The IS_UNSERIALIZED check here does not work if the string is
interned (serialized with file_cache_only=0) but unserialization
happens with file_cache_only=1. In this case the unserializde
string will be in the str area after mem, which is not included
in the script size, and which is also not accessible at this
point without threading through more information. Work around
the problem by checking for the serialized representation instead.
Currently, CE_CACHE on strings is only used with opcache interned strings. This
patch extends usage to non-opcache interned strings as well. This means that
most type strings can now make use of CE_CACHE even if opcache is not loaded,
which allows us to remove TYPE_HAS_CE kind, and fix some discrepancies
depending on whether a type stores a resolved or non-resolved name.
There are two cases where CE_CACHE will not be used:
* When opcache is not used and a permanent interned string (that is not an
internal class name) is used as a type name during the request. In this case
we can't allocate a map_ptr index for the permanent string, as it would be
not be in the permanent map_ptr index space.
* When opcache is used but the script is not cached (e.g. eval'd code or
opcache full). If opcache is used, we can't allocate additional map_ptr
indexes at runtime, because they may conflict with indexes allocated by
opcache.
In these two cases we would end up not using CE caching for property types
(argument/return types still have the separate cache slot).
zend_accel_get_class_name_map_ptr() for "self" and "parent" will
currently try to determine which class these refer to, and then
initialize the CE_CACHE on those strings.
However, this shouldn't be necessary: We already initialize
CE_CACHE on all class declaration names, so it should be covered
through that already.
When running without opcache, static_members_table is shared with
default_static_members_table. This is visible in reflection output,
because ReflectionProperty::getDefaultValue() will return the
current value, rather than the default value.
Address this by never sharing the table, which matches the behavior
we already see under opcache.
Fixes bug #80821.
Closes GH-7299.
1. Update: http://www.php.net/license/3_01.txt to https, as there is anyway server header "Location:" to https.
2. Update few license 3.0 to 3.01 as 3.0 states "php 5.1.1, 4.1.1, and earlier".
3. In some license comments is "at through the world-wide-web" while most is without "at", so deleted.
4. fixed indentation in some files before |
This is needed by both fibers and opcache (and GH-6903 also uses it),
so make it a common structure that can be used by any functionality
storing warnings/errors.
If the script will be cached in SHM (!corrupted), then we cannot
allocate the static variables on the arena. Instead do the same
thing we do during normal persistence and allocate a map ptr slot.
This reverts commit 4440ac3ed6.
This fix wasn't correct, as we do retain the arena-allocation
for the file cache only case. This will need some other way to
reconcile both modes.
During unserialization, static_variables and runtime_cache may be
allocated on the arena, and then later freed when the checkpoint
is released.
As we're only doing a single large allocation here, simply move
this to the normal allocator, so there is no interference with
other arena allocations.
This is generalization of idea, that was previously usesd for caching
resolution of class_entries in zend_type. Now very similar mechanizm is
used for general zend_string into zend_class_entry resolution.
Interned zend_string with IS_STR_CLASS_NAME_MAP_PTR GC_FLAG uses its
refcount to adress corresponding zend_class_entry cache slot.
The refcount keeps an offset to this slot from CG(map_ptr_base).
Flag may be checked by ZSTR_HAS_CE_CACHE(str), cache slot may be read by
ZSTR_GET_CE_CACHE(str) and set by ZSTR_SET_CE_CACHE(str, ce).
static_variables should be treated the same way as all other
op_array components nowadays (only static_variables_ptr is
special). There's no need to persist/serialize it is separately
per shared op_array.
Currently, dynamically declared functions and closures are inserted
into the function table under a runtime definition key, and then later
possibly renamed. When opcache is not used and a file containing a
closure is repeatedly included, this leads to a very large memory leak,
as the no longer needed closure declarations will never be freed
(https://bugs.php.net/bug.php?id=76982).
With this patch, dynamic functions are instead stored in a
dynamic_func_defs member on the op_array, which opcodes reference
by index. When the parent op_array is destroyed, the dynamic_func_defs
it contains are also destroyed (unless they are stilled used elsewhere,
e.g. because they have been bound, or are used by a live closure). This
resolves the fundamental part of the leak, though doesn't completely
fix it yet due to some arena allocations.
The main non-obvious change here is to static variable handling:
We can't destroy static_variables_ptr in destroy_op_array, as e.g.
that would clear the static variables in a dynamic function when
the op_array containing it is destroyed. Static variable destruction
is separated out for this reason (we already do static variable
destruction separately for normal functions, so we only need to
handle main scripts).
Closes GH-5595.
We already use <= for IS_SERIALIZED(), but the same general
problem can also occur for IS_UNSERIALIZED(). We don't seem to
hit this in practice prior to GH-5595 though.
