diff options
| -rw-r--r-- | Documentation/edac.txt | 152 | ||||
| -rw-r--r-- | arch/x86/include/asm/pci_x86.h | 2 | ||||
| -rw-r--r-- | arch/x86/kernel/cpu/mcheck/mce.c | 10 | ||||
| -rw-r--r-- | arch/x86/pci/legacy.c | 42 | ||||
| -rw-r--r-- | drivers/edac/Kconfig | 13 | ||||
| -rw-r--r-- | drivers/edac/Makefile | 2 | ||||
| -rw-r--r-- | drivers/edac/edac_core.h | 23 | ||||
| -rw-r--r-- | drivers/edac/edac_mc_sysfs.c | 175 | ||||
| -rw-r--r-- | drivers/edac/edac_mce.c | 61 | ||||
| -rw-r--r-- | drivers/edac/i7core_edac.c | 2078 | ||||
| -rw-r--r-- | include/linux/edac_mce.h | 31 | ||||
| -rw-r--r-- | include/linux/pci.h | 1 | ||||
| -rw-r--r-- | include/linux/pci_ids.h | 52 |
13 files changed, 2598 insertions, 44 deletions
diff --git a/Documentation/edac.txt b/Documentation/edac.txt index 79c533223762..0b875e8da969 100644 --- a/Documentation/edac.txt +++ b/Documentation/edac.txt | |||
| @@ -6,6 +6,8 @@ Written by Doug Thompson <dougthompson@xmission.com> | |||
| 6 | 7 Dec 2005 | 6 | 7 Dec 2005 |
| 7 | 17 Jul 2007 Updated | 7 | 17 Jul 2007 Updated |
| 8 | 8 | ||
| 9 | (c) Mauro Carvalho Chehab <mchehab@redhat.com> | ||
| 10 | 05 Aug 2009 Nehalem interface | ||
| 9 | 11 | ||
| 10 | EDAC is maintained and written by: | 12 | EDAC is maintained and written by: |
| 11 | 13 | ||
| @@ -717,3 +719,153 @@ unique drivers for their hardware systems. | |||
| 717 | The 'test_device_edac' sample driver is located at the | 719 | The 'test_device_edac' sample driver is located at the |
| 718 | bluesmoke.sourceforge.net project site for EDAC. | 720 | bluesmoke.sourceforge.net project site for EDAC. |
| 719 | 721 | ||
| 722 | ======================================================================= | ||
| 723 | NEHALEM USAGE OF EDAC APIs | ||
| 724 | |||
| 725 | This chapter documents some EXPERIMENTAL mappings for EDAC API to handle | ||
| 726 | Nehalem EDAC driver. They will likely be changed on future versions | ||
| 727 | of the driver. | ||
| 728 | |||
| 729 | Due to the way Nehalem exports Memory Controller data, some adjustments | ||
| 730 | were done at i7core_edac driver. This chapter will cover those differences | ||
| 731 | |||
| 732 | 1) On Nehalem, there are one Memory Controller per Quick Patch Interconnect | ||
| 733 | (QPI). At the driver, the term "socket" means one QPI. This is | ||
| 734 | associated with a physical CPU socket. | ||
| 735 | |||
| 736 | Each MC have 3 physical read channels, 3 physical write channels and | ||
| 737 | 3 logic channels. The driver currenty sees it as just 3 channels. | ||
| 738 | Each channel can have up to 3 DIMMs. | ||
| 739 | |||
| 740 | The minimum known unity is DIMMs. There are no information about csrows. | ||
| 741 | As EDAC API maps the minimum unity is csrows, the driver sequencially | ||
| 742 | maps channel/dimm into different csrows. | ||
| 743 | |||
| 744 | For example, suposing the following layout: | ||
| 745 | Ch0 phy rd0, wr0 (0x063f4031): 2 ranks, UDIMMs | ||
| 746 | dimm 0 1024 Mb offset: 0, bank: 8, rank: 1, row: 0x4000, col: 0x400 | ||
