PCI: pciehp: Flush slot control prior to reset

Since 3461a068661c we don't automatically do a pcie_wait_cmd() for
as part of pcie_write_cmd(), it needs to be called explicitly or
triggered by the next pcie_write_cmd().  However, when we do a
secondary bus reset and we're using pcie_write_cmd() to make sure
that we don't see that bus reset as a hotplug event, we really want
to make sure the update is complete.  Testing on an old Lenovo S20
system, which sets surprise hotplug for some slots, this failure to
flush results in a link down event seen by the hotplug controller
when we issue the bus reset and returns us to the undesireable
behavior of removing and re-adding the device.  Force a flush by
adding an explicit call to pcie_wait_cmd().

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Cc: stable@vger.kernel.org [3.17]
---

 drivers/pci/hotplug/pciehp_hpc.c |    1 +
 1 file changed, 1 insertion(+)


--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Fri, 2014-11-07 at 15:54 -0700, Alex Williamson wrote:
> Since 3461a068661c we don't automatically do a pcie_wait_cmd() for
> as part of pcie_write_cmd(), it needs to be called explicitly or
> triggered by the next pcie_write_cmd().  However, when we do a
> secondary bus reset and we're using pcie_write_cmd() to make sure
> that we don't see that bus reset as a hotplug event, we really want
> to make sure the update is complete.  Testing on an old Lenovo S20
> system, which sets surprise hotplug for some slots, this failure to
> flush results in a link down event seen by the hotplug controller
> when we issue the bus reset and returns us to the undesireable
> behavior of removing and re-adding the device.  Force a flush by
> adding an explicit call to pcie_wait_cmd().
> 
> Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
> Cc: stable@vger.kernel.org [3.17]
> ---
> 
>  drivers/pci/hotplug/pciehp_hpc.c |    1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
> index 0ebf754..e540966 100644
> --- a/drivers/pci/hotplug/pciehp_hpc.c
> +++ b/drivers/pci/hotplug/pciehp_hpc.c
> @@ -660,6 +660,7 @@ int pciehp_reset_slot(struct slot *slot, int probe)
>  		 pci_pcie_cap(ctrl->pcie->port) + PCI_EXP_SLTCTL, 0);
>  	if (pciehp_poll_mode)
>  		del_timer_sync(&ctrl->poll_timer);
> +	pcie_wait_cmd(ctrl);
>  
>  	pci_reset_bridge_secondary_bus(ctrl->pcie->port);
>  
> 

I think we might need a similar change on the other side of the reset,
calling pcie_wait_cmd() explicitly before we re-enable polling.  The
patch above passed about 7000 VM startup/shutdown cycles (up from zero
w/o the patch), but it did eventually get a spurious hotplug.  With the
additional pcie_wait_cmd() I'm thinking about, I've passed 8000 cycles.
TL;DR, I'm probably sending a v2 of this patch.  Thanks,

Alex

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html