Message ID | 20190227165219.31911-1-l.stach@pengutronix.de |
---|---|
State | Accepted |
Headers | show |
Series | PCI: dwc: skip MSI init if MSIs have been explicitly disabled | expand |
Hi Lucas, On Wed, Feb 27, 2019 at 1:53 PM Lucas Stach <l.stach@pengutronix.de> wrote: > > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > hierarchical API) the MSI init claims one of the controller IRQs as a > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > this line is shared with a PCIe legacy IRQ. When the line is claimed for > the MSI domain, any device trying to use this legacy IRQs will fail to > request this IRQ line. > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > as the core won't forward any legacy IRQs once any MSI has been enabled, > users wishing to use legacy IRQs already need to explictly disable MSI > support (usually via the pci=nomsi kernel commandline option). To avoid > any issues with MSI conflicting with legacy IRQs, just skip all of the > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. Thanks for the patch. > Fixes: c5925afbc58 > (PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API) Nit: the commit ID is missing the digit 7 as the first number. While at it, the Fixes lines should written as a single line Fixes: 7c5925afbc58 ("PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API") The " " symbols for the commit Subject is also missing in the commit log.
On Wed, Feb 27, 2019 at 9:25 AM Fabio Estevam <festevam@gmail.com> wrote: > > Hi Lucas, > > On Wed, Feb 27, 2019 at 1:53 PM Lucas Stach <l.stach@pengutronix.de> wrote: > > > > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > > hierarchical API) the MSI init claims one of the controller IRQs as a > > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > > this line is shared with a PCIe legacy IRQ. When the line is claimed for > > the MSI domain, any device trying to use this legacy IRQs will fail to > > request this IRQ line. > > > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > > as the core won't forward any legacy IRQs once any MSI has been enabled, > > users wishing to use legacy IRQs already need to explictly disable MSI > > support (usually via the pci=nomsi kernel commandline option). To avoid > > any issues with MSI conflicting with legacy IRQs, just skip all of the > > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > Thanks for the patch. > > > Fixes: c5925afbc58 > > (PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API) > > Nit: the commit ID is missing the digit 7 as the first number. > > While at it, the Fixes lines should written as a single line > > Fixes: 7c5925afbc58 ("PCI: dwc: Move MSI IRQs allocation to IRQ > domains hierarchical API") > > The " " symbols for the commit Subject is also missing in the commit log. Lucas, Also I think its worth having this in linux-stable for 4.20 and 4.19 (was introduced in 4.17) Thanks, Tim
On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > hierarchical API) the MSI init claims one of the controller IRQs as a > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > this line is shared with a PCIe legacy IRQ. When the line is claimed for > the MSI domain, any device trying to use this legacy IRQs will fail to > request this IRQ line. > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > as the core won't forward any legacy IRQs once any MSI has been enabled, > users wishing to use legacy IRQs already need to explictly disable MSI > support (usually via the pci=nomsi kernel commandline option). To avoid > any issues with MSI conflicting with legacy IRQs, just skip all of the > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > Fixes: c5925afbc58 > (PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API) > Signed-off-by: Lucas Stach <l.stach@pengutronix.de> > Tested-by: Tim Harvey <tharvey@gateworks.com> > --- > drivers/pci/controller/dwc/pcie-designware-host.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) I request Gustavo's ACK to proceed, thanks. Lorenzo > diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c > index 29a05759a294..f4a8494f616b 100644 > --- a/drivers/pci/controller/dwc/pcie-designware-host.c > +++ b/drivers/pci/controller/dwc/pcie-designware-host.c > @@ -433,7 +433,7 @@ int dw_pcie_host_init(struct pcie_port *pp) > if (ret) > pci->num_viewport = 2; > > - if (IS_ENABLED(CONFIG_PCI_MSI)) { > + if (IS_ENABLED(CONFIG_PCI_MSI) && pci_msi_enabled()) { > /* > * If a specific SoC driver needs to change the > * default number of vectors, it needs to implement > -- > 2.20.1 >
