释放链表时发生堆栈溢出：成功释放 100k 个节点后出现分段错误
Stack Overflow when freeing a linked list: Segmentation fault after 100k nodes successfully freed
我创建了一个链表，大约有250k 节点使用递归函数在接收到节点时添加节点。我确实验证了此列表已成功链接。
释放链表时，我使用的是从根节点开始的递归函数，并沿着链表连续释放节点迭代移动游标。
大约。成功释放 100k 个节点我收到以下段错误：
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7a8f4d9 in new_do_write (to_do=39,
data=0x7ffff7ff5000 "cursor start loop : 0x19ffca0\n340\n", fp=0x7ffff7dd4400 <_IO_2_1_stdout_>)
at fileops.c:538
538 fileops.c: No such file or directory.
(gdb)
我多次成功的事实让我假设这不太可能是我代码的 逻辑 中的错误，而是必须与内存有关分配和解除分配，更多的全局问题。
为了清楚起见，我打印了指针的分配等：
.....
cursor start loop : 0x19ffac0
currentNode start loop : (nil)
currentNode (set to cursor): 0x19ffac0
cursor children[0] point to: 0x19ffbb0
Move cursor vorward, cursor: 0x19ffbb0
free old last (currentNode): 0x19ffac0
Iterate, pass cursor : 0x19ffbb0
freecount: 87304
cursor start loop : 0x19ffbb0
currentNode start loop : (nil)
currentNode (set to cursor): 0x19ffbb0
cursor children[0] point to: 0x19ffca0
Move cursor vorward, cursor: 0x19ffca0
free old last (currentNode): 0x19ffbb0
Iterate, pass cursor : 0x19ffca0
freecount: 87305
Program received signal SIGSEGV, Segmentation fault.
Valgrind 显示堆栈溢出。我该如何解决这个问题？
==14352== Stack overflow in thread 1: can't grow stack to 0xffe801ff8
==14352==
==14352== Process terminating with default action of signal 11 (SIGSEGV)
==14352== Access not within mapped region at address 0xFFE801FF8
==14352== at 0x4EAFFC5: _IO_file_write@@GLIBC_2.2.5 (fileops.c:1254)
==14352== If you believe this happened as a result of a stack
==14352== overflow in your program's main thread (unlikely but
==14352== possible), you can try to increase the size of the
==14352== main thread stack using the --main-stacksize= flag.
==14352== The main thread stack size used in this run was 8388608.
==14352== Stack overflow in thread 1: can't grow stack to 0xffe801ff0
==14352==
==14352== Process terminating with default action of signal 11 (SIGSEGV)
==14352== Access not within mapped region at address 0xFFE801FF0
==14352== at 0x4A256B0: _vgnU_freeres (in /usr/lib/valgrind/vgpreload_core-amd64-linux.so)
==14352== If you believe this happened as a result of a stack
==14352== overflow in your program's main thread (unlikely but
==14352== possible), you can try to increase the size of the
==14352== main thread stack using the --main-stacksize= flag.
==14352== The main thread stack size used in this run was 8388608.
