| Message ID | 20250307040848.3822788-3-quic_miaoqing@quicinc.com (mailing list archive) |
|---|---|
| State | New |
| Delegated to: | Jeff Johnson |
| Headers | show |
| Series | wifi: ath11k: fix HTC rx insufficient length | expand |
Hi Miaoqing, and thanks a lot for looking into this. On Fri, Mar 07, 2025 at 12:08:48PM +0800, Miaoqing Pan wrote: > A relatively unusual race condition occurs between host software > and hardware, where the host sees the updated destination ring head > pointer before the hardware updates the corresponding descriptor. > When this situation occurs, the length of the descriptor returns 0. Can you explain how this race comes about? I worry that you may just be papering over a driver or firmware bug here so it would be good to understand how the host can see the updated pointer before the descriptor. Also do you have any suggestions for how to reproduce this more easily? Some users of the X13s (aarch64) hit this very frequently, while I've only seen it a few times. At least one user hit this consistently when roaming, and another mentioned seeing this while driving his car with the laptop on (and presumably seeing a lot of APs go by). > The current error handling method is to increment descriptor tail > pointer by 1, but 'sw_index' is not updated, causing descriptor and > skb to not correspond one-to-one, resulting in the following error: > > ath11k_pci 0006:01:00.0: HTC Rx: insufficient length, got 1488, expected 1492 > ath11k_pci 0006:01:00.0: HTC Rx: insufficient length, got 1460, expected 1484 > > To address this problem, temporarily skip processing the current > descriptor and handle it again next time. However, to prevent this > descriptor from continuously returning 0, use skb cb to set a flag. > If the length returns 0 again, this descriptor will be discarded. > > Tested-on: QCA6698AQ hw2.1 PCI WLAN.HSP.1.1-04546-QCAHSPSWPL_V1_V2_SILICONZ_IOE-1 Please also include: Reported-by: Johan Hovold <johan+linaro@kernel.org> > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218623 > Signed-off-by: Miaoqing Pan <quic_miaoqing@quicinc.com> Johan
On 3/7/2025 4:51 PM, Johan Hovold wrote: > Hi Miaoqing, > > and thanks a lot for looking into this. > > On Fri, Mar 07, 2025 at 12:08:48PM +0800, Miaoqing Pan wrote: >> A relatively unusual race condition occurs between host software >> and hardware, where the host sees the updated destination ring head >> pointer before the hardware updates the corresponding descriptor. >> When this situation occurs, the length of the descriptor returns 0. > > Can you explain how this race comes about? > > I worry that you may just be papering over a driver or firmware bug here > so it would be good to understand how the host can see the updated > pointer before the descriptor. > The entire process is handled by hardware, and host software just to read the status, so it should be a limitation of the hardware. > Also do you have any suggestions for how to reproduce this more easily? > Some users of the X13s (aarch64) hit this very frequently, while I've > only seen it a few times. At least one user hit this consistently when > roaming, and another mentioned seeing this while driving his car with > the laptop on (and presumably seeing a lot of APs go by). > We reproduced with below setup on IoT device, usually takes a few hours to several days to reproduce it. - Run TCP downlink traffic (iperf) - Run bluetooth bredr scan (bluetoothctl/power on/scan bredr) >> The current error handling method is to increment descriptor tail >> pointer by 1, but 'sw_index' is not updated, causing descriptor and >> skb to not correspond one-to-one, resulting in the following error: >> >> ath11k_pci 0006:01:00.0: HTC Rx: insufficient length, got 1488, expected 1492 >> ath11k_pci 0006:01:00.0: HTC Rx: insufficient length, got 1460, expected 1484 >> >> To address this problem, temporarily skip processing the current >> descriptor and handle it again next time. However, to prevent this >> descriptor from continuously returning 0, use skb cb to set a flag. >> If the length returns 0 again, this descriptor will be discarded. >> >> Tested-on: QCA6698AQ hw2.1 PCI WLAN.HSP.1.1-04546-QCAHSPSWPL_V1_V2_SILICONZ_IOE-1 > > Please also include: > > Reported-by: Johan Hovold <johan+linaro@kernel.org> > >> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218623 >> Signed-off-by: Miaoqing Pan <quic_miaoqing@quicinc.com> > > Johan Thanks for reporting, will update in next version.
