diff mbox series

wifi: rtw89: pci: early chips only enable 36-bit DMA on specific PCI hosts

Message ID 20240924021633.19861-1-pkshih@realtek.com (mailing list archive)
State Accepted
Commit aa70ff0945fea2ed14046273609d04725f222616
Delegated to: Kalle Valo
Headers show
Series wifi: rtw89: pci: early chips only enable 36-bit DMA on specific PCI hosts | expand

Commit Message

Ping-Ke Shih Sept. 24, 2024, 2:16 a.m. UTC
The early chips including RTL8852A, RTL8851B, RTL8852B and RTL8852BT have
interoperability problems of 36-bit DMA with some PCI hosts. Rollback
to 32-bit DMA by default, and only enable 36-bit DMA for tested platforms.

Since all Intel platforms we have can work correctly, add the vendor ID to
white list. Otherwise, list vendor/device ID of bridge we have tested.

Fixes: 1fd4b3fe52ef ("wifi: rtw89: pci: support 36-bit PCI DMA address")
Reported-by: Marcel Weißenbach <mweissenbach@ignaz.org>
Closes: https://lore.kernel.org/linux-wireless/20240918073237.Horde.VLueh0_KaiDw-9asEEcdM84@ignaz.org/T/#m07c5694df1acb173a42e1a0bab7ac22bd231a2b8
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
---
 drivers/net/wireless/realtek/rtw89/pci.c | 48 ++++++++++++++++++++----
 1 file changed, 41 insertions(+), 7 deletions(-)

Comments

Ping-Ke Shih Sept. 27, 2024, 1:34 a.m. UTC | #1
Hi Kalle,

Ping-Ke Shih <pkshih@realtek.com> wrote:
> 
> The early chips including RTL8852A, RTL8851B, RTL8852B and RTL8852BT have
> interoperability problems of 36-bit DMA with some PCI hosts. Rollback
> to 32-bit DMA by default, and only enable 36-bit DMA for tested platforms.
> 
> Since all Intel platforms we have can work correctly, add the vendor ID to
> white list. Otherwise, list vendor/device ID of bridge we have tested.
> 
> Fixes: 1fd4b3fe52ef ("wifi: rtw89: pci: support 36-bit PCI DMA address")
> Reported-by: Marcel Weißenbach <mweissenbach@ignaz.org>
> Closes:
> https://lore.kernel.org/linux-wireless/20240918073237.Horde.VLueh0_KaiDw-9asEEcdM84@ignaz.org/T/#m07c5
> 694df1acb173a42e1a0bab7ac22bd231a2b8
> Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>

Could you please queue this patch to wireless tree? Because RTL8852BE becomes
unusable for some users. I assigned this patch to you in patchwork, if you
don't think so and have other suggestions, please let me know. Thanks. 

The user has reported this patch works as expected [1], so add his tested-by:
Tested-by: Marcel Weißenbach <mweissenbach@ignaz.org>

Also, I think this should Cc stable, so
Cc: stable@vger.kernel.org

[1] https://lore.kernel.org/linux-wireless/20240927012001.Horde.vJNTbT576lzyFGWprtGx7R0@IgnazServer/T/#m01470ad9bba13e46ce9a0dd6efaffe1e176ad05e
Kalle Valo Sept. 28, 2024, 11:19 a.m. UTC | #2
Ping-Ke Shih <pkshih@realtek.com> writes:

> Ping-Ke Shih <pkshih@realtek.com> wrote:
>> 
>> The early chips including RTL8852A, RTL8851B, RTL8852B and RTL8852BT have
>> interoperability problems of 36-bit DMA with some PCI hosts. Rollback
>> to 32-bit DMA by default, and only enable 36-bit DMA for tested platforms.
>> 
>> Since all Intel platforms we have can work correctly, add the vendor ID to
>> white list. Otherwise, list vendor/device ID of bridge we have tested.
>> 
>> Fixes: 1fd4b3fe52ef ("wifi: rtw89: pci: support 36-bit PCI DMA address")
>> Reported-by: Marcel Weißenbach <mweissenbach@ignaz.org>
>> Closes:
>> https://lore.kernel.org/linux-wireless/20240918073237.Horde.VLueh0_KaiDw-9asEEcdM84@ignaz.org/T/#m07c5
>> 694df1acb173a42e1a0bab7ac22bd231a2b8
>> Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
>
> Could you please queue this patch to wireless tree? Because RTL8852BE becomes
> unusable for some users. I assigned this patch to you in patchwork, if you
> don't think so and have other suggestions, please let me know. Thanks. 
>
> The user has reported this patch works as expected [1], so add his tested-by:
> Tested-by: Marcel Weißenbach <mweissenbach@ignaz.org>
>
> Also, I think this should Cc stable, so
> Cc: stable@vger.kernel.org

Ok, I'll add all that.
Marcel Weißenbach Oct. 7, 2024, 12:25 p.m. UTC | #3
Hi there,
 
just wanted to kindly ask in which Kernel Version this patch will be implemented (given it is not yet).
 
I assume it did not make it into 6.11.2, or?
 
