The Hygon Dhyana CPU has the SMBus device with PCI device ID 0x790b, which is the same as AMD CZ SMBus device. So add Hygon Dhyana support to the i2c-piix4 driver by using the code path of AMD.
Cc: stable@vger.kernel.org # v5.0+ Signed-off-by: Pu Wen puwen@hygon.cn --- drivers/i2c/busses/i2c-piix4.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
diff --git a/drivers/i2c/busses/i2c-piix4.c b/drivers/i2c/busses/i2c-piix4.c index 90946a8..9db9d9d 100644 --- a/drivers/i2c/busses/i2c-piix4.c +++ b/drivers/i2c/busses/i2c-piix4.c @@ -289,6 +289,9 @@ static int piix4_setup_sb800(struct pci_dev *PIIX4_dev, PIIX4_dev->revision >= 0x41) || (PIIX4_dev->vendor == PCI_VENDOR_ID_AMD && PIIX4_dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS && + PIIX4_dev->revision >= 0x49) || + (PIIX4_dev->vendor == PCI_VENDOR_ID_HYGON && + PIIX4_dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS && PIIX4_dev->revision >= 0x49)) smb_en = 0x00; else @@ -361,7 +364,8 @@ static int piix4_setup_sb800(struct pci_dev *PIIX4_dev, piix4_smba, i2ccfg >> 4);
/* Find which register is used for port selection */ - if (PIIX4_dev->vendor == PCI_VENDOR_ID_AMD) { + if (PIIX4_dev->vendor == PCI_VENDOR_ID_AMD || + PIIX4_dev->vendor == PCI_VENDOR_ID_HYGON) { switch (PIIX4_dev->device) { case PCI_DEVICE_ID_AMD_KERNCZ_SMBUS: piix4_port_sel_sb800 = SB800_PIIX4_PORT_IDX_KERNCZ; @@ -794,6 +798,7 @@ static const struct pci_device_id piix4_ids[] = { { PCI_DEVICE(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_SBX00_SMBUS) }, { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_HUDSON2_SMBUS) }, { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) }, + { PCI_DEVICE(PCI_VENDOR_ID_HYGON, PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) }, { PCI_DEVICE(PCI_VENDOR_ID_SERVERWORKS, PCI_DEVICE_ID_SERVERWORKS_OSB4) }, { PCI_DEVICE(PCI_VENDOR_ID_SERVERWORKS, @@ -904,11 +909,13 @@ static int piix4_probe(struct pci_dev *dev, const struct pci_device_id *id) if ((dev->vendor == PCI_VENDOR_ID_ATI && dev->device == PCI_DEVICE_ID_ATI_SBX00_SMBUS && dev->revision >= 0x40) || - dev->vendor == PCI_VENDOR_ID_AMD) { + dev->vendor == PCI_VENDOR_ID_AMD || + dev->vendor == PCI_VENDOR_ID_HYGON) { bool notify_imc = false; is_sb800 = true;
- if (dev->vendor == PCI_VENDOR_ID_AMD && + if ((dev->vendor == PCI_VENDOR_ID_AMD || + dev->vendor == PCI_VENDOR_ID_HYGON) && dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) { u8 imc;
Hi Pu,
On Mon, 22 Apr 2019 21:10:07 +0800, Pu Wen wrote:
The Hygon Dhyana CPU has the SMBus device with PCI device ID 0x790b, which is the same as AMD CZ SMBus device. So add Hygon Dhyana support to the i2c-piix4 driver by using the code path of AMD.
Sorry for the late answer.
Cc: stable@vger.kernel.org # v5.0+
I object to adding this to stable. It's not fixing any bug, and it's far from being a one-liner. I'd rather let distributions backport it as they see fit.
Signed-off-by: Pu Wen puwen@hygon.cn
drivers/i2c/busses/i2c-piix4.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
diff --git a/drivers/i2c/busses/i2c-piix4.c b/drivers/i2c/busses/i2c-piix4.c index 90946a8..9db9d9d 100644 --- a/drivers/i2c/busses/i2c-piix4.c +++ b/drivers/i2c/busses/i2c-piix4.c @@ -289,6 +289,9 @@ static int piix4_setup_sb800(struct pci_dev *PIIX4_dev, PIIX4_dev->revision >= 0x41) || (PIIX4_dev->vendor == PCI_VENDOR_ID_AMD && PIIX4_dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS &&
PIIX4_dev->revision >= 0x49) ||
(PIIX4_dev->vendor == PCI_VENDOR_ID_HYGON &&
smb_en = 0x00;PIIX4_dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS && PIIX4_dev->revision >= 0x49))
Does the compatibility with the original AMD chipset include the revision number?
