> 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 > --- > 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 > +#include > #include > #include > #include > @@ -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 >