Skip to content

Fix block_pass JMP[N]Z optimization #20850

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

Open
iluuu1994 wants to merge 1 commit into
base: PHP-8.4
Choose a base branch
from
+15 −1
Open

Fix block_pass JMP[N]Z optimization #20850

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion Zend/Optimizer/block_pass.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1152,7 +1152,7 @@ static zend_always_inline zend_basic_block *get_next_block(const zend_cfg *cfg,
}
next_block++;
}
while (next_block->len == 0 && !(next_block->flags & ZEND_BB_PROTECTED)) {
while (next_block->len == 0 && !(next_block->flags & (ZEND_BB_TARGET|ZEND_BB_PROTECTED))) {
Copy link
Member

@arnaud-lb arnaud-lb Jan 7, 2026 •
edited
Loading

Choose a reason for hiding this comment

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

I feel that stopping at ZEND_BB_TARGET will often disable this optimization, as next_block will be ZEND_BB_TARGET most of the time (since it's the next block after an unconditional JMP, it must be a jump target), and this causes target_block == next_block to be false when there where blocks to skip here.

Maybe only in master, but would it make sense to relax the rule JMPZ(X,L1) JMP(L2) L1: -> JMPNZ(X,L2) NOP to JMPZ(X,L1) JMP(L2) L3: -> JMPNZ(X,L2) NOP with L3 being either L1 itself or part of a chain of empty predecessors to L1?

Copy link
Member Author

@iluuu1994 iluuu1994 Jan 7, 2026 •
edited
Loading

Choose a reason for hiding this comment

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

FWIW, it seems to affect only 5 of our tests. Another solution might be to immediately remove references to blocks when they become empty, and relink them to the following block, and remove ZEND_BB_TARGET from the empty block that will then be properly dead.

next_block = cfg->blocks + next_block->successors[0];
}
return next_block;
Expand Down
14 changes: 14 additions & 0 deletions ext/opcache/tests/oss-fuzz-472563272.phpt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,14 @@
--TEST--
OSS-Fuzz #472563272
--EXTENSIONS--
opcache
--INI--
opcache.enable=1
opcache.enable_cli=1
--FILE--
<?php
false || (true ? true : false) || (false ? true : false) || true;
?>
===DONE===
--EXPECT--
===DONE===