* [PATCH v7] net: airoha: npu: use aligned streaming DMA for mailbox messages
@ 2026-08-14 11:00 Daniel Pawlik
2026-08-17 1:56 ` Lorenzo Bianconi
0 siblings, 1 reply; 2+ messages in thread
From: Daniel Pawlik @ 2026-08-14 11:00 UTC (permalink / raw)
To: netdev; +Cc: lorenzo, win847, linux-arm-kernel, linux-mediatek, Daniel Pawlik
On EN7581 + MT7996 (Gemtek W1700K), mapping small caller buffers with
DMA_BIDIRECTIONAL regresses NPU version probe: the mailbox completes
successfully but WLAN_FUNC_GET_WAIT_NPU_VERSION reads as 0.0 instead of
0.1111.
Bounce mailbox traffic through a per-core cacheline-aligned buffer
allocated at probe. Map ALIGN(len, SMP_CACHE_BYTES) with
DMA_BIDIRECTIONAL while programming the original payload length into the
mailbox length register. Copy the reply only after dma_unmap_single() so
the CPU sees the NPU-written payload on non-coherent DMA.
Introduce __airoha_npu_send_msg() with an optional reply pointer so GET
callers can copy the trailing response payload once. Keep
airoha_npu_send_msg() as a wrapper that copies the full response back
into the caller buffer, matching the old bidirectional mapping.
Tested on Quantum Fiber / Gemtek W1700K (EN7581 + MT7996), kernel
6.18.44, including two cold reboots and sustained WiFi use.
Fixes: 6f884eb87a79 ("net: airoha: Fix DMA direction for NPU mailbox buffer")
Link: https://patchwork.kernel.org/project/linux-mediatek/patch/20260805070851.2885888-1-pawlik.dan@gmail.com/
Assisted-by: Cursor:composer-2
Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
---
drivers/net/ethernet/airoha/airoha_npu.c | 62 ++++++++++++++++++-----
include/linux/soc/airoha/airoha_offload.h | 1 +
2 files changed, 49 insertions(+), 14 deletions(-)
diff --git a/drivers/net/ethernet/airoha/airoha_npu.c b/drivers/net/ethernet/airoha/airoha_npu.c
index b679bed952de..2ba78bf0e97b 100644
--- a/drivers/net/ethernet/airoha/airoha_npu.c
+++ b/drivers/net/ethernet/airoha/airoha_npu.c
@@ -5,6 +5,7 @@
*/
#include <linux/devcoredump.h>
+#include <linux/dma-mapping.h>
#include <linux/firmware.h>
#include <linux/platform_device.h>
#include <linux/of_net.h>
@@ -23,6 +24,8 @@
#define NPU_EN7581_FIRMWARE_RV32_MAX_SIZE 0x200000
#define NPU_EN7581_FIRMWARE_DATA_MAX_SIZE 0x10000
#define NPU_DUMP_SIZE 512
+/* Maximum mailbox DMA payload (PPE ~28 bytes, WLAN TLV up to 24). */
+#define AIROHA_NPU_MBOX_SIZE 256
#define REG_NPU_LOCAL_SRAM 0x0
@@ -160,23 +163,33 @@ struct wlan_mbox_data {
DECLARE_FLEX_ARRAY(u8, d);
};
-static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
- void *p, int size)
+static int __airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
+ const void *data, int len, void *reply,
+ u16 reply_len)
{
- u16 core = 0; /* FIXME */
- u32 val, offset = core << 4;
+ struct airoha_npu_core *core = &npu->cores[0]; /* FIXME: core */
dma_addr_t dma_addr;
+ unsigned int map_len = ALIGN(len, SMP_CACHE_BYTES);
+ u32 val, offset = 0;
int ret;
- dma_addr = dma_map_single(npu->dev, p, size, DMA_BIDIRECTIONAL);
+ if (len <= 0 || len > AIROHA_NPU_MBOX_SIZE)
+ return -EINVAL;
+
+ if (reply && reply_len > len)
+ return -EINVAL;
+
+ spin_lock_bh(&core->lock);
+
+ memcpy(core->buf, data, len);
+
+ dma_addr = dma_map_single(npu->dev, core->buf, map_len, DMA_BIDIRECTIONAL);
ret = dma_mapping_error(npu->dev, dma_addr);
if (ret)
- return ret;
-
- spin_lock_bh(&npu->cores[core].lock);