On 28/02/2019 13:29, Lorenzo Pieralisi wrote: > On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: >> Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains >> hierarchical API) the MSI init claims one of the controller IRQs as a >> chained IRQ line for the MSI controller. On some designs, like the i.MX6, >> this line is shared with a PCIe legacy IRQ. When the line is claimed for >> the MSI domain, any device trying to use this legacy IRQs will fail to >> request this IRQ line. >> >> As MSI and legacy IRQs are already mutually exclusive on the DWC core, >> as the core won't forward any legacy IRQs once any MSI has been enabled, >> users wishing to use legacy IRQs already need to explictly disable MSI >> support (usually via the pci=nomsi kernel commandline option). To avoid >> any issues with MSI conflicting with legacy IRQs, just skip all of the >> DWC MSI initalization, including the IRQ line claim, when MSI is disabled. >> >> Fixes: c5925afbc58 >> (PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API) >> Signed-off-by: Lucas Stach <l.stach@pengutronix.de> >> Tested-by: Tim Harvey <tharvey@gateworks.com> >> --- >> drivers/pci/controller/dwc/pcie-designware-host.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) > > I request Gustavo's ACK to proceed, thanks. > > Lorenzo > >> diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c >> index 29a05759a294..f4a8494f616b 100644 >> --- a/drivers/pci/controller/dwc/pcie-designware-host.c >> +++ b/drivers/pci/controller/dwc/pcie-designware-host.c >> @@ -433,7 +433,7 @@ int dw_pcie_host_init(struct pcie_port *pp) >> if (ret) >> pci->num_viewport = 2; >> >> - if (IS_ENABLED(CONFIG_PCI_MSI)) { >> + if (IS_ENABLED(CONFIG_PCI_MSI) && pci_msi_enabled()) { >> /* >> * If a specific SoC driver needs to change the >> * default number of vectors, it needs to implement >> -- >> 2.20.1 >> Acked-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > hierarchical API) the MSI init claims one of the controller IRQs as a > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > this line is shared with a PCIe legacy IRQ. When the line is claimed for > the MSI domain, any device trying to use this legacy IRQs will fail to > request this IRQ line. > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > as the core won't forward any legacy IRQs once any MSI has been enabled, > users wishing to use legacy IRQs already need to explictly disable MSI > support (usually via the pci=nomsi kernel commandline option). To avoid > any issues with MSI conflicting with legacy IRQs, just skip all of the > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > Fixes: c5925afbc58 > (PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API) > Signed-off-by: Lucas Stach <l.stach@pengutronix.de> > Tested-by: Tim Harvey <tharvey@gateworks.com> > --- > drivers/pci/controller/dwc/pcie-designware-host.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) Applied to pci/dwc for v5.1, added CC: stable too as requested. Lorenzo > diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c > index 29a05759a294..f4a8494f616b 100644 > --- a/drivers/pci/controller/dwc/pcie-designware-host.c > +++ b/drivers/pci/controller/dwc/pcie-designware-host.c > @@ -433,7 +433,7 @@ int dw_pcie_host_init(struct pcie_port *pp) > if (ret) > pci->num_viewport = 2; > > - if (IS_ENABLED(CONFIG_PCI_MSI)) { > + if (IS_ENABLED(CONFIG_PCI_MSI) && pci_msi_enabled()) { > /* > * If a specific SoC driver needs to change the > * default number of vectors, it needs to implement > -- > 2.20.1 >
[+cc Marc, in case you have any suggestions] Hi Lucas, Nit: this has already been merged, but next time please make the subject line match the convention wrt capitalization. This is not a PCI-specific idea; all it takes is "git log --oneline <file>", and that should be common practice anywhere in the kernel. On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > hierarchical API) the MSI init claims one of the controller IRQs as a > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > this line is shared with a PCIe legacy IRQ. When the line is claimed for > the MSI domain, any device trying to use this legacy IRQs will fail to > request this IRQ line. > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > as the core won't forward any legacy IRQs once any MSI has been enabled, > users wishing to use legacy IRQs already need to explictly disable MSI > support (usually via the pci=nomsi kernel commandline option). To avoid > any issues with MSI conflicting with legacy IRQs, just skip all of the > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. Does this mean that if we have a device that uses legacy IRQs, the user has to figure out to boot with "pci=nomsi"? I don't like kernel command line parameters. If we need that parameter to make devices with legacy IRQs work, what happens without the parameter? How obvious is it that the fix is to use "pci=nomsi"? Is it impossible for Linux to figure this out and make it work automatically? If we can't do it automatically, fine, maybe we have to live with the parameter. But if there's any way we can avoid it, we should. > Fixes: c5925afbc58 > (PCI: dwc: Move MSI IRQs allocation to IRQ domains hierarchical API) > Signed-off-by: Lucas Stach <l.stach@pengutronix.de> > Tested-by: Tim Harvey <tharvey@gateworks.com> > --- > drivers/pci/controller/dwc/pcie-designware-host.