> -----Original Message-----
> From: Jason Yan <yanaijie@xxxxxxxxxx>
> Sent: Saturday, May 11, 2024 11:41 AM
> To: Li, Eric (Honggang) <Eric.H.Li@xxxxxxxx>; John Garry <john.g.garry@xxxxxxxxxx>;
> james.bottomley@xxxxxxxxxxxxxxxxxxxxx; Martin K . Petersen <martin.petersen@xxxxxxxxxx>
> Cc: linux-scsi@xxxxxxxxxxxxxxx
> Subject: Re: Issue in sas_ex_discover_dev() for multiple level of SAS expanders in a domain
>
>
> [EXTERNAL EMAIL]
>
> On 2024/5/8 16:29, Li, Eric (Honggang) wrote:
> >> -----Original Message-----
> >> From: John Garry <john.g.garry@xxxxxxxxxx>
> >> Sent: Wednesday, May 8, 2024 3:48 PM
> >> To: Li, Eric (Honggang) <Eric.H.Li@xxxxxxxx>; Jason Yan
> >> <yanaijie@xxxxxxxxxx>; james.bottomley@xxxxxxxxxxxxxxxxxxxxx; Martin
> >> K . Petersen <martin.petersen@xxxxxxxxxx>
> >> Cc: linux-scsi@xxxxxxxxxxxxxxx
> >> Subject: Re: Issue in sas_ex_discover_dev() for multiple level of SAS
> >> expanders in a domain
> >>
> >>
> >> [EXTERNAL EMAIL]
> >>
> >> On 08/05/2024 01:59, Li, Eric (Honggang) wrote:
> >>>>> Call to sas_ex_join_wide_port() makes the rest PHYs associated
> >>>>> with that existing port
> >>>> (making it become wideport) and set up sysfs between the PHY and
> >>>> port. > Set PHY_STATE_DISCOVERED would make the rest PHYs not being
> >>>> scanned/discovered again (as this wide port is already scanned).
> >>>>
> >>>> If you can just confirm that re-adding the code to set phy_state =
> >>>> DISCOVERED is good enough to see the SAS disks again, then this can
> >>>> be further discussed. >>
> >>> OK. I will work on that and keep you updated.
> >>
> >> I expect a flow like this for scanning of the downstream expander:
> >>
> >> sas_discover_new(struct domain_device *dev [upstream expander], int
> >> phy_id_a) -> sas_ex_discover_devices(single = -1) ->
> >> sas_ex_discover_dev(phy_id_b) for each phy in @dev non-vacant and
> >> non-discovered -> sas_ex_discover_expander( [downstream expander])
> >> for first phy scanned which belongs to downstream expander.
> >>
> >> And following that we have continue to scan phys in
> >> sas_ex_discover_devices(single = -1) ->
> >> sas_ex_discover_dev(phy_id_b) ->
> >> sas_ex_join_wide_port() -> for each non-vacant and non-discovered
> >> phy in phy_id_b which matches that downstream expander.
> >>
> >> Can you see why this does not actually work/occur?
> >>
> >
> > before calling sas_ex_join_wide_port(), sas_dev_present_in_domain()
> > finds the attached_sas_address of PHY (phy_id_b) is already in the domain of that root
> port, and then disable all PHYs to that downstream expander (in sas_ex_disable_port(dev,
> attached_sas_addr)) Therefore, I think we need to switch the order of function call to
> sas_ex_join_wide_port() and sas_dev_present_in_domain().
>
> Hi Eric,
>
> Can you test the following patch to see if it works?
>
> Author: Jason Yan <yanaijie@xxxxxxxxxx>
> Date: Sat May 11 11:33:35 2024 +0800
>
> scsi: libsas: Skip disable PHYs which can form wide ports
>
> Signed-off-by: Jason Yan <yanaijie@xxxxxxxxxx>
>
> diff --git a/drivers/scsi/libsas/sas_expander.c
> b/drivers/scsi/libsas/sas_expander.c
> index f6e6db8b8aba..39a86857bc52 100644
> --- a/drivers/scsi/libsas/sas_expander.c
> +++ b/drivers/scsi/libsas/sas_expander.c
> @@ -618,15 +618,17 @@ static void sas_ex_disable_port(struct domain_device *dev, u8
> *sas_addr)
> }
> }
>
> -static int sas_dev_present_in_domain(struct asd_sas_port *port,
> +static int sas_dev_present_in_domain(struct domain_device *dev,
> u8 *sas_addr)
> {
> - struct domain_device *dev;
> + struct domain_device *tmp;
>
> if (SAS_ADDR(port->sas_addr) == SAS_ADDR(sas_addr))
> return 1;
> - list_for_each_entry(dev, &port->dev_list, dev_list_node) {
> - if (SAS_ADDR(dev->sas_addr) == SAS_ADDR(sas_addr))
> + list_for_each_entry(tmp, &dev->port->dev_list, dev_list_node) {
> + if (tmp->parent == dev)
> + continue;
> + if (SAS_ADDR(tmp->sas_addr) == SAS_ADDR(sas_addr))
> return 1;
> }
> return 0;
> @@ -973,7 +975,7 @@ static int sas_ex_discover_dev(struct domain_device *dev, int phy_id)
> return 0;
> }
>
> - if (sas_dev_present_in_domain(dev->port, ex_phy->attached_sas_addr))
> + if (sas_dev_present_in_domain(dev, ex_phy->attached_sas_addr))
> sas_ex_disable_port(dev, ex_phy->attached_sas_addr);
>
> if (ex_phy->attached_dev_type == SAS_PHY_UNUSED) {
>
>
>
I am still waiting for feedback from our test team.
>From functionality, I think it should work as it skips sas_dev_present_in_domain check for the rest PHYs in the wide port.
But from logic, I think it makes sense that checking wide port is prior to checking sas address duplication_in_domain.
Just my two cents.
