2009-04-17 10:23:21

by tip-bot for Andreas Herrmann

Subject: [PATCH] x86: fixup numa_node information for AMD CPU northbridge functions

Currently the numa_node attribute for these PCI devices is 0 (it
corresponds to the numa_node for PCI bus 0). This is not a big issue
but incorrect.

This inconsistency can be fixed by reading the node number from CPU
NB function 0.

Signed-off-by: Andreas Herrmann <[email protected]>
---
arch/x86/kernel/quirks.c | 35 +++++++++++++++++++++++++++++++++++
1 files changed, 35 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/quirks.c b/arch/x86/kernel/quirks.c
index e95022e..4dd4e70 100644
--- a/arch/x86/kernel/quirks.c
+++ b/arch/x86/kernel/quirks.c
@@ -493,5 +493,40 @@ void force_hpet_resume(void)
break;
}
}
+#endif
+
+#if defined(CONFIG_PCI) && defined(CONFIG_NUMA)
+/* Set correct numa_node information for AMD NB functions */
+static void __init quirk_amd_nb_node(struct pci_dev *dev)
+{
+ u32 val;
+ struct pci_dev *nb_ht;
+ unsigned int devfn;

+ devfn = PCI_DEVFN(PCI_SLOT(dev->devfn), 0);
+ nb_ht = pci_get_slot(dev->bus, devfn);
+ if (!nb_ht)
+ return;
+ pci_read_config_dword(nb_ht, 0x60, &val);
+ set_dev_node(&dev->dev, val & 7);
+ pci_dev_put(dev);
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_ADDRMAP,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MEMCTL,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MISC,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_HT,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MAP,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_DRAM,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_LINK,
+ quirk_amd_nb_node);
#endif
--
1.6.2

2009-04-17 16:15:01

by Ingo Molnar

Subject: Re: [PATCH] x86: fixup numa_node information for AMD CPU northbridge functions

* Andreas Herrmann <[email protected]> wrote:

> Currently the numa_node attribute for these PCI devices is 0 (it
> corresponds to the numa_node for PCI bus 0). This is not a big issue
> but incorrect.
>
> This inconsistency can be fixed by reading the node number from CPU
> NB function 0.
>
> Signed-off-by: Andreas Herrmann <[email protected]>
> ---
> arch/x86/kernel/quirks.c | 35 +++++++++++++++++++++++++++++++++++
> 1 files changed, 35 insertions(+), 0 deletions(-)

applied, thanks. Note, i added this impact line to the commit:

[ Impact: fill in dev->numa_node information, to optimize DMA allocations ]

as numa_node does matter to certain DMA allocations such as
dma_generic_alloc_coherent() - we (try to) allocate node-local
memory in that case.

pci_dev::numa_node might be used for more such optimizations in the
future - so it's generally good to have this filled in correctly.

So it's more than just a fix for an inconsistency.

Ingo

2009-04-17 16:28:34

by tip-bot for Andreas Herrmann

Subject: [tip:x86/cpu] x86: fixup numa_node information for AMD CPU northbridge functions

Commit-ID: 9b94b3a19b13e094c10f65f24bc358f6ffe4eacd
Gitweb: http://git.kernel.org/tip/9b94b3a19b13e094c10f65f24bc358f6ffe4eacd
Author: Andreas Herrmann <[email protected]>
AuthorDate: Fri, 17 Apr 2009 12:07:46 +0200
Committer: Ingo Molnar <[email protected]>
CommitDate: Fri, 17 Apr 2009 18:12:36 +0200

x86: fixup numa_node information for AMD CPU northbridge functions

Currently the numa_node attribute for these PCI devices is 0 (it
corresponds to the numa_node for PCI bus 0). This is not a big issue
but incorrect.

This inconsistency can be fixed by reading the node number from CPU
NB function 0.

