2021-03-29 19:55:01

by Angela Czubak

Subject: [PATCH] resource: Prevent irqresource_disabled() from erasing flags

Do not overwrite flags as it leads to erasing triggering and polarity
information which might be useful in case of hard-coded interrupts.
This way the information can be read later on even though mapping to
APIC domain failed.

Signed-off-by: Angela Czubak <[email protected]>
---
Some Chromebooks use hard-coded interrupts in their ACPI tables.
This is an excerpt as dumped on Relm:

...
            Name (_HID, "ELAN0001")  // _HID: Hardware ID
            Name (_DDN, "Elan Touchscreen ")  // _DDN: DOS Device Name
            Name (_UID, 0x05)  // _UID: Unique ID
            Name (ISTP, Zero)
            Method (_CRS, 0, NotSerialized)  // _CRS: Current Resource Settings
            {
                Name (BUF0, ResourceTemplate ()
                {
                    I2cSerialBusV2 (0x0010, ControllerInitiated, 0x00061A80,
                        AddressingMode7Bit, "\\_SB.I2C1",
                        0x00, ResourceConsumer, , Exclusive,
                        )
                    Interrupt (ResourceConsumer, Edge, ActiveLow, Exclusive, ,, )
                    {
                        0x000000B8,
                    }
                })
                Return (BUF0) /* \_SB_.I2C1.ETSA._CRS.BUF0 */
            }
...

This interrupt is hard-coded to 0xB8 = 184 which is too high to be mapped
to IO-APIC, so no triggering information is propagated as acpi_register_gsi()
fails and irqresource_disabled() is issued, which leads to erasing triggering
and polarity information.
If that function added its flags instead of overwriting them the correct IRQ
type would be set even for the hard-coded interrupts, which allows device driver
to retrieve it.
Please, let me know if this kind of modification is acceptable.
Best Regards,
Angela

include/linux/ioport.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/ioport.h b/include/linux/ioport.h
index 55de385c839cf..647744d8514e0 100644
--- a/include/linux/ioport.h
+++ b/include/linux/ioport.h
@@ -331,7 +331,7 @@ static inline void irqresource_disabled(struct resource *res, u32 irq)
{
res->start = irq;
res->end = irq;
- res->flags = IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
+ res->flags |= IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
}

extern struct address_space *iomem_get_mapping(void);
--
2.17.1

2021-03-30 15:14:29

by Wysocki, Rafael J

Subject: Re: [PATCH] resource: Prevent irqresource_disabled() from erasing flags

On 3/29/2021 9:52 PM, Angela Czubak wrote:
> Do not overwrite flags as it leads to erasing triggering and polarity
> information which might be useful in case of hard-coded interrupts.
> This way the information can be read later on even though mapping to
> APIC domain failed.
>
> Signed-off-by: Angela Czubak <[email protected]>
> ---
> Some Chromebooks use hard-coded interrupts in their ACPI tables.
> This is an excerpt as dumped on Relm:
>
> ...
>             Name (_HID, "ELAN0001")  // _HID: Hardware ID
>             Name (_DDN, "Elan Touchscreen ")  // _DDN: DOS Device Name
>             Name (_UID, 0x05)  // _UID: Unique ID
>             Name (ISTP, Zero)
>             Method (_CRS, 0, NotSerialized)  // _CRS: Current Resource Settings
>             {
>                 Name (BUF0, ResourceTemplate ()
>                 {
>                     I2cSerialBusV2 (0x0010, ControllerInitiated, 0x00061A80,
>                         AddressingMode7Bit, "\\_SB.I2C1",
>                         0x00, ResourceConsumer, , Exclusive,
>                         )
>                     Interrupt (ResourceConsumer, Edge, ActiveLow, Exclusive, ,, )
>                     {
>                         0x000000B8,
>                     }
>                 })
>                 Return (BUF0) /* \_SB_.I2C1.ETSA._CRS.BUF0 */
>             }
> ...
>
> This interrupt is hard-coded to 0xB8 = 184 which is too high to be mapped
> to IO-APIC, so no triggering information is propagated as acpi_register_gsi()
> fails and irqresource_disabled() is issued, which leads to erasing triggering
> and polarity information.
> If that function added its flags instead of overwriting them the correct IRQ
> type would be set even for the hard-coded interrupts, which allows device driver
> to retrieve it.
> Please, let me know if this kind of modification is acceptable.

