* [PATCH 01/10] wifi: ath12k: prevent race condition in ath12k_core_hw_group_destroy()
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:06 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 02/10] wifi: ath12k: add reference counting for core attachment to hardware group Aditya Kumar Singh
` (8 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
Currently, ath12k_core_hw_group_destroy() accesses its members without any
locking mechanism. This could lead to potential issues if these members are
modified by another thread concurrently.
Hence to mitigate this, use the available mutex lock.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Fixes: 6f245ea0ec6c ("wifi: ath12k: introduce device group abstraction")
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 0c6b35aac96eb8d6660cbf30f807a04619feebda..514494dd56ad206a72c52492774d3387dae82f70 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: BSD-3-Clause-Clear
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <linux/module.h>
@@ -1712,6 +1712,7 @@ static void ath12k_core_hw_group_destroy(struct ath12k_hw_group *ag)
if (WARN_ON(!ag))
return;
+ mutex_lock(&ag->mutex);
for (i = 0; i < ag->num_devices; i++) {
ab = ag->ab[i];
if (!ab)
@@ -1719,6 +1720,7 @@ static void ath12k_core_hw_group_destroy(struct ath12k_hw_group *ag)
ath12k_core_soc_destroy(ab);
}
+ mutex_unlock(&ag->mutex);
}
static void ath12k_core_hw_group_cleanup(struct ath12k_hw_group *ag)
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 01/10] wifi: ath12k: prevent race condition in ath12k_core_hw_group_destroy()
2025-01-09 4:25 ` [PATCH 01/10] wifi: ath12k: prevent race condition in ath12k_core_hw_group_destroy() Aditya Kumar Singh
@ 2025-01-13 19:06 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:06 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> Currently, ath12k_core_hw_group_destroy() accesses its members without any
> locking mechanism. This could lead to potential issues if these members are
> modified by another thread concurrently.
>
> Hence to mitigate this, use the available mutex lock.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Fixes: 6f245ea0ec6c ("wifi: ath12k: introduce device group abstraction")
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 02/10] wifi: ath12k: add reference counting for core attachment to hardware group
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
2025-01-09 4:25 ` [PATCH 01/10] wifi: ath12k: prevent race condition in ath12k_core_hw_group_destroy() Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:06 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 03/10] wifi: ath12k: fix failed to set mhi state error during reboot with hardware grouping Aditya Kumar Singh
` (7 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
Currently, driver does not manage reference counting for attaching and
detaching cores to/from hardware groups. This can lead to issues when
multiple cores are involved. Or with same core, attach/detach is called
multiple times back to back.
Fix this issue by using reference counting.
With that, it is now ensured that the core is properly attached or detached
from the hardware group and even back to back calls will not alter the
count.
Additionally, add some debug logs during the attachment and detachment
events for better debugging and tracking.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 47 +++++++++++++++++++++++++++++++---
drivers/net/wireless/ath/ath12k/core.h | 17 ++----------
2 files changed, 45 insertions(+), 19 deletions(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 514494dd56ad206a72c52492774d3387dae82f70..299d7686616b78752164d9cb064c1805af9a1155 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -603,9 +603,49 @@ u32 ath12k_core_get_max_num_tids(struct ath12k_base *ab)
return TARGET_NUM_TIDS(SINGLE);
}
+static inline
+void ath12k_core_to_group_ref_get(struct ath12k_base *ab)
+{
+ struct ath12k_hw_group *ag = ab->ag;
+
+ lockdep_assert_held(&ag->mutex);
+
+ if (ab->hw_group_ref) {
+ ath12k_dbg(ab, ATH12K_DBG_BOOT, "core already attached to group %d\n",
+ ag->id);
+ return;
+ }
+
+ ab->hw_group_ref = true;
+ ag->num_started++;
+
+ ath12k_dbg(ab, ATH12K_DBG_BOOT, "core attached to group %d, num_started %d\n",
+ ag->id, ag->num_started);
+}
+
+static inline
+void ath12k_core_to_group_ref_put(struct ath12k_base *ab)
+{
+ struct ath12k_hw_group *ag = ab->ag;
+
+ lockdep_assert_held(&ag->mutex);
+
+ if (!ab->hw_group_ref) {
+ ath12k_dbg(ab, ATH12K_DBG_BOOT, "core already de-attached from group %d\n",
+ ag->id);
+ return;
+ }
+
+ ab->hw_group_ref = false;
+ ag->num_started--;
+
+ ath12k_dbg(ab, ATH12K_DBG_BOOT, "core de-attached from group %d, num_started %d\n",
+ ag->id, ag->num_started);
+}
+
static void ath12k_core_stop(struct ath12k_base *ab)
{
- ath12k_core_stopped(ab);
+ ath12k_core_to_group_ref_put(ab);
if (!test_bit(ATH12K_FLAG_CRASH_FLUSH, &ab->dev_flags))
ath12k_qmi_firmware_stop(ab);
@@ -841,9 +881,8 @@ static int ath12k_core_start(struct ath12k_base *ab,
/* ACPI is optional so continue in case of an error */
ath12k_dbg(ab, ATH12K_DBG_BOOT, "acpi failed: %d\n", ret);
- if (!test_bit(ATH12K_FLAG_RECOVERY, &ab->dev_flags))
- /* Indicate the core start in the appropriate group */
- ath12k_core_started(ab);
+ /* Indicate the core start in the appropriate group */
+ ath12k_core_to_group_ref_get(ab);
return 0;
diff --git a/drivers/net/wireless/ath/ath12k/core.h b/drivers/net/wireless/ath/ath12k/core.h
index 3dd01ad100c56d2b9330d9a963acba021e4d571b..58ebc56991af99de08e8ed783e98f742a687eddf 100644
--- a/drivers/net/wireless/ath/ath12k/core.h
+++ b/drivers/net/wireless/ath/ath12k/core.h
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause-Clear */
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#ifndef ATH12K_CORE_H
@@ -1052,6 +1052,7 @@ struct ath12k_base {
struct ath12k_hw_group *ag;
struct ath12k_wsi_info wsi_info;
+ bool hw_group_ref;
/* must be last */
u8 drv_priv[] __aligned(sizeof(void *));
@@ -1215,20 +1216,6 @@ static inline struct ath12k_hw_group *ath12k_ab_to_ag(struct ath12k_base *ab)
return ab->ag;
}
-static inline void ath12k_core_started(struct ath12k_base *ab)
-{
- lockdep_assert_held(&ab->ag->mutex);
-
- ab->ag->num_started++;
-}
-
-static inline void ath12k_core_stopped(struct ath12k_base *ab)
-{
- lockdep_assert_held(&ab->ag->mutex);
-
- ab->ag->num_started--;
-}
-
static inline struct ath12k_base *ath12k_ag_to_ab(struct ath12k_hw_group *ag,
u8 device_id)
{
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 02/10] wifi: ath12k: add reference counting for core attachment to hardware group
2025-01-09 4:25 ` [PATCH 02/10] wifi: ath12k: add reference counting for core attachment to hardware group Aditya Kumar Singh
@ 2025-01-13 19:06 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:06 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> Currently, driver does not manage reference counting for attaching and
> detaching cores to/from hardware groups. This can lead to issues when
> multiple cores are involved. Or with same core, attach/detach is called
> multiple times back to back.
