[adding the mpt maintainers as this seems to be a driver and not midlayer issue] On Wed, Nov 12, 2014 at 02:59:11PM -0500, Douglas Gilbert wrote: > From a correspondent and my own testing I have seen way > too many of these messages in the log: > log_info(0x31160000): originator(PL), code(0x16), sub_code(0x0000) > > That comes from either the mpt2sas or mpt3sas driver and may be > a problem with their interaction with the SCSI EH. In one case, > those messages go on forever, requiring a reboot; in my testing > (with sg_readcap) the command timeout (60 seconds) stopped them. > > > How they occur needs a bit of explaining: ATA disks are designed > to have only only initiator (host). So if you build a SAS fabric > including at least two initiators, an expander and one SATA disk, > then there is potentially a problem which SAS expanders address > with "affiliations". An affiliation is a mechanism for the > expander to remember the SAS address of the initiator (host) > that first "grabbed" the SATA disk, and rejecting any other > initiator that tries to access that SATA disk. > > That rejection, in the link layer in SAS for the STP protocol, > is a OPEN_REJECT (STP RESOURCES BUSY) response. That is *not* > a retry-able error (so the use of "busy" is unfortunate). > FreeBSD handles this correctly, Linux in some cases retries > which results in chaos plus bloated logs. > > There are mechanisms for the owner of the affiliation to clear > it so another initiator can claim it. However affiliations are > designed to thwart brute force attempts by non-owners. At best > non-owners should get one log message along the lines of: > cannot access SATA disk xxxx since another machine/HBA is > affiliated with it > > Linux properly handles SATA affiliations when it comes across > them in normal device discovery. It is the "surprise" > disappearance of an affiliation that causes instability. That > surprise is caused by a utility like smp_phy_control telling > the expander to clear the affiliation and doing a rescan on > the other machine to claim the affiliation. > > Doug Gilbert > > > -- > To unsubscribe from this list: send the line "unsubscribe linux-scsi" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html ---end quoted text--- -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
