Re: [PATCH] ppc/EEH: fix crash when adding a device in a slot with DDW

On Fri, Dec 28, 2012 at 01:18:24PM +0800, Gavin Shan wrote:

On Thu, Dec 27, 2012 at 02:34:00PM -0200, Thadeu Lima de Souza Cascardo wrote:

quoted

The DDW code uses a eeh_dev struct from the pci_dev. However, this is
not set until eeh_add_device_late is called.

Since pci_bus_add_devices is called before eeh_add_device_late, the PCI
devices are added to the bus, making drivers' probe hooks to be called.
These will call set_dma_mask, which will call the DDW code, which will
require the eeh_dev struct from pci_dev. This would result in a crash,
due to a NULL dereference.

Calling eeh_add_device_late after pci_bus_add_devices would make the
system BUG, because device files shouldn't be added to devices there
were not added to the system. So, a new function is needed to add such
files only after pci_bus_add_devices have been called.

Could you please explain for a bit how did you trigger the problem? I'm
not sure you got it while doing PCI hotplug or just saw the issue during
system bootup stage :-)

I did a DLPAR remove followed by a DLPAR add, using drmgr -r and drmgr
-a. The reason the issue is not trigerred during bootup is that there is
no driver registered yet. So no driver probe will call dma_set_mask.

quoted

Cc: stable@vger.kernel.org
Signed-off-by: Thadeu Lima de Souza Cascardo <redacted>
---
arch/powerpc/include/asm/eeh.h       |    3 +++
arch/powerpc/kernel/pci-common.c     |    7 +++++--
arch/powerpc/platforms/pseries/eeh.c |   24 +++++++++++++++++++++++-
3 files changed, 31 insertions(+), 3 deletions(-)

diff --git a/arch/powerpc/include/asm/eeh.h b/arch/powerpc/include/asm/eeh.h
index b0ef738..71aac19 100644
--- a/arch/powerpc/include/asm/eeh.h
+++ b/arch/powerpc/include/asm/eeh.h

@@ -201,6 +201,7 @@ int eeh_dev_check_failure(struct eeh_dev *edev);

void __init eeh_addr_cache_build(void);
void eeh_add_device_tree_early(struct device_node *);
void eeh_add_device_tree_late(struct pci_bus *);
+void eeh_add_device_tree_files(struct pci_bus *);

Since the function is going to add EEH specific sysfs files, its name would
be something like "eeh_add_sysfs_files" instead of "eeh_add_device_tree_files" :-)

It's indeed a better name.

quoted

void eeh_remove_bus_device(struct pci_dev *, int);

/**

@@ -240,6 +241,8 @@ static inline void eeh_add_device_tree_early(struct device_node *dn) { }

static inline void eeh_add_device_tree_late(struct pci_bus *bus) { }

+static inline void eeh_add_device_tree_files(struct pci_bus *bus) { }
+

It'd better to rename the function name to "eeh_add_sysfs_files" mentioned
as above.

quoted

static inline void eeh_remove_bus_device(struct pci_dev *dev, int purge_pe) { }

static inline void eeh_lock(void) { }

diff --git a/arch/powerpc/kernel/pci-common.c b/arch/powerpc/kernel/pci-common.c
index 7f94f76..7b1f14c 100644
--- a/arch/powerpc/kernel/pci-common.c
+++ b/arch/powerpc/kernel/pci-common.c

@@ -1480,11 +1480,14 @@ void pcibios_finish_adding_to_bus(struct pci_bus *bus)

	pcibios_allocate_bus_resources(bus);
	pcibios_claim_one_bus(bus);

+	/* Fixup EEH */
+	eeh_add_device_tree_late(bus);
+
	/* Add new devices to global lists.  Register in proc, sysfs. */
	pci_bus_add_devices(bus);

-	/* Fixup EEH */
-	eeh_add_device_tree_late(bus);
+	/* Add EEH sysfs files */
+	eeh_add_device_tree_files(bus);

The function name would be "eeh_add_sysfs_files" as above.

quoted

}
EXPORT_SYMBOL_GPL(pcibios_finish_adding_to_bus);

By the way, arch/powerpc/kernel/of_platform.c::of_pci_phb_probe is also calling
to eeh_add_device_tree_late() as well. Since you have removed part of the logic
from original eeh_add_device_tree_late(), which is add EEH specific sysfs files,
and you put that part of logic to eeh_add_device_tree_files(). So I think you
also need make the similiar change for of_pci_phb_probe() as well :-)

Good point. I will look into other call sites. However, I don't have
access to other platforms to test the patch.

quoted

diff --git a/arch/powerpc/platforms/pseries/eeh.c b/arch/powerpc/platforms/pseries/eeh.c
index 9a04322..a667a34 100644
--- a/arch/powerpc/platforms/pseries/eeh.c
+++ b/arch/powerpc/platforms/pseries/eeh.c

@@ -788,7 +788,6 @@ static void eeh_add_device_late(struct pci_dev *dev)

	dev->dev.archdata.edev = edev;

	eeh_addr_cache_insert_dev(dev);
-	eeh_sysfs_add_device(dev);
}

/**

@@ -815,6 +814,29 @@ void eeh_add_device_tree_late(struct pci_bus *bus)

EXPORT_SYMBOL_GPL(eeh_add_device_tree_late);

/**
+ * eeh_add_device_tree_files - Add EEH sysfs files for the indicated PCI bus
+ * @bus: PCI bus
+ *
+ * This routine must be used to add EEH sysfs files for PCI
+ * devices which are attached to the indicated PCI bus. The PCI bus
+ * is added after system boot through hotplug or dlpar.
+ */
+void eeh_add_device_tree_files(struct pci_bus *bus)
+{
+	struct pci_dev *dev;
+
+	list_for_each_entry(dev, &bus->devices, bus_list) {
+		eeh_sysfs_add_device(dev);
+		if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
+			struct pci_bus *subbus = dev->subordinate;
+			if (subbus)
+				eeh_add_device_tree_files(subbus);
+		}
+	}
+}
+EXPORT_SYMBOL_GPL(eeh_add_device_tree_files);
+

The function name mentioned as above.

quoted

+/**
 * eeh_remove_device - Undo EEH setup for the indicated pci device
 * @dev: pci device to be removed
 * @purge_pe: remove the PE or not

Thanks,
Gavin

Regards,
Thadeu Cascardo.