net_server always crashes booting the A4 ISO loading the ipro1000 driver

[stippi]

Hardware is a Lenovo T60. I've burned the ISO image in Ubuntu, and let the validation after burning run. On the same laptop, I run a pretty outdated revision of Haiku with no problems.

Booting the alpha4 ISO, two out of two times, the net_server crashed. I took a picture of the KDL output. From the stack trace, it looks as though net_server configures initial devices (NetServer::_ConfigureDevices()) which eventually opens the ipro1000 driver. During driver setup, it tries to figure out the MSI count (pci_msi_count()) which enters the PCI bus manager and crashes in FindDevice(). Sounds a bit as though the code that would add a device re-enters itself and FindDevice() crashes since the device that was going to be added in the first place is looking itself up in some list. But that is just gut feeling from looking at the stack crawl only, not actually at the code.

[stippi]

Picture of the KDL output after the crash

[diver]

I think this is because of a new ​ACPI version which has been committed to A4 branch in hrevr1alpha4-44682.

​http://www.haiku-files.org/haiku/development/haiku-hrevr1alpha4-44678-releasecandidate-cd.tar.xz
confirmed to work ok.

[tqh]

I'm sorry, but please don't always blame ACPI. It seems to always be the scapegoat until investigation show the reason. I think there are a lot more plausible explanations.

[diver]

Disabling local apic workarounds the problem.

[stippi]

​http://www.haiku-files.org/haiku/development/haiku-hrevr1alpha4-44678-releasecandidate-cd.tar.xz

Hm, I thought I did test the release candidate. I tested the download from the link that Matt sent me. I didn't see any commits to the alpha branch in this regard. (?)

[anevilyak]

Possibly related to #9128 ?

[stippi]

Replying to anevilyak:

Possibly related to #9128 ?

I wouldn't see how. Here is crashes before app_server even loads. It's not a hang and crashes to KDL with the same sc everytime.

[jackburton]

Happens under XenServer too, with the A4 ISO.
XenServer emulates a RealTek rtl8139 (a revision/model supported by our rtl81xx driver).

[luroh]

Yep, seeing the same on my marvell_yukon machine and the A4 iso.

[umccullough]

Unfortunately, I'm also seeing it with my RTL8111/8168B NIC here. I tried both ISO and Anyboot images burned to CD, with the same result.

Going to build my own image to double-check and test a few other things to see if I can narrow down a cause.

[diver]

debug syslog

[mmlr]

This is really the same as #8684 which has an explanation for why it's happening. So noone actually tested a CD boot before the release?

[diver]

That's weird, because one user tested haiku-hrevr1alpha4-44678-releasecandidate-cd.tar.xz and it didn't crash.

[stippi]

Maybe there is a driver that's used when booting from CD which releases the PCI device manager handle one time too many? And this driver is only used on some systems when booting from CD? Like SATA versus ATA?