用libcurl一段時間遇到莫名其妙的程序崩潰的情況,開始覺得是線程棧溢出導致的段錯誤,專門增加了線程棧的大小,還是有問題。線程也是分離的。用valgrind定位到問題可能出現在curl的調用上。
排查的時候也發現了libcurl一些額外的坑,現做個總結筆記。
1.、
線程使用libcurl訪問時,設置了超時時間,而libcurl庫不會爲這個超時信號做任何處理,信號產生而沒有信號句柄處理,可能導致程序退出。用以下選項禁止訪問超時的時候拋出超時信號。
curl_setopt(curl, CURLOPT_NOSIGNAL,1L);
2、
正常使用流程是先調用curl_global_init初始化資源,而這個函數不是線程安全的。curl_easy_init發現沒有做初始化時會自動初始化,多個線程同時進行curl_easy_init時會導致異常。curl官方建議在主線程中先做初始化。
curl_global_init(CURL_GLOBAL_ALL);
3、
用valgrind調試,做壓力測試多線程運行時,valgrind彈出類似訪問衝突的錯誤,把錯誤定位到libcurl的互斥訪問的處理上。
大致是進行訪問設計到ssl的東西的時候,需要做相應的加鎖處理,libcurl源碼沒有加這塊處理,不過給了個針對這種情況加鎖的例子,總的來說curl的示例代碼很重要很有用。
https://curl.haxx.se/libcurl/c/example.html
- #define USE_OPENSSL
- #include <stdio.h>
- #include <pthread.h>
- #include <curl/curl.h>
- #define NUMT 4
- /* we have this global to let the callback get easy access to it */
- static pthread_mutex_t *lockarray;
- #ifdef USE_OPENSSL
- #include <openssl/crypto.h>
- static void lock_callback(int mode, int type, char *file, int line)
- {
- (void)file;
- (void)line;
- if(mode & CRYPTO_LOCK) {
- pthread_mutex_lock(&(lockarray[type]));
- }
- else {
- pthread_mutex_unlock(&(lockarray[type]));
- }
- }
- static unsigned long thread_id(void)
- {
- unsigned long ret;
- ret=(unsigned long)pthread_self();
- return ret;
- }
- static void init_locks(void)
- {
- int i;
- lockarray=(pthread_mutex_t *)OPENSSL_malloc(CRYPTO_num_locks() *
- sizeof(pthread_mutex_t));
- for(i=0; i<CRYPTO_num_locks(); i++) {
- pthread_mutex_init(&(lockarray[i]), NULL);
- }
- CRYPTO_set_id_callback((unsigned long (*)())thread_id);
- CRYPTO_set_locking_callback((void (*)())lock_callback);
- }
- static void kill_locks(void)
- {
- int i;
- CRYPTO_set_locking_callback(NULL);
- for(i=0; i<CRYPTO_num_locks(); i++)
- pthread_mutex_destroy(&(lockarray[i]));
- OPENSSL_free(lockarray);
- }
- #endif
- #ifdef USE_GNUTLS
- #include <gcrypt.h>
- #include <errno.h>
- GCRY_THREAD_OPTION_PTHREAD_IMPL;
- void init_locks(void)
- {
- gcry_control(GCRYCTL_SET_THREAD_CBS);
- }
- #define kill_locks()
- #endif
- /* List of URLs to fetch.*/
- const char * const urls[]= {
- ”https://www.example.com/”,
- ”https://www2.example.com/”,
- ”https://www3.example.com/”,
- ”https://www4.example.com/”,
- };
- static void *pull_one_url(void *url)
- {
- CURL *curl;
- curl = curl_easy_init();
- curl_easy_setopt(curl, CURLOPT_URL, url);
- /* this example doesn’t verify the server’s certificate, which means we
- might be downloading stuff from an impostor */
- curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
- curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
- curl_easy_perform(curl); /* ignores error */
- curl_easy_cleanup(curl);
- return NULL;
- }
- int main(int argc, char **argv)
- {
- pthread_t tid[NUMT];
- int i;
- int error;
- (void)argc; /* we don’t use any arguments in this example */
- (void)argv;
- /* Must initialize libcurl before any threads are started */
- curl_global_init(CURL_GLOBAL_ALL);
- init_locks();
- for(i=0; i< NUMT; i++) {
- error = pthread_create(&tid[i],
