Work around BIOSes that don't report the entire Intel MCH area.
MCHBAR is not an architected PCI BAR, so MCH space is usually reported as a
PNP0C02 resource. The MCH space was once 16KB, but is 32KB in newer parts.
Some BIOSes still report a PNP0C02 resource that is only 16KB, which means
the rest of the MCH space is consumed but unreported.
This can cause resource map sanity check warnings or (theoretically) a
device conflict if we assigned the unreported space to another device.
The Intel perf event uncore driver tripped over this when it claimed the
MCH region:
resource map sanity check conflict: 0xfed10000 0xfed15fff 0xfed10000 0xfed13fff pnp 00:01
Info: mapping multiple BARs. Your kernel is fine.
To prevent this, if we find a PNP0C02 resource that covers part of the MCH
space, extend it to cover the entire space.
Link: http://lkml.kernel.org/r/[email protected]
Reported-by: Borislav Petkov <[email protected]>
Tested-by: Borislav Petkov <[email protected]>
Signed-off-by: Bjorn Helgaas <[email protected]>
Acked-by: Borislav Petkov <[email protected]>
---
drivers/pnp/quirks.c | 75 ++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 75 insertions(+)
diff --git a/drivers/pnp/quirks.c b/drivers/pnp/quirks.c
index 258fef272ea7..0d679068ef1b 100644
--- a/drivers/pnp/quirks.c
+++ b/drivers/pnp/quirks.c
@@ -15,6 +15,7 @@
#include <linux/types.h>
#include <linux/kernel.h>
+#include <linux/pci.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/pnp.h>
@@ -334,6 +335,79 @@ static void quirk_amd_mmconfig_area(struct pnp_dev *dev)
}
#endif
+/* Device IDs of parts that have 32KB MCH space */
+static const unsigned int mch_quirk_devices[] = {
+ 0x0154, /* Ivy Bridge */
+ 0x0c00, /* Haswell */
+};
+
+static struct pci_dev *get_intel_host(void)
+{
+ int i;
+ struct pci_dev *host;
+
+ for (i = 0; i < ARRAY_SIZE(mch_quirk_devices); i++) {
+ host = pci_get_device(PCI_VENDOR_ID_INTEL, mch_quirk_devices[i],
+ NULL);
+ if (host)
+ return host;
+ }
+ return NULL;
+}
+
+static void quirk_intel_mch(struct pnp_dev *dev)
+{
+ struct pci_dev *host;
+ u32 addr_lo, addr_hi;
+ struct pci_bus_region region;
+ struct resource mch;
+ struct pnp_resource *pnp_res;
+ struct resource *res;
+
+ host = get_intel_host();
+ if (!host)
+ return;
+
+ /*
+ * MCHBAR is not an architected PCI BAR, so MCH space is usually
+ * reported as a PNP0C02 resource. The MCH space was originally
+ * 16KB, but is 32KB in newer parts. Some BIOSes still report a
+ * PNP0C02 resource that is only 16KB, which means the rest of the
+ * MCH space is consumed but unreported.
+ */
+
+ /*
+ * Read MCHBAR for Host Member Mapped Register Range Base
+ * https://www-ssl.intel.com/content/www/us/en/processors/core/4th-gen-core-family-desktop-vol-2-datasheet
+ * Sec 3.1.12.
+ */
+ pci_read_config_dword(host, 0x48, &addr_lo);
+ region.start = addr_lo & ~0x7fff;
+ pci_read_config_dword(host, 0x4c, &addr_hi);
+ region.start |= (u64) addr_hi << 32;
+ region.end = region.start + 32*1024 - 1;
+
+ memset(&mch, 0, sizeof(mch));
+ mch.flags = IORESOURCE_MEM;
+ pcibios_bus_to_resource(host->bus, &mch, ®ion);
+
+ list_for_each_entry(pnp_res, &dev->resources, list) {
+ res = &pnp_res->res;
+ if (res->end < mch.start || res->start > mch.end)
+ continue; /* no overlap */
+ if (res->start == mch.start && res->end == mch.end)
+ continue; /* exact match */
+
+ dev_info(&dev->dev, FW_BUG "PNP resource %pR covers only part of %s Intel MCH; extending to %pR\n",
+ res, pci_name(host), &mch);
+ res->start = mch.start;
+ res->end = mch.end;
+ break;
+ }
+
+ pci_dev_put(host);
+}
+
/*
* PnP Quirks
* Cards or devices that need some tweaking due to incomplete resource info
@@ -364,6 +438,7 @@ static struct pnp_fixup pnp_fixups[] = {
#ifdef CONFIG_AMD_NB
{"PNP0c01", quirk_amd_mmconfig_area},
#endif
+ {"PNP0c02", quirk_intel_mch},
{""}
};
On Tue, Apr 22, 2014 at 12:17 AM, Bjorn Helgaas <[email protected]> wrote:
> Work around BIOSes that don't report the entire Intel MCH area.
