Re: [PATCH v2 01/23] NFSv4/flexfiles: reject a stripe_unit that does not fit 32 bits

[Benjamin Coddington <ben.coddington@hammerspace.com>]

On 27 Aug 2026, at 15:25, Anna Schumaker wrote:

> Hi Ben,
>
> On Fri, Aug 21, 2026, at 12:29 PM, Benjamin Coddington wrote:
>> ff_layout_alloc_lseg() decodes stripe_unit as the 64-bit value the
>> protocol defines, but every consumer treats it as a u32:
>> nfs4_ff_layout_calc_dss_id() divides by it with do_div(), which casts
>> the divisor, and ff_layout_pg_test() copies it into a u32 first.  A
>> value that does not fit is silently truncated, so the client stripes on
>> a unit the server did not specify -- or divides by zero, if the low 32
>> bits happen to be clear.
>
> I've been thinking on this. What would it take to update the users
> of stripe_unit that you mention above to treat it as a u64 instead of
> a u32? The calls to do_div() could be replaced with div64_u64() for
> example. It just feels a little more robust to me than artificially
> limiting ourselves to a 32-bit value.
>
> Thoughts?

I agree it would be better that way, storage is already u64.  I don't think
the division cost is a big deal.  I'll make that change for v3.

Ben