Centrelize common fabrics code to core drivers
Subsystems (2)
cover Centrelize common fabrics code to core drivers
Hi Christoph, Sagi, Keith and Co,
This series is intended to move common reconnection, device removal and error recovery logic to common fabrics and core drivers. In this stage, it mainly deals with the RDMA and TCP drivers which had the same logic for the above tasks. FC driver can start using this framework in the future after some preparation work that should be done.
This will ease on code maintenance and reduce the code duplication between the RDMA and TCP drivers.
Since there are some attributes that were moved to generic nvme_ctrl structure, I've measured perf impact.
NVMe PCI Perf numbers using NVIDIA's NVMe SNAP Bluefield-2 device (Fio 16 jobs, 128 iodepth, libaio):
IO size rand READ (before/after) rand WRITE (before/after)
-------- ------------------------- --------------------------
512B 2189k/2197k 2031k/2092k 1KB 2188k/2201k 2027k/2088k 2KB 2166k/2181k 2023k/2078k 4KB 2153k/2169k 2025k/2076k 8KB 1381k/1381k 1378k/1375k 16KB 669k/676k 660k/664k 32KB 320k/322k 305k/305k 64KB 157k/157k 151k/151k
As we can see, there is no impact on the performance.
Max Gurtovoy (7):
nvme: add connect_work attribute to nvme ctrl
nvme-fabrics: introduce nvmf_reconnect_or_remove API
nvme: add err_work attribute to nvme ctrl
nvme-fabrics: introduce nvmf_error_recovery API nvme/nvme-fabrics: introduce nvmf_error_recovery_work API nvme/nvme-fabrics: introduce nvmf_reconnect_ctrl_work API
nvme-fabrics: add nvmf_init_ctrl/nvmf_teardown_ctrl API
drivers/nvme/host/fabrics.c | 90 ++++++++++++++++++++++++ drivers/nvme/host/fabrics.h | 4 ++ drivers/nvme/host/fc.c | 23 ++++--- drivers/nvme/host/nvme.h | 7 ++ drivers/nvme/host/rdma.c | 134 ++++++++++-------------------------- drivers/nvme/host/tcp.c | 124 ++++++++------------------------- 6 files changed, 180 insertions(+), 202 deletions(-)
Signature · 1 line
2.18.1
PATCH 1/7 nvme: add connect_work attribute to nvme ctrl
This structure is duplicated for all the fabric controllers. Move it to common code.
--- drivers/nvme/host/fc.c | 23 ++++++++++++----------- drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 10 ++++------ drivers/nvme/host/tcp.c | 9 ++++----- 4 files changed, 21 insertions(+), 22 deletions(-)
drivers/nvme/host/fc.c +12 -11
@@ -167,7 +167,6 @@ struct nvme_fc_ctrl {struct blk_mq_tag_set tag_set;struct work_struct ioerr_work; -struct delayed_work connect_work;struct kref ref;unsigned long flags;
@@ -567,7 +566,7 @@ nvme_fc_resume_controller(struct nvme_fc_ctrl *ctrl)"NVME-FC{%d}: connectivity re-established. " "Attempting reconnect\n", ctrl->cnum); -queue_delayed_work (nvme_wq, &ctrl->connect_work,0 ); +queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work,0 );break ;case NVME_CTRL_RESETTING :
@@ -3263,7 +3262,7 @@ nvme_fc_delete_ctrl(struct nvme_ctrl *nctrl)struct nvme_fc_ctrl *ctrl =to_fc_ctrl (nctrl);cancel_work_sync (&ctrl->ioerr_work); -cancel_delayed_work_sync (&ctrl->connect_work); +cancel_delayed_work_sync (&ctrl->ctrl.connect_work);/* * kill the association on the link side. this will block * waiting for io to terminate
@@ -3300,7 +3299,8 @@ nvme_fc_reconnect_or_delete(struct nvme_fc_ctrl *ctrl, int status)else if (time_after (jiffies + recon_delay, rport->dev_loss_end)) recon_delay = rport->dev_loss_end - jiffies; -queue_delayed_work (nvme_wq, &ctrl->connect_work, recon_delay); +queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, + recon_delay); }else {if (portptr->port_state ==FC_OBJSTATE_ONLINE ) {if (status >0 && (status &NVME_SC_DNR ))
@@ -3340,12 +3340,13 @@ nvme_fc_reset_ctrl_work(struct work_struct *work) "to CONNECTING\n", ctrl->cnum);if (ctrl->rport->remoteport.port_state ==FC_OBJSTATE_ONLINE ) { -if (!queue_delayed_work (nvme_wq, &ctrl->connect_work,0 )) { +if (!queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, +0 )) {dev_err (ctrl->ctrl.device,"NVME-FC{%d}: failed to schedule connect " "after reset\n" , ctrl->cnum); }else { -flush_delayed_work (&ctrl->connect_work); +flush_delayed_work (&ctrl->ctrl.connect_work); } }else {nvme_fc_reconnect_or_delete (ctrl, -ENOTCONN );
@@ -3373,7 +3374,7 @@ nvme_fc_connect_ctrl_work(struct work_struct *work)struct nvme_fc_ctrl *ctrl =container_of (to_delayed_work (work), - struct nvme_fc_ctrl, connect_work); + struct nvme_fc_ctrl, ctrl.connect_work); ret =nvme_fc_create_association (ctrl);if (ret)
@@ -3485,7 +3486,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,kref_init (&ctrl->ref);INIT_WORK (&ctrl->ctrl.reset_work, nvme_fc_reset_ctrl_work); -INIT_DELAYED_WORK (&ctrl->connect_work, nvme_fc_connect_ctrl_work); +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_fc_connect_ctrl_work);INIT_WORK (&ctrl->ioerr_work, nvme_fc_ctrl_ioerr_work);spin_lock_init (&ctrl->lock);
@@ -3561,14 +3562,14 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, goto fail_ctrl; } -if (!queue_delayed_work (nvme_wq, &ctrl->connect_work,0 )) { +if (!queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work,0 )) {dev_err (ctrl->ctrl.device,"NVME-FC{%d}: failed to schedule initial connect\n" , ctrl->cnum); goto fail_ctrl; } -flush_delayed_work (&ctrl->connect_work); +flush_delayed_work (&ctrl->ctrl.connect_work); dev_info(ctrl->ctrl.device, "NVME-FC{%d}: new ctrl: NQN \"%s\"\n",
@@ -3580,7 +3581,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,nvme_change_ctrl_state (&ctrl->ctrl,NVME_CTRL_DELETING );cancel_work_sync (&ctrl->ioerr_work);cancel_work_sync (&ctrl->ctrl.reset_work); -cancel_delayed_work_sync (&ctrl->connect_work); +cancel_delayed_work_sync (&ctrl->ctrl.connect_work); ctrl->ctrl.opts =NULL ;
drivers/nvme/host/nvme.h +1 -0
@@ -343,6 +343,7 @@ struct nvme_ctrl {unsigned long flags;#define NVME_CTRL_FAILFAST_EXPIRED 0struct nvmf_ctrl_options *opts; +struct delayed_work connect_work;struct page *discard_page;unsigned long discard_page_busy;
drivers/nvme/host/rdma.c +4 -6
@@ -110,8 +110,6 @@ struct nvme_rdma_ctrl {struct nvme_rdma_qe async_event_sqe; -struct delayed_work reconnect_work; -struct list_head list;struct blk_mq_tag_set admin_tag_set;
@@ -1078,7 +1076,7 @@ static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl)if (nvmf_should_reconnect (&ctrl->ctrl)) {dev_info (ctrl->ctrl.device,"Reconnecting in %d seconds...\n" , ctrl->ctrl.opts->reconnect_delay); -queue_delayed_work (nvme_wq, &ctrl->reconnect_work, +queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, ctrl->ctrl.opts->reconnect_delay *HZ ); }else {nvme_delete_ctrl (&ctrl->ctrl);
@@ -1166,7 +1164,7 @@ static int nvme_rdma_setup_ctrl(struct nvme_rdma_ctrl *ctrl, bool new)static void nvme_rdma_reconnect_ctrl_work (struct work_struct *work) {struct nvme_rdma_ctrl *ctrl =container_of (to_delayed_work (work), - struct nvme_rdma_ctrl, reconnect_work); + struct nvme_rdma_ctrl, ctrl.connect_work); ++ctrl->ctrl.nr_reconnects;
@@ -2230,7 +2228,7 @@ static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {static void nvme_rdma_shutdown_ctrl (struct nvme_rdma_ctrl *ctrl,bool shutdown) {cancel_work_sync (&ctrl->err_work); -cancel_delayed_work_sync (&ctrl->reconnect_work); +cancel_delayed_work_sync (&ctrl->ctrl.connect_work);nvme_rdma_teardown_io_queues (ctrl, shutdown);blk_mq_quiesce_queue (ctrl->ctrl.admin_q);
@@ -2358,7 +2356,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, goto out_free_ctrl; } -INIT_DELAYED_WORK (&ctrl->reconnect_work, +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_rdma_reconnect_ctrl_work);INIT_WORK (&ctrl->err_work, nvme_rdma_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work);
drivers/nvme/host/tcp.c +4 -5
@@ -127,7 +127,6 @@ struct nvme_tcp_ctrl {struct nvme_ctrl ctrl;struct work_struct err_work; -struct delayed_work connect_work;struct nvme_tcp_request async_req;u32 io_queues[HCTX_MAX_TYPES ]; };
@@ -1983,7 +1982,7 @@ static void nvme_tcp_reconnect_or_remove(struct nvme_ctrl *ctrl)if (nvmf_should_reconnect (ctrl)) {dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , ctrl->opts->reconnect_delay); -queue_delayed_work (nvme_wq, &to_tcp_ctrl (ctrl)->connect_work, +queue_delayed_work (nvme_wq, &ctrl->connect_work, ctrl->opts->reconnect_delay *HZ ); }else {dev_info (ctrl->device,"Removing controller...\n" );
@@ -2066,7 +2065,7 @@ static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new)static void nvme_tcp_reconnect_ctrl_work (struct work_struct *work) {struct nvme_tcp_ctrl *tcp_ctrl =container_of (to_delayed_work (work), - struct nvme_tcp_ctrl, connect_work); + struct nvme_tcp_ctrl, ctrl.connect_work);struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; ++ctrl->nr_reconnects;
@@ -2113,7 +2112,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work)static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown) {cancel_work_sync (&to_tcp_ctrl (ctrl)->err_work); -cancel_delayed_work_sync (&to_tcp_ctrl (ctrl)->connect_work); +cancel_delayed_work_sync (&ctrl->connect_work);nvme_tcp_teardown_io_queues (ctrl, shutdown);blk_mq_quiesce_queue (ctrl->admin_q);
@@ -2513,7 +2512,7 @@ static struct nvme_ctrl *nvme_tcp_create_ctrl(struct device *dev, ctrl->ctrl.sqsize = opts->queue_size -1 ; ctrl->ctrl.kato = opts->kato; -INIT_DELAYED_WORK (&ctrl->connect_work, +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_tcp_reconnect_ctrl_work);INIT_WORK (&ctrl->err_work, nvme_tcp_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_reset_ctrl_work);
Signature · 1 line
2.18.1
Re: [PATCH 1/7] nvme: add connect_work attribute to nvme ctrl
Re: [PATCH 1/7] nvme: add connect_work attribute to nvme ctrl
In reply to Max GurtovoyQuoted 6 lines · Max Gurtovoy
This structure is duplicated for all the fabric controllers. Move it to common code.
Reviewed-by Chaitanya Kulkarni <[email protected]>Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>
Quoted 7 lines
--- drivers/nvme/host/fc.c | 23 ++++++++++++----------- drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 10 ++++------ drivers/nvme/host/tcp.c | 9 ++++----- 4 files changed, 21 insertions(+), 22 deletions(-)
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
Re: [PATCH 1/7] nvme: add connect_work attribute to nvme ctrl
In reply to Max GurtovoyQuoted 230 lines · Max Gurtovoy
This structure is duplicated for all the fabric controllers. Move it to common code.
