求问一个ngx_http_upstream_connect函数的core

liuxinglong88 · 2014-07-07T12:56:19+00:00

Hello! 2014-07-02 1:34 GMT-07:00 刘泽航: > 关于nginx_lua的主要问题就是我&#x...

求问一个ngx_http_upstream_connect函数的core

liuxinglong88

hi all

遇到一个ngx_http_upstream_connect函数的core文件

core栈如下

(gdb) bt
#0  0x00007f5d257e7715 in ?? ()
#1  0x0000000000429e7f in ngx_vslprintf (buf=0x7fffac831db7 "25.71.17, server: , request: \"GET /502774.html HTTP/1.1\", upstream: \"http://", 
    last=0x7fffac832500 "", fmt=0x90 <Address 0x90 out of bounds>, args=0x7fffac831bb8) at src/core/ngx_string.c:419
#2  0x000000000042a205 in ngx_snprintf (buf=0x7fffac831db7 "25.71.17, server: , request: \"GET /502774.html HTTP/1.1\", upstream: \"http://", max=16401, 
    fmt=0x90 <Address 0x90 out of bounds>) at src/core/ngx_string.c:118
#3  0x000000000045778c in ngx_http_log_error_handler (r=0x55fa5b0, sr=0x55fa5b0, 
    buf=0x7fffac831da3 ", upstream: \"http://25.71.17, server: , request: \"GET /502774.html HTTP/1.1\", upstream: \"http://", len=1885)
    at src/http/ngx_http_request.c:3532
#4  0x00000000004262d8 in ngx_log_error_core (level=4, log=0x6021020, err=0, fmt=0xabc6c0 "(): %d") at src/core/ngx_log.c:119
#5  0x000000000046f573 in ngx_http_upstream_reinit (u=0x2cc5bc8, r=0x55fa5b0) at src/http/ngx_http_upstream.c:1414
#6  ngx_http_upstream_connect (r=0x55fa5b0, u=0x2cc5bc8) at src/http/ngx_http_upstream.c:1242
#7  0x000000000046ddf0 in ngx_http_upstream_process_header (r=0x55fa5b0, u=0x2cc5bc8) at src/http/ngx_http_upstream.c:1566
#8  0x000000000046b56d in ngx_http_upstream_handler (ev=0x7fffac831db7) at src/http/ngx_http_upstream.c:966
#9  0x0000000000000000 in ?? ()

看到是在ngx_http_upstream_connect调用 rc = ngx_event_connect_peer(&u->peer);后没有取到peer出现了问题，

(gdb) f 6
#6  ngx_http_upstream_connect (r=0x55fa5b0, u=0x2cc5bc8) at src/http/ngx_http_upstream.c:1242
1242    in src/http/ngx_http_upstream.c
(gdb) p u->peer.name
$31 = (ngx_str_t *) 0x1888090
(gdb) p *u->peer.name
$32 = {len = 144, data = "" <Address 0x4011 out of bounds>}
(gdb) info locals
rc = -3
c = 0x7f5c8e0a0b90

此时看到rc明明是等于-3(ngx_busy),那么应该是会进入以下代码块才对啊。可是看栈信息确是进入了reinit过程？

    if (rc == NGX_BUSY) {
        ngx_log_error(NGX_LOG_ERR, r->connection->log, 0, "no live upstreams");
        ngx_http_upstream_next(r, u, NGX_HTTP_UPSTREAM_FT_NOLIVE);
        return;
    }

另外还有个疑问，为什么打印ngx_cycle一直为0x0呢？

(gdb) p ngx_cycle
$33 = (volatile ngx_cycle_t *) 0x0

agentzh

Hello!

On Sun, Jul 6, 2014 at 9:56 PM, Leo Liu wrote:
>      遇到一个ngx_http_upstream_connect函数的core文件
>
[...]
> #5  0x000000000046f573 in ngx_http_upstream_reinit (u=0x2cc5bc8,
> r=0x55fa5b0) at src/http/ngx_http_upstream.c:1414
> #6  ngx_http_upstream_connect (r=0x55fa5b0, u=0x2cc5bc8) at
> src/http/ngx_http_upstream.c:1242
> #7  0x000000000046ddf0 in ngx_http_upstream_process_header (r=0x55fa5b0,
> u=0x2cc5bc8) at src/http/ngx_http_upstream.c:1566

你使用的是何种版本的 nginx? 另外，你是否自己给它打过补丁？

我检查了好几个版本的标准 nginx 核心，其对应 C 源文件的行号没法和你的 backtrace 里的行号对应上。

Regards,
-agentzh

liuxinglong88

nginx版本是nginx/1.4.4

我本地看文件对应的行号也是错误的，甚至于core文件栈

#5 0x000000000046f573 in ngx_http_upstream_reinit (u=0x2cc5bc8, r=0x55fa5b0) at src/http/ngx_http_upstream.c:1414

函数对应的参数列表顺序也是反的？什么情况会造成这样的问题呢?

在 2014年7月7日星期一UTC+8下午1时06分01秒，agentzh写道：

Hello!

On Sun, Jul 6, 2014 at 9:56 PM, Leo Liu wrote:
> 遇到一个ngx_http_upstream_connect函数的core文件
>
[...]
> #5 0x000000000046f573 in ngx_http_upstream_reinit (u=0x2cc5bc8,
> r=0x55fa5b0) at src/http/ngx_http_upstream.c:1414
> #6 ngx_http_upstream_connect (r=0x55fa5b0, u=0x2cc5bc8) at
> src/http/ngx_http_upstream.c:1242
> #7 0x000000000046ddf0 in ngx_http_upstream_process_header (r=0x55fa5b0,
> u=0x2cc5bc8) at src/http/ngx_http_upstream.c:1566

你使用的是何种版本的 nginx? 另外，你是否自己给它打过补丁？

我检查了好几个版本的标准 nginx 核心，其对应 C 源文件的行号没法和你的 backtrace 里的行号对应上。

Regards,
-agentzh