Re: 12.3:[ZFS pointer corruption] kernel crash exporting FreeBSD src repo

On Mon, Dec 13, 2021 at 08:34:13PM +0100, Peter wrote:

!   I have a filesystem with the FreeBSD src repo clone.
!  
! When I do 
! >  git checkout-index -a --prefix=/new-filesystem/
! in 5 of 6 cases I get a kernel crash.

! 
! Fatal trap 12: page fault while in kernel mode
! cpuid = 19; apic id = 19
! fault virtual address   = 0x410
! fault code              = supervisor read data, page not present
! instruction pointer     = 0x20:0xffffffff808e9815
! stack pointer           = 0x28:0xfffffe00e667bf50
! frame pointer           = 0x28:0xfffffe00e667bff0
! code segment            = base 0x0, limit 0xfffff, type 0x1b
!                         = DPL 0, pres 1, long 1, def32 0, gran 1
! processor eflags        = interrupt enabled, resume, IOPL = 0
! current process         = 18330 (git)
! trap number             = 12

This one did finally turn out to be a known issue in the cam component.

It had haunted me already back in 2016, and made me disassemble
the whole machine, replace it with a newer (and then turning out to
be much worse) machine with lots of other issues, and I couldn't
figure what the problem actually is - it was not obvious when and
why it had appeared and I thought it to be hardware related.

Anyway, with another two changes to newer hardware (this time
hopefully to the better) I thought the issue had disappeared. No, it
hasn't - it just comes in a different disguise: now it creates
bogus data in ZFS (back then the crashes appeared within the cam
component). But the remedy/workaround is the same: changing the
cam options.


cheerio,
PMc