Inline the lookup whether a function is observed at all.
This strategy is also used for FRAMELESS calls. If the frameless call is observed, we instead allocate a call frame and push the arguments, to call the the function afterwards.
Doing so is still a performance benefit as opposed to executing individual INIT_FCALL+SEND_VAL ops. Thus, even if the frameless call turns out to be observed, the call overhead is slightly lower than before.
If the internal function is not observed at all, the unavoidable overhead is fetching the FLF zend_function pointer and the run-time cache needs to be inspected.
As part of this work, it turned out to be most viable to put the result operand on the ZEND_OP_DATA instead of ZEND_FRAMELESS_ICALL_3, allowing seamless interoperability with the DO_ICALL opcode.
This is a bit unusual in comparison to all other ZEND_OP_DATA usages, but seems to not pose problems overall.
There is also a small issue resolved: trampolines would always use the ZEND_CALL_TRAMPOLINE_SPEC_OBSERVER function due to zend_observer_fcall_op_array_extension being set to -1 too late.
The usage of the current API within an observer handler leads to bugs like https://bugs.xdebug.org/view.php?id=2232.
Given two observer handlers, next to each other. The first one is executed. It removes itself. The second observer handler gets moved to the place of the first. The first one returns. The handler in the second slot is fetched, but is now NULL, because the it's now in the slot of the first observer; i.e. the second handler is skipped and the begin/end symmetry guarantee is violated.
Providing the next handler to the caller is a zero-cost way to avoid any impact in the paths of zend_observe_fcall_begin/end.
Signed-off-by: Bob Weinand <bobwei9@hotmail.com>
There are two main motivations to this:
a) The logic for handling internal and userland observation can be unified.
b) Unwinding of observed functions on a bailout does notably not include observers. Even if users of observers were to ensure such handling themselves, it would be impossible to retain the relative ordering - either the user has to unwind all internal observed frames before the automatic unwinding (zend_observer_fcall_end_all) or afterwards, but not properly interleaved.
Signed-off-by: Bob Weinand <bobwei9@hotmail.com>
This is achieved by tracking the observers on the run_time_cache (with a fixed amount of slots, 2 for each observer).
That way round, if the run_time_cache is freed all associated observer data is as well.
This approach has been chosen, as to avoid any ABI or API breakage.
Future versions may for example choose to provide a hookable API for run_time_cache freeing or similar.
The motivation for this change is to prevent extensions from having to check executor globals for the current execute_data during function call init. A previous implementation of the observer API initialized the function call from runtime cache initialization before execute_data was allocated which is why zend_function was passed in.
But now that the observer API is implemented via opcode specialization, it makes sense to pass in the execute_data. This also keeps the API a bit more consistent for existing extensions that already hook zend_execute_ex.
Closes GH-6209
