On Wed, Jun 29, 2016 at 8:31 AM, Johannes Thumshirn <jthumshirn(a)suse.de> wrote:
On Tue, Jun 28, 2016 at 07:24:17PM -0700, Dan Williams wrote:
> On Tue, Jun 28, 2016 at 7:17 PM, yizhan <yizhan(a)redhat.com> wrote:
> > On 06/29/2016 12:26 AM, Dan Williams wrote:
> >> Do you have CONFIG_DEV_DAX_PMEM=m?
> > I also met this error with CONFIG_DEV_DAX_PMEM=m
> > Detailed log I have post at the last comment at
> > https://github.com/pmem/ndctl/issues/3
> Can you re-run the test with kernel debugging enabled and piping the
> test messages to the kernel buffer.
> Add the following to the kernel command line:
> libnvdimm.dyndbg nfit.dyndbg nd_pmem.dyndbg nd_btt.dyndbg
> dax_pmem.dyndbg dax.dyndbg
> Specify add --enable-test when configuring ndctl and then run:
> ndctl test 2>/dev/kmsg
> ...and send the log?
Here's the log. I did a dmesg -C to clear off all old stuff beforehand.
Thanks for looking into it.
[ 2738.166571] nd dax13.0: align_store: result: 0 wrote: 2097152
[ 2738.166691] dax_pmem dax13.0: bad offset: 0x100000 dax disabled
[ 2738.166694] ndbus1: dax_pmem.probe(dax13.0) = -6
[ 2738.166802] libndctl: ndctl_dax_enable: dax13.0: failed to enable
[ 2738.166820] dax13.0: expected dax enable success, region13 read-write
[ 2738.166825] namespace13.0: failed to create dax
[ 2738.166828] namespace13.0: failed to configure namespace
[ 2738.166835] ndctl-test0 failed: -6
I think there is variability in the alignment of the DMA_CMA start
address from system-to-system which is why the test is passing for me.
The recent changes to the dax unit tests   prevent this failure
by moving higher order dax / pfn testing to a memmap=ss!nn
configuration instead of nfit_test.
Specifically I think we are failing because the configuration needs
the allocation to be "2097152" (2MB) aligned, but it randomly ends up
on a 1MB alignment.