| 747 | dimm 1 1024 Mb offset: 4, bank: 8, rank: 1, row: 0x4000, col: 0x400 | ||
| 748 | Ch1 phy rd1, wr1 (0x063f4031): 2 ranks, UDIMMs | ||
| 749 | dimm 0 1024 Mb offset: 0, bank: 8, rank: 1, row: 0x4000, col: 0x400 | ||
| 750 | Ch2 phy rd3, wr3 (0x063f4031): 2 ranks, UDIMMs | ||
| 751 | dimm 0 1024 Mb offset: 0, bank: 8, rank: 1, row: 0x4000, col: 0x400 | ||
| 752 | The driver will map it as: | ||
| 753 | csrow0: channel 0, dimm0 | ||
| 754 | csrow1: channel 0, dimm1 | ||
| 755 | csrow2: channel 1, dimm0 | ||
| 756 | csrow3: channel 2, dimm0 | ||
| 757 | |||
| 758 | exports one | ||
| 759 | DIMM per csrow. | ||
| 760 | |||
| 761 | Each QPI is exported as a different memory controller. | ||
| 762 | |||
| 763 | 2) Nehalem MC has the hability to generate errors. The driver implements this | ||
| 764 | functionality via some error injection nodes: | ||
| 765 | |||
| 766 | For injecting a memory error, there are some sysfs nodes, under | ||
| 767 | /sys/devices/system/edac/mc/mc?/: | ||
| 768 | |||
| 769 | inject_addrmatch/*: | ||
| 770 | Controls the error injection mask register. It is possible to specify | ||
| 771 | several characteristics of the address to match an error code: | ||
| 772 | dimm = the affected dimm. Numbers are relative to a channel; | ||
| 773 | rank = the memory rank; | ||
| 774 | channel = the channel that will generate an error; | ||
| 775 | bank = the affected bank; | ||
| 776 | page = the page address; | ||
| 777 | column (or col) = the address column. | ||
| 778 | each of the above values can be set to "any" to match any valid value. | ||
| 779 | |||
| 780 | At driver init, all values are set to any. | ||
| 781 | |||
| 782 | For example, to generate an error at rank 1 of dimm 2, for any channel, | ||
| 783 | any bank, any page, any column: | ||
| 784 | echo 2 >/sys/devices/system/edac/mc/mc0/inject_addrmatch/dimm | ||
| 785 | echo 1 >/sys/devices/system/edac/mc/mc0/inject_addrmatch/rank | ||
| 786 | |||
| 787 | To return to the default behaviour of matching any, you can do: | ||
| 788 | echo any >/sys/devices/system/edac/mc/mc0/inject_addrmatch/dimm | ||
| 789 | echo any >/sys/devices/system/edac/mc/mc0/inject_addrmatch/rank | ||
| 790 | |||
| 791 | inject_eccmask: | ||
| 792 | specifies what bits will have troubles, | ||
| 793 | |||
| 794 | inject_section: | ||
| 795 | specifies what ECC cache section will get the error: | ||
| 796 | 3 for both | ||
| 797 | 2 for the highest | ||
| 798 | 1 for the lowest | ||
| 799 | |||
| 800 | inject_type: | ||
| 801 | specifies the type of error, being a combination of the following bits: | ||
| 802 | bit 0 - repeat | ||
| 803 | bit 1 - ecc | ||
| 804 | bit 2 - parity | ||
| 805 | |||
| 806 | inject_enable starts the error generation when something different | ||
| 807 | than 0 is written. | ||
| 808 | |||
| 809 | All inject vars can be read. root permission is needed for write. | ||
| 810 | |||
| 811 | Datasheet states that the error will only be generated after a write on an | ||
| 812 | address that matches inject_addrmatch. It seems, however, that reading will | ||
| 813 | also produce an error. | ||
