Re: [PATCH 2/4] pid: Export find_task_by_vpid for use in external modules

From: Robin Murphy
Date: Wed May 09 2018 - 07:17:43 EST


Hi Eric,

On 09/05/18 05:59, Eric W. Biederman wrote:
Kim Phillips <kim.phillips@xxxxxxx> writes:

This patch is in the context of allowing the Coresight h/w
trace driver suite to be loaded as modules. Coresight uses
find_task_by_vpid when running in direct capture mode (via sysfs)
when getting/setting the context ID comparator to trigger on
(/sys/bus/coresight/devices/<x>.etm/ctxid_pid).

Aside from my objection about how bad an interface a pid in sysfs is.
The implementation of coresight_vpid_to_pid is horrible.

The code should be just:

static inline pid_t coresight_vpid_to_pid(pid_t vpid)
{
rcu_read_lock();
pid = pid_nr(find_vpid(vpid));
rcu_read_unlock();

return pid;
}
Which takes find_task_by_vpid out of the picture.

But reading further I am seeing code writing a pid to hardware. That is
broken. That is a layering violation of the first order. Giving
implementation details like that to hardware.

Note that the value here is nothing more than a token - the CoreSight hardware doesn't actually *do* anything with it other than match it against the same value which we also stash in the CPU in much the same fashion - see CONFIG_PID_IN_CONTEXTIDR for, if you'll pardon the pun, context.

TL;DR: the CPU has a special register whose only purpose is to allow the OS help external debug tools identify the currently executing process, by writing some arbitrary identifier in there. The trace hardware can spit that identifier out into the trace stream whenever it changes, such that the user can see context switches easily. Newer trace hardware can also use it to actively filter what the capture at source, such that only the portions of interest are traced at all. We could in theory make up any old value, but as I understand it the PID is/was the most user-friendly and easily correlatable thing to hand, and it's now probably too well-established to reasonably change.

Robin.

Any chance while you are working on this you can modify this code so
that it does something sensible and defensible instead of every line of
code I read be wrong in at least one detail?

Thank you,
Eric