Hi,
I want to test the new batman V with throughput based metric, but I could not find out how to compile batman-adv against cfg80211 from linux backports.
At the moment first I am compiling cfg80211 and ath10k from the backports tree. After that I manually copy the created Modules.symvers from backports build dir to the batman-adv folder and running make with KBUILD_EXTRA_SYMBOLS, because otherwise I am getting a warning that the cfg80211_get_station symbol is missing.
Then batman-adv compiles fine. After loading the module I am getting an error with batadv_v_elp_throughput_metric_update :
[ 150.332710] Unable to handle kernel NULL pointer dereference at virtual address 00000000 [ 150.341079] pgd = 80004000 [ 150.343900] [00000000] *pgd=00000000 [ 150.347735] Internal error: Oops: 80000007 [#1] SMP ARM [ 150.347768] Modules linked in: batman_adv(O) bluetooth 6lowpan_iphc arc4 sky2 ath10k_pci(O) ath10k_core(O) ath(O) mac80211(O) ftdi_sio usbserial cfg80211(O) compat(O) [ 150.347779] CPU: 0 PID: 21 Comm: kworker/u4:1 Tainted: G O 3.14.48+yocto+g36eba1d #1 [ 150.347836] Workqueue: bat_events batadv_v_elp_throughput_metric_update [batman_adv] [ 150.347841] task: 9c0c5f80 ti: 9c1e4000 task.ti: 9c1e4000 [ 150.347848] PC is at 0x0 [ 150.347851] LR is at 0x0 [ 150.347857] pc : [<00000000>] lr : [<00000000>] psr: 20070013 [ 150.347857] sp : 9c1e5f08 ip : 0000001e fp : 9c004400 [ 150.347861] r10: 9c1e4008 r9 : 00000000 r8 : 9c004400 [ 150.347865] r7 : 00000000 r6 : 9de20500 r5 : 00000002 r4 : 9c1bc980 [ 150.347868] r3 : 96a0ee38 r2 : 00000001 r1 : 00000000 r0 : 96a0ee00 [ 150.347874] Flags: nzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment kernel [ 150.347878] Control: 10c53c7d Table: 2697804a DAC: 00000015 [ 150.347882] Process kworker/u4:1 (pid: 21, stack limit = 0x9c1e4238) [ 150.347886] Stack: (0x9c1e5f08 to 0x9c1e6000)
Regards Hannes
On Wednesday 08 June 2016 12:03:32 Harms, Hannes wrote:
Hi,
I want to test the new batman V with throughput based metric, but I could not find out how to compile batman-adv against cfg80211 from linux backports.
You cannot build against the backports modules with the batman-adv Makefile. This is only for systems which don't replace cfg80211 with an incompatible version. But you should have a look at the OpenWrt compile rules (especially NOSTDINC_FLAGS [1]). It basically uses its own compat-hack header and uses the rest from the backports compat layer + the new cfg80211 headers.
Most important here is that you don't compile against cfg80211.h of your kernel when your backports stuff uses complete different struct definitions (this is most likely what causes the crash on your system).
Kind regards, Sven
[1] https://github.com/openwrt-routing/packages/blob/master/batman-adv/Makefile
Hi Sven,
thanks for the advice. I managed somehow to integrate batman inside my backport tree and changed the Makefiles etc. Now the mesh is up and BATMAn V seems to work, at least I can ping other nodes and measure throughput.
But unfortunately now batctl always says that the mesh is not activated (no interface), which is obviously wrong. Maybe I made somewhere a mistake while porting the code.
Can you guess what goes wrong?
Best Regards, Hannes
Am 08.06.2016 um 13:20 schrieb Sven Eckelmann:
On Wednesday 08 June 2016 12:03:32 Harms, Hannes wrote:
Hi,
I want to test the new batman V with throughput based metric, but I could not find out how to compile batman-adv against cfg80211 from linux backports.
You cannot build against the backports modules with the batman-adv Makefile. This is only for systems which don't replace cfg80211 with an incompatible version. But you should have a look at the OpenWrt compile rules (especially NOSTDINC_FLAGS [1]). It basically uses its own compat-hack header and uses the rest from the backports compat layer + the new cfg80211 headers.
Most important here is that you don't compile against cfg80211.h of your kernel when your backports stuff uses complete different struct definitions (this is most likely what causes the crash on your system).
Kind regards, Sven
[1] https://github.com/openwrt-routing/packages/blob/master/batman-adv/Makefile
On Thursday 09 June 2016 15:44:03 Harms, Hannes wrote: [...]
But unfortunately now batctl always says that the mesh is not activated (no interface), which is obviously wrong. Maybe I made somewhere a mistake while porting the code.
Can you guess what goes wrong?
I don't even know what code you've ported. Was it batman-adv 2016.2? Maybe you are using an really old version and ran in an already fixed bug [1] which made it impossible to add down interfaces and later use them in B.A.T.M.A.N. V. Or there is something else because there is no "mesh is not activated" or "no interface" error in batctl/batman-adv and so I have to guess what you may have meant and what you have tried to execute.
Kind regards, Sven
b.a.t.m.a.n@lists.open-mesh.org
-
Harms, Hannes -
Sven Eckelmann