| 814 | |||
| 815 | For example, the following code will generate an error for any write access | ||
| 816 | at socket 0, on any DIMM/address on channel 2: | ||
| 817 | |||
| 818 | echo 2 >/sys/devices/system/edac/mc/mc0/inject_addrmatch/channel | ||
| 819 | echo 2 >/sys/devices/system/edac/mc/mc0/inject_type | ||
| 820 | echo 64 >/sys/devices/system/edac/mc/mc0/inject_eccmask | ||
| 821 | echo 3 >/sys/devices/system/edac/mc/mc0/inject_section | ||
| 822 | echo 1 >/sys/devices/system/edac/mc/mc0/inject_enable | ||
| 823 | dd if=/dev/mem of=/dev/null seek=16k bs=4k count=1 >& /dev/null | ||
| 824 | |||
| 825 | For socket 1, it is needed to replace "mc0" by "mc1" at the above | ||
| 826 | commands. | ||
| 827 | |||
| 828 | The generated error message will look like: | ||
| 829 | |||
| 830 | EDAC MC0: UE row 0, channel-a= 0 channel-b= 0 labels "-": NON_FATAL (addr = 0x0075b980, socket=0, Dimm=0, Channel=2, syndrome=0x00000040, count=1, Err=8c0000400001009f:4000080482 (read error: read ECC error)) | ||
| 831 | |||
| 832 | 3) Nehalem specific Corrected Error memory counters | ||
| 833 | |||
| 834 | Nehalem have some registers to count memory errors. The driver uses those | ||
| 835 | registers to report Corrected Errors on devices with Registered Dimms. | ||
| 836 | |||
| 837 | However, those counters don't work with Unregistered Dimms. As the chipset | ||
| 838 | offers some counters that also work with UDIMMS (but with a worse level of | ||
| 839 | granularity than the default ones), the driver exposes those registers for | ||
| 840 | UDIMM memories. | ||
| 841 | |||
| 842 | They can be read by looking at the contents of all_channel_counts/ | ||
| 843 | |||
| 844 | $ for i in /sys/devices/system/edac/mc/mc0/all_channel_counts/*; do echo $i; cat $i; done | ||
| 845 | /sys/devices/system/edac/mc/mc0/all_channel_counts/udimm0 | ||
| 846 | 0 | ||
| 847 | /sys/devices/system/edac/mc/mc0/all_channel_counts/udimm1 | ||
| 848 | 0 | ||
| 849 | /sys/devices/system/edac/mc/mc0/all_channel_counts/udimm2 | ||
| 850 | 0 | ||
| 851 | |||
| 852 | What happens here is that errors on different csrows, but at the same | ||
| 853 | dimm number will increment the same counter. | ||
| 854 | So, in this memory mapping: | ||
| 855 | csrow0: channel 0, dimm0 | ||
| 856 | csrow1: channel 0, dimm1 | ||
| 857 | csrow2: channel 1, dimm0 | ||
| 858 | csrow3: channel 2, dimm0 | ||
| 859 | The hardware will increment udimm0 for an error at the first dimm at either | ||
| 860 | csrow0, csrow2 or csrow3; | ||
| 861 | The hardware will increment udimm1 for an error at the second dimm at either | ||
| 862 | csrow0, csrow2 or csrow3; | ||
| 863 | The hardware will increment udimm2 for an error at the third dimm at either | ||
| 864 | csrow0, csrow2 or csrow3; | ||
| 865 | |||
| 866 | 4) Standard error counters | ||
| 867 | |||
| 868 | The standard error counters are generated when an mcelog error is received | ||
| 869 | by the driver. Since, with udimm, this is counted by software, it is | ||
| 870 | possible that some errors could be lost. With rdimm's, they displays the | ||
