| Message ID | 20201009155505.5a580ef5@xhacker.debian (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Headers | show |
| Series | PCI: dwc: fix two MSI issues | expand |
On 2020-10-09 08:55, Jisheng Zhang wrote: > Currently, dw_pcie_msi_init() allocates and maps page for msi, then > program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex > may lose power during suspend-to-RAM, so when we resume, we want to > redo the latter but not the former. If designware based driver (for > example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the > msi page will be leaked. > > As pointed out by Rob and Ard, there's no need to allocate a page for > the MSI address, we could use an address in the driver data. > > To avoid map the MSI msg again during resume, we move the map MSI msg > from dw_pcie_msi_init() to dw_pcie_host_init(). You should move the unmap there as well. As soon as you know what the relevant address would be if you *were* to do DMA to this location, then the exercise is complete. Leaving it mapped for the lifetime of the device in order to do not-DMA to it seems questionable (and represents technically incorrect API usage without at least a sync_for_cpu call before any other access to the data). Another point of note is that using streaming DMA mappings at all is a bit fragile (regardless of this change). If the host controller itself has a limited DMA mask relative to physical memory (which integrators still seem to keep doing...) then you could end up punching your MSI hole right in the middle of the SWIOTLB bounce buffer, where it's then almost *guaranteed* to interfere with real DMA :( If no DWC users have that problem and the current code is working well enough, then I see little reason not to make this partucular change to tidy up the implementation, just bear in mind that there's always the possibility of having to come back and change it yet again in future to make it more robust. I had it in mind that this trick was done with a coherent DMA allocation, which would be safe from addressing problems but would need to be kept around for the lifetime of the device, but maybe that was a different driver :/ Robin. > Suggested-by: Rob Herring <robh@kernel.org> > Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com> > Reviewed-by: Rob Herring <robh@kernel.org> > --- > drivers/pci/controller/dwc/pci-dra7xx.c | 18 +++++++++- > .../pci/controller/dwc/pcie-designware-host.c | 33 ++++++++++--------- > drivers/pci/controller/dwc/pcie-designware.h | 2 +- > 3 files changed, 36 insertions(+), 17 deletions(-) > > diff --git a/drivers/pci/controller/dwc/pci-dra7xx.c b/drivers/pci/controller/dwc/pci-dra7xx.c > index 8f0b6d644e4b..6d012d2b1e90 100644 > --- a/drivers/pci/controller/dwc/pci-dra7xx.c > +++ b/drivers/pci/controller/dwc/pci-dra7xx.c > @@ -466,7 +466,9 @@ static struct irq_chip dra7xx_pci_msi_bottom_irq_chip = { > static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) > { > struct dw_pcie *pci = to_dw_pcie_from_pp(pp); > + struct device *dev = pci->dev; > u32 ctrl, num_ctrls; > + int ret; > > pp->msi_irq_chip = &dra7xx_pci_msi_bottom_irq_chip; > > @@ -482,7 +484,21 @@ static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) > ~0); > } > > - return dw_pcie_allocate_domains(pp); > + ret = dw_pcie_allocate_domains(pp); > + if (ret) > + return ret; > + > + pp->msi_data = dma_map_single_attrs(dev, &pp->msi_msg, > + sizeof(pp->msi_msg), > + DMA_FROM_DEVICE, > + DMA_ATTR_SKIP_CPU_SYNC); > + ret = dma_mapping_error(dev, pp->msi_data); > + if (ret) { > + dev_err(dev, "Failed to map MSI data\n"); > + pp->msi_data = 0; > + dw_pcie_free_msi(pp); > + } > + return ret; > } > > static const struct dw_pcie_host_ops dra7xx_pcie_host_ops = { > diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c > index d3e9ea11ce9e..d02c7e74738d 100644 > --- a/drivers/pci/controller/dwc/pcie-designware-host.c > +++ b/drivers/pci/controller/dwc/pcie-designware-host.c > @@ -266,30 +266,23 @@ void dw_pcie_free_msi(struct pcie_port *pp) > irq_domain_remove(pp->msi_domain); > irq_domain_remove(pp->irq_domain); > > - if (pp->msi_page) > - __free_page(pp->msi_page); > + if (pp->msi_data) { > + struct dw_pcie *pci = to_dw_pcie_from_pp(pp); > + struct device *dev = pci->dev; > + > + dma_unmap_single_attrs(dev, pp->msi_data, sizeof(pp->msi_msg), > + DMA_FROM_DEVICE, DMA_ATTR_SKIP_CPU_SYNC); > + } > } > > void dw_pcie_msi_init(struct pcie_port *pp) > { > struct dw_pcie *pci = to_dw_pcie_from_pp(pp); > - struct device *dev = pci->dev; > - u64 msi_target; > + u64 msi_target = (u64)pp->msi_data; > > if (!IS_ENABLED(CONFIG_PCI_MSI)) > return; > > - pp->msi_page = alloc_page(GFP_KERNEL); > - pp->msi_data = dma_map_page(dev, pp->msi_page, 0, PAGE_SIZE, > - DMA_FROM_DEVICE); > - if (dma_mapping_error(dev, pp->msi_data)) { > - dev_err(dev, "Failed to map MSI data\n"); > - __free_page(pp->msi_page); > - pp->msi_page = NULL; > - return; > - } > - msi_target = (u64)pp->msi_data; > - > /* Program the msi_data */ > dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_LO, lower_32_bits(msi_target)); > dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_HI, upper_32_bits(msi_target)); > @@ -394,6 +387,16 @@ int dw_pcie_host_init(struct pcie_port *pp) > irq_set_chained_handler_and_data(pp->msi_irq, > dw_chained_msi_isr, > pp); > + > + pp->msi_data = dma_map_single_attrs(pci->dev, &pp->msi_msg, > + sizeof(pp->msi_msg), > + DMA_FROM_DEVICE, > + DMA_ATTR_SKIP_CPU_SYNC); > + if (dma_mapping_error(pci->dev, pp->msi_data)) { > + dev_err(pci->dev, "Failed to map MSI data\n"); > + pp->msi_data = 0; > + goto err_free_msi; > + } > } else { > ret = pp->ops->msi_host_init(pp); > if (ret < 0) > diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h > index 97c7063b9e89..9d2f511f13fa 100644 > --- a/drivers/pci/controller/dwc/pcie-designware.h > +++ b/drivers/pci/controller/dwc/pcie-designware.h > @@ -190,8 +190,8 @@ struct pcie_port { > int msi_irq; > struct irq_domain *irq_domain; > struct irq_domain *msi_domain; > + u16 msi_msg; > dma_addr_t msi_data; > - struct page *msi_page; > struct irq_chip *msi_irq_chip; > u32 num_vectors; > u32 irq_mask[MAX_MSI_CTRLS]; >
On 10/12/2020 5:07 PM, Robin Murphy wrote: > External email: Use caution opening links or attachments > > > On 2020-10-09 08:55, Jisheng Zhang wrote: >> Currently, dw_pcie_msi_init() allocates and maps page for msi, then >> program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex >> may lose power during suspend-to-RAM, so when we resume, we want to >> redo the latter but not the former. If designware based driver (for >> example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the >> msi page will be leaked. >> >> As pointed out by Rob and Ard, there's no need to allocate a page for >> the MSI address, we could use an address in the driver data. >> >> To avoid map the MSI msg again during resume, we move the map MSI msg >> from dw_pcie_msi_init() to dw_pcie_host_init(). > > You should move the unmap there as well. As soon as you know what the > relevant address would be if you *were* to do DMA to this location, then > the exercise is complete. Leaving it mapped for the lifetime of the > device in order to do not-DMA to it seems questionable (and represents > technically incorrect API usage without at least a sync_for_cpu call > before any other access to the data). > > Another point of note is that using streaming DMA mappings at all is a > bit fragile (regardless of this change). If the host controller itself > has a limited DMA mask relative to physical memory (which integrators > still seem to keep doing...) then you could end up punching your MSI > hole right in the middle of the SWIOTLB bounce buffer, where it's then > almost *guaranteed* to interfere with real DMA :( Agree with Robin. Since the MSI page is going to be locked till shutdown/reboot, wouldn't it make sense to use dma_alloc_coherent() API? Also, shouldn't we call dma_set_mask() to limit the address to only 32-bits so as to enable MSI for even those legacy PCIe devices with only 32-bit MSI capability? - Vidya Sagar > > If no DWC users have that problem and the current code is working well > enough, then I see little reason not to make this partucular change to > tidy up the implementation, just bear in mind that there's always the > possibility of having to come back and change it yet again in future to > make it more robust. I had it in mind that this trick was done with a > coherent DMA allocation, which would be safe from addressing problems > but would need to be kept around for the lifetime of the device, but > maybe that was a different driver :/ > > Robin. > >> Suggested-by: Rob Herring <robh@kernel.org> >> Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com> >> Reviewed-by: Rob Herring <robh@kernel.org> >> --- >> drivers/pci/controller/dwc/pci-dra7xx.c | 18 +++++++++- >> .../pci/controller/dwc/pcie-designware-host.c | 33 ++++++++++--------- >> drivers/pci/controller/dwc/pcie-designware.h | 2 +- >> 3 files changed, 36 insertions(+), 17 deletions(-) >> >> diff --git a/drivers/pci/controller/dwc/pci-dra7xx.c >> b/drivers/pci/controller/dwc/pci-dra7xx.c >> index 8f0b6d644e4b..6d012d2b1e90 100644 >> --- a/drivers/pci/controller/dwc/pci-dra7xx.c >> +++ b/drivers/pci/controller/dwc/pci-dra7xx.c >> @@ -466,7 +466,9 @@ static struct irq_chip >> dra7xx_pci_msi_bottom_irq_chip = { >> static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) >> { >> struct dw_pcie *pci = to_dw_pcie_from_pp(pp); >> + struct device *dev = pci->dev; >> u32 ctrl, num_ctrls; >> + int ret; >> >> pp->msi_irq_chip = &dra7xx_pci_msi_bottom_irq_chip; >> >> @@ -482,7 +484,21 @@ static int dra7xx_pcie_msi_host_init(struct >> pcie_port *pp) >> ~0); >> } >> >> - return dw_pcie_allocate_domains(pp); >> + ret = dw_pcie_allocate_domains(pp); >> + if (ret) >> + return ret; >> + >> + pp->msi_data = dma_map_single_attrs(dev, &pp->msi_msg, >> + sizeof(pp->msi_msg), >> + DMA_FROM_DEVICE, >> + DMA_ATTR_SKIP_CPU_SYNC); >> + ret = dma_mapping_error(dev, pp->msi_data); >> + if (ret) { >> + dev_err(dev, "Failed to map MSI data\n"); >> + pp->msi_data = 0; >> + dw_pcie_free_msi(pp); >> + } >> + return ret; >> } >> >> static const struct dw_pcie_host_ops dra7xx_pcie_host_ops = { >> diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c >> b/drivers/pci/controller/dwc/pcie-designware-host.c >> index d3e9ea11ce9e..d02c7e74738d 100644 >> --- a/drivers/pci/controller/dwc/pcie-designware-host.c >> +++ b/drivers/pci/controller/dwc/pcie-designware-host.c >> @@ -266,30 +266,23 @@ void dw_pcie_free_msi(struct pcie_port *pp) >> irq_domain_remove(pp->msi_domain); >> irq_domain_remove(pp->irq_domain); >> >> - if (pp->msi_page) >> - __free_page(pp->msi_page); >> + if (pp->msi_data) { >> + struct dw_pcie *pci = to_dw_pcie_from_pp(pp); >> + struct device *dev = pci->dev; >> + >> + dma_unmap_single_attrs(dev, pp->msi_data, >> sizeof(pp->msi_msg), >> + DMA_FROM_DEVICE, >> DMA_ATTR_SKIP_CPU_SYNC); >> + } >> } >> >> void dw_pcie_msi_init(struct pcie_port *pp) >> { >> struct dw_pcie *pci = to_dw_pcie_from_pp(pp); >> - struct device *dev = pci->dev; >> - u64 msi_target; >> + u64 msi_target = (u64)pp->msi_data; >> >> if (!IS_ENABLED(CONFIG_PCI_MSI)) >> return; >> >> - pp->msi_page = alloc_page(GFP_KERNEL); >> - pp->msi_data = dma_map_page(dev, pp->msi_page, 0, PAGE_SIZE, >> - DMA_FROM_DEVICE); >> - if (dma_mapping_error(dev, pp->msi_data)) { >> - dev_err(dev, "Failed to map MSI data\n"); >> - __free_page(pp->msi_page); >> - pp->msi_page = NULL; >> - return; >> - } >> - msi_target = (u64)pp->msi_data; >> - >> /* Program the msi_data */ >> dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_LO, >> lower_32_bits(msi_target)); >> dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_HI, >> upper_32_bits(msi_target)); >> @@ -394,6 +387,16 @@ int dw_pcie_host_init(struct pcie_port *pp) >> >> irq_set_chained_handler_and_data(pp->msi_irq, >> >> dw_chained_msi_isr, >> pp); >> + >> + pp->msi_data = dma_map_single_attrs(pci->dev, >> &pp->msi_msg, >> + sizeof(pp->msi_msg), >> + DMA_FROM_DEVICE, >> + >> DMA_ATTR_SKIP_CPU_SYNC); >> + if (dma_mapping_error(pci->dev, pp->msi_data)) { >> + dev_err(pci->dev, "Failed to map MSI >> data\n"); >> + pp->msi_data = 0; >> + goto err_free_msi; >> + } >> } else { >> ret = pp->ops->msi_host_init(pp); >> if (ret < 0) >> diff --git a/drivers/pci/controller/dwc/pcie-designware.h >> b/drivers/pci/controller/dwc/pcie-designware.h >> index 97c7063b9e89..9d2f511f13fa 100644 >> --- a/drivers/pci/controller/dwc/pcie-designware.h >> +++ b/drivers/pci/controller/dwc/pcie-designware.h >> @@ -190,8 +190,8 @@ struct pcie_port { >> int msi_irq; >> struct irq_domain *irq_domain; >> struct irq_domain *msi_domain; >> + u16 msi_msg; >> dma_addr_t msi_data; >> - struct page *msi_page; >> struct irq_chip *msi_irq_chip; >> u32 num_vectors; >> u32 irq_mask[MAX_MSI_CTRLS]; >>
On Mon, Oct 12, 2020 at 6:37 AM Robin Murphy <robin.murphy@arm.com> wrote: > > On 2020-10-09 08:55, Jisheng Zhang wrote: > > Currently, dw_pcie_msi_init() allocates and maps page for msi, then > > program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex > > may lose power during suspend-to-RAM, so when we resume, we want to > > redo the latter but not the former. If designware based driver (for > > example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the > > msi page will be leaked. > > > > As pointed out by Rob and Ard, there's no need to allocate a page for > > the MSI address, we could use an address in the driver data. > > > > To avoid map the MSI msg again during resume, we move the map MSI msg > > from dw_pcie_msi_init() to dw_pcie_host_init(). > > You should move the unmap there as well. As soon as you know what the > relevant address would be if you *were* to do DMA to this location, then > the exercise is complete. Leaving it mapped for the lifetime of the > device in order to do not-DMA to it seems questionable (and represents > technically incorrect API usage without at least a sync_for_cpu call > before any other access