01:10 | dung | Is it the FDT, what makes CuBoxs Linux 'upstream'? |
08:25 | ralix | morning |
08:26 | dotarray | hello ralix! |
08:27 | dotarray | i am just about to head out and get dinner :) |
08:29 | ralix | My day has just begun. I wish you bon appetit! |
12:52 | thrillerator | I have a problem with my cubox which I already posted in the forums |
12:52 | dotarray | what's wrong, thrillerator? |
12:53 | rabeeh | hi 'thrillerator' |
12:53 | thrillerator | here's the link to the thread for reference: http://www.solid-run.com/phpbb/viewtopic.php?f=6&t=839 |
12:57 | rabeeh | so your situations is exactly what xxfog had |
12:57 | rabeeh | he tried download u-boot through the serial port and nothing happened. |
12:58 | thrillerator | yes, he responded to my thread and contacted you |
12:58 | rabeeh | are you using Linux as a host PC? |
12:58 | thrillerator | Windows 7 |
12:58 | rabeeh | teraterm? |
12:58 | thrillerator | but I even tried it with Ubuntu |
12:58 | thrillerator | yeah TeraTerm |
12:59 | rabeeh | so it does finish transferring the u-boot..._uart.bin file via xmodem too? |
13:00 | thrillerator | yes it does. but after that TeraTerm doesn't show the Cubox prompt |
13:00 | thrillerator | just a blank screen |
13:00 | rabeeh | can you do another trial? |
13:01 | rabeeh | just power down and power up CuBox WITHOUT pressing the small button beneath the eSata connector |
13:01 | rabeeh | and then click enter few times on TeraTerm |
13:01 | rabeeh | does it show 'bootrom' command line? |
13:02 | thrillerator | I'm not at home right and cannot try that. but I'll note it down and report back. |
13:02 | rabeeh | let me explain the logic first |
13:03 | thrillerator | okay |
13:03 | rabeeh | there is an unwipable bootrom image inside the main processor. once powered up the CPU runs from the bootrom and according to the button state it chooses next boot source |
13:04 | rabeeh | if button is unpressed, it loads u-boot from the internal SPI flash |
13:04 | rabeeh | if button is pressed, then it loads from the serial port (which is further on connected to your PC with the micro USB cable) |
13:05 | rabeeh | the operation i'v asked is that i'm assuming is that the SPI flash is wiped; but still it will try to take u-boot from the SPI flash. |
13:05 | rabeeh | given nothing there, the bootrom code will try few times and once pressed 'enter' few times it will giveup and show you the 'bootrom' prompt |
13:08 | thrillerator | what can I do if that's the case? |
13:30 | rabeeh | if that's the case then the issue would be in downloading u-boot image through serial port. |
13:41 | thrillerator | any fix? |
13:42 | rabeeh | not on top of my mind. |
13:42 | thrillerator | damn |
13:42 | rabeeh | lets first see the 'bootrom' prompt and then we will figure out the next step |
13:42 | rabeeh | maybe there is an issue with xmodem download to CuBox; but then you have tried two platforms. |
13:43 | rabeeh | oh; sorry - are those two OS on different machines? or the same machine you run Win7/Ubuntu? |
13:43 | thrillerator | same machine, dual boot |
13:44 | rabeeh | do you have another machine by a chance that you can test? |
13:48 | thrillerator | yes I have. I'll try it with both machines later |
14:06 | dung | Is it possible to checksum after write to flash? |
14:09 | rabeeh | dung: it does checksum |
14:09 | rabeeh | the small bootrom inside the chip reads some header (which configures DDR) and checks the image if it has correct checksum and only then it copies it and continues the boot process. |
14:14 | dung | i meen after flashing. Does U-Boot check and retry until it's successful written? |
14:16 | rabeeh | nop |
14:17 | rabeeh | it doesn't |
14:17 | dung | i read about flash fs, it's its nature loosing data on write |
14:17 | rabeeh | depends which hardware device you use |
14:17 | rabeeh | typically NOR flash (which is what most SPI flash are based on) are very reliable |
14:18 | rabeeh | NAND flash on the other hand is much less reliable; that's why you need wear aware filesystem |
20:06 | yawniek | hmm also seems my network speed is somehow capped at around 100mbit with scp, cpu is 100%. is there a way to get more troughput os is that because of ssl encyption? |
20:09 | yawniek | iperf makes 490mbit |
20:09 | yawniek | also 100% cpu |
21:27 | rabeeh | yawniek: there is hardware acceleration for certain ciphers in hardware |
21:34 | Punkley_Chillin | morning |
21:36 | rabeeh | morning |
21:55 | yawniek | rabeeh: i've seen that but im not sure how to enable it or if its already built into the archlinux cubox kernel. i loaded all modules in crypto, but not much difference |
21:55 | yawniek | ive seen that it also might need a openssl patch? |
21:56 | rmull | yawniek: scp is probably going to be limited by your disk IO as well |
21:56 | rmull | Are you writing to the microsd card? |
21:56 | rmull | You can choose other crypto algorithms to see how they behave if you would like |
21:58 | yawniek | rmull: i dont think so, its a disk on esata |
21:59 | yawniek | and i also tried memory or to dev/null. also, iperf only made around 500mbit |
21:59 | yawniek | ive seen people report 800-900 |
22:01 | rabeeh | yawniek: for iperf, it depends if CuBox is the client (sender) or the server (receiver) |
22:01 | rmull | rabeeh: Why does it depend? |
22:01 | rabeeh | sender part is hardware accelerated via TCP offload engine that is part of CuBox |
22:01 | rabeeh | (that's why CuBox is also great as a NAS system) |
22:01 | rmull | That's good to know |
22:02 | rabeeh | rmull: for samba, if you enable mmap and sendfile support you should high performance on CuBox |
22:02 | rabeeh | enable jumbo frames and full wire speed takes less than 50% of the processor |
22:03 | rabeeh | full wire speed === 1000mbps |
22:03 | yawniek | rabeeh: checked both, max was 570mbit. i want it as a nas, i just received http://www.icydock.de/goods.php?id=52 to add 4 disks |
22:04 | yawniek | rabeeh: ok gonna install samba and tune the network stack, lets see |
22:04 | yawniek | is the offload engine automatically used or do i need a special driver? |
22:05 | yawniek | just realized that its kind of stupd to use 3.5" disks, two disks now use about 30 watt. but this was the cheapest esata multiplexer case |
22:11 | koboldmaki | has somebody test the network speed between two cubox's? |
23:01 | yawniek | MSS size 8948 818 MBytes 686 Mbits/sec little bit better with jumbo frames |
23:17 | yawniek | copying over samba TO the cubox into tmpfs gives around 600mbit/s and 100% cpu load. :) |