Skip to contentSkip to conversation

Bluetooth: L2CAP: Introduce minimum limit of rx_credits value

PATCH In review Pavel Shpakovskiy 3 participants 3 messages Jun 30, 2025

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Bluetooth: L2CAP: Introduce minimum limit of rx_credits value

The commit 96cd8eaa131f
("Bluetooth: L2CAP: Derive rx credits from MTU and MPS") removed the static rx_credits setup to improve BLE packet communication for high MTU values. However, due to vendor-specific issues in the Bluetooth module firmware, using low MTU values (especially less than 256 bytes) results in dynamically calculated rx_credits being too low, causing slow speeds and occasional BLE connection failures.

This change aims to improve BLE connection stability and speed for low MTU values. It is possible to tune minimum value of rx credits with debugfs handle.

---
 include/net/bluetooth/l2cap.h |  2 ++
 net/bluetooth/l2cap_core.c    | 17 +++++++++++++++--
 2 files changed, 17 insertions(+), 2 deletions(-)
include/net/bluetooth/l2cap.h +2 -0
index 4bb0eaedda180..8648d9324a654 100644
@@ -437,6 +437,8 @@ struct l2cap_conn_param_update_rsp {
 #define L2CAP_CONN_PARAM_ACCEPTED	0x0000
 #define L2CAP_CONN_PARAM_REJECTED	0x0001
 
+#define L2CAP_LE_MIN_CREDITS		10
+
 struct l2cap_le_conn_req {
 	__le16     psm;
 	__le16     scid;
net/bluetooth/l2cap_core.c +15 -2
index c88f69dde995e..392d7ba0f0737 100644
@@ -50,6 +50,8 @@ static u32 l2cap_feat_mask = L2CAP_FEAT_FIXED_CHAN | L2CAP_FEAT_UCD;
 static LIST_HEAD(chan_list);
 static DEFINE_RWLOCK(chan_list_lock);
 
+static u16 le_min_credits = L2CAP_LE_MIN_CREDITS;
+
 static struct sk_buff *l2cap_build_cmd(struct l2cap_conn *conn,
 				       u8 code, u8 ident, u16 dlen, void *data);
 static void l2cap_send_cmd(struct l2cap_conn *conn, u8 ident, u8 code, u16 len,
@@ -547,8 +549,17 @@ static __u16 l2cap_le_rx_credits(struct l2cap_chan *chan)
 	/* If we don't know the available space in the receiver buffer, give
 	 * enough credits for a full packet.
 	 */
-	if (chan->rx_avail == -1)
-		return (chan->imtu / chan->mps) + 1;
+	if (chan->rx_avail == -1) {
+		u16 rx_credits = (chan->imtu / chan->mps) + 1;
+
+		if (rx_credits < le_min_credits) {
+			rx_credits = le_min_credits;
+			BT_DBG("chan %p: set rx_credits to minimum value: %u",
+			       chan, chan->rx_credits);
+		}
+
+		return rx_credits;
+	}
 
 	/* If we know how much space is available in the receive buffer, give
 	 * out as many credits as would fill the buffer.
@@ -7661,6 +7672,8 @@ int __init l2cap_init(void)
 	l2cap_debugfs = debugfs_create_file("l2cap", 0444, bt_debugfs,
 					    NULL, &l2cap_debugfs_fops);
 
+	debugfs_create_u16("l2cap_le_min_credits", 0644, bt_debugfs,
+			   &le_min_credits);
 	return 0;
 }
 
Signature · 1 line
2.34.1
#1 lore raw

This is automated email and please do not reply to this email!

Dear submitter,

Thank you for submitting the patches to the linux bluetooth mailing list. This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=977196

---Test result---

Test Summary:

CheckPatch                    PENDING   0.30 seconds
GitLint                       PENDING   0.24 seconds
SubjectPrefix                 PASS      0.12 seconds
BuildKernel                   PASS      24.68 seconds
CheckAllWarning               PASS      26.93 seconds
CheckSparse                   PASS      30.34 seconds
BuildKernel32                 PASS      24.09 seconds
TestRunnerSetup               PASS      471.10 seconds
TestRunner_l2cap-tester       FAIL      25.49 seconds
TestRunner_iso-tester         PASS      36.43 seconds
TestRunner_bnep-tester        PASS      5.94 seconds
TestRunner_mgmt-tester        FAIL      131.35 seconds
TestRunner_rfcomm-tester      PASS      9.33 seconds
TestRunner_sco-tester         PASS      14.79 seconds
TestRunner_ioctl-tester       PASS      10.06 seconds
TestRunner_mesh-tester        FAIL      9.37 seconds
TestRunner_smp-tester         PASS      8.50 seconds
TestRunner_userchan-tester    PASS      6.22 seconds
IncrementalBuild              PENDING   0.81 seconds

Details
##############################
Test: CheckPatch - PENDING
Desc: Run checkpatch.pl script
Output:

##############################
Test: GitLint - PENDING
Desc: Run gitlint
Output:

##############################
Test: TestRunner_l2cap-tester - FAIL
Desc: Run l2cap-tester with test-runner Output:
Total: 68, Passed: 64 (94.1%), Failed: 4, Not Run: 0

Failed Test Cases

L2CAP LE Server - Success                            Failed       0.112 seconds
L2CAP Ext-Flowctl Server - Success                   Failed       0.132 seconds
L2CAP LE EATT Server - Success                       Failed       0.142 seconds
L2CAP LE EATT Server - Reject                        Failed       0.133 seconds