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c > index 29a05759a294..f4a8494f616b 100644 > --- a/drivers/pci/controller/dwc/pcie-designware-host.c > +++ b/drivers/pci/controller/dwc/pcie-designware-host.c > @@ -433,7 +433,7 @@ int dw_pcie_host_init(struct pcie_port *pp) > if (ret) > pci->num_viewport = 2; > > - if (IS_ENABLED(CONFIG_PCI_MSI)) { > + if (IS_ENABLED(CONFIG_PCI_MSI) && pci_msi_enabled()) { > /* > * If a specific SoC driver needs to change the > * default number of vectors, it needs to implement > -- > 2.20.1 >
Hi Bjorn, Am Montag, den 04.03.2019, 13:25 -0600 schrieb Bjorn Helgaas: > [+cc Marc, in case you have any suggestions] > > Hi Lucas, > > Nit: this has already been merged, but next time please make the > subject line match the convention wrt capitalization. This is not a > PCI-specific idea; all it takes is "git log --oneline <file>", and > that should be common practice anywhere in the kernel. Sorry about that. I didn't really notice the capitalization, but focused on the prefix. > On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > > hierarchical API) the MSI init claims one of the controller IRQs as a > > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > > this line is shared with a PCIe legacy IRQ. When the line is claimed for > > the MSI domain, any device trying to use this legacy IRQs will fail to > > request this IRQ line. > > > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > > as the core won't forward any legacy IRQs once any MSI has been enabled, > > users wishing to use legacy IRQs already need to explictly disable MSI > > support (usually via the pci=nomsi kernel commandline option). To avoid > > any issues with MSI conflicting with legacy IRQs, just skip all of the > > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > Does this mean that if we have a device that uses legacy IRQs, the > user has to figure out to boot with "pci=nomsi"? As long as there is only a single device connected and there are no port services things will work. If port services are active, those will start to use MSIs, breaking legacy IRQs in the process. I've asked Synopsys if there is a workaround for this, but it seems that the core is working "as designed" with no workaround for this icky behavior. > I don't like kernel command line parameters. If we need that > parameter to make devices with legacy IRQs work, what happens without > the parameter? How obvious is it that the fix is to use "pci=nomsi"? Totally non-obvious currently. Maybe this warrants a warning in the kernel log. > Is it impossible for Linux to figure this out and make it work > automatically? > > If we can't do it automatically, fine, maybe we have to live with the > parameter. But if there's any way we can avoid it, we should. I don't think there is a practical way to make this work automatically. At PCIe enumeration time we can only know if all devices in the hierarchy are able to use MSIs, but this doesn't mean the device driver will use MSIs. This is something we only get to know at driver probe time, at which point it may already be too late, as another device might already use some MSIs, so we can't revert to legacy IRQs only mode. Regards, Lucas
Hi both, On Mon, 04 Mar 2019 19:39:45 +0000, Lucas Stach <l.stach@pengutronix.de> wrote: > > Hi Bjorn, > > Am Montag, den 04.03.2019, 13:25 -0600 schrieb Bjorn Helgaas: > > [+cc Marc, in case you have any suggestions] > > > > Hi Lucas, > > > > Nit: this has already been merged, but next time please make the > > subject line match the convention wrt capitalization. This is not a > > PCI-specific idea; all it takes is "git log --oneline <file>", and > > that should be common practice anywhere in the kernel. > > Sorry about that. I didn't really notice the capitalization, but > focused on the prefix. > > > On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > > > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > > > hierarchical API) the MSI init claims one of the controller IRQs as a > > > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > > > this line is shared with a PCIe legacy IRQ. When the line is claimed for > > > the MSI domain, any device trying to use this legacy IRQs will fail to > > > request this IRQ line. > > > > > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > > > as the core won't forward any legacy IRQs once any MSI has been enabled, > > > users wishing to use legacy IRQs already need to explictly disable MSI > > > support (usually via the pci=nomsi kernel commandline option). To