Now trying to slim down the no-crash .config so I can ultimately diff it and hopefully find the option that makes the difference. And then hope that tells me something useful about the root cause of the crashes.
Show this thread
So now I reduced my kernel .config that crashes the Go test down to the point where it boots in 1s and doesn't even support PCI. Build the kernel with GCC 6.4.0, Go crashes. 5.4.0, no crash. But I also have *another* full .config that doesn't crash with 6.4.0. The plot thickens.
-
-
-
Ding ding ding it's CONFIG_OPTIMIZE_INLINING, which points back to the compiler. And now I need to figure out which of these function inlining differences is the culprit. https://mrcn.st/p/vxuE2yH4
Show this thread -
I think I'm going to wrap gcc with a script that enables/disables that inlining mode depending on a given bit of the hash of the output object file name, then build a kernel for each bit, figure out if it crashes, and narrow down the culprit object file that way.
Show this thread
End of conversation
New conversation -
-
-
Is this on bare metal or qemu/KVM/xen/.. or some such thing?
-
qemu, thankfully. I found the original problem on bare metal but thankfully it reproduces on qemu (and the host kernel doesn't seem to matter).
- Show replies
New conversation -
-
-
This is my freaking PC laptop. If only this were at least an embedded system!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Don't get lost. You're going down some deep rabbit holes!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.