>
> Fix this issue by using reference counting.
>
> With that, it is now ensured that the core is properly attached or detached
> from the hardware group and even back to back calls will not alter the
> count.
>
> Additionally, add some debug logs during the attachment and detachment
> events for better debugging and tracking.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 03/10] wifi: ath12k: fix failed to set mhi state error during reboot with hardware grouping
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
2025-01-09 4:25 ` [PATCH 01/10] wifi: ath12k: prevent race condition in ath12k_core_hw_group_destroy() Aditya Kumar Singh
2025-01-09 4:25 ` [PATCH 02/10] wifi: ath12k: add reference counting for core attachment to hardware group Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:07 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 04/10] wifi: ath12k: fix firmware assert " Aditya Kumar Singh
` (6 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
With hardware grouping, during reboot, whenever a device is removed, it
powers down itself and all its partner devices in the same group. Now this
is done by all devices and hence there is multiple power down for devices
and hence the following error messages can be seen:
ath12k_pci 0002:01:00.0: failed to set mhi state POWER_OFF(3) in current mhi state (0x0)
ath12k_pci 0002:01:00.0: failed to set mhi state: POWER_OFF(3)
ath12k_pci 0002:01:00.0: failed to set mhi state DEINIT(1) in current mhi state (0x0)
ath12k_pci 0002:01:00.0: failed to set mhi state: DEINIT(1)
ath12k_pci 0003:01:00.0: failed to set mhi state POWER_OFF(3) in current mhi state (0x0)
ath12k_pci 0003:01:00.0: failed to set mhi state: POWER_OFF(3)
ath12k_pci 0003:01:00.0: failed to set mhi state DEINIT(1) in current mhi state (0x0)
ath12k_pci 0003:01:00.0: failed to set mhi state: DEINIT(1)
ath12k_pci 0004:01:00.0: failed to set mhi state POWER_OFF(3) in current mhi state (0x0)
ath12k_pci 0004:01:00.0: failed to set mhi state: POWER_OFF(3)
To prevent this, check if the ATH12K_PCI_FLAG_INIT_DONE flag is already
set before powering down. If it is set, it indicates that another partner
device has already performed the power down, and this device can skip this
step.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/pci.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/net/wireless/ath/ath12k/pci.c b/drivers/net/wireless/ath/ath12k/pci.c
index 06cff3849ab8da3b39677bed3d6ee60af2c814d1..837be309cd45a2d037ee8c3bba8f7be0f457d6b2 100644
--- a/drivers/net/wireless/ath/ath12k/pci.c
+++ b/drivers/net/wireless/ath/ath12k/pci.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: BSD-3-Clause-Clear
/*
* Copyright (c) 2019-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <linux/module.h>
@@ -1484,6 +1484,9 @@ void ath12k_pci_power_down(struct ath12k_base *ab, bool is_suspend)
{
struct ath12k_pci *ab_pci = ath12k_pci_priv(ab);
+ if (!test_bit(ATH12K_PCI_FLAG_INIT_DONE, &ab_pci->flags))
+ return;
+
/* restore aspm in case firmware bootup fails */
ath12k_pci_aspm_restore(ab_pci);
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 03/10] wifi: ath12k: fix failed to set mhi state error during reboot with hardware grouping
2025-01-09 4:25 ` [PATCH 03/10] wifi: ath12k: fix failed to set mhi state error during reboot with hardware grouping Aditya Kumar Singh
@ 2025-01-13 19:07 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:07 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> With hardware grouping, during reboot, whenever a device is removed, it
> powers down itself and all its partner devices in the same group. Now this
> is done by all devices and hence there is multiple power down for devices
> and hence the following error messages can be seen:
>
> ath12k_pci 0002:01:00.0: failed to set mhi state POWER_OFF(3) in current mhi state (0x0)
> ath12k_pci 0002:01:00.0: failed to set mhi state: POWER_OFF(3)
> ath12k_pci 0002:01:00.0: failed to set mhi state DEINIT(1) in current mhi state (0x0)
> ath12k_pci 0002:01:00.0: failed to set mhi state: DEINIT(1)
> ath12k_pci 0003:01:00.0: failed to set mhi state POWER_OFF(3) in current mhi state (0x0)
> ath12k_pci 0003:01:00.0: failed to set mhi state: POWER_OFF(3)
> ath12k_pci 0003:01:00.0: failed to set mhi state DEINIT(1) in current mhi state (0x0)
> ath12k_pci 0003:01:00.0: failed to set mhi state: DEINIT(1)
> ath12k_pci 0004:01:00.0: failed to set mhi state POWER_OFF(3) in current mhi state (0x0)
> ath12k_pci 0004:01:00.0: failed to set mhi state: POWER_OFF(3)
>
> To prevent this, check if the ATH12K_PCI_FLAG_INIT_DONE flag is already
> set before powering down. If it is set, it indicates that another partner
> device has already performed the power down, and this device can skip this
> step.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 04/10] wifi: ath12k: fix firmware assert during reboot with hardware grouping
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (2 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 03/10] wifi: ath12k: fix failed to set mhi state error during reboot with hardware grouping Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:12 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 05/10] wifi: ath12k: fix SLUB BUG - Object already free in ath12k_reg_free() Aditya Kumar Singh
` (5 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
At present, during PCI shutdown, the power down is only executed for a
single device. However, when operating in a group, all devices need to be
powered down simultaneously. Failure to do so will result in a firmware
assertion.