avoid > > > any issues with MSI conflicting with legacy IRQs, just skip all of the > > > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > > > Does this mean that if we have a device that uses legacy IRQs, the > > user has to figure out to boot with "pci=nomsi"? > > As long as there is only a single device connected and there are no > port services things will work. If port services are active, those will > start to use MSIs, breaking legacy IRQs in the process. > > I've asked Synopsys if there is a workaround for this, but it seems > that the core is working "as designed" with no workaround for this icky > behavior. Is this the general DWC controller behaviour? Or something that is specific to a given implementation? I can't believe someone actually thought this is an acceptable behaviour... Gustavo, can you please check with your HW colleagues and let everybody know what's the official Synopsys position on this? > > > I don't like kernel command line parameters. If we need that > > parameter to make devices with legacy IRQs work, what happens without > > the parameter? How obvious is it that the fix is to use "pci=nomsi"? > > Totally non-obvious currently. Maybe this warrants a warning in the > kernel log. > > > Is it impossible for Linux to figure this out and make it work > > automatically? > > > > If we can't do it automatically, fine, maybe we have to live with the > > parameter. But if there's any way we can avoid it, we should. > > I don't think there is a practical way to make this work > automatically. > > At PCIe enumeration time we can only know if all devices in the > hierarchy are able to use MSIs, but this doesn't mean the device driver > will use MSIs. This is something we only get to know at driver probe > time, at which point it may already be too late, as another device > might already use some MSIs, so we can't revert to legacy IRQs only > mode. It is almost as the default is backward. It feels like it should default to legacy interrupts, and only enable MSIs if the user actually asks for it. We should always be able to provide legacy interrupts, as this is the usual fallback for most drivers. Thanks, M.
Hi, On 04/03/2019 20:18, Marc Zyngier wrote: > Hi both, > > On Mon, 04 Mar 2019 19:39:45 +0000, > Lucas Stach <l.stach@pengutronix.de> wrote: >> (snipped) >>>> >>>> As MSI and legacy IRQs are already mutually exclusive on the DWC core, >>>> as the core won't forward any legacy IRQs once any MSI has been enabled, >>>> users wishing to use legacy IRQs already need to explictly disable MSI >>>> support (usually via the pci=nomsi kernel commandline option). To avoid >>>> any issues with MSI conflicting with legacy IRQs, just skip all of the >>>> DWC MSI initalization, including the IRQ line claim, when MSI is disabled. >>> >>> Does this mean that if we have a device that uses legacy IRQs, the >>> user has to figure out to boot with "pci=nomsi"? >> >> As long as there is only a single device connected and there are no >> port services things will work. If port services are active, those will >> start to use MSIs, breaking legacy IRQs in the process. >> >> I've asked Synopsys if there is a workaround for this, but it seems >> that the core is working "as designed" with no workaround for this icky >> behavior. > > Is this the general DWC controller behaviour? Or something that is > specific to a given implementation? I can't believe someone actually > thought this is an acceptable behaviour... > > Gustavo, can you please check with your HW colleagues and let > everybody know what's the official Synopsys position on this? > Sure, I can ask the HW team to provide me more info about this, This can take a while. Unfortunately on my setup I only have MSI and MSI-X, therefore I can't really test what has been statemented. Regards, Gustavo
Am Mittwoch, den 06.03.2019, 09:53 +0000 schrieb Gustavo Pimentel: > Hi, > > On 04/03/2019 20:18, Marc Zyngier wrote: > > Hi both, > > > > On Mon, 04 Mar 2019 19:39:45 +0000, > > > > Lucas Stach <l.stach@pengutronix.de> wrote: > > > > > (snipped) > > > > > > > > > > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > > > > > as the core won't forward any legacy IRQs once any MSI has been enabled, > > > > > users wishing to use legacy IRQs already need to explictly disable MSI > > > > > support (usually via the pci=nomsi kernel commandline option). To avoid > > > > > any issues with MSI conflicting with legacy IRQs, just skip all of the > > > > > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > > > > > > > Does this mean that if we have a device that uses legacy IRQs, the > > > > user has to figure out to boot with "pci=nomsi"? > > > > > > As long as there is only a single device connected and there are no > > > port services things will work. If port services are active, those will > > > start