to the data). > > Another point of note is that using streaming DMA mappings at all is a > bit fragile (regardless of this change). If the host controller itself > has a limited DMA mask relative to physical memory (which integrators > still seem to keep doing...) then you could end up punching your MSI > hole right in the middle of the SWIOTLB bounce buffer, where it's then > almost *guaranteed* to interfere with real DMA :( Couldn't that happen with the current code too? alloc_page() isn't guaranteed to be DMA'able, right? > If no DWC users have that problem and the current code is working well > enough, then I see little reason not to make this partucular change to > tidy up the implementation, just bear in mind that there's always the > possibility of having to come back and change it yet again in future to > make it more robust. I had it in mind that this trick was done with a > coherent DMA allocation, which would be safe from addressing problems > but would need to be kept around for the lifetime of the device, but > maybe that was a different driver :/ Well, we're wasting 4K or 64K of memory and then leaking it is the main reason to change it. We just need any address that's not memory which PCI could access. We could possibly just take the end of (outbound) PCI memory space. Note that the DWC driver never sets up inbound translations, so it's all 1:1 mapping (though upstream could have some translation). Rob
On 2020-10-14 15:15, Rob Herring wrote: > On Mon, Oct 12, 2020 at 6:37 AM Robin Murphy <robin.murphy@arm.com> wrote: >> >> On 2020-10-09 08:55, Jisheng Zhang wrote: >>> Currently, dw_pcie_msi_init() allocates and maps page for msi, then >>> program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex >>> may lose power during suspend-to-RAM, so when we resume, we want to >>> redo the latter but not the former. If designware based driver (for >>> example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the >>> msi page will be leaked. >>> >>> As pointed out by Rob and Ard, there's no need to allocate a page for >>> the MSI address, we could use an address in the driver data. >>> >>> To avoid map the MSI msg again during resume, we move the map MSI msg >>> from dw_pcie_msi_init() to dw_pcie_host_init(). >> >> You should move the unmap there as well. As soon as you know what the >> relevant address would be if you *were* to do DMA to this location, then >> the exercise is complete. Leaving it mapped for the lifetime of the >> device in order to do not-DMA to it seems questionable (and represents >> technically incorrect API usage without at least a sync_for_cpu call >> before any other access to the data). >> >> Another point of note is that using streaming DMA mappings at all is a >> bit fragile (regardless of this change). If the host controller itself >> has a limited DMA mask relative to physical memory (which integrators >> still seem to keep doing...) then you could end up punching your MSI >> hole right in the middle of the SWIOTLB bounce buffer, where it's then >> almost *guaranteed* to interfere with real DMA :( > > Couldn't that happen with the current code too? alloc_page() isn't > guaranteed to be DMA'able, right? Indeed that's what I meant by "regardless of this change". >> If no DWC users have that problem and the current code is working well >> enough, then I see little reason not to make this partucular change to >> tidy up the implementation, just bear in mind that there's always the >> possibility of having to come back and change it yet again in future to >> make it more robust. I had it in mind that this trick was done with a >> coherent DMA allocation, which would be safe from addressing problems >> but would need to be kept around for the lifetime of the device, but >> maybe that was a different driver :/ > > Well, we're wasting 4K or 64K of memory and then leaking it is the > main reason to change it. > > We just need any address that's not memory which PCI could access. We > could possibly just take the end of (outbound) PCI memory space. Note > that the DWC driver never sets up inbound translations, so it's all > 1:1 mapping (though upstream could have some translation). Right, this patch is undeniably a better implementation of the existing approach, I just felt it worth pointing out that that approach itself has fundamental flaws which may or may not be relevant to some current and/or future users. I know for a fact that there are platforms which cripple their PCIe host bridge to 32-bit physical addressing but support having RAM above that; I don't *think* any of the ones I know of are using the dw_pcie driver, but hey, how much do I know? ;) Robin.
