Beagle time

Further to my previous post…

My Beagle is now back to its silly-buggers mode of powering up with only the power LED on. It seems whenever Linux kernel-panics, it screws up something that stops the hardware booting up for a while

Having rummaged and found an old modem serial lead, I powered up the server box and hooked into the serial port. The first time I powered up the Beagle with the original card inserted, it reported the following:

Texas Instruments X-Loader 1.5.0 (Mar 27 2011 - 17:37:56)
Beagle xM
Reading boot sector
Loading u-boot.bin from mmc


U-Boot 2011.03-rc1-00000-g9a3cc57-dirty (Apr 01 2011 - 17:41:42)

OMAP36XX/37XX-GP ES2.1, CPU-OPP2, L3-165MHz, Max CPU Clock 1 Ghz
OMAP3 Beagle board + LPDDR/NAND
I2C:   ready
DRAM:  512 MiB

Subsequently, it would stall at the “Reading boot sector” part. Has anybody else observed anything like this?

I’m thinking there’s two possibilities – the first is some sort of weird internal quirk deep inside (after the DRAM it should report no NAND), however this doesn’t explain why it acted weird and then worked fine (until my next “kernel panic”).
The second is that the kernel panic is a coincidence, and the issue shown up more readily on Linux due to loading itself from microSD, while RISC OS only needs to manage to load the ROM image. That, being, namely, that there is something not entirely right with the microSD slot and/or interface. This could, incidentally, be a contributing factor in the problem I had with BBWrite failing to write the cmos file (when it didn’t exist). It may be that whatever the issue is, it is usually a very specific set of circumstances until something happens that would appear to make it fail entirely.

I’ll see how things are tomorrow (and I’m wishing I never tried to make a replacement microSD image… that’s the last six hours wasted). I’m going to go watch Sankarea.

Best wishes,

Rick.

I’ve had wierdnesses with both BeagleBoards and Raspberry Pis with incomplete boots, which were due to insufficient power. I can’t remember what current my BBxM PSU gives, but I think I tried one at 1A and that was not enough. What power supply are you using?

Tried making a replacement Angstrom boot microSD earlier. Got a heap of device I/O errors and stuff about buffers and write failures when trying to boot from it

Hey Rick,
you should clone your micro SD first:
dd if=/dev/sda of=/dev/sdb

You could create an image and save it on your PC:
dd if=/dev/sda of=~/disk1.img

and restore it later:
dd if=disk1.img of=/dev/sda

I don’t have access to my BB xM now, otherwise I would create an image and upload it for you.
I have got AI’s disk, which I can create image of and upload it for you to test
or maybe someone else can do the original.
I don’t think there is anything wrong with your Beagle, from my experience you couldn’t
do anything on XP machine, Ubuntu nis the way anbd use:
sudo -s
to get the superuser access.

I’ve forgotten to mention this, I used Ubuntu’s gpart, deleted axisting partition from new micro SD
and created new Fat 32, sometimes the new disks don’t start from address 0×00000000 even San Disk
gets corrupted, don’t trust the name.

Theo – good grief, how much power does this thing need to get going? I normally run it from an iomega Zip power supply – it is the meatiest thing I have that actually gives a reliable +5V. Today, this morning, it is only booting as far as seeing the DRAM. I just tried jacking it into the 5V from a PC’s power supply (ought to be enough spare capacity there, it is running a harddisc and CD drive for my RiscPC, no PC motherboard) and I boot as far as… DRAM. Once, however, it saw no NAND, the SD/MMC and said an incomplete message about using a default environment.

Funny how the iomega PSU (which may be too low rated? don’t recall right now, but I think it is around 1A) runs it for days under RISC OS, but one Linux kernel panic and it all goes to hell. I’m sure Linux advocates are steaming by now, but as far as I’m concerned my Beagle has “died” twice, directly after Linux startup faults. Once is a half-assed coincidence turned into a groundless theory. Twice? That’s a little more suspicious. If/when I get the thing to reboot, rest assured, I’m not going to try a third time…

CJE – do you have a suitable power supply that would fit into a French socket? If it is a wall brick, it’ll be difficult. If it is a lump with a wire attached to each end (like a notebook’s PSU), that’s better, I could cut off anything English and fit a French plug.

joe – sorry mate, the only Linux I run around these parts is “portableubuntu” that runs within Windows (!), and that is mostly only to have a dev environment for my PVR (an ARM TI chip that is happy with ~200mA!).