>
> MCHBAR is not an architected PCI BAR, so MCH space is usually reported as a
> PNP0C02 resource. The MCH space was once 16KB, but is 32KB in newer parts.
> Some BIOSes still report a PNP0C02 resource that is only 16KB, which means
> the rest of the MCH space is consumed but unreported.
>
> This can cause resource map sanity check warnings or (theoretically) a
> device conflict if we assigned the unreported space to another device.
>
> The Intel perf event uncore driver tripped over this when it claimed the
> MCH region:
>
> resource map sanity check conflict: 0xfed10000 0xfed15fff 0xfed10000 0xfed13fff pnp 00:01
> Info: mapping multiple BARs. Your kernel is fine.
>
> To prevent this, if we find a PNP0C02 resource that covers part of the MCH
> space, extend it to cover the entire space.
>
Works for me on my Levono IvyBridge laptop.
Thanks for fixing this, Bjorn.
Acked-by: Stephane Eranian <[email protected]>
> Link: http://lkml.kernel.org/r/[email protected]
> Reported-by: Borislav Petkov <[email protected]>
> Tested-by: Borislav Petkov <[email protected]>
> Signed-off-by: Bjorn Helgaas <[email protected]>
> Acked-by: Borislav Petkov <[email protected]>
> ---
> drivers/pnp/quirks.c | 75 ++++++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 75 insertions(+)
>
> diff --git a/drivers/pnp/quirks.c b/drivers/pnp/quirks.c
> index 258fef272ea7..0d679068ef1b 100644
> --- a/drivers/pnp/quirks.c
> +++ b/drivers/pnp/quirks.c
> @@ -15,6 +15,7 @@
>
> #include <linux/types.h>
> #include <linux/kernel.h>
> +#include <linux/pci.h>
> #include <linux/string.h>
> #include <linux/slab.h>
> #include <linux/pnp.h>
> @@ -334,6 +335,79 @@ static void quirk_amd_mmconfig_area(struct pnp_dev *dev)
> }
> #endif
>
> +/* Device IDs of parts that have 32KB MCH space */
> +static const unsigned int mch_quirk_devices[] = {
> + 0x0154, /* Ivy Bridge */
> + 0x0c00, /* Haswell */
> +};
> +
> +static struct pci_dev *get_intel_host(void)
> +{
> + int i;
> + struct pci_dev *host;
> +
> + for (i = 0; i < ARRAY_SIZE(mch_quirk_devices); i++) {
> + host = pci_get_device(PCI_VENDOR_ID_INTEL, mch_quirk_devices[i],
> + NULL);
> + if (host)
> + return host;
> + }
> + return NULL;
> +}
> +
> +static void quirk_intel_mch(struct pnp_dev *dev)
> +{
> + struct pci_dev *host;
> + u32 addr_lo, addr_hi;
> + struct pci_bus_region region;
> + struct resource mch;
> + struct pnp_resource *pnp_res;
> + struct resource *res;
> +
> + host = get_intel_host();
> + if (!host)
> + return;
> +
> + /*
> + * MCHBAR is not an architected PCI BAR, so MCH space is usually
> + * reported as a PNP0C02 resource. The MCH space was originally
> + * 16KB, but is 32KB in newer parts. Some BIOSes still report a
> + * PNP0C02 resource that is only 16KB, which means the rest of the
> + * MCH space is consumed but unreported.
> + */
> +
> + /*
> + * Read MCHBAR for Host Member Mapped Register Range Base
> + * https://www-ssl.intel.com/content/www/us/en/processors/core/4th-gen-core-family-desktop-vol-2-datasheet
> + * Sec 3.1.12.
