Differences between iscsi and gscsi

[Paul Nasrat]

On 16 March 2010 15:10, Roman Rakus <rrakus at redhat.com> wrote:
> On 02/09/2010 05:58 AM, Tony Breeds wrote:
>>
>>
>>>
>>> Previous patch has a typo. Use this new one.
>>> RR
>>>
>>
>>
>>>
>>> diff -up yaboot-1.3.14/second/prom.c.iscsi yaboot-1.3.14/second/prom.c
>>> --- yaboot-1.3.14/second/prom.c.iscsi   2010-02-05 13:12:56.000000000
>>> +0100
>>> +++ yaboot-1.3.14/second/prom.c 2010-02-05 13:13:24.000000000 +0100
>>> @@ -196,7 +196,7 @@ prom_get_devtype (char *device)
>>>       char       tmp[64];
>>>
>>>       if (strstr(device, TOK_ISCSI))
>>> -         device = strcpy(tmp, "/vdevice/gscsi/disk");
>>> +         return FILE_DEVICE_ISCSI;
>>>
>>>       /* Find OF device phandle */
>>>       dev = prom_finddevice(device);
>>>
>>
>> Seems good to me.
>>
>> Yours Tony
>>
>
> And it only looks to be good and actually isn't. The problems are in places,
> where we call prom_get_devtype(), but we don't check if it returns
> FILE_DEVICE_ISCSI (only _BLOCK and _NET). Following files are affected:
> second/file.c
> second/partition.c
> second/yaboot.c
> Is it possible to just say that _BLOCK is generally the same as _ISCSI (in
> meaning how we will handle it)? Is iSCSI device handled in open firmware
> similarly as block device?
> A bits of my investigation:
> second/file.c:open_file() it looks like _BLOCK open works well when we get
> _ISCSI
> second/partition.c:get_part_type() maybe same as above
> second/yaboot.c: Here are things a bit complicated, but every time we
> compare with _NET, so if we say we will handle _ISCSI same as _BLOCK,
> nothing will change here.
>
> So, the mail question is: Is iSCSI device handled similarly as block device?

Problem is it's implementation defined and not via 1275, IIRC the
power.org specs have some documentation on the iscsi behaviour but
it's not standardized.

Paul