Hence, introduce a new ath12k_pci_hw_group_power_down() and call it during
power down. This will ensure that all partner devices are properly powered
down.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/pci.c | 23 ++++++++++++++++++++++-
1 file changed, 22 insertions(+), 1 deletion(-)
diff --git a/drivers/net/wireless/ath/ath12k/pci.c b/drivers/net/wireless/ath/ath12k/pci.c
index 837be309cd45a2d037ee8c3bba8f7be0f457d6b2..7f6521a56ffc0f1e9687c94d6829a9c1f1887661 100644
--- a/drivers/net/wireless/ath/ath12k/pci.c
+++ b/drivers/net/wireless/ath/ath12k/pci.c
@@ -1751,13 +1751,34 @@ static void ath12k_pci_remove(struct pci_dev *pdev)
ath12k_core_free(ab);
}
+static void ath12k_pci_hw_group_power_down(struct ath12k_hw_group *ag)
+{
+ struct ath12k_base *ab;
+ int i;
+
+ if (!ag)
+ return;
+
+ mutex_lock(&ag->mutex);
+
+ for (i = 0; i < ag->num_devices; i++) {
+ ab = ag->ab[i];
+ if (!ab)
+ continue;
+
+ ath12k_pci_power_down(ab, false);
+ }
+
+ mutex_unlock(&ag->mutex);
+}
+
static void ath12k_pci_shutdown(struct pci_dev *pdev)
{
struct ath12k_base *ab = pci_get_drvdata(pdev);
struct ath12k_pci *ab_pci = ath12k_pci_priv(ab);
ath12k_pci_set_irq_affinity_hint(ab_pci, NULL);
- ath12k_pci_power_down(ab, false);
+ ath12k_pci_hw_group_power_down(ab->ag);
}
static __maybe_unused int ath12k_pci_pm_suspend(struct device *dev)
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 04/10] wifi: ath12k: fix firmware assert during reboot with hardware grouping
2025-01-09 4:25 ` [PATCH 04/10] wifi: ath12k: fix firmware assert " Aditya Kumar Singh
@ 2025-01-13 19:12 ` Jeff Johnson
2025-01-16 10:43 ` Aditya Kumar Singh
0 siblings, 1 reply; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:12 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> At present, during PCI shutdown, the power down is only executed for a
> single device. However, when operating in a group, all devices need to be
> powered down simultaneously. Failure to do so will result in a firmware
> assertion.
>
> Hence, introduce a new ath12k_pci_hw_group_power_down() and call it during
> power down. This will ensure that all partner devices are properly powered
> down.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
> ---
> drivers/net/wireless/ath/ath12k/pci.c | 23 ++++++++++++++++++++++-
> 1 file changed, 22 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/wireless/ath/ath12k/pci.c b/drivers/net/wireless/ath/ath12k/pci.c
> index 837be309cd45a2d037ee8c3bba8f7be0f457d6b2..7f6521a56ffc0f1e9687c94d6829a9c1f1887661 100644
> --- a/drivers/net/wireless/ath/ath12k/pci.c
> +++ b/drivers/net/wireless/ath/ath12k/pci.c
> @@ -1751,13 +1751,34 @@ static void ath12k_pci_remove(struct pci_dev *pdev)
> ath12k_core_free(ab);
> }
>
> +static void ath12k_pci_hw_group_power_down(struct ath12k_hw_group *ag)
don't you end up calling this for every device in the group?
what prevents ath12k_pci_power_down(ab, false) from being called multiple
times for the same ab?
> +{
> + struct ath12k_base *ab;
> + int i;
> +
> + if (!ag)
> + return;
> +
> + mutex_lock(&ag->mutex);
> +
> + for (i = 0; i < ag->num_devices; i++) {
> + ab = ag->ab[i];
> + if (!ab)
> + continue;
> +
> + ath12k_pci_power_down(ab, false);
> + }
> +
> + mutex_unlock(&ag->mutex);
> +}
> +
> static void ath12k_pci_shutdown(struct pci_dev *pdev)
> {
> struct ath12k_base *ab = pci_get_drvdata(pdev);
> struct ath12k_pci *ab_pci = ath12k_pci_priv(ab);
>
> ath12k_pci_set_irq_affinity_hint(ab_pci, NULL);
> - ath12k_pci_power_down(ab, false);
> + ath12k_pci_hw_group_power_down(ab->ag);
> }
>
> static __maybe_unused int ath12k_pci_pm_suspend(struct device *dev)
>
^ permalink raw reply [flat|nested] 23+ messages in thread* Re: [PATCH 04/10] wifi: ath12k: fix firmware assert during reboot with hardware grouping
2025-01-13 19:12 ` Jeff Johnson
@ 2025-01-16 10:43 ` Aditya Kumar Singh
0 siblings, 0 replies; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-16 10:43 UTC (permalink / raw)
To: Jeff Johnson, Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy,
Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/14/25 00:42, Jeff Johnson wrote:
> On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
>> At present, during PCI shutdown, the power down is only executed for a
>> single device. However, when operating in a group, all devices need to be
>> powered down simultaneously. Failure to do so will result in a firmware
>> assertion.
>>
>> Hence, introduce a new ath12k_pci_hw_group_power_down() and call it during
>> power down. This will ensure that all partner devices are properly powered
>> down.
>>
>> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>>
>> Signed-off-by: Aditya Kumar Singh<quic_adisi@quicinc.com>
>> ---
>> drivers/net/wireless/ath/ath12k/pci.c | 23 ++++++++++++++++++++++-
>> 1 file changed, 22 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/wireless/ath/ath12k/pci.c b/drivers/net/wireless/ath/ath12k/pci.c
>> index 837be309cd45a2d037ee8c3bba8f7be0f457d6b2..7f6521a56ffc0f1e9687c94d6829a9c1f1887661 100644
>> --- a/drivers/net/wireless/ath/ath12k/pci.c
>> +++ b/drivers/net/wireless/ath/ath12k/pci.c
>> @@ -1751,13 +1751,34 @@ static void ath12k_pci_remove(struct pci_dev *pdev)
>> ath12k_core_free(ab);
>> }
>>
>> +static void ath12k_pci_hw_group_power_down(struct ath12k_hw_group *ag)
> don't you end up calling this for every device in the group?
> what prevents ath12k_pci_power_down(ab, false) from being called multiple
> times for the same ab?
That's true. ath12k_pci_power_down() has logic already that if device is
powered down, it will ignore the further call. This is handled via
previous patch.
--
Aditya
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 05/10] wifi: ath12k: fix SLUB BUG - Object already free in ath12k_reg_free()
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (3 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 04/10] wifi: ath12k: fix firmware assert " Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:21 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 06/10] wifi: ath12k: fix ath12k_core_pre_reconfigure_recovery() with grouping Aditya Kumar Singh
` (4 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
During rmmod of ath12k module with SLUB debug enabled, following print is
seen -
=============================================================================
BUG kmalloc-1k (Not tainted): Object already free
-----------------------------------------------------------------------------
Allocated in ath12k_reg_build_regd+0x94/0xa20 [ath12k] age=10470 cpu=0 pid=0
__kmalloc_noprof+0xf4/0x368
ath12k_reg_build_regd+0x94/0xa20 [ath12k]
ath12k_wmi_op_rx+0x199c/0x2c14 [ath12k]
ath12k_htc_rx_completion_handler+0x398/0x554 [ath12k]
ath12k_ce_per_engine_service+0x248/0x368 [ath12k]
ath12k_pci_ce_workqueue+0x28/0x50 [ath12k]
process_one_work+0x14c/0x28c
bh_worker+0x22c/0x27c
workqueue_softirq_action+0x80/0x90
tasklet_action+0x14/0x3c
handle_softirqs+0x108/0x240
__do_softirq+0x14/0x20
Freed in ath12k_reg_free+0x40/0x74 [ath12k] age=136 cpu=2 pid=166
kfree+0x148/0x248
ath12k_reg_free+0x40/0x74 [ath12k]
ath12k_core_hw_group_destroy+0x68/0xac [ath12k]
ath12k_core_deinit+0xd8/0x124 [ath12k]
ath12k_pci_remove+0x6c/0x130 [ath12k]
pci_device_remove+0x44/0xe8
device_remove+0x4c/0x80
device_release_driver_internal+0x1d0/0x22c
driver_detach+0x50/0x98
bus_remove_driver+0x70/0xf4
driver_unregister+0x30/0x60
pci_unregister_driver+0x24/0x9c
ath12k_pci_exit+0x18/0x24 [ath12k]
__arm64_sys_delete_module+0x1a0/0x2a8
invoke_syscall+0x48/0x110
el0_svc_common.constprop.0+0x40/0xe0
Slab 0xfffffdffc0033600 objects=10 used=6 fp=0xffff000000cdcc00 flags=0x3fffe0000000240(workingset|head|node=0|zone=0|lastcpupid=0x1ffff)
Object 0xffff000000cdcc00 @offset=19456 fp=0xffff000000cde400
[...]
