Skip to contentSkip to conversation

nvmet: add namespace-level debugfs for reservation state

PATCH Superseded Guixin Liu 3 participants 10 messages Jun 22 – Jul 7

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Guixin Liu

author
#0 lore raw

cover nvmet: add namespace-level debugfs for reservation state

This series adds debugfs support for inspecting NVMe target reservation (PR) state at the namespace level.

Patch 1 introduces per-namespace debugfs directories under each subsystem, providing the infrastructure for namespace-specific debug entries.

Patch 2 adds a 'reservation' file that exposes the persistent reservation state including enable status, generation counter, holder information, and the full registrant list.

Example output with two registered hosts and an active holder:

  $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
  enable : 1
  generation : 2
  notify_mask : 0x0
  rtype : write_exclusive
  holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
  registrants:
    hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
    hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222

Guixin Liu (2):
nvmet: add namespace-level debugfs directory
nvmet: expose reservation state through debugfs

 drivers/nvme/target/core.c    |  2 +
 drivers/nvme/target/debugfs.c | 75 +++++++++++++++++++++++++++++++++++
 drivers/nvme/target/debugfs.h |  5 +++
 drivers/nvme/target/nvmet.h   |  3 ++
 4 files changed, 85 insertions(+)
Signature · 1 line
2.43.7
author
#1 lore raw

PATCH 1/2 nvmet: add namespace-level debugfs directory

Add per-namespace debugfs directory support under the subsystem debugfs directory. Each enabled namespace gets a ns<nsid>/ directory created during nvmet_ns_enable() and removed during nvmet_ns_disable().

This provides the infrastructure for exposing namespace-specific debug information in subsequent patches.

---
 drivers/nvme/target/core.c    |  2 ++
 drivers/nvme/target/debugfs.c | 21 +++++++++++++++++++++
 drivers/nvme/target/debugfs.h |  5 +++++
 drivers/nvme/target/nvmet.h   |  3 +++
 4 files changed, 31 insertions(+)
drivers/nvme/target/core.c +2 -0
index 62dd59b9aa4f..9752913e05c6 100644
@@ -616,6 +616,7 @@ int nvmet_ns_enable(struct nvmet_ns *ns)
 	nvmet_ns_changed(subsys, ns->nsid);
 	ns->enabled = true;
 	xa_set_mark(&subsys->namespaces, ns->nsid, NVMET_NS_ENABLED);
+	nvmet_debugfs_ns_setup(ns);
 	ret = 0;
 out_unlock:
 	mutex_unlock(&subsys->lock);
@@ -642,6 +643,7 @@ void nvmet_ns_disable(struct nvmet_ns *ns)
 
 	ns->enabled = false;
 	xa_clear_mark(&subsys->namespaces, ns->nsid, NVMET_NS_ENABLED);
+	nvmet_debugfs_ns_free(ns);
 
 	list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry)
 		pci_dev_put(radix_tree_delete(&ctrl->p2p_ns_map, ns->nsid));
drivers/nvme/target/debugfs.c +21 -0
index 5dcbd5aa86e1..e6f51eb59010 100644
@@ -153,6 +153,27 @@ static int nvmet_ctrl_tls_concat_show(struct seq_file *m, void *p)
 NVMET_DEBUGFS_ATTR(nvmet_ctrl_tls_concat);
 #endif
 
+void nvmet_debugfs_ns_setup(struct nvmet_ns *ns)
+{
+	char name[16];
+	struct dentry *parent = ns->subsys->debugfs_dir;
+
+	if (!parent)
+		return;
+	snprintf(name, sizeof(name), "ns%u", ns->nsid);
+	ns->debugfs_dir = debugfs_create_dir(name, parent);
+	if (IS_ERR(ns->debugfs_dir)) {
+		ns->debugfs_dir = NULL;
+		return;
+	}
+}
+
+void nvmet_debugfs_ns_free(struct nvmet_ns *ns)
+{
+	debugfs_remove_recursive(ns->debugfs_dir);
+	ns->debugfs_dir = NULL;
+}
+
 int nvmet_debugfs_ctrl_setup(struct nvmet_ctrl *ctrl)
 {
 	char name[32];
drivers/nvme/target/debugfs.h +5 -0
index cfb8bbf6a297..b559d254fc2a 100644
@@ -14,6 +14,8 @@ int nvmet_debugfs_subsys_setup(struct nvmet_subsys *subsys);
 void nvmet_debugfs_subsys_free(struct nvmet_subsys *subsys);
 int nvmet_debugfs_ctrl_setup(struct nvmet_ctrl *ctrl);
 void nvmet_debugfs_ctrl_free(struct nvmet_ctrl *ctrl);
+void nvmet_debugfs_ns_setup(struct nvmet_ns *ns);
+void nvmet_debugfs_ns_free(struct nvmet_ns *ns);
 
