DLPX-95232 Merge conflict with upstream drgn - #75
Merged
mmaybee merged 354 commits intoAug 27, 2025
Merged
Conversation
Signed-off-by: Omar Sandoval <osandov@osandov.com>
…ols-by It'd be useful to be able to control debug info finders in addition to options. Rather than adding yet another set of command-line options, co-opt --try-symbols-by and --no-symbols-by. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Drgn has a set of globals provided in interactive mode, but they can be useful to other modes (e.g. for executing statements directly). They may also be useful to client code. Make this a separate, public helper. It does make the globals_func argument to run_interactive() a bit much, but that's ok. Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
The Python interpreter has "-c" which allows directly running code provided as a command line argument. This is useful for quick tests, without needing to write a script and execute it, and without needing to run the interactive interpreter. It can also be used in scripts or one-liners. Unfortunately, "-c" is already used in the drgn CLI. However, this functionality would be quite useful. Let's add it to the CLI using the option "-e", which is short for "execute" or "exec", the underlying Python function. Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Prior to Python 3.10, the object returned by @staticmethod apparently cannot be called unless bound to the class instance. So the new changes in the fixed commit cause the CLI to fail on Python < 3.10. We can resolve this in a variety of ways, but the one which doesn't result in any static checker warnings is the simplest: just move this function out to a module-level function. There are no users in sub-classes which need to be converted. Fixes: 4a6a9f3 ("cli: allow enabling/disabling debug info finders with --{try,no}-symbols-by") Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Do this for consistency with the other modes. Fixes: 150ee76 ("cli: add -e option to exec() code directly") Signed-off-by: Omar Sandoval <osandov@osandov.com>
With -e, args.script doesn't necessarily mean that we're running a script. That's a bug waiting to happen, so rename it and add a local variable for whether we're running a script. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Redirecting drgn's stdin (e.g., to a pipe or a file) is really broken: with the enhanced REPL, it fails with "termios.error: (25, 'Inappropriate ioctl for device')" and "AttributeError: 'UnixConsole' object has no attribute '_UnixConsole__buffer'". With the basic REPL, it prints ps1/ps2 prompts and expects extra newlines to terminate indented blocks. The standard Python REPL handles this by checking if stdin is a terminal. Unfortunately, code.interact()/_pyrepl don't do this, so let's do the check and exec manually. Reported-by: Stephen Brennan <stephen.s.brennan@oracle.com> Reported-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Omar Sandoval <osandov@osandov.com>
Also cover script mode and piping in a script and check that we set __name__, __file__, sys.path, sys.argv, prog, and the default program appropriately. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This is a shortcut for `for _ in prog.loaded_modules: pass` or `list(prog.loaded_modules())`, which is used in several test cases and is occasionally useful elsewhere. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Include the kernel version and silence the loaded module iterator warning. Signed-off-by: Omar Sandoval <osandov@osandov.com>
…_debug_info() We don't need iterator_tried_missing since we can just check num_missing (renamed from num_warnings). Signed-off-by: Omar Sandoval <osandov@osandov.com>
Instead of manually bolding the warning, always bold critical messages and log as critical or warning depending on whether we're missing the main debug info. Signed-off-by: Omar Sandoval <osandov@osandov.com>
…all_function() docstring wake_up_process() is one of the best use cases for call_function(), so use it as the first example. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Signed-off-by: Omar Sandoval <osandov@osandov.com>
My debuginfod optimizations for the Linux kernel haven't been deployed anywhere other than Fedora's debuginfod servers. On other servers, it's better to not even try since downloads will take forever. Do a sketchy check to determine whether the kernel is from Fedora and disable the debuginfod finder if not. It can still be reenabled manually with --try-symbols-by debuginfod or by a plugin. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Signed-off-by: Omar Sandoval <osandov@osandov.com>
pyroute 0.9.1 dropped support for Python < 3.9. On Python 3.6 and 3.7, it fails with an ImportError from typing that the try/except in tests/linux_kernel/helpers/test_tc.py catches by accident, but on 3.8 it fails with a TypeError: 'type' object is not subscriptable. Fix it by installing an older version of pyroute2 on those Python versions. Signed-off-by: Omar Sandoval <osandov@osandov.com>
An object can be absent for multiple reasons: e.g., the object is a placeholder, it was optimized out, or its value is represented using a DWARF opcode we don't support. We print all of these as "<absent>", which is technically correct but criminally vague. Add an absence reason to struct drgn_object/drgn.Object. For now, this only has one possible value, "OTHER", but it will be made more specific in follow-up commits. Signed-off-by: Omar Sandoval <osandov@osandov.com>
The most common ways users will run into absent objects are optimized
out local variables and inlined functions. Add an absence reason for
this and use it for DWARF, then use it to display such objects with more
familiar terminology ("<optimized out>").
