B.A.T.M.A.N

b.a.t.m.a.n@lists.open-mesh.org

15 participants
4308 discussions

[B.A.T.M.A.N.] [RFC 0/2] batman-adv: netlink restructuring, part 2
by Sven Eckelmann 05 Nov '18

05 Nov '18

Hi, Jiri Pirko called the batman-adv configuration interface a while back "a huge mistake" [1] and suggested genl as alternative. But it is not completely clear what he meant by that. I had a look at the implementation in team and devlink (he is maintainer of both). Unfortunately, both work in slightly different ways. But they seem to have in common that there are different classes of configuration options/params which then have their own set of commands. But each set of command works for multiple options/params. I have tried to replicate something similar with this implementation. It might be important to point out that it requires multiple messages when trying to change multiple options (e.g. modification of gw_bandwidth_down and gw_bandwidth_up require two messages). This should be more like devlink and not like team. But it directly sends the new options back via genlmsg_multicast to all subscribers of the "config" mcast group. There is also currently no nesting of attributes. The code is basically around the triplet name-type-value. All three are required for set operations, name is required for get operations and nothing of the triplet is required for dump operations (because it will create messages containing this triplet). The different classes of options also require as input: * mesh interface (the batadv device) - meshif_index * hard interface specific (the enslaved devices) - meshif_index - hardif_index * vlan specific (vlan on top of the mesh interface) - meshif_index - vlan id Luckily, Jiri Pirko told us that he will have a look at patches. Here are two patches with an initial implementation. I might have to split them futher - maybe three patches for the infrastructure part and a couple for the import of the sysfs options in the new infrastructure. I am happy to receive suggestions how I should split them. There is also some potential in reducing the shared code for the .doit commands by moving the BATADV_ATTR_MESH_IFINDEX -> bat_priv code to .pre and .post handlers. Kind regards, Sven [1] https://www.open-mesh.org/issues/300 Sven Eckelmann (2): batman-adv: Add infrastructure for netlink config manipulation batman-adv: Convert existing sysfs options to netlink include/uapi/linux/batman_adv.h | 55 + net/batman-adv/Makefile | 1 + net/batman-adv/netlink.c | 50 +- net/batman-adv/netlink.h | 6 + net/batman-adv/netlink_cfg.c | 2096 +++++++++++++++++++++++++++++++ net/batman-adv/netlink_cfg.h | 78 ++ 6 files changed, 2280 insertions(+), 6 deletions(-) create mode 100644 net/batman-adv/netlink_cfg.c create mode 100644 net/batman-adv/netlink_cfg.h -- 2.19.1

2 7

[B.A.T.M.A.N.] [RFC 0/2] batctl: netlink restructuring, part 3
by Sven Eckelmann 04 Nov '18

04 Nov '18

As mentioned earlier, the netlink config restructuring in batctl will be splitted in multiple steps. The first step was already merged and the RFC changes for the netlink interface in batman-adv were just posted. Since it doesn't make sense to have a full implementation in batctl when the batman-adv part is not yet in a finished state. But it also is useless to have only the batman-adv RFC part without a userspace tool. The (by default disabled) config subcommand is (my) solution for this problem. It allows developers to play around with this interface and assign new values for specific keys and to read the current config state. Make sure that you enable it during the build: $ make clean; make CONFIG_config=y Kind regards, Sven Sven Eckelmann (2): batctl: Add generic netlink configuration subcommand batctl: Add support for config mcast group in event monitor Makefile | 1 + batman_adv.h | 55 ++++++ config.c | 542 +++++++++++++++++++++++++++++++++++++++++++++++++++ event.c | 100 ++++++++++ man/batctl.8 | 37 ++++ netlink.c | 3 + 6 files changed, 738 insertions(+) create mode 100644 config.c -- 2.19.1

1 2

[B.A.T.M.A.N.] [PATCH] batctl: Fix minor typos in manpage
by Sven Eckelmann 03 Nov '18