 int __init nvmet_init_debugfs(void);
 void nvmet_exit_debugfs(void);
@@ -30,6 +32,9 @@ static inline int nvmet_debugfs_ctrl_setup(struct nvmet_ctrl *ctrl)
 }
 static inline void nvmet_debugfs_ctrl_free(struct nvmet_ctrl *ctrl) {}
 
+static inline void nvmet_debugfs_ns_setup(struct nvmet_ns *ns) {}
+static inline void nvmet_debugfs_ns_free(struct nvmet_ns *ns) {}
+
 static inline int __init nvmet_init_debugfs(void)
 {
     return 0;
drivers/nvme/target/nvmet.h +3 -0
index 3305a88684ec..142a78d9160f 100644
@@ -128,6 +128,9 @@ struct nvmet_ns {
 	u8			csi;
 	struct nvmet_pr		pr;
 	struct xarray		pr_per_ctrl_refs;
+#ifdef CONFIG_NVME_TARGET_DEBUGFS
+	struct dentry		*debugfs_dir;
+#endif
 };
 
 static inline struct nvmet_ns *to_nvmet_ns(struct config_item *item)
Signature · 1 line
2.43.7
author
#2 lore raw

PATCH 2/2 nvmet: expose reservation state through debugfs

In reply to Guixin Liu

Add a 'reservation' debugfs file under each namespace directory that shows the persistent reservation state, including enable status, generation counter, notify mask, current holder info, and the full registrant list with hostid and reservation key.

The output uses rcu_read_lock() for safe access to the holder and registrant_list, consistent with other PR read paths.

---
 drivers/nvme/target/debugfs.c | 54 +++++++++++++++++++++++++++++++++++
 1 file changed, 54 insertions(+)
drivers/nvme/target/debugfs.c +54 -0
index e6f51eb59010..96d81d446d13 100644
@@ -153,6 +153,58 @@ static int nvmet_ctrl_tls_concat_show(struct seq_file *m, void *p)
 NVMET_DEBUGFS_ATTR(nvmet_ctrl_tls_concat);
 #endif
 
+static const char *const nvmet_pr_type_names[] = {
+	[NVME_PR_WRITE_EXCLUSIVE]		= "write_exclusive",
+	[NVME_PR_EXCLUSIVE_ACCESS]		= "exclusive_access",
+	[NVME_PR_WRITE_EXCLUSIVE_REG_ONLY]	= "write_exclusive_reg_only",
+	[NVME_PR_EXCLUSIVE_ACCESS_REG_ONLY]	= "exclusive_access_reg_only",
+	[NVME_PR_WRITE_EXCLUSIVE_ALL_REGS]	= "write_exclusive_all_regs",
+	[NVME_PR_EXCLUSIVE_ACCESS_ALL_REGS]	= "exclusive_access_all_regs",
+};
+
+static const char *nvmet_pr_type_to_str(enum nvme_pr_type type)
+{
+	if (type < ARRAY_SIZE(nvmet_pr_type_names) &&
+	    nvmet_pr_type_names[type])
+		return nvmet_pr_type_names[type];
+	return "unknown";
+}
+
+static int nvmet_ns_pr_show(struct seq_file *m, void *p)
+{
+	struct nvmet_ns *ns = m->private;
+	struct nvmet_pr *pr = &ns->pr;
+	struct nvmet_pr_registrant *holder, *reg;
+
+	seq_printf(m, "enable : %d\n", pr->enable);
+	if (!pr->enable)
+		return 0;
+
+	seq_printf(m, "generation : %u\n", atomic_read(&pr->generation));
+	seq_printf(m, "notify_mask : 0x%lx\n", pr->notify_mask);
+
+	rcu_read_lock();
+	holder = rcu_dereference(pr->holder);
+	if (holder) {
+		seq_printf(m, "rtype : %s\n",
+			   nvmet_pr_type_to_str(holder->rtype));
+		seq_printf(m, "holder : hostid=%pUb, rkey=0x%llx\n",
+			   &holder->hostid, holder->rkey);
+	} else {
+		seq_puts(m, "holder : none\n");
+	}
+
+	seq_puts(m, "registrants:\n");
+	list_for_each_entry_rcu(reg, &pr->registrant_list, entry) {
+		seq_printf(m, "  hostid=%pUb, rkey=0x%llx\n",
+			   &reg->hostid, reg->rkey);
+	}
+	rcu_read_unlock();
+
+	return 0;
+}
+NVMET_DEBUGFS_ATTR(nvmet_ns_pr);
+
 void nvmet_debugfs_ns_setup(struct nvmet_ns *ns)
 {
 	char name[16];
@@ -166,6 +218,8 @@ void nvmet_debugfs_ns_setup(struct nvmet_ns *ns)
 		ns->debugfs_dir = NULL;
 		return;
 	}
+	debugfs_create_file("reservation", 0400, ns->debugfs_dir, ns,
+			    &nvmet_ns_pr_fops);
 }
 
