Hi, On Fri, Sep 16, 2016 at 09:59:14AM +0200, Aleksander Morgado wrote: > Hey, > > Code-wise review of the patch only here; see comments below. > > On Thu, Sep 15, 2016 at 6:57 PM, Stefan Eichenberger <[email protected]> > wrote: > > > > The ME909u-512 does theoretically support the DHCP feature over AT but if > > this is done right after the NDISDUP command, it seems that the firmware > > has some trouble with it because the firmware will not attach the > > Ethernet header afterwards. This can be seen by doing a tcpdump while > > pinging e.g. 8.8.8.8. The request has a size of 98 Bytes, while the > > replies only have a size of 84 Bytes. The first 14 Bytes are missing > > which is exactly the size of the Ethernet header. > > > > To reproduce this issue, I've tried the command sequence manually: > > mmcli -m 0 --command 'AT^NDISDUP=1,1,"gprs.swisscom.ch"' > > mmcli -m 0 --command "AT+CREG?" > > mmcli -m 0 --command "AT^NDISSTATQRY?" > > mmcli -m 0 --command "AT+CGREG?" > > mmcli -m 0 --command "AT^DHCP?" > > > > The error appears in with this sequence > > > > mmcli -m 0 --command 'AT^NDISDUP=1,1,"gprs.swisscom.ch"' > > mmcli -m 0 --command "AT+CREG?" > > mmcli -m 0 --command "AT^NDISSTATQRY?" > > mmcli -m 0 --command "AT+CGREG?" > > > > The error does not appear in this sequence > > > > This commit applies to 1.6.0 and adds a udev option for this modem, > > which disables the DHCP over AT command feature. > > > > So calling AT^DHCP actually modifies the behavior of the network > interface? Have we seen something like this in any other Huawei modem?
I have to say the ME909u has fw version 12.636.11.01.00. We also use a ME609 with fw version 12.107.08.01.00 where it works correctly and with the ME909s which has as far as I know a HiSilicon chipset it works too. > > > Signed-off-by: Stefan Eichenberger <[email protected]> > > --- > > plugins/huawei/77-mm-huawei-net-port-types.rules | 3 ++ > > plugins/huawei/mm-broadband-bearer-huawei.c | 50 > > ++++++++++++++++++------ > > 2 files changed, 42 insertions(+), 11 deletions(-) > > > > diff --git a/plugins/huawei/77-mm-huawei-net-port-types.rules > > b/plugins/huawei/77-mm-huawei-net-port-types.rules > > index f60f1f8..d35f2d5 100644 > > --- a/plugins/huawei/77-mm-huawei-net-port-types.rules > > +++ b/plugins/huawei/77-mm-huawei-net-port-types.rules > > @@ -6,6 +6,9 @@ ENV{ID_VENDOR_ID}!="12d1", GOTO="mm_huawei_port_types_end" > > # MU609 does not support getportmode (crashes modem with default firmware) > > ATTRS{idProduct}=="1573", ENV{ID_MM_HUAWEI_DISABLE_GETPORTMODE}="1" > > > > +# MU909u does not support DHCP properly, it can happen that the Ethernet > > frames do not attach the Ethernet header afterwards. > > +ATTRS{idProduct}=="1573", ENV{ID_MM_HUAWEI_DISABLE_DHCP}="1" > > + > > # Mark the modem and at port flags for ModemManager > > SUBSYSTEMS=="usb", ATTRS{bInterfaceClass}=="ff", > > ATTRS{bInterfaceSubClass}=="01", ATTRS{bInterfaceProtocol}=="01", > > ENV{ID_MM_HUAWEI_MODEM_PORT}="1" > > SUBSYSTEMS=="usb", ATTRS{bInterfaceClass}=="ff", > > ATTRS{bInterfaceSubClass}=="01", ATTRS{bInterfaceProtocol}=="02", > > ENV{ID_MM_HUAWEI_AT_PORT}="1" > > diff --git a/plugins/huawei/mm-broadband-bearer-huawei.c > > b/plugins/huawei/mm-broadband-bearer-huawei.c > > index 60a91e5..11782c3 100644 > > --- a/plugins/huawei/mm-broadband-bearer-huawei.c > > +++ b/plugins/huawei/mm-broadband-bearer-huawei.c > > @@ -465,17 +465,34 @@ connect_3gpp_context_step (Connect3gppContext *ctx) > > g_object_ref (ctx->self)); > > return; > > > > - case CONNECT_3GPP_CONTEXT_STEP_IP_CONFIG: > > - mm_base_modem_at_command_full (ctx->modem, > > - ctx->primary, > > - "^DHCP?", > > - 3, > > - FALSE, > > - FALSE, > > - NULL, > > - > > (GAsyncReadyCallback)connect_dhcp_check_ready, > > - g_object_ref (ctx->self)); > > - return; > > + case CONNECT_3GPP_CONTEXT_STEP_IP_CONFIG: { > > + GUdevClient *client; > > + GUdevDevice *data_device; > > + > > Don't call it "data_device"; as it's not; it's a tty_device. > > > + // ME909u has a problem with DHCP over AT. If it's done right > > after NDSIDUP > > + // the modem doesn't send the Ethernet header anymore which > > confuses the network stack > > No comments with //, use only /* this */ even for single line comments :) > > > + client = g_udev_client_new (NULL); > > + data_device = (g_udev_client_query_by_subsystem_and_name ( > > + client, > > + "tty", > > + mm_port_get_device > > (&ctx->primary->parent.parent))); > > No "&primary->parent.parent". I'm assuming you need a MMPort from a > MMPortAt; so you should do: > mm_port_get_device (MM_PORT (ctx->primary)) > > > + if (!data_device || !g_udev_device_get_property_as_boolean > > (data_device, "ID_MM_HUAWEI_DISABLE_DHCP")) { > > Shouldn't we actually error out if !data_device? > > > + mm_base_modem_at_command_full (ctx->modem, > > + ctx->primary, > > + "^DHCP?", > > + 3, > > + FALSE, > > + FALSE, > > + NULL, > > + > > (GAsyncReadyCallback)connect_dhcp_check_ready, > > + g_object_ref (ctx->self)); > > Both client and data_device are leaking here. > > > + return; > > + } > > + > > Both client and data_device are leaking here. > > > + mm_info("This device (%s) does not support DHCP over AT", > > mm_port_get_device (ctx->data)); > > You could set here the ctx->ipv4_config value with > MM_BEARER_IP_METHOD_DHCP, so that you don't need to change the next > step. I think this would also be clearer; ipv4_config will be set > always in the CONNECT_3GPP_CONTEXT_STEP_IP_CONFIG step, regardless of > whether the modem supports ^DHCP or not. > > > + ctx->step ++; > > + /* Fall down to the next step */ > > + } > > > > case CONNECT_3GPP_CONTEXT_STEP_LAST: > > /* Clear context */ > > @@ -489,6 +506,17 @@ connect_3gpp_context_step (Connect3gppContext *ctx) > > mm_bearer_connect_result_new (ctx->data, > > ctx->ipv4_config, NULL), > > (GDestroyNotify)mm_bearer_connect_result_unref); > > } > > + else { > > + MMBearerIpConfig *ipv4_config; > > + > > + ipv4_config = mm_bearer_ip_config_new (); > > + mm_bearer_ip_config_set_method (ipv4_config, > > MM_BEARER_IP_METHOD_DHCP); > > + g_simple_async_result_set_op_res_gpointer ( > > + ctx->result, > > + mm_bearer_connect_result_new (ctx->data, ipv4_config, > > NULL), > > + (GDestroyNotify)mm_bearer_connect_result_unref); > > + g_object_unref (ipv4_config); > > + } > > } > > > > connect_3gpp_context_complete_and_free (ctx); > > -- > > 2.0.4 > > > Thanks a lot for the review so far, I will fix the findings as soon as it seems to be okay to have this dirty workaround. Regards, Stefan _______________________________________________ ModemManager-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/modemmanager-devel