We might want to distinguish between truly optimized out values and
unsaved registers, but we can improve that later.
See osandov#488.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Rather than failing hard on an unknown DWARF expression opcode, log a warning and return an absent object with a NOT_IMPLEMENTED absence reason. Since DW_OP_(GNU_)entry_value is somewhat common and we already know about it, we still classify it as OPTIMIZED_OUT rather than NOT_IMPLEMENTED. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Famous last words: "it hasn't changed since v2.6.38, so let's hardcode it for now." Linux 6.15 changed the value of DCACHE_MOUNTED, which broke path_lookup(). I sent a patch converting the DCACHE_* flags to an enum [1], but we need to unbreak 6.15 (with a version check, unfortunately) until that lands. [1]: https://lore.kernel.org/linux-fsdevel/177665a082f048cf536b9cd6af467b3be6b6e6ed.1744141838.git.osandov@fb.com/T/#u Signed-off-by: Omar Sandoval <osandov@osandov.com>
Linux 6.15 renamed kernfs_node::parent to __parent, which broke a couple of kernfs helpers. Add a kernfs_parent() helper that handles the change and use it in the existing helpers. Signed-off-by: Omar Sandoval <osandov@osandov.com>
We already have this internally, so let's make it public. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This one needed updates to the kernfs helpers and path_lookup(). Signed-off-by: Omar Sandoval <osandov@osandov.com>
I'm experimenting with kprobes-based breakpoint support via kmodify (not coming soon but eventually). Enable CONFIG_KPROBES so I can test it. Signed-off-by: Omar Sandoval <osandov@osandov.com>
I think I only used drgn_module_find_or_create() instead of drgn_module_find_or_create_relocatable() because at some point earlier in development, the latter didn't have new_ret. Change it so we can make drgn_module_find_or_create() static. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This is more type safe and more consistent with the Python API. (drgn_module_find() still exists as a static function in debug_info.c.) Signed-off-by: Omar Sandoval <osandov@osandov.com>
argparse apparently interprets % in the help string as %-style formatting. We don't want that, but we also don't want to escape it in the string literal and make drgndoc deal with unescaping it, so escape it in the framework right before we pass it to argparse. Signed-off-by: Omar Sandoval <osandov@osandov.com>
The nastiest part of this one is handling the ambiguities between expressions, command line flags, pipelines, and redirections. Other than that, the expression syntax that crash supports is very limited, so it's fairly simple. Signed-off-by: Omar Sandoval <osandov@osandov.com>
…lpers The crash tree command visits red-black tree entries in pre-order by default. Add helpers to do this, which are straightforward variants of the existing in-order helpers. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This way we can make kind default to "*", which will be more convenient for most of the existing callers and the upcoming list and tree commands. Signed-off-by: Omar Sandoval <osandov@osandov.com>
There are a few functions in drgn.commands._builtin.crash.structunion that will also be needed for the list and tree commands. Move them to drgn.commands.crash. For now, they're considered private, but they could be made public if we needed. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This one has nice correspondence to existing drgn helpers, although the various offset options are a bit of a pain to generate drgn code for. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This should only be logged if we don't find a uprobe consumer, but we're always logging it because we never mark that we found a consumer. Fixes: a5da128 ("tools/fsrefs.py: check for references from uprobes") Signed-off-by: Omar Sandoval <osandov@osandov.com>
Linux kernel commit cac5cefbade9 ("sched/smp: Make SMP unconditional")
(in v6.17) changed init_task.comm from "swapper" to "swapper/0" on !SMP
kernels. We have a couple of tests that check for the old behavior.