This issue arises because in ath12k_core_hw_group_destroy(), each device
calls ath12k_core_soc_destroy() for itself and all its partners within the
same group. Since ath12k_core_hw_group_destroy() is invoked for each
device, this results in a double free condition, eventually causing the
SLUB bug.
To resolve this, a new member regd_freed is introduced in the ath12k_base
object. Once regd is freed, regd_freed is set to true. This ensures that
in the removal context of other devices, regd is not freed again if
regd_freed is already true. And since there could be a race condition to
read this member, guard ath12k_core_soc_destroy() with the mutext lock.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Fixes: 6f245ea0ec6c ("wifi: ath12k: introduce device group abstraction")
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 2 ++
drivers/net/wireless/ath/ath12k/core.h | 1 +
drivers/net/wireless/ath/ath12k/reg.c | 8 +++++++-
drivers/net/wireless/ath/ath12k/wmi.c | 4 +++-
4 files changed, 13 insertions(+), 2 deletions(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 299d7686616b78752164d9cb064c1805af9a1155..72e6e3a0cf7be03b20b7421866c479dfcb8038ff 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1757,7 +1757,9 @@ static void ath12k_core_hw_group_destroy(struct ath12k_hw_group *ag)
if (!ab)
continue;
+ mutex_lock(&ab->core_lock);
ath12k_core_soc_destroy(ab);
+ mutex_unlock(&ab->core_lock);
}
mutex_unlock(&ag->mutex);
}
diff --git a/drivers/net/wireless/ath/ath12k/core.h b/drivers/net/wireless/ath/ath12k/core.h
index 58ebc56991af99de08e8ed783e98f742a687eddf..cc1bfcc1e65c87e30d86dad4c0bcd1905e6a2f51 100644
--- a/drivers/net/wireless/ath/ath12k/core.h
+++ b/drivers/net/wireless/ath/ath12k/core.h
@@ -961,6 +961,7 @@ struct ath12k_base {
* This may or may not be used during the runtime
*/
struct ieee80211_regdomain *new_regd[MAX_RADIOS];
+ bool regd_freed;
/* Current DFS Regulatory */
enum ath12k_dfs_region dfs_region;
diff --git a/drivers/net/wireless/ath/ath12k/reg.c b/drivers/net/wireless/ath/ath12k/reg.c
index 439d61f284d89222e79c05d6cff8e85d0d315aad..b4d7fa1a04ca0e72728e8989c29b82d089171fc2 100644
--- a/drivers/net/wireless/ath/ath12k/reg.c
+++ b/drivers/net/wireless/ath/ath12k/reg.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: BSD-3-Clause-Clear
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <linux/rtnetlink.h>
#include "core.h"
@@ -777,8 +777,14 @@ void ath12k_reg_free(struct ath12k_base *ab)
{
int i;
+ if (ab->regd_freed)
+ return;
+
for (i = 0; i < ab->hw_params->max_radios; i++) {
kfree(ab->default_regd[i]);
kfree(ab->new_regd[i]);
+ ab->default_regd[i] = NULL;
+ ab->new_regd[i] = NULL;
+ ab->regd_freed = true;
}
}
diff --git a/drivers/net/wireless/ath/ath12k/wmi.c b/drivers/net/wireless/ath/ath12k/wmi.c
index 4dd6cdf84571d3652cd03281ffa6486e3d340c42..1de6ed6cceaee3a22de63a2369358fe53fb0d638 100644
--- a/drivers/net/wireless/ath/ath12k/wmi.c
+++ b/drivers/net/wireless/ath/ath12k/wmi.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: BSD-3-Clause-Clear
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <linux/skbuff.h>
#include <linux/ctype.h>
@@ -5950,6 +5950,8 @@ static int ath12k_reg_chan_list_event(struct ath12k_base *ab, struct sk_buff *sk
/* This regd would be applied during mac registration */
ab->default_regd[pdev_idx] = regd;
}
+
+ ab->regd_freed = false;
ab->dfs_region = reg_info->dfs_region;
spin_unlock(&ab->base_lock);
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 05/10] wifi: ath12k: fix SLUB BUG - Object already free in ath12k_reg_free()
2025-01-09 4:25 ` [PATCH 05/10] wifi: ath12k: fix SLUB BUG - Object already free in ath12k_reg_free() Aditya Kumar Singh
@ 2025-01-13 19:21 ` Jeff Johnson
2025-01-20 8:38 ` Aditya Kumar Singh
0 siblings, 1 reply; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:21 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> During rmmod of ath12k module with SLUB debug enabled, following print is
> seen -
>
> =============================================================================
> BUG kmalloc-1k (Not tainted): Object already free
> -----------------------------------------------------------------------------
>
> Allocated in ath12k_reg_build_regd+0x94/0xa20 [ath12k] age=10470 cpu=0 pid=0
> __kmalloc_noprof+0xf4/0x368
> ath12k_reg_build_regd+0x94/0xa20 [ath12k]
> ath12k_wmi_op_rx+0x199c/0x2c14 [ath12k]
> ath12k_htc_rx_completion_handler+0x398/0x554 [ath12k]
> ath12k_ce_per_engine_service+0x248/0x368 [ath12k]
> ath12k_pci_ce_workqueue+0x28/0x50 [ath12k]
> process_one_work+0x14c/0x28c
> bh_worker+0x22c/0x27c
> workqueue_softirq_action+0x80/0x90
> tasklet_action+0x14/0x3c
> handle_softirqs+0x108/0x240
> __do_softirq+0x14/0x20
> Freed in ath12k_reg_free+0x40/0x74 [ath12k] age=136 cpu=2 pid=166
> kfree+0x148/0x248
> ath12k_reg_free+0x40/0x74 [ath12k]
> ath12k_core_hw_group_destroy+0x68/0xac [ath12k]
> ath12k_core_deinit+0xd8/0x124 [ath12k]
> ath12k_pci_remove+0x6c/0x130 [ath12k]
> pci_device_remove+0x44/0xe8
> device_remove+0x4c/0x80
> device_release_driver_internal+0x1d0/0x22c
> driver_detach+0x50/0x98
> bus_remove_driver+0x70/0xf4
> driver_unregister+0x30/0x60
> pci_unregister_driver+0x24/0x9c
> ath12k_pci_exit+0x18/0x24 [ath12k]
> __arm64_sys_delete_module+0x1a0/0x2a8
> invoke_syscall+0x48/0x110
> el0_svc_common.constprop.0+0x40/0xe0
> Slab 0xfffffdffc0033600 objects=10 used=6 fp=0xffff000000cdcc00 flags=0x3fffe0000000240(workingset|head|node=0|zone=0|lastcpupid=0x1ffff)
> Object 0xffff000000cdcc00 @offset=19456 fp=0xffff000000cde400
> [...]