to use MSIs, breaking legacy IRQs in the process. > > > > > > I've asked Synopsys if there is a workaround for this, but it seems > > > that the core is working "as designed" with no workaround for this icky > > > behavior. > > > > Is this the general DWC controller behaviour? Or something that is > > specific to a given implementation? I can't believe someone actually > > thought this is an acceptable behaviour... > > > > Gustavo, can you please check with your HW colleagues and let > > everybody know what's the official Synopsys position on this? > > > > Sure, I can ask the HW team to provide me more info about this, This can take a > while. Unfortunately on my setup I only have MSI and MSI-X, therefore I can't > really test what has been statemented. At least this behavior was confirmed by Joao Pinto at Synopsys last time the issue was discussed. http://lists.infradead.org/pipermail/linux-arm-kernel/2017-April/499591.html Regards, Lucas
On Mon, Mar 4, 2019 at 12:18 PM Marc Zyngier <marc.zyngier@arm.com> wrote: > > Hi both, > > On Mon, 04 Mar 2019 19:39:45 +0000, > Lucas Stach <l.stach@pengutronix.de> wrote: > > > > Hi Bjorn, > > > > Am Montag, den 04.03.2019, 13:25 -0600 schrieb Bjorn Helgaas: > > > [+cc Marc, in case you have any suggestions] > > > > > > Hi Lucas, > > > > > > Nit: this has already been merged, but next time please make the > > > subject line match the convention wrt capitalization. This is not a > > > PCI-specific idea; all it takes is "git log --oneline <file>", and > > > that should be common practice anywhere in the kernel. > > > > Sorry about that. I didn't really notice the capitalization, but > > focused on the prefix. > > > > > On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > > > > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > > > > hierarchical API) the MSI init claims one of the controller IRQs as a > > > > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > > > > this line is shared with a PCIe legacy IRQ. When the line is claimed for > > > > the MSI domain, any device trying to use this legacy IRQs will fail to > > > > request this IRQ line. > > > > > > > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > > > > as the core won't forward any legacy IRQs once any MSI has been enabled, > > > > users wishing to use legacy IRQs already need to explictly disable MSI > > > > support (usually via the pci=nomsi kernel commandline option). To avoid > > > > any issues with MSI conflicting with legacy IRQs, just skip all of the > > > > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > > > > > Does this mean that if we have a device that uses legacy IRQs, the > > > user has to figure out to boot with "pci=nomsi"? > > > > As long as there is only a single device connected and there are no > > port services things will work. If port services are active, those will > > start to use MSIs, breaking legacy IRQs in the process. > > > > I've asked Synopsys if there is a workaround for this, but it seems > > that the core is working "as designed" with no workaround for this icky > > behavior. > > Is this the general DWC controller behaviour? Or something that is > specific to a given implementation? I can't believe someone actually > thought this is an acceptable behaviour... > > Gustavo, can you please check with your HW colleagues and let > everybody know what's the official Synopsys position on this? > > > > > > I don't like kernel command line parameters. If we need that > > > parameter to make devices with legacy IRQs work, what happens without > > > the parameter? How obvious is it that the fix is to use "pci=nomsi"? > > > > Totally non-obvious currently. Maybe this warrants a warning in the > > kernel log. > > > > > Is it impossible for Linux to figure this out and make it work > > > automatically? > > > > > > If we can't do it automatically, fine, maybe we have to live with the > > > parameter. But if there's any way we can avoid it, we should. > > > > I don't think there is a practical way to make this work > > automatically. > > > > At PCIe enumeration time we can only know if all devices in the > > hierarchy are able to use MSIs, but this doesn't mean the device driver > > will use MSIs. This is something we only get to know at driver probe > > time, at which point it may already be too late, as another device > > might already use some MSIs, so we can't revert to legacy IRQs only > > mode. > > It is almost as the default is backward. It feels like it should > default to legacy interrupts, and only enable MSIs if the user > actually asks for it. We should always be able to provide legacy > interrupts, as this is the usual fallback for most drivers. > Marc, I would agree with 'should default to legacy interrupts' as that works everywhere AFAIK and likely allows for improved performance. Tim
