[PATCH 1/1] pci: reset all pci endpoints to stop on going dma

Post by Li, Zhen-Hua
This is an update of the patch
https://lkml.org/lkml/2014/10/10/37
This patch is doing the reset works before the kdump kernel boots.

If I have said it once I have said it a thousand times.

Nacked-by: "Eric W. Biederman" <***@xmission.com>
Nacked-by: "Eric W. Biederman" <***@xmission.com>
Nacked-by: "Eric W. Biederman" <***@xmission.com>

It is absolutely inappropriate to do this in the kernel that has
crashed.

Either reserve a chunk of the iommu for use by the crash dump kernel,
or figure out how to do this during boot up.

But this is absolutely and totally inappropriate to do in crash_kexec.
The failure modes are all wrong. It will impact the reliability of our
crash dumps.

If an the code for a linux architecture is not structured in such a way
as to make it easy or straight forward to do this my sympathies you are
going to have do fix that linux port to be able to do things during
boot.

But things called from crash_kexec should be absolute necessities and I
do not see this as coming anywhere close to being something that is
impossible to do in the target kernel.

Eric

Post by Li, Zhen-Hua
On a Linux system with iommu supported and many PCI devices on it,
when kernel crashed and the kdump kernel boots with intel_iommu=on,
there may be some unexpected DMA requests on this adapter, which will
dmar: DRHD: handling fault status reg 102
dmar: DMAR:[DMA Read] Request device [41:00.0] fault addr fff81000
DMAR:[fault reason 01] Present bit in root entry is clear
This bug may happen on *any* PCI device.
static struct context_entry * device_to_context_entry(
struct intel_iommu *iommu, u8 bus, u8 devfn)
{
......
set_root_present(root);
......
}
device driver
intel_alloc_coherent
__intel_map_single
domain_context_mapping
domain_context_mapping_one
device_to_context_entry
This means, the present bit in root entry will not be set until the device
driver is loaded.
But in the kdump kernel, hardware devices are not aware that control has
transferred to the second kernel, and those drivers must initialize again.
Consequently there may be unexpected DMA requests from devices activity
initiated in the first kernel leading to the DMA Remapping errors in the
second kernel.
To fix this DMAR fault, we need to reset the bus that this device on. Reset
the device itself does not work.
https://lkml.org/lkml/2014/9/30/55
As in discussion, this bug may happen on *any* device, so we need to reset
all pci devices.
https://lkml.org/lkml/2013/5/14/9
According to the previous discussion, On sparc, the IOMMU is initialized
before PCI devices are enumerated, this patch does the resetting works before
the kdump kernel boots, so it can also fix the problems on sparc.
Add support for legacy PCI devices.
Use pci_try_reset_bus instead of do_downstream_device_reset.
Reset all PCI/PCIe deviecs in the first kernel, before kdump kernel boots.
Randy Wright corrects some misunderstanding in this description.
---
drivers/pci/pci.c | 82 +++++++++++++++++++++++++++++++++++++++++++++++++++++
include/linux/pci.h | 6 ++++
kernel/kexec.c | 2 ++
3 files changed, 90 insertions(+)
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 625a4ac..aa9192a 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -23,6 +23,7 @@
#include <linux/device.h>
#include <linux/pm_runtime.h>
#include <linux/pci_hotplug.h>
+#include <linux/crash_dump.h>
#include <asm-generic/pci-bridge.h>
#include <asm/setup.h>
#include "pci.h"
@@ -4466,6 +4467,87 @@ void __weak pci_fixup_cardbus(struct pci_bus *bus)
}
EXPORT_SYMBOL(pci_fixup_cardbus);
+/*
+ * Return true if dev is PCI root port or downstream port whose child is PCI
+ * endpoint except VGA device.
+ */
+static int __pci_dev_need_reset(struct pci_dev *dev)
+{
+ struct pci_bus *subordinate;
+ struct pci_dev *child;
+
+ if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
+ return 0;
+
+ if (pci_is_pcie(dev)) {
+ if ((pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT) &&
+ (pci_pcie_type(dev) != PCI_EXP_TYPE_DOWNSTREAM))
+ return 0;
+ }
+
+ subordinate = dev->subordinate;
+ list_for_each_entry(child, &subordinate->devices, bus_list) {
+ /* Don't reset switch, bridge, VGA device */
+ if ((child->hdr_type == PCI_HEADER_TYPE_BRIDGE) ||
+ ((child->class >> 16) == PCI_BASE_CLASS_BRIDGE) ||
+ ((child->class >> 16) == PCI_BASE_CLASS_DISPLAY))
+ return 0;
+
+ if (pci_is_pcie(child)) {
+ if ((pci_pcie_type(child) == PCI_EXP_TYPE_UPSTREAM) ||
+ (pci_pcie_type(child) == PCI_EXP_TYPE_PCI_BRIDGE))
+ return 0;
+ }
+ }
+
+ return 1;
+}
+
+struct pci_dev_reset_entry {
+ struct list_head list;
+ struct pci_dev *dev;
+};
+int pci_reset_endpoints(void)
+{
+ struct pci_dev *dev = NULL;
+ struct pci_dev_reset_entry *pdev_entry, *tmp;
+ struct pci_bus *subordinate = NULL;
+ int has_it;
+
+ LIST_HEAD(pdev_list);
+
+
+ for_each_pci_dev(dev) {
+ subordinate = dev->subordinate;
+ if (!subordinate || list_empty(&subordinate->devices))
+ continue;
+
+ has_it = 0;
+ list_for_each_entry(pdev_entry, &pdev_list, list) {
+ if (dev == pdev_entry->dev) {
+ has_it = 1;
+ break;
+ }
+ }
+ if (has_it)
+ continue;
+
+ if (__pci_dev_need_reset(dev)) {
+ pdev_entry = kmalloc(sizeof(*pdev_entry), GFP_KERNEL);
+ pdev_entry->dev = dev;
+ list_add(&pdev_entry->list, &pdev_list);
+ }
+ }
+
+ list_for_each_entry_safe(pdev_entry, tmp, &pdev_list, list) {
+ pci_try_reset_bus(pdev_entry->dev->subordinate);
+ kfree(pdev_entry);
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(pci_reset_endpoints);
+
static int __init pci_setup(char *str)
{
while (str) {
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 5be8db4..1cf7207 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -1869,4 +1869,10 @@ static inline bool pci_is_dev_assigned(struct pci_dev *pdev)
{
return (pdev->dev_flags & PCI_DEV_FLAGS_ASSIGNED) == PCI_DEV_FLAGS_ASSIGNED;
}
+
+/*
+ * Reset all pci devices by resetting the buses.
+ */
+int pci_reset_endpoints(void);
+
#endif /* LINUX_PCI_H */
diff --git a/kernel/kexec.c b/kernel/kexec.c
index 2abf9f6..986e8f7 100644
--- a/kernel/kexec.c
+++ b/kernel/kexec.c
@@ -36,6 +36,7 @@
#include <linux/syscore_ops.h>
#include <linux/compiler.h>
#include <linux/hugetlb.h>
+#include <linux/pci.h>
#include <asm/page.h>
#include <asm/uaccess.h>
@@ -1474,6 +1475,7 @@ void crash_kexec(struct pt_regs *regs)
if (kexec_crash_image) {
struct pt_regs fixed_regs;
+ pci_reset_endpoints();
crash_setup_regs(&fixed_regs, regs);
crash_save_vmcoreinfo();
machine_crash_shutdown(&fixed_regs);

Gavin Shan

2014-10-19 23:20:08 UTC

Post by Li, Zhen-Hua
This is an update of the patch
https://lkml.org/lkml/2014/10/10/37
This patch is doing the reset works before the kdump kernel boots.

If I have said it once I have said it a thousand times.
It is absolutely inappropriate to do this in the kernel that has
crashed.
Either reserve a chunk of the iommu for use by the crash dump kernel,
or figure out how to do this during boot up.
But this is absolutely and totally inappropriate to do in crash_kexec.
The failure modes are all wrong. It will impact the reliability of our
crash dumps.
If an the code for a linux architecture is not structured in such a way
as to make it easy or straight forward to do this my sympathies you are
going to have do fix that linux port to be able to do things during
boot.
But things called from crash_kexec should be absolute necessities and I
do not see this as coming anywhere close to being something that is
impossible to do in the target kernel.

Zhenhua, Intrestingly, PPC has mechanism to reset the root port for kdump case
in order to drop pending DMA traffic. Maybe you can implement similar thing.
However, this kind of reset would be platform/board specific and needs support
from underly firmware. More details could be found from arch/powerpc/
platforms/powernv/pci-ioda.c

if (is_kdump_kernel()) {
pr_info(" Issue PHB reset ...\n");
ioda_eeh_phb_reset(hose, EEH_RESET_FUNDAMENTAL);
ioda_eeh_phb_reset(hose, OPAL_DEASSERT_RESET);
}

Please check if it's the thing you're looking for. I didn't understand your
requirements 100%

Thanks,
Gavin

Post by Eric W. Biederman
Eric

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
More majordomo info at http://vger.kernel.org/majordomo-info.html

Li, ZhenHua

2014-10-20 01:46:46 UTC

Post by Li, Zhen-Hua
This is an update of the patch
https://lkml.org/lkml/2014/10/10/37
This patch is doing the reset works before the kdump kernel boots.

Zhenhua, Intrestingly, PPC has mechanism to reset the root port for kdump case
in order to drop pending DMA traffic. Maybe you can implement similar thing.
However, this kind of reset would be platform/board specific and needs support
from underly firmware. More details could be found from arch/powerpc/
platforms/powernv/pci-ioda.c
if (is_kdump_kernel()) {
pr_info(" Issue PHB reset ...\n");
ioda_eeh_phb_reset(hose, EEH_RESET_FUNDAMENTAL);
ioda_eeh_phb_reset(hose, OPAL_DEASSERT_RESET);
}
Please check if it's the thing you're looking for. I didn't understand your
requirements 100%
Thanks,
Gavin

Gavin,
I was trying to fix a DMAR fault for iommu. On X86 platform, iommu is
initialized after pci , while on powerpc iommu is before pci. So I
tried to do the operation in the first kernel, and this can work on all
platforms.

But according to Eric's mail, it is inappropriate doing this in
crash_kexec. So I have to figure out how to do this in the kdump kernel.

As it is clear the powerpc has already do the resetting during booting,
so I only need to implement a similar mechanism on X86 system.

Thanks
Zhenhua

Post by Eric W. Biederman
Eric

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
More majordomo info at http://vger.kernel.org/majordomo-info.html

Li, ZhenHua

2014-10-20 02:06:01 UTC

Post by Li, ZhenHua

Post by Li, Zhen-Hua
This is an update of the patch
https://lkml.org/lkml/2014/10/10/37
This patch is doing the reset works before the kdump kernel boots.

Zhenhua, Intrestingly, PPC has mechanism to reset the root port for kdump case
in order to drop pending DMA traffic. Maybe you can implement similar thing.
However, this kind of reset would be platform/board specific and needs support
from underly firmware. More details could be found from arch/powerpc/
platforms/powernv/pci-ioda.c
if (is_kdump_kernel()) {
pr_info(" Issue PHB reset ...\n");
ioda_eeh_phb_reset(hose, EEH_RESET_FUNDAMENTAL);
ioda_eeh_phb_reset(hose, OPAL_DEASSERT_RESET);
}
Please check if it's the thing you're looking for. I didn't understand your
requirements 100%
Thanks,
Gavin

Sorry I made mistake. it should be sparc, not powerpc.

Post by Li, ZhenHua
But according to Eric's mail, it is inappropriate doing this in
crash_kexec. So I have to figure out how to do this in the kdump kernel.
As it is clear the powerpc has already do the resetting during booting,
so I only need to implement a similar mechanism on X86 system.
Thanks
Zhenhua

Post by Eric W. Biederman
Eric

Post by Li, Zhen-Hua
On a Linux system with iommu supported and many PCI devices on it,
when kernel crashed and the kdump kernel boots with intel_iommu=on,
there may be some unexpected DMA requests on this adapter, which will
dmar: DRHD: handling fault status reg 102
dmar: DMAR:[DMA Read] Request device [41:00.0] fault addr fff81000
DMAR:[fault reason 01] Present bit in root entry is clear
This bug may happen on *any* PCI device.
static struct context_entry * device_to_context_entry(
struct intel_iommu *iommu, u8 bus, u8 devfn)
{
......
set_root_present(root);
......
}
device driver
intel_alloc_coherent
__intel_map_single
domain_context_mapping
domain_context_mapping_one
device_to_context_entry
This means, the present bit in root entry will not be set until the device
driver is loaded.
But in the kdump kernel, hardware devices are not aware that control has
transferred to the second kernel, and those drivers must initialize again.
Consequently there may be unexpected DMA requests from devices activity
initiated in the first kernel leading to the DMA Remapping errors in the
second kernel.
To fix this DMAR fault, we need to reset the bus that this device on. Reset
the device itself does not work.
https://lkml.org/lkml/2014/9/30/55
As in discussion, this bug may happen on *any* device, so we need to reset
all pci devices.
https://lkml.org/lkml/2013/5/14/9
According to the previous discussion, On sparc, the IOMMU is initialized
before PCI devices are enumerated, this patch does the resetting works before
the kdump kernel boots, so it can also fix the problems on sparc.
Add support for legacy PCI devices.
Use pci_try_reset_bus instead of do_downstream_device_reset.
Reset all PCI/PCIe deviecs in the first kernel, before kdump kernel boots.
Randy Wright corrects some misunderstanding in this description.
---
drivers/pci/pci.c | 82
+++++++++++++++++++++++++++++++++++++++++++++++++++++
include/linux/pci.h | 6 ++++
kernel/kexec.c | 2 ++
3 files changed, 90 insertions(+)
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 625a4ac..aa9192a 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -23,6 +23,7 @@
#include <linux/device.h>
#include <linux/pm_runtime.h>
#include <linux/pci_hotplug.h>
+#include <linux/crash_dump.h>
#include <asm-generic/pci-bridge.h>
#include <asm/setup.h>
#include "pci.h"
@@ -4466,6 +4467,87 @@ void __weak pci_fixup_cardbus(struct pci_bus *bus)
}
EXPORT_SYMBOL(pci_fixup_cardbus);
+/*
+ * Return true if dev is PCI root port or downstream port whose child is PCI
+ * endpoint except VGA device.
+ */
+static int __pci_dev_need_reset(struct pci_dev *dev)
+{
+ struct pci_bus *subordinate;
+ struct pci_dev *child;
+
+ if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
+ return 0;
+
+ if (pci_is_pcie(dev)) {
+ if ((pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT) &&
+ (pci_pcie_type(dev) != PCI_EXP_TYPE_DOWNSTREAM))
+ return 0;
+ }
+
+ subordinate = dev->subordinate;
+ list_for_each_entry(child, &subordinate->devices, bus_list) {
+ /* Don't reset switch, bridge, VGA device */
+ if ((child->hdr_type == PCI_HEADER_TYPE_BRIDGE) ||
+ ((child->class >> 16) == PCI_BASE_CLASS_BRIDGE) ||
+ ((child->class >> 16) == PCI_BASE_CLASS_DISPLAY))
+ return 0;
+
+ if (pci_is_pcie(child)) {
+ if ((pci_pcie_type(child) == PCI_EXP_TYPE_UPSTREAM) ||
+ (pci_pcie_type(child) == PCI_EXP_TYPE_PCI_BRIDGE))
+ return 0;
+ }
+ }
+
+ return 1;
+}
+
+struct pci_dev_reset_entry {
+ struct list_head list;
+ struct pci_dev *dev;
+};
+int pci_reset_endpoints(void)
+{
+ struct pci_dev *dev = NULL;
+ struct pci_dev_reset_entry *pdev_entry, *tmp;
+ struct pci_bus *subordinate = NULL;
+ int has_it;
+
+ LIST_HEAD(pdev_list);
+
+
+ for_each_pci_dev(dev) {
+ subordinate = dev->subordinate;
+ if (!subordinate || list_empty(&subordinate->devices))
+ continue;
+
+ has_it = 0;
+ list_for_each_entry(pdev_entry, &pdev_list, list) {
+ if (dev == pdev_entry->dev) {
+ has_it = 1;
+ break;
+ }
+ }
+ if (has_it)
+ continue;
+
+ if (__pci_dev_need_reset(dev)) {
+ pdev_entry = kmalloc(sizeof(*pdev_entry), GFP_KERNEL);
+ pdev_entry->dev = dev;
+ list_add(&pdev_entry->list, &pdev_list);
+ }
+ }
+
+ list_for_each_entry_safe(pdev_entry, tmp, &pdev_list, list) {
+ pci_try_reset_bus(pdev_entry->dev->subordinate);
+ kfree(pdev_entry);
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(pci_reset_endpoints);
+
static int __init pci_setup(char *str)
{
while (str) {
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 5be8db4..1cf7207 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -1869,4 +1869,10 @@ static inline bool pci_is_dev_assigned(struct pci_dev *pdev)
{
return (pdev->dev_flags & PCI_DEV_FLAGS_ASSIGNED) ==
PCI_DEV_FLAGS_ASSIGNED;
}
+
+/*
+ * Reset all pci devices by resetting the buses.
+ */
+int pci_reset_endpoints(void);
+
#endif /* LINUX_PCI_H */
diff --git a/kernel/kexec.c b/kernel/kexec.c
index 2abf9f6..986e8f7 100644
--- a/kernel/kexec.c
+++ b/kernel/kexec.c
@@ -36,6 +36,7 @@
#include <linux/syscore_ops.h>
#include <linux/compiler.h>
#include <linux/hugetlb.h>
+#include <linux/pci.h>
#include <asm/page.h>
#include <asm/uaccess.h>
@@ -1474,6 +1475,7 @@ void crash_kexec(struct pt_regs *regs)
if (kexec_crash_image) {
struct pt_regs fixed_regs;
+ pci_reset_endpoints();
crash_setup_regs(&fixed_regs, regs);
crash_save_vmcoreinfo();
machine_crash_shutdown(&fixed_regs);

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
More majordomo info at http://vger.kernel.org/majordomo-info.html

Li, ZhenHua

2014-10-20 01:34:00 UTC

Post by Li, Zhen-Hua
This is an update of the patch
https://lkml.org/lkml/2014/10/10/37
This patch is doing the reset works before the kdump kernel boots.

If I have said it once I have said it a thousand times.

I am very sorry for your name. I got your name by running the script
./scripts/get_maintainer.pl, maybe you need to update your name in the
git system.

Post by Eric W. Biederman
It is absolutely inappropriate to do this in the kernel that has
crashed.
Either reserve a chunk of the iommu for use by the crash dump kernel,
or figure out how to do this during boot up.
But this is absolutely and totally inappropriate to do in crash_kexec.
The failure modes are all wrong. It will impact the reliability of our
crash dumps.
If an the code for a linux architecture is not structured in such a way
as to make it easy or straight forward to do this my sympathies you are
going to have do fix that linux port to be able to do things during
boot.
But things called from crash_kexec should be absolute necessities and I
do not see this as coming anywhere close to being something that is
impossible to do in the target kernel.
Eric

So now do the resetting during the kdump kernel boots is the only
choice. I will try to figure out how to fix this.

Thank you very much.

Zhenhua