Edit report at https://bugs.php.net/bug.php?id=70973&edit=1
ID: 70973
Comment by: cv at jet9 dot ru
Reported by: cv at jet9 dot ru
Summary: Segmentation fault when no huge pages available
Status: Feedback
Type: Bug
Package: opcache
Operating System: CentOS 6
PHP Version: 7.0.0RC7
Assigned To: laruence
Block user comment: N
Private report: N
New Comment:
No, the bug still exists. With the patch applied the function accel_move_code_to_huge_pages()
returns and segfault raises later:
(gdb) br /usr/src/debug/php-7.0.0RC8/ext/opcache/ZendAccelerator.c:2620
(gdb) run
Breakpoint 1, accel_startup (extension=<value optimized out>) at
/usr/src/debug/php-7.0.0RC8/ext/opcache/ZendAccelerator.c:2620
2620 if (!ZCG(accel_directives).enable_cli &&
(gdb) n
2619 accel_startup_ok = 0;
(gdb)
2620 if (!ZCG(accel_directives).enable_cli &&
(gdb)
2621 strcmp(sapi_module.name, "cli") == 0) {
(gdb)
2620 if (!ZCG(accel_directives).enable_cli &&
(gdb)
2622 zps_startup_failure("Opcode Caching is disabled for CLI",
NULL, accelerator_remove_cb);
(gdb)
Program received signal SIGSEGV, Segmentation fault.
zend_llist_del_element (l=0x9dc8e0, element=0x0, compare=0x7fffedf40590
<accelerator_remove_cb>) at /usr/src/debug/php-7.0.0RC8/Zend/zend_llist.c:91
91 {
Previous Comments:
------------------------------------------------------------------------
[2015-11-26 04:10:56] laruence@php.net
I can not reproduce this simply, so could you please test with this patch?
diff --git a/ext/opcache/ZendAccelerator.c b/ext/opcache/ZendAccelerator.c
index e23e751..ae10d12 100644
--- a/ext/opcache/ZendAccelerator.c
+++ b/ext/opcache/ZendAccelerator.c
@@ -2531,7 +2531,7 @@ static int accel_remap_huge_pages(void *start, size_t size, const char *name,
si
PROT_READ | PROT_WRITE | PROT_EXEC,
MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED,
-1, 0);
- if (-1 == madvise(start, size, MADV_HUGEPAGE)) {
+ if (ret == MAP_FAILED || -1 == madvise(start, size, MADV_HUGEPAGE)) {
munmap(mem, size);
return -1;
}
------------------------------------------------------------------------
[2015-11-25 21:31:51] cv at jet9 dot ru
Description:
------------
When opcache.huge_code_pages=1 and no huge pages available, php gots segfault:
jet9.ru$ strace /own/bin/php
...
mmap(0x400000, 2097152, PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS|0x40000,
-1, 0) = -1 ENOMEM (Cannot allocate memory)
munmap(0x7fa18485f000, 2097152) = 0
close(3) = 0
munmap(0x7fa1a0800000, 4096) = 0
--- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0x5d72c0} ---
+++ killed by SIGSEGV +++
Segmentation fault
jet9.ru$ gdb /own/bin/php
...
Program received signal SIGSEGV, Segmentation fault.
zend_llist_del_element (l=0x9dc8e0, element=0x0, compare=0x7fffedf40590
<accelerator_remove_cb>) at /usr/src/debug/php-7.0.0RC8/Zend/zend_llist.c:91
I guess somewhere's lost mmap() returning value check when dealing with huge pages.
------------------------------------------------------------------------
--
Edit this bug report at https://bugs.php.net/bug.php?id=70973&edit=1