Best wishes,

Rick.

I should add… I can usually boot RISC OS with keyboard, mouse, USB stick, and s-video…
…but right now I’m trying with nothing attached, just hoping for it to get as far as turning its LEDs on.

Rick, I now run a 5A @5V supply but that’s only because I wired up two USB hubs in line with the BB itself. Seem to remember from TI docs the board itself requires ~1.5A at boot plus whatever’s connected to USB. Perhaps 2.5-3A would be a good place to start. Most power bricks use either a figure eight or kettle lead, so easy enough to plug into french CE sockets.

mail received btw, I’ll send your battery tomorrrow :)

• The BeagleBoard is shipped without any power supply. It requires a regulated 5 volts DC input, and has a barrel connector of the most common size. The BeagleBoard itself will consume something like 1 amp, but may need more in some circumstances, most notably at start-up. So you really don’t want to give it less than a 2 amp supply. Bear in mind that these figures don’t take into account the power drain of any attached USB devices. With a ‘classic’ BeagleBoard it’s best to use a hub that has its own mains power supply, so a 2 amp supply for the board will be fine. But with the BeagleBoard-xM a 3 amp supply is recommended to ensure that there is enough power available for the integrated USB hub.

http://www.riscosopen.org/wiki/documentation/show/How%20to%20get%20started%20with%20a%20BeagleBoard

Hi Rick yes we do one.
AIUI they recommend 2.5Amp capacity. We have sourced some 5V 2.5Amp with Fig8 input and even have continental fig8 leads in stock! They aren’t cheap though, PSU + Fig 8 lead + shipping to France 27.00 GBP
It isn’t yet on our website but you can order online securely by selecting:
http://cjemicros.co.uk/micros/individual/newprodpages/prodinfo.php?prodcode=VAR-UNLISTED4D
at the checkout mention in the ‘Additional Messages’ section what it is you want and we adjust the order appropriately.

Unlisted item/Custom sale

[…]

Usually available from stock

Fits CJE’s oft-cited reputation regarding stock!

Thanks Chris. I’ll get one when my pay comes in around 5th-10th June (depending on when La Banque Postale gets their act together). Great to know it has a fig. 8 lead. So long as you’ll guarantee a Beagle xM will start up on this supply, then I’ll go for it. Certainly, it would be worth making sure the power supply is not a problem before contacting Beagle for an RMA.

I will phone you (calls to UK landlines are free over the VoIP line), certainly paying in GBP from an EUR card is a tad more complicated than the method listed on your site, so it’ll be better to do it “in person” (sort of) as I need to set an amount here and provide you with a credit card number, but if the limit is not good it’ll be rejected and you can tell me straight away so we can try again. If this sounds confusing, it is a “virtual credit card” – take a look at http://www.heyrick.co.uk/blog/index.php?diary=20100907 (second section). And no, my real credit card does not work in situations where I cannot physically enter my PIN. It’s a security thing that is 50% very useful (eBay and fraud? no worries) and 50% annoying (I use my other card, a little French debit card for Amazon else I’d have to register a new card for every purchase!).

Still don’t get why the PC’s PSU didn’t offer enough juice to boot the thing. Surely the +5V rail there has plenty of amps to go around – if it can run Intel slabs, it can sure as hell run a Beagle, no?

Still (part 2) doesn’t explain why this nonsense follows a Linux kernel panic. I am attempting to start the board with nothing plugged in. This doesn’t mention the fact that through the week (and during development of BeagLEDs) I started and used RISC OS numerous times, booting flawlessly with stuff connected.
Explanations as to “WTF!?” on a postcard to……

Best wishes,

Rick.

Joe,

check his B.LOG at: www.heyrick.co.uk or www.heyrick.co.uk/blog/index.php? diary=00000000

Yikes, the eight-zeroes is the magic date for listing everything. There’s a lot of rubbish in there (well, it is a blog), it might be better to just go to www.heyrick.co.uk/blog/ and look at the last five entries listed on the right column (in desktop view).

