Opened 4 months ago

Closed 4 months ago

#17857 closed bug (fixed)

"General Protection Error" during boot

Reported by: widespace_ Owned by: nobody
Priority: normal Milestone: R1/beta4
Component: System/Kernel Version: R1/Development
Keywords: Cc:
Blocked By: Blocking:
Platform: x86-64

Description

On hrev56326 trying to boot results in the kernel panicking with a "General Protection Error" without "turning on" any of the boot indicators. It does not appear to generate a syslog.

Attachments (1)

20220804_171051.jpg (2.8 MB ) - added by widespace_ 4 months ago.
Image of KDL output

Change History (11)

by widespace_, 4 months ago

Attachment: 20220804_171051.jpg added

Image of KDL output

comment:1 by widespace_, 4 months ago

I just checked and hrev56325 boots fine so the problematic change is between these two revisions.

comment:2 by korli, 4 months ago

Thanks. How much RAM on this computer? https://github.com/haiku/haiku/blob/e83979afa4cf62fd7980a863da80be11644feabe/src/system/kernel/fs/EntryCache.cpp#L87 looks like it checks against 1MB instead of 1GB.

comment:3 by widespace_, 4 months ago

The computer has 32GB of RAM.

comment:4 by waddlesplash, 4 months ago

shaka reported something similar via IRC: https://matrix.org/_matrix/media/r0/download/matrix.org/cptKGazpyXmnOZZLmeHyiItT/ima_9f0a5f2.jpeg

korli, how does that check 1MB? 1024^3 is the number of bytes in 1GB. Anyway 32GB of RAM is obviously more than enough for this. I'm not sure how we wound up with a crash here, in the worst case we'd get a NULL dereference but this isn't even that.

comment:5 by waddlesplash, 4 months ago

Seems more likely that the refactoring changes somehow caused this, really, not the last commit to have larger cache sizes.

comment:6 by waddlesplash, 4 months ago

Aha, it's the destructor being called. I need to add a NULL set.

comment:7 by waddlesplash, 4 months ago

Should be fixed in hrev56330, please test.

comment:8 by korli, 4 months ago

korli, how does that check 1MB?

Right, read too much strange code. The coding style didn't help either, "1024 * 1024 * 1024" would be more readable...

in reply to:  7 comment:9 by widespace_, 4 months ago

Replying to waddlesplash:

Should be fixed in hrev56330, please test.

It appears to be fixed, hrev56332 boots fine.

comment:10 by korli, 4 months ago

Milestone: UnscheduledR1/beta4
Resolution: fixed
Status: newclosed
Note: See TracTickets for help on using tickets.