The dmaengine core assumes that async DMA devices will only be removed
when they not used anymore, or it assumes dma_async_device_unregister()
will only be called by dma driver exit routines. But this assumption is
not true for the IOAT driver, which calls dma_async_device_unregister()
from ioat_remove(). So current IOAT driver doesn't support device
hot-removal because it may cause system crash to hot-remove an inuse
IOAT device.
To support CPU socket hot-removal, all PCI devices, including IOAT
devices embedded in the socket, will be hot-removed. The idea solution
is to enhance the dmaengine core and IOAT driver to support hot-removal,
but that's too hard.
This patch implements a hack to disable IOAT devices under hotplug-capable
CPU socket so it won't break socket hot-removal.
Signed-off-by: Jiang Liu <[email protected]>
---
Hi Vinod,
This is a resend of v1, no actual changes except the email title.
Dan and Dave have no objection to this approach, though they haven't
explicitly acked the patch:)
Thanks!
Gerry
---
drivers/dma/ioat/pci.c | 34 ++++++++++++++++++++++++++++++++++
1 file changed, 34 insertions(+)
diff --git a/drivers/dma/ioat/pci.c b/drivers/dma/ioat/pci.c
index 76f0dc688a19..3b8c9b03f4b3 100644
--- a/drivers/dma/ioat/pci.c
+++ b/drivers/dma/ioat/pci.c
@@ -27,6 +27,7 @@
#include <linux/interrupt.h>
#include <linux/dca.h>
#include <linux/slab.h>
+#include <linux/acpi.h>
#include "dma.h"
#include "dma_v2.h"
#include "registers.h"
@@ -148,6 +149,34 @@ alloc_ioatdma(struct pci_dev *pdev, void __iomem *iobase)
return d;
}
+/*
+ * The dmaengine core assumes that async DMA devices will only be removed
+ * when they not used anymore, or it assumes dma_async_device_unregister()
+ * will only be called by dma driver exit routines. But this assumption is
+ * not true for the IOAT driver, which calls dma_async_device_unregister()
+ * from ioat_remove(). So current IOAT driver doesn't support device
+ * hot-removal because it may cause system crash to hot-remove inuse IOAT
+ * devices.
+ *
+ * This is a hack to disable IOAT devices under ejectable PCI host bridge
+ * so it won't break PCI host bridge hot-removal.
+ */
+static bool ioat_pci_has_ejectable_acpi_ancestor(struct pci_dev *pdev)
+{
+#ifdef CONFIG_ACPI
+ struct pci_bus *bus = pdev->bus;
+ struct acpi_device *adev;
+
+ while (bus->parent)
+ bus = bus->parent;
+ for (adev = ACPI_COMPANION(bus->bridge); adev; adev = adev->parent)
+ if (adev->flags.ejectable)
+ return true;
+#endif
+
+ return false;
+}
+
static int ioat_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
{
void __iomem * const *iomap;
@@ -155,6 +184,11 @@ static int ioat_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
struct ioatdma_device *device;
int err;
+ if (ioat_pci_has_ejectable_acpi_ancestor(pdev)) {
+ dev_dbg(&pdev->dev, "ignore ejectable IOAT device.\n");
+ return -ENODEV;
+ }
+
err = pcim_enable_device(pdev);
if (err)
return err;
--
1.7.10.4
On Wed, Jul 08, 2015 at 03:41:42PM +0800, Jiang Liu wrote:
> The dmaengine core assumes that async DMA devices will only be removed
> when they not used anymore, or it assumes dma_async_device_unregister()
> will only be called by dma driver exit routines. But this assumption is
> not true for the IOAT driver, which calls dma_async_device_unregister()
> from ioat_remove(). So current IOAT driver doesn't support device
> hot-removal because it may cause system crash to hot-remove an inuse
> IOAT device.
>
> To support CPU socket hot-removal, all PCI devices, including IOAT
> devices embedded in the socket, will be hot-removed. The idea solution
> is to enhance the dmaengine core and IOAT driver to support hot-removal,
> but that's too hard.
>
> This patch implements a hack to disable IOAT devices under hotplug-capable
> CPU socket so it won't break socket hot-removal.
Applied now, please ensure you use the right subsystem name
--
~Vinod