| 871 | contents of the registers | ||
diff --git a/arch/x86/include/asm/pci_x86.h b/arch/x86/include/asm/pci_x86.h index 8d8797eae5d7..cd2a31dc5fb8 100644 --- a/arch/x86/include/asm/pci_x86.h +++ b/arch/x86/include/asm/pci_x86.h | |||
| @@ -53,6 +53,8 @@ extern int pcibios_last_bus; | |||
| 53 | extern struct pci_bus *pci_root_bus; | 53 | extern struct pci_bus *pci_root_bus; |
| 54 | extern struct pci_ops pci_root_ops; | 54 | extern struct pci_ops pci_root_ops; |
| 55 | 55 | ||
| 56 | void pcibios_scan_specific_bus(int busn); | ||
| 57 | |||
| 56 | /* pci-irq.c */ | 58 | /* pci-irq.c */ |
| 57 | 59 | ||
| 58 | struct irq_info { | 60 | struct irq_info { |
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 707165dbc203..18cc42562250 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c | |||
| @@ -36,6 +36,7 @@ | |||
| 36 | #include <linux/fs.h> | 36 | #include <linux/fs.h> |
| 37 | #include <linux/mm.h> | 37 | #include <linux/mm.h> |
| 38 | #include <linux/debugfs.h> | 38 | #include <linux/debugfs.h> |
| 39 | #include <linux/edac_mce.h> | ||
| 39 | 40 | ||
| 40 | #include <asm/processor.h> | 41 | #include <asm/processor.h> |
| 41 | #include <asm/hw_irq.h> | 42 | #include <asm/hw_irq.h> |
| @@ -169,6 +170,15 @@ void mce_log(struct mce *mce) | |||
| 169 | entry = rcu_dereference_check_mce(mcelog.next); | 170 | entry = rcu_dereference_check_mce(mcelog.next); |
| 170 | for (;;) { | 171 | for (;;) { |
| 171 | /* | 172 | /* |
| 173 | * If edac_mce is enabled, it will check the error type | ||
| 174 | * and will process it, if it is a known error. | ||
| 175 | * Otherwise, the error will be sent through mcelog | ||
| 176 | * interface | ||
| 177 | */ | ||
| 178 | if (edac_mce_parse(mce)) | ||
| 179 | return; | ||
| 180 | |||
| 181 | /* | ||
| 172 | * When the buffer fills up discard new entries. | 182 | * When the buffer fills up discard new entries. |
| 173 | * Assume that the earlier errors are the more | 183 | * Assume that the earlier errors are the more |
| 174 | * interesting ones: | 184 | * interesting ones: |
diff --git a/arch/x86/pci/legacy.c b/arch/x86/pci/legacy.c index 0db5eaf54560..8d460eaf524f 100644 --- a/arch/x86/pci/legacy.c +++ b/arch/x86/pci/legacy.c | |||
| @@ -11,28 +11,14 @@ | |||
| 11 | */ | 11 | */ |
| 12 | static void __devinit pcibios_fixup_peer_bridges(void) | 12 | static void __devinit pcibios_fixup_peer_bridges(void) |
| 13 | { | 13 | { |
| 14 | int n, devfn; | 14 | int n; |
| 15 | long node; | ||
| 16 | 15 | ||
| 17 | if (pcibios_last_bus <= 0 || pcibios_last_bus > 0xff) | 16 | if (pcibios_last_bus <= 0 || pcibios_last_bus > 0xff) |
| 18 | return; | 17 | return; |
| 19 | DBG("PCI: Peer bridge fixup\n"); | 18 | DBG("PCI: Peer bridge fixup\n"); |
| 20 | 19 | ||
| 21 | for (n=0; n <= pcibios_last_bus; n++) { | 20 | for (n=0; n <= pcibios_last_bus; n++) |
| 22 | u32 l; | 21 | pcibios_scan_specific_bus(n); |
| 23 | if (pci_find_bus(0, n)) | ||
| 24 | continue; | ||
| 25 | node = get_mp_bus_to_node(n); | ||
| 26 | for (devfn = 0; devfn < 256; devfn += 8) { | ||
