diff mbox series

[v6] ACPI: skip IRQ override on AMD Zen platforms

Message ID 20220712020058.90374-1-gch981213@gmail.com
State Accepted
Commit 9946e39fe8d0a5da9eb947d8e40a7ef204ba016e
Headers show
Series [v6] ACPI: skip IRQ override on AMD Zen platforms | expand

Commit Message

Chuanhong Guo July 12, 2022, 2 a.m. UTC 
IRQ override isn't needed on modern AMD Zen systems.
There's an active low keyboard IRQ on AMD Ryzen 6000 and it will stay
this way on newer platforms. This IRQ override breaks keyboards for
almost all Ryzen 6000 laptops currently on the market.

Skip this IRQ override for all AMD Zen platforms because this IRQ
override is supposed to be a workaround for buggy ACPI DSDT and we can't
have a long list of all future AMD CPUs/Laptops in the kernel code.
If a device with buggy ACPI DSDT shows up, a separated list containing
just them should be created.

Suggested-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
---
Change sice v5: reworked

 drivers/acpi/resource.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

Comments

Mario Limonciello July 14, 2022, 7:59 p.m. UTC | #1 
[Public]



> -----Original Message-----
> From: Rafael J. Wysocki <rafael@kernel.org>
> Sent: Thursday, July 14, 2022 13:36
> To: Chuanhong Guo <gch981213@gmail.com>
> Cc: ACPI Devel Maling List <linux-acpi@vger.kernel.org>; Tighe Donnelly
> <tighe.donnelly@protonmail.com>; Limonciello, Mario
> <Mario.Limonciello@amd.com>; Rafael J. Wysocki <rafael@kernel.org>; Len
> Brown <lenb@kernel.org>; open list <linux-kernel@vger.kernel.org>
> Subject: Re: [PATCH v6] ACPI: skip IRQ override on AMD Zen platforms
> 
> On Tue, Jul 12, 2022 at 4:01 AM Chuanhong Guo <gch981213@gmail.com>
> wrote:
> >
> > IRQ override isn't needed on modern AMD Zen systems.
> > There's an active low keyboard IRQ on AMD Ryzen 6000 and it will stay
> > this way on newer platforms. This IRQ override breaks keyboards for
> > almost all Ryzen 6000 laptops currently on the market.
> >
> > Skip this IRQ override for all AMD Zen platforms because this IRQ
> > override is supposed to be a workaround for buggy ACPI DSDT and we can't
> > have a long list of all future AMD CPUs/Laptops in the kernel code.
> > If a device with buggy ACPI DSDT shows up, a separated list containing
> > just them should be created.
> >
> > Suggested-by: Mario Limonciello <mario.limonciello@amd.com>
> > Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
> 
> I need an ACK from Mario here.

I believe this is fine, but give me a few days to check with others.
This isn't 5.19 material, and so I think an ack next week should be fine for it.

> 
> > ---
> > Change sice v5: reworked
> >
> >  drivers/acpi/resource.c | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> >
> > diff --git a/drivers/acpi/resource.c b/drivers/acpi/resource.c
> > index c2d494784425..510cdec375c4 100644
> > --- a/drivers/acpi/resource.c
> > +++ b/drivers/acpi/resource.c
> > @@ -416,6 +416,16 @@ static bool acpi_dev_irq_override(u32 gsi, u8
> triggering, u8 polarity,
> >  {
> >         int i;
> >
> > +#ifdef CONFIG_X86
> > +       /*
> > +        * IRQ override isn't needed on modern AMD Zen systems and
> > +        * this override breaks active low IRQs on AMD Ryzen 6000 and
> > +        * newer systems. Skip it.
> > +        */
> > +       if (boot_cpu_has(X86_FEATURE_ZEN))
> > +               return false;
> > +#endif
> > +
> >         for (i = 0; i < ARRAY_SIZE(skip_override_table); i++) {
> >                 const struct irq_override_cmp *entry = &skip_override_table[i];
> >
> > --
> > 2.36.1
> >
Jiri Slaby Sept. 28, 2022, 6:20 a.m. UTC | #2 
Hi,

On 12. 07. 22, 4:00, Chuanhong Guo wrote:
> IRQ override isn't needed on modern AMD Zen systems.
> There's an active low keyboard IRQ on AMD Ryzen 6000 and it will stay
> this way on newer platforms. This IRQ override breaks keyboards for
> almost all Ryzen 6000 laptops currently on the market.
> 
> Skip this IRQ override for all AMD Zen platforms because this IRQ
> override is supposed to be a workaround for buggy ACPI DSDT and we can't
> have a long list of all future AMD CPUs/Laptops in the kernel code.
> If a device with buggy ACPI DSDT shows up, a separated list containing
> just them should be created.