"Kalle Valo" kvalo@kernel.org – 2024年9月28日 20:19
> Ping-Ke Shih <pkshih@realtek.com> writes:

> > Ping-Ke Shih <pkshih@realtek.com> wrote:
> >> 
> >> The early chips including RTL8852A, RTL8851B, RTL8852B and RTL8852BT have
> >> interoperability problems of 36-bit DMA with some PCI hosts. Rollback
> >> to 32-bit DMA by default, and only enable 36-bit DMA for tested platforms.
> >> 
> >> Since all Intel platforms we have can work correctly, add the vendor ID to
> >> white list. Otherwise, list vendor/device ID of bridge we have tested.
> >> 
> >> Fixes: 1fd4b3fe52ef ("wifi: rtw89: pci: support 36-bit PCI DMA address")
> >> Reported-by: Marcel Weißenbach <mweissenbach@ignaz.org>
> >> Closes:
> >> https://lore.kernel.org/linux-wireless/20240918073237.Horde.VLueh0_KaiDw-9asEEcdM84@ignaz.org/T/#m07c5
> >> 694df1acb173a42e1a0bab7ac22bd231a2b8
> >> Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
> >
> > Could you please queue this patch to wireless tree? Because RTL8852BE becomes
> > unusable for some users. I assigned this patch to you in patchwork, if you
> > don't think so and have other suggestions, please let me know. Thanks. 
> >
> > The user has reported this patch works as expected [1], so add his tested-by:
> > Tested-by: Marcel Weißenbach <mweissenbach@ignaz.org>
> >
> > Also, I think this should Cc stable, so
> > Cc: stable@vger.kernel.org

> Ok, I'll add all that.

>
Kalle Valo Oct. 11, 2024, 12:43 p.m. UTC | #4
Marcel Weißenbach <mweissenbach@ignaz.org> writes:

> Hi there,

Please don't top post.

> just wanted to kindly ask in which Kernel Version this patch will be implemented (given it is not yet).

This is not yet applied because I have been busy.
Kalle Valo Oct. 17, 2024, 2:23 p.m. UTC | #5
Ping-Ke Shih <pkshih@realtek.com> wrote:

> The early chips including RTL8852A, RTL8851B, RTL8852B and RTL8852BT have
> interoperability problems of 36-bit DMA with some PCI hosts. Rollback
> to 32-bit DMA by default, and only enable 36-bit DMA for tested platforms.
> 
> Since all Intel platforms we have can work correctly, add the vendor ID to
> white list. Otherwise, list vendor/device ID of bridge we have tested.
> 
> Fixes: 1fd4b3fe52ef ("wifi: rtw89: pci: support 36-bit PCI DMA address")
> Reported-by: Marcel Weißenbach <mweissenbach@ignaz.org>
> Closes: https://lore.kernel.org/linux-wireless/20240918073237.Horde.VLueh0_KaiDw-9asEEcdM84@ignaz.org/T/#m07c5694df1acb173a42e1a0bab7ac22bd231a2b8
> Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
> Tested-by: Marcel Weißenbach <mweissenbach@ignaz.org>

Patch applied to wireless.git, thanks.

aa70ff0945fe wifi: rtw89: pci: early chips only enable 36-bit DMA on specific PCI hosts
diff mbox series

Patch

diff --git a/drivers/net/wireless/realtek/rtw89/pci.c b/drivers/net/wireless/realtek/rtw89/pci.c
index 02afeb3acce4..5aef7fa37878 100644
--- a/drivers/net/wireless/realtek/rtw89/pci.c
+++ b/drivers/net/wireless/realtek/rtw89/pci.c
@@ -3026,24 +3026,54 @@  static void rtw89_pci_declaim_device(struct rtw89_dev *rtwdev,
 	pci_disable_device(pdev);
 }
 
-static void rtw89_pci_cfg_dac(struct rtw89_dev *rtwdev)
+static bool rtw89_pci_chip_is_manual_dac(struct rtw89_dev *rtwdev)
 {
-	struct rtw89_pci *rtwpci = (struct rtw89_pci *)rtwdev->priv;
 	const struct rtw89_chip_info *chip = rtwdev->chip;
 
-	if (!rtwpci->enable_dac)
-		return;
-
 	switch (chip->chip_id) {
 	case RTL8852A:
 	case RTL8852B:
 	case RTL8851B:
 	case RTL8852BT:
-		break;
+		return true;
 	default:
-		return;
+		return false;
+	}
+}
+
+static bool rtw89_pci_is_dac_compatible_bridge(struct rtw89_dev *rtwdev)
+{
+	struct rtw89_pci *rtwpci = (struct rtw89_pci *)rtwdev->priv;
+	struct pci_dev *bridge = pci_upstream_bridge(rtwpci->pdev);
+
+	if (!rtw89_pci_chip_is_manual_dac(rtwdev))
+		return true;
+
+	if (!bridge)
+		return false;
+
+	switch (bridge->vendor) {
+	case PCI_VENDOR_ID_INTEL:
+		return true;
+	case PCI_VENDOR_ID_ASMEDIA:
+		if (bridge->device == 0x2806)
+			return true;
+		break;
 	}
 
+	return false;
+}
+
+static void rtw89_pci_cfg_dac(struct rtw89_dev *rtwdev)
+{
+	struct rtw89_pci *rtwpci = (struct rtw89_pci *)rtwdev->priv;
+
+	if (!rtwpci->enable_dac)
+		return;
+
+	if (!rtw89_pci_chip_is_manual_dac(rtwdev))
+		return;
+
 	rtw89_pci_config_byte_set(rtwdev, RTW89_PCIE_L1_CTRL, RTW89_PCIE_BIT_EN_64BITS);
 }
 
@@ -3061,6 +3091,9 @@  static int rtw89_pci_setup_mapping(struct rtw89_dev *rtwdev,
 		goto err;
 	}
 
+	if (!rtw89_pci_is_dac_compatible_bridge(rtwdev))
+		goto no_dac;
+
 	ret = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(36));
 	if (!ret) {
 		rtwpci->enable_dac = true;
@@ -3073,6 +3106,7 @@  static int rtw89_pci_setup_mapping(struct rtw89_dev *rtwdev,
 			goto err_release_regions;
 		}
 	}
+no_dac:
 
 	resource_len = pci_resource_len(pdev, bar_id);
 	rtwpci->mmap = pci_iomap(pdev, bar_id, resource_len);