==14352==
==14352== HEAP SUMMARY:
==14352== in use at exit: 62,486,368 bytes in 278,957 blocks
==14352== total heap usage: 366,221 allocs, 87,264 frees, 82,034,192 bytes allocated
==14352==
==14352== LEAK SUMMARY:
==14352== definitely lost: 0 bytes in 0 blocks
==14352== indirectly lost: 0 bytes in 0 blocks
==14352== possibly lost: 0 bytes in 0 blocks
==14352== still reachable: 62,486,368 bytes in 278,957 blocks
==14352== suppressed: 0 bytes in 0 blocks
==14352== Rerun with --leak-check=full to see details of leaked memory
==14352==
==14352== For counts of detected and suppressed errors, rerun with: -v
==14352== Use --track-origins=yes to see where uninitialised values come from
==14352== ERROR SUMMARY: 4 errors from 2 contexts (suppressed: 0 from 0)
Segmentation fault
这是我的代码
头文件
// define global pointer variable root
extern struct node* rootNS;
源文件
struct node* rootNS;
// create root node in a function (not main)
// at first function call send root to linked list that stores all created nodes
if (x == 0)
{
// allocate storage
rootNS = malloc(sizeof(struct node));
rootNS->children[0] = NULL;
x = 1;
nodestore(rootNS);
}
// function creating the linked list, nodes are past to the function
bool nodestore(struct node* nodeAD)
{
// for debugging: Create nodecounter
static int nodecounter = 0;
nodecounter +=1;
// create cursor at first function call
static struct node* cursor;
if (nodeAD == rootNS)
{
cursor = nodeAD;
if (nodecounter == 1)
{
cursor->children[0] = NULL;
}
return true;
}
// append list
// point last node (cursor) to new node
cursor->children[0] = nodeAD;
// set cursor to new node = last node
cursor = nodeAD;
return true;
}
// function freeing the linked list
int unloader(struct node* cursor)
{
// for debugging: Creat nodecounter
static int nodecounter = -1;
nodecounter +=1;
// create helper pointer
struct node* currentNode = NULL;
// check if end of linked list has been reached
if(cursor->children[0] == NULL)
{
free(cursor);
free(root);
return true;
}
else
{
// keep track of old last node before moving cursor
currentNode = cursor;
// move cursor forward
cursor = cursor->children[0];
// free old last node
free(currentNode);
// iterate passing new last node (cursor)
return unloader(cursor);
}
}
你的 unloader 函数是递归的，所以每次调用 unloader 都会产生一个额外的堆栈帧......在 100k 次函数调用时，你简单地，嗯，随着堆栈的增长而溢出堆栈大。我不知道你有什么 compiler/system，但是 ~100k 堆栈帧将为 2MB 堆栈提供~20 字节的帧大小（IIRC 是 Windows 上的默认值），这听起来不错查看您的代码。
解决此问题的唯一合理方法是将 unloader 更改为迭代函数（不再递归）。有多种方法可以增加默认堆栈大小，但具体方法取决于您的 compiler/system（在任何情况下通常都不推荐）。
我创建了一个链表，大约有250k 节点使用递归函数在接收到节点时添加节点。我确实验证了此列表已成功链接。
释放链表时，我使用的是从根节点开始的递归函数，并沿着链表连续释放节点迭代移动游标。
大约。成功释放 100k 个节点我收到以下段错误：
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7a8f4d9 in new_do_write (to_do=39,
data=0x7ffff7ff5000 "cursor start loop : 0x19ffca0\n340\n", fp=0x7ffff7dd4400 <_IO_2_1_stdout_>)
at fileops.c:538
538 fileops.c: No such file or directory.
(gdb)
我多次成功的事实让我假设这不太可能是我代码的 逻辑 中的错误，而是必须与内存有关分配和解除分配，更多的全局问题。
为了清楚起见，我打印了指针的分配等：
.....
cursor start loop : 0x19ffac0
currentNode start loop : (nil)
currentNode (set to cursor): 0x19ffac0
cursor children[0] point to: 0x19ffbb0
Move cursor vorward, cursor: 0x19ffbb0
free old last (currentNode): 0x19ffac0
Iterate, pass cursor : 0x19ffbb0
freecount: 87304
cursor start loop : 0x19ffbb0
currentNode start loop : (nil)
currentNode (set to cursor): 0x19ffbb0
cursor children[0] point to: 0x19ffca0
Move cursor vorward, cursor: 0x19ffca0
free old last (currentNode): 0x19ffbb0
Iterate, pass cursor : 0x19ffca0
freecount: 87305
Program received signal SIGSEGV, Segmentation fault.
Valgrind 显示堆栈溢出。我该如何解决这个问题？
==14352== Stack overflow in thread 1: can't grow stack to 0xffe801ff8
==14352==
==14352== Process terminating with default action of signal 11 (SIGSEGV)
==14352== Access not within mapped region at address 0xFFE801FF8
==14352== at 0x4EAFFC5: _IO_file_write@@GLIBC_2.2.5 (fileops.c:1254)
==14352== If you believe this happened as a result of a stack
==14352== overflow in your program's main thread (unlikely but
==14352== possible), you can try to increase the size of the
==14352== main thread stack using the --main-stacksize= flag.