03 Nov '18

Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- man/batctl.8 | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/man/batctl.8 b/man/batctl.8 index d5fc5302ffd5d6a5ab1dfcb164c57a9f7f8e398f..3e210281fcf4d5d3c1d76b1f9c02459f4828979e 100644 --- a/man/batctl.8 +++ b/man/batctl.8 @@ -338,7 +338,7 @@ given batctl will not replace the MAC addresses with bat\-host names in the outp This command starts a throughput test entirely controlled by batman module in kernel space: the computational resources needed to align memory and copy data between user and kernel space that are required by other user space tools may -represent a bootleneck on some low profile device. +represent a bottleneck on some low profile device. The test consist of the transfer of 14 MB of data between the two nodes. The protocol used to transfer the data is somehow similar to TCP, but simpler: some @@ -346,9 +346,9 @@ TCP features are still missing, thus protocol performances could be worst. Since a fixed amount of data is transferred the experiment duration depends on the network conditions. The experiment can be interrupted with CTRL + C. At the end of a successful experiment the throughput in KBytes per second is returned, -togheter with the experiment duration in millisecond and the amount of bytes +together with the experiment duration in millisecond and the amount of bytes transferred. If too many packets are lost or the specified MAC address is not -reachable, a message notifing the error is returned instead of the result. +reachable, a message notifying the error is returned instead of the result. .RE .br .SH FILES

1 1

[B.A.T.M.A.N.] [PATCH v3 00/42] batctl: pre-netlink restructuring, part 1
by Sven Eckelmann 03 Nov '18

03 Nov '18

Hi, the batctl command already has support to get (previously debugfs) tables via netlink. It does this while still being able to fall back to the old debugfs tables on kernels which don't support the new generic netlink family. Something similar should be done in the future for the settings which are currently part of sysfs. But we can already see that the current integration of netlink in batctl only done in a single file - netlink.c But this file is already starting to be so big that working with it is rather cumbersome. So as first steps: * restructure command registration * move commands in separate files (which should store their actual implementation) * add a new helper command to receive multicast group messages from the kernel More things will follow in the future. Changes: v3: * correctly split ap_isolation and bonding patch * fix copyright headers for new files * move more routing_algo specific code to its command specific file * fix build problems with musl in OpenWrt 18.06 * switch Makefile's OBJ + OBJ_BISECT to obj-y and obj-n * add way to "freely" disable some commands for size optimization reasons - it is for example now possible to reduce the size of the batctl on mips_24kc_musl from 64604 bytes to 14252 bytes when only the interface commands are required (I know, stupid example) - the binary to generate the debug tables output for the same architectur used 26988 bytes v2: * Add missing orig_interval.c Kind regards, Sven Sven Eckelmann (42): batctl: Drop unused define SOCKET_PATH_FMT batctl: Use common code organization for statistics batctl: Drop legacy vis_* related warning messages batctl: Move loglevel command to separate file batctl: Move log command to separate file batctl: Move gw_mode command to separate file batctl: Move routing_algo command to separate file batctl: Rename tp_meter to throughputmeter batctl: Introduce datastructure for subcommands batctl: Add per command flags batctl: Use command structure for remaining subcommands batctl: Use getopt to parse main options batctl: Store usage line next to command batctl: Prepare command infrastructure for shared functions batctl: Add type to command to structure usage output batctl: Convert debug table to command infrastructure batctl: Convert sysfs settings to command infrastructure batctl: Move backbonetable debug table to own file batctl: Move claimtable debug table to own file batctl: Move dat_cache debug table to own file batctl: Move gateways debug table to own file batctl: Move mcast_flags debug table to own file batctl: Move nc_nodes debug table to own file batctl: Move neighbors debug table to own file batctl: Move originators debug table to own file batctl: Move transglobal debug table to own file batctl: Move translocal debug table to own file batctl: Move aggregation setting to own file batctl: Move ap_isolation setting to own file batctl: Move bonding setting to own file batctl: Move bridge_loop_avoidance setting to own file batctl: Move distributed_arp_table setting to own file batctl: Move fragmentation setting to own file batctl: Move isolation_mark setting to own file batctl: Move multicast_mode setting to own file batctl: Move network_coding setting to own file batctl: Move orig_interval setting to own file batctl: Use external netlink socket for debug tables batctl: Move routing_algo specific code it command source file batctl: Add command to monitor for netlink events batctl: Simplify enabling for bisect_iv subcommand batctl: Allow to enable/disable subcommands Makefile | 91 ++- translate.h => aggregation.c | 15 +- ap_isolation.c | 33 + backbonetable.c | 126 ++++ bisect_iv.c | 6 +- bisect_iv.h | 4 - traceroute.h => bonding.c | 15 +- bridge_loop_avoidance.c | 33 + claimtable.c | 131 ++++ dat_cache.c | 147 +++++ debug.c | 218 +------ debug.h | 38 +- distributed_arp_table.c | 33 + event.c | 246 ++++++++ ping.h => fragmentation.c | 15 +- functions.c | 28 +- functions.h | 1 - gateways.c | 175 +++++ gw_mode.c | 171 +++++ interface.c | 23 +- ioctl.h | 28 - isolation_mark.c | 35 + log.c | 73 +++ loglevel.c | 147 +++++ main.c | 274 ++++---- main.h | 65 +- man/batctl.8 | 4 + mcast_flags.c | 180 ++++++ multicast_mode.c | 33 + interface.h => nc_nodes.c | 14 +- neighbors.c | 142 +++++ netlink.c | 1332 ++++----------------------------------- netlink.h | 55 +- network_coding.c | 33 + orig_interval.c | 35 + originators.c | 228 +++++++ ping.c | 10 +- routing_algo.c | 263 ++++++++ ioctl.c => statistics.c | 10 +- sys.c | 432 +------------ sys.h | 39 +- tcpdump.c | 5 +- tcpdump.h | 2 - tp_meter.c => throughputmeter.c | 10 +- tp_meter.h | 28 - traceroute.c | 10 +- transglobal.c | 162 +++++ translate.c | 8 +- translocal.c | 158 +++++ 49 files changed, 3156 insertions(+), 2208 deletions(-) rename translate.h => aggregation.c (67%) create mode 100644 ap_isolation.c create mode 100644 backbonetable.c rename traceroute.h => bonding.c (68%) create mode 100644 bridge_loop_avoidance.c create mode 100644 claimtable.c create mode 100644 dat_cache.c create mode 100644 distributed_arp_table.c create mode 100644 event.c rename ping.h => fragmentation.c (67%) create mode 100644 gateways.c create mode 100644 gw_mode.c delete mode 100644 ioctl.h create mode 100644 isolation_mark.c create mode 100644 log.c create mode 100644 loglevel.c create mode 100644 mcast_flags.c create mode 100644 multicast_mode.c rename interface.h => nc_nodes.c (73%) create mode 100644 neighbors.c create mode 100644 network_coding.c create mode 100644 orig_interval.c create mode 100644 originators.c create mode 100644 routing_algo.c rename ioctl.c => statistics.c (90%) rename tp_meter.c => throughputmeter.c (97%) delete mode 100644 tp_meter.h create mode 100644 transglobal.c create mode 100644 translocal.c -- 2.11.0

