On Mon, Oct 1, 2012 at 12:14 PM, Dan Williams <djbw@xxxxxx> wrote: > On Mon, Oct 1, 2012 at 11:22 AM, Gwendal Grignou <gwendal@xxxxxxxxxx> wrote: >> This set of patches improve ATA transport classes integration with SCSI >> objects. >> >> Before [2.6.x] >> >> Ata and scsi transport class where separated: >> `--0000:09:00.0 >> | `--ata1 >> | | `--port_port >> | | `--link1 >> | | | `--dev1.0 >> | | | `--dev1.1 >> | `--ata2 >> | ... >> | `--host0 >> | | `--scsi_host >> | | | `--host0 >> | | `--target0:0:0 >> | | | `--0:0:0:0 >> | | | | `--block >> | | | | | `--sda >> | | | | | | `--sda1 >> >> In 3.2, Lin - in commit 9a6d6a2ddabbd32c07f6a38b659e5f3db319fa5a - addressed >> the issue of linking the ata port with the scsi host object by placing the >> scsi_host object under ata port objects. >> >> However to be more consistent with other transport, this patch does the opposite: >> >> For instance, with SAS transport, We have >> `--0000:0b:00.0 >> | `--host6 >> | | `--phy-6:0 >> | | `--phy-6:1 >> ... >> | | `--port-6:0 >> | | | `--end_device-6:0 >> | | | | `--sas_device >> | | | | | `--end_device-6:0 >> | | | | `--sas_end_device >> | | | | | `--end_device-6:0 >> | | | | `--target6:0:0 >> | | | | | `--6:0:0:0 >> | | | | | | `--block >> | | | | | | | `--sdb >> ... >> | | `--port-6:1 >> | | | `--end_device-6:1 >> ... >> phy and port have to be separated, sas_port are created dynamically. >> >> For ata, all objects are created at initialization time, so the layout is: >> `--0000:09:00.0 >> | `--host0 >> | | `--port1 >> | | | `--link1 >> | | | | `--dev1.0 >> | | | | | `--target0:0:0 >> | | | | | | `--0:0:0:0 >> | | | | | | | `--block >> | | | | | | | | `--sda >> >> If we have a port multiplier, more links are created. >> `--0000:09:00.0 >> ... >> | `--host4 >> | | `--port5 >> | | | `--link5 >> | | | | `--dev5.0 >> [device for the port multiplier] >> | | | `--link5.0 >> | | | | `--dev5.0.0 >> | | | | | `--target4:0:0 >> | | | | | | `--4:0:0:0 >> | | | | | | | `--block >> | | | | | | | | `--sdc >> [disk in port 0 of the port multiplier] >> ... >> | | | `--link5.2 >> | | | | `--dev5.2.0 >> | | | | | `--target4:2:0 >> | | | | | | `--4:2:0:0 >> | | | | | | | `--block >> | | | | | | | | `--sde >> [disk in port 2 of the port multiplier] >> >> In consequence, the path of a scsi device becomes: >> .../0000:00:1f.2/host0/port1/link1/dev1.0/target0:0:0/0:0:0:0 >> dev1.0 indicates the master device [0] in ata port 1. >> ata1 being under host0, we know the reliationships between the scsi_host id and >> ata port id. >> >> or when a port multiplier is present: for instance the device in port 4 of the >> port multiplier: >> .../0000:00:06.0/0000:09:00.0/host5/port6/link6.4/dev6.4.0/target5:4:0/5:4:0:0 > > What's the benefit of this? + To unify ata transport sysfs topology with other scsi transport. + To easily map a ata_port with its associated scsi_host structure. > From a PM perspective now it seems we'll > have the parent hardware port suspended before all the scsi_hosts on > that port, There is a one to one mapping between ata port and scsi-host, so it works. All ata ports must be suspended before the parent hardware is. Gwendal. > which defeats the original purpose of putting the ata_port > in the PM hierarchy. > > -- > Dan -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html