This breaks pads on IdeaPad 5 Flex:
https://bugzilla.suse.com/show_bug.cgi?id=1203794

 > [    1.058135] hid-generic 0020:1022:0001.0001: hidraw0: SENSOR HUB 
HID v0.00 Device [hid-amdsfh 1022:0001] on pcie_mp2_amd
 > [    2.038937] i2c_designware AMDI0010:00: controller timed out
 > [    2.146627] i2c_designware AMDI0010:03: controller timed out
 > [    6.166859] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
 > [    8.279604] i2c_designware AMDI0010:03: controller timed out
 > [   12.310897] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
 > [   14.429372] i2c_designware AMDI0010:03: controller timed out
 > [   18.462629] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
 > [   20.579183] i2c_designware AMDI0010:03: controller timed out
 > [   24.598703] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
 > [   25.629071] i2c_hid_acpi i2c-MSFT0001:00: can't add hid device: -61
 > [   25.629430] i2c_hid_acpi: probe of i2c-MSFT0001:00 failed with 
error -61

The diff of good and bad dmesgs:
-ACPI: IRQ 10 override to edge, high
-ACPI: IRQ 6 override to edge, high

The diff of /proc/interrupts:
      6: ...  IR-IO-APIC    [-6-fasteoi-]    {+6-edge+}      AMDI0010:03
     10: ...  IR-IO-APIC   [-10-fasteoi-]   {+10-edge+}      AMDI0010:00

And:
   i2c_designware: /devices/platform/AMDI0010:00
   i2c_designware: /devices/platform/AMDI0010:03


So the if needs to be fine-tuned, apparently. Maybe introduce some list 
as suggested in the commit log. Based on the below?

DMI says:
   System Info: #14
     Manufacturer: "LENOVO"
     Product: "82RA"
     Version: "IdeaPad Flex 5 16ALC7"
     Serial: "PW02359K"
     UUID: 6b2d54d9-cd80-11ec-83eb-e00af665fbac
     Wake-up: 0x06 (Power Switch)
   Board Info: #15
     Manufacturer: "LENOVO"
     Product: "LNVNB161216"
     Version: "SDK0T76463 WIN"
     Serial: "PW02359K"
     Asset Tag: "No Asset Tag"
     Type: 0x0a (Motherboard)
     Features: 0x09
       Hosting Board
       Replaceable
     Location: "Chassis Location Unknown"
   Chassis Info: #16
     Manufacturer: "LENOVO"
     Version: "IdeaPad Flex 5 16ALC7"
     Serial: "PW02359K"
     Asset Tag: "No Asset Tag"
     Type: 0x1f (Other)
     Bootup State: 0x03 (Safe)
     Power Supply State: 0x03 (Safe)
     Thermal State: 0x01 (Other)
     Security Status: 0x01 (Other)

> Suggested-by: Mario Limonciello <mario.limonciello@amd.com>
> Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
> ---
> Change sice v5: reworked
> 
>   drivers/acpi/resource.c | 10 ++++++++++
>   1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/acpi/resource.c b/drivers/acpi/resource.c
> index c2d494784425..510cdec375c4 100644
> --- a/drivers/acpi/resource.c
> +++ b/drivers/acpi/resource.c
> @@ -416,6 +416,16 @@ static bool acpi_dev_irq_override(u32 gsi, u8 triggering, u8 polarity,
>   {
>   	int i;
>   
> +#ifdef CONFIG_X86
> +	/*
> +	 * IRQ override isn't needed on modern AMD Zen systems and
> +	 * this override breaks active low IRQs on AMD Ryzen 6000 and
> +	 * newer systems. Skip it.
> +	 */
> +	if (boot_cpu_has(X86_FEATURE_ZEN))
> +		return false;
> +#endif
> +
>   	for (i = 0; i < ARRAY_SIZE(skip_override_table); i++) {
>   		const struct irq_override_cmp *entry = &skip_override_table[i];
>   