 void nvmet_debugfs_ns_free(struct nvmet_ns *ns)
Signature · 1 line
2.43.7
#3 lore raw
In reply to Guixin Liu
Quoted 12 lines · Guixin Liu

This series adds debugfs support for inspecting NVMe target reservation (PR) state at the namespace level.

Patch 1 introduces per-namespace debugfs directories under each subsystem, providing the infrastructure for namespace-specific debug entries.

Patch 2 adds a 'reservation' file that exposes the persistent reservation state including enable status, generation counter, holder information, and the full registrant list.

Example output with two registered hosts and an active holder:

Quoted 9 lines
  $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
  enable : 1
  generation : 2
  notify_mask : 0x0
  rtype : write_exclusive
  holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
  registrants:
    hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
    hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

author
#4 lore raw
Quoted 22 lines · Daniel Wagner
Quoted 12 lines · Guixin Liu

This series adds debugfs support for inspecting NVMe target reservation (PR) state at the namespace level.

Patch 1 introduces per-namespace debugfs directories under each subsystem, providing the infrastructure for namespace-specific debug entries.

Patch 2 adds a 'reservation' file that exposes the persistent reservation state including enable status, generation counter, holder information, and the full registrant list.

Example output with two registered hosts and an active holder:

Quoted 9 lines
   $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
   enable : 1
   generation : 2
   notify_mask : 0x0
   rtype : write_exclusive
   holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
   registrants:
     hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
     hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222
Quoted 3 lines

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

How about this:

1. One holer, multi registrants:
  enable=1
  generation=2
  notify_mask=0x0
  rtype=write_exclusive holder=11111111-1111-1111-1111-111111111111,0x1111 reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

2. No holder, multi registrants:
  enable=1
  generation=1
  notify_mask=0x0
  rtype=none
  holder=none
  reg=11111111-1111-1111-1111-111111111111,0x1111
  reg=22222222-2222-2222-2222-222222222222,0x2222

3. No registrant:
  enable=1
  generation=0
  notify_mask=0x0
  rtype=none
  holder=none

4. Reservation not enabled:
  enable=0

Otherwise, the notify_mask can be stringization:

  notify_mask=reg_preempted,resv_released,resv_preempted

Best Regards,
Guixin Liu

author
#5 lore raw
Quoted 58 lines · Guixin Liu
Quoted 22 lines · Daniel Wagner
Quoted 12 lines · Guixin Liu

This series adds debugfs support for inspecting NVMe target reservation (PR) state at the namespace level.

Patch 1 introduces per-namespace debugfs directories under each subsystem, providing the infrastructure for namespace-specific debug entries.

Patch 2 adds a 'reservation' file that exposes the persistent reservation state including enable status, generation counter, holder information, and the full registrant list.

Example output with two registered hosts and an active holder:

Quoted 9 lines
   $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
   enable : 1
   generation : 2
   notify_mask : 0x0
   rtype : write_exclusive
   holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
   registrants:
     hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
     hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222
Quoted 3 lines

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

How about this:

1. One holer, multi registrants:
  enable=1
  generation=2
  notify_mask=0x0
  rtype=write_exclusive holder=11111111-1111-1111-1111-111111111111,0x1111 reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

2. No holder, multi registrants:
  enable=1
  generation=1
  notify_mask=0x0
  rtype=none
  holder=none
  reg=11111111-1111-1111-1111-111111111111,0x1111
  reg=22222222-2222-2222-2222-222222222222,0x2222

3. No registrant:
  enable=1
  generation=0
  notify_mask=0x0
  rtype=none
  holder=none

4. Reservation not enabled:
  enable=0

Quoted 7 lines

Otherwise, the notify_mask can be stringization:

  notify_mask=reg_preempted,resv_released,resv_preempted

Best Regards,
Guixin Liu

Hi Daniel,
    What do you think?

