<div dir="ltr"><div><div>Hi Aleksander, > I saw your merge request and replied about that there, the problem you > had with ModemManager not detecting the cdc-wdm port as a QMI port was > indeed the kernel driver name, "qmi_wwan_q" vs "qmi_wwan": > <a href="https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/merge_requests/406" rel="noreferrer" target="_blank">https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/merge_requests/406</a></div> Thanks for confirmation, no worries about the merge request, I wasn't convinced this is the way to fix it neither. </div>Hi Carl, > For you are using EC21, the upstream qmi_wwan driver is enough. > If the EC21's VID and PID is not in the qmi_wwan.c products table, you just need add it, as next: > {QMI_QUIRK_SET_DTR(0x2c7c, 0x0121, 4)}, /* Quectel EC21 Mini PCIe */ > > If you are using LTE-A or 5G modules, it is recomment to use MBIM (set by at+qcfg="usbnet", 2 ) <div> > (for these modems, it is recomment or mandatory to use "QMAP" fucntion, but not well support by the upsream qmi_wwan)</div><div> </div><div>Thanks for comment, I am using kernel 5.4.8 - I've just checked that upstream qmi_wwan driver handles EG21 without any issues, so just like you've said there is no point to add the qmi_wwan_q driver. Please update the "LTE&5G Linux USB Driver User Guide" doc so that in future no one else falls into these issues. </div><div>For those who wants to use qmi_wwan_q driver with a Modem Manager, here is the hint how the workaround patch should look like - I am basically replacing strcmp checks with has_prefix checks against the 'qmi_wwan' driver name to accept the 'qmi_wwan_q' driver as well: diff --git a/src/kerneldevice/mm-kernel-device-udev.c b/src/kerneldevice/mm-kernel-device-udev.c index 4b06969e..10abefc4 100644 --- a/src/kerneldevice/mm-kernel-device-udev.c +++ b/src/kerneldevice/mm-kernel-device-udev.c @@ -87,7 +87,7 @@ get_device_ids (GUdevDevice *device, success = TRUE; goto out; } else if (g_str_has_prefix (parent_subsys, "usb") && - (!g_strcmp0 (g_udev_device_get_driver (parent), "qmi_wwan") || + (g_str_has_prefix (g_udev_device_get_driver (parent), "qmi_wwan") || !g_strcmp0 (g_udev_device_get_driver (parent), "cdc_mbim"))) { /* Need to look for vendor/product in the parent of the QMI/MBIM device */ GUdevDevice *qmi_parent; diff --git a/src/mm-plugin.c b/src/mm-plugin.c index 7150b037..06128c93 100644 --- a/src/mm-plugin.c +++ b/src/mm-plugin.c @@ -316,7 +316,7 @@ apply_pre_probing_filters (MMPlugin *self, for (j = 0; drivers[j]; j++) { /* If we match the QMI driver: unsupported */ - if (g_str_equal (drivers[j], "qmi_wwan")) { + if (g_str_has_prefix (drivers[j], "qmi_wwan")) { mm_dbg ("(%s) [%s] filtered by implicit QMI driver", self->priv->name, mm_kernel_device_get_name (port)); @@ -801,7 +801,7 @@ mm_plugin_supports_port (MMPlugin *self, probe_run_flags |= MM_PORT_PROBE_QCDM; } else { /* cdc-wdm ports... */ - if (self->priv->qmi && !g_strcmp0 (mm_kernel_device_get_driver (port), "qmi_wwan")) + if (self->priv->qmi && g_str_has_prefix (mm_kernel_device_get_driver (port), "qmi_wwan")) probe_run_flags |= MM_PORT_PROBE_QMI; else if (self->priv->mbim && !g_strcmp0 (mm_kernel_device_get_driver (port), "cdc_mbim")) probe_run_flags |= MM_PORT_PROBE_MBIM; @@ -980,7 +980,7 @@ mm_plugin_create_modem (MMPlugin *self, #if defined WITH_QMI if (MM_IS_BROADBAND_MODEM_QMI (modem) && port_type == MM_PORT_TYPE_NET && - g_strcmp0 (driver, "qmi_wwan") != 0) { + !g_str_has_prefix (driver, "qmi_wwan")) { /* Non-QMI net ports are ignored in QMI modems */ mm_dbg ("(%s/%s): ignoring non-QMI net port in QMI modem", subsys, name); force_ignored = TRUE; @@ -989,7 +989,7 @@ mm_plugin_create_modem (MMPlugin *self, if (!MM_IS_BROADBAND_MODEM_QMI (modem) && port_type == MM_PORT_TYPE_NET && - g_strcmp0 (driver, "qmi_wwan") == 0) { + g_str_has_prefix (driver, "qmi_wwan")) { /* QMI net ports are ignored in non-QMI modems */ mm_dbg ("(%s/%s): ignoring QMI net port in non-QMI modem", subsys, name); force_ignored = TRUE; @@ -997,7 +997,7 @@ mm_plugin_create_modem (MMPlugin *self, } #else if (port_type == MM_PORT_TYPE_NET && - g_strcmp0 (driver, "qmi_wwan") == 0) { + g_str_has_prefix (driver, "qmi_wwan")) { /* QMI net ports are ignored if QMI support not built */ mm_dbg ("(%s/%s): ignoring QMI net port as QMI support isn't available", subsys, name); force_ignored = TRUE; </div><div>This was checked for Modem Manager v1.12.10. </div><div> </div><div>Thanks a lot for help guys! </div><div> </div><div>Best Regards, </div><div>Rafał </div></div> <div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Thu, Nov 26, 2020 at 10:09 AM Aleksander Morgado <<a href="mailto:aleksander@aleksander.es" target="_blank">aleksander@aleksander.es</a>> wrote: </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hey Carl, > Quectel has lot of customers, and most of them are use system like openwrt, android, others embedded -linux. > and the kernel version used is usually is V3.x,V4.x. > it will spend very much time (for quectel-self and customers): > if we appiled patch to the upstream qmi_wwan driver, and back-port the driver to the customer's kernel. > So we directly povide qmi_wwan_q.c, suitable for ALL kernel versions (V3.2 and later), the customers just need 'copy + paste' it. > That's a fair enough reason I guess. May I suggest, though, that instead of always telling your customers to use the "qmi_wwan_q" driver, you could tell them something like "the EC21 (0x2c7c, 0x0121) is supported in the Linux kernel since v4.9"? Your driver should be a fallback case when the user is using an old kernel that doesn't support the module by default, your driver should not be the default IMO. Keeping a list with your module versions and the kernel versions where support was introduced would be a great source of information for your customers ;) -- Aleksander <a href="https://aleksander.es" rel="noreferrer" target="_blank">https://aleksander.es</a> </blockquote></div>