stack smashing detected with 'nvme sanitize-log /dev/nvme0'

[Daniel Wagner]

On Thu, Jul 27, 2023 at 09:37:05AM +0800, Guangwu Zhang wrote:
> Can not reproduce the bug with our test environment.

Thanks for the quick test. It looks like it depends on the SDD invovled, the
ones from MAXIO:

# nvme id-ctrl -H /dev/nvme0
NVME Identify Controller:
vid       : 0x1e4b
ssvid     : 0x1e4b

Both reports have this SDD in common. Other SDDs in the same systems did not
cause the stack smash.

> > >  b8f6446b6853 ("nvme-pci: fix DMA direction of unmapping integrity data")
> >
> > This commit only fixes DMA UNMAP direction for integrity data, but is
> > there integrity data involved for 'nvme sanitize-log /dev/nvme0'?

Don't think so. As I said I just saw this fix and I was wondering if the changes
which came in v5.2 in this area uncoverd a bug. As Christoph suggested, I need
to figure out if device behaves correctly for example with playing with the
IOMMU.