thanks,
Jiri Slaby Sept. 28, 2022, 8:31 a.m. UTC | #3 
On 28. 09. 22, 8:20, Jiri Slaby wrote:
> Hi,
> 
> On 12. 07. 22, 4:00, Chuanhong Guo wrote:
>> IRQ override isn't needed on modern AMD Zen systems.
>> There's an active low keyboard IRQ on AMD Ryzen 6000 and it will stay
>> this way on newer platforms. This IRQ override breaks keyboards for
>> almost all Ryzen 6000 laptops currently on the market.
>>
>> Skip this IRQ override for all AMD Zen platforms because this IRQ
>> override is supposed to be a workaround for buggy ACPI DSDT and we can't
>> have a long list of all future AMD CPUs/Laptops in the kernel code.
>> If a device with buggy ACPI DSDT shows up, a separated list containing
>> just them should be created.
> 
> This breaks pads on IdeaPad 5 Flex:
> https://bugzilla.suse.com/show_bug.cgi?id=1203794
> 
>  > [    1.058135] hid-generic 0020:1022:0001.0001: hidraw0: SENSOR HUB 
> HID v0.00 Device [hid-amdsfh 1022:0001] on pcie_mp2_amd
>  > [    2.038937] i2c_designware AMDI0010:00: controller timed out
>  > [    2.146627] i2c_designware AMDI0010:03: controller timed out
>  > [    6.166859] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
>  > [    8.279604] i2c_designware AMDI0010:03: controller timed out
>  > [   12.310897] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
>  > [   14.429372] i2c_designware AMDI0010:03: controller timed out
>  > [   18.462629] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
>  > [   20.579183] i2c_designware AMDI0010:03: controller timed out
>  > [   24.598703] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
>  > [   25.629071] i2c_hid_acpi i2c-MSFT0001:00: can't add hid device: -61
>  > [   25.629430] i2c_hid_acpi: probe of i2c-MSFT0001:00 failed with 
> error -61
> 
> The diff of good and bad dmesgs:
> -ACPI: IRQ 10 override to edge, high
> -ACPI: IRQ 6 override to edge, high
> 
> The diff of /proc/interrupts:
>       6: ...  IR-IO-APIC    [-6-fasteoi-]    {+6-edge+}      AMDI0010:03
>      10: ...  IR-IO-APIC   [-10-fasteoi-]   {+10-edge+}      AMDI0010:00
> 
> And:
>    i2c_designware: /devices/platform/AMDI0010:00
>    i2c_designware: /devices/platform/AMDI0010:03
> 
> 
> So the if needs to be fine-tuned, apparently. Maybe introduce some list 
> as suggested in the commit log. Based on the below?

Something like the attached. It's:
1) untested yet
2) contains more debug messaging
3) contains both cases for ACPI_ACTIVE_* as I don't know the original 
polarity

I don't know how widely this is spread -- maybe it would be worth a 
commandline parameter so that people can work around this until this is 
fixed by a DMI entry permanently?

> DMI says:
>    System Info: #14
>      Manufacturer: "LENOVO"
>      Product: "82RA"
>      Version: "IdeaPad Flex 5 16ALC7"
>      Serial: "PW02359K"
>      UUID: 6b2d54d9-cd80-11ec-83eb-e00af665fbac
>      Wake-up: 0x06 (Power Switch)
>    Board Info: #15
>      Manufacturer: "LENOVO"
>      Product: "LNVNB161216"
>      Version: "SDK0T76463 WIN"
>      Serial: "PW02359K"
>      Asset Tag: "No Asset Tag"
>      Type: 0x0a (Motherboard)
>      Features: 0x09
>        Hosting Board
>        Replaceable
>      Location: "Chassis Location Unknown"
>    Chassis Info: #16
>      Manufacturer: "LENOVO"
>      Version: "IdeaPad Flex 5 16ALC7"
>      Serial: "PW02359K"
>      Asset Tag: "No Asset Tag"
>      Type: 0x1f (Other)
>      Bootup State: 0x03 (Safe)
>      Power Supply State: 0x03 (Safe)
>      Thermal State: 0x01 (Other)
>      Security Status: 0x01 (Other)
> 
>> Suggested-by: Mario Limonciello <mario.limonciello@amd.com>
>> Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
>> ---
>> Change sice v5: reworked
>>
>>   drivers/acpi/resource.c | 10 ++++++++++
>>   1 file changed, 10 insertions(+)
>>
>> diff --git a/drivers/acpi/resource.c b/drivers/acpi/resource.c
>> index c2d494784425..510cdec375c4 100644
>> --- a/drivers/acpi/resource.c
>> +++ b/drivers/acpi/resource.c
>> @@ -416,6 +416,16 @@ static bool acpi_dev_irq_override(u32 gsi, u8 
>> triggering, u8 polarity,
>>   {
>>       int i;
>> +#ifdef CONFIG_X86
>> +    /*
>> +     * IRQ override isn't needed on modern AMD Zen systems and
>> +     * this override breaks active low IRQs on AMD Ryzen 6000 and
>> +     * newer systems. Skip it.
>> +     */
>> +    if (boot_cpu_has(X86_FEATURE_ZEN))
>> +        return false;
>> +#endif
>> +
>>       for (i = 0; i < ARRAY_SIZE(skip_override_table); i++) {
>>           const struct irq_override_cmp *entry = &skip_override_table[i];
> 
> thanks,
Jiri Slaby Sept. 28, 2022, 8:57 a.m. UTC | #4 
On 28. 09. 22, 10:31, Jiri Slaby wrote:
> Something like the attached. It's:
> 1) untested yet
> 2) contains more debug messaging
> 3) contains both cases for ACPI_ACTIVE_* as I don't know the original 
> polarity

