On 08/08/2017 06:06 PM, Alan Stern wrote:
> On Tue, 8 Aug 2017, Hannes Reinecke wrote:
>
>> On 08/08/2017 05:25 PM, Alan Stern wrote:
>>> On Tue, 8 Aug 2017, Hannes Reinecke wrote:
>>>
>>>> When checking the model and vendor string we need to use the
>>>> minimum value of either string, otherwise we'll miss out on
>>>> wildcard matches.
>>>> And we should avoid matching anything with zero size as this
>>>> match will be incorrect.
>>>> Without this patch certain Hitachi arrays will not be presenting
>>>> VPD pages correctly.
>>>>
>>>> Fixes: 5e7ff2c ("SCSI: fix new bug in scsi_dev_info_list string matching")
>>>> Signed-off-by: Hannes Reinecke <hare@xxxxxxxx>
>>>
>>> This is very questionable. As described in commit b704f70ce200 ("SCSI:
>>> fix bug in scsi_dev_info_list matching"), there are devices which have
>>> vendor or model strings of length 0, and there are entries in the
>>> global device list like that too. With this patch, such devices and
>>> such entries would never match anything.
>>>
>>> Also, the idea of using the minimum length of either string doesn't
>>> sound right. It would allow a device with string "abc" to match an
>>> entry in the list with string "abcd", which would be wrong.
>>>
>>> Suggestion:
>>>
>>> Allow a device string to match a list entry if the lengths
>>> are equal or the list entry is shorter (this would be a
>>> wildcard match), but not if the list entry is longer. Here
>>> the device string length is after leading and trailing blanks
>>> have been removed.
>>>
>>> As a corollary, allow an empty device string to match an empty
>>> list entry but nothing else.
>>>
>>> Alan Stern
>>>
>> Well, so far we haven't encountered any of those devices.
>
> But what if we do in the future? The code should be correct for
> anything.
>
>> I _have_ validated the patch with the known problematic inquiry strings,
>> namely
>>
>> "HITACHI" "OPEN-V"
>> "Inateck" " "
>> " " "Scanner"
>> "Promise" "STEX"
>>
>> and it gives the correct result for all of these.
>
> How can " " "Scanner" give the correct result? Oh, I see -- the
> patch description and the code comment both say that empty strings
> never match, but in fact they do -- you allow a match if either the
> vendor or model string is empty, but not if they are both empty.
>
Yeah, should've clarified that.
>> If you have any inquiry strings for which this does _not_ produce the
>> correct result please let me know.
>
> What about "HITA" "OPEN-V"? Or "" "OPEN-V"? Both of these will match
> the "HITACHI" "OPEN-" entry, but neither of them should.
>
Hmm. "HITA" should not match the "HITACHI"; the search string is always
the full string, and only the devinfo string should be treated as partial.
Some goes for "".
> What about "ABCD" "Scanner"? This will match the " " "Scanner"
> entry, but it shouldn't.
>
Hmm. Might be. But now we can test this with the scsi_debug module.
I'll be checking and report the results.
Cheers,
Hannes
--
Dr. Hannes Reinecke Teamlead Storage & Networking
hare@xxxxxxx +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
HRB 21284 (AG Nürnberg)