Reviewed-by Chaitanya Kulkarni <[email protected]>Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>---
drivers/nvme/host/fc.c | 23 ++++++++++++----------- drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 10 ++++------ drivers/nvme/host/tcp.c | 9 ++++----- 4 files changed, 21 insertions(+), 22 deletions(-)drivers/nvme/host/fc.c +12
-11@@ -167,7 +167,6 @@ struct nvme_fc_ctrl {struct blk_mq_tag_set tag_set;struct work_struct ioerr_work; -struct delayed_work connect_work;struct kref ref;unsigned long flags;@@ -567,7 +566,7 @@ nvme_fc_resume_controller(struct nvme_fc_ctrl *ctrl)"NVME-FC{%d}: connectivity re-established. " "Attempting reconnect\n", ctrl->cnum); -queue_delayed_work (nvme_wq, &ctrl->connect_work,0 ); +queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work,0 );break ;case NVME_CTRL_RESETTING :@@ -3263,7 +3262,7 @@ nvme_fc_delete_ctrl(struct nvme_ctrl *nctrl)struct nvme_fc_ctrl *ctrl =to_fc_ctrl (nctrl);cancel_work_sync (&ctrl->ioerr_work); -cancel_delayed_work_sync (&ctrl->connect_work); +cancel_delayed_work_sync (&ctrl->ctrl.connect_work);/* * kill the association on the link side. this will block * waiting for io to terminate @@ -3300,7 +3299,8 @@ nvme_fc_reconnect_or_delete(struct nvme_fc_ctrl *ctrl, int status)else if (time_after (jiffies + recon_delay, rport->dev_loss_end)) recon_delay = rport->dev_loss_end - jiffies; -queue_delayed_work (nvme_wq, &ctrl->connect_work, recon_delay); +queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, + recon_delay); }else {if (portptr->port_state ==FC_OBJSTATE_ONLINE ) {if (status >0 && (status &NVME_SC_DNR ))@@ -3340,12 +3340,13 @@ nvme_fc_reset_ctrl_work(struct work_struct *work) "to CONNECTING\n", ctrl->cnum);if (ctrl->rport->remoteport.port_state ==FC_OBJSTATE_ONLINE ) { -if (!queue_delayed_work (nvme_wq, &ctrl->connect_work,0 )) { +if (!queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, +0 )) {dev_err (ctrl->ctrl.device,"NVME-FC{%d}: failed to schedule connect " "after reset\n" , ctrl->cnum); }else { -flush_delayed_work (&ctrl->connect_work); +flush_delayed_work (&ctrl->ctrl.connect_work); } }else {nvme_fc_reconnect_or_delete (ctrl, -ENOTCONN );@@ -3373,7 +3374,7 @@ nvme_fc_connect_ctrl_work(struct work_struct *work)struct nvme_fc_ctrl *ctrl =container_of (to_delayed_work (work), - struct nvme_fc_ctrl, connect_work); + struct nvme_fc_ctrl, ctrl.connect_work); ret =nvme_fc_create_association (ctrl);if (ret)@@ -3485,7 +3486,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,kref_init (&ctrl->ref);INIT_WORK (&ctrl->ctrl.reset_work, nvme_fc_reset_ctrl_work); -INIT_DELAYED_WORK (&ctrl->connect_work, nvme_fc_connect_ctrl_work); +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_fc_connect_ctrl_work);INIT_WORK (&ctrl->ioerr_work, nvme_fc_ctrl_ioerr_work);spin_lock_init (&ctrl->lock);@@ -3561,14 +3562,14 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, goto fail_ctrl; } -if (!queue_delayed_work (nvme_wq, &ctrl->connect_work,0 )) { +if (!queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work,0 )) {dev_err (ctrl->ctrl.device,"NVME-FC{%d}: failed to schedule initial connect\n" , ctrl->cnum); goto fail_ctrl; } -flush_delayed_work (&ctrl->connect_work); +flush_delayed_work (&ctrl->ctrl.connect_work); dev_info(ctrl->ctrl.device, "NVME-FC{%d}: new ctrl: NQN \"%s\"\n",@@ -3580,7 +3581,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,nvme_change_ctrl_state (&ctrl->ctrl,NVME_CTRL_DELETING );cancel_work_sync (&ctrl->ioerr_work);cancel_work_sync (&ctrl->ctrl.reset_work); -cancel_delayed_work_sync (&ctrl->connect_work); +cancel_delayed_work_sync (&ctrl->ctrl.connect_work); ctrl->ctrl.opts =NULL ;drivers/nvme/host/nvme.h +1
-0@@ -343,6 +343,7 @@ struct nvme_ctrl {unsigned long flags;#define NVME_CTRL_FAILFAST_EXPIRED 0struct nvmf_ctrl_options *opts; +struct delayed_work connect_work;struct page *discard_page;unsigned long discard_page_busy;drivers/nvme/host/rdma.c +4
-6@@ -110,8 +110,6 @@ struct nvme_rdma_ctrl {struct nvme_rdma_qe async_event_sqe; -struct delayed_work reconnect_work; -struct list_head list;struct blk_mq_tag_set admin_tag_set;@@ -1078,7 +1076,7 @@ static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl)if (nvmf_should_reconnect (&ctrl->ctrl)) {dev_info (ctrl->ctrl.device,"Reconnecting in %d seconds...\n" , ctrl->ctrl.opts->reconnect_delay); -queue_delayed_work (nvme_wq, &ctrl->reconnect_work, +queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, ctrl->ctrl.opts->reconnect_delay *HZ ); }else {nvme_delete_ctrl (&ctrl->ctrl);@@ -1166,7 +1164,7 @@ static int nvme_rdma_setup_ctrl(struct nvme_rdma_ctrl *ctrl, bool new)static void nvme_rdma_reconnect_ctrl_work (struct work_struct *work) {struct nvme_rdma_ctrl *ctrl =container_of (to_delayed_work (work), - struct nvme_rdma_ctrl, reconnect_work); + struct nvme_rdma_ctrl, ctrl.connect_work); ++ctrl->ctrl.nr_reconnects;@@ -2230,7 +2228,7 @@ static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {static void nvme_rdma_shutdown_ctrl (struct nvme_rdma_ctrl *ctrl,bool shutdown) {cancel_work_sync (&ctrl->err_work); -cancel_delayed_work_sync (&ctrl->reconnect_work); +cancel_delayed_work_sync (&ctrl->ctrl.connect_work);nvme_rdma_teardown_io_queues (ctrl, shutdown);blk_mq_quiesce_queue (ctrl->ctrl.admin_q);@@ -2358,7 +2356,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, goto out_free_ctrl; } -INIT_DELAYED_WORK (&ctrl->reconnect_work, +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_rdma_reconnect_ctrl_work);INIT_WORK (&ctrl->err_work, nvme_rdma_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work);drivers/nvme/host/tcp.c +4
-5@@ -127,7 +127,6 @@ struct nvme_tcp_ctrl {struct nvme_ctrl ctrl;struct work_struct err_work; -struct delayed_work connect_work;struct nvme_tcp_request async_req;u32 io_queues[HCTX_MAX_TYPES ]; };@@ -1983,7 +1982,7 @@ static void nvme_tcp_reconnect_or_remove(struct nvme_ctrl *ctrl)if (nvmf_should_reconnect (ctrl)) {dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , ctrl->opts->reconnect_delay); -queue_delayed_work (nvme_wq, &to_tcp_ctrl (ctrl)->connect_work, +queue_delayed_work (nvme_wq, &ctrl->connect_work, ctrl->opts->reconnect_delay *HZ ); }else {dev_info (ctrl->device,"Removing controller...\n" );@@ -2066,7 +2065,7 @@ static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new)static void nvme_tcp_reconnect_ctrl_work (struct work_struct *work) {struct nvme_tcp_ctrl *tcp_ctrl =container_of (to_delayed_work (work), - struct nvme_tcp_ctrl, connect_work); + struct nvme_tcp_ctrl, ctrl.connect_work);struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; ++ctrl->nr_reconnects;@@ -2113,7 +2112,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work)static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown) {cancel_work_sync (&to_tcp_ctrl (ctrl)->err_work); -cancel_delayed_work_sync (&to_tcp_ctrl (ctrl)->connect_work); +cancel_delayed_work_sync (&ctrl->connect_work);nvme_tcp_teardown_io_queues (ctrl, shutdown);blk_mq_quiesce_queue (ctrl->admin_q);@@ -2513,7 +2512,7 @@ static struct nvme_ctrl *nvme_tcp_create_ctrl(struct device *dev, ctrl->ctrl.sqsize = opts->queue_size -1 ; ctrl->ctrl.kato = opts->kato; -INIT_DELAYED_WORK (&ctrl->connect_work, +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_tcp_reconnect_ctrl_work);INIT_WORK (&ctrl->err_work, nvme_tcp_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_reset_ctrl_work);Signature · 1 line
2.18.1
--
Himanshu Madhani Oracle Linux Engineering
PATCH 2/7 nvme-fabrics: introduce nvmf_reconnect_or_remove API
In reply to Max GurtovoyThis logic is duplicated today for RDMA and TCP controllers. Move it to the fabrics driver and export it as a new API.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
--- drivers/nvme/host/fabrics.c | 21 +++++++++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 +++---------------------- drivers/nvme/host/tcp.c | 26 +++----------------------- 4 files changed, 28 insertions(+), 45 deletions(-)
drivers/nvme/host/fabrics.c +21 -0
@@ -472,6 +472,27 @@ bool nvmf_should_reconnect(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_should_reconnect); +void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl) +{ +/* If we are resetting/deleting then do nothing */ +if (ctrl->state !=NVME_CTRL_CONNECTING ) { +WARN_ON_ONCE (ctrl->state ==NVME_CTRL_NEW || + ctrl->state ==NVME_CTRL_LIVE ); +return ; + } + +if (nvmf_should_reconnect (ctrl)) { +dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , + ctrl->opts->reconnect_delay); +queue_delayed_work (nvme_wq, &ctrl->connect_work, + ctrl->opts->reconnect_delay *HZ ); + }else { +dev_info (ctrl->device,"Removing controller...\n" ); +nvme_delete_ctrl (ctrl); + } +} +EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +/** * nvmf_register_transport() - NVMe Fabrics Library registration function. * @ops: Transport ops instance to be registered to the
drivers/nvme/host/fabrics.h +1 -0
@@ -188,6 +188,7 @@ void nvmf_unregister_transport(struct nvmf_transport_ops *ops);void nvmf_free_options (struct nvmf_ctrl_options *opts);int nvmf_get_address (struct nvme_ctrl *ctrl,char *buf,int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl); +void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);
drivers/nvme/host/rdma.c +3 -22
@@ -1064,25 +1064,6 @@ static void nvme_rdma_free_ctrl(struct nvme_ctrl *nctrl)kfree (ctrl); } -static void nvme_rdma_reconnect_or_remove (struct nvme_rdma_ctrl *ctrl) -{ -/* If we are resetting/deleting then do nothing */ -if (ctrl->ctrl.state !=NVME_CTRL_CONNECTING ) { -WARN_ON_ONCE (ctrl->ctrl.state ==NVME_CTRL_NEW || - ctrl->ctrl.state ==NVME_CTRL_LIVE ); -return ; - } - -if (nvmf_should_reconnect (&ctrl->ctrl)) { -dev_info (ctrl->ctrl.device,"Reconnecting in %d seconds...\n" , - ctrl->ctrl.opts->reconnect_delay); -queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, - ctrl->ctrl.opts->reconnect_delay *HZ ); - }else { -nvme_delete_ctrl (&ctrl->ctrl); - } -} -static int nvme_rdma_setup_ctrl (struct nvme_rdma_ctrl *ctrl,bool new) {int ret;
@@ -1181,7 +1162,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) requeue:dev_info (ctrl->ctrl.device,"Failed reconnect attempt %d\n" , ctrl->ctrl.nr_reconnects); -nvme_rdma_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (&ctrl->ctrl); }static void nvme_rdma_error_recovery_work (struct work_struct *work)
@@ -1202,7 +1183,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)return ; } -nvme_rdma_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (&ctrl->ctrl); }static void nvme_rdma_error_recovery (struct nvme_rdma_ctrl *ctrl)
@@ -2265,7 +2246,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) out_fail: ++ctrl->ctrl.nr_reconnects; -nvme_rdma_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (&ctrl->ctrl); }static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = {
drivers/nvme/host/tcp.c +3 -23
@@ -1970,26 +1970,6 @@ static void nvme_tcp_teardown_io_queues(struct nvme_ctrl *ctrl,nvme_tcp_destroy_io_queues (ctrl, remove); } -static void nvme_tcp_reconnect_or_remove (struct nvme_ctrl *ctrl) -{ -/* If we are resetting/deleting then do nothing */ -if (ctrl->state !=NVME_CTRL_CONNECTING ) { -WARN_ON_ONCE (ctrl->state ==NVME_CTRL_NEW || - ctrl->state ==NVME_CTRL_LIVE ); -return ; - } - -if (nvmf_should_reconnect (ctrl)) { -dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , - ctrl->opts->reconnect_delay); -queue_delayed_work (nvme_wq, &ctrl->connect_work, - ctrl->opts->reconnect_delay *HZ ); - }else { -dev_info (ctrl->device,"Removing controller...\n" ); -nvme_delete_ctrl (ctrl); - } -} -static int nvme_tcp_setup_ctrl (struct nvme_ctrl *ctrl,bool new) {struct nvmf_ctrl_options *opts = ctrl->opts;
@@ -2083,7 +2063,7 @@ static void nvme_tcp_reconnect_ctrl_work(struct work_struct *work) requeue:dev_info (ctrl->device,"Failed reconnect attempt %d\n" , ctrl->nr_reconnects); -nvme_tcp_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (ctrl); }static void nvme_tcp_error_recovery_work (struct work_struct *work)
@@ -2106,7 +2086,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work)return ; } -nvme_tcp_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (ctrl); }static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown)
@@ -2150,7 +2130,7 @@ static void nvme_reset_ctrl_work(struct work_struct *work) out_fail: ++ctrl->nr_reconnects; -nvme_tcp_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (ctrl); }static void nvme_tcp_free_ctrl (struct nvme_ctrl *nctrl)
Signature · 1 line
2.18.1
Re: [PATCH 2/7] nvme-fabrics: introduce nvmf_reconnect_or_remove API
Quoted 8 lines · Max Gurtovoy
This logic is duplicated today for RDMA and TCP controllers. Move it to the fabrics driver and export it as a new API.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>
Looks good.
