Re: cgdstrategy: divide fault in supervisor mode

[mlelstv%serpens.de@localhost (Michael van Elst)]

alnsn%yandex.ru@localhost (Alexander Nasonov) writes:

>All mount/umount worked but when I ran reboot, the system trapped here:

>fatal integer divide fault in supervisor mode
>trap type 8 code 0 rip ffffffff808db36f cs 8 rflags 10246 cr2 efd...
>curlwp 0xfffffe81163b4a40 pid 276.1 lowest kstack 0xfffffe8117343...
>kernel: integer divide fault trap, code=0
>Stopped in pid 276.1 (reboot) at        netbsd:cgdstrategy+0x26:

>        if (bp->b_blkno < 0 ||
>            (bp->b_bcount % dg->dg_secsize) != 0 ||

>ffffffff808db36b:       89 c8                   mov    %ecx,%eax
>ffffffff808db36d:       31 d2                   xor    %edx,%edx
>ffffffff808db36f:       f7 77 40                divl   0x40(%rdi)


That would require dg_secsize to be 0 which is difficult to do
because the drivers initialize the value and the common disk_set_info()
function fixes up a zero value.

But maybe the dg pointer is bad? Please have a look at the %rdi
register.

N.B. there are some rare failure paths in getcgd_softc() that would
return a NULL pointer that isn't checked. If the kernel maps zeros
at NULL this could trigger a divide error here.


-- 
-- 
                                Michael van Elst
Internet: mlelstv%serpens.de@localhost
                                "A potential Snark may lurk in every tree."