1 43

[B.A.T.M.A.N.] [PATCH maint] batman-adv: Use explicit tvlv padding for ELP packets
by Sven Eckelmann 03 Nov '18

03 Nov '18

The announcement messages of batman-adv COMPAT_VERSION 15 have the possibility to announce additional information via a dynamic TVLV part. This part is optional for the ELP packets and currently not parsed by the Linux implementation. Still out-of-tree versions are using it to transport things like neighbor hashes to optimize the rebroadcast behavior. Since the ELP broadcast packets are smaller than the minimal ethernet packet, it often has to be padded. This is often done (as specified in RFC894) with octets of zero and thus work perfectly fine with the TVLV part (making it a zero length and thus empty). But not all ethernet compatible hardware seems to follow this advice. To avoid ambiguous situations when parsing the TVLV header, just force the 4 bytes (TVLV length + padding) after the required ELP header to zero. Fixes: a4b88af77e28 ("batman-adv: ELP - adding basic infrastructure") Reported-by: Linus Lüssing <linus.luessing(a)c0d3.blue> Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- net/batman-adv/bat_v_elp.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/net/batman-adv/bat_v_elp.c b/net/batman-adv/bat_v_elp.c index 9f481cfd..e8090f09 100644 --- a/net/batman-adv/bat_v_elp.c +++ b/net/batman-adv/bat_v_elp.c @@ -352,19 +352,21 @@ static void batadv_v_elp_periodic_work(struct work_struct *work) */ int batadv_v_elp_iface_enable(struct batadv_hard_iface *hard_iface) { + static const size_t tvlv_padding = sizeof(__be32); struct batadv_elp_packet *elp_packet; unsigned char *elp_buff; u32 random_seqno; size_t size; int res = -ENOMEM; - size = ETH_HLEN + NET_IP_ALIGN + BATADV_ELP_HLEN; + size = ETH_HLEN + NET_IP_ALIGN + BATADV_ELP_HLEN + tvlv_padding; hard_iface->bat_v.elp_skb = dev_alloc_skb(size); if (!hard_iface->bat_v.elp_skb) goto out; skb_reserve(hard_iface->bat_v.elp_skb, ETH_HLEN + NET_IP_ALIGN); - elp_buff = skb_put_zero(hard_iface->bat_v.elp_skb, BATADV_ELP_HLEN); + elp_buff = skb_put_zero(hard_iface->bat_v.elp_skb, + BATADV_ELP_HLEN + tvlv_padding); elp_packet = (struct batadv_elp_packet *)elp_buff; elp_packet->packet_type = BATADV_ELP; -- 2.19.1

