[Bug 227116] CURRENT doesn't boot with integer divide fault in uma_startup_count

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Sat Mar 31 14:48:57 UTC 2018


https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=227116

--- Comment #9 from Daniel Kolesa <daniel at octaforge.org> ---
Actually, I managed to obtain a kernel-dbg.txz of the same snapshot and kgdb it
on my server, which runs HardenedBSD 11-STABLE.

This is the output:

(kgdb) list *uma_startup_count+0xe6
0xffffffff80e13296 is in uma_startup_count (/usr/src/sys/vm/uma_core.c:1827).
1822    
1823            bucket_init();
1824    
1825            booted = UMA_STARTUP;
1826    
1827    #ifdef UMA_DEBUG
1828            printf("UMA startup complete.\n");
1829    #endif
1830    }
1831    

The printed lines appear to be wrong. The actual line (1827) though appears
correct, it's consistent with what I guessed from my disassembly.

1827            pages += howmany(zones,
1828                UMA_SLAB_SPACE / roundup2(zsize, UMA_BOOT_ALIGN));

Sadly, without a dump, it won't be possible inspect the memory, and I realized
I cannot obtain a crash dump as the kernel panic happens immediately after it's
loaded, before userspace and therefore before any swap/dump device could be set
up.

-- 
You are receiving this mail because:
You are the assignee for the bug.


More information about the freebsd-bugs mailing list