| 27 | if (!raw_pci_read(0, n, devfn, PCI_VENDOR_ID, 2, &l) && | ||
| 28 | l != 0x0000 && l != 0xffff) { | ||
| 29 | DBG("Found device at %02x:%02x [%04x]\n", n, devfn, l); | ||
| 30 | printk(KERN_INFO "PCI: Discovered peer bus %02x\n", n); | ||
| 31 | pci_scan_bus_on_node(n, &pci_root_ops, node); | ||
| 32 | break; | ||
| 33 | } | ||
| 34 | } | ||
| 35 | } | ||
| 36 | } | 22 | } |
| 37 | 23 | ||
| 38 | int __init pci_legacy_init(void) | 24 | int __init pci_legacy_init(void) |
| @@ -50,6 +36,28 @@ int __init pci_legacy_init(void) | |||
| 50 | return 0; | 36 | return 0; |
| 51 | } | 37 | } |
| 52 | 38 | ||
| 39 | void pcibios_scan_specific_bus(int busn) | ||
| 40 | { | ||
| 41 | int devfn; | ||
| 42 | long node; | ||
| 43 | u32 l; | ||
| 44 | |||
| 45 | if (pci_find_bus(0, busn)) | ||
| 46 | return; | ||
| 47 | |||
| 48 | node = get_mp_bus_to_node(busn); | ||
| 49 | for (devfn = 0; devfn < 256; devfn += 8) { | ||
| 50 | if (!raw_pci_read(0, busn, devfn, PCI_VENDOR_ID, 2, &l) && | ||
| 51 | l != 0x0000 && l != 0xffff) { | ||
| 52 | DBG("Found device at %02x:%02x [%04x]\n", busn, devfn, l); | ||
| 53 | printk(KERN_INFO "PCI: Discovered peer bus %02x\n", busn); | ||
| 54 | pci_scan_bus_on_node(busn, &pci_root_ops, node); | ||
| 55 | return; | ||
| 56 | } | ||
| 57 | } | ||
| 58 | } | ||
| 59 | EXPORT_SYMBOL_GPL(pcibios_scan_specific_bus); | ||
| 60 | |||
| 53 | int __init pci_subsys_init(void) | 61 | int __init pci_subsys_init(void) |
| 54 | { | 62 | { |
| 55 | /* | 63 | /* |
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig index 55c9c59b3f71..aedef7941b22 100644 --- a/drivers/edac/Kconfig +++ b/drivers/edac/Kconfig | |||
| @@ -69,6 +69,9 @@ config EDAC_MM_EDAC | |||
| 69 | occurred so that a particular failing memory module can be | 69 | occurred so that a particular failing memory module can be |
| 70 | replaced. If unsure, select 'Y'. | 70 | replaced. If unsure, select 'Y'. |
| 71 | 71 | ||
| 72 | config EDAC_MCE | ||
| 73 | bool | ||
| 74 | |||
| 72 | config EDAC_AMD64 | 75 | config EDAC_AMD64 |
| 73 | tristate "AMD64 (Opteron, Athlon64) K8, F10h, F11h" | 76 | tristate "AMD64 (Opteron, Athlon64) K8, F10h, F11h" |
| 74 | depends on EDAC_MM_EDAC && K8_NB && X86_64 && PCI && EDAC_DECODE_MCE | 77 | depends on EDAC_MM_EDAC && K8_NB && X86_64 && PCI && EDAC_DECODE_MCE |
| @@ -166,6 +169,16 @@ config EDAC_I5400 | |||
| 166 | Support for error detection and correction the Intel | 169 | Support for error detection and correction the Intel |
| 167 | i5400 MCH chipset (Seaburg). | 170 | i5400 MCH chipset (Seaburg). |
| 168 | 171 | ||
| 172 | config EDAC_I7CORE | ||
| 173 | tristate "Intel i7 Core (Nehalem) processors" | ||
| 174 | depends on EDAC_MM_EDAC && PCI && X86 | ||
| 175 | select EDAC_MCE | ||
| 176 | help | ||
| 177 | Support for error detection and correction the Intel | ||
| 178 | i7 Core (Nehalem) Integrated Memory Controller that exists on | ||
| 179 | newer processors like i7 Core, i7 Core Extreme, Xeon 35xx | ||
| 180 | and Xeon 55xx processors. | ||
| 181 | |||
| 169 | config EDAC_I82860 | 182 | config EDAC_I82860 |