==14352== The main thread stack size used in this run was 8388608.
==14352== Stack overflow in thread 1: can't grow stack to 0xffe801ff0
==14352==
==14352== Process terminating with default action of signal 11 (SIGSEGV)
==14352== Access not within mapped region at address 0xFFE801FF0
==14352== at 0x4A256B0: _vgnU_freeres (in /usr/lib/valgrind/vgpreload_core-amd64-linux.so)
==14352== If you believe this happened as a result of a stack
==14352== overflow in your program's main thread (unlikely but
==14352== possible), you can try to increase the size of the
==14352== main thread stack using the --main-stacksize= flag.
==14352== The main thread stack size used in this run was 8388608.
==14352==
==14352== HEAP SUMMARY:
==14352== in use at exit: 62,486,368 bytes in 278,957 blocks
==14352== total heap usage: 366,221 allocs, 87,264 frees, 82,034,192 bytes allocated
==14352==
==14352== LEAK SUMMARY:
==14352== definitely lost: 0 bytes in 0 blocks
==14352== indirectly lost: 0 bytes in 0 blocks
==14352== possibly lost: 0 bytes in 0 blocks
==14352== still reachable: 62,486,368 bytes in 278,957 blocks
==14352== suppressed: 0 bytes in 0 blocks
==14352== Rerun with --leak-check=full to see details of leaked memory
==14352==
==14352== For counts of detected and suppressed errors, rerun with: -v
==14352== Use --track-origins=yes to see where uninitialised values come from
==14352== ERROR SUMMARY: 4 errors from 2 contexts (suppressed: 0 from 0)
Segmentation fault
这是我的代码
头文件
// define global pointer variable root
extern struct node* rootNS;
源文件
struct node* rootNS;
// create root node in a function (not main)
// at first function call send root to linked list that stores all created nodes
if (x == 0)
{
// allocate storage
rootNS = malloc(sizeof(struct node));
rootNS->children[0] = NULL;
x = 1;
nodestore(rootNS);
}
// function creating the linked list, nodes are past to the function
bool nodestore(struct node* nodeAD)
{
// for debugging: Create nodecounter
static int nodecounter = 0;
nodecounter +=1;
// create cursor at first function call
static struct node* cursor;
if (nodeAD == rootNS)
{
cursor = nodeAD;
if (nodecounter == 1)
{
cursor->children[0] = NULL;
}
return true;
}
// append list
// point last node (cursor) to new node
cursor->children[0] = nodeAD;
// set cursor to new node = last node
cursor = nodeAD;
return true;
}
// function freeing the linked list
int unloader(struct node* cursor)
{
// for debugging: Creat nodecounter
static int nodecounter = -1;
nodecounter +=1;
// create helper pointer
struct node* currentNode = NULL;
// check if end of linked list has been reached
if(cursor->children[0] == NULL)
{
free(cursor);
free(root);
return true;
}
else
{
// keep track of old last node before moving cursor
currentNode = cursor;
// move cursor forward
cursor = cursor->children[0];
// free old last node
free(currentNode);
// iterate passing new last node (cursor)
return unloader(cursor);
}
}
你的 unloader 函数是递归的，所以每次调用 unloader 都会产生一个额外的堆栈帧......在 100k 次函数调用时，你简单地，嗯，随着堆栈的增长而溢出堆栈大。我不知道你有什么 compiler/system，但是 ~100k 堆栈帧将为 2MB 堆栈提供~20 字节的帧大小（IIRC 是 Windows 上的默认值），这听起来不错查看您的代码。
解决此问题的唯一合理方法是将 unloader 更改为迭代函数（不再递归）。有多种方法可以增加默认堆栈大小，但具体方法取决于您的 compiler/system（在任何情况下通常都不推荐）。