@@ -735,8 +735,14 @@ struct se_device *target_alloc_device(struct se_hba *hba, const char *name)
dev->queue_cnt = nr_cpu_ids;
for (i = 0; i < dev->queue_cnt; i++) {
- INIT_LIST_HEAD(&dev->queues[i].state_list);
- spin_lock_init(&dev->queues[i].lock);
+ struct se_device_queue *q;
+
+ q = &dev->queues[i];
+ INIT_LIST_HEAD(&q->state_list);
+ spin_lock_init(&q->lock);
+
+ init_llist_head(&q->sq.cmd_list);
+ INIT_WORK(&q->sq.work, target_queued_submit_work);
}
dev->se_hba = hba;
@@ -153,6 +153,7 @@ void target_qf_do_work(struct work_struct *work);
bool target_check_wce(struct se_device *dev);
bool target_check_fua(struct se_device *dev);
void __target_execute_cmd(struct se_cmd *, bool);
+void target_queued_submit_work(struct work_struct *work);
/* target_core_stat.c */
void target_stat_setup_dev_default_groups(struct se_device *);
@@ -41,6 +41,7 @@
#include <trace/events/target.h>
static struct workqueue_struct *target_completion_wq;
+static struct workqueue_struct *target_submission_wq;
static struct kmem_cache *se_sess_cache;
struct kmem_cache *se_ua_cache;
struct kmem_cache *t10_pr_reg_cache;
@@ -129,8 +130,15 @@ int init_se_kmem_caches(void)
if (!target_completion_wq)
goto out_free_lba_map_mem_cache;
+ target_submission_wq = alloc_workqueue("target_submission",
+ WQ_MEM_RECLAIM, 0);
+ if (!target_submission_wq)
+ goto out_free_completion_wq;
+
return 0;
+out_free_completion_wq:
+ destroy_workqueue(target_completion_wq);
out_free_lba_map_mem_cache:
kmem_cache_destroy(t10_alua_lba_map_mem_cache);
out_free_lba_map_cache:
@@ -153,6 +161,7 @@ int init_se_kmem_caches(void)
void release_se_kmem_caches(void)
{
+ destroy_workqueue(target_submission_wq);
destroy_workqueue(target_completion_wq);
kmem_cache_destroy(se_sess_cache);
kmem_cache_destroy(se_ua_cache);
@@ -1380,7 +1389,6 @@ void transport_init_se_cmd(
{
INIT_LIST_HEAD(&cmd->se_delayed_node);
INIT_LIST_HEAD(&cmd->se_qf_node);
- INIT_LIST_HEAD(&cmd->se_cmd_list);
INIT_LIST_HEAD(&cmd->state_list);
init_completion(&cmd->t_transport_stop_comp);
cmd->free_compl = NULL;
@@ -1827,6 +1835,87 @@ int target_submit_cmd(struct se_cmd *se_cmd, struct se_session *se_sess,
}
EXPORT_SYMBOL(target_submit_cmd);
+void target_queued_submit_work(struct work_struct *work)
+{
+ struct se_cmd_queue *sq = container_of(work, struct se_cmd_queue, work);
+ struct se_cmd *se_cmd, *next_cmd;
+ struct llist_node *cmd_list;
+ unsigned char *cdb;
+
+ cmd_list = llist_del_all(&sq->cmd_list);
+ if (!cmd_list)
+ /* Previous call took what we were queued to submit */
+ return;
+
+ cmd_list = llist_reverse_order(cmd_list);
+ llist_for_each_entry_safe(se_cmd, next_cmd, cmd_list, se_cmd_list) {
+ cdb = se_cmd->se_tfo->get_cdb(se_cmd);
+ target_submit(se_cmd, cdb);
+ }
+}
+
+static void target_queue_cmd_work(struct se_cmd_queue *q, struct se_cmd *se_cmd,
+ int cpu)
+{
+ llist_add(&se_cmd->se_cmd_list, &q->cmd_list);
+ queue_work_on(cpu, target_submission_wq, &q->work);
+}
+
+/**
+ * target_queue_cmd_submit - queue the cmd to run on the LIO workqueue
+ * @se_cmd: command descriptor to submit
+ * @se_sess: associated se_sess for endpoint
+ * @sense: pointer to SCSI sense buffer
+ * @unpacked_lun: unpacked LUN to reference for struct se_lun
+ * @data_length: fabric expected data transfer length
+ * @task_attr: SAM task attribute
+ * @data_dir: DMA data direction
+ * @flags: flags for command submission from target_sc_flags_tables
+ * @sgl: struct scatterlist memory for unidirectional mapping
+ * @sgl_count: scatterlist count for unidirectional mapping
+ * @sgl_bidi: struct scatterlist memory for bidirectional READ mapping
+ * @sgl_bidi_count: scatterlist count for bidirectional READ mapping
+ * @sgl_prot: struct scatterlist memory protection information
+ * @sgl_prot_count: scatterlist count for protection information
+ *
+ * Task tags are supported if the caller has set @se_cmd->tag.