| 170 | tristate "Intel 82860" | 183 | tristate "Intel 82860" |
| 171 | depends on EDAC_MM_EDAC && PCI && X86_32 | 184 | depends on EDAC_MM_EDAC && PCI && X86_32 |
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile index bc5dc232a0fb..ca6b1bb24ccc 100644 --- a/drivers/edac/Makefile +++ b/drivers/edac/Makefile | |||
| @@ -8,6 +8,7 @@ | |||
| 8 | 8 | ||
| 9 | obj-$(CONFIG_EDAC) := edac_stub.o | 9 | obj-$(CONFIG_EDAC) := edac_stub.o |
| 10 | obj-$(CONFIG_EDAC_MM_EDAC) += edac_core.o | 10 | obj-$(CONFIG_EDAC_MM_EDAC) += edac_core.o |
| 11 | obj-$(CONFIG_EDAC_MCE) += edac_mce.o | ||
| 11 | 12 | ||
| 12 | edac_core-objs := edac_mc.o edac_device.o edac_mc_sysfs.o edac_pci_sysfs.o | 13 | edac_core-objs := edac_mc.o edac_device.o edac_mc_sysfs.o edac_pci_sysfs.o |
| 13 | edac_core-objs += edac_module.o edac_device_sysfs.o | 14 | edac_core-objs += edac_module.o edac_device_sysfs.o |
| @@ -23,6 +24,7 @@ obj-$(CONFIG_EDAC_CPC925) += cpc925_edac.o | |||
| 23 | obj-$(CONFIG_EDAC_I5000) += i5000_edac.o | 24 | obj-$(CONFIG_EDAC_I5000) += i5000_edac.o |
| 24 | obj-$(CONFIG_EDAC_I5100) += i5100_edac.o | 25 | obj-$(CONFIG_EDAC_I5100) += i5100_edac.o |
| 25 | obj-$(CONFIG_EDAC_I5400) += i5400_edac.o | 26 | obj-$(CONFIG_EDAC_I5400) += i5400_edac.o |
| 27 | obj-$(CONFIG_EDAC_I7CORE) += i7core_edac.o | ||
| 26 | obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o | 28 | obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o |
| 27 | obj-$(CONFIG_EDAC_E752X) += e752x_edac.o | 29 | obj-$(CONFIG_EDAC_E752X) += e752x_edac.o |
| 28 | obj-$(CONFIG_EDAC_I82443BXGX) += i82443bxgx_edac.o | 30 | obj-$(CONFIG_EDAC_I82443BXGX) += i82443bxgx_edac.o |
diff --git a/drivers/edac/edac_core.h b/drivers/edac/edac_core.h index 001b2e797fb3..efca9343d26a 100644 --- a/drivers/edac/edac_core.h +++ b/drivers/edac/edac_core.h | |||
| @@ -341,12 +341,30 @@ struct csrow_info { | |||
| 341 | struct channel_info *channels; | 341 | struct channel_info *channels; |
| 342 | }; | 342 | }; |
| 343 | 343 | ||
| 344 | struct mcidev_sysfs_group { | ||
| 345 | const char *name; /* group name */ | ||
| 346 | struct mcidev_sysfs_attribute *mcidev_attr; /* group attributes */ | ||
| 347 | }; | ||
| 348 | |||
| 349 | struct mcidev_sysfs_group_kobj { | ||
| 350 | struct list_head list; /* list for all instances within a mc */ | ||
| 351 | |||
| 352 | struct kobject kobj; /* kobj for the group */ | ||
| 353 | |||
| 354 | struct mcidev_sysfs_group *grp; /* group description table */ | ||
| 355 | struct mem_ctl_info *mci; /* the parent */ | ||
| 356 | }; | ||
| 357 | |||
| 344 | /* mcidev_sysfs_attribute structure | 358 | /* mcidev_sysfs_attribute structure |
| 345 | * used for driver sysfs attributes and in mem_ctl_info | 359 | * used for driver sysfs attributes and in mem_ctl_info |
| 346 | * sysfs top level entries | 360 | * sysfs top level entries |
| 347 | */ | 361 | */ |
| 348 | struct mcidev_sysfs_attribute { | 362 | struct mcidev_sysfs_attribute { |
| 349 | struct attribute attr; | 363 | /* It should use either attr or grp */ |
| 364 | struct attribute attr; | ||