I’ve just walked across the fields and around town (err, village (and that’s being generous)). Maybe six miles. Tired, but it was pleasant. Didn’t think about anything geeky the entire time. In fact I was cultivating a story I’d like to put together with Blender if I was any good at it.
Chris – you think £27 for a powerpack is expensive? It’s about how much Amazon wants for “Blender for dummies”. It’s just mom’s assortment of “cosies” and “murder mysteries” that come in used/new for a single solitary penny…

Oh well.

Best wishes,

Rick.

Hi,

Status update. My board is still not starting. It has been suggested that there might be a hardware error with reading from microSD cards, and perhaps the evidence does indeed point to this – all from the beginning the supplied Linux had ‘issues’ (I couldn’t configure anything, complained of missing configuration files) and now it just plain won’t work at all. Perhaps the reason RISC OS ran so well is because it only needed to get as far as loading the image and then RISC OS ignores the microSD card. This may even go as far as explaining the problems I was having with BBwriter? I have now a debug version of BBwriter, but nothing to use it on. Pffft…

Joe is sending a clone of his original microSD, so I will try my board with that and using either a Beeb power supply or one from an A3000 (depending on which I lay my hands on first). Running an ‘80s TTL-logic micro, or a large ARM board, ought to be sufficient oomph to get the board going – though at this point I would be surprised (and not entirely sure for how long I’d trust it). We’ll see…

If it doesn’t start up with that combo, I will need to RMA it back home to get poked by people way more experienced than I am. To be honest, I rather miss the days when computers were built with logic chips. Okay, things like YouTube and my daily dose of animé weren’t possible on a Beeb, but from a diagnostic point of view you could follow every single pulse and understood exactly what it did and why. I’m so glad I grew up with that stuff because as amazing as today’s little ARM devices are (and, really, a half gig of memory stuck on top of a chip containing at least four processors (ARM, DSP, Neon, VFP…) and the whole lot smaller than a euro coin, it’s pretty bloody amazing. But, if it is borked, it is borked, and there is roughly diddly/squat an average geek can do about it. We can feel the power of the ARM, but I believe, even when coding assembler to bare metal, something is still lacking because these days there is just such a large disconnect between the coding and the underlying hardware. I wonder if today’s assembler-level coders can actually truly understand what is going on inside the machine?

The Beagle guys let me know that if my RMA request is accepted (when I make it, that is), I can use any mail service that does full point-to-point tracking and insurance. This is good, for I checked DHL, FedEx, and UPS and none of them seem to have the concept of it doesn’t need to be there tomorrow for God’s sake which could explain why the average economy price was in the ballpark of €180. La Poste ought to be able to offer me something in the realms of reality…

Anyway… This was supposed to be a short status update, not a blog post. ;-)

Best wishes,

Rick.

all from the beginning the supplied Linux had ‘issues’ (I couldn’t configure anything, complained of missing configuration files)

Hi Ricardo,
my xM was behaving exactly the same way, there is nothing wrong with the Beagle itself,
they give you the unit as it is, it’s up to you what you want to run on it.
The boys from beagle had to pay someone to develop the OS, the less you demand the less you pay.

I checked DHL, FedEx, and UPS and none of them seem to have the concept of it doesn’t need to be there tomorrow for God’s sake which could explain why the average economy price was in the ballpark of €180.

Ive got better idea, you send me yours, I’ll send you mine, all sorted.

I have to say that we’ve had to replace a few beagles which have died in (possibly) similar ways. Please let us know how you get on with RMA-ing though – we’ve generally just taken the hit, which is always painful.

Rick, had the same when fiddling around with multiple µSD cards last year. Thought I’d broken the slot itself as nothing I did seemed to help. In the end I downloaded the original Angstrom distro from the BB website, got me going again in no time. Been a bit wary about sd flashing ever since.

“Are you powering the board via the USB OTG port? If you are there are two issues.

1) There is not enough power to power the board form the OTG port because of the onboard USB hub. You will need to use a double ended USB connector or a DC power supply. 2) If you have the OTG cable plugged in you will get a kernel panic no matter how you have the board power. There is an issue with the validation image supporting the OTG port. It will result in a kernel panic.

Gerald"
Answer from the Beagleboard xM experts from here:
http://markmail.org/message/rxvxs5n6xqn3sjzy

I downloaded the original Angstrom distro from the BB website

Patric, which MLO, Ubut etc. have you used, some extra information would be very useful.