Sagi Grimberg
maintainerRe: [PATCH 2/7] nvme-fabrics: introduce nvmf_reconnect_or_remove API
In reply to Max GurtovoyQuoted 32 lines
This logic is duplicated today for RDMA and TCP controllers. Move it to the fabrics driver and export it as a new API.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>--- drivers/nvme/host/fabrics.c | 21 +++++++++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 +++---------------------- drivers/nvme/host/tcp.c | 26 +++----------------------- 4 files changed, 28 insertions(+), 45 deletions(-)drivers/nvme/host/fabrics.c +8
-0@@ -472,6 +472,27 @@ bool nvmf_should_reconnect(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_should_reconnect); +void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl) +{ +/* If we are resetting/deleting then do nothing */ +if (ctrl->state !=NVME_CTRL_CONNECTING ) { +WARN_ON_ONCE (ctrl->state ==NVME_CTRL_NEW || + ctrl->state ==NVME_CTRL_LIVE ); +return ; + }
drivers/nvme/host/fabrics.c bool nvmf_should_reconnect(struct nvme_ctrl *ctrl) · line 492
Quoted 10 lines
drivers/nvme/host/fabrics.c +10
-0+ +if (nvmf_should_reconnect (ctrl)) { +dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , + ctrl->opts->reconnect_delay); +queue_delayed_work (nvme_wq, &ctrl->connect_work, + ctrl->opts->reconnect_delay *HZ ); + }else { +dev_info (ctrl->device,"Removing controller...\n" ); +nvme_delete_ctrl (ctrl); + }
We need to look at fc here as it does this differently. Can it converge somehow? if not we can add a callback for a driver to override it.
Re: [PATCH 2/7] nvme-fabrics: introduce nvmf_reconnect_or_remove API
Quoted 37 lines · Sagi Grimberg
Quoted 37 lines
This logic is duplicated today for RDMA and TCP controllers. Move it to the fabrics driver and export it as a new API.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>---
drivers/nvme/host/fabrics.c| 21 +++++++++++++++++++++drivers/nvme/host/fabrics.h| 1 + drivers/nvme/host/rdma.c | 25 +++---------------------- drivers/nvme/host/tcp.c | 26 +++-----------------------4 files changed, 28 insertions(+), 45 deletions(-)drivers/nvme/host/fabrics.c +13
-0@@ -472,6 +472,27 @@ bool nvmf_should_reconnect(struct nvme_ctrl *ctrl) } EXPORT_SYMBOL_GPL(nvmf_should_reconnect); +void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) +{ + /* If we are resetting/deleting then do nothing */ + if (ctrl->state != NVME_CTRL_CONNECTING) { + WARN_ON_ONCE(ctrl->state == NVME_CTRL_NEW || + ctrl->state == NVME_CTRL_LIVE); + return; + } + + if (nvmf_should_reconnect(ctrl)) { + dev_info(ctrl->device, "Reconnecting in %d seconds...\n", + ctrl->opts->reconnect_delay); + queue_delayed_work(nvme_wq, &ctrl->connect_work, + ctrl->opts->reconnect_delay * HZ);
Quoted 8 lines
drivers/nvme/host/fabrics.c bool nvmf_should_reconnect(struct nvme_ctrl *ctrl) · line 492
Quoted 4 lines
drivers/nvme/host/fabrics.c +4
-0+ } else { + dev_info(ctrl->device, "Removing controller...\n"); + nvme_delete_ctrl(ctrl); + }We need to look at fc here as it does this differently. Can it converge somehow? if not we can add a callback for a driver to override it.
we need to start somewhere.
Adding a callback to override this logic is not a problem and i thought about it but decided not to do so.
James and other FC developers evaluate moving FC code using more core functionality to ease on this driver.
This work can be done also after merging this series.
This series is not related to that.
Re: [PATCH 2/7] nvme-fabrics: introduce nvmf_reconnect_or_remove API
In reply to Max GurtovoyQuoted 6 lines · Max Gurtovoy
This logic is duplicated today for RDMA and TCP controllers. Move it to the fabrics driver and export it as a new API.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Quoted 9 lines
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>--- drivers/nvme/host/fabrics.c | 21 +++++++++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 +++---------------------- drivers/nvme/host/tcp.c | 26 +++----------------------- 4 files changed, 28 insertions(+), 45 deletions(-)
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
Re: [PATCH 2/7] nvme-fabrics: introduce nvmf_reconnect_or_remove API
In reply to Max GurtovoyQuoted 177 lines · Max Gurtovoy
This logic is duplicated today for RDMA and TCP controllers. Move it to the fabrics driver and export it as a new API.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>---
drivers/nvme/host/fabrics.c | 21 +++++++++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 +++---------------------- drivers/nvme/host/tcp.c | 26 +++----------------------- 4 files changed, 28 insertions(+), 45 deletions(-)drivers/nvme/host/fabrics.c +21
-0@@ -472,6 +472,27 @@ bool nvmf_should_reconnect(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_should_reconnect); +void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl) +{ +/* If we are resetting/deleting then do nothing */ +if (ctrl->state !=NVME_CTRL_CONNECTING ) { +WARN_ON_ONCE (ctrl->state ==NVME_CTRL_NEW || + ctrl->state ==NVME_CTRL_LIVE ); +return ; + } + +if (nvmf_should_reconnect (ctrl)) { +dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , + ctrl->opts->reconnect_delay); +queue_delayed_work (nvme_wq, &ctrl->connect_work, + ctrl->opts->reconnect_delay *HZ ); + }else { +dev_info (ctrl->device,"Removing controller...\n" ); +nvme_delete_ctrl (ctrl); + } +} +EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +/** * nvmf_register_transport() - NVMe Fabrics Library registration function. * @ops: Transport ops instance to be registered to the drivers/nvme/host/fabrics.h +1
-0@@ -188,6 +188,7 @@ void nvmf_unregister_transport(struct nvmf_transport_ops *ops);void nvmf_free_options (struct nvmf_ctrl_options *opts);int nvmf_get_address (struct nvme_ctrl *ctrl,char *buf,int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl); +void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);drivers/nvme/host/rdma.c +0
-19@@ -1064,25 +1064,6 @@ static void nvme_rdma_free_ctrl(struct nvme_ctrl *nctrl)kfree (ctrl); } -static void nvme_rdma_reconnect_or_remove (struct nvme_rdma_ctrl *ctrl) -{ -/* If we are resetting/deleting then do nothing */ -if (ctrl->ctrl.state !=NVME_CTRL_CONNECTING ) { -WARN_ON_ONCE (ctrl->ctrl.state ==NVME_CTRL_NEW || - ctrl->ctrl.state ==NVME_CTRL_LIVE ); -return ; - } - -if (nvmf_should_reconnect (&ctrl->ctrl)) { -dev_info (ctrl->ctrl.device,"Reconnecting in %d seconds...\n" , - ctrl->ctrl.opts->reconnect_delay); -queue_delayed_work (nvme_wq, &ctrl->ctrl.connect_work, - ctrl->ctrl.opts->reconnect_delay *HZ ); - }else { -nvme_delete_ctrl (&ctrl->ctrl); - } -} -static int nvme_rdma_setup_ctrl (struct nvme_rdma_ctrl *ctrl,bool new) {int ret;@@ -1181,7 +1162,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)requeue:
dev_info(ctrl->ctrl.device, "Failed reconnect attempt %d\n", ctrl->ctrl.nr_reconnects);drivers/nvme/host/rdma.c +1
-1-nvme_rdma_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (&ctrl->ctrl);}
static void
nvme_rdma_error_recovery_work(struct work_struct *work)drivers/nvme/host/rdma.c +1
-1@@ -1202,7 +1183,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)return ; } -nvme_rdma_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (&ctrl->ctrl); }static void nvme_rdma_error_recovery (struct nvme_rdma_ctrl *ctrl)@@ -2265,7 +2246,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
out_fail:++ctrl->ctrl.nr_reconnects;drivers/nvme/host/rdma.c +1
-1-nvme_rdma_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (&ctrl->ctrl);}
static const struct
nvme_ctrl_opsnvme_rdma_ctrl_ops= {drivers/nvme/host/tcp.c +0
-20@@ -1970,26 +1970,6 @@ static void nvme_tcp_teardown_io_queues(struct nvme_ctrl *ctrl,nvme_tcp_destroy_io_queues (ctrl, remove); } -static void nvme_tcp_reconnect_or_remove (struct nvme_ctrl *ctrl) -{ -/* If we are resetting/deleting then do nothing */ -if (ctrl->state !=NVME_CTRL_CONNECTING ) { -WARN_ON_ONCE (ctrl->state ==NVME_CTRL_NEW || - ctrl->state ==NVME_CTRL_LIVE ); -return ; - } - -if (nvmf_should_reconnect (ctrl)) { -dev_info (ctrl->device,"Reconnecting in %d seconds...\n" , - ctrl->opts->reconnect_delay); -queue_delayed_work (nvme_wq, &ctrl->connect_work, - ctrl->opts->reconnect_delay *HZ ); - }else { -dev_info (ctrl->device,"Removing controller...\n" ); -nvme_delete_ctrl (ctrl); - } -} -static int nvme_tcp_setup_ctrl (struct nvme_ctrl *ctrl,bool new) {struct nvmf_ctrl_options *opts = ctrl->opts;@@ -2083,7 +2063,7 @@ static void nvme_tcp_reconnect_ctrl_work(struct work_struct *work)requeue:
dev_info(ctrl->device, "Failed reconnect attempt %d\n", ctrl->nr_reconnects);drivers/nvme/host/tcp.c +1
-1-nvme_tcp_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (ctrl);}
static void
nvme_tcp_error_recovery_work(struct work_struct *work)drivers/nvme/host/tcp.c +1
-1@@ -2106,7 +2086,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work)return ; } -nvme_tcp_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (ctrl); }static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown)@@ -2150,7 +2130,7 @@ static void nvme_reset_ctrl_work(struct work_struct *work)
out_fail:++ctrl->nr_reconnects;drivers/nvme/host/tcp.c +1
-1-nvme_tcp_reconnect_or_remove (ctrl); +nvmf_reconnect_or_remove (ctrl);}
Quoted 5 lines
static void
nvme_tcp_free_ctrl(struct nvme_ctrl *nctrl)Signature · 1 line
2.18.1
--
Himanshu Madhani Oracle Linux Engineering
PATCH 3/7 nvme: add err_work attribute to nvme ctrl
In reply to Max GurtovoyThis structure is duplicated for RDMA/TCP fabric controllers. Move it to common code.
FC controllers might use this attribute in the future instead of a local ioerr_work attribute.
--- drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 9 ++++----- drivers/nvme/host/tcp.c | 12 +++++------- 3 files changed, 10 insertions(+), 12 deletions(-)
drivers/nvme/host/nvme.h +1 -0
@@ -344,6 +344,7 @@ struct nvme_ctrl {#define NVME_CTRL_FAILFAST_EXPIRED 0struct nvmf_ctrl_options *opts;struct delayed_work connect_work; +struct work_struct err_work;struct page *discard_page;unsigned long discard_page_busy;
drivers/nvme/host/rdma.c +4 -5
@@ -106,7 +106,6 @@ struct nvme_rdma_ctrl {/* other member variables */ struct blk_mq_tag_set tag_set; -struct work_struct err_work;struct nvme_rdma_qe async_event_sqe;
@@ -1168,7 +1167,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)static void nvme_rdma_error_recovery_work (struct work_struct *work) {struct nvme_rdma_ctrl *ctrl =container_of (work, - struct nvme_rdma_ctrl, err_work); + struct nvme_rdma_ctrl, ctrl.err_work);nvme_stop_keep_alive (&ctrl->ctrl);nvme_rdma_teardown_io_queues (ctrl, false);
@@ -1192,7 +1191,7 @@ static void nvme_rdma_error_recovery(struct nvme_rdma_ctrl *ctrl)return ;dev_warn (ctrl->ctrl.device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &ctrl->err_work); +queue_work (nvme_reset_wq, &ctrl->ctrl.err_work); }static void nvme_rdma_end_request (struct nvme_rdma_request *req)
@@ -2208,7 +2207,7 @@ static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {static void nvme_rdma_shutdown_ctrl (struct nvme_rdma_ctrl *ctrl,bool shutdown) { -cancel_work_sync (&ctrl->err_work); +cancel_work_sync (&ctrl->ctrl.err_work);cancel_delayed_work_sync (&ctrl->ctrl.connect_work);nvme_rdma_teardown_io_queues (ctrl, shutdown);
@@ -2339,7 +2338,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev,INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_rdma_reconnect_ctrl_work); -INIT_WORK (&ctrl->err_work, nvme_rdma_error_recovery_work); +INIT_WORK (&ctrl->ctrl.err_work, nvme_rdma_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work); ctrl->ctrl.queue_count = opts->nr_io_queues + opts->nr_write_queues +
drivers/nvme/host/tcp.c +5 -7
@@ -126,7 +126,6 @@ struct nvme_tcp_ctrl {struct sockaddr_storage src_addr;struct nvme_ctrl ctrl; -struct work_struct err_work;struct nvme_tcp_request async_req;u32 io_queues[HCTX_MAX_TYPES ]; };
@@ -486,7 +485,7 @@ static void nvme_tcp_error_recovery(struct nvme_ctrl *ctrl)return ;dev_warn (ctrl->device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &to_tcp_ctrl (ctrl)->err_work); +queue_work (nvme_reset_wq, &ctrl->err_work); } static int nvme_tcp_process_nvme_cqe(struct nvme_tcp_queue *queue,
@@ -2068,9 +2067,8 @@ static void nvme_tcp_reconnect_ctrl_work(struct work_struct *work)static void nvme_tcp_error_recovery_work (struct work_struct *work) { -struct nvme_tcp_ctrl *tcp_ctrl =container_of (work, - struct nvme_tcp_ctrl, err_work); -struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; +struct nvme_ctrl *ctrl =container_of (work, + struct nvme_ctrl, err_work);nvme_stop_keep_alive (ctrl);nvme_tcp_teardown_io_queues (ctrl, false);
@@ -2091,7 +2089,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work)static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown) { -cancel_work_sync (&to_tcp_ctrl (ctrl)->err_work); +cancel_work_sync (&ctrl->err_work);cancel_delayed_work_sync (&ctrl->connect_work);nvme_tcp_teardown_io_queues (ctrl, shutdown);
@@ -2494,7 +2492,7 @@ static struct nvme_ctrl *nvme_tcp_create_ctrl(struct device *dev,INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_tcp_reconnect_ctrl_work); -INIT_WORK (&ctrl->err_work, nvme_tcp_error_recovery_work); +INIT_WORK (&ctrl->ctrl.err_work, nvme_tcp_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_reset_ctrl_work);if (!(opts->mask &NVMF_OPT_TRSVCID )) {
Signature · 1 line
2.18.1
Re: [PATCH 3/7] nvme: add err_work attribute to nvme ctrl
Re: [PATCH 3/7] nvme: add err_work attribute to nvme ctrl
In reply to Max GurtovoyQuoted 134 lines · Max Gurtovoy
This structure is duplicated for RDMA/TCP fabric controllers. Move it to common code.
