] Error while running command use: uninitialized constant Msf::Ui::Console::CommandDispatcher::Payload
Did you mean? Msf::Payload
Call stack:
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/modules.rb:769:in `cmd_use'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:581:in `run_command'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:530:in `block in run_single'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:524:in `each'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:524:in `run_single'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/shell.rb:168:in `run'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/metasploit/framework/command/console.rb:48:in `start'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/metasploit/framework/command/base.rb:82:in `start'
msfconsole:23:in `<main>'
I'm not sure if it makes sense to raise a better error to help with debugging this scenario in the future - feel free to close the issue if there's nothing immediately actionable from this issue. I'll keep an eye open to see if it happens again, just wanted to raise the issue for visibility - thanks! :+1:
Just another datapoint :+1:
Hit this again on another CI run, this time failing on a top level constant:
module Msf
module Ui
module Console
module ModuleCommandDispatcher
include Msf::Ui::Console::CommandDispatcher
include Msf::Ui::Console::ModuleArgumentParsing <---
Stack trace:
Error while running command use: wrong number of arguments (given 0, expected 2)
Call stack:
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/zeitwerk-2.6.8/lib/zeitwerk/loader/helpers.rb:118:in `cpath'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/zeitwerk-2.6.8/lib/zeitwerk/loader/callbacks.rb:12:in `on_file_autoloaded'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/zeitwerk-2.6.8/lib/zeitwerk/kernel.rb:31:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/module_command_dispatcher.rb:15:in `<module:ModuleCommandDispatcher>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/module_command_dispatcher.rb:12:in `<module:Console>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/module_command_dispatcher.rb:5:in `<module:Ui>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/module_command_dispatcher.rb:4:in `<module:Msf>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/module_command_dispatcher.rb:3:in `<top (required)>'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/bootsnap-1.16.0/lib/bootsnap/load_path_cache/core_ext/kernel_require.rb:32:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/bootsnap-1.16.0/lib/bootsnap/load_path_cache/core_ext/kernel_require.rb:32:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/zeitwerk-2.6.8/lib/zeitwerk/kernel.rb:30:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/payload.rb:11:in `<class:Payload>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/payload.rb:10:in `<module:CommandDispatcher>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/payload.rb:6:in `<module:Console>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/payload.rb:5:in `<module:Ui>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/payload.rb:4:in `<module:Msf>'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/payload.rb:3:in `<top (required)>'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/bootsnap-1.16.0/lib/bootsnap/load_path_cache/core_ext/kernel_require.rb:32:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/bootsnap-1.16.0/lib/bootsnap/load_path_cache/core_ext/kernel_require.rb:32:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/vendor/bundle/ruby/3.0.0/gems/zeitwerk-2.6.8/lib/zeitwerk/kernel.rb:30:in `require'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/msf/ui/console/command_dispatcher/modules.rb:769:in `cmd_use'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:581:in `run_command'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:530:in `block in run_single'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:524:in `each'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/dispatcher_shell.rb:524:in `run_single'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/rex/ui/text/shell.rb:168:in `run'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/metasploit/framework/command/console.rb:48 in `start'
/Users/runner/work/metasploit-framework/metasploit-framework/lib/metasploit/framework/command/base.rb:82:in `start'
msfconsole:23:in `<main>'
Environment:
Image: macos-11
ruby-version: 3.0.2
bundler: 2.2.33
If there's any extra logging I could add to help aid with debugging the issue I'd be happy to wire that up - unfortunately I don't know enough about zeitwerk's internals to jump to a solution just yet 😄
How strange!
I believe it is good that it happens in CI, because normally execution is more constrained in there. Sometimes, not even multi-threaded. Is your test suite in CI multi-threaded?
Something that could help is logging the activity of the loader by issuing loader.log!
(docs about logging).
Is that the CI of a Rails application or of a library? Do you know where is the loader instantiated?
I dislike closing issues without understanding their root cause. The root cause here could be Zeitwerk, but could also be Ruby, or Metasploit. We just don't know.
On paper, that callback is protected by require
. If Zeitwerk had a threading issue, it is unlikely that apps are not hitting it and is showing up in the issue tracker.
We don't have a way to reproduce or feedback that could help me guide debugging.
So, by now I'll close. If there's extra data some day, please do not hesitate to followup.
Agreed, thanks! Will keep an eye out for this in the future, and hopefully I'll catch it with the Zeitwerk debugger enabled one day to shed more light here :+1:
Hi there; Thanks for the project :+1:
We spotted this stacktrace on our CI setup that we've never seen before:
Unfortunately I don't have specific replication steps, it's the first time I've seen the issue in thousands of CI runs. I'm guessing the cref returned nil here:
https://github.com/fxn/zeitwerk/blob/ed7a7c8792cea5ca10a83241fa699939cd598000/lib/zeitwerk/loader/callbacks.rb#L10-L12
I noticed the other
on_dir_autoloaded
code has a surrounding mutex, so maybe something similar is missing here - not sure. Potentially adding a guard clause/error for this specific scenario might be useful to help with future debugging