diff mbox series

qla2xxx: Avoid PCI IRQ affinity mapping when multiqueue is not supported.

Message ID 20190216004255.28245-1-hmadhani@marvell.com (mailing list archive)
State Mainlined
Commit f3e026951771bceb17319a4d0d6121ca58746c88
Headers show
Series qla2xxx: Avoid PCI IRQ affinity mapping when multiqueue is not supported. | expand

Commit Message

Himanshu Madhani Feb. 16, 2019, 12:42 a.m. UTC 
From: Giridhar Malavali <gmalavali@marvell.com>

This patch fixes warning seen when BLK-MQ is enabled and hardware
does not support MQ. This will result into driver requesting MSIx
vectors which are equal or less than pre_desc via PCI IRQ Affinity
infrastructure.

    [   19.746300] qla2xxx [0000:00:00.0]-0005: : QLogic Fibre Channel HBA Driver: 10.00.00.12-k.
    [   19.746599] qla2xxx [0000:02:00.0]-001d: : Found an ISP2432 irq 18 iobase 0x(____ptrval____).
    [   20.203186] ------------[ cut here ]------------
    [   20.203306] WARNING: CPU: 8 PID: 268 at drivers/pci/msi.c:1273 pci_irq_get_affinity+0xf4/0x120
    [   20.203481] Modules linked in: tg3 ptp qla2xxx(+) pps_core sg libphy scsi_transport_fc flash loop autofs4
    [   20.203700] CPU: 8 PID: 268 Comm: systemd-udevd Not tainted 5.0.0-rc5-00358-gdf3865f #113
    [   20.203830] Call Trace:
    [   20.203933]  [0000000000461bb0] __warn+0xb0/0xe0
    [   20.204090]  [00000000006c8f34] pci_irq_get_affinity+0xf4/0x120
    [   20.204219]  [000000000068c764] blk_mq_pci_map_queues+0x24/0x120
    [   20.204396]  [00000000007162f4] scsi_map_queues+0x14/0x40
    [   20.204626]  [0000000000673654] blk_mq_update_queue_map+0x94/0xe0
    [   20.204698]  [0000000000676ce0] blk_mq_alloc_tag_set+0x120/0x300
    [   20.204869]  [000000000071077c] scsi_add_host_with_dma+0x7c/0x300
    [   20.205419]  [00000000100ead54] qla2x00_probe_one+0x19d4/0x2640 [qla2xxx]
    [   20.205621]  [00000000006b3c88] pci_device_probe+0xc8/0x160
    [   20.205697]  [0000000000701c0c] really_probe+0x1ac/0x2e0
    [   20.205770]  [0000000000701f90] driver_probe_device+0x50/0x100
    [   20.205843]  [0000000000702134] __driver_attach+0xf4/0x120
    [   20.205913]  [0000000000700644] bus_for_each_dev+0x44/0x80
    [   20.206081]  [0000000000700c98] bus_add_driver+0x198/0x220
    [   20.206300]  [0000000000702950] driver_register+0x70/0x120
    [   20.206582]  [0000000010248224] qla2x00_module_init+0x224/0x284 [qla2xxx]
    [   20.206857] ---[ end trace b1de7a3f79fab2c2 ]---

The fix is to check if the hardware does not have Multi Queue capabiltiy, use
pci_alloc_irq_vectors() call instead of pci_alloc_irq_affinity().

Fixes: f664a3cc17b7d ("scsi: kill off the legacy IO path")
Cc: stable@vger.kernel.org #4.19
Signed-off-by: Giridhar Malavali <gmalavali@marvell.com>
Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
---
Hi Martin, 

This patch fixes issue, if BLK-MQ is enabled, driver will not map IRQ affinity
if the hardware does not support MQ. This was reported by Meelis Ross here 

https://marc.info/?l=linux-scsi&m=154980175720501&w=2

Please apply this on top of my earlier series for inclusion in 5.1/scsi-queue

https://marc.info/?l=linux-scsi&m=155027035917217&w=2

Thanks,
Himanshu
---
 drivers/scsi/qla2xxx/qla_isr.c | 2 +-
 drivers/scsi/qla2xxx/qla_os.c  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Comments

Martin K. Petersen Feb. 16, 2019, 3:46 a.m. UTC | #1 
Himanshu,

> This patch fixes warning seen when BLK-MQ is enabled and hardware does
> not support MQ. This will result into driver requesting MSIx vectors
> which are equal or less than pre_desc via PCI IRQ Affinity
> infrastructure.

Applied to 5.1/scsi-queue, thanks!
diff mbox series

Patch

diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index b5ae76869d5b..89ee82e6773b 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -3431,7 +3431,7 @@  qla24xx_enable_msix(struct qla_hw_data *ha, struct rsp_que *rsp)
 		min_vecs++;
 	}
 
-	if (USER_CTRL_IRQ(ha)) {
+	if (USER_CTRL_IRQ(ha) || !ha->mqiobase) {
 		/* user wants to control IRQ setting for target mode */
 		ret = pci_alloc_irq_vectors(ha->pdev, min_vecs,
 		    ha->msix_count, PCI_IRQ_MSIX);
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index 7bf23943c815..84cd6bff8b12 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -7181,7 +7181,7 @@  static int qla2xxx_map_queues(struct Scsi_Host *shost)
 	scsi_qla_host_t *vha = (scsi_qla_host_t *)shost->hostdata;
 	struct blk_mq_queue_map *qmap = &shost->tag_set.map[0];
 
-	if (USER_CTRL_IRQ(vha->hw))
+	if (USER_CTRL_IRQ(vha->hw) || !vha->hw->mqiobase)
 		rc = blk_mq_map_queues(qmap);
 	else
 		rc = blk_mq_pci_map_queues(qmap, vha->hw->pdev, 0);