> + */
> + pci_read_config_dword(host, 0x48, &addr_lo);
> + region.start = addr_lo & ~0x7fff;
> + pci_read_config_dword(host, 0x4c, &addr_hi);
> + region.start |= (u64) addr_hi << 32;
> + region.end = region.start + 32*1024 - 1;
> +
> + memset(&mch, 0, sizeof(mch));
> + mch.flags = IORESOURCE_MEM;
> + pcibios_bus_to_resource(host->bus, &mch, ®ion);
> +
> + list_for_each_entry(pnp_res, &dev->resources, list) {
> + res = &pnp_res->res;
> + if (res->end < mch.start || res->start > mch.end)
> + continue; /* no overlap */
> + if (res->start == mch.start && res->end == mch.end)
> + continue; /* exact match */
> +
> + dev_info(&dev->dev, FW_BUG "PNP resource %pR covers only part of %s Intel MCH; extending to %pR\n",
> + res, pci_name(host), &mch);
> + res->start = mch.start;
> + res->end = mch.end;
> + break;
> + }
> +
> + pci_dev_put(host);
> +}
> +
> /*
> * PnP Quirks
> * Cards or devices that need some tweaking due to incomplete resource info
> @@ -364,6 +438,7 @@ static struct pnp_fixup pnp_fixups[] = {
> #ifdef CONFIG_AMD_NB
> {"PNP0c01", quirk_amd_mmconfig_area},
> #endif
> + {"PNP0c02", quirk_intel_mch},
> {""}
> };
>
>
* Stephane Eranian <[email protected]> wrote:
> On Tue, Apr 22, 2014 at 12:17 AM, Bjorn Helgaas <[email protected]> wrote:
> > Work around BIOSes that don't report the entire Intel MCH area.
> >
> > MCHBAR is not an architected PCI BAR, so MCH space is usually reported as a
> > PNP0C02 resource. The MCH space was once 16KB, but is 32KB in newer parts.
> > Some BIOSes still report a PNP0C02 resource that is only 16KB, which means
> > the rest of the MCH space is consumed but unreported.
> >
> > This can cause resource map sanity check warnings or (theoretically) a
> > device conflict if we assigned the unreported space to another device.
> >
> > The Intel perf event uncore driver tripped over this when it claimed the
> > MCH region:
> >
> > resource map sanity check conflict: 0xfed10000 0xfed15fff 0xfed10000 0xfed13fff pnp 00:01
> > Info: mapping multiple BARs. Your kernel is fine.
> >
> > To prevent this, if we find a PNP0C02 resource that covers part of the MCH
> > space, extend it to cover the entire space.
> >
> Works for me on my Levono IvyBridge laptop.
> Thanks for fixing this, Bjorn.
> Acked-by: Stephane Eranian <[email protected]>
Just curious, what problems triggered on your laptop: only the
warnings, or did something get mapped to the undeclared area,
causing other misbehavor?
Thanks,
Ingo
On Thu, Apr 24, 2014 at 9:20 AM, Ingo Molnar <[email protected]> wrote:
>
> * Stephane Eranian <[email protected]> wrote:
>
>> On Tue, Apr 22, 2014 at 12:17 AM, Bjorn Helgaas <[email protected]> wrote:
>> > Work around BIOSes that don't report the entire Intel MCH area.
>> >
>> > MCHBAR is not an architected PCI BAR, so MCH space is usually reported as a
>> > PNP0C02 resource. The MCH space was once 16KB, but is 32KB in newer parts.
>> > Some BIOSes still report a PNP0C02 resource that is only 16KB, which means
>> > the rest of the MCH space is consumed but unreported.
>> >
>> > This can cause resource map sanity check warnings or (theoretically) a
>> > device conflict if we assigned the unreported space to another device.
>> >
>> > The Intel perf event uncore driver tripped over this when it claimed the
>> > MCH region:
>> >
>> > resource map sanity check conflict: 0xfed10000 0xfed15fff 0xfed10000 0xfed13fff pnp 00:01
>> > Info: mapping multiple BARs. Your kernel is fine.
>> >
>> > To prevent this, if we find a PNP0C02 resource that covers part of the MCH
>> > space, extend it to cover the entire space.
>> >
>> Works for me on my Levono IvyBridge laptop.
>> Thanks for fixing this, Bjorn.
>> Acked-by: Stephane Eranian <[email protected]>
>
> Just curious, what problems triggered on your laptop: only the
> warnings, or did something get mapped to the undeclared area,
> causing other misbehavor?
>
In my case, it was just the warning. Everything appeared to worked correctly.