>
> This issue arises because in ath12k_core_hw_group_destroy(), each device
> calls ath12k_core_soc_destroy() for itself and all its partners within the
> same group. Since ath12k_core_hw_group_destroy() is invoked for each
> device, this results in a double free condition, eventually causing the
> SLUB bug.
>
> To resolve this, a new member regd_freed is introduced in the ath12k_base
> object. Once regd is freed, regd_freed is set to true. This ensures that
> in the removal context of other devices, regd is not freed again if
> regd_freed is already true. And since there could be a race condition to
> read this member, guard ath12k_core_soc_destroy() with the mutext lock.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Fixes: 6f245ea0ec6c ("wifi: ath12k: introduce device group abstraction")
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
> ---
> drivers/net/wireless/ath/ath12k/core.c | 2 ++
> drivers/net/wireless/ath/ath12k/core.h | 1 +
> drivers/net/wireless/ath/ath12k/reg.c | 8 +++++++-
> drivers/net/wireless/ath/ath12k/wmi.c | 4 +++-
> 4 files changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
> index 299d7686616b78752164d9cb064c1805af9a1155..72e6e3a0cf7be03b20b7421866c479dfcb8038ff 100644
> --- a/drivers/net/wireless/ath/ath12k/core.c
> +++ b/drivers/net/wireless/ath/ath12k/core.c
> @@ -1757,7 +1757,9 @@ static void ath12k_core_hw_group_destroy(struct ath12k_hw_group *ag)
> if (!ab)
> continue;
>
> + mutex_lock(&ab->core_lock);
> ath12k_core_soc_destroy(ab);
> + mutex_unlock(&ab->core_lock);
> }
> mutex_unlock(&ag->mutex);
> }
> diff --git a/drivers/net/wireless/ath/ath12k/core.h b/drivers/net/wireless/ath/ath12k/core.h
> index 58ebc56991af99de08e8ed783e98f742a687eddf..cc1bfcc1e65c87e30d86dad4c0bcd1905e6a2f51 100644
> --- a/drivers/net/wireless/ath/ath12k/core.h
> +++ b/drivers/net/wireless/ath/ath12k/core.h
> @@ -961,6 +961,7 @@ struct ath12k_base {
> * This may or may not be used during the runtime
> */
> struct ieee80211_regdomain *new_regd[MAX_RADIOS];
> + bool regd_freed;
>
> /* Current DFS Regulatory */
> enum ath12k_dfs_region dfs_region;
> diff --git a/drivers/net/wireless/ath/ath12k/reg.c b/drivers/net/wireless/ath/ath12k/reg.c
> index 439d61f284d89222e79c05d6cff8e85d0d315aad..b4d7fa1a04ca0e72728e8989c29b82d089171fc2 100644
> --- a/drivers/net/wireless/ath/ath12k/reg.c
> +++ b/drivers/net/wireless/ath/ath12k/reg.c
> @@ -1,7 +1,7 @@
> // SPDX-License-Identifier: BSD-3-Clause-Clear
> /*
> * Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
> - * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
> + * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
> */
> #include <linux/rtnetlink.h>
> #include "core.h"
> @@ -777,8 +777,14 @@ void ath12k_reg_free(struct ath12k_base *ab)
> {
> int i;
>
> + if (ab->regd_freed)
> + return;
> +
> for (i = 0; i < ab->hw_params->max_radios; i++) {
> kfree(ab->default_regd[i]);
> kfree(ab->new_regd[i]);
> + ab->default_regd[i] = NULL;
> + ab->new_regd[i] = NULL;
> + ab->regd_freed = true;
since it is loop invariant, should this last assignment be outside the loop,
either before or after the loop?
but then again, why is a flag needed since setting the pointers to NULL should
already show they are freed, and any race conditions with those pointers would
also exist with the new flag (which you have addressed with the locking change).
> }
> }
> diff --git a/drivers/net/wireless/ath/ath12k/wmi.c b/drivers/net/wireless/ath/ath12k/wmi.c
> index 4dd6cdf84571d3652cd03281ffa6486e3d340c42..1de6ed6cceaee3a22de63a2369358fe53fb0d638 100644
> --- a/drivers/net/wireless/ath/ath12k/wmi.c
> +++ b/drivers/net/wireless/ath/ath12k/wmi.c
> @@ -1,7 +1,7 @@
> // SPDX-License-Identifier: BSD-3-Clause-Clear
> /*
> * Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
> - * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
> + * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
> */
> #include <linux/skbuff.h>
> #include <linux/ctype.h>
> @@ -5950,6 +5950,8 @@ static int ath12k_reg_chan_list_event(struct ath12k_base *ab, struct sk_buff *sk
> /* This regd would be applied during mac registration */
> ab->default_regd[pdev_idx] = regd;
> }
> +
> + ab->regd_freed = false;
> ab->dfs_region = reg_info->dfs_region;
> spin_unlock(&ab->base_lock);
>
>
^ permalink raw reply [flat|nested] 23+ messages in thread* Re: [PATCH 05/10] wifi: ath12k: fix SLUB BUG - Object already free in ath12k_reg_free()
2025-01-13 19:21 ` Jeff Johnson
@ 2025-01-20 8:38 ` Aditya Kumar Singh
0 siblings, 0 replies; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-20 8:38 UTC (permalink / raw)
To: Jeff Johnson, Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy,
Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/14/25 00:51, Jeff Johnson wrote:
>> diff --git a/drivers/net/wireless/ath/ath12k/reg.c b/drivers/net/wireless/ath/ath12k/reg.c
>> index 439d61f284d89222e79c05d6cff8e85d0d315aad..b4d7fa1a04ca0e72728e8989c29b82d089171fc2 100644
>> --- a/drivers/net/wireless/ath/ath12k/reg.c
>> +++ b/drivers/net/wireless/ath/ath12k/reg.c
>> @@ -1,7 +1,7 @@
>> // SPDX-License-Identifier: BSD-3-Clause-Clear
>> /*
>> * Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
>> - * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
>> + * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
>> */
>> #include <linux/rtnetlink.h>
>> #include "core.h"
>> @@ -777,8 +777,14 @@ void ath12k_reg_free(struct ath12k_base *ab)
>> {
>> int i;
>>
>> + if (ab->regd_freed)
>> + return;
>> +
>> for (i = 0; i < ab->hw_params->max_radios; i++) {
>> kfree(ab->default_regd[i]);
>> kfree(ab->new_regd[i]);
>> + ab->default_regd[i] = NULL;
>> + ab->new_regd[i] = NULL;
>> + ab->regd_freed = true;
> since it is loop invariant, should this last assignment be outside the loop,
> either before or after the loop?
