Re: [PATCH] scsi: avoid use of reclaimed reference

[James Bottomley]

On Tue, 2013-11-12 at 18:09 -0800, David Decotigny wrote:
> I was considering the following scenario wherein the "if
> (scsi_device_created(sdev))" test at the end would test garbage at
> best (or unmapped data):

Well, no, the counting isn't right:

>        if (!(sdev = scsi_device_lookup_by_target(starget, 0))) {  // not found
>                sdev = scsi_alloc_sdev(starget, 0, NULL);    // -> ref cnt = 2

1

>                if (scsi_device_get(sdev)) {  // -> ref cnt = 3

2

>                }
>                ...
>        }
>       ...
>                        res = scsi_probe_and_add_lun(starget,    // ->
> ref cnt = 1

No idea what you think here, where were the other puts?  If starget,lun
is sdev, then the count goes to 3 here otherwise it stays at 2 if it
isn't reported in the scan.
      ...
>        scsi_device_put(sdev);  // -> reclaimed

No, it goes to either 2 or 1 here.  If it goes to 1 it's because the
sdev was never probed and thus it remains in the created state.

>        if (scsi_device_created(sdev))  // test on garbage or unmapped data (#PF)

Which means this test passes and it gets garbage collected by
__scsi_remove_device().  Otherwise we exit with refcount 2.

James


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/