[PATCH] wifi: mt76: mt7928: only do CLC hw type sanity check on MT7925

[PATCH] wifi: mt76: mt7928: only do CLC hw type sanity check on MT7925

[JB Tsai]

From: Charlie-cy Wu <Charlie-cy.Wu@mediatek.com>

mt7925_load_clc() is shared by MT7925, MT7927 and MT7928 (0x7935 is
folded into 0x7928 in pci.c). The hw type sanity check it does on every
CLC region is MT7925-specific twice over: the encap bits live at
efuse offset MT_EE_HW_TYPE (0xa71) in the MT7925 layout, and only
MT7925 CLC blobs carry that hw type in their t0/t2 header. On
MT7927/MT7928 the driver therefore reads a byte that means nothing on
those parts and compares it against a header field that was never
populated, so a mismatch silently skips every CLC region and leaves
phy->clc[] empty - no country/regd table gets installed at all.

Gate both the efuse read and the two header checks on is_mt7925(), so
MT7927/MT7928 accept the CLC regions from their own firmware as-is.
No functional change for MT7925: PCIe still reads 0xa71 and filters by
encap type, USB still runs with hw_encap = 0 as before.

Fixes: 356b39285559 ("wifi: mt76: connac3: refactor CLC header type validation logic")

Signed-off-by: Charlie-cy Wu <Charlie-cy.Wu@mediatek.com>
---
 .../net/wireless/mediatek/mt76/mt7925/mcu.c    | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/wireless/mediatek/mt76/mt7925/mcu.c b/drivers/net/wireless/mediatek/mt76/mt7925/mcu.c
index bd63789ab8f7..bf2cb6172cc8 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7925/mcu.c
+++ b/drivers/net/wireless/mediatek/mt76/mt7925/mcu.c
@@ -855,7 +855,7 @@ static int mt7925_load_clc(struct mt792x_dev *dev, const char *fw_name)
 	if (!mt7925_regd_clc_supported(dev))
 		return 0;
 
-	if (mt76_is_mmio(&dev->mt76)) {
+	if (is_mt7925(&dev->mt76) && mt76_is_mmio(&dev->mt76)) {
 		ret = mt7925_mcu_read_eeprom(dev, MT_EE_HW_TYPE, &hw_encap);
 		if (ret)
 			return ret;
@@ -906,14 +906,16 @@ static int mt7925_load_clc(struct mt792x_dev *dev, const char *fw_name)
 		if (phy->clc[clc->idx])
 			continue;
 
-		/* header content sanity */
-		if ((clc->idx == MT792x_CLC_BE_CTRL || clc->idx == MT792x_CLC_REGD) &&
-		    u8_get_bits(clc->t2.type, MT_EE_HW_TYPE_ENCAP) != hw_encap)
-			continue;
+		/* header content sanity, only MT7925 encodes hw type in CLC */
+		if (is_mt7925(&dev->mt76)) {
+			if ((clc->idx == MT792x_CLC_BE_CTRL || clc->idx == MT792x_CLC_REGD) &&
+			    u8_get_bits(clc->t2.type, MT_EE_HW_TYPE_ENCAP) != hw_encap)
+				continue;
 
-		if (clc->idx != MT792x_CLC_BE_CTRL && clc->idx != MT792x_CLC_REGD &&
-		    u8_get_bits(clc->t0.type, MT_EE_HW_TYPE_ENCAP) != hw_encap)
-			continue;
+			if (clc->idx != MT792x_CLC_BE_CTRL && clc->idx != MT792x_CLC_REGD &&
+			    u8_get_bits(clc->t0.type, MT_EE_HW_TYPE_ENCAP) != hw_encap)
+				continue;
+		}
 
 		phy->clc[clc->idx] = devm_kmemdup(mdev->dev, clc,
 						  le32_to_cpu(clc->len),
-- 
2.18.0

Re: [PATCH] wifi: mt76: mt7928: only do CLC hw type sanity check on MT7925

[Felix Fietkau]

On 09.09.26 07:57, JB Tsai wrote:
> From: Charlie-cy Wu <Charlie-cy.Wu@mediatek.com>
> 
> mt7925_load_clc() is shared by MT7925, MT7927 and MT7928 (0x7935 is
> folded into 0x7928 in pci.c). The hw type sanity check it does on every
> CLC region is MT7925-specific twice over: the encap bits live at
> efuse offset MT_EE_HW_TYPE (0xa71) in the MT7925 layout, and only
> MT7925 CLC blobs carry that hw type in their t0/t2 header. On
> MT7927/MT7928 the driver therefore reads a byte that means nothing on
> those parts and compares it against a header field that was never
> populated, so a mismatch silently skips every CLC region and leaves
> phy->clc[] empty - no country/regd table gets installed at all.
> 
> Gate both the efuse read and the two header checks on is_mt7925(), so
> MT7927/MT7928 accept the CLC regions from their own firmware as-is.
> No functional change for MT7925: PCIe still reads 0xa71 and filters by
> encap type, USB still runs with hw_encap = 0 as before.
> 
> Fixes: 356b39285559 ("wifi: mt76: connac3: refactor CLC header type validation logic")
Patch doesn't apply, and I can't find the commit mentioned above.

- Felix