>
> but then again, why is a flag needed since setting the pointers to NULL should
> already show they are freed, and any race conditions with those pointers would
> also exist with the new flag (which you have addressed with the locking change).
Well, looks like, this flag is not needed. I will remove this in next
version. Thanks for pointing it out!
--
Aditya
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 06/10] wifi: ath12k: fix ath12k_core_pre_reconfigure_recovery() with grouping
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (4 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 05/10] wifi: ath12k: fix SLUB BUG - Object already free in ath12k_reg_free() Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:22 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 07/10] wifi: ath12k: fix ATH12K_FLAG_REGISTERED flag handling Aditya Kumar Singh
` (3 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
Currently, ath12k_core_pre_reconfigure_recovery() reconfigures all radios
within the same group. During grouping and driver going for a recovery,
this function is called as many times as there are devices in the group.
Consequently, it performs the same reconfiguration multiple times, which
is unnecessary.
To prevent this, add a check to continue if the action has already been
taken.
To simplify the management of various flags, the reason for hardware queues
being stopped is used as a check instead of introducing a new flag.
While at it, also add missing wiphy locks. Wiphy lock is required since
ath12k_mac_drain_tx() which is called by
ath12k_core_pre_reconfigure_recovery() needs this lock to be held by the
caller.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 72e6e3a0cf7be03b20b7421866c479dfcb8038ff..5700fc661ac380b6c01d0571595d27fb1ab7c8c5 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1279,6 +1279,18 @@ static void ath12k_core_pre_reconfigure_recovery(struct ath12k_base *ab)
if (!ah || ah->state == ATH12K_HW_STATE_OFF)
continue;
+ wiphy_lock(ah->hw->wiphy);
+
+ /* If queue 0 is stopped, it is safe to assume that all
+ * other queues are stopped by driver via
+ * ieee80211_stop_queues() below. This means, there is
+ * no need to stop it again and hence continue
+ */
+ if (ieee80211_queue_stopped(ah->hw, 0)) {
+ wiphy_unlock(ah->hw->wiphy);
+ continue;
+ }
+
ieee80211_stop_queues(ah->hw);
for (j = 0; j < ah->num_radio; j++) {
@@ -1301,6 +1313,8 @@ static void ath12k_core_pre_reconfigure_recovery(struct ath12k_base *ab)
idr_destroy(&ar->txmgmt_idr);
wake_up(&ar->txmgmt_empty_waitq);
}
+
+ wiphy_unlock(ah->hw->wiphy);
}
wake_up(&ab->wmi_ab.tx_credits_wq);
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 06/10] wifi: ath12k: fix ath12k_core_pre_reconfigure_recovery() with grouping
2025-01-09 4:25 ` [PATCH 06/10] wifi: ath12k: fix ath12k_core_pre_reconfigure_recovery() with grouping Aditya Kumar Singh
@ 2025-01-13 19:22 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:22 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> Currently, ath12k_core_pre_reconfigure_recovery() reconfigures all radios
> within the same group. During grouping and driver going for a recovery,
> this function is called as many times as there are devices in the group.
> Consequently, it performs the same reconfiguration multiple times, which
> is unnecessary.
>
> To prevent this, add a check to continue if the action has already been
> taken.
>
> To simplify the management of various flags, the reason for hardware queues
> being stopped is used as a check instead of introducing a new flag.
>
> While at it, also add missing wiphy locks. Wiphy lock is required since
> ath12k_mac_drain_tx() which is called by
> ath12k_core_pre_reconfigure_recovery() needs this lock to be held by the
> caller.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 07/10] wifi: ath12k: fix ATH12K_FLAG_REGISTERED flag handling
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (5 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 06/10] wifi: ath12k: fix ath12k_core_pre_reconfigure_recovery() with grouping Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:23 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 08/10] wifi: ath12k: handle ath12k_core_restart() with hardware grouping Aditya Kumar Singh
` (2 subsequent siblings)
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
Commit a5686ae820fa ("wifi: ath12k: move ATH12K_FLAG_REGISTERED handling to
ath12k_mac_register()") relocated the setting of the ATH12K_FLAG_REGISTERED
flag to the ath12k_mac_register() function. However, this function only
accesses the first device (ab) via ag->ab[0], resulting in the flag being
set only for the first device in the group. Similarly,
ath12k_mac_unregister() only unsets the flag for the first device. The flag
should actually be set for all devices in the group to avoid issues during
recovery.
Hence, move setting and clearing of this flag in the function
ath12k_core_hw_group_start() and ath12k_core_hw_group_stop() respectively.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Fixes: a5686ae820fa ("wifi: ath12k: move ATH12K_FLAG_REGISTERED handling to ath12k_mac_register()")
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 5 +++++
drivers/net/wireless/ath/ath12k/mac.c | 6 +-----
2 files changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 5700fc661ac380b6c01d0571595d27fb1ab7c8c5..b67ef79e62b3fbb5667cb627cf565998a35f3c49 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -920,6 +920,9 @@ static void ath12k_core_hw_group_stop(struct ath12k_hw_group *ag)
ab = ag->ab[i];
if (!ab)
continue;
+
+ clear_bit(ATH12K_FLAG_REGISTERED, &ab->dev_flags);
+
ath12k_core_device_cleanup(ab);
}
@@ -1025,6 +1028,8 @@ static int ath12k_core_hw_group_start(struct ath12k_hw_group *ag)
mutex_lock(&ab->core_lock);
+ set_bit(ATH12K_FLAG_REGISTERED, &ab->dev_flags);
+
ret = ath12k_core_pdev_create(ab);
if (ret) {
ath12k_err(ab, "failed to create pdev core %d\n", ret);
diff --git a/drivers/net/wireless/ath/ath12k/mac.c b/drivers/net/wireless/ath/ath12k/mac.c
index 48d110e2a7ded61c4094b0ce7e5bbb50b94d5cd4..1ff141ee9e94fed6ac954c0e411a0f8cedb96035 100644
--- a/drivers/net/wireless/ath/ath12k/mac.c
+++ b/drivers/net/wireless/ath/ath12k/mac.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: BSD-3-Clause-Clear
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <net/mac80211.h>
@@ -11251,8 +11251,6 @@ int ath12k_mac_register(struct ath12k_hw_group *ag)
goto err;
}
- set_bit(ATH12K_FLAG_REGISTERED, &ab->dev_flags);
-
return 0;
err:
@@ -11273,8 +11271,6 @@ void ath12k_mac_unregister(struct ath12k_hw_group *ag)
struct ath12k_hw *ah;
int i;
- clear_bit(ATH12K_FLAG_REGISTERED, &ab->dev_flags);
-
for (i = ath12k_get_num_hw(ab) - 1; i >= 0; i--) {
ah = ath12k_ab_to_ah(ab, i);
if (!ah)
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 07/10] wifi: ath12k: fix ATH12K_FLAG_REGISTERED flag handling
2025-01-09 4:25 ` [PATCH 07/10] wifi: ath12k: fix ATH12K_FLAG_REGISTERED flag handling Aditya Kumar Singh
@ 2025-01-13 19:23 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:23 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> Commit a5686ae820fa ("wifi: ath12k: move ATH12K_FLAG_REGISTERED handling to
> ath12k_mac_register()") relocated the setting of the ATH12K_FLAG_REGISTERED
> flag to the ath12k_mac_register() function. However, this function only
> accesses the first device (ab) via ag->ab[0], resulting in the flag being
> set only for the first device in the group. Similarly,
> ath12k_mac_unregister() only unsets the flag for the first device. The flag
> should actually be set for all devices in the group to avoid issues during
> recovery.
