Re: [Scst-devel] Initiators in multiple initiator groups?

[Vladislav B. <vs...@vl...>]

matthew patton, on 04/09/2014 01:08 PM wrote:
>> If the SCST core looks up an initiator name the lookup process stops
>>    when the first initiator match is found.
>>
>> In other words, any WWN
>
>>     that is not a pattern containing wildcards should only occur once in
>>     scst.conf.
>
> may I respectfully submit this logic is flawed? Because it is normal, indeed customary across the rest of the industry to group masking based on LUNs, not initiators; though both are possible. If one has to break it out by individual initiators then the file can become VERY long. Furthermore answering the question "what initiators have have access to this LUN" becomes a lot more involved.

No, there's nothing flawed in this logic. What Bart meant is that it is a bad idea to 
create several wildcards matching to the same initiator, because order of the matching 
is not defined.

> Ideally the parser should handle multiple UNION sets of permissions. and then reduce that into memory data structures that are indexed by LUN and separately indexed by initiator.

This is how it's done.

Vlad