Hallo, I am experimenting about the Linux compatibility of a particular USB modem. The modem has Vendor ID:Product ID = 047e:2892. From this piece of information and various other sources, I gathered that the suitable driver would be "agrsm". So, I compiled a vanilla 2.6.27.7 kernel (ALSA 1.0.17) after adding the EXPORT_SYMBOLs to hda_codec.c as detailed in HOWTO-Agere-11c11040-HDA.html; downloaded and compiled agrsm-HDA-20080721.tar.bz2; copied the agr*.ko modules in place, created the device node, booted, modprobed, and tried to access the modem through /dev/ttyAGS3 with minicom and pppd, only to get a "No such device" error. Some investigations led to the discovery of the NONGPL switch in agrsoftmodem.c: from what I've seen, support for USB modems was disabled some time ago, and can be successfully reenabled only for kernels prior to 2.6.25. So I switched to a vanilla 2.6.24.7 (the highest pre-2.6.25 version), compiled agrsm-HDA-20080721-ALSA15.tar.bz2 (as the ALSA version in that kernel is 1.0.15), and went through the above routine again. This time, after modprobing agr*.ko, a whole new lot of previously unseen USB-related debug output showed up in dmesg, which I interpreted as a good sign. In fact, I was able to access the modem and establish a PPP session. I could even manage an inbound connection with mgetty. However, after a little more testing, the modem ceased to respond, while dmesg revealed a timeout on the USB bus: [...] Bulk Patch downloaded already! USS_CMD_CHANGE_CLK command passed Usb_BoostArmClock(): ARMCDR Reg= e, clock running @ 16 Mhz BALA IsoUsb_StartIsoDma usb 1-1: mgetty timed out on ep0out len=0/0 reset enpoint 0x1 retVal -110 usb 1-1: mgetty timed out on ep0out len=0/0 reset enpoint 0x81 retVal -110 x_usb_start_dma pRecDmabuffer c0da8220 pPlayDmabuffer c0da8e20 [...] Further connection attempts caused the machine to hang with a "not syncing: fatal exception in interrupt" kernel panic. This behaviour is quite reproducible: boot the machine, use the modem for a while for inbound and outbound connections, and the kernel will go bananas. Having some experience with isochronous transfers in other USB devices, I am inclined to think that such timeouts as the one reported in dmesg are usually caused by buggy hardware rather than a buggy driver. Could this be my case? Or have I just stumbled upon a known bug in the agrsm driver? Or maybe dial-in and dial-out don't mix well on this kind of hardware? Could someone share his/her (in)success story? Also, to let me gain some more insight, could someone confirm that snd-hda-intel.ko is not really needed in the USB modem case, although agrmodem.ko depends on it? Many thanks, David