On Mon, 12 Oct 2020 at 13:38, Robin Murphy <robin.murphy@arm.com> wrote: > > On 2020-10-09 08:55, Jisheng Zhang wrote: > > Currently, dw_pcie_msi_init() allocates and maps page for msi, then > > program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex > > may lose power during suspend-to-RAM, so when we resume, we want to > > redo the latter but not the former. If designware based driver (for > > example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the > > msi page will be leaked. > > > > As pointed out by Rob and Ard, there's no need to allocate a page for > > the MSI address, we could use an address in the driver data. > > > > To avoid map the MSI msg again during resume, we move the map MSI msg > > from dw_pcie_msi_init() to dw_pcie_host_init(). > > You should move the unmap there as well. As soon as you know what the > relevant address would be if you *were* to do DMA to this location, then > the exercise is complete. Leaving it mapped for the lifetime of the > device in order to do not-DMA to it seems questionable (and represents > technically incorrect API usage without at least a sync_for_cpu call > before any other access to the data). > > Another point of note is that using streaming DMA mappings at all is a > bit fragile (regardless of this change). If the host controller itself > has a limited DMA mask relative to physical memory (which integrators > still seem to keep doing...) then you could end up punching your MSI > hole right in the middle of the SWIOTLB bounce buffer, where it's then > almost *guaranteed* to interfere with real DMA :( > Wouldn't it be the unmap you are suggesting that would create this problem? If the bounce buffer is never released, the fake MSI doorbell address can never conflict with any other DMA mappings. > If no DWC users have that problem and the current code is working well > enough, then I see little reason not to make this partucular change to > tidy up the implementation, just bear in mind that there's always the > possibility of having to come back and change it yet again in future to > make it more robust. I had it in mind that this trick was done with a > coherent DMA allocation, which would be safe from addressing problems > but would need to be kept around for the lifetime of the device, but > maybe that was a different driver :/ > > Robin. > > > Suggested-by: Rob Herring <robh@kernel.org> > > Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com> > > Reviewed-by: Rob Herring <robh@kernel.org> > > --- > > drivers/pci/controller/dwc/pci-dra7xx.c | 18 +++++++++- > > .../pci/controller/dwc/pcie-designware-host.c | 33 ++++++++++--------- > > drivers/pci/controller/dwc/pcie-designware.h | 2 +- > > 3 files changed, 36 insertions(+), 17 deletions(-) > > > > diff --git a/drivers/pci/controller/dwc/pci-dra7xx.c b/drivers/pci/controller/dwc/pci-dra7xx.c > > index 8f0b6d644e4b..6d012d2b1e90 100644 > > --- a/drivers/pci/controller/dwc/pci-dra7xx.c > > +++ b/drivers/pci/controller/dwc/pci-dra7xx.c > > @@ -466,7 +466,9 @@ static struct irq_chip dra7xx_pci_msi_bottom_irq_chip = { > > static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) > > { > > struct dw_pcie *pci = to_dw_pcie_from_pp(pp); > > + struct device *dev = pci->dev; > > u32 ctrl, num_ctrls; > > + int ret; > > > > pp->msi_irq_chip = &dra7xx_pci_msi_bottom_irq_chip; > > > > @@ -482,7 +484,21 @@ static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) > > ~0); > > } > > > > - return dw_pcie_allocate_domains(pp); > > + ret = dw_pcie_allocate_domains(pp); > > + if (ret) > > + return ret; > > + > > + pp->msi_data = dma_map_single_attrs(dev, &pp->msi_msg, > > + sizeof(pp->msi_msg), > > + DMA_FROM_DEVICE, > > + DMA_ATTR_SKIP_CPU_SYNC); > > + ret = dma_mapping_error(dev, pp->msi_data); > > + if (ret) { > > + dev_err(dev, "Failed to map MSI data\n"); > > + pp->msi_data = 0; > > + dw_pcie_free_msi(pp); > > + } > > + return ret; > > } > > > > static const struct dw_pcie_host_ops dra7xx_pcie_host_ops = { > > diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c > > index d3e9ea11ce9e..d02c7e74738d 100644 > > --- a/drivers/pci/controller/dwc/pcie-designware-host.c > > +++ b/drivers/pci/controller/dwc/pcie-designware-host.c > > @@ -266,30 +266,23 @@ void dw_pcie_free_msi(struct