On 15.04.2021 17:59, Tim Deegan wrote:
> At 12:42 +0200 on 12 Apr (1618231332), Jan Beulich wrote:
>> Some of them have entries with stale comments. Rather than correcting
>> these comments, re-arrange how these arrays get populated, shrinking
>> their sizes at the same time (by omitting trailing NULL entries: Use
>> dedicated element initializers, serving the purpose of what the
>> comments did so far. This then also makes these arrays independent of
>> the actual ordering of the individual SH_type_*.
>>
>> While tightening respective ASSERT()s in hash_{vcpu,domain}_foreach(),
>> also tighten related ones in shadow_hash_{insert,delete}().
>>
>> Signed-off-by: Jan Beulich <jbeul...@suse.com>
>
> Looks good, but please leave the arrays at full size. With the full
> array, a bug could lead to an assertion failure or NULL deref; with
> a short array it could mean following a bogus funtion pointer.
>
> With that change, Acked-by: Tim Deegan <t...@xen.org>
Thanks, but let me ask back about which of the two possble meanings
of "full" you mean: Dimensioned by SH_type_unused, or dimensioned
by SH_type_max_shadow + 1? The former would leave the arrays as they
are now, while the latter would shrink them a little.
Jan