On Mon, Mar 4, 2019 at 20:18:13, Marc Zyngier <marc.zyngier@arm.com> wrote: Hi, > Hi both, > > On Mon, 04 Mar 2019 19:39:45 +0000, > Lucas Stach <l.stach@pengutronix.de> wrote: > > > > Hi Bjorn, > > > > Am Montag, den 04.03.2019, 13:25 -0600 schrieb Bjorn Helgaas: > > > [+cc Marc, in case you have any suggestions] > > > > > > Hi Lucas, > > > > > > Nit: this has already been merged, but next time please make the > > > subject line match the convention wrt capitalization. This is not a > > > PCI-specific idea; all it takes is "git log --oneline <file>", and > > > that should be common practice anywhere in the kernel. > > > > Sorry about that. I didn't really notice the capitalization, but > > focused on the prefix. > > > > > On Wed, Feb 27, 2019 at 05:52:19PM +0100, Lucas Stach wrote: > > > > Since 7c5925afbc58 (PCI: dwc: Move MSI IRQs allocation to IRQ domains > > > > hierarchical API) the MSI init claims one of the controller IRQs as a > > > > chained IRQ line for the MSI controller. On some designs, like the i.MX6, > > > > this line is shared with a PCIe legacy IRQ. When the line is claimed for > > > > the MSI domain, any device trying to use this legacy IRQs will fail to > > > > request this IRQ line. > > > > > > > > As MSI and legacy IRQs are already mutually exclusive on the DWC core, > > > > as the core won't forward any legacy IRQs once any MSI has been enabled, > > > > users wishing to use legacy IRQs already need to explictly disable MSI > > > > support (usually via the pci=nomsi kernel commandline option). To avoid > > > > any issues with MSI conflicting with legacy IRQs, just skip all of the > > > > DWC MSI initalization, including the IRQ line claim, when MSI is disabled. > > > > > > Does this mean that if we have a device that uses legacy IRQs, the > > > user has to figure out to boot with "pci=nomsi"? > > > > As long as there is only a single device connected and there are no > > port services things will work. If port services are active, those will > > start to use MSIs, breaking legacy IRQs in the process. > > > > I've asked Synopsys if there is a workaround for this, but it seems > > that the core is working "as designed" with no workaround for this icky > > behavior. > > Is this the general DWC controller behaviour? Or something that is > specific to a given implementation? I can't believe someone actually > thought this is an acceptable behaviour... > > Gustavo, can you please check with your HW colleagues and let > everybody know what's the official Synopsys position on this? > Sorry for the delay, according to HW colleagues, there are no special restrictions on receiving/dealing all the interrupts types at the same time. In other words, for example, it's possible to have by design the DW PCIe controller working with three EPs each one with a different type of interrupt enabled (legacy, MSI, MSI-X). Unfortunately, I never tested this kind of scenario due to my prototype set up restrictions. Regards, Gustavo > > > > > I don't like kernel command line parameters. If we need that > > > parameter to make devices with legacy IRQs work, what happens without > > > the parameter? How obvious is it that the fix is to use "pci=nomsi"? > > > > Totally non-obvious currently. Maybe this warrants a warning in the > > kernel log. > > > > > Is it impossible for Linux to figure this out and make it work > > > automatically? > > > > > > If we can't do it automatically, fine, maybe we have to live with the > > > parameter. But if there's any way we can avoid it, we should. > > > > I don't think there is a practical way to make this work > > automatically. > > > > At PCIe enumeration time we can only know if all devices in the > > hierarchy are able to use MSIs, but this doesn't mean the device driver > > will use MSIs. This is something we only get to know at driver probe > > time, at which point it may already be too late, as another device > > might already use some MSIs, so we can't revert to legacy IRQs only > > mode. > > It is almost as the default is backward. It feels like it should > default to legacy interrupts, and only enable MSIs if the user > actually asks for it. We should always be able to provide legacy > interrupts, as this is the usual fallback for most drivers. > > Thanks, > > M. > > -- > Jazz is not dead, it just smell funny.
diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c index 29a05759a294..f4a8494f616b 100644 --- a/drivers/pci/controller/dwc/pcie-designware-host.c +++ b/drivers/pci/controller/dwc/pcie-designware-host.c @@ -433,7 +433,7 @@ int dw_pcie_host_init(struct pcie_port *pp) if (ret) pci->num_viewport = 2; - if (IS_ENABLED(CONFIG_PCI_MSI)) { + if (IS_ENABLED(CONFIG_PCI_MSI) && pci_msi_enabled()) { /* * If a specific SoC driver needs to change the * default number of vectors, it needs to implement