Now, I have DSDT which says for both of them:
                 IRQ (Level, ActiveLow, ExclusiveAndWake, )

So the patch should actually have:
         { lenovo_laptop, 6, ACPI_LEVEL_SENSITIVE, ACPI_ACTIVE_LOW, 0, 
true },
         { lenovo_laptop, 10, ACPI_LEVEL_SENSITIVE, ACPI_ACTIVE_LOW, 0, 
true },

thanks,
Chuanhong Guo Sept. 30, 2022, 2:45 a.m. UTC | #5 
Hi!

On Wed, Sep 28, 2022 at 4:31 PM Jiri Slaby <jirislaby@kernel.org> wrote:
> > This breaks pads on IdeaPad 5 Flex:
> > https://bugzilla.suse.com/show_bug.cgi?id=1203794
> >
> >  > [    1.058135] hid-generic 0020:1022:0001.0001: hidraw0: SENSOR HUB
> > HID v0.00 Device [hid-amdsfh 1022:0001] on pcie_mp2_amd
> >  > [    2.038937] i2c_designware AMDI0010:00: controller timed out
> >  > [    2.146627] i2c_designware AMDI0010:03: controller timed out
> >  > [    6.166859] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
> >  > [    8.279604] i2c_designware AMDI0010:03: controller timed out
> >  > [   12.310897] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
> >  > [   14.429372] i2c_designware AMDI0010:03: controller timed out
> >  > [   18.462629] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
> >  > [   20.579183] i2c_designware AMDI0010:03: controller timed out
> >  > [   24.598703] i2c_hid_acpi i2c-MSFT0001:00: failed to reset device: -61
> >  > [   25.629071] i2c_hid_acpi i2c-MSFT0001:00: can't add hid device: -61
> >  > [   25.629430] i2c_hid_acpi: probe of i2c-MSFT0001:00 failed with
> > error -61
> >
> > The diff of good and bad dmesgs:
> > -ACPI: IRQ 10 override to edge, high
> > -ACPI: IRQ 6 override to edge, high
> >
> > The diff of /proc/interrupts:
> >       6: ...  IR-IO-APIC    [-6-fasteoi-]    {+6-edge+}      AMDI0010:03
> >      10: ...  IR-IO-APIC   [-10-fasteoi-]   {+10-edge+}      AMDI0010:00
> >
> > And:
> >    i2c_designware: /devices/platform/AMDI0010:00
> >    i2c_designware: /devices/platform/AMDI0010:03

Oops...

> > So the if needs to be fine-tuned, apparently. Maybe introduce some list
> > as suggested in the commit log. Based on the below?
>
> Something like the attached. It's:
> 1) untested yet
> 2) contains more debug messaging
> 3) contains both cases for ACPI_ACTIVE_* as I don't know the original
> polarity

The patch in your attachment looks good to me. But I think
"lenovo_laptop" is a bit too generic. Maybe name it
lenovo_82ra instead?

>
> I don't know how widely this is spread -- maybe it would be worth a
> commandline parameter so that people can work around this until this is
> fixed by a DMI entry permanently?

That's a good idea :)

--
Regards,
Chuanhong Guo
diff mbox series

Patch

diff --git a/drivers/acpi/resource.c b/drivers/acpi/resource.c
index c2d494784425..510cdec375c4 100644
--- a/drivers/acpi/resource.c
+++ b/drivers/acpi/resource.c
@@ -416,6 +416,16 @@  static bool acpi_dev_irq_override(u32 gsi, u8 triggering, u8 polarity,
 {
 	int i;
 
+#ifdef CONFIG_X86
+	/*
+	 * IRQ override isn't needed on modern AMD Zen systems and
+	 * this override breaks active low IRQs on AMD Ryzen 6000 and
+	 * newer systems. Skip it.
+	 */
+	if (boot_cpu_has(X86_FEATURE_ZEN))
+		return false;
+#endif
+
 	for (i = 0; i < ARRAY_SIZE(skip_override_table); i++) {
 		const struct irq_override_cmp *entry = &skip_override_table[i];