Some firmware Q2E24 nand blasting results
Here's a first pass focusing on Q2E24 nand blasting between 2 XOs (one B4 and one XO-1). These tests only used the new nb-clone OFW command of an 8.2.1-760 image. I'll try testing the nb-update and nb-secure later (hopefully once I get a 3rd XO unlocked for unsigned firmware testing). Q2E24 general observation: First key press at ok prompt is dropped (kept catching me out and making initial typos) not sure if this is worth a ticket. SUMMARY: No failures over 4 blasts, performance seemed fast (under quiet conditions) and/or robust (under weak network signal conditions). DETAILS: Nand blaster nb-clone results, quiet network environment with ~8 APs visible, XOs tested in both close and far locations, with ears up and ears closed. After each test the receiving XO was booted up and tested to be working. XOs next to each other, ears up. Sender 1007937 KiB sent in 677 s (1487 KiB/s) Receiver Net 672.0s 1202Kib/s Rd 46.10s FEC 7.5s CRC 19.6s Wr 95.2s Er 14.6s = 16m to complete XOs next to each other, ears closed. Sender 1007937 KiB sent in 649 s (1553 Kibs) Receiver Net 648.3s 1246KiB/s Rd 47.0s FEC 11.1s CRC 19.6s Wr 94.10s Er 14.5s = 15m to complete XOs ~7m away plus 2 walls, ears up. Sender 1007937 KiB sent in 649 s (1553 Kibs) Receiver Net 1143.9s 706Kib/s Rd 47.1s FEC 30.3s CRC 19.6s Wr 95.0s Er 14.5s = 14m to complete, saw occasional clumps purple partial blocks on first pass (~10%). XOs ~7m away plus 2 walls, ears closed. Sender 1007937 KiB sent in 649 s (1553 Kibs) Receiver Net 4294.3s 188KiB/s Rd 46.10s FEC 115.9s CRC 19.7s Wr 94.7s Er 14.6s = 78m to complete, heavy amount of purple partials (~40%) some yellow missing blocks, many passes. Looks like a very useful trick to have available! --Gary ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Some firmware Q2E24 nand blasting results
Indeed it is excellent stuff ... I've not flawed it here in several tests. Well done Mitch. -- James Cameronmailto:qu...@us.netrek.org http://quozl.netrek.org/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Some firmware Q2E24 nand blasting results
Yes, this is really nice. Will make flashing thousands of xo's a lot easier if we can't get our custom image preinstalled (not so covert plea slipped in here somehow). Worked fine over here in the tests done so far. thanks, /Ties On Fri, Dec 12, 2008 at 2:05 PM, James Cameron qu...@laptop.org wrote: Indeed it is excellent stuff ... I've not flawed it here in several tests. Well done Mitch. -- James Cameronmailto:qu...@us.netrek.org http://quozl.netrek.org/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Some firmware Q2E24 nand blasting results
Gary C Martin wrote: Q2E24 general observation: First key press at ok prompt is dropped (kept catching me out and making initial typos) not sure if this is worth a ticket. Here's my guess for what is causing the loss of first key: To get to the ok prompt, you have to type to ESC key. That could be interpreted one of two ways: a) Press and immediately release the ESC key or b) Hold down the ESC key until you are sure that OFW recognizes it. In case (b), the keyboard auto-repeats the ESC char so OFW sees N repetitions. The type ESC to interrupt handler takes the first one, and the next N-1 repetitions go to the normal OFW command line editor. ESC is a prefix char for editing commands like delete word - the editor generally follows EMACS conventions in that respect. ESC-ESC is an unimplemented command, so nothing obvious happens for paired ESC-ESC. But if N-1 is odd, the left over ESC puts the line editor in waiting for the completion character for the ESC prefix state. So the first printable character you type is taken as the completion of the ESC-whatever command. Even if that first character represents a valid editing command, you aren't likely to notice its effect, since the command line is empty at the time, so the editing command has nothing to do - no word to delete, no place to move forward or backward, etc. I never have this problem because I don't hold down ESC, I just type it once. As it turns out, the startup jingle is a good marker for when OFW is ready to accept the ESC - as soon as I hear the jingle start, I reach for the ESC key and type it once. Mitch ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Some firmware Q2E24 nand blasting results
Here's an opportunity for you folks to exercise some creativity: Develop an efficient logistics procedure for NANDblasting thousands of machines effectively. Where to put the machines (tables, floor, shelves, ...), the power adapters (or is it okay to use battery power), the boxes as they are being unpacked/repacked, how many people you need to do N units at a time, how to organize those people so they don't get in each other's way, etc. ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Some firmware Q2E24 nand blasting results
I've been able to reproduce Gary's symptom by waiting for the jingle and then holding down the ESC key. Simple rule, don't hold it down. Doesn't hurt much. -- James Cameronmailto:qu...@us.netrek.org http://quozl.netrek.org/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel