aacraid: reply queue mapping to CPUs based of IRQ affinity
Subsystems (3)
PATCH v2 1/1 aacraid: reply queue mapping to CPUs based of IRQ affinity
Fix the IO hang that arises because of MSIx vector not having a mapped online CPU upon receiving completion. This patch sets up a reply queue mapping to CPUs based on the IRQ affinity retrieved using pci_irq_get_affinity() API.
aac_setup_reply_map() is an explicit mapping for internally generated (non-SCSI) cmds.
The SCSI cmds take the blk_mq route, and the non-SCSI cmds are mapped to the reply_map.
--- drivers/scsi/aacraid/aacraid.h | 1 + drivers/scsi/aacraid/comminit.c | 32 ++++++++++++++++++++++++++++++++ drivers/scsi/aacraid/commsup.c | 6 +++++- drivers/scsi/aacraid/linit.c | 25 +++++++++++++++++++++++++ drivers/scsi/aacraid/src.c | 13 +++++++++++-- 5 files changed, 74 insertions(+), 3 deletions(-)
drivers/scsi/aacraid/aacraid.h +1 -0
@@ -1678,6 +1678,7 @@ struct aac_devu32 handle_pci_error;bool init_reset;u8 soft_reset_support; +unsigned int *reply_map; };#define aac_adapter_interrupt (dev) \
drivers/scsi/aacraid/comminit.c +32 -0
@@ -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 +663,30 @@ struct aac_dev *aac_init_adapter(struct aac_dev *dev)return dev; } +/* +* aac_setup_reply_map - This is an explicit mapping for +* internally generated (non-SCSI) cmds which need to be +* serviced outside of IO requests. +* The SCSI cmds take the blk_mq mechanism, +* and the non-SCSI cmds are mapped to the reply_map. +*/ +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 ; +}
drivers/scsi/aacraid/commsup.c +5 -1
@@ -223,8 +223,12 @@ int aac_fib_setup(struct aac_dev * dev)struct fib *aac_fib_alloc_tag (struct aac_dev *dev,struct scsi_cmnd *scmd) {struct fib *fibptr; +u32 blk_tag; +int i; - fibptr = &dev->fibs[scsi_cmd_to_rq (scmd)->tag]; + blk_tag =blk_mq_unique_tag (scsi_cmd_to_rq (scmd)); + i =blk_mq_unique_tag_to_tag (blk_tag); + fibptr = &dev->fibs[i];/* * Null out fields that depend on being zero at the start of * each I/O
drivers/scsi/aacraid/linit.c +25 -0
@@ -34,6 +34,7 @@#include <linux/delay.h> #include <linux/kthread.h> #include <linux/msdos_partition.h> +#include <linux/blk-mq-pci.h> #include <scsi/scsi.h> #include <scsi/scsi_cmnd.h>
@@ -505,6 +506,16 @@ static int aac_slave_configure(struct scsi_device *sdev)return 0 ; } +static void aac_map_queues (struct Scsi_Host *shost) +{ +struct aac_dev *aac = (struct aac_dev *)shost->hostdata; + +blk_mq_pci_map_queues (&shost->tag_set.map[HCTX_TYPE_DEFAULT ], + aac->pdev,0 ); +} + + +/** * aac_change_queue_depth - alter queue depths * @sdev: SCSI device we are considering
@@ -1489,6 +1500,7 @@ static struct scsi_host_template aac_driver_template = { .bios_param = aac_biosparm, .shost_groups = aac_host_groups, .slave_configure = aac_slave_configure, + .map_queues = aac_map_queues, .change_queue_depth = aac_change_queue_depth, .sdev_groups = aac_dev_groups, .eh_abort_handler = aac_eh_abort,
@@ -1668,6 +1680,14 @@ 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);
@@ -1776,6 +1796,8 @@ static int aac_probe_one(struct pci_dev *pdev, const struct pci_device_id *id) shost->max_lun =AAC_MAX_LUN ;pci_set_drvdata (pdev, shost); + shost->nr_hw_queues = aac->max_msix; + shost->host_tagset =1 ; error =scsi_add_host (shost, &pdev->dev);if (error)
@@ -1797,6 +1819,8 @@ 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 +1942,7 @@ 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);
drivers/scsi/aacraid/src.c +11 -2
@@ -493,6 +493,8 @@ static int aac_src_deliver_message(struct fib *fib)#endif u16 vector_no; +struct scsi_cmnd *scmd; +u32 blk_tag;atomic_inc (&q->numpending);
@@ -505,8 +507,15 @@ static int aac_src_deliver_message(struct fib *fib)if ((dev->comm_interface ==AAC_COMM_MESSAGE_TYPE3 ) && dev->sa_firmware) vector_no =aac_get_vector (dev); -else - vector_no = fib->vector_no; +else { +if (!fib->vector_no || !fib->callback_data) { + vector_no = dev->reply_map[raw_smp_processor_id ()]; + }else { + scmd = (struct scsi_cmnd *)fib->callback_data; + blk_tag =blk_mq_unique_tag (scsi_cmd_to_rq (scmd)); + vector_no =blk_mq_unique_tag_to_hwq (blk_tag); + } + }if (native_hba) {if (fib->flags &FIB_CONTEXT_FLAG_NATIVE_HBA_TMF ) {
Signature · 1 line
2.29.0
kernel test robot
build botFYI: kernel test robot notices the stable kernel rule is not satisfied.
Hi,
Thanks for your patch.
FYI: kernel test robot notices the stable kernel rule is not satisfied.
Rule: 'Cc: [email protected]' or 'commit <sha1> upstream.'
Subject: [PATCH v2] aacraid: reply queue mapping to CPUs based of IRQ affinity
The check is based on https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests
Quoted 9 lines · Sagar Biradar
Fix the IO hang that arises because of MSIx vector not having a mapped online CPU upon receiving completion. This patch sets up a reply queue mapping to CPUs based on the IRQ affinity retrieved using
pci_irq_get_affinity()API.
aac_setup_reply_map()is an explicit mapping for internally generated (non-SCSI) cmds.
The SCSI cmds take theblk_mqroute, and the non-SCSI cmds are mapped to thereply_map.
This now looks better.
I would still prefer if no reply_map was used even for internal commands. As I see, you have two alternatives (to using reply_map):
- instead of using a driver-internal reply_map, lookup CPU->HW queue mapping for internal commands by using shost->tag_set.map[HCTX_TYPE_DEFAULT].mq_map[raw_smp_processor_id()]
Ideally when we finally support reserved commands for SCSI ML we will have a better solution for this.
- if it is possible to send driver internal commands on a specific HW queue always, then reserve a dedicated HW queue for them (and always send on that HW queue). You may reserve this HW queue by omitting 1x HW queue from pci_alloc_irq_vectors_affinity() for affinity spread
Quoted 108 lines
Reviewed-by Gilbert Wu <[email protected]>Signed-off-by Sagar Biradar <[email protected]>--- drivers/scsi/aacraid/aacraid.h | 1 + drivers/scsi/aacraid/comminit.c | 32 ++++++++++++++++++++++++++++++++ drivers/scsi/aacraid/commsup.c | 6 +++++- drivers/scsi/aacraid/linit.c | 25 +++++++++++++++++++++++++ drivers/scsi/aacraid/src.c | 13 +++++++++++-- 5 files changed, 74 insertions(+), 3 deletions(-)drivers/scsi/aacraid/aacraid.h +1
-0@@ -1678,6 +1678,7 @@ struct aac_devu32 handle_pci_error;bool init_reset;u8 soft_reset_support; +unsigned int *reply_map; };#define aac_adapter_interrupt (dev) \drivers/scsi/aacraid/comminit.c +32
-0@@ -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 +663,30 @@ struct aac_dev *aac_init_adapter(struct aac_dev *dev)return dev; } +/* +* aac_setup_reply_map - This is an explicit mapping for +* internally generated (non-SCSI) cmds which need to be +* serviced outside of IO requests. +* The SCSI cmds take the blk_mq mechanism, +* and the non-SCSI cmds are mapped to the reply_map. +*/ +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 ; +}drivers/scsi/aacraid/commsup.c +5
-1@@ -223,8 +223,12 @@ int aac_fib_setup(struct aac_dev * dev)struct fib *aac_fib_alloc_tag (struct aac_dev *dev,struct scsi_cmnd *scmd) {struct fib *fibptr; +u32 blk_tag; +int i; - fibptr = &dev->fibs[scsi_cmd_to_rq (scmd)->tag]; + blk_tag =blk_mq_unique_tag (scsi_cmd_to_rq (scmd)); + i =blk_mq_unique_tag_to_tag (blk_tag); + fibptr = &dev->fibs[i];/* * Null out fields that depend on being zero at the start of * each I/O drivers/scsi/aacraid/linit.c +1
-0@@ -34,6 +34,7 @@#include <linux/delay.h> #include <linux/kthread.h> #include <linux/msdos_partition.h> +#include <linux/blk-mq-pci.h> #include <scsi/scsi.h> #include <scsi/scsi_cmnd.h>
drivers/scsi/aacraid/linit.c static int aac_slave_configure(struct scsi_device *sdev) · line 511
Quoted 6 lines
drivers/scsi/aacraid/linit.c +3
-0@@ -505,6 +506,16 @@ static int aac_slave_configure(struct scsi_device *sdev)return 0 ; } +static void aac_map_queues (struct Scsi_Host *shost) +{ +struct aac_dev *aac = (struct aac_dev *)shost->hostdata;
I don't think that you need a explicit casting ...
Quoted 90 lines
drivers/scsi/aacraid/linit.c +21
-0+ +blk_mq_pci_map_queues (&shost->tag_set.map[HCTX_TYPE_DEFAULT ], + aac->pdev,0 ); +} + + +/** * aac_change_queue_depth - alter queue depths * @sdev: SCSI device we are considering @@ -1489,6 +1500,7 @@ static struct scsi_host_template aac_driver_template = { .bios_param = aac_biosparm, .shost_groups = aac_host_groups, .slave_configure = aac_slave_configure, + .map_queues = aac_map_queues, .change_queue_depth = aac_change_queue_depth, .sdev_groups = aac_dev_groups, .eh_abort_handler = aac_eh_abort,@@ -1668,6 +1680,14 @@ 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);@@ -1776,6 +1796,8 @@ static int aac_probe_one(struct pci_dev *pdev, const struct pci_device_id *id) shost->max_lun =AAC_MAX_LUN ;pci_set_drvdata (pdev, shost); + shost->nr_hw_queues = aac->max_msix; + shost->host_tagset =1 ; error =scsi_add_host (shost, &pdev->dev);if (error)@@ -1797,6 +1819,8 @@ 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 +1942,7 @@ 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);drivers/scsi/aacraid/src.c +11
-2@@ -493,6 +493,8 @@ static int aac_src_deliver_message(struct fib *fib)#endif u16 vector_no; +struct scsi_cmnd *scmd; +u32 blk_tag;atomic_inc (&q->numpending);@@ -505,8 +507,15 @@ static int aac_src_deliver_message(struct fib *fib)if ((dev->comm_interface ==AAC_COMM_MESSAGE_TYPE3 ) && dev->sa_firmware) vector_no =aac_get_vector (dev); -else - vector_no = fib->vector_no; +else { +if (!fib->vector_no || !fib->callback_data) { + vector_no = dev->reply_map[raw_smp_processor_id ()]; + }else { + scmd = (struct scsi_cmnd *)fib->callback_data; + blk_tag =blk_mq_unique_tag (scsi_cmd_to_rq (scmd)); + vector_no =blk_mq_unique_tag_to_hwq (blk_tag); + } + }if (native_hba) {if (fib->flags &FIB_CONTEXT_FLAG_NATIVE_HBA_TMF ) {