Skip to content

Zend: Fix GH-22257 type confusion in Exception::getTraceAsString(). #22263

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
devnexen wants to merge 6 commits into from
+56 −6
Closed

Zend: Fix GH-22257 type confusion in Exception::getTraceAsString(). #22263

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
05e9698
Zend: Fix GH-22257 type confusion in Exception::getTraceAsString().
devnexen Jun 9, 2026
303ee1b
apply @ndossche suggestion
devnexen Jun 9, 2026
fd6d35d
CREDITS when due
devnexen Jun 10, 2026
2ec1f96
suggestion 2
devnexen Jun 10, 2026
2b8c0a0
defer typed property default restore dtor via var_push_dtor_value().
devnexen Jun 11, 2026
eb42872
address arnaud feedback
devnexen Jun 25, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 29 additions & 0 deletions Zend/tests/gh22257.phpt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,29 @@
--TEST--
GH-22257 (Type confusion / OOB read unserializing an Exception with a non-array trace)
--CREDITS--
Igor Sak-Sakovskiy (Positive Technologies)
--FILE--
<?php
/* A crafted, deliberately truncated payload makes the nested value of the typed
* "array $trace" property fail to unserialize. On that failure path the slot used
* to keep the half-built (non-array) value, and the partially-built Exception was
* then exposed to getTraceAsString() through SplHeap's delayed __unserialize(),
* type-confusing the object as a HashTable. The slot is now reset to the property
* default, so the run completes without an out-of-bounds read. */
$n = "\x00";
try {
unserialize(
'O:9:"Exception":1:{s:16:"' . $n . 'Exception' . $n . 'trace";' .
'O:8:"stdClass":2:{s:1:"0";O:10:"SplMaxHeap":2:{i:0;a:0:{}i:1;a:2:{' .
's:5:"flags";i:0;s:13:"heap_elements";a:2:{i:0;s:0:"";i:1;R:1;}}}z}}'
);
} catch (\Throwable $e) {
for (; $e; $e = $e->getPrevious()) {
printf("%s: %s\n", $e::class, $e->getMessage());
}
}
echo "OK\n";
?>
--EXPECTF--
Warning: unserialize(): Error at offset %d of %d bytes in %s on line %d
OK
15 changes: 11 additions & 4 deletions ext/standard/var_unserializer.re
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -677,10 +677,17 @@ second_try:
}

if (!php_var_unserialize_internal(data, p, max, var_hash)) {
if (info && Z_ISREF_P(data)) {
/* Add type source even if we failed to unserialize.
* The data is still stored in the property. */
ZEND_REF_ADD_TYPE_SOURCE(Z_REF_P(data), info);
if (info) {
if (Z_ISREF_P(data)) {
ZEND_REF_ADD_TYPE_SOURCE(Z_REF_P(data), info);
} else {
/* "partially unserialized" value might violate the property
* declared type so we restore the default
*/
zval *tmp = &obj->ce->default_properties_table[OBJ_PROP_TO_NUM(info->offset)];
zval_ptr_dtor(data);

@ndossche ndossche Jun 10, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I assume evil destructors can cause issues here, so a temporary copy may be necessary (i.e. the zval garbage; trick)
I wonder though if "restoring after violation" is the right approach, but that may be a much bigger change

@devnexen devnexen Jun 10, 2026

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Went with var_push_dtor_value rather than the inline zval garbage, it moves the old value into the deferred-dtor list. So nothing is destroyed during the unwind or the delayed calls wdyt ?

@ndossche ndossche Jun 11, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That makes sense

ZVAL_COPY_OR_DUP(data, tmp);

@arnaud-lb arnaud-lb Jun 25, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nit: could use ZVAL_COPY_VALUE_PROP() like in _object_properties_init()

@devnexen devnexen Jun 25, 2026 •
edited
Loading

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think ZVAL_COPY_OR_DUP_PROP is then more appropriate due to ref count (esp for other cases than spl).

}

@arnaud-lb arnaud-lb Jun 25, 2026 •
edited
Loading

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In theory the same problem happens after zend_verify_prop_assignable_by_ref() bellow, but right now the call to SplMaxHeap->__unserialize() (and thus Exception->getTraceAsString()) is prevented due to EG(exception) being set. The prop is undef, but IS_PROP_UNINIT is not set. Maybe update that code path too?

Edit: the if (Z_ISREF_P(data)) { case above could also leave a prop assigned to a wrong type

@devnexen devnexen Jun 25, 2026

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For the if (Z_ISREF_P(data)) case, left it for now since it's gated by EG(exception) too, feels like follow up material, wdyt ?

@arnaud-lb arnaud-lb Jun 26, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, that's fine

}
goto failure;
}
Expand Down
Loading