1 1

[B.A.T.M.A.N.] Question on 'batctl o'
by John Gorkos 02 Nov '18

02 Nov '18

What does the leading asterisk indicate by an originator in 'batctl o'? i.e.: [B.A.T.M.A.N. adv 2017.3, MainIF/MAC: wlan0/00:c0:ca:97:4c:34 (bat0/e2:9d:52:8c:34:83 BATMAN_IV)] Originator last-seen (#/255) Nexthop [outgoingIF] * 00:0f:00:54:9b:d4 0.130s (197) 00:0f:00:54:9b:d4 [ wlan0] * b8:27:eb:a0:2a:53 0.830s (225) b8:27:eb:a0:2a:53 [ wlan0] Regards, John Gorkos

2 1

[B.A.T.M.A.N.] [PATCH v3] alfred: Request MAC resolution for IPv4 address not in ARP cache
by Jonathan Haws 01 Nov '18

01 Nov '18

When using IPv4, if the remote server is not yet in the ARP cache, the MAC resolution will fail and data appear to not be shared via alfred. Add a routine (modified from batctl sources) to request MAC resolution by simply sending a datagram to the discard port (UDP/9). This adds the remote MAC to the ARP cache, resulting in successful MAC resolution. Fixes: c7da798113a2 ("alfred: IPv4 multicast distribution support.") Signed-off-by: Jonathan Haws <jhaws(a)sdl.usu.edu> --- util.c | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-) diff --git a/util.c b/util.c index dd3f00f..0794792 100644 --- a/util.c +++ b/util.c @@ -30,6 +30,7 @@ #include <sys/ioctl.h> #include <sys/time.h> #include <time.h> +#include <unistd.h> #include "alfred.h" int time_diff(struct timespec *tv1, struct timespec *tv2, @@ -80,11 +81,35 @@ bool is_valid_ether_addr(uint8_t addr[ETH_ALEN]) return true; } +static void ipv4_request_mac_resolve(const alfred_addr *addr) +{ + const struct sockaddr *sockaddr; + struct sockaddr_in inet4; + size_t sockaddr_len; + int sock; + char t = 0; + + sock = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); + if (sock < 0) + return; + + memset(&inet4, 0, sizeof(inet4)); + inet4.sin_family = AF_INET; + inet4.sin_port = htons(9); + inet4.sin_addr.s_addr = addr->ipv4.s_addr; + sockaddr = (const struct sockaddr *)&inet4; + sockaddr_len = sizeof(inet4); + + sendto(sock, &t, sizeof(t), 0, sockaddr, sockaddr_len); + close(sock); +} + int ipv4_arp_request(struct interface *interface, const alfred_addr *addr, struct ether_addr *mac) { struct arpreq arpreq; struct sockaddr_in *sin; + int retries = 1; memset(&arpreq, 0, sizeof(arpreq)); memset(mac, 0, ETH_ALEN); @@ -96,8 +121,13 @@ int ipv4_arp_request(struct interface *interface, const alfred_addr *addr, strncpy(arpreq.arp_dev, interface->interface, sizeof(arpreq.arp_dev)); arpreq.arp_dev[sizeof(arpreq.arp_dev) - 1] = '\0'; - if (ioctl(interface->netsock, SIOCGARP, &arpreq) < 0) - return -1; + while ((ioctl(interface->netsock, SIOCGARP, &arpreq) < 0) || !(arpreq.arp_flags & ATF_COM)) { + ipv4_request_mac_resolve(addr); + usleep(200000); + + if (retries-- == 0) + break; + } if (arpreq.arp_flags & ATF_COM) { memcpy(mac, arpreq.arp_ha.sa_data, sizeof(*mac)); -- 2.17.1

3 2

[B.A.T.M.A.N.] [PATCH 0/9] batman-adv: Add inconsistent nl dump detection
by Sven Eckelmann 30 Oct '18

