Re: [PATCH 06/13] qla4xxx: Prevent other port reinitialization during remove_adapter

[Mike Christie <michaelc@cs.wisc.edu>]

On 03/15/2011 04:21 AM, Vikas Chaudhary wrote:
>> -----Original Message-----
>> From: linux-scsi-owner@vger.kernel.org [mailto:linux-scsi-
>> owner@vger.kernel.org] On Behalf Of Mike Christie
>> Sent: Tuesday, March 15, 2011 3:37 AM
>> To: Vikas Chaudhary
>> Cc: James Bottomley; linux-scsi@vger.kernel.org; Ravi Anand; Lalit
>> Chandivade; Karen Higgins
>> Subject: Re: [PATCH 06/13] qla4xxx: Prevent other port reinitialization
>> during remove_adapter
>>
>> On 03/14/2011 12:24 AM, Vikas Chaudhary wrote:
>>> index 6068f80..59c5552 100644
>>> --- a/drivers/scsi/qla4xxx/ql4_os.c
>>> +++ b/drivers/scsi/qla4xxx/ql4_os.c
>>> @@ -27,6 +27,11 @@ static char qla4xxx_version_str[40];
>>>    static struct kmem_cache *srb_cachep;
>>>
>>>    /*
>>> + * List of host adapters
>>> + */
>>> +static struct klist qla4xxx_hostlist;
>>> +
>>
>>
>>
>>>
>>>    /**
>>> + * qla4xxx_prevent_other_port_reinit - Mark the other ISP-4xxx port to
>> indicate
>>> + * that the driver is being removed, so that the other port will not
>>> + * re-initialize while in the process of removing the ha due to driver
>> unload
>>> + * or hba hotplug.
>>
>> For docbook comments you use a one line description on the first line,
>> then add more informative stuff after the params.
>>
>
> Ok, I will be sending updated patch in next email.
>
>>
>>> + * @ha: pointer to adapter structure
>>> + **/
>>> +static void qla4xxx_prevent_other_port_reinit(struct scsi_qla_host *ha)
>>> +{
>>> +       struct scsi_qla_host *ha_listp;
>>> +       struct klist_iter i;
>>> +       struct klist_node *n;
>>> +
>>> +       klist_iter_init(&qla4xxx_hostlist,&i);
>>> +       while ((n = klist_next(&i)) != NULL) {
>>> +               ha_listp = container_of(n, struct scsi_qla_host, node);
>>> +               if (ha == ha_listp)
>>> +                       continue;
>>> +
>>> +               if ((pci_domain_nr(ha->pdev->bus) ==
>>> +                   pci_domain_nr(ha_listp->pdev->bus))&&
>>> +                   (ha->pdev->bus->number ==
>>> +                   ha_listp->pdev->bus->number)&&
>>> +                   (PCI_SLOT(ha->pdev->devfn) ==
>>> +                   PCI_SLOT(ha_listp->pdev->devfn))) {
>>> +
>>
>> Is there a qla4xxx specific reason why you need to prevent other ports
>>from re-initing when removing a port? Does the hardware cause the other
>> ports to be disrupted and that causes them to run the re-init code?
>
> Yes, its ISP40XX specific. While unloading driver we make call to function
> "qla4xxx_hw_reset" from "qla4xxx_free_adapter". When driver do hw_reset
> for one port, hardware will generate interrupt for other ports to inform
> reset is about to happen. After this interrupt driver will schedule reset
> for other port by setting dpc flag "DPC_RESET_HA_INTR" in function
> "qla4xxx_intr_handler" this will cause re-initialization of other port
> after reset, but while unloading driver we don't want to schedule reset
> for the other port as other port does not need re-initialization".
>

Ok.

I think that the driver should not be needing to manage a list of hosts 
though. We have the driver model related lists already. There is also 
the pci bus stuff.  So I think you should be using one of those.