Replace them with different checks that work regardless of kernel.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
This is our preferred format, but repr() returns a single-quoted string. Add a hacky helper function to return a double-quoted repr when it's preferrable and use it where appropriate. Signed-off-by: Omar Sandoval <osandov@osandov.com>
argparse allows specifying a different name per argument by passing a tuple as the metavar. Handle that. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This is needed for the crash mod command. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Since we just added the helper for module-specific taints, let's add one for the global taints. Signed-off-by: Omar Sandoval <osandov@osandov.com>
I wasn't planning on implementing -s or -S at first, but they're easy enough to do with the module API. Signed-off-by: Omar Sandoval <osandov@osandov.com>
These are trivial. Signed-off-by: Omar Sandoval <osandov@osandov.com>
We need a couple of ioctls for upcoming test cases. Annoyingly, the
_IO{,R,W,WR} macros used to define ioctl numbers are
architecture-dependent. Add Python functions based on these macros,
generated from the kernel source code, along with the script used to
generate them.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Instead, require that the file already exists when mkswap() is called, like mkswap(8), and get the file or partition size automatically. This will simplify upcoming tests. Signed-off-by: Omar Sandoval <osandov@osandov.com>
Upcoming tests for swap helpers need to write a partition table. We could use an external command, but the only one provided in BusyBox is fdisk, which is a pain to script. The format is simple enough, so add a function to do it ourselves. Signed-off-by: Omar Sandoval <osandov@osandov.com>
These are needed for the crash kmem -i and swap commands. Signed-off-by: Omar Sandoval <osandov@osandov.com>
This one wasn't super high priority, but I implemented the necessary helpers while working on the kmem -i command, and it doesn't have any options, so it was quick. Signed-off-by: Omar Sandoval <osandov@osandov.com>
mmaybee
force-pushed
the
dlpx/pr/mmaybee/ea2b8bac-b860-436f-a705-4f194be02b33
branch
from
August 27, 2025 15:10
c11a228 to
2a0c489
Compare
mmaybee
marked this pull request as ready for review
August 27, 2025 15:38
sebroy
approved these changes
Aug 27, 2025
There was a problem hiding this comment.
Thank you for doing this! We could probably follow this up with a change to restore that deleted file so that it stops being a merge conflict in the future. For now, let's move forward with this to unblock us.
That said, it looks like the package doesn't build in your pre-push run: https://selfservice-jenkins.eng-tools-prd.aws.delphixcloud.com/job/linux-pkg/job/develop/job/build-package/job/drgn/job/pre-push/44/console
I'm interpreting this as perhaps this requires a new version of libkdumpfile, which is another linux-pkg package that hasn't been updated in a long time stuck behind its own merge conflict.
delphix/libkdumpfile#39 is merged now, so we should be able to depend on |
prakashsurya
approved these changes
Aug 27, 2025
mmaybee
deleted the
dlpx/pr/mmaybee/ea2b8bac-b860-436f-a705-4f194be02b33
branch
August 27, 2025 21:06
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
10 participants
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.
Problem
There is a small confict with the auto-merge with the upstream drgn repo.
The file
.pre-commit-config.yamlhas been deleted by us, but was modifiedupstream.
Solution
Delete file from incomming merge.
Testing Done
https://selfservice-jenkins.eng-tools-prd.aws.delphixcloud.com/job/appliance-build-orchestrator-pre-push/12033