Re: [patch,v2 05/10] sd: use alloc_disk_node

[Bart Van Assche <bvanassche@acm.org>]

On 11/05/12 15:12, Jeff Moyer wrote:
> Bart Van Assche <bvanassche@acm.org> writes:
>
>> On 11/02/12 22:45, Jeff Moyer wrote:
>>> Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
>>> ---
>>>    drivers/scsi/sd.c |    2 +-
>>>    1 files changed, 1 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
>>> index 12f6fdf..8deb915 100644
>>> --- a/drivers/scsi/sd.c
>>> +++ b/drivers/scsi/sd.c
>>> @@ -2714,7 +2714,7 @@ static int sd_probe(struct device *dev)
>>>    	if (!sdkp)
>>>    		goto out;
>>>
>>> -	gd = alloc_disk(SD_MINORS);
>>> +	gd = alloc_disk_node(SD_MINORS, dev_to_node(dev));
>>>    	if (!gd)
>>>    		goto out_free;
>>
>> shost->numa_node can be another NUMA node than dev_to_node(dev). Have
>> you considered using shost->numa_node here ?
>
> It can?  How?

E.g. if the LLD allows the user to specify the value of numa_node and 
passes that value to scsi_host_alloc_node() (see also 
http://lkml.org/lkml/2012/10/23/477 for further information).

> Just so I'm clear, you're suggesting I use the scsi_device's host
> pointer to get to the Scsi_Host, and that *will* be filled in that this
> point, right?

As far as I can see the sdev->host pointer is set in scsi_alloc_sdev() 
and that happens before sd_probe() is invoked.

Bart.