Best Regards,
Guixin Liu

#6 lore raw
Quoted 41 lines · Guixin Liu
Quoted 41 lines
Quoted 11 lines
Quoted 11 lines

Example output with two registered hosts and an active holder:

   $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
   enable : 1
   generation : 2
   notify_mask : 0x0
   rtype : write_exclusive
   holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
   registrants:
     hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
     hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222
Quoted 3 lines

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

How about this:

1. One holer, multi registrants:
  enable=1
  generation=2
  notify_mask=0x0
  rtype=write_exclusive holder=11111111-1111-1111-1111-111111111111,0x1111 reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

2. No holder, multi registrants:
  enable=1
  generation=1
  notify_mask=0x0
  rtype=none
  holder=none
  reg=11111111-1111-1111-1111-111111111111,0x1111
  reg=22222222-2222-2222-2222-222222222222,0x2222

3. No registrant:
  enable=1
  generation=0
  notify_mask=0x0
  rtype=none
  holder=none

Quoted 6 lines
Quoted 6 lines

4. Reservation not enabled:
  enable=0

Otherwise, the notify_mask can be stringization:

  notify_mask=reg_preempted,resv_released,resv_preempted

Sorry, missed your previous post. This looks easier to parser (*), but I think it's up to Shiinchiro to make the call.

Thanks,
Daniel

(*) I asked a LLM to generate one and it looks straight forward.

#!/bin/bash

# Clean plaintext data with email/quote formatting completely removed DATA=$(cat << 'EOF'

  19   enable=1
  20   generation=2
  21   notify_mask=0x0
  22   rtype=write_exclusive
  23   holder=11111111-1111-1111-1111-111111111111,0x1111
  24   reg=11111111-1111-1111-1111-111111111111,0x1111
  25   reg=22222222-2222-2222-2222-222222222222,0x2222
  26 
  27 2. No holder, multi registrants:
  28   enable=1
  29   generation=1
  30   notify_mask=0x0
  31   rtype=none
  32   holder=none
  33   reg=11111111-1111-1111-1111-111111111111,0x1111
  34   reg=22222222-2222-2222-2222-222222222222,0x2222
  35 
  36 3. No registrant:
  37   enable=1
  38   generation=0
  39   notify_mask=0x0
  40   rtype=none
  41   holder=none

42
43 4. Reservation not enabled:
44 enable=0
EOF
)

# Declare an associative array to store parsed data declare -A INI_DATA

# Tracks the active section namespace
current_section="1. Default"

# Read line-by-line
while IFS= read -r raw_line; do

    # Strip leading line numbers (e.g., "  19 ") and trim surrounding spaces
    line=$(echo "$raw_line" | sed -E 's/^[[:space:]]*[0-9]+[[:space:]]+//; s/^[[:space:]]+//; s/[[:space:]]+$//')

    # Skip empty lines
    [[ -z "$line" ]] && continue

    # Check if the line is a section header (e.g., "2. No holder, multi registrants:")
    if [[ "$line" =~ ^[0-9]+\. ]]; then
        current_section="$line"
        continue
    fi

    # Parse Key=Value configurations
    if [[ "$line" == *"="* ]]; then
        key="${line%%=*}"
        val="${line#*=}"
        
        # Build a compound key: "SectionName|||KeyName"
        map_key="${current_section}|||${key}"

        # If key already exists in this section, append value using a pipe '|' delimiter
        if [[ -n "${INI_DATA[$map_key]}" ]]; then
            INI_DATA[$map_key]+="|$val"
        else
            INI_DATA[$map_key]="$val"
        fi
    fi

done <<< "$DATA"

# --- Demonstrate Accessing the Generated Arrays --- echo "=== PARSED DATA STRUCTURE ==="

for compound_key in "${!INI_DATA[@]}"; do

    section="${compound_key%%|||*}"
    key="${compound_key##*|||}"
    raw_value="${INI_DATA[$compound_key]}"

    echo "[$section]"
    
