* [PATCH] wifi: ath12k: fix HE/EHT capability handling on big endian
@ 2026-03-17 10:59 Alexander Wilhelm
2026-03-18 0:17 ` Jeff Johnson
0 siblings, 1 reply; 3+ messages in thread
From: Alexander Wilhelm @ 2026-03-17 10:59 UTC (permalink / raw)
To: Jeff Johnson; +Cc: linux-wireless, ath12k, linux-kernel
Currently the driver uses u32 data types for the HE/EHT capabilities in
CPU‑native order. However, the ieee80211.h header defines these fields as
u8 arrays. This causes the ieee80211 registration failure on big‑endian
platforms, as shown in the following log:
ath12k_pci 0001:01:00.0: BAR 0: assigned [mem 0xc00000000-0xc001fffff 64bit]
ath12k_pci 0001:01:00.0: MSI vectors: 1
ath12k_pci 0001:01:00.0: Hardware name: qcn9274 hw2.0
ath12k_pci 0001:01:00.0: qmi dma allocation failed (29360128 B type 1), will try later with small size
ath12k_pci 0001:01:00.0: memory type 10 not supported
ath12k_pci 0001:01:00.0: chip_id 0x0 chip_family 0xb board_id 0x1005 soc_id 0x401a2200
ath12k_pci 0001:01:00.0: fw_version 0x111300d6 fw_build_timestamp 2024-08-06 08:43 fw_build_id QC_IMAGE_VERSION_STRING=WLAN.WBE.1.1.1-00214-QCAHKSWPL_SILICONZ-1
ath12k_pci 0001:01:00.0: leaving PCI ASPM disabled to avoid MHI M2 problems
ath12k_pci 0001:01:00.0: Invalid module id 2
ath12k_pci 0001:01:00.0: failed to parse tlv -22
ath12k_pci 0001:01:00.0: ieee80211 registration failed: -22
ath12k_pci 0001:01:00.0: failed register the radio with mac80211: -22
ath12k_pci 0001:01:00.0: failed to create pdev core: -22
ath12k_pci 0001:01:00.0: qmi failed set mode request, mode: 4, err = -110
ath12k_pci 0001:01:00.0: qmi failed to send wlan mode off
Use the __le32 data type for the HE/EHT capabilities instead and avoid
swapping, so that both platform endiannesses are supported.
Signed-off-by: Alexander Wilhelm <alexander.wilhelm@westermo.com>
---
drivers/net/wireless/ath/ath12k/core.h | 8 ++---
drivers/net/wireless/ath/ath12k/wmi.c | 58 ++++++++++++++++------------------
drivers/net/wireless/ath/ath12k/wmi.h | 4 +--
3 files changed, 34 insertions(+), 36 deletions(-)
diff --git a/drivers/net/wireless/ath/ath12k/core.h b/drivers/net/wireless/ath/ath12k/core.h
index 59c193b24764..8481015dcda6 100644
--- a/drivers/net/wireless/ath/ath12k/core.h
+++ b/drivers/net/wireless/ath/ath12k/core.h
@@ -788,13 +788,13 @@ struct ath12k_band_cap {
u32 phy_id;
u32 max_bw_supported;
u32 ht_cap_info;
- u32 he_cap_info[2];
+ __le32 he_cap_info[2];
u32 he_mcs;
- u32 he_cap_phy_info[PSOC_HOST_MAX_PHY_SIZE];
+ __le32 he_cap_phy_info[PSOC_HOST_MAX_PHY_SIZE];
struct ath12k_wmi_ppe_threshold_arg he_ppet;
u16 he_6ghz_capa;
- u32 eht_cap_mac_info[WMI_MAX_EHTCAP_MAC_SIZE];
- u32 eht_cap_phy_info[WMI_MAX_EHTCAP_PHY_SIZE];
+ __le32 eht_cap_mac_info[WMI_MAX_EHTCAP_MAC_SIZE];
+ __le32 eht_cap_phy_info[WMI_MAX_EHTCAP_PHY_SIZE];
u32 eht_mcs_20_only;
u32 eht_mcs_80;
u32 eht_mcs_160;
diff --git a/drivers/net/wireless/ath/ath12k/wmi.c b/drivers/net/wireless/ath/ath12k/wmi.c
index 65a05a9520ff..f5cd1eb27685 100644
--- a/drivers/net/wireless/ath/ath12k/wmi.c
+++ b/drivers/net/wireless/ath/ath12k/wmi.c
@@ -487,12 +487,11 @@ ath12k_pull_mac_phy_cap_svc_ready_ext(struct ath12k_wmi_pdev *wmi_handle,
cap_band->phy_id = le32_to_cpu(mac_caps->phy_id);
cap_band->max_bw_supported = le32_to_cpu(mac_caps->max_bw_supported_2g);
cap_band->ht_cap_info = le32_to_cpu(mac_caps->ht_cap_info_2g);
- cap_band->he_cap_info[0] = le32_to_cpu(mac_caps->he_cap_info_2g);
- cap_band->he_cap_info[1] = le32_to_cpu(mac_caps->he_cap_info_2g_ext);
+ cap_band->he_cap_info[0] = mac_caps->he_cap_info_2g;
+ cap_band->he_cap_info[1] = mac_caps->he_cap_info_2g_ext;
cap_band->he_mcs = le32_to_cpu(mac_caps->he_supp_mcs_2g);
- for (i = 0; i < WMI_MAX_HECAP_PHY_SIZE; i++)
- cap_band->he_cap_phy_info[i] =
- le32_to_cpu(mac_caps->he_cap_phy_info_2g[i]);
+ memcpy(&cap_band->he_cap_phy_info, &mac_caps->he_cap_phy_info_2g,
+ sizeof(u32) * WMI_MAX_HECAP_PHY_SIZE);
cap_band->he_ppet.numss_m1 = le32_to_cpu(mac_caps->he_ppet2g.numss_m1);
cap_band->he_ppet.ru_bit_mask = le32_to_cpu(mac_caps->he_ppet2g.ru_info);
@@ -508,12 +507,11 @@ ath12k_pull_mac_phy_cap_svc_ready_ext(struct ath12k_wmi_pdev *wmi_handle,
cap_band->max_bw_supported =
le32_to_cpu(mac_caps->max_bw_supported_5g);
cap_band->ht_cap_info = le32_to_cpu(mac_caps->ht_cap_info_5g);
- cap_band->he_cap_info[0] = le32_to_cpu(mac_caps->he_cap_info_5g);
- cap_band->he_cap_info[1] = le32_to_cpu(mac_caps->he_cap_info_5g_ext);
+ cap_band->he_cap_info[0] = mac_caps->he_cap_info_5g;
+ cap_band->he_cap_info[1] = mac_caps->he_cap_info_5g_ext;
cap_band->he_mcs = le32_to_cpu(mac_caps->he_supp_mcs_5g);
- for (i = 0; i < WMI_MAX_HECAP_PHY_SIZE; i++)
- cap_band->he_cap_phy_info[i] =
- le32_to_cpu(mac_caps->he_cap_phy_info_5g[i]);
+ memcpy(&cap_band->he_cap_phy_info, &mac_caps->he_cap_phy_info_5g,
+ sizeof(u32) * WMI_MAX_HECAP_PHY_SIZE);
cap_band->he_ppet.numss_m1 = le32_to_cpu(mac_caps->he_ppet5g.numss_m1);
cap_band->he_ppet.ru_bit_mask = le32_to_cpu(mac_caps->he_ppet5g.ru_info);
@@ -526,12 +524,11 @@ ath12k_pull_mac_phy_cap_svc_ready_ext(struct ath12k_wmi_pdev *wmi_handle,
cap_band->max_bw_supported =
le32_to_cpu(mac_caps->max_bw_supported_5g);
cap_band->ht_cap_info = le32_to_cpu(mac_caps->ht_cap_info_5g);
- cap_band->he_cap_info[0] = le32_to_cpu(mac_caps->he_cap_info_5g);
- cap_band->he_cap_info[1] = le32_to_cpu(mac_caps->he_cap_info_5g_ext);
+ cap_band->he_cap_info[0] = mac_caps->he_cap_info_5g;
+ cap_band->he_cap_info[1] = mac_caps->he_cap_info_5g_ext;
cap_band->he_mcs = le32_to_cpu(mac_caps->he_supp_mcs_5g);
- for (i = 0; i < WMI_MAX_HECAP_PHY_SIZE; i++)
- cap_band->he_cap_phy_info[i] =
- le32_to_cpu(mac_caps->he_cap_phy_info_5g[i]);
+ memcpy(&cap_band->he_cap_phy_info, &mac_caps->he_cap_phy_info_5g,
+ sizeof(u32) * WMI_MAX_HECAP_PHY_SIZE);
cap_band->he_ppet.numss_m1 = le32_to_cpu(mac_caps->he_ppet5g.numss_m1);
cap_band->he_ppet.ru_bit_mask = le32_to_cpu(mac_caps->he_ppet5g.ru_info);
@@ -2226,14 +2223,13 @@ int ath12k_wmi_send_peer_assoc_cmd(struct ath12k *ar,
cmd->peer_phymode = cpu_to_le32(arg->peer_phymode);
/* Update 11ax capabilities */
- cmd->peer_he_cap_info = cpu_to_le32(arg->peer_he_cap_macinfo[0]);
- cmd->peer_he_cap_info_ext = cpu_to_le32(arg->peer_he_cap_macinfo[1]);
+ cmd->peer_he_cap_info = arg->peer_he_cap_macinfo[0];
+ cmd->peer_he_cap_info_ext = arg->peer_he_cap_macinfo[1];
cmd->peer_he_cap_info_internal = cpu_to_le32(arg->peer_he_cap_macinfo_internal);
cmd->peer_he_caps_6ghz = cpu_to_le32(arg->peer_he_caps_6ghz);
cmd->peer_he_ops = cpu_to_le32(arg->peer_he_ops);
- for (i = 0; i < WMI_MAX_HECAP_PHY_SIZE; i++)
- cmd->peer_he_cap_phy[i] =
- cpu_to_le32(arg->peer_he_cap_phyinfo[i]);
+ memcpy(cmd->peer_he_cap_phy, arg->peer_he_cap_phyinfo,
+ sizeof(u32) * WMI_MAX_HECAP_PHY_SIZE);
cmd->peer_ppet.numss_m1 = cpu_to_le32(arg->peer_ppet.numss_m1);
cmd->peer_ppet.ru_info = cpu_to_le32(arg->peer_ppet.ru_bit_mask);
for (i = 0; i < WMI_MAX_NUM_SS; i++)
@@ -5034,17 +5030,17 @@ static void ath12k_wmi_eht_caps_parse(struct ath12k_pdev *pdev, u32 band,
u8 i;
if (band == NL80211_BAND_6GHZ)
- support_320mhz = cap_band->eht_cap_phy_info[0] &
- IEEE80211_EHT_PHY_CAP0_320MHZ_IN_6GHZ;
+ support_320mhz = le32_to_cpu(cap_band->eht_cap_phy_info[0]) &
+ IEEE80211_EHT_PHY_CAP0_320MHZ_IN_6GHZ;
- for (i = 0; i < WMI_MAX_EHTCAP_MAC_SIZE; i++)
- cap_band->eht_cap_mac_info[i] = le32_to_cpu(cap_mac_info[i]);
+ memcpy(cap_band->eht_cap_mac_info, cap_mac_info,
+ sizeof(u32) * WMI_MAX_EHTCAP_MAC_SIZE);
- for (i = 0; i < WMI_MAX_EHTCAP_PHY_SIZE; i++)
- cap_band->eht_cap_phy_info[i] = le32_to_cpu(cap_phy_info[i]);
+ memcpy(cap_band->eht_cap_phy_info, cap_phy_info,
+ sizeof(u32) * WMI_MAX_EHTCAP_PHY_SIZE);
if (band == NL80211_BAND_6GHZ)
- cap_band->eht_cap_phy_info[0] |= support_320mhz;
+ cap_band->eht_cap_phy_info[0] |= cpu_to_le32(support_320mhz);
cap_band->eht_mcs_20_only = le32_to_cpu(supp_mcs[0]);
cap_band->eht_mcs_80 = le32_to_cpu(supp_mcs[1]);
@@ -5132,10 +5128,12 @@ static int ath12k_wmi_tlv_mac_phy_caps_ext(struct ath12k_base *ab, u16 tag,
if (ab->hw_params->single_pdev_only) {
if (caps->hw_mode_id == WMI_HOST_HW_MODE_SINGLE) {
- support_320mhz = le32_to_cpu(caps->eht_cap_phy_info_5ghz[0]) &
- IEEE80211_EHT_PHY_CAP0_320MHZ_IN_6GHZ;
+ support_320mhz =
+ le32_to_cpu(caps->eht_cap_phy_info_5ghz[0]) &
+ IEEE80211_EHT_PHY_CAP0_320MHZ_IN_6GHZ;
cap_band = &ab->pdevs[0].cap.band[NL80211_BAND_6GHZ];
- cap_band->eht_cap_phy_info[0] |= support_320mhz;
+ cap_band->eht_cap_phy_info[0] |=
+ cpu_to_le32(support_320mhz);
}
if (ab->wmi_ab.preferred_hw_mode != le32_to_cpu(caps->hw_mode_id))
diff --git a/drivers/net/wireless/ath/ath12k/wmi.h b/drivers/net/wireless/ath/ath12k/wmi.h
index 5ba9b7d3a888..ea680a1a5464 100644
--- a/drivers/net/wireless/ath/ath12k/wmi.h
+++ b/drivers/net/wireless/ath/ath12k/wmi.h
@@ -3911,11 +3911,11 @@ struct ath12k_wmi_peer_assoc_arg {
u8 peer_mac[ETH_ALEN];
bool he_flag;
- u32 peer_he_cap_macinfo[2];
+ __le32 peer_he_cap_macinfo[2];
u32 peer_he_cap_macinfo_internal;
u32 peer_he_caps_6ghz;
u32 peer_he_ops;
- u32 peer_he_cap_phyinfo[WMI_HOST_MAX_HECAP_PHY_SIZE];
+ __le32 peer_he_cap_phyinfo[WMI_HOST_MAX_HECAP_PHY_SIZE];
u32 peer_he_mcs_count;
u32 peer_he_rx_mcs_set[WMI_HOST_MAX_HE_RATE_SET];
u32 peer_he_tx_mcs_set[WMI_HOST_MAX_HE_RATE_SET];
---
base-commit: 702847e8cfd51856836a282db2073defd7cfd80c
change-id: 20260317-fix-he-eht-capabilities-on-big-endian-d941c42f65e5
Best regards,
--
Alexander Wilhelm <alexander.wilhelm@westermo.com>
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] wifi: ath12k: fix HE/EHT capability handling on big endian
2026-03-17 10:59 [PATCH] wifi: ath12k: fix HE/EHT capability handling on big endian Alexander Wilhelm
@ 2026-03-18 0:17 ` Jeff Johnson
2026-03-18 7:15 ` Alexander Wilhelm
0 siblings, 1 reply; 3+ messages in thread
From: Jeff Johnson @ 2026-03-18 0:17 UTC (permalink / raw)
To: Alexander Wilhelm, Jeff Johnson; +Cc: linux-wireless, ath12k, linux-kernel
On 3/17/2026 3:59 AM, Alexander Wilhelm wrote:
> Currently the driver uses u32 data types for the HE/EHT capabilities in
> CPU‑native order. However, the ieee80211.h header defines these fields as
> u8 arrays. This causes the ieee80211 registration failure on big‑endian
> platforms, as shown in the following log:
>
> ath12k_pci 0001:01:00.0: BAR 0: assigned [mem 0xc00000000-0xc001fffff 64bit]
> ath12k_pci 0001:01:00.0: MSI vectors: 1
> ath12k_pci 0001:01:00.0: Hardware name: qcn9274 hw2.0
> ath12k_pci 0001:01:00.0: qmi dma allocation failed (29360128 B type 1), will try later with small size
> ath12k_pci 0001:01:00.0: memory type 10 not supported
> ath12k_pci 0001:01:00.0: chip_id 0x0 chip_family 0xb board_id 0x1005 soc_id 0x401a2200
> ath12k_pci 0001:01:00.0: fw_version 0x111300d6 fw_build_timestamp 2024-08-06 08:43 fw_build_id QC_IMAGE_VERSION_STRING=WLAN.WBE.1.1.1-00214-QCAHKSWPL_SILICONZ-1
> ath12k_pci 0001:01:00.0: leaving PCI ASPM disabled to avoid MHI M2 problems
> ath12k_pci 0001:01:00.0: Invalid module id 2
> ath12k_pci 0001:01:00.0: failed to parse tlv -22
> ath12k_pci 0001:01:00.0: ieee80211 registration failed: -22
> ath12k_pci 0001:01:00.0: failed register the radio with mac80211: -22
> ath12k_pci 0001:01:00.0: failed to create pdev core: -22
> ath12k_pci 0001:01:00.0: qmi failed set mode request, mode: 4, err = -110
> ath12k_pci 0001:01:00.0: qmi failed to send wlan mode off
>
> Use the __le32 data type for the HE/EHT capabilities instead and avoid
> swapping, so that both platform endiannesses are supported.
>
> Signed-off-by: Alexander Wilhelm <alexander.wilhelm@westermo.com>
With this series applied I see with make W=1 C=1:
drivers/net/wireless/ath/ath12k/debugfs.c:413:65: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/ath/ath12k/debugfs.c:413:65: expected unsigned int [usertype] v
drivers/net/wireless/ath/ath12k/debugfs.c:413:65: got restricted __le32
From existing code:
f5c90ff80b4c0 (Sowmiya Sree Elavalagan 2025-01-30 11:41:04 +0530 413) extra_mcs_supported = u32_get_bits(cap_band->he_cap_info[1],
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] wifi: ath12k: fix HE/EHT capability handling on big endian
2026-03-18 0:17 ` Jeff Johnson
@ 2026-03-18 7:15 ` Alexander Wilhelm
0 siblings, 0 replies; 3+ messages in thread
From: Alexander Wilhelm @ 2026-03-18 7:15 UTC (permalink / raw)
To: Jeff Johnson; +Cc: Jeff Johnson, linux-wireless, ath12k, linux-kernel
On Tue, Mar 17, 2026 at 05:17:43PM -0700, Jeff Johnson wrote:
> On 3/17/2026 3:59 AM, Alexander Wilhelm wrote:
> > Currently the driver uses u32 data types for the HE/EHT capabilities in
> > CPU‑native order. However, the ieee80211.h header defines these fields as
> > u8 arrays. This causes the ieee80211 registration failure on big‑endian
> > platforms, as shown in the following log:
> >
> > ath12k_pci 0001:01:00.0: BAR 0: assigned [mem 0xc00000000-0xc001fffff 64bit]
> > ath12k_pci 0001:01:00.0: MSI vectors: 1
> > ath12k_pci 0001:01:00.0: Hardware name: qcn9274 hw2.0
> > ath12k_pci 0001:01:00.0: qmi dma allocation failed (29360128 B type 1), will try later with small size
> > ath12k_pci 0001:01:00.0: memory type 10 not supported
> > ath12k_pci 0001:01:00.0: chip_id 0x0 chip_family 0xb board_id 0x1005 soc_id 0x401a2200
> > ath12k_pci 0001:01:00.0: fw_version 0x111300d6 fw_build_timestamp 2024-08-06 08:43 fw_build_id QC_IMAGE_VERSION_STRING=WLAN.WBE.1.1.1-00214-QCAHKSWPL_SILICONZ-1
> > ath12k_pci 0001:01:00.0: leaving PCI ASPM disabled to avoid MHI M2 problems
> > ath12k_pci 0001:01:00.0: Invalid module id 2
> > ath12k_pci 0001:01:00.0: failed to parse tlv -22
> > ath12k_pci 0001:01:00.0: ieee80211 registration failed: -22
> > ath12k_pci 0001:01:00.0: failed register the radio with mac80211: -22
> > ath12k_pci 0001:01:00.0: failed to create pdev core: -22
> > ath12k_pci 0001:01:00.0: qmi failed set mode request, mode: 4, err = -110
> > ath12k_pci 0001:01:00.0: qmi failed to send wlan mode off
> >
> > Use the __le32 data type for the HE/EHT capabilities instead and avoid
> > swapping, so that both platform endiannesses are supported.
> >
> > Signed-off-by: Alexander Wilhelm <alexander.wilhelm@westermo.com>
>
> With this series applied I see with make W=1 C=1:
> drivers/net/wireless/ath/ath12k/debugfs.c:413:65: warning: incorrect type in argument 1 (different base types)
> drivers/net/wireless/ath/ath12k/debugfs.c:413:65: expected unsigned int [usertype] v
> drivers/net/wireless/ath/ath12k/debugfs.c:413:65: got restricted __le32
>
> From existing code:
> f5c90ff80b4c0 (Sowmiya Sree Elavalagan 2025-01-30 11:41:04 +0530 413) extra_mcs_supported = u32_get_bits(cap_band->he_cap_info[1],
Sorry, I had `ATH12K_DEBUGFS` disabled when I ran the sparse check. I’ll
fix this in the next version.
Best regards
Alexander Wilhelm
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-03-18 7:15 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-03-17 10:59 [PATCH] wifi: ath12k: fix HE/EHT capability handling on big endian Alexander Wilhelm
2026-03-18 0:17 ` Jeff Johnson
2026-03-18 7:15 ` Alexander Wilhelm
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox