Re: [PATCH] libnvdimm/labels: Prevent integer overflow in __nd_label_validate()

[Bryam Vargas]

On 2026-06-21, David Laight wrote:
> The same could be done for nslot - any value above 64k is pretty much
> guaranteed to be garbage

I took that up in v2, but it does not hold against the code, so v3 drops it.

The allocation it was meant to bound -- ndd->data in nd_label_data_init() --
is kvzalloc(config_size), not nslot-derived, so capping nslot shrinks nothing.
And the cap is unsafe: on ND_NSINDEX_INIT the kernel writes
nslot = nvdimm_num_label_slots(ndd) = config_size / label_size, which is above
64K once config_size is past ~8.4MB.  A 64K cap then rejects labels the kernel
itself wrote, so a freshly-formatted large device fails its own next probe.

The (u64) cast in patch 1 already makes the bound exact, so the overflow is
closed without the cap.  v3 keeps the cast; the labelsize-shift UB the review
also turned up is a separate fix, not a stand-in for the cap.

Thanks,
Bryam