| Message ID | 1467188859-28188-2-git-send-email-okaya@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Wed, Jun 29, 2016 at 10:27 AM, Sinan Kaya <okaya@codeaurora.org> wrote: > The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce > resource requirements") omitted the initially assigned POSSIBLE penalty > when the IRQ is active. It would be good to say what can go wrong with that here. > The original code would assign the POSSIBLE value divided by the number > of possible IRQs during initialization. > > Later, if the IRQ is chosen as the active IRQ or if the IRQ is in use > by ISA; additional penalties get added. Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 6/29/2016 9:13 AM, Rafael J. Wysocki wrote: > On Wed, Jun 29, 2016 at 10:27 AM, Sinan Kaya <okaya@codeaurora.org> wrote: >> The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce >> resource requirements") omitted the initially assigned POSSIBLE penalty >> when the IRQ is active. > > It would be good to say what can go wrong with that here. > I can add more description. Here is a first attempt. Incorrect calculation of penalty leads to ACPI code assigning the wrong interrupt number to PCI INTx interrupts. This would not be as bad as it sounds in theory. You would just cause the interrupts to be shared and observe performance penalty. However, some drivers like parallel port driver doesn't like interrupt sharing as in this example and causes all other PCI drivers sharing the interrupt to malfunction. The issue has not been caught because the behavior is platform specific and depends on the peripheral drivers sharing the IRQ. I can claim that this could be a BIOS bug. if interrupt 7 is not good for PCI, it shouldn't have been listed in the possible PCI interrupts to begin with. Given this is an existing platform, I don't think we have the luxury to request all BIOS to be updated. This bugfix is needed to support existing platforms. Feel free to request more information if the above description is not clear. >> The original code would assign the POSSIBLE value divided by the number >> of possible IRQs during initialization. >> >> Later, if the IRQ is chosen as the active IRQ or if the IRQ is in use >> by ISA; additional penalties get added. > > Thanks, > Rafael > -- > To unsubscribe from this list: send the line "unsubscribe linux-pci" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >
On Wed, Jun 29, 2016 at 8:47 PM, Sinan Kaya <okaya@codeaurora.org> wrote: > On 6/29/2016 9:13 AM, Rafael J. Wysocki wrote: >> On Wed, Jun 29, 2016 at 10:27 AM, Sinan Kaya <okaya@codeaurora.org> wrote: >>> The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce >>> resource requirements") omitted the initially assigned POSSIBLE penalty >>> when the IRQ is active. >> >> It would be good to say what can go wrong with that here. >> > > I can add more description. Here is a first attempt. > > Incorrect calculation of penalty leads to ACPI code assigning the wrong > interrupt number to PCI INTx interrupts. > > This would not be as bad as it sounds in theory. You would just cause the > interrupts to be shared and observe performance penalty. > > However, some drivers like parallel port driver doesn't like interrupt > sharing as in this example and causes all other PCI drivers sharing the interrupt > to malfunction. > > The issue has not been caught because the behavior is platform specific > and depends on the peripheral drivers sharing the IRQ. > > I can claim that this could be a BIOS bug. if interrupt 7 is not good for PCI, > it shouldn't have been listed in the possible PCI interrupts to begin with. > Given this is an existing platform, I don't think we have the luxury to request > all BIOS to be updated. This bugfix is needed to support existing platforms. > > > Feel free to request more information if the above description is not clear. It is clear enough. I can add it to the changelog when applying the patch. > >>> The original code would assign the POSSIBLE value divided by the number >>> of possible IRQs during initialization. >>> >>> Later, if the IRQ is chosen as the active IRQ or if the IRQ is in use >>> by ISA; additional penalties get added. Does "later" here mean "later in that code path" or "in a later patch"? Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 6/29/2016 5:19 PM, Rafael J. Wysocki wrote: > On Wed, Jun 29, 2016 at 8:47 PM, Sinan Kaya <okaya@codeaurora.org> wrote: >> On 6/29/2016 9:13 AM, Rafael J. Wysocki wrote: >>> On Wed, Jun 29, 2016 at 10:27 AM, Sinan Kaya <okaya@codeaurora.org> wrote: >>>> The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce >>>> resource requirements") omitted the initially assigned POSSIBLE penalty >>>> when the IRQ is active. >>> >>> It would be good to say what can go wrong with that here. >>> >> >> I can add more description. Here is a first attempt. >> >> Incorrect calculation of penalty leads to ACPI code assigning the wrong >> interrupt number to PCI INTx interrupts. >> >> This would not be as bad as it sounds in theory. You would just cause the >> interrupts to be shared and observe performance penalty. >> >> However, some drivers like parallel port driver doesn't like interrupt >> sharing as in this example and causes all other PCI drivers sharing the interrupt >> to malfunction. >> >> The issue has not been caught because the behavior is platform specific >> and depends on the peripheral drivers sharing the IRQ. >> >> I can claim that this could be a BIOS bug. if interrupt 7 is not good for PCI, >> it shouldn't have been listed in the possible PCI interrupts to begin with. >> Given this is an existing platform, I don't think we have the luxury to request >> all BIOS to be updated. This bugfix is needed to support existing platforms. >> >> >> Feel free to request more information if the above description is not clear. > > It is clear enough. I can add it to the changelog when applying the patch. OK > >> >>>> The original code would assign the POSSIBLE value divided by the number >>>> of possible IRQs during initialization. >>>> >>>> Later, if the IRQ is chosen as the active IRQ or if the IRQ is in use >>>> by ISA; additional penalties get added. > > Does "later" here mean "later in that code path" or "in a later patch"? "later in that code path" > > Thanks, > Rafael >
