[kamailio/kamailio] Segmentation fault when using http_hdr with http_async_client (#760)

List overview All Threads
Download

newer

older

git:master:3fec5152: topos: error...

git:master:48195546: core: tcp...

AlainDubois78

29 Aug 2016 29 Aug '16

12:39 p.m.

While using httpa_sync_client, I'm trying to retrieve the Content-Type of the body I got in a 200OK after an HTTP get using $var(my_content_type) =$ http_hdr(Content-Type); Then When I launch Kamailio, I get a seg fault at the configuration file validation step because in http_async_client_mod.c the line 'return pv_api.parse_hdr_name(sp, in)' is trying to use parse_hdr_name before it being initialized. I noticed that when I remove my previous line, the initialization is done but after the configuration file validation. Any idea if I'm doing something wrong?

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760

Attachments:

attachment.html (text/html — 2.2 KB)

Show replies by date

Camille Oudot

29 Aug 29 Aug

1:05 p.m.

Hi, is the pv module loaded before the http_async_client module? If not can you try?

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243118242

AlainDubois78

3:11 p.m.

Sorry, I forgot to say it but yes I verified that the pv module is loaded. Actually I added traces in the mod_init function of the http_async_client_module and when I remove my faulty line I can see the full init taking place including the pv_api being filled. When I put back my faulty line, the traces I added are not displayed the seg fault occuring before the init.

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243153452

Daniel-Constantin Mierla

3:19 p.m.

If you get a segfault, can you paste here the backtrace taken with gdb from corefile? It will reveal the execution stack and the point where it crashes.

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243155917

AlainDubois78

3:29 p.m.

Here you are: #0 0x0000000000000000 in ?? () #1 0x00007f1a90b8f3b3 in w_pv_parse_hdr_name (sp=0x7f1a9f82b128, in=0x7ffcf7e142b0) at http_async_client_mod.c:681 #2 0x00000000004d382f in pv_parse_spec2 (in=in@entry=0x7f1a9f82b110, e=e@entry=0x7f1a9f82b128, silent=silent@entry=0) at pvapi.c:873 #3 0x00000000004d5e61 in pv_cache_add (name=name@entry=0x7ffcf7e143b0) at pvapi.c:282 #4 0x00000000004d65cb in pv_cache_get (name=name@entry=0x9ffbb0 <s_tmp>) at pvapi.c:356 #5 0x0000000000668025 in yyparse () at cfg.y:2596 #6 0x000000000042362a in main (argc=6, argv=0x7ffcf7e14ef8) at main.c:2119

(gdb) up #1 0x00007f1a90b8f3b3 in w_pv_parse_hdr_name (sp=0x7f1a9f82b128, in=0x7ffcf7e142b0) at http_async_client_mod.c:681 681 return pv_api.parse_hdr_name(sp, in);

Note that I'm not using the latest version of http_async_client yet. I verifyed that sp and in are correct but actually pv_api is not: print pv_api $1 = {get_reason = 0x0, get_hdr = 0x0, parse_hdr_name = 0x0, get_status = 0x0, get_msg_body = 0x0, get_body_size = 0x0, get_msg_buf = 0x0, get_msg_len = 0x0}

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243158993

Daniel-Constantin Mierla

3:30 p.m.

Can you provide also the output from gdb for next commands:

``` frame 1 info locals list ```

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243159435

AlainDubois78

3:34 p.m.

Results: (gdb) frame 1 #1 0x00007f1a90b8f3b3 in w_pv_parse_hdr_name (sp=0x7f1a9f82b128, in=0x7ffcf7e142b0) at http_async_client_mod.c:681 681 return pv_api.parse_hdr_name(sp, in); (gdb) info locals No locals. (gdb) list 676 AH_WRAP_GET_PV(ah_get_body_size, get_body_size) 677 AH_WRAP_GET_PV(ah_get_msg_buf, get_msg_buf) 678 AH_WRAP_GET_PV(ah_get_msg_len, get_msg_len) 679 680 static int w_pv_parse_hdr_name(pv_spec_p sp, str *in) { 681 return pv_api.parse_hdr_name(sp, in); 682 } 683 684 static int ah_get_ok(struct sip_msg *msg, pv_param_t *param, pv_value_t *res) { 685 if (ah_reply) {

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243160387

Camille Oudot

30 Aug 30 Aug

9:44 a.m.

Closed #760 via 597242dfbfd883d136aa49d88266fc9edb6b8e09.

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#event-771891396

Camille Oudot

9:44 a.m.

Reopened #760.

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#event-771891635

Camille Oudot

9:44 a.m.

Hi, we could reproduce the issue. When using instructions as

xlog("$http_hdr(Content-Type)");

the PV's internal name is parsed at runtime, and the PV API has been initialized already. When using

$var(ct) = $http_hdr(Content-Type)

it's parsed at start time, before the API was initialized. I sent a patch that does the init sooner, and should fix the issue

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243388344

Camille Oudot

31 Aug 31 Aug

10:03 a.m.

Unless there is still an issue despite the fix, I'm going to close this issue. @miconda, I'm planning to backport it to 4.4 branch too, but this patch will now make kamailio fail to start if the **http_async_client** module is loaded before **pv**. This is the way it always have been documented, but until now it worked no matter what the loading order was. Is it ok to backport it anyway?

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243718303

Daniel-Constantin Mierla

10:06 a.m.

For me it's ok to backport.

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243718848

Camille Oudot

10:13 a.m.

Closed #760.

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#event-773430119

AlainDubois78

1:52 p.m.

Camille, your fix works for me as well! Many thanks

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/kamailio/kamailio/issues/760#issuecomment-243770796

3139

Age (days ago)

3141

Last active (days ago)

sr-dev@lists.kamailio.org

13 comments

3 participants

tags (0)

participants (3)

AlainDubois78
Camille Oudot
Daniel-Constantin Mierla