Senor Ricardo is your RISC OS still booting?

Patric, which MLO, Ubut etc. have you used, some extra information would be very useful.

Phew, that’s like more than a year ago and mine’s a Rev A or B board anyway.
There used to be a downloadable Angstrom validation image containing everything needed
on the BB website, no idea if that’s still the case. Guess I could dig it up on my Laptop somewhere?

Validation image (xM) here:
http://circuitco.com/support/index.php?title=BeagleBoard-xM
(God only knows how many times I’ve burned off an image these last few days)

And for older Beagles, complete with glaring typo, here:
http://circuitco.com/support/index.php?title=BeagleBoard

Andrew – I just reimaged the original validation SD (so one that was known to work). Not hooked it to serial (other computer, long, complicated) but the behaviour was the same. Power LED, sod all else. :-/
I will try Joe’s uSD when it arrives (I have a tiny spark of optimism in that I’m using Windows to make the images, and Windows is generally crap at anything non-Windowsy), but I think the quickest route back to RISC OS would be to take him up on his swapsies offer (Joe – do you really want a non-functional board!?).

It’s a shame I’m not kitted out with cool equipment and a better brain – it would be interesting to try to determine what is actually going wrong. Is it a flaw in the design? The hardware? The manufacturing? I could blame sod’s law, but with your comment about similar problems, you know, it makes one wonder. In a Hyouka (Chitanda Eru – click for explanation picture) kind of way. ;-)

One final question, Andrew, couldn’t you send back bad boards? I just hope, as you have “taken the hit” (ouch!) that you are able to get them at something akin to “trade price”!

Monsieur Richard,
I’ve left response on your latest blog that I was willing to swap
because we have DigiKeys around the corner here, you have to
send the paper work as well.
I’ve read somewhere, that when creating micro SD card for
Beagleboard xM, manually, we copy MLO file to Fat 32 partition
first, to be sure that MLO starts from the address 0, while booting,
Most of the memory cards, when formatted on Windows systems
start after first 2 MB, that’s how the partition is created, when aligned
to MB, when we align the partition to None, while creating it, we have
got better chance that MLO starts from the very first bit on the card.
I created a few cards for my xM unit but never ever, card created on
Windows system worked, I had many problems on linux/Ubuntu,
but at least the board was booting to some stage, where I could read
how far the booting process went.

Hi Rick,
The serial port readout after DRAM: 512MiB, should look like this:

NAND: 0 MiB
MMC: OMAP SD/MMC: 0

• Warning – readenv() failed, using default environment

In: serial
Out: serial
Err: serial
Beagle xM Rev C
No EEPROM on expansion board
Die ID #147c00029ff800000168301015020023
Hit any key to stop autoboot: 0
The user button is currently NOT pressed.
SD/MMC found on device 0
reading uEnv.txt

134 bytes read
Loaded environment from uEnv.txt
Importing environment from mmc …
Loading file “/boot/uImage” from mmc device 0:2 (xxa2)
3194256 bytes read
Booting from mmc …

1. Booting kernel from Legacy Image at 80200000 …

and very, very long list.

I was told that since boards were “dev boards”, any damage was unlikely to be covered under any kind of warranty. And no, we don’t get any significant trade discounts. It is this kind of thing that has to be considered when working out product pricing (ie. warranty carries an inherent cost, divided across sales). I just have a small cardboard box with boards gathering dust, and dreams of finding someone who can fix them! It isn’t an endemic fault, and rev C’s seem better than B’s, but it does happen, unfortunately.

I reckon you contacted DigiKey but it does surprise me a little. Last I heard was they’re rather generous re faulty boards (in most cases due to user error aka wrong voltage/polarity).

I was told that since boards were “dev boards”, any damage was unlikely to be covered under any kind of warranty.

Just out of interest – how does this go under British/European law? Simply calling it a dev board may get around the necessity of various emissions tests (etc), but if something fails in a couple of weeks when treated respectably it isn’t “damage” so much as being (and here’s a phrase the British govt enjoys abusing) “not fit for purpose”. ;-)

Certainly, for the price of the things, they are a little less disposable than the likes of the RPi, so you would hope that they would be a little more resiliant.

I, of course, am hoping and praying that Joe’s uSD will boot up fine, then I can say unkind things to Windows, though I have my doubts as to whether or not it’ll work. Fingers crossed, eh?

