Re: [PATCH] target: generate correct string identifiers for PR OUT transport IDs

["Maurizio Lombardi" <mlombard@bsdbackstore.eu>]

On Fri Jun 27, 2025 at 7:40 PM CEST, Dmitry Bogdanov wrote:
> On Fri, Jun 27, 2025 at 04:52:29PM +0200, Maurizio Lombardi wrote:
>> 
>> Fix target_parse_pr_out_transport_id() to
>> return a dynamically allocated string representing the
>> transport ID in a standardized, human-readable format
>> (e.g., naa.xxxxxxxx...) for various SCSI protocol types
>> (SAS, FCP, SRP, SBP).
>
> There is no a single standard how to represent TransportId. There are
> several standards for the same port address. But TransportId itself is
> the single standard to represent different port names in binary format.
> So, the most correct solution would be have TransportId for ACLs and
> to match them.

That's right. I will avoid using the word "standardized", which is not
correct in this context.

>
>> 
>> Previously, the function returned a pointer to the raw binary buffer.
>> The caller would then compared it to a human-readable string,
>> which obviously always failed.
>>
>> Now, the function constructs a string using kasprintf()
>> based on the protocol's offset and format:
>> 
>> * SAS, FCP, SBP: 64-bit identifier
>> * SRP: 128-bit identifier
>> * iSCSI: duplicates the iqn string to match the new allocation behavior
>
> The caller compares it with a specifically prepared string buffer by
> fabric module in its own way. And they are not corresponded any
> standard and, that especially important, to your code.
> Mostly that string contains only hex-representation without any prefixes.
> You may grep target_setup_session to check which transport
> generates what a string.

Thanks for the hint. I was misled by rtslib which enforces those
prefixes. I will fix it and submit a V2.

Maurizio