On Wed, Jun 29, 2016 at 11:21 PM, Sinan Kaya <okaya@codeaurora.org> wrote: > On 6/29/2016 5:19 PM, Rafael J. Wysocki wrote: >> On Wed, Jun 29, 2016 at 8:47 PM, Sinan Kaya <okaya@codeaurora.org> wrote: >>> On 6/29/2016 9:13 AM, Rafael J. Wysocki wrote: >>>> On Wed, Jun 29, 2016 at 10:27 AM, Sinan Kaya <okaya@codeaurora.org> wrote: >>>>> The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce >>>>> resource requirements") omitted the initially assigned POSSIBLE penalty >>>>> when the IRQ is active. >>>> >>>> It would be good to say what can go wrong with that here. >>>> >>> >>> I can add more description. Here is a first attempt. >>> >>> Incorrect calculation of penalty leads to ACPI code assigning the wrong >>> interrupt number to PCI INTx interrupts. >>> >>> This would not be as bad as it sounds in theory. You would just cause the >>> interrupts to be shared and observe performance penalty. >>> >>> However, some drivers like parallel port driver doesn't like interrupt >>> sharing as in this example and causes all other PCI drivers sharing the interrupt >>> to malfunction. >>> >>> The issue has not been caught because the behavior is platform specific >>> and depends on the peripheral drivers sharing the IRQ. >>> >>> I can claim that this could be a BIOS bug. if interrupt 7 is not good for PCI, >>> it shouldn't have been listed in the possible PCI interrupts to begin with. >>> Given this is an existing platform, I don't think we have the luxury to request >>> all BIOS to be updated. This bugfix is needed to support existing platforms. >>> >>> >>> Feel free to request more information if the above description is not clear. >> >> It is clear enough. I can add it to the changelog when applying the patch. > > OK > >> >>> >>>>> The original code would assign the POSSIBLE value divided by the number >>>>> of possible IRQs during initialization. >>>>> >>>>> Later, if the IRQ is chosen as the active IRQ or if the IRQ is in use >>>>> by ISA; additional penalties get added. >> >> Does "later" here mean "later in that code path" or "in a later patch"? > > "later in that code path" OK I'm hoping we'll hear from the reporter shortly. Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, Jun 29, 2016 at 11:21 PM, Sinan Kaya <okaya@codeaurora.org> wrote: > On 6/29/2016 5:19 PM, Rafael J. Wysocki wrote: >> On Wed, Jun 29, 2016 at 8:47 PM, Sinan Kaya <okaya@codeaurora.org> wrote: >>> On 6/29/2016 9:13 AM, Rafael J. Wysocki wrote: >>>> On Wed, Jun 29, 2016 at 10:27 AM, Sinan Kaya <okaya@codeaurora.org> wrote: >>>>> The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce >>>>> resource requirements") omitted the initially assigned POSSIBLE penalty >>>>> when the IRQ is active. >>>> >>>> It would be good to say what can go wrong with that here. >>>> >>> >>> I can add more description. Here is a first attempt. >>> >>> Incorrect calculation of penalty leads to ACPI code assigning the wrong >>> interrupt number to PCI INTx interrupts. >>> >>> This would not be as bad as it sounds in theory. You would just cause the >>> interrupts to be shared and observe performance penalty. >>> >>> However, some drivers like parallel port driver doesn't like interrupt >>> sharing as in this example and causes all other PCI drivers sharing the interrupt >>> to malfunction. >>> >>> The issue has not been caught because the behavior is platform specific >>> and depends on the peripheral drivers sharing the IRQ. >>> >>> I can claim that this could be a BIOS bug. if interrupt 7 is not good for PCI, >>> it shouldn't have been listed in the possible PCI interrupts to begin with. >>> Given this is an existing platform, I don't think we have the luxury to request >>> all BIOS to be updated. This bugfix is needed to support existing platforms. >>> >>> >>> Feel free to request more information if the above description is not clear. >> >> It is clear enough. I can add it to the changelog when applying the patch. > > OK BTW, care to add Fixes: tags to these patches? Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/acpi/pci_link.c b/drivers/acpi/pci_link.c index 8fc7323..f2b69e3 100644 --- a/drivers/acpi/pci_link.c +++ b/drivers/acpi/pci_link.c @@ -470,6 +470,7 @@ static int acpi_irq_pci_sharing_penalty(int irq) { struct acpi_pci_link *link; int penalty = 0; + int i; list_for_each_entry(link, &acpi_link_list, list) { /* @@ -478,18 +479,14 @@ static int acpi_irq_pci_sharing_penalty(int irq) */ if (link->irq.active && link->irq.active == irq) penalty += PIRQ_PENALTY_PCI_USING; - else { - int i; - - /* - * If a link is inactive, penalize the IRQs it - * might use, but not as severely. - */ - for (i = 0; i < link->irq.possible_count; i++) - if (link->irq.possible[i] == irq) - penalty += PIRQ_PENALTY_PCI_POSSIBLE / - link->irq.possible_count; - } + + /* + * penalize the IRQs PCI might use, but not as severely. + */ + for (i = 0; i < link->irq.possible_count; i++) + if (link->irq.possible[i] == irq) + penalty += PIRQ_PENALTY_PCI_POSSIBLE / + link->irq.possible_count; } return penalty;
The change introduced in commit 103544d86976 ("ACPI,PCI,IRQ: reduce resource requirements") omitted the initially assigned POSSIBLE penalty when the IRQ is active. The original code would assign the POSSIBLE value divided by the number of possible IRQs during initialization. Later, if the IRQ is chosen as the active IRQ or if the IRQ is in use by ISA; additional penalties get added. Signed-off-by: Sinan Kaya <okaya@codeaurora.org> --- drivers/acpi/pci_link.c | 21 +++++++++------------ 1 file changed, 9 insertions(+), 12 deletions(-)