| Message ID | 1517415535-21850-1-git-send-email-poza@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Wed, Jan 31, 2018 at 09:48:55PM +0530, Oza Pawandeep wrote: > Current dpc driver acknowledge the interrupt in deferred work, which works > okay since LPI are edge triggered. > But when RP does not have MSI support, port service driver falls back to > legacy GIC SPI interrupts, and with current code we do not acknowledge the > interrupt and we get dpc interrupt storm. > This patch acknowledges the interrupt in interrupt handler. > > Signed-off-by: Oza Pawandeep <poza@codeaurora.org> Thanks, looks good to me. Reviewed-by: Keith Busch <keith.busch@intel.com>
Hi Oza, On Wed, Jan 31, 2018 at 09:48:55PM +0530, Oza Pawandeep wrote: > Current dpc driver acknowledge the interrupt in deferred work, which works > okay since LPI are edge triggered. > But when RP does not have MSI support, port service driver falls back to > legacy GIC SPI interrupts, and with current code we do not acknowledge the > interrupt and we get dpc interrupt storm. > This patch acknowledges the interrupt in interrupt handler. Would you mind rephrasing this changelog in generic PCIe terms? E.g., "LPI" and "GIC SPI" are meaningless to me (I assume they're arm64-specific things), but I don't think it's any arm64-specific thing that motivates this change. IIUC, the issue should occur any time we're not using MSI, so we should be able to describe it in purely PCIe terms. s/dpc/DPC/ in text. It looks like this changelog is intended to be multiple paragraphs; if so, please leave a blank line between them. It's a little awkward to infer "this is a new paragraph" from the fact that "the previous line didn't look full" :) Please rebase to my "master" branch (v4.16-rc1). There was some minor change that made this not apply cleanly. > Signed-off-by: Oza Pawandeep <poza@codeaurora.org> > > diff --git a/drivers/pci/pcie/pcie-dpc.c b/drivers/pci/pcie/pcie-dpc.c > index 2d976a6..570b561 100644 > --- a/drivers/pci/pcie/pcie-dpc.c > +++ b/drivers/pci/pcie/pcie-dpc.c > @@ -134,7 +134,7 @@ static void interrupt_event_handler(struct work_struct *work) > } > > pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, > - PCI_EXP_DPC_STATUS_TRIGGER | PCI_EXP_DPC_STATUS_INTERRUPT); > + PCI_EXP_DPC_STATUS_TRIGGER); > } > > static void dpc_rp_pio_print_tlp_header(struct device *dev, > @@ -277,6 +277,8 @@ static irqreturn_t dpc_irq(int irq, void *context) > > schedule_work(&dpc->work); > } > + pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, > + PCI_EXP_DPC_STATUS_INTERRUPT); > return IRQ_HANDLED; > } > > -- > Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm Technologies, Inc., > a Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project. >
On 2018-03-01 21:08, Bjorn Helgaas wrote: > Hi Oza, > > On Wed, Jan 31, 2018 at 09:48:55PM +0530, Oza Pawandeep wrote: >> Current dpc driver acknowledge the interrupt in deferred work, which >> works >> okay since LPI are edge triggered. >> But when RP does not have MSI support, port service driver falls back >> to >> legacy GIC SPI interrupts, and with current code we do not acknowledge >> the >> interrupt and we get dpc interrupt storm. >> This patch acknowledges the interrupt in interrupt handler. > > Would you mind rephrasing this changelog in generic PCIe terms? E.g., > "LPI" and "GIC SPI" are meaningless to me (I assume they're > arm64-specific things), but I don't think it's any arm64-specific > thing that motivates this change. IIUC, the issue should occur any > time we're not using MSI, so we should be able to describe it in > purely PCIe terms. > > s/dpc/DPC/ in text. > > It looks like this changelog is intended to be multiple paragraphs; if > so, please leave a blank line between them. It's a little awkward to > infer "this is a new paragraph" from the fact that "the previous line > didn't look full" :) > > Please rebase to my "master" branch (v4.16-rc1). There was some minor > change that made this not apply cleanly. > sure will do these changes and re-post the same. Regards, Oza. >> Signed-off-by: Oza Pawandeep <poza@codeaurora.org> >> >> diff --git a/drivers/pci/pcie/pcie-dpc.c b/drivers/pci/pcie/pcie-dpc.c >> index 2d976a6..570b561 100644 >> --- a/drivers/pci/pcie/pcie-dpc.c >> +++ b/drivers/pci/pcie/pcie-dpc.c >> @@ -134,7 +134,7 @@ static void interrupt_event_handler(struct >> work_struct *work) >> } >> >> pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, >> - PCI_EXP_DPC_STATUS_TRIGGER | PCI_EXP_DPC_STATUS_INTERRUPT); >> + PCI_EXP_DPC_STATUS_TRIGGER); >> } >> >> static void dpc_rp_pio_print_tlp_header(struct device *dev, >> @@ -277,6 +277,8 @@ static irqreturn_t dpc_irq(int irq, void *context) >> >> schedule_work(&dpc->work); >> } >> + pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, >> + PCI_EXP_DPC_STATUS_INTERRUPT); >> return IRQ_HANDLED; >> } >> >> -- >> Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm >> Technologies, Inc., >> a Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a >> Linux Foundation Collaborative Project. >>
On 2018-03-01 21:08, Bjorn Helgaas wrote: > Hi Oza, > > On Wed, Jan 31, 2018 at 09:48:55PM +0530, Oza Pawandeep wrote: >> Current dpc driver acknowledge the interrupt in deferred work, which >> works >> okay since LPI are edge triggered. >> But when RP does not have MSI support, port service driver falls back >> to >> legacy GIC SPI interrupts, and with current code we do not acknowledge >> the >> interrupt and we get dpc interrupt storm. >> This patch acknowledges the interrupt in interrupt handler. > > Would you mind rephrasing this changelog in generic PCIe terms? E.g., > "LPI" and "GIC SPI" are meaningless to me (I assume they're > arm64-specific things), but I don't think it's any arm64-specific > thing that motivates this change. IIUC, the issue should occur any > time we're not using MSI, so we should be able to describe it in > purely PCIe terms. > > s/dpc/DPC/ in text. > > It looks like this changelog is intended to be multiple paragraphs; if > so, please leave a blank line between them. It's a little awkward to > infer "this is a new paragraph" from the fact that "the previous line > didn't look full" :) > > Please rebase to my "master" branch (v4.16-rc1). There was some minor > change that made this not apply cleanly. > >> Signed-off-by: Oza Pawandeep <poza@codeaurora.org> >> >> diff --git a/drivers/pci/pcie/pcie-dpc.c b/drivers/pci/pcie/pcie-dpc.c >> index 2d976a6..570b561 100644 >> --- a/drivers/pci/pcie/pcie-dpc.c >> +++ b/drivers/pci/pcie/pcie-dpc.c >> @@ -134,7 +134,7 @@ static void interrupt_event_handler(struct >> work_struct *work) >> } >> >> pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, >> - PCI_EXP_DPC_STATUS_TRIGGER | PCI_EXP_DPC_STATUS_INTERRUPT); >> + PCI_EXP_DPC_STATUS_TRIGGER); >> } >> >> static void dpc_rp_pio_print_tlp_header(struct device *dev, >> @@ -277,6 +277,8 @@ static irqreturn_t dpc_irq(int irq, void *context) >> >> schedule_work(&dpc->work); >> } >> + pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, >> + PCI_EXP_DPC_STATUS_INTERRUPT); >> return IRQ_HANDLED; >> } >> >> -- >> Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm >> Technologies, Inc., >> a Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a >> Linux Foundation Collaborative Project. >> Hi Bjorn, Can I make this path on top of DPC/AER patch series ? Regards, Oza.
diff --git a/drivers/pci/pcie/pcie-dpc.c b/drivers/pci/pcie/pcie-dpc.c index 2d976a6..570b561 100644 --- a/drivers/pci/pcie/pcie-dpc.c +++ b/drivers/pci/pcie/pcie-dpc.c @@ -134,7 +134,7 @@ static void interrupt_event_handler(struct work_struct *work) } pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, - PCI_EXP_DPC_STATUS_TRIGGER | PCI_EXP_DPC_STATUS_INTERRUPT); + PCI_EXP_DPC_STATUS_TRIGGER); } static void dpc_rp_pio_print_tlp_header(struct device *dev, @@ -277,6 +277,8 @@ static irqreturn_t dpc_irq(int irq, void *context) schedule_work(&dpc->work); } + pci_write_config_word(pdev, dpc->cap_pos + PCI_EXP_DPC_STATUS, + PCI_EXP_DPC_STATUS_INTERRUPT); return IRQ_HANDLED; }
Current dpc driver acknowledge the interrupt in deferred work, which works okay since LPI are edge triggered. But when RP does not have MSI support, port service driver falls back to legacy GIC SPI interrupts, and with current code we do not acknowledge the interrupt and we get dpc interrupt storm. This patch acknowledges the interrupt in interrupt handler. Signed-off-by: Oza Pawandeep <poza@codeaurora.org>