From the quick look it should not be problematic, but it needs to be
checked more carefully.

Mika, what do you think?


> include/linux/ioport.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/include/linux/ioport.h b/include/linux/ioport.h
> index 55de385c839cf..647744d8514e0 100644
> --- a/include/linux/ioport.h
> +++ b/include/linux/ioport.h
> @@ -331,7 +331,7 @@ static inline void irqresource_disabled(struct resource *res, u32 irq)
> {
> res->start = irq;
> res->end = irq;
> - res->flags = IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> + res->flags |= IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> }
>
> extern struct address_space *iomem_get_mapping(void);

2021-03-30 15:54:09

by Mika Westerberg

Subject: Re: [PATCH] resource: Prevent irqresource_disabled() from erasing flags

Hi,

On Tue, Mar 30, 2021 at 05:09:42PM +0200, Rafael J. Wysocki wrote:
> On 3/29/2021 9:52 PM, Angela Czubak wrote:
> > Do not overwrite flags as it leads to erasing triggering and polarity
> > information which might be useful in case of hard-coded interrupts.
> > This way the information can be read later on even though mapping to
> > APIC domain failed.
> >
> > Signed-off-by: Angela Czubak <[email protected]>
> > ---
> > Some Chromebooks use hard-coded interrupts in their ACPI tables.
> > This is an excerpt as dumped on Relm:
> >
> > ...
> > ? ? ? ? ? ? Name (_HID, "ELAN0001") ?// _HID: Hardware ID
> > ? ? ? ? ? ? Name (_DDN, "Elan Touchscreen ") ?// _DDN: DOS Device Name
> > ? ? ? ? ? ? Name (_UID, 0x05) ?// _UID: Unique ID
> > ? ? ? ? ? ? Name (ISTP, Zero)
> > ? ? ? ? ? ? Method (_CRS, 0, NotSerialized) ?// _CRS: Current Resource Settings
> > ? ? ? ? ? ? {
> > ? ? ? ? ? ? ? ? Name (BUF0, ResourceTemplate ()
> > ? ? ? ? ? ? ? ? {
> > ? ? ? ? ? ? ? ? ? ? I2cSerialBusV2 (0x0010, ControllerInitiated, 0x00061A80,
> > ? ? ? ? ? ? ? ? ? ? ? ? AddressingMode7Bit, "\\_SB.I2C1",
> > ? ? ? ? ? ? ? ? ? ? ? ? 0x00, ResourceConsumer, , Exclusive,
> > ? ? ? ? ? ? ? ? ? ? ? ? )
> > ? ? ? ? ? ? ? ? ? ? Interrupt (ResourceConsumer, Edge, ActiveLow, Exclusive, ,, )
> > ? ? ? ? ? ? ? ? ? ? {
> > ? ? ? ? ? ? ? ? ? ? ? ? 0x000000B8,
> > ? ? ? ? ? ? ? ? ? ? }
> > ? ? ? ? ? ? ? ? })
> > ? ? ? ? ? ? ? ? Return (BUF0) /* \_SB_.I2C1.ETSA._CRS.BUF0 */
> > ? ? ? ? ? ? }
> > ...
> >
> > This interrupt is hard-coded to 0xB8 = 184 which is too high to be mapped
> > to IO-APIC, so no triggering information is propagated as acpi_register_gsi()
> > fails and irqresource_disabled() is issued, which leads to erasing triggering
> > and polarity information.
> > If that function added its flags instead of overwriting them the correct IRQ
> > type would be set even for the hard-coded interrupts, which allows device driver
> > to retrieve it.
> > Please, let me know if this kind of modification is acceptable.
>
> From the quick look it should not be problematic, but it needs to be checked
> more carefully.
>
> Mika, what do you think?