I just have a small cardboard box with boards gathering dust, and dreams of finding someone who can fix them!

How much have you looked into this? Have you hooked ’em to a serial port? Do you have JTAG?

I have been reading the OMAP’s TRM and it would appear that the boot loader contains a number of “dead loops” that are branched into upon specific faults. The idea being that you hook up JTAG, start up the chip, and where it drops to dead loop will give you an indication of what is going wrong. [list on p3541]

In my specific case, it would appear that something is being loaded from SD, perhaps not correctly, as I think (but can’t yet confirm) that the on-chip loader is incapable of outputting any text messages. That part is done by the initial loader (the MLO file) which is somehow failing. When? Where? How? It might be failing at the part where it tries to load in U-boot? I’m kinda blind without JTAG (I have one, but it’s for a 3.3V board, not the OMAP’s 1.8V). There may be a big latency shift between what appears on the serial port and what the loader is actually doing. Having said that, I’ve been nice to the board since I got it, I even installed RISC OS! As to the specifics of what is going wrong, it’s NMFP – I do software. Somebody else looks after the nuts’n’bolts side.

Just for what it is worth, by the way, this idea that the MLO must be the first file written is either patently false, or the ROM startup code is buggy. The TRM states, and I quote:
The ROM code checks each entry in the root directory until either the booting file is found or the entry is empty (first byte is 00h), or when the end of the root directory is reached. Entries with the ATTR_LONG_NAME attribute (LFN) and first byte at E5h (erased file) are ignored.
[DM3730 TRM p3583]

Joe – the address zero stuff is not relevant. The OMAP has a boot mode where it can read raw data from offset zero or (IIRC) offset 128. These are when using the SD as a raw data device without an actual filesystem on it.
By necessity, the first sector of a FAT formatted device holds the boot record that identifies the disc and information about it. As for writing the MLO first to be at address zero, that simply isn’t possible. I don’t have a program that tells me where on a disc a file is actually stored (as an offset from the start), however if you think about it: there’s the boot record, the FAT, the root directory… This isn’t necessarily all entirely relevant (ie, I think the root directory is somewhere in the middle, or is that ADFS?) but once all of that stuff has been sorted out, then we’re ready to think about actual files!
As far as I can see, the MLO’s criteria is that it needs to be on a valid active FAT disc/partition with no funny stuff, plus as a regular (non LFN) entry in the root directory of the FAT device. This is, assuming, of course, that the datasheet is accurate.

Aaaanyway … Just got back from work so I’m going to have a bath and forget about just about everything for a while… Maybe I’m sulking ‘cos the Euro lottery is something insane like €135,000,000 (!) now. And me? Didn’t win a bean. Oh well, there’s always Tuesday.
<wanders off, dreaming of renting a little place in east Kyoto and haunting all the locations of Haruhi Suzumiya – diversion: follow link to see what heaven looks like in 2D and 3D realities, side by side>

Just out of interest – how does this go under British/European law?

If it was a ‘consumer’ transaction rather than a ‘trade’ one, then what they call it does not remove or reduce the consumers legal rights. But buyers in business transactions have very little rights other than those in the sellers ‘contract’ which includes their terms and conditions. Farnell have always been a Distributor and to buy from them you are supposed to be a business. Though this has now been muddied by them selling the Raspberry Pi to anyone.
The Raspberry Pi foundation are I think a bit week in the business area. The Beta boards they put on ebay they listed as ‘NEW’ but the description said ‘No Warranty’ breaching I think both the Sale of Goods Act and ebay’s rules. I pointed this out to them and they said because it was a ‘development’ unit they were correct. I still disagree. They also said just prior to launch that end user units would be ‘No Warranty’, they responded that yes the Sale of Goods act did apply but there would be no warranty! Quite what they mean I don’t know.
As they are not selling them it is now up to RS & Farnell, I don’t think they are making the same mistake.

Just to confirm on the power supply front, my BBxM PSU is 5V 1A (a huge linear PSU lump) which works. I originally tried it with 5V 300mA and that fell over somewhere in the boot process. At some point I plan to make some USB to 2.1mm DC adaptors, so I can power things from tiny USB phone chargers (SMPSUs) and cut down on the transformer lumps I have lying around.