- NULL, /* default attributes please */
- pull_one_url,
- (void *)urls[i]);
- if(0 != error)
- fprintf(stderr, ”Couldn’t run thread number %d, errno %d\n”, i, error);
- else
- fprintf(stderr, ”Thread %d, gets %s\n”, i, urls[i]);
- }
- /* now wait for all threads to terminate */
- for(i=0; i< NUMT; i++) {
- error = pthread_join(tid[i], NULL);
- fprintf(stderr, ”Thread %d terminated\n”, i);
- }
- kill_locks();
- return 0;
- }
#define USE_OPENSSL
#include <stdio.h>
#include <pthread.h>
#include <curl/curl.h>
#define NUMT 4
/* we have this global to let the callback get easy access to it */
static pthread_mutex_t *lockarray;
#ifdef USE_OPENSSL
#include <openssl/crypto.h>
static void lock_callback(int mode, int type, char *file, int line)
{
(void)file;
(void)line;
if(mode & CRYPTO_LOCK) {
pthread_mutex_lock(&(lockarray[type]));
}
else {
pthread_mutex_unlock(&(lockarray[type]));
}
}
static unsigned long thread_id(void)
{
unsigned long ret;
ret=(unsigned long)pthread_self();
return ret;
}
static void init_locks(void)
{
int i;
lockarray=(pthread_mutex_t *)OPENSSL_malloc(CRYPTO_num_locks() *
sizeof(pthread_mutex_t));
for(i=0; i<CRYPTO_num_locks(); i++) {
pthread_mutex_init(&(lockarray[i]), NULL);
}
CRYPTO_set_id_callback((unsigned long (*)())thread_id);
CRYPTO_set_locking_callback((void (*)())lock_callback);
}
static void kill_locks(void)
{
int i;
CRYPTO_set_locking_callback(NULL);
for(i=0; i<CRYPTO_num_locks(); i++)
pthread_mutex_destroy(&(lockarray[i]));
OPENSSL_free(lockarray);
}
#endif
#ifdef USE_GNUTLS
#include <gcrypt.h>
#include <errno.h>
GCRY_THREAD_OPTION_PTHREAD_IMPL;
void init_locks(void)
{
gcry_control(GCRYCTL_SET_THREAD_CBS);
}
#define kill_locks()
#endif
/* List of URLs to fetch.*/
const char * const urls[]= {
"https://www.example.com/",
"https://www2.example.com/",
"https://www3.example.com/",
"https://www4.example.com/",
};
static void *pull_one_url(void *url)
{
CURL *curl;
curl = curl_easy_init();
curl_easy_setopt(curl, CURLOPT_URL, url);
/* this example doesn't verify the server's certificate, which means we
might be downloading stuff from an impostor */
curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
curl_easy_perform(curl); /* ignores error */
curl_easy_cleanup(curl);
return NULL;
}
int main(int argc, char **argv)
{
pthread_t tid[NUMT];
int i;
int error;
(void)argc; /* we don't use any arguments in this example */
(void)argv;
/* Must initialize libcurl before any threads are started */
curl_global_init(CURL_GLOBAL_ALL);
init_locks();
for(i=0; i< NUMT; i++) {
error = pthread_create(&tid[i],
NULL, /* default attributes please */
pull_one_url,
(void *)urls[i]);
if(0 != error)
fprintf(stderr, "Couldn't run thread number %d, errno %d\n", i, error);
else
fprintf(stderr, "Thread %d, gets %s\n", i, urls[i]);
}
/* now wait for all threads to terminate */
for(i=0; i< NUMT; i++) {
error = pthread_join(tid[i], NULL);
fprintf(stderr, "Thread %d terminated\n", i);
}
kill_locks();
return 0;
}
未完待續。。。