I think it makes sense. We still set IORESOURCE_DISABLED unconditionally
so this should not cause issues. In theory at least :)

> > include/linux/ioport.h | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/include/linux/ioport.h b/include/linux/ioport.h
> > index 55de385c839cf..647744d8514e0 100644
> > --- a/include/linux/ioport.h
> > +++ b/include/linux/ioport.h
> > @@ -331,7 +331,7 @@ static inline void irqresource_disabled(struct resource *res, u32 irq)
> > {
> > res->start = irq;
> > res->end = irq;
> > - res->flags = IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> > + res->flags |= IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> > }
> > extern struct address_space *iomem_get_mapping(void);
>

2021-04-07 20:54:10

by Angela Czubak

Subject: Re: [PATCH] resource: Prevent irqresource_disabled() from erasing flags

On Tue, Mar 30, 2021 at 5:50 PM Mika Westerberg
<[email protected]> wrote:
>
> Hi,
>
> On Tue, Mar 30, 2021 at 05:09:42PM +0200, Rafael J. Wysocki wrote:
> > On 3/29/2021 9:52 PM, Angela Czubak wrote:
> > > Do not overwrite flags as it leads to erasing triggering and polarity
> > > information which might be useful in case of hard-coded interrupts.
> > > This way the information can be read later on even though mapping to
> > > APIC domain failed.
> > >
> > > Signed-off-by: Angela Czubak <[email protected]>
> > > ---
> > > Some Chromebooks use hard-coded interrupts in their ACPI tables.
> > > This is an excerpt as dumped on Relm:
> > >
> > > ...
> > > Name (_HID, "ELAN0001") // _HID: Hardware ID
> > > Name (_DDN, "Elan Touchscreen ") // _DDN: DOS Device Name
> > > Name (_UID, 0x05) // _UID: Unique ID
> > > Name (ISTP, Zero)
> > > Method (_CRS, 0, NotSerialized) // _CRS: Current Resource Settings
> > > {
> > > Name (BUF0, ResourceTemplate ()
> > > {
> > > I2cSerialBusV2 (0x0010, ControllerInitiated, 0x00061A80,
> > > AddressingMode7Bit, "\\_SB.I2C1",
> > > 0x00, ResourceConsumer, , Exclusive,
> > > )
> > > Interrupt (ResourceConsumer, Edge, ActiveLow, Exclusive, ,, )
> > > {
> > > 0x000000B8,
> > > }
> > > })
> > > Return (BUF0) /* \_SB_.I2C1.ETSA._CRS.BUF0 */
> > > }
> > > ...
> > >
> > > This interrupt is hard-coded to 0xB8 = 184 which is too high to be mapped
> > > to IO-APIC, so no triggering information is propagated as acpi_register_gsi()
> > > fails and irqresource_disabled() is issued, which leads to erasing triggering
> > > and polarity information.
> > > If that function added its flags instead of overwriting them the correct IRQ
> > > type would be set even for the hard-coded interrupts, which allows device driver
> > > to retrieve it.
> > > Please, let me know if this kind of modification is acceptable.
> >
> > From the quick look it should not be problematic, but it needs to be checked
> > more carefully.
> >
> > Mika, what do you think?
>
> I think it makes sense. We still set IORESOURCE_DISABLED unconditionally
> so this should not cause issues. In theory at least :)
>
Is there anything else you would need me to do regarding the patch?
I suppose there are more platforms that could benefit from not erasing
the flags, so if this patch is fit for upstream, can we continue the
process?

