Re: [PATCH V4 0/2] ndctl: Add daxctl support for the new "famfs" mode of devdax

[John Groves <John@groves.net>]

On 26/02/09 03:13PM, Alison Schofield wrote:
> On Sun, Jan 18, 2026 at 10:36:02PM +0000, John Groves wrote:
> > From: John Groves <john@groves.net>
> > 
> > No change since V2 - re-sending due to technical challenges.
> > 
> > No change since V1 - reposting as V2 to keep this with the related
> > kernel (dax and fuse) patches and libfuse patches.
> > 
> > This short series adds support and tests to daxctl for famfs[1]. The
> > famfs kernel patch series, under the same "compound cover" as this
> > series, adds a new 'fsdev_dax' driver for devdax. When that driver
> > is bound (instead of device_dax), the device is in 'famfs' mode rather
> > than 'devdax' mode.
> > 
> 
> Hi John, 
> 
> I fired this all up and ran it. It got through all but it's last test
> case before failing.
> 
> Three things appended:
> 1) the diff I applied to daxctl-famfs.sh to run the test
> 2) testlog.txt output of the test
> 3) RIP: 0010:is_free_buddy_page+0x39/0x60 kernel log
> 
> 
> 1) Diff I applied to execute the test:
> 
> diff --git a/test/daxctl-famfs.sh b/test/daxctl-famfs.sh
> index 12fbfefa3144..a4e8d87b9762 100755
> --- a/test/daxctl-famfs.sh
> +++ b/test/daxctl-famfs.sh
> @@ -9,6 +9,17 @@ rc=77
>  
>  trap 'cleanup $LINENO' ERR
>  
> +# Use cxl-test module to get the DAX device of the CXL auto region,
> +# which also makes this test NON destructive.
> +#
> +# The $CXL list below is a delay because find_daxdev() was not
> +# finding the DAX region without it.
> +#
> +modprobe -r cxl-test
> +modprobe cxl-test
> +$CXL list
> +
> +
>  daxdev=""
>  original_mode=""

[snip]

Thank you Allison! I'm still working to reproduce the BUG you found (and 
cover this in my jenkins automation), but I think I have the fix already: 
when existing famfs mode (unbinding drivers/dax/fsdev.c), a
fsdev_clear_folio_state() call is needed. I've added that for v8 as a devm
action...

I'm not 100% up to speed on how to use the test infrastructure, and may
ping you with questions later...

Regards,
John