Regression : broadcom570x blocks on boot

[Barrett]

From my last update i have my system which is not booting at all. It sometimes block on rocket, but in majority of cases it show just a black screen. Anyway, it's booting with safe mode activated (without video support).

I tested various revs, and i can say the regression was between :

45c56fdb6b6ac5a882d597d3337e860cd726a11a

and

4377c66f87368091b4d7ceed5aa08c2bbdb6b134

The first is a working rev, the second is non-working.

My pc is an Acer 5738Z. Unfortunately, is not created any syslog to attach.

[Barrett]

An update...after opening the ticket i rebooted, and i become crazy because it wasn't booting. I've then tested those revs :

c9a56147d2de5fcdd44da46e02c22770070cdd5c

e3ca61af3262f112788c14a93129736f543a83af

I rebooted and both were not working. So i launched linux, and then before to log-in rebooted to Haiku, it booted.

I don't know what exactly is, but basically after the image is written, i have to reboot two times to get it booting.

[Barrett]

Ok, it's going to be a lot more strange. Sometimes boot, sometimes not. I started the pc, Haiku booted, then rebooted and it blocked at the rocket icon. If someone have some further test just let me know.

[umccullough]

Is it just an issue of warm vs cold boot? Sometimes the hardware gets left in a bad state if the driver doesn't clean up properly.

Also, I know I've run into situations where booting into Windows or Linux seems to configure the hardware in such a way that it will reboot into Haiku without problem...

[korli]

If you could boot in safe mode and start servers one by one, this would help to see which driver is misbehaving. Why do you think it's ACPI related?

[Barrett]

@umccullough

Really...don't have idea of what it's. Sometimes Haiku boot correctly, but at the next reboot it block, sometimes i have to pass through linux, sometimes i can pray god for a day but it won't boot. What i can say, seems that in september end revs there's not this problem.

@korli

At this point i'm not sure anymore it's acpi related, I thinked of it, because revs before the ACPI changes apparently booted correctly, but it looks like i was deceived by this strange behaviour.

I'll do those tests, trying also to blacklist the acpi driver, and let you know...

[Barrett]

I tested a bit more, and blacklisting the acpi_battery driver, solve the problem.

[Barrett]

I'm full of false positives, very sorry about that, also without the drivers it continue to boot intermittently, with apparently no sense at all.

I tried to start the servers one by one with haiku booted in safe mode, but i'm unable to identify which component is causing the problem. Might it be some conflict between drivers?

[Barrett]

BCM5x driver debug screenshot

[Barrett]

I've finally found the source of the problem enabling on-screen debug. The BCM5x block up, unfortunately i don't know why this happens intermittently.

[korli]

Could you eventually find the first non working revision? I suppose that, when you disable the driver broadcom570x, the problem doesn't occur anymore?

[Barrett]

Yes disabling the driver, the problem disappear. I was far from the pc with this hardware, i'll try to find the non working rev in the next weeks.

[Barrett]

(In #12222) The same here with a BCM5784M, it seems a problem of the BCM* drivers or the FreeBSD compatibility layer. Duplicate of #10310.

[waddlesplash]

Still an issue after the FreeBSD 11 upgrade?

[waddlesplash]

Fixed in hrev52204; the hang location is even identical to #10673.

[nielx]

Assign tickets with status=closed and resolution=fixed within the R1/beta2 development window to the R1/beta2 Milestone