> > > include/linux/ioport.h | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/include/linux/ioport.h b/include/linux/ioport.h
> > > index 55de385c839cf..647744d8514e0 100644
> > > --- a/include/linux/ioport.h
> > > +++ b/include/linux/ioport.h
> > > @@ -331,7 +331,7 @@ static inline void irqresource_disabled(struct resource *res, u32 irq)
> > > {
> > > res->start = irq;
> > > res->end = irq;
> > > - res->flags = IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> > > + res->flags |= IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> > > }
> > > extern struct address_space *iomem_get_mapping(void);
> >

2021-04-07 20:59:20

by Rafael J. Wysocki

Subject: Re: [PATCH] resource: Prevent irqresource_disabled() from erasing flags

On Wed, Apr 7, 2021 at 1:01 PM Angela Czubak <[email protected]> wrote:
>
> On Tue, Mar 30, 2021 at 5:50 PM Mika Westerberg
> <[email protected]> wrote:
> >
> > Hi,
> >
> > On Tue, Mar 30, 2021 at 05:09:42PM +0200, Rafael J. Wysocki wrote:
> > > On 3/29/2021 9:52 PM, Angela Czubak wrote:
> > > > Do not overwrite flags as it leads to erasing triggering and polarity
> > > > information which might be useful in case of hard-coded interrupts.
> > > > This way the information can be read later on even though mapping to
> > > > APIC domain failed.
> > > >
> > > > Signed-off-by: Angela Czubak <[email protected]>
> > > > ---
> > > > Some Chromebooks use hard-coded interrupts in their ACPI tables.
> > > > This is an excerpt as dumped on Relm:
> > > >
> > > > ...
> > > > Name (_HID, "ELAN0001") // _HID: Hardware ID
> > > > Name (_DDN, "Elan Touchscreen ") // _DDN: DOS Device Name
> > > > Name (_UID, 0x05) // _UID: Unique ID
> > > > Name (ISTP, Zero)
> > > > Method (_CRS, 0, NotSerialized) // _CRS: Current Resource Settings
> > > > {
> > > > Name (BUF0, ResourceTemplate ()
> > > > {
> > > > I2cSerialBusV2 (0x0010, ControllerInitiated, 0x00061A80,
> > > > AddressingMode7Bit, "\\_SB.I2C1",
> > > > 0x00, ResourceConsumer, , Exclusive,
> > > > )
> > > > Interrupt (ResourceConsumer, Edge, ActiveLow, Exclusive, ,, )
> > > > {
> > > > 0x000000B8,
> > > > }
> > > > })
> > > > Return (BUF0) /* \_SB_.I2C1.ETSA._CRS.BUF0 */
> > > > }
> > > > ...
> > > >
> > > > This interrupt is hard-coded to 0xB8 = 184 which is too high to be mapped
> > > > to IO-APIC, so no triggering information is propagated as acpi_register_gsi()
> > > > fails and irqresource_disabled() is issued, which leads to erasing triggering
> > > > and polarity information.
> > > > If that function added its flags instead of overwriting them the correct IRQ
> > > > type would be set even for the hard-coded interrupts, which allows device driver
> > > > to retrieve it.
> > > > Please, let me know if this kind of modification is acceptable.
> > >
> > > From the quick look it should not be problematic, but it needs to be checked
> > > more carefully.
> > >
> > > Mika, what do you think?
> >
> > I think it makes sense. We still set IORESOURCE_DISABLED unconditionally
> > so this should not cause issues. In theory at least :)
> >
> Is there anything else you would need me to do regarding the patch?

Yes, please resend it with a CC to [email protected] for more
visibility.

> I suppose there are more platforms that could benefit from not erasing
> the flags, so if this patch is fit for upstream, can we continue the
> process?
>
> > > > include/linux/ioport.h | 2 +-
> > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/include/linux/ioport.h b/include/linux/ioport.h
> > > > index 55de385c839cf..647744d8514e0 100644
> > > > --- a/include/linux/ioport.h
> > > > +++ b/include/linux/ioport.h
> > > > @@ -331,7 +331,7 @@ static inline void irqresource_disabled(struct resource *res, u32 irq)
> > > > {
> > > > res->start = irq;
> > > > res->end = irq;
> > > > - res->flags = IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> > > > + res->flags |= IORESOURCE_IRQ | IORESOURCE_DISABLED | IORESOURCE_UNSET;
> > > > }
> > > > extern struct address_space *iomem_get_mapping(void);
> > >