Commit 201aed67 authored Apr 27, 2011 by Brian King Committed by James Bottomley May 01, 2011

[SCSI] ibmvscsi: Improve CRQ reset reliability



This patch fixes an issue seen where an event occurs
which causes the ibmvscsi driver to reset its CRQ. Upon
re-registering its CRQ, it receives H_CLOSED, indicating
the Virtual I/O Server is not yet ready to receive commands.
This resulted in the ibmvscsi driver essentially offlining
the adapter and not recovering. The fix is to re-enable
our interrupt so that when the Virtual I/O server is ready
and sends a CRQ init, we will be able to receive it and
resume initialization of the VSCSI adapter.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

parent 4d4dd706

drivers/scsi/ibmvscsi/ibmvscsi.c

+1 −2

Original line number	Diff line number	Diff line
		@@ -1849,8 +1849,7 @@ static void ibmvscsi_do_work(struct ibmvscsi_host_data *hostdata)
		rc = ibmvscsi_ops->reset_crq_queue(&hostdata->queue, hostdata);
		if (!rc)
		rc = ibmvscsi_ops->send_crq(hostdata, 0xC001000000000000LL, 0);
		if (!rc)
		rc = vio_enable_interrupts(to_vio_dev(hostdata->dev));
		vio_enable_interrupts(to_vio_dev(hostdata->dev));
		} else if (hostdata->reenable_crq) {
		smp_rmb();
		action = "enable";