On Tue, May 12, 2015 at 1:20 PM, Praveen Murali <pmurali@xxxxxxxxxxxx> wrote: >> > -----Original Message----- >> > From: Dan Williams [mailto:dan.j.williams@xxxxxxxxx] >> > Sent: Monday, May 11, 2015 3:44 PM >> > To: Praveen Murali >> > Cc: James Bottomley (James.Bottomley@xxxxxxxxxxxxxxxxxxxxx); linux- >> > scsi@xxxxxxxxxxxxxxx >> > Subject: Re: sysfs group not found for kobject on mvsas drive removal >> > >> > So I wonder if this is a new warning, because as far as I can tell >> > we've always been double deleting the devices below sas_port. It >> > seems we should delete all the devices from rphy down before deleting >> > the parent port. Something like this (excuse the whitespace >> > mangling). Only compile tested as I don't have a SAS setup handy. >> > > I tested that patch on my setup with the Marvell SAS HBA and it seems to be working well; all the warnings are gone. Cool, hmm, makes me nervous when untested patches work right the first time :-). I wonder if we are still managing the reference counts correctly and not leaking devices. Previously we would get 2 calls to put_device() on the sas_port device. Now I think we only get one. Can you confirm that sas_port_release() is still being called. /me also wonders why the scsi_transport_class has "port->dev.parent = get_device(parent);" all over the place. Parents should be pinned as long as children are registered. James, do you recall why these extra/explicit references are taken? -- 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