When a method is inherited, the static variables will now always
use the initial values, rather than the values at the time of
inheritance. As such, behavior no longer depends on whether
inheritance happens before or after a method has been called.
This is implemented by always keeping static_variables as the
original values, and static_variables_ptr as the modified copy.
Closes GH-6705.
This is a new transparent technology that eliminates overhead of PHP class inheritance.
PHP classes are compiled and cached (by opcahce) separately, however their "linking" was done at run-time - on each request. The process of "linking" may involve a number of compatibility checks and borrowing methods/properties/constants form parent and traits. This takes significant time, but the result is the same on each request.
Inheritance Cache performs "linking" for unique set of all the depending classes (parent, interfaces, traits, property types, method types involved into compatibility checks) once and stores result in opcache shared memory. As a part of the this patch, I removed limitations for immutable classes (unresolved constants, typed properties and covariant type checks). So now all classes stored in opcache are "immutable". They may be lazily loaded into process memory, if necessary, but this usually occurs just once (on first linking).
The patch shows 8% improvement on Symphony "Hello World" app.
The `zend_system_id` is a (true global) system ID that fingerprints a process state. When extensions add engine hooks during MINIT/startup, entropy is added the system ID for each hook. This allows extensions to identify that changes have been made to the engine since the last PHP process restart.
Closes GH-5871
op_arrays can be shared on two levels: Either the op_array is
completely shared, or it is distinct but shares all members
(apart from static_variables).
The the op_array is distinct, we need to make sure to properly
initialize the MAP_PTR structures.
These systems has two versions of the iovec interface dependent on
compiler flags passed, the legacy version causing little build issue.
Closes GH-5939.
From an engine perspective, named parameters mainly add three
concepts:
* The SEND_* opcodes now accept a CONST op2, which is the
argument name. For now, it is looked up by linear scan and
runtime cached.
* This may leave UNDEF arguments on the stack. To avoid having
to deal with them in other places, a CHECK_UNDEF_ARGS opcode
is used to either replace them with defaults, or error.
* For variadic functions, EX(extra_named_params) are collected
and need to be freed based on ZEND_CALL_HAS_EXTRA_NAMED_PARAMS.
RFC: https://wiki.php.net/rfc/named_params
Closes GH-5357.
If opcache.record_warnings is enabled, opcache will record
compilation warnings and replay them when the file is included
again. The primary use case I have in mind for this is automated
testing of the opcache file cache.
This resolves bug #76535.
If the class is already linked, we need to serialize and
unserialize INDIRECTed static properties. Normally these would
be set up when copying from cache.
Instead of having a completely independent encoding for type list
entries. This is going to use more memory, but I'm not particularly
concerned about that, as type unions that contain multiple classes
should be uncommon. On the other hand, this allows us to treat
top-level types and types inside lists mostly the same.
A new ZEND_TYPE_FOREACH macros allows to transparently treat list
and non-list types the same way. I'm not using it everywhere it could be
used for now, just the places that seemed most obvious.
Of course, this will make any future type system changes much simpler,
as it will not be necessary to duplicate all logic two times.
Fix file cache serialization of property types. I'm changing the
overall type serialization format to perform additional adjustments
in order to yield a plausible pointer for zend_type, rather than
using an entirely separate serialization format, as was previously
done. That would have been annoying to extend to the case of CE
pointers.
If the script was serialized as file_cache_only (thus non-immutable)
and then gets unserialized into SHM, we need to allocate a new
run_time_cache slot and can't use the serialized arena pointer.
According to RFC: https://wiki.php.net/rfc/union_types_v2
The type representation now makes use of both the pointer payload
and the type mask at the same time. Additionall, zend_type_list is
introduced as a new kind of pointer payload, which is used to store
multiple class types. Each of the class types is a tagged pointer,
which may be either a class name or class entry. The latter is only
used for typed properties, while arguments/returns will instead use
cache slots. A type list can contain a mix of both names and CEs at
the same time, as not all classes may be resolvable.
One thing this is missing is support for union types in arginfo
and stubs, which I want to handle separately.
I've also dropped the special object code from the JIT implementation
for now -- I plan to add this back in a different form at a later time.
For now I did not want to include non-trivial JIT changes together
with large functional changes.
Another possible piece of follow-up work is to implement "iterable"
as an internal alias for "array|Traversable". I believe this will
eliminate quite a few special-cases that had to be implemented.
Closes GH-4838.
We now store the pointer payload and the type mask separately. This
is in preparation for union types, where we will be using both at
the same time.
To avoid increasing the size of arginfo structures, the
pass_by_reference and is_variadic fields are now stored as part of
the type_mask (8-bit are reserved for custom use).
Different types of pointer payloads are distinguished based on bits
in the type_mask.
