Re: [Bluez-users] pairing on x86_64

[Volker Fritzsch <kokor.hekkus@gmx.de>]

[-- Attachment #1: Type: text/plain, Size: 2184 bytes --]

Hi Marcel.

> what is the sizeof(unsigned char) and sizeof(char) on x86_64?
printf says 1 (for both).

> > Feb 21 16:13:57 dionysos hcid[5088]: ** io_security_event
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[0]: 4
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[1]: 22
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[2]: 6
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[3]: 121
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[4]: -109
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[5]: 50
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[6]: -32
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[7]: 7
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf[8]: 0
> > Feb 21 16:13:57 dionysos hcid[5088]: glib-ectomy * buf points to:
> > -1073745232 Feb 21 16:13:57 dionysos hcid[5088]: security * buf[0] = 0
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[1] = 0
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[2] = 0
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[3] = 0
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[4] = -109
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[5] = 50
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[6] = -32
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[7] = 7
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf[8] = 0
> > Feb 21 16:13:57 dionysos hcid[5088]: security * buf points to -1073745232
> > Feb 21 16:13:57 dionysos hcid[5088]: ** type not HCI_EVENT_PKT
>
> I don't know why this happens.
I couldn't understand too, but with your sizeof(char) question you made me 
curious.

On x86_64, int has a size of 4 and size_t takes 8 (while on x86 both have a 
size of 4).
By passing &len to g_io_channel_read (which takes a size_t *buffer_read as 
argument), we blank out the first four bytes of buf - which sits directly 
after len in memory ('oh wonderful world...').

> This can be a problem that the call of the PIN helper program fails.
This problem is also solved with the change from int len to size_t len.

So, please trash my first patch and take this instead.

Regards
 Volker

[-- Attachment #2: patch --]
[-- Type: text/x-diff, Size: 338 bytes --]

--- ../orig/bluez-utils-2.4/hcid/security.c	2003-07-24 19:34:34.000000000 +0200
+++ hcid/security.c	2004-02-21 20:47:35.000000000 +0100
@@ -356,7 +356,8 @@
 {
 	char buf[HCI_MAX_EVENT_SIZE], *ptr = buf;
 	struct hci_dev_info *di = (void *) data;
-	int len, type, dev;
+	int type, dev;
+	size_t len;
 	hci_event_hdr *eh;
 	GIOError err;