| 365 | struct mcidev_sysfs_group *grp; /* Points to a group of attributes */ | ||
| 366 | |||
| 367 | /* Ops for show/store values at the attribute - not used on group */ | ||
| 350 | ssize_t (*show)(struct mem_ctl_info *,char *); | 368 | ssize_t (*show)(struct mem_ctl_info *,char *); |
| 351 | ssize_t (*store)(struct mem_ctl_info *, const char *,size_t); | 369 | ssize_t (*store)(struct mem_ctl_info *, const char *,size_t); |
| 352 | }; | 370 | }; |
| @@ -424,6 +442,9 @@ struct mem_ctl_info { | |||
| 424 | /* edac sysfs device control */ | 442 | /* edac sysfs device control */ |
| 425 | struct kobject edac_mci_kobj; | 443 | struct kobject edac_mci_kobj; |
| 426 | 444 | ||
| 445 | /* list for all grp instances within a mc */ | ||
| 446 | struct list_head grp_kobj_list; | ||
| 447 | |||
| 427 | /* Additional top controller level attributes, but specified | 448 | /* Additional top controller level attributes, but specified |
| 428 | * by the low level driver. | 449 | * by the low level driver. |
| 429 | * | 450 | * |
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c index 418b65f1a1da..c200c2fd43ea 100644 --- a/drivers/edac/edac_mc_sysfs.c +++ b/drivers/edac/edac_mc_sysfs.c | |||
| @@ -557,6 +557,8 @@ static ssize_t mcidev_show(struct kobject *kobj, struct attribute *attr, | |||
| 557 | struct mem_ctl_info *mem_ctl_info = to_mci(kobj); | 557 | struct mem_ctl_info *mem_ctl_info = to_mci(kobj); |
| 558 | struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr); | 558 | struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr); |
| 559 | 559 | ||
| 560 | debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info); | ||
| 561 | |||
| 560 | if (mcidev_attr->show) | 562 | if (mcidev_attr->show) |
| 561 | return mcidev_attr->show(mem_ctl_info, buffer); | 563 | return mcidev_attr->show(mem_ctl_info, buffer); |
| 562 | 564 | ||
| @@ -569,6 +571,8 @@ static ssize_t mcidev_store(struct kobject *kobj, struct attribute *attr, | |||
| 569 | struct mem_ctl_info *mem_ctl_info = to_mci(kobj); | 571 | struct mem_ctl_info *mem_ctl_info = to_mci(kobj); |
| 570 | struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr); | 572 | struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr); |
| 571 | 573 | ||
| 574 | debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info); | ||
| 575 | |||
| 572 | if (mcidev_attr->store) | 576 | if (mcidev_attr->store) |
| 573 | return mcidev_attr->store(mem_ctl_info, buffer, count); | 577 | return mcidev_attr->store(mem_ctl_info, buffer, count); |
| 574 | 578 | ||
| @@ -726,28 +730,118 @@ void edac_mc_unregister_sysfs_main_kobj(struct mem_ctl_info *mci) | |||
| 726 | 730 | ||
| 727 | #define EDAC_DEVICE_SYMLINK "device" | 731 | #define EDAC_DEVICE_SYMLINK "device" |
| 728 | 732 | ||
| 733 | #define grp_to_mci(k) (container_of(k, struct mcidev_sysfs_group_kobj, kobj)->mci) | ||
| 734 | |||
| 735 | /* MCI show/store functions for top most object */ | ||
| 736 | static ssize_t inst_grp_show(struct kobject *kobj, struct attribute *attr, | ||
| 737 | char *buffer) | ||
| 738 | { | ||
| 739 | struct mem_ctl_info *mem_ctl_info = grp_to_mci(kobj); | ||