FC controllers might use this attribute in the future instead of a local
ioerr_workattribute.Reviewed-by Chaitanya Kulkarni <[email protected]>Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>---
drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 9 ++++----- drivers/nvme/host/tcp.c | 12 +++++------- 3 files changed, 10 insertions(+), 12 deletions(-)drivers/nvme/host/nvme.h +1
-0@@ -344,6 +344,7 @@ struct nvme_ctrl {#define NVME_CTRL_FAILFAST_EXPIRED 0struct nvmf_ctrl_options *opts;struct delayed_work connect_work; +struct work_struct err_work;struct page *discard_page;unsigned long discard_page_busy;drivers/nvme/host/rdma.c +4
-5@@ -106,7 +106,6 @@ struct nvme_rdma_ctrl {/* other member variables */ struct blk_mq_tag_set tag_set; -struct work_struct err_work;struct nvme_rdma_qe async_event_sqe;@@ -1168,7 +1167,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)static void nvme_rdma_error_recovery_work (struct work_struct *work) {struct nvme_rdma_ctrl *ctrl =container_of (work, - struct nvme_rdma_ctrl, err_work); + struct nvme_rdma_ctrl, ctrl.err_work);nvme_stop_keep_alive (&ctrl->ctrl);nvme_rdma_teardown_io_queues (ctrl, false);@@ -1192,7 +1191,7 @@ static void nvme_rdma_error_recovery(struct nvme_rdma_ctrl *ctrl)return ;dev_warn (ctrl->ctrl.device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &ctrl->err_work); +queue_work (nvme_reset_wq, &ctrl->ctrl.err_work); }static void nvme_rdma_end_request (struct nvme_rdma_request *req)@@ -2208,7 +2207,7 @@ static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {static void nvme_rdma_shutdown_ctrl (struct nvme_rdma_ctrl *ctrl,bool shutdown) { -cancel_work_sync (&ctrl->err_work); +cancel_work_sync (&ctrl->ctrl.err_work);cancel_delayed_work_sync (&ctrl->ctrl.connect_work);nvme_rdma_teardown_io_queues (ctrl, shutdown);@@ -2339,7 +2338,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev,INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_rdma_reconnect_ctrl_work); -INIT_WORK (&ctrl->err_work, nvme_rdma_error_recovery_work); +INIT_WORK (&ctrl->ctrl.err_work, nvme_rdma_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work); ctrl->ctrl.queue_count = opts->nr_io_queues + opts->nr_write_queues +drivers/nvme/host/tcp.c +5
-7@@ -126,7 +126,6 @@ struct nvme_tcp_ctrl {struct sockaddr_storage src_addr;struct nvme_ctrl ctrl; -struct work_struct err_work;struct nvme_tcp_request async_req;u32 io_queues[HCTX_MAX_TYPES ]; };@@ -486,7 +485,7 @@ static void nvme_tcp_error_recovery(struct nvme_ctrl *ctrl)return ;dev_warn (ctrl->device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &to_tcp_ctrl (ctrl)->err_work); +queue_work (nvme_reset_wq, &ctrl->err_work); } static int nvme_tcp_process_nvme_cqe(struct nvme_tcp_queue *queue,@@ -2068,9 +2067,8 @@ static void nvme_tcp_reconnect_ctrl_work(struct work_struct *work)static void nvme_tcp_error_recovery_work (struct work_struct *work) { -struct nvme_tcp_ctrl *tcp_ctrl =container_of (work, - struct nvme_tcp_ctrl, err_work); -struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; +struct nvme_ctrl *ctrl =container_of (work, + struct nvme_ctrl, err_work);nvme_stop_keep_alive (ctrl);nvme_tcp_teardown_io_queues (ctrl, false);@@ -2091,7 +2089,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work)static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown) { -cancel_work_sync (&to_tcp_ctrl (ctrl)->err_work); +cancel_work_sync (&ctrl->err_work);cancel_delayed_work_sync (&ctrl->connect_work);nvme_tcp_teardown_io_queues (ctrl, shutdown);@@ -2494,7 +2492,7 @@ static struct nvme_ctrl *nvme_tcp_create_ctrl(struct device *dev,INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_tcp_reconnect_ctrl_work); -INIT_WORK (&ctrl->err_work, nvme_tcp_error_recovery_work); +INIT_WORK (&ctrl->ctrl.err_work, nvme_tcp_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_reset_ctrl_work);if (!(opts->mask &NVMF_OPT_TRSVCID )) {Signature · 1 line
2.18.1
--
Himanshu Madhani Oracle Linux Engineering
PATCH 4/7 nvme-fabrics: introduce nvmf_error_recovery API
In reply to Max GurtovoyError recovery mechanism is duplicated in RDMA and TCP transports. Move this logic to common code.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
--- drivers/nvme/host/fabrics.c | 10 ++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 ++++++++----------------- drivers/nvme/host/tcp.c | 19 +++++-------------- 4 files changed, 24 insertions(+), 31 deletions(-)
drivers/nvme/host/fabrics.c +10 -0
@@ -493,6 +493,16 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +void nvmf_error_recovery (struct nvme_ctrl *ctrl) +{ +if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING )) +return ; + +dev_warn (ctrl->device,"starting error recovery\n" ); +queue_work (nvme_reset_wq, &ctrl->err_work); +} +EXPORT_SYMBOL_GPL (nvmf_error_recovery); +/** * nvmf_register_transport() - NVMe Fabrics Library registration function. * @ops: Transport ops instance to be registered to the
drivers/nvme/host/fabrics.h +1 -0
@@ -189,6 +189,7 @@ void nvmf_free_options(struct nvmf_ctrl_options *opts);int nvmf_get_address (struct nvme_ctrl *ctrl,char *buf,int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl); +void nvmf_error_recovery (struct nvme_ctrl *ctrl);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);
drivers/nvme/host/rdma.c +8 -17
@@ -1185,15 +1185,6 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)nvmf_reconnect_or_remove (&ctrl->ctrl); } -static void nvme_rdma_error_recovery (struct nvme_rdma_ctrl *ctrl) -{ -if (!nvme_change_ctrl_state (&ctrl->ctrl,NVME_CTRL_RESETTING )) -return ; - -dev_warn (ctrl->ctrl.device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &ctrl->ctrl.err_work); -} -static void nvme_rdma_end_request (struct nvme_rdma_request *req) {struct request *rq =blk_mq_rq_from_pdu (req);
@@ -1215,7 +1206,7 @@ static void nvme_rdma_wr_error(struct ib_cq *cq, struct ib_wc *wc,"%s for CQE 0x%p failed with status %s (%d)\n" , op, wc->wr_cqe, ib_wc_status_msg(wc->status), wc->status); -nvme_rdma_error_recovery (ctrl); +nvmf_error_recovery (&ctrl->ctrl); }static void nvme_rdma_memreg_done (struct ib_cq *cq,struct ib_wc *wc)
@@ -1715,7 +1706,7 @@ static void nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,dev_err (queue->ctrl->ctrl.device,"got bad command_id %#x on QP %#x\n" , cqe->command_id, queue->qp->qp_num); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return ; } req =blk_mq_rq_to_pdu (rq);
@@ -1729,7 +1720,7 @@ static void nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,dev_err (queue->ctrl->ctrl.device,"Bogus remote invalidation for rkey %#x\n" , req->mr ? req->mr->rkey :0 ); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl); } }else if (req->mr) {int ret;
@@ -1739,7 +1730,7 @@ static void nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,dev_err (queue->ctrl->ctrl.device,"Queueing INV WR for rkey %#x failed (%d)\n" , req->mr->rkey, ret); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl); }/* the local invalidation completion will end the request */ return ;
@@ -1766,7 +1757,7 @@ static void nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)if (unlikely (wc->byte_len < len)) {dev_err (queue->ctrl->ctrl.device,"Unexpected nvme completion length(%d)\n" , wc->byte_len); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return ; }
@@ -1936,7 +1927,7 @@ static int nvme_rdma_cm_handler(struct rdma_cm_id *cm_id,case RDMA_CM_EVENT_TIMEWAIT_EXIT :dev_dbg (queue->ctrl->ctrl.device,"disconnect received - connection closed\n" ); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);break ;case RDMA_CM_EVENT_DEVICE_REMOVAL :/* device removal is handled via the ib_client API */
@@ -1944,7 +1935,7 @@ static int nvme_rdma_cm_handler(struct rdma_cm_id *cm_id,default :dev_err (queue->ctrl->ctrl.device,"Unexpected RDMA CM event (%d)\n" , ev->event); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);break ; }
@@ -2000,7 +1991,7 @@ nvme_rdma_timeout(struct request *rq, bool reserved)* LIVE state should trigger the normal error recovery which will * handle completing this request. */ -nvme_rdma_error_recovery (ctrl); +nvmf_error_recovery (&ctrl->ctrl);return BLK_EH_RESET_TIMER ; }
drivers/nvme/host/tcp.c +5 -14
@@ -479,15 +479,6 @@ static void nvme_tcp_init_recv_ctx(struct nvme_tcp_queue *queue) queue->ddgst_remaining =0 ; } -static void nvme_tcp_error_recovery (struct nvme_ctrl *ctrl) -{ -if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING )) -return ; - -dev_warn (ctrl->device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &ctrl->err_work); -} -static int nvme_tcp_process_nvme_cqe (struct nvme_tcp_queue *queue,struct nvme_completion *cqe) {
@@ -499,7 +490,7 @@ static int nvme_tcp_process_nvme_cqe(struct nvme_tcp_queue *queue,dev_err (queue->ctrl->ctrl.device,"got bad cqe.command_id %#x on queue %d\n" , cqe->command_id,nvme_tcp_queue_id (queue)); -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return -EINVAL ; }
@@ -541,7 +532,7 @@ static int nvme_tcp_handle_c2h_data(struct nvme_tcp_queue *queue,dev_err (queue->ctrl->ctrl.device,"queue %d tag %#x SUCCESS set but not last PDU\n" ,nvme_tcp_queue_id (queue), rq->tag); -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return -EPROTO ; }
@@ -850,7 +841,7 @@ static int nvme_tcp_recv_skb(read_descriptor_t *desc, struct sk_buff *skb,dev_err (queue->ctrl->ctrl.device,"receive failed: %d\n" , result); queue->rd_enabled = false; -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return result; } }
@@ -898,7 +889,7 @@ static void nvme_tcp_state_change(struct sock *sk)case TCP_LAST_ACK :case TCP_FIN_WAIT1 :case TCP_FIN_WAIT2 : -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);break ;default : dev_info(queue->ctrl->ctrl.device,
@@ -2252,7 +2243,7 @@ nvme_tcp_timeout(struct request *rq, bool reserved)* LIVE state should trigger the normal error recovery which will * handle completing this request. */ -nvme_tcp_error_recovery (ctrl); +nvmf_error_recovery (ctrl);return BLK_EH_RESET_TIMER ; }
Signature · 1 line
2.18.1
Re: [PATCH 4/7] nvme-fabrics: introduce nvmf_error_recovery API
Quoted 9 lines · Max Gurtovoy
Error recovery mechanism is duplicated in RDMA and TCP transports. Move this logic to common code.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Chaitanya Kulkarni <[email protected]>Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>
Quoted 7 lines
--- drivers/nvme/host/fabrics.c | 10 ++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 ++++++++----------------- drivers/nvme/host/tcp.c | 19 +++++-------------- 4 files changed, 24 insertions(+), 31 deletions(-)
_Technically_ we shouldn't start reconnecting for command errors; reconnection only should start once KATO expires.
But that might be a topic for a different discussion.