+ goto unlock;
regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(0) + offset, dma_addr);
- regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(1) + offset, size);
+ regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(1) + offset, len);
regmap_read(npu->regmap, REG_CR_MBQ0_CTRL(2) + offset, &val);
regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(2) + offset, val + 1);
val = FIELD_PREP(MBOX_MSG_FUNC_ID, func_id) | MBOX_MSG_WAIT_RSP;
@@ -189,13 +202,23 @@ static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
if (!ret && FIELD_GET(MBOX_MSG_STATUS, val) != NPU_MBOX_SUCCESS)
ret = -EINVAL;
- spin_unlock_bh(&npu->cores[core].lock);
+ dma_unmap_single(npu->dev, dma_addr, map_len, DMA_BIDIRECTIONAL);
- dma_unmap_single(npu->dev, dma_addr, size, DMA_BIDIRECTIONAL);
+ /* Copy the trailing reply_len bytes of the response. */
+ if (!ret && reply)
+ memcpy(reply, core->buf + len - reply_len, reply_len);
+unlock:
+ spin_unlock_bh(&core->lock);
return ret;
}
+static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
+ void *data, int len)
+{
+ return __airoha_npu_send_msg(npu, func_id, data, len, data, len);
+}
+
static int airoha_npu_load_firmware(struct device *dev, void __iomem *addr,
const char *fw_name, int fw_max_size)
{
@@ -497,9 +520,8 @@ static int airoha_npu_wlan_msg_get(struct airoha_npu *npu, int ifindex,
wlan_data->func_type = NPU_OP_GET;
wlan_data->func_id = func_id;
- err = airoha_npu_send_msg(npu, NPU_FUNC_WIFI, wlan_data, len);
- if (!err)
- memcpy(data, wlan_data->d, data_len);
+ err = __airoha_npu_send_msg(npu, NPU_FUNC_WIFI, wlan_data, len,
+ data, data_len);
kfree(wlan_data);
return err;
@@ -770,6 +792,18 @@ static int airoha_npu_probe(struct platform_device *pdev)
if (err)
return err;
+ for (i = 0; i < ARRAY_SIZE(npu->cores); i++) {
+ struct airoha_npu_core *core = &npu->cores[i];
+ void *raw;
+
+ raw = devm_kmalloc(dev, AIROHA_NPU_MBOX_SIZE + SMP_CACHE_BYTES,
+ GFP_KERNEL);
+ if (!raw)
+ return -ENOMEM;
+
+ core->buf = PTR_ALIGN(raw, SMP_CACHE_BYTES);
+ }
+
err = airoha_npu_run_firmware(dev, base, &res);
if (err)
return dev_err_probe(dev, err, "failed to run npu firmware\n");
diff --git a/include/linux/soc/airoha/airoha_offload.h b/include/linux/soc/airoha/airoha_offload.h
index 7589fccfeef6..f635766e3dba 100644
--- a/include/linux/soc/airoha/airoha_offload.h
+++ b/include/linux/soc/airoha/airoha_offload.h
@@ -173,6 +173,7 @@ struct airoha_npu {
/* protect concurrent npu memory accesses */
spinlock_t lock;
struct work_struct wdt_work;
+ void *buf; /* mailbox DMA bounce buffer */
} cores[NPU_NUM_CORES];
int irqs[NPU_NUM_IRQ];
--
2.55.0
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH v7] net: airoha: npu: use aligned streaming DMA for mailbox messages
2026-08-14 11:00 [PATCH v7] net: airoha: npu: use aligned streaming DMA for mailbox messages Daniel Pawlik
@ 2026-08-17 1:56 ` Lorenzo Bianconi
0 siblings, 0 replies; 2+ messages in thread
From: Lorenzo Bianconi @ 2026-08-17 1:56 UTC (permalink / raw)
To: Daniel Pawlik; +Cc: netdev, win847, linux-arm-kernel, linux-mediatek
[-- Attachment #1: Type: text/plain, Size: 6855 bytes --]
> On EN7581 + MT7996 (Gemtek W1700K), mapping small caller buffers with
> DMA_BIDIRECTIONAL regresses NPU version probe: the mailbox completes
> successfully but WLAN_FUNC_GET_WAIT_NPU_VERSION reads as 0.0 instead of
> 0.1111.