    # If the value contains our delimiter '|', turn it into a real Bash array
    if [[ "$raw_value" == *"|"* ]]; then
        IFS='|' read -r -a val_array <<< "$raw_value"
        echo "  Key '$key' is an ARRAY (${#val_array[@]} elements):"
        for i in "${!val_array[@]}"; do
            echo "    -> index [$i]: ${val_array[$i]}"
        done
    else
        echo "  Key '$key' is a STRING:"
        echo "    -> value: $raw_value"
    fi
    echo "----------------------------------------"

done

=== PARSED DATA STRUCTURE ===
[3. No registrant:]
Key 'rtype' is a STRING:

    -> value: none
----------------------------------------
[3. No registrant:]
  Key 'generation' is a STRING:
    -> value: 0
----------------------------------------
[1. Default]
  Key 'generation' is a STRING:
    -> value: 2
----------------------------------------
[2. No holder, multi registrants:]
  Key 'holder' is a STRING:
    -> value: none
----------------------------------------
[2. No holder, multi registrants:]
  Key 'reg' is an ARRAY (2 elements):
    -> index [0]: 11111111-1111-1111-1111-111111111111,0x1111
    -> index [1]: 22222222-2222-2222-2222-222222222222,0x2222
----------------------------------------
[2. No holder, multi registrants:]
  Key 'enable' is a STRING:
    -> value: 1
----------------------------------------
[1. Default]
  Key 'enable' is a STRING:
    -> value: 1
----------------------------------------
[1. Default]
  Key 'notify_mask' is a STRING:
    -> value: 0x0
----------------------------------------
[2. No holder, multi registrants:]
  Key 'generation' is a STRING:
    -> value: 1
----------------------------------------
[3. No registrant:]
  Key 'notify_mask' is a STRING:
    -> value: 0x0
----------------------------------------
[1. Default]
  Key 'holder' is a STRING:
    -> value: 11111111-1111-1111-1111-111111111111,0x1111
----------------------------------------
[3. No registrant:]
  Key 'holder' is a STRING:
    -> value: none
----------------------------------------
[2. No holder, multi registrants:]
  Key 'rtype' is a STRING:
    -> value: none
----------------------------------------
[3. No registrant:]
  Key 'enable' is a STRING:
    -> value: 1
----------------------------------------
[4. Reservation not enabled:]
  Key 'enable' is a STRING:
    -> value: 0
----------------------------------------
[1. Default]
  Key 'rtype' is a STRING:
    -> value: write_exclusive
----------------------------------------
[2. No holder, multi registrants:]
  Key 'notify_mask' is a STRING:
    -> value: 0x0
----------------------------------------
[1. Default]
  Key 'reg' is an ARRAY (2 elements):
    -> index [0]: 11111111-1111-1111-1111-111111111111,0x1111
    -> index [1]: 22222222-2222-2222-2222-222222222222,0x2222

----------------------------------------

author
#7 lore raw
Quoted 225 lines · Daniel Wagner
Quoted 41 lines · Guixin Liu
Quoted 41 lines
Quoted 11 lines
Quoted 11 lines

Example output with two registered hosts and an active holder:

    $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
    enable : 1
    generation : 2
    notify_mask : 0x0
    rtype : write_exclusive
    holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
    registrants:
      hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
      hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222
Quoted 3 lines

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

How about this:

1. One holer, multi registrants:
  enable=1
  generation=2
  notify_mask=0x0
  rtype=write_exclusive holder=11111111-1111-1111-1111-111111111111,0x1111 reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

2. No holder, multi registrants:
  enable=1
  generation=1
  notify_mask=0x0
  rtype=none
  holder=none reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

3. No registrant:
  enable=1
  generation=0
  notify_mask=0x0
  rtype=none
  holder=none

Quoted 6 lines
Quoted 6 lines

4. Reservation not enabled:
  enable=0

Otherwise, the notify_mask can be stringization:

  notify_mask=reg_preempted,resv_released,resv_preempted

Sorry, missed your previous post. This looks easier to parser (*), but I think it's up to Shiinchiro to make the call.

Thanks,
Daniel

(*) I asked a LLM to generate one and it looks straight forward.

