Re: [PATCH 3/8] scsi: libsas: always unregister the old device if going to discover new

[Jason Yan]

On 2018/5/31 23:09, John Garry wrote:
> On 29/05/2018 03:23, Jason Yan wrote:
> > If we went into sas_rediscover_dev() the attached_sas_addr was already
> > insured not to be zero. So it's unnecessary to check if the
> > attached_sas_addr is zero.
> >
> > And although if the sas address is not changed, we always have to
> > unregister the old device when we are going to register a new one. We
> > cannot just leave the device there and bring up the new.
> >
> > Signed-off-by: Jason Yan <yanaijie@xxxxxxxxxx>
> > CC: chenxiang <chenxiang66@xxxxxxxxxxxxx>
> > CC: John Garry <john.garry@xxxxxxxxxx>
> > CC: Johannes Thumshirn <jthumshirn@xxxxxxx>
> > CC: Ewan Milne <emilne@xxxxxxxxxx>
> > CC: Christoph Hellwig <hch@xxxxxx>
> > CC: Tomas Henzl <thenzl@xxxxxxxxxx>
> > CC: Dan Williams <dan.j.williams@xxxxxxxxx>
> > CC: Hannes Reinecke <hare@xxxxxxxx>
> > ---
> >  drivers/scsi/libsas/sas_expander.c | 13 +++++--------
> >  1 file changed, 5 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/scsi/libsas/sas_expander.c
> > b/drivers/scsi/libsas/sas_expander.c
> > index 8b7114348def..629c580d906b 100644
> > --- a/drivers/scsi/libsas/sas_expander.c
> > +++ b/drivers/scsi/libsas/sas_expander.c
> > @@ -2054,14 +2054,11 @@ static int sas_rediscover_dev(struct
> > domain_device *dev, int phy_id, bool last)
> >          return res;
> >      }
> >
> > -    /* delete the old link */
> > -    if (SAS_ADDR(phy->attached_sas_addr) &&
> > -        SAS_ADDR(sas_addr) != SAS_ADDR(phy->attached_sas_addr)) {
> > -        SAS_DPRINTK("ex %016llx phy 0x%x replace %016llx\n",
> > -                SAS_ADDR(dev->sas_addr), phy_id,
> > -                SAS_ADDR(phy->attached_sas_addr));
> > -        sas_unregister_devs_sas_addr(dev, phy_id, last);
> > -    }
>
> The preceeding checks in code check for no device/comm fail or SATA
> flutter.
>
> If we're rediscovering the device and the SAS address has not changed,
> then why previously still try to discover a new device? I'm guessing
> sas_discover_new() had no affect in this case, since maybe since the PHY
> was already discovered.

When we went here, means it is not flutter, something must change,
either the device type or the phy address. Then we call
sas_discover_new(). And sas_discover_new() sure *have* effect in this
case. Please check sas_discover_new() carefully.

 But that would not make sense since you say "we
> are going to register a new one". Or, if we are always going to register
> a new one, how did we ensure we always unregistered the old device
> previously (when SAS address did not change)?

If SAS address did not change, the device type must changed, otherwise
it will be a "flutter" and won't get here. So if the device type
changed, do we have a reason to keep the device? I don't think so.

> > +    /* we always have to delete the old device when we went here */
> > +    SAS_DPRINTK("ex %016llx phy 0x%x replace %016llx\n",
> > +            SAS_ADDR(dev->sas_addr), phy_id,
> > +            SAS_ADDR(phy->attached_sas_addr));
> > +    sas_unregister_devs_sas_addr(dev, phy_id, last);
> >
> >      return sas_discover_new(dev, phy_id);
> >  }
>
>
>
> .