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
Re: [PATCH 4/7] nvme-fabrics: introduce nvmf_error_recovery API
In reply to Max GurtovoyQuoted 214 lines · Max Gurtovoy
Error recovery mechanism is duplicated in RDMA and TCP transports. Move this logic to common code.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Chaitanya Kulkarni <[email protected]>Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>---
drivers/nvme/host/fabrics.c | 10 ++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/rdma.c | 25 ++++++++----------------- drivers/nvme/host/tcp.c | 19 +++++-------------- 4 files changed, 24 insertions(+), 31 deletions(-)drivers/nvme/host/fabrics.c +10
-0@@ -493,6 +493,16 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +void nvmf_error_recovery (struct nvme_ctrl *ctrl) +{ +if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING )) +return ; + +dev_warn (ctrl->device,"starting error recovery\n" ); +queue_work (nvme_reset_wq, &ctrl->err_work); +} +EXPORT_SYMBOL_GPL (nvmf_error_recovery); +/** * nvmf_register_transport() - NVMe Fabrics Library registration function. * @ops: Transport ops instance to be registered to the drivers/nvme/host/fabrics.h +1
-0@@ -189,6 +189,7 @@ void nvmf_free_options(struct nvmf_ctrl_options *opts);int nvmf_get_address (struct nvme_ctrl *ctrl,char *buf,int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl); +void nvmf_error_recovery (struct nvme_ctrl *ctrl);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);drivers/nvme/host/rdma.c +8
-17@@ -1185,15 +1185,6 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)nvmf_reconnect_or_remove (&ctrl->ctrl); } -static void nvme_rdma_error_recovery (struct nvme_rdma_ctrl *ctrl) -{ -if (!nvme_change_ctrl_state (&ctrl->ctrl,NVME_CTRL_RESETTING )) -return ; - -dev_warn (ctrl->ctrl.device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &ctrl->ctrl.err_work); -} -static void nvme_rdma_end_request (struct nvme_rdma_request *req) {struct request *rq =blk_mq_rq_from_pdu (req);@@ -1215,7 +1206,7 @@ static void nvme_rdma_wr_error(struct ib_cq *cq, struct ib_wc *wc,"%s for CQE 0x%p failed with status %s (%d)\n" , op, wc->wr_cqe, ib_wc_status_msg(wc->status), wc->status); -nvme_rdma_error_recovery (ctrl); +nvmf_error_recovery (&ctrl->ctrl); }static void nvme_rdma_memreg_done (struct ib_cq *cq,struct ib_wc *wc)@@ -1715,7 +1706,7 @@ static void nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,dev_err (queue->ctrl->ctrl.device,"got bad command_id %#x on QP %#x\n" , cqe->command_id, queue->qp->qp_num); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return ; } req =blk_mq_rq_to_pdu (rq);@@ -1729,7 +1720,7 @@ static void nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,dev_err (queue->ctrl->ctrl.device,"Bogus remote invalidation for rkey %#x\n" , req->mr ? req->mr->rkey :0 ); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl); } }else if (req->mr) {int ret;@@ -1739,7 +1730,7 @@ static void nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,dev_err (queue->ctrl->ctrl.device,"Queueing INV WR for rkey %#x failed (%d)\n" , req->mr->rkey, ret); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl); }/* the local invalidation completion will end the request */ return ;@@ -1766,7 +1757,7 @@ static void nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)if (unlikely (wc->byte_len < len)) {dev_err (queue->ctrl->ctrl.device,"Unexpected nvme completion length(%d)\n" , wc->byte_len); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return ; }@@ -1936,7 +1927,7 @@ static int nvme_rdma_cm_handler(struct rdma_cm_id *cm_id,case RDMA_CM_EVENT_TIMEWAIT_EXIT :dev_dbg (queue->ctrl->ctrl.device,"disconnect received - connection closed\n" ); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);break ;case RDMA_CM_EVENT_DEVICE_REMOVAL :/* device removal is handled via the ib_client API */ @@ -1944,7 +1935,7 @@ static int nvme_rdma_cm_handler(struct rdma_cm_id *cm_id,default :dev_err (queue->ctrl->ctrl.device,"Unexpected RDMA CM event (%d)\n" , ev->event); -nvme_rdma_error_recovery (queue->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);break ; }@@ -2000,7 +1991,7 @@ nvme_rdma_timeout(struct request *rq, bool reserved)* LIVE state should trigger the normal error recovery which will * handle completing this request. */ -nvme_rdma_error_recovery (ctrl); +nvmf_error_recovery (&ctrl->ctrl);return BLK_EH_RESET_TIMER ; }drivers/nvme/host/tcp.c +5
-14@@ -479,15 +479,6 @@ static void nvme_tcp_init_recv_ctx(struct nvme_tcp_queue *queue) queue->ddgst_remaining =0 ; } -static void nvme_tcp_error_recovery (struct nvme_ctrl *ctrl) -{ -if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING )) -return ; - -dev_warn (ctrl->device,"starting error recovery\n" ); -queue_work (nvme_reset_wq, &ctrl->err_work); -} -static int nvme_tcp_process_nvme_cqe (struct nvme_tcp_queue *queue,struct nvme_completion *cqe) {@@ -499,7 +490,7 @@ static int nvme_tcp_process_nvme_cqe(struct nvme_tcp_queue *queue,dev_err (queue->ctrl->ctrl.device,"got bad cqe.command_id %#x on queue %d\n" , cqe->command_id,nvme_tcp_queue_id (queue)); -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return -EINVAL ; }@@ -541,7 +532,7 @@ static int nvme_tcp_handle_c2h_data(struct nvme_tcp_queue *queue,dev_err (queue->ctrl->ctrl.device,"queue %d tag %#x SUCCESS set but not last PDU\n" ,nvme_tcp_queue_id (queue), rq->tag); -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return -EPROTO ; }@@ -850,7 +841,7 @@ static int nvme_tcp_recv_skb(read_descriptor_t *desc, struct sk_buff *skb,dev_err (queue->ctrl->ctrl.device,"receive failed: %d\n" , result); queue->rd_enabled = false; -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);return result; } }@@ -898,7 +889,7 @@ static void nvme_tcp_state_change(struct sock *sk)case TCP_LAST_ACK :case TCP_FIN_WAIT1 :case TCP_FIN_WAIT2 : -nvme_tcp_error_recovery (&queue->ctrl->ctrl); +nvmf_error_recovery (&queue->ctrl->ctrl);break ;default : dev_info(queue->ctrl->ctrl.device,@@ -2252,7 +2243,7 @@ nvme_tcp_timeout(struct request *rq, bool reserved)* LIVE state should trigger the normal error recovery which will * handle completing this request. */ -nvme_tcp_error_recovery (ctrl); +nvmf_error_recovery (ctrl);return BLK_EH_RESET_TIMER ; }Signature · 1 line
2.18.1
--
Himanshu Madhani Oracle Linux Engineering
PATCH 5/7 nvme/nvme-fabrics: introduce nvmf_error_recovery_work API
In reply to Max GurtovoyError recovery work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce 2 new ctrl ops to teardown IO and admin queue.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
--- drivers/nvme/host/fabrics.c | 23 +++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/nvme.h | 4 +++ drivers/nvme/host/rdma.c | 56 ++++++++++++++++--------------------- drivers/nvme/host/tcp.c | 56 +++++++++++++++---------------------- 5 files changed, 75 insertions(+), 65 deletions(-)
drivers/nvme/host/fabrics.c +23 -0
@@ -493,6 +493,29 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +void nvmf_error_recovery_work (struct work_struct *work) +{ +struct nvme_ctrl *ctrl =container_of (work, + struct nvme_ctrl, err_work); + +nvme_stop_keep_alive (ctrl); + ctrl->ops->teardown_ctrl_io_queues (ctrl); +/* unquiesce to fail fast pending requests */ +nvme_start_queues (ctrl); + ctrl->ops->teardown_ctrl_admin_queue (ctrl); +blk_mq_unquiesce_queue (ctrl->admin_q); + +if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_CONNECTING )) { +/* state change failure is ok if we started ctrl delete */ +WARN_ON_ONCE (ctrl->state !=NVME_CTRL_DELETING && + ctrl->state !=NVME_CTRL_DELETING_NOIO ); +return ; + } + +nvmf_reconnect_or_remove (ctrl); +} +EXPORT_SYMBOL_GPL (nvmf_error_recovery_work); +void nvmf_error_recovery (struct nvme_ctrl *ctrl) {if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING ))
drivers/nvme/host/fabrics.h +1 -0
@@ -190,6 +190,7 @@ int nvmf_get_address(struct nvme_ctrl *ctrl, char *buf, int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);void nvmf_error_recovery (struct nvme_ctrl *ctrl); +void nvmf_error_recovery_work (struct work_struct *work);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);
drivers/nvme/host/nvme.h +4 -0
@@ -493,6 +493,10 @@ struct nvme_ctrl_ops {void (*submit_async_event)(struct nvme_ctrl *ctrl);void (*delete_ctrl)(struct nvme_ctrl *ctrl);int (*get_address)(struct nvme_ctrl *ctrl,char *buf,int size); + +/* Fabrics only */ +void (*teardown_ctrl_io_queues)(struct nvme_ctrl *ctrl); +void (*teardown_ctrl_admin_queue)(struct nvme_ctrl *ctrl); };/*
drivers/nvme/host/rdma.c +24 -32
86 lines not on this page · raw message
drivers/nvme/host/tcp.c +23 -33
86 lines not on this page · raw message
Signature · 1 line
2.18.1
Re: [PATCH 5/7] nvme/nvme-fabrics: introduce nvmf_error_recovery_work API
Quoted 10 lines · Max Gurtovoy
Error recovery work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce 2 new ctrl ops to teardown IO and admin queue.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>---
I didn't like the odd convention"_nvme", since we are under time crunch :-
Sagi Grimberg
maintainerRe: [PATCH 5/7] nvme/nvme-fabrics: introduce nvmf_error_recovery_work API
In reply to Max GurtovoyQuoted 9 lines · Max Gurtovoy
Error recovery work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce 2 new ctrl ops to teardown IO and admin queue.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>
Quoted 7 lines
--- drivers/nvme/host/fabrics.c | 23 +++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/nvme.h | 4 +++ drivers/nvme/host/rdma.c | 56 ++++++++++++++++--------------------- drivers/nvme/host/tcp.c | 56 +++++++++++++++---------------------- 5 files changed, 75 insertions(+), 65 deletions(-)
Diffstat dry stats are not in your favor...
Quoted 19 lines
drivers/nvme/host/fabrics.c +10
-0@@ -493,6 +493,29 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +void nvmf_error_recovery_work (struct work_struct *work) +{ +struct nvme_ctrl *ctrl =container_of (work, + struct nvme_ctrl, err_work); + +nvme_stop_keep_alive (ctrl); + ctrl->ops->teardown_ctrl_io_queues (ctrl); +/* unquiesce to fail fast pending requests */ +nvme_start_queues (ctrl); + ctrl->ops->teardown_ctrl_admin_queue (ctrl);
drivers/nvme/host/fabrics.c void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) · line 515
Quoted 10 lines
drivers/nvme/host/fabrics.c +10
-0+blk_mq_unquiesce_queue (ctrl->admin_q); + +if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_CONNECTING )) { +/* state change failure is ok if we started ctrl delete */ +WARN_ON_ONCE (ctrl->state !=NVME_CTRL_DELETING && + ctrl->state !=NVME_CTRL_DELETING_NOIO ); +return ; + } + +nvmf_reconnect_or_remove (ctrl);
We need James to provide feedback how can this be useful for FC.
Quoted 23 lines
drivers/nvme/host/fabrics.c +3
-0+} +EXPORT_SYMBOL_GPL (nvmf_error_recovery_work); +void nvmf_error_recovery (struct nvme_ctrl *ctrl) {if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING ))drivers/nvme/host/fabrics.h +1
-0@@ -190,6 +190,7 @@ int nvmf_get_address(struct nvme_ctrl *ctrl, char *buf, int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);void nvmf_error_recovery (struct nvme_ctrl *ctrl); +void nvmf_error_recovery_work (struct work_struct *work);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);drivers/nvme/host/nvme.h +0
-0
drivers/nvme/host/nvme.h struct nvme_ctrl_ops { · line 499
Quoted 7 lines
drivers/nvme/host/nvme.h +4
-0@@ -493,6 +493,10 @@ struct nvme_ctrl_ops {void (*submit_async_event)(struct nvme_ctrl *ctrl);void (*delete_ctrl)(struct nvme_ctrl *ctrl);int (*get_address)(struct nvme_ctrl *ctrl,char *buf,int size); + +/* Fabrics only */ +void (*teardown_ctrl_io_queues)(struct nvme_ctrl *ctrl); +void (*teardown_ctrl_admin_queue)(struct nvme_ctrl *ctrl);
This becomes strange that we have teardown without a setup callback...