IIUC you are suggesting the issue is if we allocate a small buffer it could fit
in the same cacheline of a neighbour object so if the DMA reads/writes the same
cacheline, it could invalidate data, right? (I am not sure if this device is
DMA-coherent or not).
If so, in the airoha_npu_send_msg() callers, I guess it is enough to allocate
something like:
kzalloc(max(len, SMP_CACHE_BYTES))
and avoid the extra copy in __airoha_npu_send_msg(). What do you think?
IIRC kzalloc() aligns to SMP_CACHE_BYTES is len >= SMP_CACHE_BYTES.
Regards,
Lorenzo
>
> Bounce mailbox traffic through a per-core cacheline-aligned buffer
> allocated at probe. Map ALIGN(len, SMP_CACHE_BYTES) with
> DMA_BIDIRECTIONAL while programming the original payload length into the
> mailbox length register. Copy the reply only after dma_unmap_single() so
> the CPU sees the NPU-written payload on non-coherent DMA.
>
> Introduce __airoha_npu_send_msg() with an optional reply pointer so GET
> callers can copy the trailing response payload once. Keep
> airoha_npu_send_msg() as a wrapper that copies the full response back
> into the caller buffer, matching the old bidirectional mapping.
>
> Tested on Quantum Fiber / Gemtek W1700K (EN7581 + MT7996), kernel
> 6.18.44, including two cold reboots and sustained WiFi use.
>
> Fixes: 6f884eb87a79 ("net: airoha: Fix DMA direction for NPU mailbox buffer")
> Link: https://patchwork.kernel.org/project/linux-mediatek/patch/20260805070851.2885888-1-pawlik.dan@gmail.com/
> Assisted-by: Cursor:composer-2
> Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
> ---
> drivers/net/ethernet/airoha/airoha_npu.c | 62 ++++++++++++++++++-----
> include/linux/soc/airoha/airoha_offload.h | 1 +
> 2 files changed, 49 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/net/ethernet/airoha/airoha_npu.c b/drivers/net/ethernet/airoha/airoha_npu.c
> index b679bed952de..2ba78bf0e97b 100644
> --- a/drivers/net/ethernet/airoha/airoha_npu.c
> +++ b/drivers/net/ethernet/airoha/airoha_npu.c
> @@ -5,6 +5,7 @@
> */
>
> #include <linux/devcoredump.h>
> +#include <linux/dma-mapping.h>
> #include <linux/firmware.h>
> #include <linux/platform_device.h>
> #include <linux/of_net.h>
> @@ -23,6 +24,8 @@
> #define NPU_EN7581_FIRMWARE_RV32_MAX_SIZE 0x200000
> #define NPU_EN7581_FIRMWARE_DATA_MAX_SIZE 0x10000
> #define NPU_DUMP_SIZE 512
> +/* Maximum mailbox DMA payload (PPE ~28 bytes, WLAN TLV up to 24). */
> +#define AIROHA_NPU_MBOX_SIZE 256
>
> #define REG_NPU_LOCAL_SRAM 0x0
>
> @@ -160,23 +163,33 @@ struct wlan_mbox_data {
> DECLARE_FLEX_ARRAY(u8, d);
> };
>
> -static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
> - void *p, int size)
> +static int __airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
> + const void *data, int len, void *reply,
> + u16 reply_len)
> {
> - u16 core = 0; /* FIXME */
> - u32 val, offset = core << 4;
> + struct airoha_npu_core *core = &npu->cores[0]; /* FIXME: core */
> dma_addr_t dma_addr;
> + unsigned int map_len = ALIGN(len, SMP_CACHE_BYTES);
> + u32 val, offset = 0;
> int ret;
>
> - dma_addr = dma_map_single(npu->dev, p, size, DMA_BIDIRECTIONAL);
> + if (len <= 0 || len > AIROHA_NPU_MBOX_SIZE)
> + return -EINVAL;
> +
> + if (reply && reply_len > len)
> + return -EINVAL;
> +
> + spin_lock_bh(&core->lock);
> +
> + memcpy(core->buf, data, len);
