Skip to contentSkip to conversation

Small Patch for scotest.c

DISCUSSION Brian Smith 2 participants 3 messages Jun 7, 2012 – Jun 10, 2012 Bluetooth drivers

lore
Share & tools
lore

Brian Smith

author
#0 lore raw

Hi there.
Here is a tiny 3-line patch to scotest.c to get it working. (I've been trying to get bluez working on a Raspberry Pi, hitting USB problems, having a workng scotest is helpful). The problem is that bdaddr never gets initialized and tends to have random data in it, preventing the client/server end from matching up.

344a345,346

Quoted 2 lines
      hci_devba(0, &bdaddr);
Signature · 2 lines
Brian Smith.
"The bells of clocktowers stitch the sleeper's dreams together." - Memory Palace
#1 lore raw

Hi Brian,

Quoted 10 lines · Brian Smith

Hi there.
Here is a tiny 3-line patch to scotest.c to get it working. (I've been trying to get bluez working on a Raspberry Pi, hitting USB problems, having a workng scotest is helpful). The problem is that bdaddr never gets initialized and tends to have random data in it, preventing the client/server end from matching up.

344a345,346

Quoted 2 lines
      hci_devba(0, &bdaddr);

Please only send patches in unified diff format. If you want them applied upstream they should be created using "git format-patch" and preferably sent with "git send-email".

However, in this case I don't think your patch is quite right even if it was in the right format. Looking at bdaddr in scotest.c it's a static variable and to my understanding those should (according to the C standard) always get implicitly initialized to 0 if the code itself doesn't do so. At least gcc should follow this, so which compiler are you really using?

Furthermore, assuming that this (zero-initialized bdaddr) is how scotest behaves for most people (as it's existed many years and you're the first one to face the issue), it's the same as using BDADDR_ANY. Therefore, I'd just go ahead and remove the bdaddr variable and replace the places where it was used with BDADDR_ANY (and please format the patch like I described above). Thanks.

Johan

author
#2 lore raw
Quoted 28 lines · Johan Hedberg

Hi Brian,

Quoted 7 lines · Brian Smith

Hi there.
Here is a tiny 3-line patch to scotest.c to get it working. (I've been trying to get bluez working on a Raspberry Pi, hitting USB problems, having a workng scotest is helpful). The problem is that bdaddr never gets initialized and tends to have random data in it, preventing the client/server end from matching up.

Quoted 5 lines

344a345,346

Quoted 3 lines
      hci_devba(0, &bdaddr);

Please only send patches in unified diff format. If you want them applied upstream they should be created using "git format-patch" and preferably sent with "git send-email".

However, in this case I don't think your patch is quite right even if it was in the right format. Looking at bdaddr in scotest.c it's a static variable and to my understanding those should (according to the C standard) always get implicitly initialized to 0 if the code itself doesn't do so. At least gcc should follow this, so which compiler are you really using?

Quoted 10 lines

Furthermore, assuming that this (zero-initialized bdaddr) is how scotest behaves for most people (as it's existed many years and you're the first one to face the issue), it's the same as using BDADDR_ANY. Therefore, I'd just go ahead and remove the bdaddr variable and replace the places where it was used with BDADDR_ANY (and please format the patch like I described above). Thanks.

Johan

Hi Johan.
Thanks for taking the time to look at my email and the feedback. Yes, you are right about the static variable initialization, bdaddr starts as 0:0:0:0:0:0, but my understanding of what hci_devba() does was wrong. It changes baddrr from 0 to the MAC address of hci0, the local bluetooth adapter, which does actually seem to make scotest work properly (In practice it certainly does from the testing I've done. I can imagine that when both the server and client MAC address are 0/BDADDR_ANY it could cause problems). So I think the general idea of the patch is right at least. (It just came from comparing scotest.c with l2test.c really). Also I'm not the first person to see this, I found this unresolved issue when googling the problem initially:

http://comments.gmane.org/gmane.linux.bluez.kernel/3122

Signature · 2 lines
Brian Smith.
"The bells of clocktowers stitch the sleeper's dreams together." - Memory Palace