Quoted 90 lines
drivers/nvme/host/nvme.h +0
-0};/* drivers/nvme/host/rdma.c +23
-31@@ -1031,6 +1031,11 @@ static void nvme_rdma_teardown_admin_queue(struct nvme_rdma_ctrl *ctrl,nvme_rdma_destroy_admin_queue (ctrl, remove); } +static void _nvme_rdma_teardown_admin_queue (struct nvme_ctrl *ctrl) +{ +nvme_rdma_teardown_admin_queue (to_rdma_ctrl (ctrl), false); +} +static void nvme_rdma_teardown_io_queues (struct nvme_rdma_ctrl *ctrl,bool remove) {@@ -1046,6 +1051,11 @@ static void nvme_rdma_teardown_io_queues(struct nvme_rdma_ctrl *ctrl, } } +static void _nvme_rdma_teardown_io_queues (struct nvme_ctrl *ctrl) +{ +nvme_rdma_teardown_io_queues (to_rdma_ctrl (ctrl), false); +} +static void nvme_rdma_free_ctrl (struct nvme_ctrl *nctrl) {struct nvme_rdma_ctrl *ctrl =to_rdma_ctrl (nctrl);@@ -1164,27 +1174,6 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)nvmf_reconnect_or_remove (&ctrl->ctrl); } -static void nvme_rdma_error_recovery_work (struct work_struct *work) -{ -struct nvme_rdma_ctrl *ctrl =container_of (work, - struct nvme_rdma_ctrl, ctrl.err_work); - -nvme_stop_keep_alive (&ctrl->ctrl); -nvme_rdma_teardown_io_queues (ctrl, false); -nvme_start_queues (&ctrl->ctrl); -nvme_rdma_teardown_admin_queue (ctrl, false); -blk_mq_unquiesce_queue (ctrl->ctrl.admin_q); - -if (!nvme_change_ctrl_state (&ctrl->ctrl,NVME_CTRL_CONNECTING )) { -/* state change failure is ok if we started ctrl delete */ -WARN_ON_ONCE (ctrl->ctrl.state !=NVME_CTRL_DELETING && - ctrl->ctrl.state !=NVME_CTRL_DELETING_NOIO ); -return ; - } - -nvmf_reconnect_or_remove (&ctrl->ctrl); -} -static void nvme_rdma_end_request (struct nvme_rdma_request *req) {struct request *rq =blk_mq_rq_from_pdu (req);@@ -2240,16 +2229,19 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) }static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = { - .name ="rdma" , - .module =THIS_MODULE , - .flags =NVME_F_FABRICS |NVME_F_METADATA_SUPPORTED , - .reg_read32 = nvmf_reg_read32, - .reg_read64 = nvmf_reg_read64, - .reg_write32 = nvmf_reg_write32, - .free_ctrl = nvme_rdma_free_ctrl, - .submit_async_event = nvme_rdma_submit_async_event, - .delete_ctrl = nvme_rdma_delete_ctrl, - .get_address = nvmf_get_address, + .name ="rdma" , + .module =THIS_MODULE , + .flags =NVME_F_FABRICS | +NVME_F_METADATA_SUPPORTED , + .reg_read32 = nvmf_reg_read32, + .reg_read64 = nvmf_reg_read64, + .reg_write32 = nvmf_reg_write32, + .free_ctrl = nvme_rdma_free_ctrl, + .submit_async_event = nvme_rdma_submit_async_event, + .delete_ctrl = nvme_rdma_delete_ctrl, + .get_address = nvmf_get_address, + .teardown_ctrl_io_queues = _nvme_rdma_teardown_io_queues, + .teardown_ctrl_admin_queue = _nvme_rdma_teardown_admin_queue, };/*
drivers/nvme/host/rdma.c static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, · line 2324
Quoted 5 lines
drivers/nvme/host/rdma.c +1
-1@@ -2329,7 +2321,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev,INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvme_rdma_reconnect_ctrl_work); -INIT_WORK (&ctrl->ctrl.err_work, nvme_rdma_error_recovery_work); +INIT_WORK (&ctrl->ctrl.err_work, nvmf_error_recovery_work);
This initialization needs to move to the core or fabrics lib.
Re: [PATCH 5/7] nvme/nvme-fabrics: introduce nvmf_error_recovery_work API
Quoted 78 lines · Sagi Grimberg
Quoted 9 lines · Max Gurtovoy
Error recovery work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce 2 new ctrl ops to teardown IO and admin queue.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>Quoted 7 lines
---
drivers/nvme/host/fabrics.c| 23 +++++++++++++++drivers/nvme/host/fabrics.h| 1 + drivers/nvme/host/nvme.h | 4 +++ drivers/nvme/host/rdma.c | 56 ++++++++++++++++--------------------- drivers/nvme/host/tcp.c | 56 +++++++++++++++----------------------5 files changed, 75 insertions(+), 65 deletions(-)Diffstat dry stats are not in your favor...
Quoted 19 lines
drivers/nvme/host/fabrics.c +9
-0@@ -493,6 +493,29 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) } EXPORT_SYMBOL_GPL(nvmf_reconnect_or_remove); +void nvmf_error_recovery_work(struct work_struct *work) +{ + struct nvme_ctrl *ctrl = container_of(work, + struct nvme_ctrl, err_work); + + nvme_stop_keep_alive(ctrl); + ctrl->ops->teardown_ctrl_io_queues(ctrl); + /* unquiesce to fail fast pending requests */ + nvme_start_queues(ctrl); + ctrl->ops->teardown_ctrl_admin_queue(ctrl);drivers/nvme/host/fabrics.c void nvmf_reconnect_or_remove(struct nvme_ctrl · line 515
Quoted 10 lines
drivers/nvme/host/fabrics.c +10
-0+ blk_mq_unquiesce_queue(ctrl->admin_q); + + if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_CONNECTING)) { + /* state change failure is ok if we started ctrl delete */ + WARN_ON_ONCE(ctrl->state != NVME_CTRL_DELETING && + ctrl->state != NVME_CTRL_DELETING_NOIO); + return; + } + + nvmf_reconnect_or_remove(ctrl);We need James to provide feedback how can this be useful for FC.
Quoted 24 lines
drivers/nvme/host/fabrics.c +3
-0+} +EXPORT_SYMBOL_GPL (nvmf_error_recovery_work); + void nvmf_error_recovery(struct nvme_ctrl *ctrl) {if (!nvme_change_ctrl_state(ctrl,
NVME_CTRL_RESETTING))drivers/nvme/host/fabrics.h +1
-0@@ -190,6 +190,7 @@ int nvmf_get_address(struct nvme_ctrl *ctrl, char *buf, int size); bool nvmf_should_reconnect(struct nvme_ctrl *ctrl); void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl); void nvmf_error_recovery(struct nvme_ctrl *ctrl); +void nvmf_error_recovery_work (struct work_struct *work); bool nvmf_ip_options_match(struct nvme_ctrl *ctrl, struct nvmf_ctrl_options *opts);diff --git
a/drivers/nvme/host/nvme.hb/drivers/nvme/host/nvme.hindexf9e1ce93d61d..1573edf6e97f100644drivers/nvme/host/nvme.h +0
-0@@ -493,6 +493,10 @@ struct nvme_ctrl_ops { void (*submit_async_event)(struct nvme_ctrl *ctrl);
Quoted 8 lines
drivers/nvme/host/nvme.h struct nvme_ctrl_ops { · line 499
Quoted 6 lines
drivers/nvme/host/nvme.h +4
-0void (*delete_ctrl)(struct nvme_ctrl *ctrl); int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size); + + /* Fabrics only */ + void (*teardown_ctrl_io_queues)(struct nvme_ctrl *ctrl); + void (*teardown_ctrl_admin_queue)(struct nvme_ctrl *ctrl);This becomes strange that we have teardown without a setup callback...
We can do it incrementally.
It's not the first time we do it :)
Quoted 88 lines
Quoted 88 lines
drivers/nvme/host/nvme.h +0
-0}; /*drivers/nvme/host/rdma.c +4
-0@@ -1031,6 +1031,11 @@ static void nvme_rdma_teardown_admin_queue(struct nvme_rdma_ctrl *ctrl, nvme_rdma_destroy_admin_queue(ctrl, remove); } +static void _nvme_rdma_teardown_admin_queue(struct nvme_ctrl *ctrl) +{ + nvme_rdma_teardown_admin_queue(to_rdma_ctrl(ctrl), false); +} + static void nvme_rdma_teardown_io_queues(struct nvme_rdma_ctrl *ctrl, bool remove){
drivers/nvme/host/rdma.c +4
-0@@ -1046,6 +1051,11 @@ static void nvme_rdma_teardown_io_queues(struct nvme_rdma_ctrl *ctrl, } } +static void _nvme_rdma_teardown_io_queues(struct nvme_ctrl *ctrl) +{ + nvme_rdma_teardown_io_queues(to_rdma_ctrl(ctrl), false); +} + static void nvme_rdma_free_ctrl(struct nvme_ctrl *nctrl) {struct
nvme_rdma_ctrl*ctrl =to_rdma_ctrl(nctrl);drivers/nvme/host/rdma.c +0
-20@@ -1164,27 +1174,6 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) nvmf_reconnect_or_remove(&ctrl->ctrl); } -static void nvme_rdma_error_recovery_work(struct work_struct *work) -{ - struct nvme_rdma_ctrl *ctrl = container_of(work, - struct nvme_rdma_ctrl, ctrl.err_work); - - nvme_stop_keep_alive(&ctrl->ctrl); - nvme_rdma_teardown_io_queues(ctrl, false); - nvme_start_queues(&ctrl->ctrl); - nvme_rdma_teardown_admin_queue(ctrl, false); - blk_mq_unquiesce_queue(ctrl->ctrl.admin_q); - - if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_CONNECTING)) { - /* state change failure is ok if we started ctrl delete */ - WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING && - ctrl->ctrl.state != NVME_CTRL_DELETING_NOIO); - return; - } - - nvmf_reconnect_or_remove(&ctrl->ctrl); -} - static void nvme_rdma_end_request(struct nvme_rdma_request *req) {struct request *rq =
blk_mq_rq_from_pdu(req);drivers/nvme/host/rdma.c +13
-10@@ -2240,16 +2229,19 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) } static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = { - .name = "rdma", - .module = THIS_MODULE, - .flags = NVME_F_FABRICS | NVME_F_METADATA_SUPPORTED, - .reg_read32 = nvmf_reg_read32, - .reg_read64 = nvmf_reg_read64, - .reg_write32 = nvmf_reg_write32, - .free_ctrl = nvme_rdma_free_ctrl, - .submit_async_event = nvme_rdma_submit_async_event, - .delete_ctrl = nvme_rdma_delete_ctrl, - .get_address = nvmf_get_address, + .name = "rdma", + .module = THIS_MODULE, + .flags = NVME_F_FABRICS | + NVME_F_METADATA_SUPPORTED, + .reg_read32 = nvmf_reg_read32, + .reg_read64 = nvmf_reg_read64, + .reg_write32 = nvmf_reg_write32, + .free_ctrl = nvme_rdma_free_ctrl, + .submit_async_event = nvme_rdma_submit_async_event, + .delete_ctrl = nvme_rdma_delete_ctrl, + .get_address = nvmf_get_address, + .teardown_ctrl_io_queues = _nvme_rdma_teardown_io_queues, + .teardown_ctrl_admin_queue = _nvme_rdma_teardown_admin_queue, }; /*
Quoted 8 lines
drivers/nvme/host/rdma.c static struct nvme_ctrl · line 2324
Quoted 6 lines
drivers/nvme/host/rdma.c +1
-1@@ -2329,7 +2321,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, INIT_DELAYED_WORK(&ctrl->ctrl.connect_work, nvme_rdma_reconnect_ctrl_work); - INIT_WORK(&ctrl->ctrl.err_work, nvme_rdma_error_recovery_work); + INIT_WORK(&ctrl->ctrl.err_work, nvmf_error_recovery_work);This initialization needs to move to the core or fabrics lib.
It's done in the next patches.
Re: [PATCH 5/7] nvme/nvme-fabrics: introduce nvmf_error_recovery_work API
In reply to Max GurtovoyQuoted 68 lines · Max Gurtovoy
Error recovery work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce 2 new ctrl ops to teardown IO and admin queue.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>--- drivers/nvme/host/fabrics.c | 23 +++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/nvme.h | 4 +++ drivers/nvme/host/rdma.c | 56 ++++++++++++++++--------------------- drivers/nvme/host/tcp.c | 56 +++++++++++++++---------------------- 5 files changed, 75 insertions(+), 65 deletions(-)drivers/nvme/host/fabrics.c +23
-0@@ -493,6 +493,29 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); +void nvmf_error_recovery_work (struct work_struct *work) +{ +struct nvme_ctrl *ctrl =container_of (work, + struct nvme_ctrl, err_work); + +nvme_stop_keep_alive (ctrl); + ctrl->ops->teardown_ctrl_io_queues (ctrl); +/* unquiesce to fail fast pending requests */ +nvme_start_queues (ctrl); + ctrl->ops->teardown_ctrl_admin_queue (ctrl); +blk_mq_unquiesce_queue (ctrl->admin_q); + +if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_CONNECTING )) { +/* state change failure is ok if we started ctrl delete */ +WARN_ON_ONCE (ctrl->state !=NVME_CTRL_DELETING && + ctrl->state !=NVME_CTRL_DELETING_NOIO ); +return ; + } + +nvmf_reconnect_or_remove (ctrl); +} +EXPORT_SYMBOL_GPL (nvmf_error_recovery_work); +void nvmf_error_recovery (struct nvme_ctrl *ctrl) {if (!nvme_change_ctrl_state (ctrl,NVME_CTRL_RESETTING ))drivers/nvme/host/fabrics.h +1
-0@@ -190,6 +190,7 @@ int nvmf_get_address(struct nvme_ctrl *ctrl, char *buf, int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);void nvmf_error_recovery (struct nvme_ctrl *ctrl); +void nvmf_error_recovery_work (struct work_struct *work);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);drivers/nvme/host/nvme.h +0
-0
drivers/nvme/host/nvme.h struct nvme_ctrl_ops { · line 502
Quoted 10 lines
drivers/nvme/host/nvme.h +4
-0@@ -493,6 +493,10 @@ struct nvme_ctrl_ops {void (*submit_async_event)(struct nvme_ctrl *ctrl);void (*delete_ctrl)(struct nvme_ctrl *ctrl);int (*get_address)(struct nvme_ctrl *ctrl,char *buf,int size); + +/* Fabrics only */ +void (*teardown_ctrl_io_queues)(struct nvme_ctrl *ctrl); +void (*teardown_ctrl_admin_queue)(struct nvme_ctrl *ctrl); };/*
As noted by Sagi, this abstraction really is awkward. Why not have
void (*teardown_ctrl_queues)(struct nvme_ctrl *ctrl, int qid);
and let the callback figure out whether it's for the admin queue (ie qid == 8) or the normal I/O queue?
