Re: [PATCH V2 05/13] pm80xx : Support for char device.

[Douglas Gilbert <dgilbert@interlog.com>]

On 2020-03-11 1:08 p.m., Jinpu Wang wrote:
> On Tue, Jan 28, 2020 at 10:43 AM <Deepak.Ukey@microchip.com> wrote:
>>
>>
>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>>
>> On 22/01/2020 08:50, Deepak.Ukey@microchip.com wrote:
>>> -r--r--r-- 1 root root 4096 Jan 21 12:05 running_disparity_error_count
>>> ***
>>> -r--r--r-- 1 root root 4096 Jan 21 12:05 sas_address
>>> lrwxrwxrwx 1 root root    0 Jan 21 11:45 subsystem ->
>>> ../../../../../../../class/sas_phy
>>> -r--r--r-- 1 root root 4096 Jan 21 12:05 target_port_protocols
>>> -rw-r--r-- 1 root root 4096 Jan 21 11:45 uevent
>>>
>>> Maybe the other stuff provided in the patches are useful, I don't know.
>>> But debugfs seems better for that.
>>>
>>>        - 0006-pm80xx-sysfs-attribute-for-number-of-phys
>>>        - 0007-pm80xx-IOCTL-functionality-to-get-phy-status gets things like Programmed Link Rate, Negotiated Link Rate, PHY Identifier
>>>        - 0008-pm80xx-IOCTL-functionality-to-get-phy-error provides other things like Invalid Dword Error Count, Disparity Error Count
>>>        - Thanks for addressing it. We can get this info from /sys/class/sas_phy and /sys/class/sas_port so we will drop these above mentioned three patches from the next              - patch series.
>>>
> 
>>>
>>>        - 0009-pm80xx-IOCTL-functionality-for-GPIO
>>>        - 0013-pm80xx-IOCTL-functionality-for-TWI-device
>>>        - For the above patches management utility passes command specific information to driver through IOCTL structure, which used by driver to frame the command and         - send to FW.  We are using the IOCTL interface for the same. Please let us know your thought.
>>
>> So I specifically questioned the SGPIO patch and why it would have an IOCTL, as this function is supported in kernel libsas/SAS transport code as an SMP function.
>>>   Thank you for your suggestions. We will make use of function supported in libsas.
> 
> So basically you only need IOCTL for GPIO and TWI devices, others can
> implement via libsas interface or from sysfs directly.
> 
> I would like to suggest you do send out other changes without the
> IOCTL parts first, and consider again Is it really needed by the user
> to control GPIO and TWI, and if there is other way to do it?
> 
> Sorry, I don't have a better suggestion!

LSI SAS HBAs (LSI now owned by Broadcom) implement an internal ** SMP
target. It can be seen here:

# ls /dev/bsg
3:0:0:0  3:0:3:0  8:0:0:0  8:0:0:3	     end_device-3:1    expander-3:0
3:0:1:0  4:0:0:0  8:0:0:1  8:0:0:4	     end_device-3:1:0  expander-3:1
3:0:2:0  7:0:0:0  8:0:0:2  end_device-3:0:1  end_device-3:2    sas_host3

It is the last device node: "sas_host3". How do I know it is a SMP target?
Because this works:

# smp_read_gpio /dev/bsg/sas_host3
Read GPIO register response:
   GPIO_CFG[0]:
     version: 0
     GPIO enable: 1
     cfg register count: 2
     gp register count: 1
     supported drive count: 16

When you work out what LSI are doing with this, perhaps you could write
an article about it and make it publicly available.
It is always a good idea to see how your competitors solve problems :-)

Doug Gilbert


** I call it "internal" because it is not seen when doing a SMP DISCOVER
    on a SAS expander to which that SAS HBA is connected.