>
> Hence, move setting and clearing of this flag in the function
> ath12k_core_hw_group_start() and ath12k_core_hw_group_stop() respectively.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Fixes: a5686ae820fa ("wifi: ath12k: move ATH12K_FLAG_REGISTERED handling to ath12k_mac_register()")
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 08/10] wifi: ath12k: handle ath12k_core_restart() with hardware grouping
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (6 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 07/10] wifi: ath12k: fix ATH12K_FLAG_REGISTERED flag handling Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:23 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 09/10] wifi: ath12k: handle ath12k_core_reset() " Aditya Kumar Singh
2025-01-09 4:25 ` [PATCH 10/10] wifi: ath12k: reset MLO global memory during recovery Aditya Kumar Singh
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
Currently, when ath12k_core_restart() is called and the ab->is_reset flag
is set, it invokes ieee80211_restart_hw() for all hardware in the same
group. However, other hardware might still be in the recovery process,
making this call inappropriate with grouping into picture.
To address this, add a condition to check if the group is ready. If the
group is not ready, do not call ieee80211_restart_hw().
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index b67ef79e62b3fbb5667cb627cf565998a35f3c49..0a9e35695f760799273eeba32b889375232eedc0 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1394,12 +1394,22 @@ static void ath12k_core_restart(struct work_struct *work)
ath12k_dbg(ab, ATH12K_DBG_BOOT, "reset success\n");
}
+ mutex_lock(&ab->ag->mutex);
+
+ if (!ath12k_core_hw_group_start_ready(ab->ag)) {
+ mutex_unlock(&ab->ag->mutex);
+ goto exit_restart;
+ }
+
for (i = 0; i < ath12k_get_num_hw(ab); i++) {
ah = ath12k_ab_to_ah(ab, i);
ieee80211_restart_hw(ah->hw);
}
+
+ mutex_unlock(&ab->ag->mutex);
}
+exit_restart:
complete(&ab->restart_completed);
}
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 08/10] wifi: ath12k: handle ath12k_core_restart() with hardware grouping
2025-01-09 4:25 ` [PATCH 08/10] wifi: ath12k: handle ath12k_core_restart() with hardware grouping Aditya Kumar Singh
@ 2025-01-13 19:23 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:23 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> Currently, when ath12k_core_restart() is called and the ab->is_reset flag
> is set, it invokes ieee80211_restart_hw() for all hardware in the same
> group. However, other hardware might still be in the recovery process,
> making this call inappropriate with grouping into picture.
>
> To address this, add a condition to check if the group is ready. If the
> group is not ready, do not call ieee80211_restart_hw().
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 09/10] wifi: ath12k: handle ath12k_core_reset() with hardware grouping
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (7 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 08/10] wifi: ath12k: handle ath12k_core_restart() with hardware grouping Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:24 ` Jeff Johnson
2025-01-09 4:25 ` [PATCH 10/10] wifi: ath12k: reset MLO global memory during recovery Aditya Kumar Singh
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
Currently, in ath12k_core_reset(), the device is powered up immediately
after a power down. However, with hardware grouping, when one device
asserts, all partner devices also asserts. If there is a delay in
processing these asserts, by the time this device powers up, other devices
might still be asserting, leading to an overall recovery failure.
To prevent this issue, ensure all asserts for a group are processed before
initiating the power-up sequence.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 30 +++++++++++++++++++++++++++---
1 file changed, 27 insertions(+), 3 deletions(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 0a9e35695f760799273eeba32b889375232eedc0..ce261151f5b887656f9582e3337a2d5e5236bbc1 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1416,7 +1416,8 @@ static void ath12k_core_restart(struct work_struct *work)
static void ath12k_core_reset(struct work_struct *work)
{
struct ath12k_base *ab = container_of(work, struct ath12k_base, reset_work);
- int reset_count, fail_cont_count;
+ struct ath12k_hw_group *ag = ab->ag;
+ int reset_count, fail_cont_count, i;
long time_left;
if (!(test_bit(ATH12K_FLAG_QMI_FW_READY_COMPLETE, &ab->dev_flags))) {
@@ -1475,9 +1476,32 @@ static void ath12k_core_reset(struct work_struct *work)
ath12k_hif_ce_irq_disable(ab);
ath12k_hif_power_down(ab, false);
- ath12k_hif_power_up(ab);
- ath12k_dbg(ab, ATH12K_DBG_BOOT, "reset started\n");
+ /* prepare for power up */
+ ab->qmi.num_radios = U8_MAX;
+ ab->single_chip_mlo_supp = false;
+
+ mutex_lock(&ag->mutex);
+ ath12k_core_to_group_ref_put(ab);
+
+ if (ag->num_started > 0) {
+ ath12k_dbg(ab, ATH12K_DBG_BOOT,
+ "waiting for %d partner device(s) to reset\n",
+ ag->num_started);
+ mutex_unlock(&ag->mutex);
+ return;
+ }
+
+ for (i = 0; i < ag->num_devices; i++) {
+ ab = ag->ab[i];
+ if (!ab)
+ continue;
+
+ ath12k_hif_power_up(ab);
+ ath12k_dbg(ab, ATH12K_DBG_BOOT, "reset started\n");
+ }
+
+ mutex_unlock(&ag->mutex);
}
int ath12k_core_pre_init(struct ath12k_base *ab)
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 09/10] wifi: ath12k: handle ath12k_core_reset() with hardware grouping
2025-01-09 4:25 ` [PATCH 09/10] wifi: ath12k: handle ath12k_core_reset() " Aditya Kumar Singh
@ 2025-01-13 19:24 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:24 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> Currently, in ath12k_core_reset(), the device is powered up immediately
> after a power down. However, with hardware grouping, when one device
> asserts, all partner devices also asserts. If there is a delay in
> processing these asserts, by the time this device powers up, other devices
> might still be asserting, leading to an overall recovery failure.