pcie_port *pp) > > irq_domain_remove(pp->msi_domain); > > irq_domain_remove(pp->irq_domain); > > > > - if (pp->msi_page) > > - __free_page(pp->msi_page); > > + if (pp->msi_data) { > > + struct dw_pcie *pci = to_dw_pcie_from_pp(pp); > > + struct device *dev = pci->dev; > > + > > + dma_unmap_single_attrs(dev, pp->msi_data, sizeof(pp->msi_msg), > > + DMA_FROM_DEVICE, DMA_ATTR_SKIP_CPU_SYNC); > > + } > > } > > > > void dw_pcie_msi_init(struct pcie_port *pp) > > { > > struct dw_pcie *pci = to_dw_pcie_from_pp(pp); > > - struct device *dev = pci->dev; > > - u64 msi_target; > > + u64 msi_target = (u64)pp->msi_data; > > > > if (!IS_ENABLED(CONFIG_PCI_MSI)) > > return; > > > > - pp->msi_page = alloc_page(GFP_KERNEL); > > - pp->msi_data = dma_map_page(dev, pp->msi_page, 0, PAGE_SIZE, > > - DMA_FROM_DEVICE); > > - if (dma_mapping_error(dev, pp->msi_data)) { > > - dev_err(dev, "Failed to map MSI data\n"); > > - __free_page(pp->msi_page); > > - pp->msi_page = NULL; > > - return; > > - } > > - msi_target = (u64)pp->msi_data; > > - > > /* Program the msi_data */ > > dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_LO, lower_32_bits(msi_target)); > > dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_HI, upper_32_bits(msi_target)); > > @@ -394,6 +387,16 @@ int dw_pcie_host_init(struct pcie_port *pp) > > irq_set_chained_handler_and_data(pp->msi_irq, > > dw_chained_msi_isr, > > pp); > > + > > + pp->msi_data = dma_map_single_attrs(pci->dev, &pp->msi_msg, > > + sizeof(pp->msi_msg), > > + DMA_FROM_DEVICE, > > + DMA_ATTR_SKIP_CPU_SYNC); > > + if (dma_mapping_error(pci->dev, pp->msi_data)) { > > + dev_err(pci->dev, "Failed to map MSI data\n"); > > + pp->msi_data = 0; > > + goto err_free_msi; > > + } > > } else { > > ret = pp->ops->msi_host_init(pp); > > if (ret < 0) > > diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h > > index 97c7063b9e89..9d2f511f13fa 100644 > > --- a/drivers/pci/controller/dwc/pcie-designware.h > > +++ b/drivers/pci/controller/dwc/pcie-designware.h > > @@ -190,8 +190,8 @@ struct pcie_port { > > int msi_irq; > > struct irq_domain *irq_domain; > > struct irq_domain *msi_domain; > > + u16 msi_msg; > > dma_addr_t msi_data; > > - struct page *msi_page; > > struct irq_chip *msi_irq_chip; > > u32 num_vectors; > > u32 irq_mask[MAX_MSI_CTRLS]; > > > > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
On 2020-10-14 17:52, Ard Biesheuvel wrote: > On Mon, 12 Oct 2020 at 13:38, Robin Murphy <robin.murphy@arm.com> wrote: >> >> On 2020-10-09 08:55, Jisheng Zhang wrote: >>> Currently, dw_pcie_msi_init() allocates and maps page for msi, then >>> program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex >>> may lose power during suspend-to-RAM, so when we resume, we want to >>> redo the latter but not the former. If designware based driver (for >>> example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the >>> msi page will be leaked. >>> >>> As pointed out by Rob and Ard, there's no need to allocate a page for >>> the MSI address, we could use an address in the driver data. >>> >>> To avoid map the MSI msg again during resume, we move the map MSI msg >>> from dw_pcie_msi_init() to dw_pcie_host_init(). >> >> You should move the unmap there as well. As soon as you know what the >> relevant address would be if you *were* to do DMA to this location, then >> the exercise is complete. Leaving it mapped for the lifetime of the >> device in order to do not-DMA to it seems questionable (and represents >> technically incorrect API usage without at least a sync_for_cpu call >> before any other access to the data). >> >> Another point of note is that using streaming DMA mappings at all is a >> bit fragile (regardless of this change). If the host controller itself >> has a limited DMA mask relative to physical memory (which integrators >> still seem to keep doing...) then you could end up punching your MSI >> hole right in the middle of the SWIOTLB bounce buffer, where it's then >> almost *guaranteed* to interfere with real DMA :( >> > > Wouldn't it be the unmap you are suggesting that would create this > problem? If the bounce buffer is never released, the fake MSI doorbell > address can never conflict with any other DMA mappings. True, it's just that hogging a SWIOTLB slot for effectively the entire lifetime of the system is such an anathema that it didn't even cross my mind... Ugh, so I guess the patch as it stands does actually work out more robust than I gave it credit for, even if the reason is horrid... oh well :) Robin. >> If no DWC users have that problem and the current code is working well >> enough, then I see little reason not to make this partucular change to >> tidy up the implementation, just bear in mind that there's always the >> possibility of having to come back and change it yet again in future to >> make it more robust. I had it in mind that this trick was done with a >> coherent DMA allocation, which would be safe from addressing problems >> but would need to be kept around for the lifetime of the device, but >> maybe that was a different driver :/ >> >> Robin. >> >>> Suggested-by: Rob Herring <robh@kernel.org> >>> Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com> >>> Reviewed-by: Rob Herring <robh@kernel.org> >>> --- >>> drivers/pci/controller/dwc/pci-dra7xx.c | 18 +++++++++- >>> .../pci/controller/dwc/pcie-designware-host.c | 33 ++++++++++--------- >>> drivers/pci/controller/dwc/pcie-designware.h | 2 +- >>> 3 files changed, 36 insertions(+), 17 deletions(-) >>> >>> diff --git a/drivers/pci/controller/dwc/pci-dra7xx.c b/drivers/pci/controller/dwc/pci-dra7xx.c >>> index 8f0b6d644e4b..6d012d2b1e90 100644 >>> --- a/drivers/pci/controller/dwc/pci-dra7xx.c >>> +++ b/drivers/pci/controller/dwc/pci-dra7xx.c >>> @@ -466,7 +466,9 @@ static struct irq_chip dra7xx_pci_msi_bottom_irq_chip = { >>> static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) >>> { >>> struct dw_pcie *pci = to_dw_pcie_from_pp(pp); >>> + struct device *dev = pci->dev; >>> u32 ctrl, num_ctrls; >>> + int ret; >>> >>> pp->msi_irq_chip = &dra7xx_pci_msi_bottom_irq_chip; >>> >>> @@ -482,7 +484,21 @@ static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) >>> ~0); >>> } >>> >>> - return dw_pcie_allocate_domains(pp); >>> + ret = dw_pcie_allocate_domains(pp); >>> + if (ret) >>> + return ret; >>> + >>> + pp->msi_data = dma_map_single_attrs(dev, &pp->msi_msg, >>> + sizeof(pp->msi_msg), >>> + DMA_FROM_DEVICE, >>> + DMA_ATTR_SKIP_CPU_SYNC); >>> + ret = dma_mapping_error(dev, pp->msi_data); >>> + if (ret) { >>> + dev_err(dev, "Failed to map MSI data\n"); >>> + pp->msi_data = 0; >>> + dw_pcie_free_msi(pp); >>> + } >>> + return ret; >>> } >>> >>> static const struct dw_pcie_host_ops dra7xx_pcie_host_ops = { >>> diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c >>> index d3e9ea11ce9e..d02c7e74738d 100644 >>> --- a/drivers/pci/controller/dwc/pcie-designware-host.c >>> +++ b/drivers/pci/controller/dwc/pcie-designware-host.c >>> @@ -266,30 +266,23 @@ void dw_pcie_free_msi(struct pcie_port *pp) >>> irq_domain_remove(pp->msi_domain); >>> irq_domain_remove(pp->irq_domain); >>> >>> - if (pp->msi_page) >>> - __free_page(pp->msi_page); >>> + if (pp->msi_data) { >>> + struct dw_pcie *pci = to_dw_pcie_from_pp(pp); >>> + struct device *dev = pci->dev; >>> + >>> + dma_unmap_single_attrs(dev, pp->msi_data, sizeof(pp->msi_msg), >>> + DMA_FROM_DEVICE, DMA_ATTR_SKIP_CPU_SYNC); >>> + } >>> } >>> >>> void dw_pcie_msi_init(struct pcie_port *pp) >>> { >>> struct dw_pcie *pci = to_dw_pcie_from_pp(pp); >>> - struct device *dev = pci->dev; >>> - u64 msi_target; >>> + u64 msi_target = (u64)pp->msi_data; >>> >>> if (!IS_ENABLED(CONFIG_PCI_MSI)) >>> return; >>> >>> - pp->msi_page = alloc_page(GFP_KERNEL); >>> - pp->msi_data = dma_map_page(dev, pp->msi_page, 0, PAGE_SIZE, >>> - DMA_FROM_DEVICE); >>> - if (dma_mapping_error(dev, pp->msi_data)) { >>> - dev_err(dev, "Failed to map MSI data\n"); >>> - __free_page(pp->msi_page); >>> - pp->msi_page = NULL; >>> - return; >>> - } >>> - msi_target = (u64)pp->msi_data; >>> - >>> /* Program the msi_data */ >>> dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_LO, lower_32_bits(msi_target)); >>> dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_HI, upper_32_bits(msi_target)); >>> @@ -394,6 +387,16 @@ int dw_pcie_host_init(struct pcie_port *pp) >>> irq_set_chained_handler_and_data(pp->msi_irq, >>> dw_chained_msi_isr, >>> pp); >>> + >>> + pp->msi_data = dma_map_single_attrs(pci->dev, &pp->msi_msg, >>> + sizeof(pp->msi_msg), >>> + DMA_FROM_DEVICE, >>> + DMA_ATTR_SKIP_CPU_SYNC); >>> + if (dma_mapping_error(pci->dev, pp->msi_data)) { >>> + dev_err(pci->dev, "Failed to map MSI data\n"); >>> + pp->msi_data = 0; >>> + goto err_free_msi; >>> + } >>> } else { >>> ret = pp->ops->msi_host_init(pp); >>> if (ret < 0) >>> diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h >>> index 97c7063b9e89..9d2f511f13fa 100644 >>> --- a/drivers/pci/controller/dwc/pcie-designware.h >>> +++ b/drivers/pci/controller/dwc/pcie-designware.h >>> @@ -190,8 +190,8 @@ struct pcie_port { >>> int msi_irq; >>> struct irq_domain *irq_domain; >>> struct irq_domain *msi_domain; >>> + u16 msi_msg; >>> dma_addr_t msi_data; >>> - struct page *msi_page; >>> struct irq_chip *msi_irq_chip; >>> u32 num_vectors; >>> u32 irq_mask[MAX_MSI_CTRLS]; >>> >> >> _______________________________________________ >> linux-arm-kernel mailing list >> linux-arm-kernel@lists.infradead.org >> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
diff --git a/drivers/pci/controller/dwc/pci-dra7xx.c b/drivers/pci/controller/dwc/pci-dra7xx.c index 8f0b6d644e4b..6d012d2b1e90 100644 --- a/drivers/pci/controller/dwc/pci-dra7xx.c +++ b/drivers/pci/controller/dwc/pci-dra7xx.c @@ -466,7 +466,9 @@ static struct irq_chip dra7xx_pci_msi_bottom_irq_chip = { static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) { struct dw_pcie *pci = to_dw_pcie_from_pp(pp); + struct device *dev = pci->dev; u32 ctrl, num_ctrls; + int ret; pp->msi_irq_chip = &dra7xx_pci_msi_bottom_irq_chip; @@ -482,7 +484,21 @@ static int dra7xx_pcie_msi_host_init(struct pcie_port *pp) ~0); } - return dw_pcie_allocate_domains(pp); + ret = dw_pcie_allocate_domains(pp); + if (ret) + return ret; + + pp->msi_data = dma_map_single_attrs(dev, &pp->msi_msg, + sizeof(pp->msi_msg), + DMA_FROM_DEVICE, + DMA_ATTR_SKIP_CPU_SYNC); + ret = dma_mapping_error(dev, pp->msi_data); + if (ret) { + dev_err(dev, "Failed to map MSI data\n"); + pp->msi_data = 0; + dw_pcie_free_msi(pp); + } + return ret; } static const struct dw_pcie_host_ops dra7xx_pcie_host_ops = { diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c index d3e9ea11ce9e..d02c7e74738d 100644 --- a/drivers/pci/controller/dwc/pcie-designware-host.c +++ b/drivers/pci/controller/dwc/pcie-designware-host.c @@ -266,30 +266,23 @@ void dw_pcie_free_msi(struct pcie_port *pp) irq_domain_remove(pp->msi_domain); irq_domain_remove(pp->irq_domain); - if (pp->msi_page) - __free_page(pp->msi_page); + if (pp->msi_data) { + struct dw_pcie *pci = to_dw_pcie_from_pp(pp); + struct device *dev = pci->dev; + + dma_unmap_single_attrs(dev, pp->msi_data, sizeof(pp->msi_msg), + DMA_FROM_DEVICE, DMA_ATTR_SKIP_CPU_SYNC); + } } void dw_pcie_msi_init(struct pcie_port *pp) { struct dw_pcie *pci = to_dw_pcie_from_pp(pp); - struct device *dev = pci->dev; - u64 msi_target; + u64 msi_target = (u64)pp->msi_data; if (!IS_ENABLED(CONFIG_PCI_MSI)) return; - pp->msi_page = alloc_page(GFP_KERNEL); - pp->msi_data = dma_map_page(dev, pp->msi_page, 0, PAGE_SIZE, - DMA_FROM_DEVICE); - if (dma_mapping_error(dev, pp->msi_data)) { - dev_err(dev, "Failed to map MSI data\n"); - __free_page(pp->msi_page); - pp->msi_page = NULL; - return; - } - msi_target = (u64)pp->msi_data; - /* Program the msi_data */ dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_LO, lower_32_bits(msi_target)); dw_pcie_writel_dbi(pci, PCIE_MSI_ADDR_HI, upper_32_bits(msi_target)); @@ -394,6 +387,16 @@ int dw_pcie_host_init(struct pcie_port *pp) irq_set_chained_handler_and_data(pp->msi_irq, dw_chained_msi_isr, pp); + + pp->msi_data = dma_map_single_attrs(pci->dev, &pp->msi_msg, + sizeof(pp->msi_msg), + DMA_FROM_DEVICE, + DMA_ATTR_SKIP_CPU_SYNC); + if (dma_mapping_error(pci->dev, pp->msi_data)) { + dev_err(pci->dev, "Failed to map MSI data\n"); + pp->msi_data = 0; + goto err_free_msi; + } } else { ret = pp->ops->msi_host_init(pp); if (ret < 0) diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h index 97c7063b9e89..9d2f511f13fa 100644 --- a/drivers/pci/controller/dwc/pcie-designware.h +++ b/drivers/pci/controller/dwc/pcie-designware.h @@ -190,8 +190,8 @@ struct pcie_port { int msi_irq; struct irq_domain *irq_domain; struct irq_domain *msi_domain; + u16 msi_msg; dma_addr_t msi_data; - struct page *msi_page; struct irq_chip *msi_irq_chip; u32 num_vectors; u32 irq_mask[MAX_MSI_CTRLS];