![]() |
DSMG600.InstallingOpenSlug HistoryHide minor edits - Show changes to markup June 16, 2006, at 04:22 AM
by -- Clarified the kernel rebuild
Changed lines 45-46 from:
Wireless NIC supportto:
Wireless NICChanged lines 49-53 from:
Wired NIC supportDoesn't work, no driver in the feeds. You'll need to install a couple of patches to get the via-velocity driver to work in big-endian mode, actually. One patch set patches the driver, the other exports a symbol in the kernel that's needed in order to load the via-velocity driver as a module. Once both are added, you'll need to edit the defconfig file to specify that you want the velocity driver built as a module. Build a new kernel, flash the new kernel. Then install the kernel-module you built. Note that there are a couple of parameters you might like to pass into the via-velocity module (I use the "options" statement in the /etc/modprobe.d/eth0 file). The first choice is a compromise between taking a hit for 7 unaligned accesses per IP frame received, versus a memory copy of each IP frame to align it. I've opted to align it, add the flag "IP_byte_align=1". The other is an annoyance. If the interrupt handler gets too many frames stacked up, it complains by logging a message to the console. Adding the flag "int_works=64" minimizes the chatter. I've only seen it do this when under extreme load -- NFS copy to and and NFS copy from happening simultaneously. And despite the complaints printed to the console, the NFS transfers completed without error. to:
Wired NICDoesn't work, no driver in the feeds. You'll need to install a couple of patches to get the via-velocity driver to work in big-endian mode, actually. One patch set patches the driver, the other exports a symbol in the kernel that's needed in order to load the via-velocity driver as a module. Once both are added, you'll need to edit the defconfig file to specify that you want the velocity driver built as a module. Build a new kernel, flash the new kernel (the kernel build and flash are only required once, in order to ensure that the kernel is exporting the symbol that the module needs). Then install the kernel-module you built. Note that there are a couple of parameters you might like to pass into the via-velocity module (I use the "options" statement in the /etc/modprobe.d/eth0 file). The first choice is a compromise between taking a hit for 7 unaligned accesses per IP frame received, versus a memory copy of each IP frame to align it. I've opted to align it, add the flag "IP_byte_align=1". The other is an annoyance. If the interrupt handler gets too many frames stacked up, it complains by logging a message to the console. Adding the flag "int_works=64" minimizes the chatter. I've only seen it do this when under extreme load -- NFS copy to and and NFS copy from happening simultaneously. And despite the complaints printed to the console, the NFS transfers completed without error. June 16, 2006, at 02:23 AM
by -- fconfig, not fcconfig
Changed lines 37-38 from:
to:
June 16, 2006, at 02:19 AM
by -- typographical and formatting corrections
Changed lines 4-5 from:
to:
Changed lines 13-16 from:
NOTE: The currently-selling model is the RevB? -- this more recent model is based on the MIPS chip. This article is specific to the RevA? device. The DSM-G600 RevA? device is an ARM-based device, similar to the NSLU-2 -- except that it has double the flash, double the RAM, a gigabit network interface, a wireless network card, and an internal IDE port. When I stumbled across one of the old RevA? units on a recent business trip (in Detroit, of all places), I snapped one up -- for $130 USD, how could one go wrong? to:
NOTE: The currently-selling model is the RevB -- this more recent model is based on the MIPS chip. This article is specific to the RevA device. The DSM-G600 RevA device is an ARM-based device, similar to the NSLU-2 -- except that it has double the flash, double the RAM, a gigabit network interface, a wireless network card, and an internal IDE port. When I stumbled across one of the older RevA units on a recent business trip (in Detroit, of all places), I snapped one up -- for $130 USD, how could one go wrong? Changed lines 19-22 from:
to:
Changed lines 41-44 from:
LEDs? and ButtonsThey don't work. You'll need to install some patches, and build a new kernel. As of this writing, the patches provide support for the POWER and WLAN LEDs? only, the RESET button works as it does on the slug (meaning instant power-off), and there is experimental support for the POWER button (hold it depressed for 2 - 3 seconds to trigger a controlled shutdown and power-off). You'll have to build a new kernel, and reflash the kernel as you did originally (no root filesystem changes to enable the LEDs? and buttons, so you don't need to reflash that. to:
LEDs and ButtonsThey don't work. You'll need to install some patches, and build a new kernel. As of this writing, the patches provide support for the POWER and WLAN LEDs only, the RESET button works as it does on the slug (meaning instant power-off), and there is experimental support for the POWER button (hold it depressed for 2 - 3 seconds to trigger a controlled shutdown and power-off). You'll have to build a new kernel, and reflash the kernel as you did originally (no root filesystem changes to enable the LEDs and buttons, so you don't need to reflash that. June 16, 2006, at 02:15 AM
by -- Installing a custom OS on the DSM-G600
Added lines 1-54:
Installing OpenSlug 3.10 on the DSM-G600This article pertains to: (:table border=0 width=100% bgcolor=#ffdddd:) (:cell:) This article is intended for advanced users! If you are not willing to risk "bricking" your DSM-G600, then DO NOT attempt this. Do this at your own risk. ~mwester, 15 June 2006
(:tableend:) NOTE: The currently-selling model is the RevB? -- this more recent model is based on the MIPS chip. This article is specific to the RevA? device. The DSM-G600 RevA? device is an ARM-based device, similar to the NSLU-2 -- except that it has double the flash, double the RAM, a gigabit network interface, a wireless network card, and an internal IDE port. When I stumbled across one of the old RevA? units on a recent business trip (in Detroit, of all places), I snapped one up -- for $130 USD, how could one go wrong? Once you start working with the device, you'll discover that it has some quirks. First and foremost -- you can't do anything without a serial port. It lacks the Redboot upgrade mode that we love so much on the NSLU2 -- and even worse than that, Redboot on the device cannot drive either the wireless NIC or the built-in NIC! So, firmware needs to be loaded via the "xmodem" protocol on the console.
LEDs? and ButtonsThey don't work. You'll need to install some patches, and build a new kernel. As of this writing, the patches provide support for the POWER and WLAN LEDs? only, the RESET button works as it does on the slug (meaning instant power-off), and there is experimental support for the POWER button (hold it depressed for 2 - 3 seconds to trigger a controlled shutdown and power-off). You'll have to build a new kernel, and reflash the kernel as you did originally (no root filesystem changes to enable the LEDs? and buttons, so you don't need to reflash that. Wireless NIC supportInstall the madwifi-ng package. The dependency list is incomplete, so don't forget to manually install the wireless-tools package. This should get the wireless NIC working, at least in no security or WEP mode. I've not tested with any stronger security. Wired NIC supportDoesn't work, no driver in the feeds. You'll need to install a couple of patches to get the via-velocity driver to work in big-endian mode, actually. One patch set patches the driver, the other exports a symbol in the kernel that's needed in order to load the via-velocity driver as a module. Once both are added, you'll need to edit the defconfig file to specify that you want the velocity driver built as a module. Build a new kernel, flash the new kernel. Then install the kernel-module you built. Note that there are a couple of parameters you might like to pass into the via-velocity module (I use the "options" statement in the /etc/modprobe.d/eth0 file). The first choice is a compromise between taking a hit for 7 unaligned accesses per IP frame received, versus a memory copy of each IP frame to align it. I've opted to align it, add the flag "IP_byte_align=1". The other is an annoyance. If the interrupt handler gets too many frames stacked up, it complains by logging a message to the console. Adding the flag "int_works=64" minimizes the chatter. I've only seen it do this when under extreme load -- NFS copy to and and NFS copy from happening simultaneously. And despite the complaints printed to the console, the NFS transfers completed without error. Have fun! |