> +
> + dma_addr = dma_map_single(npu->dev, core->buf, map_len, DMA_BIDIRECTIONAL);
> ret = dma_mapping_error(npu->dev, dma_addr);
> if (ret)
> - return ret;
> -
> - spin_lock_bh(&npu->cores[core].lock);
> + goto unlock;
>
> regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(0) + offset, dma_addr);
> - regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(1) + offset, size);
> + regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(1) + offset, len);
> regmap_read(npu->regmap, REG_CR_MBQ0_CTRL(2) + offset, &val);
> regmap_write(npu->regmap, REG_CR_MBQ0_CTRL(2) + offset, val + 1);
> val = FIELD_PREP(MBOX_MSG_FUNC_ID, func_id) | MBOX_MSG_WAIT_RSP;
> @@ -189,13 +202,23 @@ static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
> if (!ret && FIELD_GET(MBOX_MSG_STATUS, val) != NPU_MBOX_SUCCESS)
> ret = -EINVAL;
>
> - spin_unlock_bh(&npu->cores[core].lock);
> + dma_unmap_single(npu->dev, dma_addr, map_len, DMA_BIDIRECTIONAL);
>
> - dma_unmap_single(npu->dev, dma_addr, size, DMA_BIDIRECTIONAL);
> + /* Copy the trailing reply_len bytes of the response. */
> + if (!ret && reply)
> + memcpy(reply, core->buf + len - reply_len, reply_len);
> +unlock:
> + spin_unlock_bh(&core->lock);
>
> return ret;
> }
>
> +static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
> + void *data, int len)
> +{
> + return __airoha_npu_send_msg(npu, func_id, data, len, data, len);
> +}
> +
> static int airoha_npu_load_firmware(struct device *dev, void __iomem *addr,
> const char *fw_name, int fw_max_size)
> {
> @@ -497,9 +520,8 @@ static int airoha_npu_wlan_msg_get(struct airoha_npu *npu, int ifindex,
> wlan_data->func_type = NPU_OP_GET;
> wlan_data->func_id = func_id;
>
> - err = airoha_npu_send_msg(npu, NPU_FUNC_WIFI, wlan_data, len);
> - if (!err)
> - memcpy(data, wlan_data->d, data_len);
> + err = __airoha_npu_send_msg(npu, NPU_FUNC_WIFI, wlan_data, len,
> + data, data_len);
> kfree(wlan_data);
>
> return err;
> @@ -770,6 +792,18 @@ static int airoha_npu_probe(struct platform_device *pdev)
> if (err)
> return err;
>
> + for (i = 0; i < ARRAY_SIZE(npu->cores); i++) {
> + struct airoha_npu_core *core = &npu->cores[i];
> + void *raw;
> +
> + raw = devm_kmalloc(dev, AIROHA_NPU_MBOX_SIZE + SMP_CACHE_BYTES,
> + GFP_KERNEL);
> + if (!raw)
> + return -ENOMEM;
> +
> + core->buf = PTR_ALIGN(raw, SMP_CACHE_BYTES);
> + }
> +
> err = airoha_npu_run_firmware(dev, base, &res);
> if (err)
> return dev_err_probe(dev, err, "failed to run npu firmware\n");
> diff --git a/include/linux/soc/airoha/airoha_offload.h b/include/linux/soc/airoha/airoha_offload.h
> index 7589fccfeef6..f635766e3dba 100644
> --- a/include/linux/soc/airoha/airoha_offload.h
> +++ b/include/linux/soc/airoha/airoha_offload.h
> @@ -173,6 +173,7 @@ struct airoha_npu {
> /* protect concurrent npu memory accesses */
> spinlock_t lock;
> struct work_struct wdt_work;
> + void *buf; /* mailbox DMA bounce buffer */
> } cores[NPU_NUM_CORES];
>
> int irqs[NPU_NUM_IRQ];
> --
> 2.55.0
>
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2026-08-17 1:56 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-08-14 11:00 [PATCH v7] net: airoha: npu: use aligned streaming DMA for mailbox messages Daniel Pawlik
2026-08-17 1:56 ` Lorenzo Bianconi
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox