[PATCH] wifi: rtw88: 8821c: Add HP single-antenna quirk

[PATCH] wifi: rtw88: 8821c: Add HP single-antenna quirk

[not-eexto]

Certain HP laptop platforms with RTL8821CE (PCI ID 10ec:c821, subsystem
103c:831a) are manufactured with only a single physical antenna wire
connected to the Aux/S1 port (Port 2). As documented in the OEM Windows
INF file (netrtwlane.inf):
  "single ant(connect to right one S1)"

However, the eFuse on these chipsets reports RFE Type 2 (Main antenna,
Port 1), causing the hardware SPDT RF switch to route to an unconnected
port under Linux. This leads to severe signal attenuation (~ -86 dBm)
and frequent beacon-loss disconnections.

Add an automatic quirk in rtw8821c_read_efuse() for HP subsystem
10ec:c821 [103c:831a] that overrides efuse->rfe_option and
efuse->rfe_option_full to 4 (Aux), routing the RF path to the physically
connected antenna.

Signed-off-by: not-eexto <p17232763@gmail.com>
---
 rtw8821c.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drivers/net/wireless/realtek/rtw88/rtw8821c.c b/drivers/net/wireless/realtek/rtw88/rtw8821c.c
index 9f0964d..32c7de5 100644
--- a/drivers/net/wireless/realtek/rtw88/rtw8821c.c
+++ b/drivers/net/wireless/realtek/rtw88/rtw8821c.c
@@ -15,6 +15,9 @@
 #include "debug.h"
 #include "bf.h"
 #include "regd.h"
+#if IS_ENABLED(CONFIG_PCI)
+#include <linux/pci.h>
+#endif
 
 static const s8 lna_gain_table_0[8] = {22, 8, -6, -22, -31, -40, -46, -52};
 static const s8 lna_gain_table_1[16] = {10, 6, 2, -2, -6, -10, -14, -17,
@@ -56,6 +59,21 @@ static int rtw8821c_read_efuse(struct rtw_dev *rtwdev, u8 *log_map)
 
 	efuse->rfe_option = map->rfe_option & 0x1f;
 	efuse->rfe_option_full = map->rfe_option;
+
+#if IS_ENABLED(CONFIG_PCI)
+	if (rtw_hci_type(rtwdev) == RTW_HCI_TYPE_PCIE && dev_is_pci(rtwdev->dev)) {
+		struct pci_dev *pdev = to_pci_dev(rtwdev->dev);
+
+		if (pdev->vendor == PCI_VENDOR_ID_REALTEK &&
+		    pdev->device == 0xc821 &&
+		    pdev->subsystem_vendor == 0x103c &&
+		    pdev->subsystem_device == 0x831a) {
+			efuse->rfe_option = 4;
+			efuse->rfe_option_full = 4;
+			rtw_info(rtwdev, "HP quirk auto-detected (10ec:c821 [103c:831a]): forced RFE option 4 (Aux antenna S1)\n");
+		}
+	}
+#endif
 	efuse->rf_board_option = map->rf_board_option;
 	efuse->crystal_cap = map->xtal_k;
 	efuse->pa_type_2g = map->pa_type;
-- 
2.43.0

[PATCH] wifi: rtw88: 8821c: Add HP single-antenna quirk

[not-eexto]

Certain HP laptop platforms with RTL8821CE (PCI ID 10ec:c821, subsystem
103c:831a) are manufactured with only a single physical antenna wire
connected to the Aux/S1 port (Port 2). As documented in the OEM Windows
INF file (netrtwlane.inf):
  "single ant(connect to right one S1)"

However, the eFuse on these chipsets reports RFE Type 2 (Main antenna,
Port 1), causing the hardware SPDT RF switch to route to an unconnected
port under Linux. This leads to severe signal attenuation (~ -86 dBm)
and frequent beacon-loss disconnections.

Add an automatic quirk in rtw8821c_read_efuse() for HP subsystem
10ec:c821 [103c:831a] that overrides efuse->rfe_option and
efuse->rfe_option_full to 4 (Aux), routing the RF path to the physically
connected antenna.

Signed-off-by: not-eexto <eexto@aol.com>
---
 rtw8821c.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drivers/net/wireless/realtek/rtw88/rtw8821c.c b/drivers/net/wireless/realtek/rtw88/rtw8821c.c
index 9f0964d..32c7de5 100644
--- a/drivers/net/wireless/realtek/rtw88/rtw8821c.c
+++ b/drivers/net/wireless/realtek/rtw88/rtw8821c.c
@@ -15,6 +15,9 @@
 #include "debug.h"
 #include "bf.h"
 #include "regd.h"
+#if IS_ENABLED(CONFIG_PCI)
+#include <linux/pci.h>
+#endif
 
 static const s8 lna_gain_table_0[8] = {22, 8, -6, -22, -31, -40, -46, -52};
 static const s8 lna_gain_table_1[16] = {10, 6, 2, -2, -6, -10, -14, -17,
@@ -56,6 +59,21 @@ static int rtw8821c_read_efuse(struct rtw_dev *rtwdev, u8 *log_map)
 
 	efuse->rfe_option = map->rfe_option & 0x1f;
 	efuse->rfe_option_full = map->rfe_option;
+
+#if IS_ENABLED(CONFIG_PCI)
+	if (rtw_hci_type(rtwdev) == RTW_HCI_TYPE_PCIE && dev_is_pci(rtwdev->dev)) {
+		struct pci_dev *pdev = to_pci_dev(rtwdev->dev);
+
+		if (pdev->vendor == PCI_VENDOR_ID_REALTEK &&
+		    pdev->device == 0xc821 &&
+		    pdev->subsystem_vendor == 0x103c &&
+		    pdev->subsystem_device == 0x831a) {
+			efuse->rfe_option = 4;
+			efuse->rfe_option_full = 4;
+			rtw_info(rtwdev, "HP quirk auto-detected (10ec:c821 [103c:831a]): forced RFE option 4 (Aux antenna S1)\n");
+		}
+	}
+#endif
 	efuse->rf_board_option = map->rf_board_option;
 	efuse->crystal_cap = map->xtal_k;
 	efuse->pa_type_2g = map->pa_type;
-- 
2.43.0

RE: [PATCH] wifi: rtw88: 8821c: Add HP single-antenna quirk

[Ping-Ke Shih]

not-eexto <eexto@aol.com> wrote:
> Certain HP laptop platforms with RTL8821CE (PCI ID 10ec:c821, subsystem
> 103c:831a) are manufactured with only a single physical antenna wire
> connected to the Aux/S1 port (Port 2). As documented in the OEM Windows
> INF file (netrtwlane.inf):
>   "single ant(connect to right one S1)"
> 
> However, the eFuse on these chipsets reports RFE Type 2 (Main antenna,
> Port 1), causing the hardware SPDT RF switch to route to an unconnected
> port under Linux. This leads to severe signal attenuation (~ -86 dBm)
> and frequent beacon-loss disconnections.
> 
> Add an automatic quirk in rtw8821c_read_efuse() for HP subsystem
> 10ec:c821 [103c:831a] that overrides efuse->rfe_option and
> efuse->rfe_option_full to 4 (Aux), routing the RF path to the physically
> connected antenna.
> 
> Signed-off-by: not-eexto <eexto@aol.com>

Should use your full real name.

You sent the same patch by different mail address. I'll drop previous one.

> ---
>  rtw8821c.c | 18 ++++++++++++++++++
>  1 file changed, 18 insertions(+)
> 
> diff --git a/drivers/net/wireless/realtek/rtw88/rtw8821c.c
> b/drivers/net/wireless/realtek/rtw88/rtw8821c.c
> index 9f0964d..32c7de5 100644
> --- a/drivers/net/wireless/realtek/rtw88/rtw8821c.c
> +++ b/drivers/net/wireless/realtek/rtw88/rtw8821c.c
> @@ -15,6 +15,9 @@
>  #include "debug.h"
>  #include "bf.h"
>  #include "regd.h"
> +#if IS_ENABLED(CONFIG_PCI)
> +#include <linux/pci.h>
> +#endif

If this is the only one platform which needs this kind of quirk. Add an entry
to rtw_pci_quirks[].

> 
>  static const s8 lna_gain_table_0[8] = {22, 8, -6, -22, -31, -40, -46, -52};
>  static const s8 lna_gain_table_1[16] = {10, 6, 2, -2, -6, -10, -14, -17,
> @@ -56,6 +59,21 @@ static int rtw8821c_read_efuse(struct rtw_dev *rtwdev, u8 *log_map)
> 
>         efuse->rfe_option = map->rfe_option & 0x1f;
>         efuse->rfe_option_full = map->rfe_option;
> +
> +#if IS_ENABLED(CONFIG_PCI)
> +       if (rtw_hci_type(rtwdev) == RTW_HCI_TYPE_PCIE && dev_is_pci(rtwdev->dev)) {
> +               struct pci_dev *pdev = to_pci_dev(rtwdev->dev);
> +
> +               if (pdev->vendor == PCI_VENDOR_ID_REALTEK &&
> +                   pdev->device == 0xc821 &&
> +                   pdev->subsystem_vendor == 0x103c &&
> +                   pdev->subsystem_device == 0x831a) {
> +                       efuse->rfe_option = 4;
> +                       efuse->rfe_option_full = 4;

I don't see rtw88 has ' rfe_option_full' ?

> +                       rtw_info(rtwdev, "HP quirk auto-detected (10ec:c821 [103c:831a]): forced RFE
> option 4 (Aux antenna S1)\n");
> +               }
> +       }
> +#endif
>         efuse->rf_board_option = map->rf_board_option;
>         efuse->crystal_cap = map->xtal_k;
>         efuse->pa_type_2g = map->pa_type;
> --
> 2.43.0

Re: [PATCH] wifi: rtw88: 8821c: Add HP single-antenna quirk

[Paolo Lazzaroni]

On Sun, Sep 13, 2026 at 02:15 UTC, Ping-Ke Shih <pkshih@realtek.com> wrote:
> Should use your full real name.
> 
> You sent the same patch by different mail address. I'll drop previous one.

Both patch submissions were identical, but please ignore the submission from
p17232763@gmail.com. I will use this address (eexto@aol.com) with my full
legal name (Paolo Lazzaroni) for all future submissions and v2.

> I don't see rtw88 has ' rfe_option_full' ?

Apologies, the previous patch was based on an out-of-tree backport tree.
I have checked out rtw-next where rfe_option_full does not exist, and will
only set efuse->rfe_option in v2.

> If this is the only one platform which needs this kind of quirk. Add an entry
> to rtw_pci_quirks[].

Regarding rtw_pci_quirks[]:

The reason I matched the PCI subsystem ID (10ec:c821 [103c:831a]) rather
than DMI in rtw_pci_quirks[] is that HP assigns this specific subsystem
ID across a family of laptop platforms (e.g. HP 15-da, 15-db, and related
chassis) to denote the single physical antenna connected to Aux/S1.
This is documented in the OEM Windows driver INF (netrtwlane.inf):
  "PCI\VEN_10EC&DEV_C821&SUBSYS_831A103C ;;single ant(connect to right one S1)"

If we use DMI matching in rtw_pci_quirks[], it would only match this specific
laptop SKU ("HP Laptop 15-da0xxx" / "4XW76EA#ABZ"), leaving other HP models
sharing this identical 103c:831a single-antenna card with severe signal loss.

However, I understand that inspecting PCI structures inside rtw8821c_read_efuse()
violates bus abstraction for rtw8821c.c.

Would you prefer:
Option A: Keep the PCI subsystem match, but handle it cleanly in rtw_pci_probe()
          (e.g., setting a flag or rfe_option override in rtwdev before chip setup)?
Option B: Add a DMI entry to rtw_pci_quirks[] matching this platform?

RE: [PATCH] wifi: rtw88: 8821c: Add HP single-antenna quirk

[Ping-Ke Shih]

Paolo Lazzaroni <eexto@aol.com> wrote:
> > If this is the only one platform which needs this kind of quirk. Add an entry
> > to rtw_pci_quirks[].
> 
> Regarding rtw_pci_quirks[]:
> 
> The reason I matched the PCI subsystem ID (10ec:c821 [103c:831a]) rather
> than DMI in rtw_pci_quirks[] is that HP assigns this specific subsystem
> ID across a family of laptop platforms (e.g. HP 15-da, 15-db, and related
> chassis) to denote the single physical antenna connected to Aux/S1.
> This is documented in the OEM Windows driver INF (netrtwlane.inf):
>   "PCI\VEN_10EC&DEV_C821&SUBSYS_831A103C ;;single ant(connect to right one S1)"
> 
> If we use DMI matching in rtw_pci_quirks[], it would only match this specific
> laptop SKU ("HP Laptop 15-da0xxx" / "4XW76EA#ABZ"), leaving other HP models
> sharing this identical 103c:831a single-antenna card with severe signal loss.
> 
> However, I understand that inspecting PCI structures inside rtw8821c_read_efuse()
> violates bus abstraction for rtw8821c.c.
> 
> Would you prefer:
> Option A: Keep the PCI subsystem match, but handle it cleanly in rtw_pci_probe()
>           (e.g., setting a flag or rfe_option override in rtwdev before chip setup)?
> Option B: Add a DMI entry to rtw_pci_quirks[] matching this platform?

I prefer option A, but yes, please handle it cleanly in pci.c. 
You can flag a quirk at pci.c, and update rfe_option in 8821c.c.

Ping-Ke