| Message ID | 20230103073401.1256736-3-chenhuacai@loongson.cn |
|---|---|
| State | New |
| Headers | show |
| Series | PCI: Add two Loongson's LS7A quirks | expand |
On Tue, Jan 03, 2023 at 03:34:01PM +0800, Huacai Chen wrote: > cc27b735ad3a7557 ("PCI/portdrv: Turn off PCIe services during shutdown") > causes poweroff/reboot failure on systems with LS7A chipset. We found > that if we remove "pci_command &= ~PCI_COMMAND_MASTER" in do_pci_disable > _device(), it can work well. The hardware engineer says that the root > cause is that CPU is still accessing PCIe devices while poweroff/reboot, Did you ever figure out what these CPU accesses are? If we call the Root Port .shutdown() method, and later access a downstream device, that seems like a problem in itself. At least, we should understand exactly *why* we access that downstream device. To be clear, cc27b735ad3a does not cause the failure. IIUC, the cause is: - CPU issues MMIO read to device below Root Port - LS7A Root Port fails to forward transaction to secondary bus because of LS7A Bus Master defect - CPU hangs waiting for response to MMIO read > and if we disable the Bus Master Bit at this time, the PCIe controller > doesn't forward requests to downstream devices, and also does not send > TIMEOUT to CPU, which causes CPU wait forever (hardware deadlock). This > behavior is a PCIe protocol violation (Bus Master should not be involved > in CPU MMIO transactions), and it will be fixed in new revisions of > hardware (add timeout mechanism for CPU read request, whether or not Bus > Master bit is cleared). > > On some x86 platforms, radeon/amdgpu devices can cause similar problems > [1][2]. Once before I wanted to make a single patch to solve "all of > these problems" together, but it seems unreasonable because maybe they > are not exactly the same problem. I don't know what any of these problems are. Neither one of these bug reports has a root cause analysis, and it's not obvious how they're connected to this patch. > So, this patch add a new function > pcie_portdrv_shutdown(), a slight modified copy of pcie_portdrv_remove() > dedicated for the shutdown path, and then add a quirk just for LS7A to > avoid clearing Bus Master bit in pcie_portdrv_shutdown(). Leave other > platforms behave as before. Nit: don't break function names across lines ("do_pci_disable_device()"). > [1] https://bugs.freedesktop.org/show_bug.cgi?id=97980 > [2] https://bugs.freedesktop.org/show_bug.cgi?id=98638 > > Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> > --- > drivers/pci/controller/pci-loongson.c | 17 +++++++++++++++++ > drivers/pci/pcie/portdrv.c | 21 +++++++++++++++++++-- > include/linux/pci.h | 1 + > 3 files changed, 37 insertions(+), 2 deletions(-) > > diff --git a/drivers/pci/controller/pci-loongson.c b/drivers/pci/controller/pci-loongson.c > index 759ec211c17b..641308ba4126 100644 > --- a/drivers/pci/controller/pci-loongson.c > +++ b/drivers/pci/controller/pci-loongson.c > @@ -93,6 +93,23 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > DEV_PCIE_PORT_2, loongson_mrrs_quirk); > > +static void loongson_bmaster_quirk(struct pci_dev *pdev) > +{ > + /* > + * Some Loongson PCIe ports will cause CPU deadlock if disable > + * the Bus Master bit during poweroff/reboot. This is not actually true, as far as I can see. It's not turning off Bus Master that causes the problem; it's the MMIO read to a downstream device when the Root Port has bus mastering disabled that causes the problem. > + */ > + struct pci_host_bridge *bridge = pci_find_host_bridge(pdev->bus); > + > + bridge->no_dis_bmaster = 1; > +} > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > + DEV_PCIE_PORT_0, loongson_bmaster_quirk); > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > + DEV_PCIE_PORT_1, loongson_bmaster_quirk); > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > + DEV_PCIE_PORT_2, loongson_bmaster_quirk); > + > static void loongson_pci_pin_quirk(struct pci_dev *pdev) > { > pdev->pin = 1 + (PCI_FUNC(pdev->devfn) & 3); > diff --git a/drivers/pci/pcie/portdrv.c b/drivers/pci/pcie/portdrv.c > index 2cc2e60bcb39..96f45c444422 100644 > --- a/drivers/pci/pcie/portdrv.c > +++ b/drivers/pci/pcie/portdrv.c > @@ -501,7 +501,6 @@ static void pcie_port_device_remove(struct pci_dev *dev) > { > device_for_each_child(&dev->dev, NULL, remove_iter); > pci_free_irq_vectors(dev); > - pci_disable_device(dev); > } > > /** > @@ -727,6 +726,24 @@ static void pcie_portdrv_remove(struct pci_dev *dev) > } > > pcie_port_device_remove(dev); > + > + pci_disable_device(dev); > +} > + > +static void pcie_portdrv_shutdown(struct pci_dev *dev) > +{ > + struct pci_host_bridge *bridge = pci_find_host_bridge(dev->bus); > + > + if (pci_bridge_d3_possible(dev)) { > + pm_runtime_forbid(&dev->dev); > + pm_runtime_get_noresume(&dev->dev); > + pm_runtime_dont_use_autosuspend(&dev->dev); > + } > + > + pcie_port_device_remove(dev); > + > + if (!bridge->no_dis_bmaster) > + pci_disable_device(dev); > } > > static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev, > @@ -777,7 +794,7 @@ static struct pci_driver pcie_portdriver = { > > .probe = pcie_portdrv_probe, > .remove = pcie_portdrv_remove, > - .shutdown = pcie_portdrv_remove, > + .shutdown = pcie_portdrv_shutdown, > > .err_handler = &pcie_portdrv_err_handler, > > diff --git a/include/linux/pci.h b/include/linux/pci.h > index 3df2049ec4a8..a64dbcb89231 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -573,6 +573,7 @@ struct pci_host_bridge { > unsigned int ignore_reset_delay:1; /* For entire hierarchy */ > unsigned int no_ext_tags:1; /* No Extended Tags */ > unsigned int no_inc_mrrs:1; /* No Increase MRRS */ > + unsigned int no_dis_bmaster:1; /* No Disable Bus Master */ > unsigned int native_aer:1; /* OS may use PCIe AER */ > unsigned int native_pcie_hotplug:1; /* OS may use PCIe hotplug */ > unsigned int native_shpc_hotplug:1; /* OS may use SHPC hotplug */ > -- > 2.31.1 >
Hi, Bjorn, On Thu, Jan 5, 2023 at 2:37 AM Bjorn Helgaas <helgaas@kernel.org> wrote: > > On Tue, Jan 03, 2023 at 03:34:01PM +0800, Huacai Chen wrote: > > cc27b735ad3a7557 ("PCI/portdrv: Turn off PCIe services during shutdown") > > causes poweroff/reboot failure on systems with LS7A chipset. We found > > that if we remove "pci_command &= ~PCI_COMMAND_MASTER" in do_pci_disable > > _device(), it can work well. The hardware engineer says that the root > > cause is that CPU is still accessing PCIe devices while poweroff/reboot, > > Did you ever figure out what these CPU accesses are? If we call the > Root Port .shutdown() method, and later access a downstream device, > that seems like a problem in itself. At least, we should understand > exactly *why* we access that downstream device. Maybe I failed to get the key point, but from my point of view, the root cause is clear in previous discussions: https://lore.kernel.org/linux-pci/CAAhV-H5uT+wDRkVbW_o1hG2u0rtv6FFABTymL1VdjMMD_UEN+Q@mail.gmail.com/ https://lore.kernel.org/linux-pci/20220617113708.GA1177054@bhelgaas/ https://lore.kernel.org/linux-pci/CAAhV-H6raQnXZ4ZZRq19cugew26wXYONctcFO0392gZ00LC6bw@mail.gmail.com/ > > To be clear, cc27b735ad3a does not cause the failure. IIUC, the cause > is: cc27b735ad3a is not a bug, we refer to it just because we observe problems after it. > > - CPU issues MMIO read to device below Root Port > > - LS7A Root Port fails to forward transaction to secondary bus > because of LS7A Bus Master defect > > - CPU hangs waiting for response to MMIO read > > > and if we disable the Bus Master Bit at this time, the PCIe controller > > doesn't forward requests to downstream devices, and also does not send > > TIMEOUT to CPU, which causes CPU wait forever (hardware deadlock). This > > behavior is a PCIe protocol violation (Bus Master should not be involved > > in CPU MMIO transactions), and it will be fixed in new revisions of > > hardware (add timeout mechanism for CPU read request, whether or not Bus > > Master bit is cleared). > > > > On some x86 platforms, radeon/amdgpu devices can cause similar problems > > [1][2]. Once before I wanted to make a single patch to solve "all of > > these problems" together, but it seems unreasonable because maybe they > > are not exactly the same problem. > > I don't know what any of these problems are. Neither one of these bug > reports has a root cause analysis, and it's not obvious how they're > connected to this patch. radeon/amdgpu devices cause problems and this patch can solve it. But..., it is very hard to distinguish "whether they are exactly the same problem". > > > So, this patch add a new function > > pcie_portdrv_shutdown(), a slight modified copy of pcie_portdrv_remove() > > dedicated for the shutdown path, and then add a quirk just for LS7A to > > avoid clearing Bus Master bit in pcie_portdrv_shutdown(). Leave other > > platforms behave as before. > > Nit: don't break function names across lines ("do_pci_disable_device()"). OK, I will improve it. > > > [1] https://bugs.freedesktop.org/show_bug.cgi?id=97980 > > [2] https://bugs.freedesktop.org/show_bug.cgi?id=98638 > > > > Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> > > --- > > drivers/pci/controller/pci-loongson.c | 17 +++++++++++++++++ > > drivers/pci/pcie/portdrv.c | 21 +++++++++++++++++++-- > > include/linux/pci.h | 1 + > > 3 files changed, 37 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/pci/controller/pci-loongson.c b/drivers/pci/controller/pci-loongson.c > > index 759ec211c17b..641308ba4126 100644 > > --- a/drivers/pci/controller/pci-loongson.c > > +++ b/drivers/pci/controller/pci-loongson.c > > @@ -93,6 +93,23 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > > DEV_PCIE_PORT_2, loongson_mrrs_quirk); > > > > +static void loongson_bmaster_quirk(struct pci_dev *pdev) > > +{ > > + /* > > + * Some Loongson PCIe ports will cause CPU deadlock if disable > > + * the Bus Master bit during poweroff/reboot. > > This is not actually true, as far as I can see. > > It's not turning off Bus Master that causes the problem; it's the MMIO > read to a downstream device when the Root Port has bus mastering > disabled that causes the problem. OK, I will improve the comments. Huacai > > > + */ > > + struct pci_host_bridge *bridge = pci_find_host_bridge(pdev->bus); > > + > > + bridge->no_dis_bmaster = 1; > > +} > > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > > + DEV_PCIE_PORT_0, loongson_bmaster_quirk); > > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > > + DEV_PCIE_PORT_1, loongson_bmaster_quirk); > > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, > > + DEV_PCIE_PORT_2, loongson_bmaster_quirk); > > + > > static void loongson_pci_pin_quirk(struct pci_dev *pdev) > > { > > pdev->pin = 1 + (PCI_FUNC(pdev->devfn) & 3); > > diff --git a/drivers/pci/pcie/portdrv.c b/drivers/pci/pcie/portdrv.c > > index 2cc2e60bcb39..96f45c444422 100644 > > --- a/drivers/pci/pcie/portdrv.c > > +++ b/drivers/pci/pcie/portdrv.c > > @@ -501,7 +501,6 @@ static void pcie_port_device_remove(struct pci_dev *dev) > > { > > device_for_each_child(&dev->dev, NULL, remove_iter); > > pci_free_irq_vectors(dev); > > - pci_disable_device(dev); > > } > > > > /** > > @@ -727,6 +726,24 @@ static void pcie_portdrv_remove(struct pci_dev *dev) > > } > > > > pcie_port_device_remove(dev); > > + > > + pci_disable_device(dev); > > +} > > + > > +static void pcie_portdrv_shutdown(struct pci_dev *dev) > > +{ > > + struct pci_host_bridge *bridge = pci_find_host_bridge(dev->bus); > > + > > + if (pci_bridge_d3_possible(dev)) { > > + pm_runtime_forbid(&dev->dev); > > + pm_runtime_get_noresume(&dev->dev); > > + pm_runtime_dont_use_autosuspend(&dev->dev); > > + } > > + > > + pcie_port_device_remove(dev); > > + > > + if (!bridge->no_dis_bmaster) > > + pci_disable_device(dev); > > } > > > > static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev, > > @@ -777,7 +794,7 @@ static struct pci_driver pcie_portdriver = { > > > > .probe = pcie_portdrv_probe, > > .remove = pcie_portdrv_remove, > > - .shutdown = pcie_portdrv_remove, > > + .shutdown = pcie_portdrv_shutdown, > > > > .err_handler = &pcie_portdrv_err_handler, > > > > diff --git a/include/linux/pci.h b/include/linux/pci.h > > index 3df2049ec4a8..a64dbcb89231 100644 > > --- a/include/linux/pci.h > > +++ b/include/linux/pci.h > > @@ -573,6 +573,7 @@ struct pci_host_bridge { > > unsigned int ignore_reset_delay:1; /* For entire hierarchy */ > > unsigned int no_ext_tags:1; /* No Extended Tags */ > > unsigned int no_inc_mrrs:1; /* No Increase MRRS */ > > + unsigned int no_dis_bmaster:1; /* No Disable Bus Master */ > > unsigned int native_aer:1; /* OS may use PCIe AER */ > > unsigned int native_pcie_hotplug:1; /* OS may use PCIe hotplug */ > > unsigned int native_shpc_hotplug:1; /* OS may use SHPC hotplug */ > > -- > > 2.31.1 > >
On Thu, Jan 05, 2023 at 10:49:53AM +0800, Huacai Chen wrote: > On Thu, Jan 5, 2023 at 2:37 AM Bjorn Helgaas <helgaas@kernel.org> wrote: > > On Tue, Jan 03, 2023 at 03:34:01PM +0800, Huacai Chen wrote: > > > cc27b735ad3a7557 ("PCI/portdrv: Turn off PCIe services during shutdown") > > > causes poweroff/reboot failure on systems with LS7A chipset. We found > > > that if we remove "pci_command &= ~PCI_COMMAND_MASTER" in do_pci_disable > > > _device(), it can work well. The hardware engineer says that the root > > > cause is that CPU is still accessing PCIe devices while poweroff/reboot, > > > > Did you ever figure out what these CPU accesses are? If we call the > > Root Port .shutdown() method, and later access a downstream device, > > that seems like a problem in itself. At least, we should understand > > exactly *why* we access that downstream device. > > Maybe I failed to get the key point, but from my point of view, the > root cause is clear in previous discussions: > https://lore.kernel.org/linux-pci/CAAhV-H5uT+wDRkVbW_o1hG2u0rtv6FFABTymL1VdjMMD_UEN+Q@mail.gmail.com/ > https://lore.kernel.org/linux-pci/20220617113708.GA1177054@bhelgaas/ > https://lore.kernel.org/linux-pci/CAAhV-H6raQnXZ4ZZRq19cugew26wXYONctcFO0392gZ00LC6bw@mail.gmail.com/ That's great, but the root cause should be summarized here in the commit log. > > To be clear, cc27b735ad3a does not cause the failure. IIUC, the cause > > is: > > cc27b735ad3a is not a bug, we refer to it just because we observe > problems after it. Right. But you said "cc27b735ad3a ... causes failure," which is not quite true. cc27b735ad3a may *expose* an LS7A hardware defect that previously didn't cause a problem, but I don't want to blame cc27b735ad3a for that hardware issue. Bjorn
On Thu, Jan 5, 2023 at 12:01 PM Bjorn Helgaas <helgaas@kernel.org> wrote: > > On Thu, Jan 05, 2023 at 10:49:53AM +0800, Huacai Chen wrote: > > On Thu, Jan 5, 2023 at 2:37 AM Bjorn Helgaas <helgaas@kernel.org> wrote: > > > On Tue, Jan 03, 2023 at 03:34:01PM +0800, Huacai Chen wrote: > > > > cc27b735ad3a7557 ("PCI/portdrv: Turn off PCIe services during shutdown") > > > > causes poweroff/reboot failure on systems with LS7A chipset. We found > > > > that if we remove "pci_command &= ~PCI_COMMAND_MASTER" in do_pci_disable > > > > _device(), it can work well. The hardware engineer says that the root > > > > cause is that CPU is still accessing PCIe devices while poweroff/reboot, > > > > > > Did you ever figure out what these CPU accesses are? If we call the > > > Root Port .shutdown() method, and later access a downstream device, > > > that seems like a problem in itself. At least, we should understand > > > exactly *why* we access that downstream device. > > > > Maybe I failed to get the key point, but from my point of view, the > > root cause is clear in previous discussions: > > https://lore.kernel.org/linux-pci/CAAhV-H5uT+wDRkVbW_o1hG2u0rtv6FFABTymL1VdjMMD_UEN+Q@mail.gmail.com/ > > https://lore.kernel.org/linux-pci/20220617113708.GA1177054@bhelgaas/ > > https://lore.kernel.org/linux-pci/CAAhV-H6raQnXZ4ZZRq19cugew26wXYONctcFO0392gZ00LC6bw@mail.gmail.com/ > > That's great, but the root cause should be summarized here in the > commit log. OK, I will update the commit log. > > > > To be clear, cc27b735ad3a does not cause the failure. IIUC, the cause > > > is: > > > > cc27b735ad3a is not a bug, we refer to it just because we observe > > problems after it. > > Right. But you said "cc27b735ad3a ... causes failure," which is not > quite true. cc27b735ad3a may *expose* an LS7A hardware defect that > previously didn't cause a problem, but I don't want to blame > cc27b735ad3a for that hardware issue. OK, got it. Huacai > > Bjorn
diff --git a/drivers/pci/controller/pci-loongson.c b/drivers/pci/controller/pci-loongson.c index 759ec211c17b..641308ba4126 100644 --- a/drivers/pci/controller/pci-loongson.c +++ b/drivers/pci/controller/pci-loongson.c @@ -93,6 +93,23 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, DEV_PCIE_PORT_2, loongson_mrrs_quirk); +static void loongson_bmaster_quirk(struct pci_dev *pdev) +{ + /* + * Some Loongson PCIe ports will cause CPU deadlock if disable + * the Bus Master bit during poweroff/reboot. + */ + struct pci_host_bridge *bridge = pci_find_host_bridge(pdev->bus); + + bridge->no_dis_bmaster = 1; +} +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, + DEV_PCIE_PORT_0, loongson_bmaster_quirk); +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, + DEV_PCIE_PORT_1, loongson_bmaster_quirk); +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, + DEV_PCIE_PORT_2, loongson_bmaster_quirk); + static void loongson_pci_pin_quirk(struct pci_dev *pdev) { pdev->pin = 1 + (PCI_FUNC(pdev->devfn) & 3); diff --git a/drivers/pci/pcie/portdrv.c b/drivers/pci/pcie/portdrv.c index 2cc2e60bcb39..96f45c444422 100644 --- a/drivers/pci/pcie/portdrv.c +++ b/drivers/pci/pcie/portdrv.c @@ -501,7 +501,6 @@ static void pcie_port_device_remove(struct pci_dev *dev) { device_for_each_child(&dev->dev, NULL, remove_iter); pci_free_irq_vectors(dev); - pci_disable_device(dev); } /** @@ -727,6 +726,24 @@ static void pcie_portdrv_remove(struct pci_dev *dev) } pcie_port_device_remove(dev); + + pci_disable_device(dev); +} + +static void pcie_portdrv_shutdown(struct pci_dev *dev) +{ + struct pci_host_bridge *bridge = pci_find_host_bridge(dev->bus); + + if (pci_bridge_d3_possible(dev)) { + pm_runtime_forbid(&dev->dev); + pm_runtime_get_noresume(&dev->dev); + pm_runtime_dont_use_autosuspend(&dev->dev); + } + + pcie_port_device_remove(dev); + + if (!bridge->no_dis_bmaster) + pci_disable_device(dev); } static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev, @@ -777,7 +794,7 @@ static struct pci_driver pcie_portdriver = { .probe = pcie_portdrv_probe, .remove = pcie_portdrv_remove, - .shutdown = pcie_portdrv_remove, + .shutdown = pcie_portdrv_shutdown, .err_handler = &pcie_portdrv_err_handler, diff --git a/include/linux/pci.h b/include/linux/pci.h index 3df2049ec4a8..a64dbcb89231 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -573,6 +573,7 @@ struct pci_host_bridge { unsigned int ignore_reset_delay:1; /* For entire hierarchy */ unsigned int no_ext_tags:1; /* No Extended Tags */ unsigned int no_inc_mrrs:1; /* No Increase MRRS */ + unsigned int no_dis_bmaster:1; /* No Disable Bus Master */ unsigned int native_aer:1; /* OS may use PCIe AER */ unsigned int native_pcie_hotplug:1; /* OS may use PCIe hotplug */ unsigned int native_shpc_hotplug:1; /* OS may use SHPC hotplug */
cc27b735ad3a7557 ("PCI/portdrv: Turn off PCIe services during shutdown") causes poweroff/reboot failure on systems with LS7A chipset. We found that if we remove "pci_command &= ~PCI_COMMAND_MASTER" in do_pci_disable _device(), it can work well. The hardware engineer says that the root cause is that CPU is still accessing PCIe devices while poweroff/reboot, and if we disable the Bus Master Bit at this time, the PCIe controller doesn't forward requests to downstream devices, and also does not send TIMEOUT to CPU, which causes CPU wait forever (hardware deadlock). This behavior is a PCIe protocol violation (Bus Master should not be involved in CPU MMIO transactions), and it will be fixed in new revisions of hardware (add timeout mechanism for CPU read request, whether or not Bus Master bit is cleared). On some x86 platforms, radeon/amdgpu devices can cause similar problems [1][2]. Once before I wanted to make a single patch to solve "all of these problems" together, but it seems unreasonable because maybe they are not exactly the same problem. So, this patch add a new function pcie_portdrv_shutdown(), a slight modified copy of pcie_portdrv_remove() dedicated for the shutdown path, and then add a quirk just for LS7A to avoid clearing Bus Master bit in pcie_portdrv_shutdown(). Leave other platforms behave as before. [1] https://bugs.freedesktop.org/show_bug.cgi?id=97980 [2] https://bugs.freedesktop.org/show_bug.cgi?id=98638 Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> --- drivers/pci/controller/pci-loongson.c | 17 +++++++++++++++++ drivers/pci/pcie/portdrv.c | 21 +++++++++++++++++++-- include/linux/pci.h | 1 + 3 files changed, 37 insertions(+), 2 deletions(-)