#!/bin/bash

# Clean plaintext data with email/quote formatting completely removed DATA=$(cat << 'EOF'

   19   enable=1
   20   generation=2
   21   notify_mask=0x0
   22   rtype=write_exclusive
   23   holder=11111111-1111-1111-1111-111111111111,0x1111
   24   reg=11111111-1111-1111-1111-111111111111,0x1111
   25   reg=22222222-2222-2222-2222-222222222222,0x2222
   26
   27 2. No holder, multi registrants:
   28   enable=1
   29   generation=1
   30   notify_mask=0x0
   31   rtype=none
   32   holder=none
   33   reg=11111111-1111-1111-1111-111111111111,0x1111
   34   reg=22222222-2222-2222-2222-222222222222,0x2222
   35
   36 3. No registrant:
   37   enable=1
   38   generation=0
   39   notify_mask=0x0
   40   rtype=none
   41   holder=none
   42
   43 4. Reservation not enabled:
   44   enable=0

EOF
)

# Declare an associative array to store parsed data declare -A INI_DATA

# Tracks the active section namespace
current_section="1. Default"

# Read line-by-line
while IFS= read -r raw_line; do

     # Strip leading line numbers (e.g., "  19 ") and trim surrounding spaces
     line=$(echo "$raw_line" | sed -E 's/^[[:space:]]*[0-9]+[[:space:]]+//; s/^[[:space:]]+//; s/[[:space:]]+$//')

     # Skip empty lines
     [[ -z "$line" ]] && continue

     # Check if the line is a section header (e.g., "2. No holder, multi registrants:")
     if [[ "$line" =~ ^[0-9]+\. ]]; then
         current_section="$line"
         continue
     fi

     # Parse Key=Value configurations
     if [[ "$line" == *"="* ]]; then
         key="${line%%=*}"
         val="${line#*=}"
         
         # Build a compound key: "SectionName|||KeyName"
         map_key="${current_section}|||${key}"

         # If key already exists in this section, append value using a pipe '|' delimiter
         if [[ -n "${INI_DATA[$map_key]}" ]]; then
             INI_DATA[$map_key]+="|$val"
         else
             INI_DATA[$map_key]="$val"
         fi
     fi

done <<< "$DATA"

# --- Demonstrate Accessing the Generated Arrays --- echo "=== PARSED DATA STRUCTURE ==="

for compound_key in "${!INI_DATA[@]}"; do

     section="${compound_key%%|||*}"
     key="${compound_key##*|||}"
     raw_value="${INI_DATA[$compound_key]}"

     echo "[$section]"
     
     # If the value contains our delimiter '|', turn it into a real Bash array
     if [[ "$raw_value" == *"|"* ]]; then
         IFS='|' read -r -a val_array <<< "$raw_value"
         echo "  Key '$key' is an ARRAY (${#val_array[@]} elements):"
         for i in "${!val_array[@]}"; do
             echo "    -> index [$i]: ${val_array[$i]}"
         done
     else
         echo "  Key '$key' is a STRING:"
         echo "    -> value: $raw_value"
     fi
     echo "----------------------------------------"

done

=== PARSED DATA STRUCTURE ===
[3. No registrant:]
Key 'rtype' is a STRING:

     -> value: none
----------------------------------------
[3. No registrant:]
   Key 'generation' is a STRING:
     -> value: 0
----------------------------------------
[1. Default]
   Key 'generation' is a STRING:
     -> value: 2
----------------------------------------
[2. No holder, multi registrants:]
   Key 'holder' is a STRING:
     -> value: none
----------------------------------------
[2. No holder, multi registrants:]
   Key 'reg' is an ARRAY (2 elements):
     -> index [0]: 11111111-1111-1111-1111-111111111111,0x1111
     -> index [1]: 22222222-2222-2222-2222-222222222222,0x2222
----------------------------------------
[2. No holder, multi registrants:]
   Key 'enable' is a STRING:
     -> value: 1
----------------------------------------
[1. Default]
   Key 'enable' is a STRING:
     -> value: 1
----------------------------------------
[1. Default]
   Key 'notify_mask' is a STRING:
     -> value: 0x0
----------------------------------------
[2. No holder, multi registrants:]
   Key 'generation' is a STRING:
     -> value: 1
----------------------------------------
[3. No registrant:]
   Key 'notify_mask' is a STRING:
     -> value: 0x0
----------------------------------------
[1. Default]
   Key 'holder' is a STRING:
     -> value: 11111111-1111-1111-1111-111111111111,0x1111
----------------------------------------
[3. No registrant:]
   Key 'holder' is a STRING:
     -> value: none
----------------------------------------
[2. No holder, multi registrants:]
   Key 'rtype' is a STRING:
     -> value: none
----------------------------------------
[3. No registrant:]
   Key 'enable' is a STRING:
     -> value: 1
----------------------------------------
[4. Reservation not enabled:]
   Key 'enable' is a STRING:
     -> value: 0
----------------------------------------
[1. Default]
   Key 'rtype' is a STRING:
     -> value: write_exclusive
----------------------------------------
[2. No holder, multi registrants:]
   Key 'notify_mask' is a STRING:
     -> value: 0x0
----------------------------------------
[1. Default]
   Key 'reg' is an ARRAY (2 elements):
     -> index [0]: 11111111-1111-1111-1111-111111111111,0x1111
     -> index [1]: 22222222-2222-2222-2222-222222222222,0x2222
Quoted 1 line