else @@ -361,7 +364,8 @@ static int piix4_setup_sb800(struct pci_dev *PIIX4_dev, piix4_smba, i2ccfg >> 4); /* Find which register is used for port selection */
- if (PIIX4_dev->vendor == PCI_VENDOR_ID_AMD) {
- if (PIIX4_dev->vendor == PCI_VENDOR_ID_AMD ||
switch (PIIX4_dev->device) { case PCI_DEVICE_ID_AMD_KERNCZ_SMBUS: piix4_port_sel_sb800 = SB800_PIIX4_PORT_IDX_KERNCZ;PIIX4_dev->vendor == PCI_VENDOR_ID_HYGON) {
@@ -794,6 +798,7 @@ static const struct pci_device_id piix4_ids[] = { { PCI_DEVICE(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_SBX00_SMBUS) }, { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_HUDSON2_SMBUS) }, { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) },
- { PCI_DEVICE(PCI_VENDOR_ID_HYGON, PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) }, { PCI_DEVICE(PCI_VENDOR_ID_SERVERWORKS, PCI_DEVICE_ID_SERVERWORKS_OSB4) }, { PCI_DEVICE(PCI_VENDOR_ID_SERVERWORKS,
@@ -904,11 +909,13 @@ static int piix4_probe(struct pci_dev *dev, const struct pci_device_id *id) if ((dev->vendor == PCI_VENDOR_ID_ATI && dev->device == PCI_DEVICE_ID_ATI_SBX00_SMBUS && dev->revision >= 0x40) ||
dev->vendor == PCI_VENDOR_ID_AMD) {
dev->vendor == PCI_VENDOR_ID_AMD ||
bool notify_imc = false; is_sb800 = true;dev->vendor == PCI_VENDOR_ID_HYGON) {
if (dev->vendor == PCI_VENDOR_ID_AMD &&
if ((dev->vendor == PCI_VENDOR_ID_AMD ||
dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) { u8 imc;dev->vendor == PCI_VENDOR_ID_HYGON) &&
Patch looks good. I assume you have tested it on real hardware?
I would like you to also document the new supported chipset in drivers/i2c/busses/Kconfig and Documentation/i2c/busses/i2c-piix4 as well as in the header comment of i2c-piix4.c itself. I know it seems redundant but it helps the user know which driver they need.
Thanks,
On 2019/4/26 17:38, Jean Delvare wrote:
Sorry for the late answer.
Thanks for the reply!
Cc: stable@vger.kernel.org # v5.0+
I object to adding this to stable. It's not fixing any bug, and it's far from being a one-liner. I'd rather let distributions backport it as they see fit.
Okay, will not add this patch to stable.
PIIX4_dev->revision >= 0x49) ||
(PIIX4_dev->vendor == PCI_VENDOR_ID_HYGON &&
smb_en = 0x00;PIIX4_dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS && PIIX4_dev->revision >= 0x49))
Does the compatibility with the original AMD chipset include the revision number?
The revision number may not be needed here, although it indeed larger than 0x49 on Hygon platform.
if (dev->vendor == PCI_VENDOR_ID_AMD &&
if ((dev->vendor == PCI_VENDOR_ID_AMD ||
dev->device == PCI_DEVICE_ID_AMD_KERNCZ_SMBUS) { u8 imc;dev->vendor == PCI_VENDOR_ID_HYGON) &&
Patch looks good. I assume you have tested it on real hardware?
Yes, I have tested it on Hygon hardware. It works well.
I would like you to also document the new supported chipset in drivers/i2c/busses/Kconfig and Documentation/i2c/busses/i2c-piix4 as well as in the header comment of i2c-piix4.c itself. I know it seems redundant but it helps the user know which driver they need.
Because Hygon uses the same PCI device ID of AMD's, so is it appropriate to document with the name "Hygon CZ" or just "Hygon"?
On Fri, 2019-04-26 at 22:23 +0800, Pu Wen wrote:
On 2019/4/26 17:38, Jean Delvare wrote:
I would like you to also document the new supported chipset in drivers/i2c/busses/Kconfig and Documentation/i2c/busses/i2c-piix4 as well as in the header comment of i2c-piix4.c itself. I know it seems redundant but it helps the user know which driver they need.
Because Hygon uses the same PCI device ID of AMD's, so is it appropriate to document with the name "Hygon CZ" or just "Hygon"?
"Hygon CZ" please, as Hygon is the vendor not the device, and I can imagine you may create more devices in the future.
Thanks,
On 2019/4/27 0:48, Jean Delvare wrote:
On Fri, 2019-04-26 at 22:23 +0800, Pu Wen wrote:
On 2019/4/26 17:38, Jean Delvare wrote:
I would like you to also document the new supported chipset in drivers/i2c/busses/Kconfig and Documentation/i2c/busses/i2c-piix4 as well as in the header comment of i2c-piix4.c itself. I know it seems redundant but it helps the user know which driver they need.
Because Hygon uses the same PCI device ID of AMD's, so is it appropriate to document with the name "Hygon CZ" or just "Hygon"?
"Hygon CZ" please, as Hygon is the vendor not the device, and I can
Okay.
imagine you may create more devices in the future.
Surely.
linux-stable-mirror@lists.linaro.org