Re: [PATCH 18/26] target/iscsi: Allocate session IDs from an IDA

[Mike Christie]

On 07/31/2018 01:15 PM, Matthew Wilcox wrote:
> On Mon, Jul 30, 2018 at 10:03:21PM -0400, Martin K. Petersen wrote:
>>
>> Mike,
>>
>>>> How about I take it through my tree to minimise the amount of rebasing
>>>> I'll need to do?  I'm already dependent on the nvdimm tree and I'd rather
>>>> not depend on the scsi tree as well.  I'll queue it up in front of my
>>>> IDA change to maximise its backportability.
>>>
>>> Ccing Martin, because he has been handling target patches.
>>
>> That's fine with me.
>>
>> Acked-by: Martin K. Petersen <martin.petersen@xxxxxxxxxx>
> 
> Thanks, Martin.  Mike, can I have your Signed-off-by: on the original patch?

Yes,

Signed-off-by: Mike Christie <mchristi@xxxxxxxxxx>


> What Fixes: line should this have?  As far as I can tell, the bug is present
> all the way back to its introduction, so:
> 
> Fixes: e48354ce078c ("iscsi-target: Add iSCSI fabric support for target v4.1")
> 
> yes?

The session_index bug was in that patch, but I think it correctly kfreed
the session in __iscsi_target_login_thread.


I think it was this one that added the early kfree bug:

commit 0957627a99604f379d35831897a2aa15272528fc
Author: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx>
Date:   Fri Nov 4 11:36:38 2011 -0700

    iscsi-target: Fix sess allocation leak in iscsi_login_zero_tsih_s1

Dan's tools probably just did not catch the iscsi_login_zero_tsih_s1 ->
iscsi_login_set_conn_values call that sets conn->sess then later
kfree(conn->sess) call in the error handling in __iscsi_target_login_thread.