+ *
+ * Returns non zero to signal active I/O shutdown failure. All other
+ * setup exceptions will be returned as a SCSI CHECK_CONDITION response,
+ * but still return zero here.
+ *
+ * This may only be called from interrupt context if the caller's
+ * queue_status and release_cmd callouts do not block.
+ *
+ * This assumes internal allocation of fabric payload buffer by target-core.
+ */
+int
+target_queue_cmd_submit(struct se_cmd *se_cmd, struct se_session *se_sess,
+ unsigned char *sense, u64 unpacked_lun, u32 data_length,
+ int task_attr, int data_dir, int flags,
+ struct scatterlist *sgl, u32 sgl_count,
+ struct scatterlist *sgl_bidi, u32 sgl_bidi_count,
+ struct scatterlist *sgl_prot, u32 sgl_prot_count)
+{
+ struct se_device *se_dev;
+ int cpu, ret;
+
+ ret = target_submit_prep(se_cmd, se_sess, sense, unpacked_lun,
+ data_length, task_attr, data_dir, flags,
+ sgl, sgl_count, sgl_bidi, sgl_bidi_count,
+ sgl_prot, sgl_prot_count);
+ if (ret < 0)
+ return ret;
+ else if (ret > 0)
+ return 0;
+
+ cpu = se_cmd->cpuid;
+ se_dev = se_cmd->se_dev;
+ target_queue_cmd_work(&se_dev->queues[cpu].sq, se_cmd, cpu);
+ return 0;
+}
+EXPORT_SYMBOL_GPL(target_queue_cmd_submit);
+
static void target_complete_tmr_failure(struct work_struct *work)
{
struct se_cmd *se_cmd = container_of(work, struct se_cmd, work);
@@ -487,7 +487,7 @@ struct se_cmd {
/* Only used for internal passthrough and legacy TCM fabric modules */
struct se_session *se_sess;
struct se_tmr_req *se_tmr_req;
- struct list_head se_cmd_list;
+ struct llist_node se_cmd_list;
struct completion *free_compl;
struct completion *abrt_compl;
const struct target_core_fabric_ops *se_tfo;
@@ -764,9 +764,15 @@ struct se_dev_stat_grps {
struct config_group scsi_lu_group;
};
+struct se_cmd_queue {
+ struct llist_head cmd_list;
+ struct work_struct work;
+};
+
struct se_device_queue {
struct list_head state_list;
spinlock_t lock;
+ struct se_cmd_queue sq;
};
struct se_device {
@@ -80,6 +80,7 @@ struct target_core_fabric_ops {
int (*queue_status)(struct se_cmd *);
void (*queue_tm_rsp)(struct se_cmd *);
void (*aborted_task)(struct se_cmd *);
+ unsigned char *(*get_cdb)(struct se_cmd *);
/*
* fabric module calls for target_core_fabric_configfs.c
*/
@@ -160,6 +161,12 @@ int target_submit_cmd_map_sgls(struct se_cmd *, struct se_session *,
struct scatterlist *, u32);
int target_submit_cmd(struct se_cmd *, struct se_session *, unsigned char *,
unsigned char *, u64, u32, int, int, int);
+int target_queue_cmd_submit(struct se_cmd *se_cmd, struct se_session *se_sess,
+ unsigned char *sense, u64 unpacked_lun, u32 data_length,
+ int task_attr, int data_dir, int flags, struct scatterlist *sgl,
+ u32 sgl_count, struct scatterlist *sgl_bidi, u32 sgl_bidi_count,
+ struct scatterlist *sgl_prot, u32 sgl_prot_count);
+
int target_submit_tmr(struct se_cmd *se_cmd, struct se_session *se_sess,
unsigned char *sense, u64 unpacked_lun,
void *fabric_tmr_ptr, unsigned char tm_type,
loop and vhost-scsi do their target cmd submission from driver workqueues. This allows them to avoid an issue where the backend may block waiting for resources like tags/requests, mem/locks, etc and that ends up blocking their entire submission path and for the case of vhost-scsi both the submission and completion path. This patch adds a helper drivers can use to submit from a lio workqueue. This code will then be extended in the next patches to fix the plugging of backend devices. Note: I'm only converting vhost/loop initially, but the workqueue based submission will work for other drivers and have similar benefits where the main target loops will not end up blocking one some backend resource. I'll port others when I have more time to test as I think we might want to make it configurable for some drivers. Signed-off-by: Mike Christie <michael.christie@oracle.com> --- drivers/target/target_core_device.c | 10 ++- drivers/target/target_core_internal.h | 1 + drivers/target/target_core_transport.c | 91 +++++++++++++++++++++++++- include/target/target_core_base.h | 8 ++- include/target/target_core_fabric.h | 7 ++ 5 files changed, 113 insertions(+), 4 deletions(-)