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
PATCH 6/7 nvme/nvme-fabrics: introduce nvmf_reconnect_ctrl_work API
In reply to Max GurtovoyReconnect work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce a new ctrl op to setup a ctrl.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
--- drivers/nvme/host/fabrics.c | 24 ++++++++++++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 26 ++++---------------------- drivers/nvme/host/tcp.c | 27 ++++----------------------- 5 files changed, 34 insertions(+), 45 deletions(-)
drivers/nvme/host/fabrics.c +24 -0
@@ -526,6 +526,30 @@ void nvmf_error_recovery(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_error_recovery); +void nvmf_reconnect_ctrl_work (struct work_struct *work) +{ +struct nvme_ctrl *ctrl =container_of (to_delayed_work (work), + struct nvme_ctrl, connect_work); + + ++ctrl->nr_reconnects; + +if (ctrl->ops->setup_ctrl (ctrl)) + goto requeue; + +dev_info (ctrl->device,"Successfully reconnected (%d attempt)\n" , + ctrl->nr_reconnects); + + ctrl->nr_reconnects =0 ; + +return ; + +requeue: +dev_info (ctrl->device,"Failed reconnect attempt %d\n" , + ctrl->nr_reconnects); +nvmf_reconnect_or_remove (ctrl); +} +EXPORT_SYMBOL_GPL (nvmf_reconnect_ctrl_work); +/** * nvmf_register_transport() - NVMe Fabrics Library registration function. * @ops: Transport ops instance to be registered to the
drivers/nvme/host/fabrics.h +1 -0
@@ -191,6 +191,7 @@ bool nvmf_should_reconnect(struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);void nvmf_error_recovery (struct nvme_ctrl *ctrl);void nvmf_error_recovery_work (struct work_struct *work); +void nvmf_reconnect_ctrl_work (struct work_struct *work);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);
drivers/nvme/host/nvme.h +1 -0
@@ -497,6 +497,7 @@ struct nvme_ctrl_ops {/* Fabrics only */ void (*teardown_ctrl_io_queues)(struct nvme_ctrl *ctrl);void (*teardown_ctrl_admin_queue)(struct nvme_ctrl *ctrl); +int (*setup_ctrl)(struct nvme_ctrl *ctrl); };/*
drivers/nvme/host/rdma.c +4 -22
@@ -1151,27 +1151,9 @@ static int nvme_rdma_setup_ctrl(struct nvme_rdma_ctrl *ctrl, bool new)return ret; } -static void nvme_rdma_reconnect_ctrl_work (struct work_struct *work) +static int _nvme_rdma_setup_ctrl (struct nvme_ctrl *ctrl) { -struct nvme_rdma_ctrl *ctrl =container_of (to_delayed_work (work), - struct nvme_rdma_ctrl, ctrl.connect_work); - - ++ctrl->ctrl.nr_reconnects; - -if (nvme_rdma_setup_ctrl (ctrl, false)) - goto requeue; - -dev_info (ctrl->ctrl.device,"Successfully reconnected (%d attempts)\n" , - ctrl->ctrl.nr_reconnects); - - ctrl->ctrl.nr_reconnects =0 ; - -return ; - -requeue: -dev_info (ctrl->ctrl.device,"Failed reconnect attempt %d\n" , - ctrl->ctrl.nr_reconnects); -nvmf_reconnect_or_remove (&ctrl->ctrl); +return nvme_rdma_setup_ctrl (to_rdma_ctrl (ctrl), false); }static void nvme_rdma_end_request (struct nvme_rdma_request *req)
@@ -2242,6 +2224,7 @@ static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = { .get_address = nvmf_get_address, .teardown_ctrl_io_queues = _nvme_rdma_teardown_io_queues, .teardown_ctrl_admin_queue = _nvme_rdma_teardown_admin_queue, + .setup_ctrl = _nvme_rdma_setup_ctrl, };/*
@@ -2319,8 +2302,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, goto out_free_ctrl; } -INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, - nvme_rdma_reconnect_ctrl_work); +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvmf_reconnect_ctrl_work);INIT_WORK (&ctrl->ctrl.err_work, nvmf_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work);
drivers/nvme/host/tcp.c +4 -23
@@ -2042,28 +2042,9 @@ static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new)return ret; } -static void nvme_tcp_reconnect_ctrl_work (struct work_struct *work) +static int _nvme_tcp_setup_ctrl (struct nvme_ctrl *ctrl) { -struct nvme_tcp_ctrl *tcp_ctrl =container_of (to_delayed_work (work), - struct nvme_tcp_ctrl, ctrl.connect_work); -struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; - - ++ctrl->nr_reconnects; - -if (nvme_tcp_setup_ctrl (ctrl, false)) - goto requeue; - -dev_info (ctrl->device,"Successfully reconnected (%d attempt)\n" , - ctrl->nr_reconnects); - - ctrl->nr_reconnects =0 ; - -return ; - -requeue: -dev_info (ctrl->device,"Failed reconnect attempt %d\n" , - ctrl->nr_reconnects); -nvmf_reconnect_or_remove (ctrl); +return nvme_tcp_setup_ctrl (ctrl, false); }static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown)
@@ -2435,6 +2416,7 @@ static const struct nvme_ctrl_ops nvme_tcp_ctrl_ops = { .get_address = nvmf_get_address, .teardown_ctrl_io_queues = _nvme_tcp_teardown_io_queues, .teardown_ctrl_admin_queue = _nvme_tcp_teardown_admin_queue, + .setup_ctrl = _nvme_tcp_setup_ctrl, }; static bool
@@ -2471,8 +2453,7 @@ static struct nvme_ctrl *nvme_tcp_create_ctrl(struct device *dev, ctrl->ctrl.sqsize = opts->queue_size -1 ; ctrl->ctrl.kato = opts->kato; -INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, - nvme_tcp_reconnect_ctrl_work); +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvmf_reconnect_ctrl_work);INIT_WORK (&ctrl->ctrl.err_work, nvmf_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_reset_ctrl_work);
Signature · 1 line
2.18.1
Re: [PATCH 6/7] nvme/nvme-fabrics: introduce nvmf_reconnect_ctrl_work API
Quoted 8 lines · Max Gurtovoy
Reconnect work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce a new ctrl op to setup a ctrl.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>
Same as previous patch :-
Looks good.
Sagi Grimberg
maintainerRe: [PATCH 6/7] nvme/nvme-fabrics: introduce nvmf_reconnect_ctrl_work API
In reply to Max Gurtovoydrivers/nvme/host/tcp.c static const struct nvme_ctrl_ops nvme_tcp_ctrl_ops = { · line 2419
Quoted 7 lines
static void
nvme_tcp_teardown_ctrl(struct nvme_ctrl *ctrl, bool shutdown)drivers/nvme/host/tcp.c +1
-0@@ -2435,6 +2416,7 @@ static const struct nvme_ctrl_ops nvme_tcp_ctrl_ops = { .get_address = nvmf_get_address, .teardown_ctrl_io_queues = _nvme_tcp_teardown_io_queues, .teardown_ctrl_admin_queue = _nvme_tcp_teardown_admin_queue, + .setup_ctrl = _nvme_tcp_setup_ctrl,
This asymmetry is not great, what made you choose to go this way?
Re: [PATCH 6/7] nvme/nvme-fabrics: introduce nvmf_reconnect_ctrl_work API
Quoted 10 lines · Sagi Grimberg
drivers/nvme/host/tcp.c static const struct nvme_ctrl_ops · line 2420
Quoted 8 lines
static void
nvme_tcp_teardown_ctrl(struct nvme_ctrl *ctrl, bool shutdown)drivers/nvme/host/tcp.c +1
-0@@ -2435,6 +2416,7 @@ static const struct nvme_ctrl_ops nvme_tcp_ctrl_ops = { .get_address = nvmf_get_address, .teardown_ctrl_io_queues = _nvme_tcp_teardown_io_queues, .teardown_ctrl_admin_queue = _nvme_tcp_teardown_admin_queue, + .setup_ctrl = _nvme_tcp_setup_ctrl,This asymmetry is not great, what made you choose to go this way?
There are more duplicated flows we can work on in the future.
Re: [PATCH 6/7] nvme/nvme-fabrics: introduce nvmf_reconnect_ctrl_work API
In reply to Max GurtovoyQuoted 99 lines · Max Gurtovoy
Reconnect work is duplicated in RDMA and TCP transports. Move this logic to common code. For that, introduce a new ctrl op to setup a ctrl.
Also update the RDMA/TCP transport drivers to use this API and remove the duplicated code.
Reviewed-by Israel Rukshin <[email protected]>Signed-off-by Max Gurtovoy <[email protected]>--- drivers/nvme/host/fabrics.c | 24 ++++++++++++++++++++++++ drivers/nvme/host/fabrics.h | 1 + drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/rdma.c | 26 ++++---------------------- drivers/nvme/host/tcp.c | 27 ++++----------------------- 5 files changed, 34 insertions(+), 45 deletions(-)drivers/nvme/host/fabrics.c +24
-0@@ -526,6 +526,30 @@ void nvmf_error_recovery(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_error_recovery); +void nvmf_reconnect_ctrl_work (struct work_struct *work) +{ +struct nvme_ctrl *ctrl =container_of (to_delayed_work (work), + struct nvme_ctrl, connect_work); + + ++ctrl->nr_reconnects; + +if (ctrl->ops->setup_ctrl (ctrl)) + goto requeue; + +dev_info (ctrl->device,"Successfully reconnected (%d attempt)\n" , + ctrl->nr_reconnects); + + ctrl->nr_reconnects =0 ; + +return ; + +requeue: +dev_info (ctrl->device,"Failed reconnect attempt %d\n" , + ctrl->nr_reconnects); +nvmf_reconnect_or_remove (ctrl); +} +EXPORT_SYMBOL_GPL (nvmf_reconnect_ctrl_work); +/** * nvmf_register_transport() - NVMe Fabrics Library registration function. * @ops: Transport ops instance to be registered to the drivers/nvme/host/fabrics.h +1
-0@@ -191,6 +191,7 @@ bool nvmf_should_reconnect(struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);void nvmf_error_recovery (struct nvme_ctrl *ctrl);void nvmf_error_recovery_work (struct work_struct *work); +void nvmf_reconnect_ctrl_work (struct work_struct *work);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);drivers/nvme/host/nvme.h +1
-0@@ -497,6 +497,7 @@ struct nvme_ctrl_ops {/* Fabrics only */ void (*teardown_ctrl_io_queues)(struct nvme_ctrl *ctrl);void (*teardown_ctrl_admin_queue)(struct nvme_ctrl *ctrl); +int (*setup_ctrl)(struct nvme_ctrl *ctrl); };/* drivers/nvme/host/rdma.c +1
-14@@ -1151,27 +1151,9 @@ static int nvme_rdma_setup_ctrl(struct nvme_rdma_ctrl *ctrl, bool new)return ret; } -static void nvme_rdma_reconnect_ctrl_work (struct work_struct *work) +static int _nvme_rdma_setup_ctrl (struct nvme_ctrl *ctrl) { -struct nvme_rdma_ctrl *ctrl =container_of (to_delayed_work (work), - struct nvme_rdma_ctrl, ctrl.connect_work); - - ++ctrl->ctrl.nr_reconnects; - -if (nvme_rdma_setup_ctrl (ctrl, false)) - goto requeue; - -dev_info (ctrl->ctrl.device,"Successfully reconnected (%d attempts)\n" , - ctrl->ctrl.nr_reconnects); - - ctrl->ctrl.nr_reconnects =0 ; -
drivers/nvme/host/rdma.c static int nvme_rdma_setup_ctrl(struct nvme_rdma_ctrl *ctrl, bool new) · line 1159
Quoted 10 lines
drivers/nvme/host/rdma.c +1
-6-return ; - -requeue: -dev_info (ctrl->ctrl.device,"Failed reconnect attempt %d\n" , - ctrl->ctrl.nr_reconnects); -nvmf_reconnect_or_remove (&ctrl->ctrl); +return nvme_rdma_setup_ctrl (to_rdma_ctrl (ctrl), false); }static void nvme_rdma_end_request (struct nvme_rdma_request *req)
Really? Can't we separate nvme_rdma_setup_ctrl() such that we have two distinct functions (one for new == true, and one for new == false)? Or, alternatively, setting up the tagset in nvme_rdma_init_ctrl() and kill the 'new' argument altogether?
