[Mesa-dev] [PATCH 2/3] util/disk_cache: hash pointer size and gpu name into cache keys

[Nicolai Hähnle]

On 21.03.2017 11:23, Timothy Arceri wrote:
>
>
> On 21/03/17 21:01, Nicolai Hähnle wrote:
>> On 21.03.2017 10:22, Timothy Arceri wrote:
>>>
>>>
>>> On 21/03/17 19:25, Nicolai Hähnle wrote:
>>>> On 21.03.2017 06:14, Timothy Arceri wrote:
>>>>>  void
>>>>>  disk_cache_compute_key(struct disk_cache *cache, const void *data,
>>>>> size_t size,
>>>>>                         cache_key key)
>>>>>  {
>>>>>     struct mesa_sha1 ctx;
>>>>>
>>>>>     _mesa_sha1_init(&ctx);
>>>>>     _mesa_sha1_update(&ctx, cache->timestamp, cache->timestamp_size);
>>>>> +   _mesa_sha1_update(&ctx, cache->gpu_name, cache->gpu_name_size);
>>>>
>>>> For consistency, please add the same + 1 as in the previous patch.
>>>
>>>
>>> I was trying to appease the people complaining about weakening the hash
>>> by salting it with this driver identifying data.
>>
>> Huh, I guess I missed that discussion due to vacation and massive
>> backlog, but I don't see why that would weaken anything. Do you have a
>> link?
>
> I don't think it would make much difference either but I'm sick of all
> the "you should do it this way" shader cache discussions.
>
> https://lists.freedesktop.org/archives/mesa-dev/2017-March/148029.html

I skimmed this mail just now.

The disk cache doesn't use salts in any way as far as I'm aware (calling 
the addition of driver timestamps a "salt" is a fairly clear 
misunderstanding of what a salt is), and even if it were, the "analysis" 
of salts in that email makes no sense.

I think it's safe to conclude that that mail is full of bullshit.

Nicolai