[ Impact: fill in dev->numa_node information, to optimize DMA allocations ]

Signed-off-by: Andreas Herrmann <[email protected]>
Cc: [email protected]
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>


---
arch/x86/kernel/quirks.c | 37 +++++++++++++++++++++++++++++++++++++
1 files changed, 37 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/quirks.c b/arch/x86/kernel/quirks.c
index e95022e..94ad0c0 100644
--- a/arch/x86/kernel/quirks.c
+++ b/arch/x86/kernel/quirks.c
@@ -493,5 +493,42 @@ void force_hpet_resume(void)
break;
}
}
+#endif
+
+#if defined(CONFIG_PCI) && defined(CONFIG_NUMA)
+/* Set correct numa_node information for AMD NB functions */
+static void __init quirk_amd_nb_node(struct pci_dev *dev)
+{
+ struct pci_dev *nb_ht;
+ unsigned int devfn;
+ u32 val;
+
+ devfn = PCI_DEVFN(PCI_SLOT(dev->devfn), 0);
+ nb_ht = pci_get_slot(dev->bus, devfn);
+ if (!nb_ht)
+ return;
+
+ pci_read_config_dword(nb_ht, 0x60, &val);
+ set_dev_node(&dev->dev, val & 7);
+ pci_dev_put(dev);
+}

+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_ADDRMAP,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MEMCTL,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MISC,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_HT,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MAP,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_DRAM,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC,
+ quirk_amd_nb_node);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_LINK,
+ quirk_amd_nb_node);
#endif

2009-04-17 19:22:19

by Yinghai Lu

Subject: Re: [tip:x86/cpu] x86: fixup numa_node information for AMD CPU northbridge functions

On Fri, Apr 17, 2009 at 9:27 AM, tip-bot for Andreas Herrmann
<[email protected]> wrote:
> Commit-ID: ?9b94b3a19b13e094c10f65f24bc358f6ffe4eacd
> Gitweb: ? ? http://git.kernel.org/tip/9b94b3a19b13e094c10f65f24bc358f6ffe4eacd
> Author: ? ? Andreas Herrmann <[email protected]>
> AuthorDate: Fri, 17 Apr 2009 12:07:46 +0200
> Committer: ?Ingo Molnar <[email protected]>
> CommitDate: Fri, 17 Apr 2009 18:12:36 +0200
>
> x86: fixup numa_node information for AMD CPU northbridge functions
>
> Currently the numa_node attribute for these PCI devices is 0 (it
> corresponds to the numa_node for PCI bus 0). This is not a big issue
> but incorrect.
>
> This inconsistency can be fixed by reading the node number from CPU
> NB function 0.
>
> [ Impact: fill in dev->numa_node information, to optimize DMA allocations ]
>
> Signed-off-by: Andreas Herrmann <[email protected]>
> Cc: [email protected]
> LKML-Reference: <[email protected]>
> Signed-off-by: Ingo Molnar <[email protected]>
>
>
> ---
> ?arch/x86/kernel/quirks.c | ? 37 +++++++++++++++++++++++++++++++++++++
> ?1 files changed, 37 insertions(+), 0 deletions(-)
>
> diff --git a/arch/x86/kernel/quirks.c b/arch/x86/kernel/quirks.c
> index e95022e..94ad0c0 100644
> --- a/arch/x86/kernel/quirks.c
> +++ b/arch/x86/kernel/quirks.c
> @@ -493,5 +493,42 @@ void force_hpet_resume(void)
> ? ? ? ? ? ? ? ?break;
> ? ? ? ?}
> ?}
> +#endif
> +
> +#if defined(CONFIG_PCI) && defined(CONFIG_NUMA)
> +/* Set correct numa_node information for AMD NB functions */
> +static void __init quirk_amd_nb_node(struct pci_dev *dev)
> +{
> + ? ? ? struct pci_dev *nb_ht;
> + ? ? ? unsigned int devfn;
> + ? ? ? u32 val;
> +
> + ? ? ? devfn = PCI_DEVFN(PCI_SLOT(dev->devfn), 0);
> + ? ? ? nb_ht = pci_get_slot(dev->bus, devfn);
> + ? ? ? if (!nb_ht)
> + ? ? ? ? ? ? ? return;
> +
> + ? ? ? pci_read_config_dword(nb_ht, 0x60, &val);
> + ? ? ? set_dev_node(&dev->dev, val & 7);
> + ? ? ? pci_dev_put(dev);
> +}
>
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_ADDRMAP,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MEMCTL,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MISC,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_HT,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MAP,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_DRAM,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_LINK,
> + ? ? ? ? ? ? ? ? ? ? ? quirk_amd_nb_node);
> ?#endif
> --

wonder if need to check if the node is online aka there is memory on
that node...

looks like if that node is not online, should use -1 instead for 64bit.

YH