On 28/03/2023 22:41, Sagar Biradar wrote:
Fix the IO hang that arises because of MSIx vector not
having a mapped online CPU upon receiving completion.
What about if the CPU targeted goes offline while the IO is in-flight?
This patch sets up a reply queue mapping to CPUs based on the
IRQ affinity retrieved using pci_irq_get_affinity() API.
blk-mq already does what you want here, including handling for the case
I mention above. It maintains a CPU -> HW queue mapping, and using a
reply map in the LLD is the old way of doing this.
Could you instead follow the example in commit 664f0dce2058 ("scsi:
mpt3sas: Add support for shared host tagset for CPU hotplug"), and
expose the HW queues to the upper layer? You can alternatively check the
example of any SCSI driver which sets shost->host_tagset for this.
Thanks,
John
Reviewed-by: Gilbert Wu <gilbert.wu@xxxxxxxxxxxxx>
Signed-off-by: Sagar Biradar <Sagar.Biradar@xxxxxxxxxxxxx>
---
drivers/scsi/aacraid/aacraid.h | 1 +
drivers/scsi/aacraid/comminit.c | 25 +++++++++++++++++++++++++
drivers/scsi/aacraid/linit.c | 11 +++++++++++
drivers/scsi/aacraid/src.c | 2 +-
4 files changed, 38 insertions(+), 1 deletion(-)
diff --git a/drivers/scsi/aacraid/aacraid.h b/drivers/scsi/aacraid/aacraid.h
index 5e115e8b2ba4..4a23f9fab61f 100644
--- a/drivers/scsi/aacraid/aacraid.h
+++ b/drivers/scsi/aacraid/aacraid.h
@@ -1678,6 +1678,7 @@ struct aac_dev
u32 handle_pci_error;
bool init_reset;
u8 soft_reset_support;
+ unsigned int *reply_map;
};
#define aac_adapter_interrupt(dev) \
diff --git a/drivers/scsi/aacraid/comminit.c b/drivers/scsi/aacraid/comminit.c
index bd99c5492b7d..6fc323844a31 100644
--- a/drivers/scsi/aacraid/comminit.c
+++ b/drivers/scsi/aacraid/comminit.c
@@ -33,6 +33,8 @@
#include "aacraid.h"
+void aac_setup_reply_map(struct aac_dev *dev);
+
struct aac_common aac_config = {
.irq_mod = 1
};
@@ -630,6 +632,9 @@ struct aac_dev *aac_init_adapter(struct aac_dev *dev)
if (aac_is_src(dev))
aac_define_int_mode(dev);
+
+ aac_setup_reply_map(dev);
+
/*
* Ok now init the communication subsystem
*/
@@ -658,3 +663,23 @@ struct aac_dev *aac_init_adapter(struct aac_dev *dev)
return dev;
}
+void aac_setup_reply_map(struct aac_dev *dev)
+{
+ const struct cpumask *mask;
+ unsigned int i, cpu = 1;
+
+ for (i = 1; i < dev->max_msix; i++) {
+ mask = pci_irq_get_affinity(dev->pdev, i);
+ if (!mask)
+ goto fallback;
+
+ for_each_cpu(cpu, mask) {
+ dev->reply_map[cpu] = i;
+ }
+ }
+ return;
+
+fallback:
+ for_each_possible_cpu(cpu)
+ dev->reply_map[cpu] = 0;
+}
diff --git a/drivers/scsi/aacraid/linit.c b/drivers/scsi/aacraid/linit.c
index 5ba5c18b77b4..af60c7d26407 100644
--- a/drivers/scsi/aacraid/linit.c
+++ b/drivers/scsi/aacraid/linit.c
@@ -1668,6 +1668,14 @@ static int aac_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
goto out_free_host;
}
+ aac->reply_map = kzalloc(sizeof(unsigned int) * nr_cpu_ids,
+ GFP_KERNEL);
+ if (!aac->reply_map) {
+ error = -ENOMEM;
+ dev_err(&pdev->dev, "reply_map allocation failed\n");
+ goto out_free_host;
+ }
+
spin_lock_init(&aac->fib_lock);
mutex_init(&aac->ioctl_mutex);
@@ -1797,6 +1805,8 @@ static int aac_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
aac->comm_addr, aac->comm_phys);
kfree(aac->queues);
aac_adapter_ioremap(aac, 0);
+ /* By now we should have configured the reply_map */
+ kfree(aac->reply_map);
kfree(aac->fibs);
kfree(aac->fsa_dev);
out_free_host:
@@ -1918,6 +1928,7 @@ static void aac_remove_one(struct pci_dev *pdev)
aac_adapter_ioremap(aac, 0);
+ kfree(aac->reply_map);
kfree(aac->fibs);
kfree(aac->fsa_dev);
diff --git a/drivers/scsi/aacraid/src.c b/drivers/scsi/aacraid/src.c
index 11ef58204e96..e84ec60a655b 100644
--- a/drivers/scsi/aacraid/src.c
+++ b/drivers/scsi/aacraid/src.c
@@ -506,7 +506,7 @@ static int aac_src_deliver_message(struct fib *fib)
&& dev->sa_firmware)
vector_no = aac_get_vector(dev);
else
- vector_no = fib->vector_no;
+ vector_no = dev->reply_map[raw_smp_processor_id()];
if (native_hba) {
if (fib->flags & FIB_CONTEXT_FLAG_NATIVE_HBA_TMF) {