30 Oct '18

The netlink dump functionality transfers a large number of entries from the kernel to userspace. It is rather likely that the transfer has to interrupted and later continued. During that time, it can happen that either new entries are added or removed. The userspace could than either receive some entries multiple times or miss entries. Commit 670dc2833d14 ("netlink: advertise incomplete dumps") introduced a mechanism to inform userspace about this problem. Userspace can then decide whether it is necessary or not to retry dumping the information again. The netlink dump functions have to be switched to exclusive locks to avoid changes while the current message is prepared. And an external generation sequence counter is introduced which tracks all modifications of the list/hash. There some netlink dump commands which were not yet modified. These datastructures use subentries (or entries for multiple main object) which makes mapping to a generation sequence counter more problematic: * originator * neighbor list (multiple interfaces make it harder) * global translation table Also the batadv_algo stuff was ignored because it doesn't even have locks and is not supposed to be modified at runtime. Kind regards, Sven Sven Eckelmann (9): batman-adv: Add compat for genl_dump_check_consistent batman-adv: Add inconsistent gateway netlink dump detection batman-adv: Add inconsistent hardif netlink dump detection batman-adv: Store modification counter via hash helpers batman-adv: Add inconsistent backbone netlink dump detection batman-adv: Add inconsistent claim netlink dump detection batman-adv: Add inconsistent dat netlink dump detection batman-adv: Add inconsistent local TT netlink dump detection batman-adv: Add inconsistent multicast netlink dump detection compat-include/net/genetlink.h | 45 ++++++++++++++ net/batman-adv/bat_iv_ogm.c | 24 +++++--- net/batman-adv/bat_v.c | 26 +++++--- net/batman-adv/bridge_loop_avoidance.c | 82 +++++++++++++++----------- net/batman-adv/distributed-arp-table.c | 42 +++++++------ net/batman-adv/gateway_client.c | 3 + net/batman-adv/hard-interface.c | 3 + net/batman-adv/hash.c | 2 + net/batman-adv/hash.h | 6 ++ net/batman-adv/main.c | 3 + net/batman-adv/main.h | 1 + net/batman-adv/multicast.c | 51 ++++++++-------- net/batman-adv/netlink.c | 24 ++++---- net/batman-adv/translation-table.c | 41 +++++++------ net/batman-adv/types.h | 5 +- 15 files changed, 233 insertions(+), 125 deletions(-) create mode 100644 compat-include/net/genetlink.h -- 2.19.1

1 9

[B.A.T.M.A.N.] alfred and batadv-vis issue
by gary 30 Oct '18

30 Oct '18

Hello guys, I setup a testbed like this. (BBN = backbone node) Switch ---------->BBN1 --------------->MP1 | ------->BBN2---------------->MP2 MP1 and MP2 may select BBN1 or BBN2 as gateway. On both MP1 and MP2, I enable Alfred and batadv-vis as follows: alfred -i br0 -4 224.0.0.1 -m & batadv-vis -i bat0 -s when I run batadv-vis again on MP1, I can't get the info from MP2. If I ping the ip address of MP2 from MP1, MP1 will get MP2's mac address by arp. And then I run batadv-vis again, I can get the data from MP2 now. Is there any configure or method to get MP2's info without ping test? Thanks, Gary

3 21

[B.A.T.M.A.N.] [PATCH maint] alfred: Fix detection of own packets for IPv4 mode
by Sven Eckelmann 29 Oct '18

29 Oct '18

The incoming packet address is checked for a match against the local interface addresses to avoid processing its own packets. The IPv4 implementation used the same code but only initialized 4 of the 16 bytes of the address in the recv function. The interface initialization code in netsock_set_interfaces set all unused bytes to zero but recv_alfred_packet was modified to use 12 random bytes from the stack. Both functions must work the same way and first set the address bytes to zero and overwrite the actual used bytes with the address bytes. Otherwise, the result of netsock_set_interfaces for own packets is random in the IPv4 implementation. Fixes: c7da798113a2 ("alfred: IPv4 multicast distribution support.") Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- Cc: Jonathan Haws <jhaws(a)sdl.usu.edu> --- recv.c | 1 + 1 file changed, 1 insertion(+) diff --git a/recv.c b/recv.c index 59d759c..5ff4bb5 100644 --- a/recv.c +++ b/recv.c @@ -416,6 +416,7 @@ int recv_alfred_packet(struct globals *globals, struct interface *interface, packet = (struct alfred_tlv *)buf; + memset(&alfred_source, 0, sizeof(alfred_source)); if (globals->ipv4mode) { memcpy(&alfred_source, &source4.sin_addr, sizeof(source4.sin_addr)); } else { -- 2.19.1

2 1

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

B.A.T.M.A.N