Repository navigation
Improve random number generation - #258
Merged
Merged
Conversation
Previously, subshells would inherit the same random number sequence as the parent shell, which could lead to unexpected behavior in scripts that rely on randomness. This change ensures that each subshell generates a different sequence of random numbers. The sequence is still deterministic if the user seeds $RANDOM to a specific value. When subshells are forked after seeding to the same value, subshells in the same position in the fork tree will yield the same sequence, but other subshells will yield different sequences. This is achieved by saving the given seed in the `seed_candidate` variable, which is then updated deterministically when a subshell is forked. Fixes #256
This refactoring extracts the generate_seed function from variable.c to util.c, so that it can be used from path.c as well.
The `num` variable is now initialized using `generate_seed()` instead of relying only on the current time. The `n` variable has been replaced with a `salt` variable that is initialized with the current process ID, which helps to reduce the likelihood of generating the same random number in different subshells. The random number generation formula has also been updated to use a larger multiplier, which should help to produce a more diverse set of random numbers. One notable change is that `num` is no longer updated when the temporary file is created successfully in the first iteration. This means that successive calls to `create_temporary_file` will result in same-named temporary files being created as long as the previously created file is removed before the next call, which is often the case in practice.
The `generate_seed' function now uses `clock_gettime` with `CLOCK_REALTIME` to generate a more precise seed for random number generation. If `clock_gettime` is not available, it falls back to using `time(NULL)`. This should provide more entropy for the seed, especially helpful when the function is called multiple times in quick succession.
In environments where uintptr_t is defined, we can use the address of the string buffer to mix into the seed for generating temporary file names. This helps to reduce the likelihood of collisions when multiple subshells sharing the same seed are creating files simultaneously.
The existing algorithm for generating temporary file names is not very secure because the random number seed is derived from public information like the process ID and the current time. This makes it possible for an attacker to guess the file names and create files with those names before the shell does, which can lead all the 100 attempts to fail. This commit adds a call to getentropy() to generate a random salt after 8 failed attempts. Since getentropy() is designed to provide cryptographically secure random numbers, this should make it much harder for an attacker to guess the file names and cause collisions. getentropy() is used only after 8 failed attempts to avoid unnecessarily consuming entropy from the system.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
[Summary by GitHub Copilot]
This pull request makes significant improvements to the
$RANDOMspecial variable in the shell, ensuring that random number sequences are independent between subshells and more secure. It also introduces new configuration checks and enhances the randomness of temporary file names. Comprehensive tests have been added to validate the new randomization behavior.Random Number Generation Improvements:
$RANDOMvariable now produces independent sequences of random numbers, preventing subshells from repeating the parent shell’s sequence. This is achieved by introducing a new reseeding mechanism and seed management logic. (NEWS,NEWS.ja,variable.c,exec.c,variable.h, [1] [2] [3] [4] [5] [6] [7] [8] [9]util.c,util.h, [1] [2] [3] [4]Configuration and Portability:
getentropyandclock_gettimefor improved randomness and portability. (configure, configureR1299-R1362)Temporary File Randomization:
path.c, [1] [2] [3] [4] [5]Testing Enhancements:
$RANDOMproduces different values in subshells, when re-seeded, and across shell invocations, as well as to ensure correct behavior when setting or unsetting the variable. (tests/random-y.tst, [1] [2] [3] [4]Code Modernization:
util.c,util.h,variable.h, [1] [2] [3]These changes together make random number generation in the shell more robust, secure, and predictable for scripting use cases.