drgn.helpers.linux.mm: have cmdline() and environ() return None for k…

[pcc]

…ernel tasks

Sometimes I need to enumerate all tasks in the system and print their command lines, among other information. This fails for kernel tasks (i.e. tasks without mm):

Traceback (most recent call last):
File "", line 2, in
File "[...]/drgn/helpers/linux/mm.py", line 1276, in cmdline
arg_start = mm.arg_start.value_()
^^^^^^^^^^^^^^^^^^^^^
_drgn.FaultError: address is not mapped: 0x140

It seems reasonable for this not to cause an exception but rather to return a value that the caller can conveniently recover from (consider also the behavior of cat /proc/2/cmdline). Therefore, change cmdline() to return None for kernel tasks. Do the same for environ() for consistency.

[osandov]

This seems reasonable. I'm debating whether I'd prefer returning an empty list rather than None. That would be more backwards compatible (e.g., type-checked scripts wouldn't start getting errors now that cmdline() can return None), and still in the spirit of /proc/2/cmdline and /proc/2/environ being empty files. I could be swayed either way.

[pcc]

I also considered returning an empty list. None seemed more accurate though, since it is possible for a userspace task to have an empty environment or (before commit dcd46d897adb70d63e025f175a00a89797d31a43) an empty argv. My gut feeling is that Python type checking is unlikely to be used in the sort of ad-hoc scripts that use drgn, so I wouldn't expect this to be likely to break anything.

[osandov]

Yeah, distinguishing between kernel threads and truly empty seems like a good enough reason for this, so please update with the nitpicks and I'll merge this, thanks!