bin/150723: bsnmpd(1): add knowledge of "ada" disks (ATA via SCSI CAM)

[Alexander Motin]

Jeremy Chadwick wrote:
> On Thu, Sep 23, 2010 at 07:11:35PM +0300, Shteryana Shopova wrote:
>> I am seeing a similar problem with removable USB disks - libdevinfo(3)
>> reports that the name of the device is umass0 and as a result the disk
>> name da0 does not match the device entry. Thus I also see the same
>> "snmpd[3632]: disk_OS_get_disks: device 'da0' not in device list"
>> messages in the system log. In addition two entries are present for
>> the disk in the hrDeviceTable -
>>
>> hrDeviceDescr[57] = umass0: Kingston DataTraveler 2.0, clas
>> hrDeviceDescr[154] = da0:
>>
>> As I don't have an "ada" disk to test with, could you please show me
>> the output of
>>
>> #devinfo -v
>> #camcontrol devlist -v
>>
>> and also the complete output when walking the hrDeviceTable on your system.
> 
> No problem.  Due to the amount of output, I felt it'd be prudent to put
> them up on the web for easier review.  Here you go:
> 
> http://jdc.parodius.com/freebsd/150723/camcontrol.txt
> http://jdc.parodius.com/freebsd/150723/devinfo.txt
> http://jdc.parodius.com/freebsd/150723/snmpwalk.txt
> 
> It's interesting that the adaX disks don't show up in devinfo.
> 
> On a different system where ahci.ko and SCSI CAM *is not* in use,
> devinfo shows the disks (ad4 and ad6) under their appropriate ataX
> devices.  Sadly I don't have any native SCSI (non-AHCI) systems, so I
> can't see if disks there appear correctly -- but from your experience,
> it sounds like they don't.
> 
> Furthermore, it appears the FreeNAS folks also see this problem (scroll
> down near the bottom):
> 
> http://sourceforge.net/apps/phpbb/freenas/viewtopic.php?f=57&t=1001
> 
> I'm adding mav@ to the CC list since he might have some ideas regarding
> the issue.  It sounds like CAM(4) might not be doing something it
> should (maybe with devclass(9)?  I'm just guessing).

The answer is simple, CAM doesn't uses newbus infrastructure. None of
CAM devices (ada, da, cd, ...) ever appeared in devinfo. There were
plans of respecting CAM refactoring, but it is a huge project, which
nobody was brave enough to start yet.

-- 
Alexander Motin