diff --git a/drivers/net/wireless/ath/ath11k/ce.c b/drivers/net/wireless/ath/ath11k/ce.c index e66e86bdec20..2573f8c7a994 100644 --- a/drivers/net/wireless/ath/ath11k/ce.c +++ b/drivers/net/wireless/ath/ath11k/ce.c @@ -1,7 +1,7 @@ // SPDX-License-Identifier: BSD-3-Clause-Clear /* * Copyright (c) 2018-2019 The Linux Foundation. All rights reserved. - * Copyright (c) 2021-2023 Qualcomm Innovation Center, Inc. All rights reserved. + * Copyright (c) 2021-2023, 2025 Qualcomm Innovation Center, Inc. All rights reserved. */ #include "dp_rx.h" @@ -387,18 +387,36 @@ static int ath11k_ce_completed_recv_next(struct ath11k_ce_pipe *pipe, ath11k_hal_srng_access_begin(ab, srng); - desc = ath11k_hal_srng_dst_get_next_entry(ab, srng); + desc = ath11k_hal_srng_dst_peek(ab, srng); if (!desc) { ret = -EIO; goto err; } *nbytes = ath11k_hal_ce_dst_status_get_length(desc); - if (*nbytes == 0) { - ret = -EIO; - goto err; + if (unlikely(*nbytes == 0)) { + struct ath11k_skb_rxcb *rxcb = + ATH11K_SKB_RXCB(pipe->dest_ring->skb[sw_index]); + + /* A relatively unusual race condition occurs between host + * software and hardware, where the host sees the updated + * destination ring head pointer before the hardware updates + * the corresponding descriptor. + * + * Temporarily skip processing the current descriptor and handle + * it again next time. However, to prevent this descriptor from + * continuously returning 0, set 'is_desc_len0' flag. If the + * length returns 0 again, this descriptor will be discarded. + */ + if (!rxcb->is_desc_len0) { + rxcb->is_desc_len0 = true; + ret = -EIO; + goto err; + } } + ath11k_hal_srng_dst_next(ab, srng); + *skb = pipe->dest_ring->skb[sw_index]; pipe->dest_ring->skb[sw_index] = NULL; @@ -430,8 +448,8 @@ static void ath11k_ce_recv_process_cb(struct ath11k_ce_pipe *pipe) dma_unmap_single(ab->dev, ATH11K_SKB_RXCB(skb)->paddr, max_nbytes, DMA_FROM_DEVICE); - if (unlikely(max_nbytes < nbytes)) { - ath11k_warn(ab, "rxed more than expected (nbytes %d, max %d)", + if (unlikely(max_nbytes < nbytes || !nbytes)) { + ath11k_warn(ab, "rxed invalid length (nbytes %d, max %d)", nbytes, max_nbytes); dev_kfree_skb_any(skb); continue; diff --git a/drivers/net/wireless/ath/ath11k/core.h b/drivers/net/wireless/ath/ath11k/core.h index 1a3d0de4afde..c8614c5c6493 100644 --- a/drivers/net/wireless/ath/ath11k/core.h +++ b/drivers/net/wireless/ath/ath11k/core.h @@ -128,6 +128,7 @@ struct ath11k_skb_rxcb { bool is_continuation; bool is_mcbc; bool is_eapol; + bool is_desc_len0; struct hal_rx_desc *rx_desc; u8 err_rel_src; u8 err_code;
A relatively unusual race condition occurs between host software and hardware, where the host sees the updated destination ring head pointer before the hardware updates the corresponding descriptor. When this situation occurs, the length of the descriptor returns 0. The current error handling method is to increment descriptor tail pointer by 1, but 'sw_index' is not updated, causing descriptor and skb to not correspond one-to-one, resulting in the following error: ath11k_pci 0006:01:00.0: HTC Rx: insufficient length, got 1488, expected 1492 ath11k_pci 0006:01:00.0: HTC Rx: insufficient length, got 1460, expected 1484 To address this problem, temporarily skip processing the current descriptor and handle it again next time. However, to prevent this descriptor from continuously returning 0, use skb cb to set a flag. If the length returns 0 again, this descriptor will be discarded. Tested-on: QCA6698AQ hw2.1 PCI WLAN.HSP.1.1-04546-QCAHSPSWPL_V1_V2_SILICONZ_IOE-1 Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218623 Signed-off-by: Miaoqing Pan <quic_miaoqing@quicinc.com> --- drivers/net/wireless/ath/ath11k/ce.c | 32 ++++++++++++++++++++------ drivers/net/wireless/ath/ath11k/core.h | 1 + 2 files changed, 26 insertions(+), 7 deletions(-)