The 5.5 kernel includes a mvsas driver backported from 2.6.25, so I hear. In 2.6.31, Marvell contributed a new driver. It has novel features like error handling. On my cluster of 160 machines, the driver shipped in 5.5 dies about once per day in this code: case SAS_PROTOCOL_SATA | SAS_PROTOCOL_STP:{ dev_printk(KERN_DEBUG, &pdev->dev, "STP Abort! \n"); spin_lock_irqsave(&task->task_state_lock, flags); if (task->task_state_flags & SAS_TASK_NEED_DEV_RESET) { /* TODO */ ; } Could you please backport the new driver with error handling? I'd be happy to be a tester.
(In reply to comment #0) > The 5.5 kernel includes a mvsas driver backported from 2.6.25, so I hear. > > In 2.6.31, Marvell contributed a new driver. It has novel features like error > handling. > > On my cluster of 160 machines, the driver shipped in 5.5 dies about once per > day in this code: > > case SAS_PROTOCOL_SATA | SAS_PROTOCOL_STP:{ > dev_printk(KERN_DEBUG, &pdev->dev, "STP Abort! \n"); > spin_lock_irqsave(&task->task_state_lock, flags); > if (task->task_state_flags & SAS_TASK_NEED_DEV_RESET) { > /* TODO */ > ; > } > > Could you please backport the new driver with error handling? I'd be happy to > be a tester. Unfortunately there has been irregular and incomplete maintenance of the mvsas driver upstream by Marvell. This leads to lack of confidence in any updates. I have followed the mvsas driver on linux-scsi and many issues have gone unanswered by Marvell. Should Marvell start maintaining the mvsas driver more dilligently, such updates will be more likely. Rob
Some recent posts on lkml seem to indicate some promise with the latest mvsas as of 2.6.37-rc4 or so. Any chance this might appear in RHEL 5 or 6?
(In reply to comment #3) > Some recent posts on lkml seem to indicate some promise with the latest mvsas > as of 2.6.37-rc4 or so. Any chance this might appear in RHEL 5 or 6? Unlikely, unfortunately.