opensolaris has a fault management architecture. smartd has the detection capability to report faults into this architecture.

There is a simple API for reporting faults. ​http://www.opensolaris.org/os/community/fm/

detailed documentation for this is available: ​http://docs.sun.com/app/docs/doc/819-3196/gemfu?l=en&a=view

smartd has knowledge of the state of all the hard drives in the system, and as such it is a perfect abstraction layer between the hardware and services or software that would make use of such information.

Unfortunately, although it is a daemon, it currently behaves in a cron-like fashion, by querying the HDD at specified intervals and optionally sending emails when a fault is detected.

My proposal is to add a Unix socket interface to smartd so that the last-checked HDD state and drive inventory can be queried from external software. I am proposing a simple query interface: 1-line command, 1-line response.

What smartd should provide on the socket:

• a hello, I am running message; smartd version string included
• a command to shut down
• a command to reload the configs
• a command to provide internal daemon statistics, like daemon runtime, etc

For the drive data that should also be:

• a command to enumerate the drives monitored by smartd
• a command to provide SMART variables for the specified drive (including SCSI/ATA specifics)

Optionally, these might make sense too:

• a command to get/change device settings (-g -s in smartctl)
• a command to deal with self-tests, SMART-GPT logs

When scanning for devices, smartd uses the FreeBSD cam subsystem to iterate through all of the devices on the system. When it encounters a SATA port multipler (e.g., /dev/pmp0) it sends commands to that device that eventually timeout. This delays boot by several minutes on a large system.

Running "smartd -q onechceck" produces errors like this for each port multiplier encountered,

Device: /dev/pmp8, opened (pass53:mvsch10:0:15:0): ATA_IDENTIFY. ACB: ec 00 00 00 00 40 00 00 00 00 01 00 (pass53:mvsch10:0:15:0): CAM status: ATA Status Error (pass53:mvsch10:0:15:0): ATA status: 51 (DRDY SERV ERR), error: 04 (ABRT ) (pass53:mvsch10:0:15:0): RES: 51 04 00 00 00 00 00 00 00 00 00 (pass53:mvsch10:0:15:0): ATAPI_IDENTIFY. ACB: a1 00 00 00 00 40 00 00 00 00 01 00 (pass53:mvsch10:0:15:0): CAM status: Command timeout Device: /dev/pmp8, not ATA, no IDENTIFY DEVICE Structure Unable to register ATA device /dev/pmp8

These devices should be avoided on FreeBSD. The attached patch avoids them when running "smartd -d onecheck" but it is not clear if that is sufficient or even the right place to make such a change.