Quoted 42 lines
drivers/nvme/host/rdma.c +2
-2@@ -2242,6 +2224,7 @@ static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = { .get_address = nvmf_get_address, .teardown_ctrl_io_queues = _nvme_rdma_teardown_io_queues, .teardown_ctrl_admin_queue = _nvme_rdma_teardown_admin_queue, + .setup_ctrl = _nvme_rdma_setup_ctrl, };/* @@ -2319,8 +2302,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, goto out_free_ctrl; } -INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, - nvme_rdma_reconnect_ctrl_work); +INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvmf_reconnect_ctrl_work);INIT_WORK (&ctrl->ctrl.err_work, nvmf_error_recovery_work);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work);drivers/nvme/host/tcp.c +1
-14@@ -2042,28 +2042,9 @@ static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new)return ret; } -static void nvme_tcp_reconnect_ctrl_work (struct work_struct *work) +static int _nvme_tcp_setup_ctrl (struct nvme_ctrl *ctrl) { -struct nvme_tcp_ctrl *tcp_ctrl =container_of (to_delayed_work (work), - struct nvme_tcp_ctrl, ctrl.connect_work); -struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; - - ++ctrl->nr_reconnects; - -if (nvme_tcp_setup_ctrl (ctrl, false)) - goto requeue; - -dev_info (ctrl->device,"Successfully reconnected (%d attempt)\n" , - ctrl->nr_reconnects); - - ctrl->nr_reconnects =0 ;
drivers/nvme/host/tcp.c static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new) · line 2049
Quoted 10 lines
drivers/nvme/host/tcp.c +1
-7- -return ; - -requeue: -dev_info (ctrl->device,"Failed reconnect attempt %d\n" , - ctrl->nr_reconnects); -nvmf_reconnect_or_remove (ctrl); +return nvme_tcp_setup_ctrl (ctrl, false); } Same argument here; I'd rather modify nvme_tcp_setup_ctrl() to drop the
'new' argument and allowing it to be used as the callback directly.
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
PATCH 7/7 nvme-fabrics: add nvmf_init_ctrl/nvmf_teardown_ctrl API
In reply to Max GurtovoyCentralize the initalization and teardown of fabrics specific settings. For now, only used by RDMA and TCP fabric transports.
Also, convert the reconnect_work and error_recovery_work to be static functions since they are not used outside the fabrics driver anymore.
--- drivers/nvme/host/fabrics.c | 20 ++++++++++++++++---- drivers/nvme/host/fabrics.h | 4 ++-- drivers/nvme/host/rdma.c | 7 ++----- drivers/nvme/host/tcp.c | 7 ++----- 4 files changed, 22 insertions(+), 16 deletions(-)
drivers/nvme/host/fabrics.c +16 -4
@@ -493,7 +493,7 @@ void nvmf_reconnect_or_remove(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_reconnect_or_remove); -void nvmf_error_recovery_work (struct work_struct *work) +static void nvmf_error_recovery_work (struct work_struct *work) {struct nvme_ctrl *ctrl =container_of (work, struct nvme_ctrl, err_work);
@@ -514,7 +514,6 @@ void nvmf_error_recovery_work(struct work_struct *work)nvmf_reconnect_or_remove (ctrl); } -EXPORT_SYMBOL_GPL (nvmf_error_recovery_work);void nvmf_error_recovery (struct nvme_ctrl *ctrl) {
@@ -526,7 +525,7 @@ void nvmf_error_recovery(struct nvme_ctrl *ctrl) }EXPORT_SYMBOL_GPL (nvmf_error_recovery); -void nvmf_reconnect_ctrl_work (struct work_struct *work) +static void nvmf_reconnect_ctrl_work (struct work_struct *work) {struct nvme_ctrl *ctrl =container_of (to_delayed_work (work), struct nvme_ctrl, connect_work);
@@ -548,7 +547,20 @@ void nvmf_reconnect_ctrl_work(struct work_struct *work) ctrl->nr_reconnects);nvmf_reconnect_or_remove (ctrl); } -EXPORT_SYMBOL_GPL (nvmf_reconnect_ctrl_work); + +void nvmf_init_ctrl (struct nvme_ctrl *ctrl) +{ +INIT_DELAYED_WORK (&ctrl->connect_work, nvmf_reconnect_ctrl_work); +INIT_WORK (&ctrl->err_work, nvmf_error_recovery_work); +} +EXPORT_SYMBOL_GPL (nvmf_init_ctrl); + +void nvmf_teardown_ctrl (struct nvme_ctrl *ctrl) +{ +cancel_work_sync (&ctrl->err_work); +cancel_delayed_work_sync (&ctrl->connect_work); +} +EXPORT_SYMBOL_GPL (nvmf_teardown_ctrl);/** * nvmf_register_transport() - NVMe Fabrics Library registration function.
drivers/nvme/host/fabrics.h +2 -2
@@ -190,8 +190,8 @@ int nvmf_get_address(struct nvme_ctrl *ctrl, char *buf, int size);bool nvmf_should_reconnect (struct nvme_ctrl *ctrl);void nvmf_reconnect_or_remove (struct nvme_ctrl *ctrl);void nvmf_error_recovery (struct nvme_ctrl *ctrl); -void nvmf_error_recovery_work (struct work_struct *work); -void nvmf_reconnect_ctrl_work (struct work_struct *work); +void nvmf_init_ctrl (struct nvme_ctrl *ctrl); +void nvmf_teardown_ctrl (struct nvme_ctrl *ctrl);bool nvmf_ip_options_match (struct nvme_ctrl *ctrl,struct nvmf_ctrl_options *opts);
drivers/nvme/host/rdma.c +2 -5
@@ -2169,9 +2169,7 @@ static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {static void nvme_rdma_shutdown_ctrl (struct nvme_rdma_ctrl *ctrl,bool shutdown) { -cancel_work_sync (&ctrl->ctrl.err_work); -cancel_delayed_work_sync (&ctrl->ctrl.connect_work); - +nvmf_teardown_ctrl (&ctrl->ctrl);nvme_rdma_teardown_io_queues (ctrl, shutdown);blk_mq_quiesce_queue (ctrl->ctrl.admin_q);if (shutdown)
@@ -2302,8 +2300,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, goto out_free_ctrl; } -INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvmf_reconnect_ctrl_work); -INIT_WORK (&ctrl->ctrl.err_work, nvmf_error_recovery_work); +nvmf_init_ctrl (&ctrl->ctrl);INIT_WORK (&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work); ctrl->ctrl.queue_count = opts->nr_io_queues + opts->nr_write_queues +
drivers/nvme/host/tcp.c +2 -5
@@ -2049,9 +2049,7 @@ static int _nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl)static void nvme_tcp_teardown_ctrl (struct nvme_ctrl *ctrl,bool shutdown) { -cancel_work_sync (&ctrl->err_work); -cancel_delayed_work_sync (&ctrl->connect_work); - +nvmf_teardown_ctrl (ctrl);nvme_tcp_teardown_io_queues (ctrl, shutdown);blk_mq_quiesce_queue (ctrl->admin_q);if (shutdown)
@@ -2453,8 +2451,7 @@ static struct nvme_ctrl *nvme_tcp_create_ctrl(struct device *dev, ctrl->ctrl.sqsize = opts->queue_size -1 ; ctrl->ctrl.kato = opts->kato; -INIT_DELAYED_WORK (&ctrl->ctrl.connect_work, nvmf_reconnect_ctrl_work); -INIT_WORK (&ctrl->ctrl.err_work, nvmf_error_recovery_work); +nvmf_init_ctrl (&ctrl->ctrl);INIT_WORK (&ctrl->ctrl.reset_work, nvme_reset_ctrl_work);if (!(opts->mask &NVMF_OPT_TRSVCID )) {
Signature · 1 line
2.18.1
Sagi Grimberg
maintainerRe: [PATCH 7/7] nvme-fabrics: add nvmf_init_ctrl/nvmf_teardown_ctrl API
drivers/nvme/host/fabrics.c void nvmf_reconnect_ctrl_work(struct work_struct *work) · line 561
Quoted 5 lines
drivers/nvme/host/fabrics.c +4
-0+void nvmf_teardown_ctrl (struct nvme_ctrl *ctrl) +{ +cancel_work_sync (&ctrl->err_work); +cancel_delayed_work_sync (&ctrl->connect_work);
Name doesn't represent what is done in the function at all.. possible names:nvmf_uninit_ctrlnvmf_deinit_ctrlnvmf_fini_ctrl
...
Re: [PATCH 7/7] nvme-fabrics: add nvmf_init_ctrl/nvmf_teardown_ctrl API
Quoted 9 lines · Sagi Grimberg
drivers/nvme/host/fabrics.c void nvmf_reconnect_ctrl_work(struct work_struct *work) · line 561
Quoted 5 lines
drivers/nvme/host/fabrics.c +4
-0+void nvmf_teardown_ctrl (struct nvme_ctrl *ctrl) +{ + cancel_work_sync(&ctrl->err_work); + cancel_delayed_work_sync(&ctrl->connect_work);Name doesn't represent what is done in the function at all.. possible names:
nvmf_uninit_ctrl
Ok, i'll change it to nvmf_uninit_ctrl
Quoted 3 lines
nvmf_deinit_ctrlnvmf_fini_ctrl
...
Quoted 38 lines · Max Gurtovoy
Hi Christoph, Sagi, Keith and Co,
This series is intended to move common reconnection, device removal and error recovery logic to common fabrics and core drivers. In this stage, it mainly deals with the RDMA and TCP drivers which had the same logic for the above tasks. FC driver can start using this framework in the future after some preparation work that should be done.This will ease on code maintenance and reduce the code duplication between the RDMA and TCP drivers.
Since there are some attributes that were moved to generic
nvme_ctrlstructure, I've measured perf impact.NVMe PCI Perf numbers using NVIDIA's NVMe SNAP Bluefield-2 device (Fio 16 jobs, 128 iodepth, libaio):
IO size rand READ (before/after) rand WRITE (before/after)
-------- ------------------------- --------------------------512B 2189k/2197k 2031k/2092k 1KB 2188k/2201k 2027k/2088k 2KB 2166k/2181k 2023k/2078k 4KB 2153k/2169k 2025k/2076k 8KB 1381k/1381k 1378k/1375k 16KB 669k/676k 660k/664k 32KB 320k/322k 305k/305k 64KB 157k/157k 151k/151kAs we can see, there is no impact on the performance.
Max Gurtovoy (7):
nvme: addconnect_workattribute to nvme ctrl
nvme-fabrics: introducenvmf_reconnect_or_removeAPI
nvme: adderr_workattribute to nvme ctrl
nvme-fabrics: introducenvmf_error_recoveryAPI nvme/nvme-fabrics: introducenvmf_error_recovery_workAPI nvme/nvme-fabrics: introducenvmf_reconnect_ctrl_workAPI
nvme-fabrics: add nvmf_init_ctrl/nvmf_teardown_ctrl API
Quoted 8 lines
drivers/nvme/host/fabrics.c | 90 ++++++++++++++++++++++++ drivers/nvme/host/fabrics.h | 4 ++ drivers/nvme/host/fc.c | 23 ++++--- drivers/nvme/host/nvme.h | 7 ++ drivers/nvme/host/rdma.c | 134 ++++++++++-------------------------- drivers/nvme/host/tcp.c | 124 ++++++++------------------------- 6 files changed, 180 insertions(+), 202 deletions(-)
Thanks Max.
I've been thinking we needed to do this for a while as well. FC should be part of this, and up front, so the abstraction is tested properly. I'd rather merge now than spin things again. I'll go through this over the next day or so.
-- james
Christoph Hellwig
maintainerQuoted 4 lines · James Smart
I've been thinking we needed to do this for a while as well. FC should be part of this, and up front, so the abstraction is tested properly. I'd rather merge now than spin things again. I'll go through this over the next day or so.
I'd really love tp pick this up quickly so that it doesn't go stale again if we could get a few reviews. The merge window is about to close.
Quoted 6 lines
NVMe PCI Perf numbers using NVIDIA's NVMe SNAP Bluefield-2 device (Fio 16 jobs, 128 iodepth, libaio):
IO size rand READ (before/after) rand WRITE (before/after)
-------- ------------------------- --------------------------
Quoted 9 lines
512B 2189k/2197k 2031k/2092k 1KB 2188k/2201k 2027k/2088k 2KB 2166k/2181k 2023k/2078k 4KB 2153k/2169k 2025k/2076k 8KB 1381k/1381k 1378k/1375k 16KB 669k/676k 660k/664k 32KB 320k/322k 305k/305k 64KB 157k/157k 151k/151k
Can you please share latency numbers for the purpose of documentation on this thread ?
Sagi Grimberg
maintainerQuoted 6 lines · Max Gurtovoy
Hi Christoph, Sagi, Keith and Co,
This series is intended to move common reconnection, device removal and error recovery logic to common fabrics and core drivers. In this stage, it mainly deals with the RDMA and TCP drivers which had the same logic for the above tasks. FC driver can start using this framework in the future after some preparation work that should be done.
This is a good start, but I think we should also see how we can fit pci here as well.