Boot loader death land on Samsung RV510

[michel]

1. this laptop boots, installs and runs Haiku alpha3 flawlessly. This is the last version of Haiku that it will boot.

2. The Alpha4 and Nightly CDs and USB sticks I have tried to boot on it always result in a BLDL error, regardless of whether they were created from ISO or AnyBoot. These same CDs successfully boot on a really ancient Dell laptop. Most recent nightly I've tried was 46601

Samsung tech specs: ​http://www.samsung.com/uk/consumer/pc-peripherals/notebook-computers/essential/NP-RV510-A0AUK-spec

listdev output: ​http://haikuware.com/details/rv510

The computer has been upgraded to 6 GB RAM, but removing the new 4GB does not solve the problem.

Photo of BLDL output attached.

[waddlesplash]

OK, so R1A3 was around hrev42210 and R1A4 was around hrev44699. The change occured in src/system/boot, so here's the commit log starting at R1A4 and going backwards: ​http://cgit.haiku-os.org/haiku/log/src/system/boot?id=9c50f5eded2b92be18cba2dfd3834d473f245a60

I'll look into this more later.

[waddlesplash]

OK, after filtering out a bunch of commits, here's a possible commit blamelist (there are some PPC commits still in this list, oh well):

​https://gist.github.com/waddlesplash/cec92df5dc9d1a222b40 (The line at the top is the command I used to generate this.)

Michel, can you verify that hrev41861 boots but hrev44624 does not? (Use the closest images that are at ​http://haiku-files.org/haiku/development/ if those exact ones aren't there.)

[michel]

Replying to waddlesplash:

Michel, can you verify that hrev41861 boots but hrev44624 does not? (Use the closest images that are at ​http://haiku-files.org/haiku/development/ if those exact ones aren't there.)

The closest ones I found were hrev41539 - boots hrev44415 - does not boot

FWIW I am also struggling to get USB working on FreeDOS on the same computer - no matter how I disable everything, the drivers keep insisting that there is something already trying to handle it (except that it doesn't). Linux, of course, just overrides everything.

[umccullough]

That's a LOT of changes over the course of several years...

Maybe do some binary searching until you get it down to something manageable?

[waddlesplash]

Replying to michel:

FWIW I am also struggling to get USB working on FreeDOS on the same computer - no matter how I disable everything, the drivers keep insisting that there is something already trying to handle it (except that it doesn't).

I highly doubt this has anything to do with BLDL, but thanks for the info. :)

Replying to umccullough:

That's a LOT of changes over the course of several years... Maybe do some binary searching until you get it down to something manageable?

No, I was thinking using the "halving" method: try the middle and go from there.

So Michel, can you try hrev42458? There's a build of that exact revision.

[umccullough]

Replying to waddlesplash:

No, I was thinking using the "halving" method: try the middle and go from there.

Yes, that's what binary searching is.

[michel]

Replying to waddlesplash:

So Michel, can you try hrev42458? There's a build of that exact revision.

hrev42458 does not boot.

So let me see if I understand correctly. I go halfway down again, and again until I find an image that boots, then back up until I can report that "x boots, but x + 1 does not"?

Damn, that's a lot of bandwidth. Give me a few days, I'll get back to you.

[waddlesplash]

So let me see if I understand correctly. I go halfway down again, and again until I find an image that boots, then back up until I can report that "x boots, but x + 1 does not"?

Yes, that's the state of things. So the next thing to try is hrev41708.

[michel]

OK, that wasn't so hard. hrev51539 boots. hrev42421 (the next one on the list) does not.

[waddlesplash]

OK, down to 29 commits: ​https://gist.github.com/waddlesplash/cec92df5dc9d1a222b40

Looking at commit size, the obvious culprit is probably hrev42092, as it's the largest and most significant change in there. I don't have a Haiku build system set up; can someone who does please build images hrev42091 and hrev42092 and upload them somewhere for Michel to test?

[waddlesplash]

Michel, you also posted here that an Acer laptop of yours didn't work either: ​http://www.haiku-os.org/blog/waddlesplash/2014-01-03_usability_introduction#comment-27887

Does hrev41539 boot on this laptop but not hrev42421?

[umccullough]

To be clear, Haiku R1 Alpha3 was actually hrev42244 - so if it works with the alpha3 images, and not with hrev42421 nightly, then that reduces the range of commits down to something manageable.

There are some commits that may not have been merged into alpha3 from trunk, however, so some care should be taken to verify if any bootloader-related commits missed that branch since hrev41539.

[umccullough]

And some more info: BLDL didn't even exist until hrev42092 (that is actually what that commit was about). Before that commit, there was no boot loader debug land, and it didn't detect page faults in the bootloader - so the bug didn't necessarily get introduced at that time, but rather it was revealed.

Also, while hrev42092 preceded alpha3, it was apparently not merged into the alpha3 branch before it was released.

As such, I'm not sure we can claim that there was a "commit that broke Haiku" - but rather a commit that revealed an otherwise-ignored bug in the bootloader that has yet to be resolved.

If the cause can't be determined based on the screenshot (which is actually a bit unclear at the top right), then my suggestion is to start compiling a list of known-failing machines and get something into a dev's hands to debug directly with.

[michel]

Replying to waddlesplash:

Michel, you also posted here that an Acer laptop of yours didn't work either:

Netbook. Acer Aspire One A110L

Does hrev41539 boot on this laptop but not hrev42421?

My life should be so simple. That one only started doing the BLDL sometime after PM was introduced. Apologies, my post there was imprecise. Anyway, it's just a toy. The Samsung is a production computer.

I'll try to make a better screenshot and/or transcribe the screenshot data for you.

[umccullough]

To be sure we are on the same page, you're booting the Anyboot images from a USB stick right?

[michel]

ISO on CD-RW. My favourite burning sw on my Mac takes one look at Anyboot and says "what dat?"

I have used Anyboot on USB sticks in the past and it makes no difference. I've also used Installer to put a working Haiku installation from my Dell onto a stick. Put it in the Samsung and you get the same BLDL.

[kerobaros]

I am getting the same BLDL when trying to boot Haiku on a Lenovo Ideapad S10e. Tried an anyboot nightly of rev49143 written to a couple different USB sticks. I then booted a VM of r1alpha4.1 in VirtualBox, which worked perfectly, and installed Haiku from there onto a USB stick. Same BLDL.

Any ideas on what I should test next? I'd love to give this netbook some new life.

[knorr]

Im getting the same problem on my old Samsung RV410. Haiku r1beta2. USB-stick tested to work with QEMU (qemu-system-x86_64 /dev/sdc -m 1G)