[Crash-utility] [PATCH v2] struct: Fix handing of percpu symbols (when symbolic argument only)

[Dave Anderson]

----- Original Message -----
> ----- Original Message -----
> > Hi Dave,
> 
> > 
> > Sorry about the delay. I completely forgot about this one!
> > Please note I have only addressed the case when the argument to the
> > 'struct' command is not an address:
> > 
> 
> And lastly, the symbol:cpuspec argument to the struct command is meant
> to be the address of the structure, not a pointer to the address:
> 
>   crash> help struct
> 
>   NAME
>     struct - structure contents
> 
>   SYNOPSIS
>     struct struct_name[.member[,member]][-o][-l offset][-rfuxdp]
>            [address | symbol][:cpuspec] [count | -c count]
> 
>   ... [ cut ] ...
> 
>            symbol  symbolic reference to the address of a structure.
>          :cpuspec  CPU specification for a per-cpu address or symbol:
>                      :             CPU of the currently selected task.
>                      :a[ll]        all CPUs.
>                      :#[-#][,...]  CPU list(s), e.g. "1,3,5", "1-3",
>                                  or "1,3,5-7,10".
>   ...
> 
> So I'm worried that if you wanted to see a structure member that actually
> is a pointer, then the patch is going to dereference it and print bogus
> data.

Hi Aaron,

Given that you are essentially specifying a completely new argument
type to the struct command, I'm going to have to NAK this patch.

To do what you want, honestly I don't really feel that it's asking too much
to make it a two-step process, i.e, get the address first, and then apply it
to the struct command.  

However, if you really feel it's worth it, I suppose you could create a 
new "indirect-pointer-argument-type" for the struct command that explicitly
specifies that it is a pointer to the target address.

Thanks,
  Dave