No subject

For other "conditionally unconditional" things, we don't have a
corresponding check on user->foo.

For conditional stuff, non-NULLness of user->foo is used to track
whether we decided to dump the corresponding record; for consistency
here, if we have system_supports_poe() && err == 0, then that's
sufficient (though in prior versions of this code, POR_EL0 dumping was
conditional and so the extra check did do something...)


In any case, if some allocation fails then we splat out with a SIGSEGV
before modifying the user task state to deliver the signal (in
setup_return() etc.)

If The user's POR_EL0 value is being clobbered before we get here, we
would save the wrong value -- so the code would be broken anyway.


So, as Joey says, this is probably fine, but the user->poe_offset check
looks superfluous.  The kernel will splat on us here and kill the thread
if it's NULL anyway.

[...]

Cheers
---Dave