##############################
Test: TestRunner_mgmt-tester - FAIL
Desc: Run mgmt-tester with test-runner Output:
Total: 490, Passed: 484 (98.8%), Failed: 2, Not Run: 4

Failed Test Cases

LL Privacy - Add Device 2 (2 Devices to AL)          Failed       0.189 seconds
LL Privacy - Set Flags 3 (2 Devices to RL)           Failed       0.201 seconds

##############################
Test: TestRunner_mesh-tester - FAIL
Desc: Run mesh-tester with test-runner
Output:

BUG: KASAN: slab-use-after-free in run_timer_softirq+0x76b/0x7d0
WARNING: CPU: 0 PID: 66 at kernel/workqueue.c:2257 __queue_work+0x93e/0xba0

Total: 10, Passed: 8 (80.0%), Failed: 2, Not Run: 0

Failed Test Cases

Mesh - Send cancel - 1                               Failed       0.144 seconds
Mesh - Send cancel - 2                               Timed out    2.014 seconds

##############################
Test: IncrementalBuild - PENDING
Desc: Incremental build with the patches in the series Output:

---
Regards,
Linux Bluetooth

#2 lore raw
In reply to Pavel Shpakovskiy

Hi Pavel,

Quoted 9 lines · Pavel Shpakovskiy

The commit 96cd8eaa131f
("Bluetooth: L2CAP: Derive rx credits from MTU and MPS") removed the static rx_credits setup to improve BLE packet communication for high MTU values. However, due to vendor-specific issues in the Bluetooth module firmware, using low MTU values (especially less than 256 bytes) results in dynamically calculated rx_credits being too low, causing slow speeds and occasional BLE connection failures.

You will have to be more specific here, what is the use case and model that doesn't work depending on the number of credits? If the idea is to just disable flow control to allow the remote side to pipe more data then the MTU that sort of defeats the purpose of using CoC, but maybe the use case requires or the remote side is too slow to process the updates of credits?

Quoted 39 lines

This change aims to improve BLE connection stability and speed for low MTU values. It is possible to tune minimum value of rx credits with debugfs handle.

---
 include/net/bluetooth/l2cap.h |  2 ++
 net/bluetooth/l2cap_core.c    | 17 +++++++++++++++--
 2 files changed, 17 insertions(+), 2 deletions(-)
include/net/bluetooth/l2cap.h +2 -0
index 4bb0eaedda180..8648d9324a654 100644
@@ -437,6 +437,8 @@ struct l2cap_conn_param_update_rsp {
 #define L2CAP_CONN_PARAM_ACCEPTED      0x0000
 #define L2CAP_CONN_PARAM_REJECTED      0x0001
 
+#define L2CAP_LE_MIN_CREDITS           10
+
 struct l2cap_le_conn_req {
        __le16     psm;
        __le16     scid;
net/bluetooth/l2cap_core.c +2 -0
index c88f69dde995e..392d7ba0f0737 100644
@@ -50,6 +50,8 @@ static u32 l2cap_feat_mask = L2CAP_FEAT_FIXED_CHAN | L2CAP_FEAT_UCD;
 static LIST_HEAD(chan_list);
 static DEFINE_RWLOCK(chan_list_lock);
 
+static u16 le_min_credits = L2CAP_LE_MIN_CREDITS;
+
 static struct sk_buff *l2cap_build_cmd(struct l2cap_conn *conn,
                                       u8 code, u8 ident, u16 dlen, void *data);
 static void l2cap_send_cmd(struct l2cap_conn *conn, u8 ident, u8 code, u16 len,
@@ -547,8 +549,17 @@ static __u16 l2cap_le_rx_credits(struct l2cap_chan *chan)
        /* If we don't know the available space in the receiver buffer, give
         * enough credits for a full packet.

net/bluetooth/l2cap_core.c static __u16 l2cap_le_rx_credits(struct l2cap_chan *chan) · line 558

Quoted 10 lines
net/bluetooth/l2cap_core.c +7 -2
         */
-       if (chan->rx_avail == -1)
-               return (chan->imtu / chan->mps) + 1;
+       if (chan->rx_avail == -1) {
+               u16 rx_credits = (chan->imtu / chan->mps) + 1;
+
+               if (rx_credits < le_min_credits) {
+                       rx_credits = le_min_credits;
+                       BT_DBG("chan %p: set rx_credits to minimum value: %u",
+                              chan, chan->rx_credits);

This doesn't make much sense in my opinion, if we want to disable flow control then we shall allow the remote to pipe as many packets without waiting for more credits, note though rx_credits handling changes after receiving the first packet then the credits are updated based on the socket receiving buffer:

https://github.com/bluez/bluetooth-next/commit/ce60b9231b66710b6ee24042ded26efee120ecfc

So perhaps it is the socket receiving buffer that needs to be adjusted, which is something the process has control over.

Quoted 19 lines
net/bluetooth/l2cap_core.c +6 -1
+               }
+
+               return rx_credits;
+       }
 
        /* If we know how much space is available in the receive buffer, give
         * out as many credits as would fill the buffer.
@@ -7661,6 +7672,8 @@ int __init l2cap_init(void)
        l2cap_debugfs = debugfs_create_file("l2cap", 0444, bt_debugfs,
                                            NULL, &l2cap_debugfs_fops);
 
+       debugfs_create_u16("l2cap_le_min_credits", 0644, bt_debugfs,
+                          &le_min_credits);
        return 0;
 }
 
--

2.34.1

Signature · 1 line
Luiz Augusto von Dentz