----------------------------------------

OK, this seems feasible.

Hi Shin'ichiro:
    Does this work for you?

Best Regards,
Guixin Liu

#8 lore raw
Quoted 47 lines · Guixin Liu
Quoted 44 lines · Daniel Wagner
Quoted 43 lines · Guixin Liu
Quoted 43 lines
Quoted 11 lines
Quoted 11 lines

Example output with two registered hosts and an active holder:

    $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
    enable : 1
    generation : 2
    notify_mask : 0x0
    rtype : write_exclusive
    holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
    registrants:
      hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
      hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222
Quoted 3 lines

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

How about this:

1. One holer, multi registrants:
  enable=1
  generation=2
  notify_mask=0x0
  rtype=write_exclusive holder=11111111-1111-1111-1111-111111111111,0x1111 reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

2. No holder, multi registrants:
  enable=1
  generation=1
  notify_mask=0x0
  rtype=none
  holder=none reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

3. No registrant:
  enable=1
  generation=0
  notify_mask=0x0
  rtype=none
  holder=none

4. Reservation not enabled:
  enable=0

...

Quoted 2 lines

Hi Shin'ichiro:
    Does this work for you?

Hi Guixin, yes, I think it is good enough. If this is sysfs, "one value per file" and "array by files in a directory" would be guided, but this is debugfs readonly data. So, it sounds reasonable to put some data structure into a single debufs attribute file. Uevent like key=value format sounds good. It is a bit wacky to use comma to hold both host id and key for a holder and a registrant, but I cannot think of better solution.

author
#9 lore raw
Quoted 48 lines · Shin'ichiro Kawasaki
Quoted 46 lines · Guixin Liu
Quoted 44 lines · Daniel Wagner
Quoted 43 lines · Guixin Liu
Quoted 43 lines
Quoted 11 lines
Quoted 11 lines

Example output with two registered hosts and an active holder:

     $ cat /sys/kernel/debug/nvmet/testnqn/ns1/reservation
     enable : 1
     generation : 2
     notify_mask : 0x0
     rtype : write_exclusive
     holder : hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
     registrants:
       hostid=11111111-1111-1111-1111-111111111111, rkey=0x1111
       hostid=22222222-2222-2222-2222-222222222222, rkey=0x2222
Quoted 3 lines

I expect blktests will make use of this interface eventually. The parser for this wont be dead simple, sure doable but not simple. If you see a way to make a bit simpler I wouldn't mind.

How about this:

1. One holer, multi registrants:
  enable=1
  generation=2
  notify_mask=0x0
  rtype=write_exclusive holder=11111111-1111-1111-1111-111111111111,0x1111 reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

2. No holder, multi registrants:
  enable=1
  generation=1
  notify_mask=0x0
  rtype=none
  holder=none reg=11111111-1111-1111-1111-111111111111,0x1111 reg=22222222-2222-2222-2222-222222222222,0x2222

3. No registrant:
  enable=1
  generation=0
  notify_mask=0x0
  rtype=none
  holder=none

4. Reservation not enabled:
  enable=0

...

Quoted 8 lines
Quoted 2 lines

Hi Shin'ichiro:
    Does this work for you?

Hi Guixin, yes, I think it is good enough. If this is sysfs, "one value per file" and "array by files in a directory" would be guided, but this is debugfs readonly data. So, it sounds reasonable to put some data structure into a single debufs attribute file. Uevent like key=value format sounds good. It is a bit wacky to use comma to hold both host id and key for a holder and a registrant, but I cannot think of better solution.

OK, I will change this in v2, thanks.

Best Regards,
Guixin Liu