>
> To prevent this issue, ensure all asserts for a group are processed before
> initiating the power-up sequence.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 10/10] wifi: ath12k: reset MLO global memory during recovery
2025-01-09 4:25 [PATCH 00/10] wifi: ath12k: fixes for rmmod and recovery issues with hardware grouping Aditya Kumar Singh
` (8 preceding siblings ...)
2025-01-09 4:25 ` [PATCH 09/10] wifi: ath12k: handle ath12k_core_reset() " Aditya Kumar Singh
@ 2025-01-09 4:25 ` Aditya Kumar Singh
2025-01-13 19:25 ` Jeff Johnson
9 siblings, 1 reply; 23+ messages in thread
From: Aditya Kumar Singh @ 2025-01-09 4:25 UTC (permalink / raw)
To: Kalle Valo, Jeff Johnson, Karthikeyan Periyasamy, Harshitha Prem
Cc: Jeff Johnson, Kalle Valo, linux-wireless, ath12k, linux-kernel,
Aditya Kumar Singh
When operating with multiple devices grouped together, the firmware stores
data related to the state machine of each partner device in the MLO global
memory region. If the firmware crashes, it updates the state to 'crashed'.
During recovery, this memory is shared with the firmware again, and upon
detecting the 'crashed' state, it reasserts. This leads to a loop of
firmware asserts and it never recovers.
Hence to fix this issue, once all devices in the group have been asserted
and powered down, reset the MLO global memory region.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
---
drivers/net/wireless/ath/ath12k/core.c | 3 +++
drivers/net/wireless/ath/ath12k/qmi.c | 24 +++++++++++++++++++++++-
drivers/net/wireless/ath/ath12k/qmi.h | 4 +++-
3 files changed, 29 insertions(+), 2 deletions(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index ce261151f5b887656f9582e3337a2d5e5236bbc1..0475c80befc82f59b2384d68b382b38d06a83bfe 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1492,6 +1492,9 @@ static void ath12k_core_reset(struct work_struct *work)
return;
}
+ /* Prepare MLO global memory region for power up */
+ ath12k_qmi_reset_mlo_mem(ag);
+
for (i = 0; i < ag->num_devices; i++) {
ab = ag->ab[i];
if (!ab)
diff --git a/drivers/net/wireless/ath/ath12k/qmi.c b/drivers/net/wireless/ath/ath12k/qmi.c
index 5c3563383fabba779b0afd885802637fd5c53656..bc14fa106b27015b35641caac51fcb4b808e094d 100644
--- a/drivers/net/wireless/ath/ath12k/qmi.c
+++ b/drivers/net/wireless/ath/ath12k/qmi.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: BSD-3-Clause-Clear
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <linux/elf.h>
@@ -2440,6 +2440,28 @@ int ath12k_qmi_respond_fw_mem_request(struct ath12k_base *ab)
return ret;
}
+void ath12k_qmi_reset_mlo_mem(struct ath12k_hw_group *ag)
+{
+ struct target_mem_chunk *mlo_chunk;
+ int i;
+
+ lockdep_assert_held(&ag->mutex);
+
+ if (!ag->mlo_mem.init_done || ag->num_started)
+ return;
+
+ for (i = 0; i < ARRAY_SIZE(ag->mlo_mem.chunk); i++) {
+ mlo_chunk = &ag->mlo_mem.chunk[i];
+
+ if (mlo_chunk->v.addr)
+ /* TODO: Mode 0 recovery is the default mode hence resetting the
+ * whole memory region for now. Once Mode 1 support is added, this
+ * needs to be handled properly
+ */
+ memset(mlo_chunk->v.addr, 0, mlo_chunk->size);
+ }
+}
+
static void ath12k_qmi_free_mlo_mem_chunk(struct ath12k_base *ab,
struct target_mem_chunk *chunk,
int idx)
diff --git a/drivers/net/wireless/ath/ath12k/qmi.h b/drivers/net/wireless/ath/ath12k/qmi.h
index 45d7c3fcafdd7a0afa3d193cf612d255ca5a5e2e..1406ec372592b076dd0a5caef32044db238e352e 100644
--- a/drivers/net/wireless/ath/ath12k/qmi.h
+++ b/drivers/net/wireless/ath/ath12k/qmi.h
@@ -1,7 +1,7 @@
/* SPDX-License-Identifier: BSD-3-Clause-Clear */
/*
* Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2025 Qualcomm Innovation Center, Inc. All rights reserved.
*/
#ifndef ATH12K_QMI_H
@@ -41,6 +41,7 @@
#define ATH12K_BOARD_ID_DEFAULT 0xFF
struct ath12k_base;
+struct ath12k_hw_group;
enum ath12k_qmi_file_type {
ATH12K_QMI_FILE_TYPE_BDF_GOLDEN = 0,
@@ -621,5 +622,6 @@ void ath12k_qmi_deinit_service(struct ath12k_base *ab);
int ath12k_qmi_init_service(struct ath12k_base *ab);
void ath12k_qmi_free_resource(struct ath12k_base *ab);
void ath12k_qmi_trigger_host_cap(struct ath12k_base *ab);
+void ath12k_qmi_reset_mlo_mem(struct ath12k_hw_group *ag);
#endif
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH 10/10] wifi: ath12k: reset MLO global memory during recovery
2025-01-09 4:25 ` [PATCH 10/10] wifi: ath12k: reset MLO global memory during recovery Aditya Kumar Singh
@ 2025-01-13 19:25 ` Jeff Johnson
0 siblings, 0 replies; 23+ messages in thread
From: Jeff Johnson @ 2025-01-13 19:25 UTC (permalink / raw)
To: Aditya Kumar Singh, Kalle Valo, Jeff Johnson,
Karthikeyan Periyasamy, Harshitha Prem
Cc: Kalle Valo, linux-wireless, ath12k, linux-kernel
On 1/8/2025 8:25 PM, Aditya Kumar Singh wrote:
> When operating with multiple devices grouped together, the firmware stores
> data related to the state machine of each partner device in the MLO global
> memory region. If the firmware crashes, it updates the state to 'crashed'.
> During recovery, this memory is shared with the firmware again, and upon
> detecting the 'crashed' state, it reasserts. This leads to a loop of
> firmware asserts and it never recovers.
>
> Hence to fix this issue, once all devices in the group have been asserted
> and powered down, reset the MLO global memory region.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 23+ messages in thread