| 740 | struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr); | ||
| 741 | |||
| 742 | debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info); | ||
| 743 | |||
| 744 | if (mcidev_attr->show) | ||
| 745 | return mcidev_attr->show(mem_ctl_info, buffer); | ||
| 746 | |||
| 747 | return -EIO; | ||
| 748 | } | ||
| 749 | |||
| 750 | static ssize_t inst_grp_store(struct kobject *kobj, struct attribute *attr, | ||
| 751 | const char *buffer, size_t count) | ||
| 752 | { | ||
| 753 | struct mem_ctl_info *mem_ctl_info = grp_to_mci(kobj); | ||
| 754 | struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr); | ||
| 755 | |||
| 756 | debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info); | ||
| 757 | |||
| 758 | if (mcidev_attr->store) | ||
| 759 | return mcidev_attr->store(mem_ctl_info, buffer, count); | ||
| 760 | |||
| 761 | return -EIO; | ||
| 762 | } | ||
| 763 | |||
| 764 | /* No memory to release for this kobj */ | ||
| 765 | static void edac_inst_grp_release(struct kobject *kobj) | ||
| 766 | { | ||
| 767 | struct mcidev_sysfs_group_kobj *grp; | ||
| 768 | struct mem_ctl_info *mci; | ||
| 769 | |||
| 770 | debugf1("%s()\n", __func__); | ||
| 771 | |||
| 772 | grp = container_of(kobj, struct mcidev_sysfs_group_kobj, kobj); | ||
| 773 | mci = grp->mci; | ||
| 774 | |||
| 775 | kobject_put(&mci->edac_mci_kobj); | ||
| 776 | } | ||
| 777 | |||
| 778 | /* Intermediate show/store table */ | ||
| 779 | static struct sysfs_ops inst_grp_ops = { | ||
| 780 | .show = inst_grp_show, | ||
| 781 | .store = inst_grp_store | ||
| 782 | }; | ||
| 783 | |||
| 784 | /* the kobj_type instance for a instance group */ | ||
| 785 | static struct kobj_type ktype_inst_grp = { | ||
| 786 | .release = edac_inst_grp_release, | ||
| 787 | .sysfs_ops = &inst_grp_ops, | ||
| 788 | }; | ||
| 789 | |||
| 790 | |||
| 729 | /* | 791 | /* |
| 730 | * edac_create_mci_instance_attributes | 792 | * edac_create_mci_instance_attributes |
| 731 | * create MC driver specific attributes at the topmost level | 793 | * create MC driver specific attributes bellow an specified kobj |
| 732 | * directory of this mci instance. | 794 | * This routine calls itself recursively, in order to create an entire |
| 795 | * object tree. | ||
| 733 | */ | 796 | */ |
| 734 | static int edac_create_mci_instance_attributes(struct mem_ctl_info *mci) | 797 | static int edac_create_mci_instance_attributes(struct mem_ctl_info *mci, |
| 798 | struct mcidev_sysfs_attribute *sysfs_attrib, | ||
| 799 | struct kobject *kobj) | ||
| 735 | { | 800 | { |
| 736 | int err; | 801 | int err; |
| 737 | struct mcidev_sysfs_attribute *sysfs_attrib; | ||
| 738 | 802 | ||
| 739 | /* point to the start of the array and iterate over it | 803 | debugf1("%s()\n", __func__); |
| 740 | * adding each attribute listed to this mci instance's kobject | 804 | |
| 741 | */ | 805 | while (sysfs_attrib) { |
| 742 | sysfs_attrib = mci->mc_driver_sysfs_attributes; | 806 | if (sysfs_attrib->grp) { |
| 807 | struct mcidev_sysfs_group_kobj *grp_kobj; | ||
| 808 | |||
| 809 | grp_kobj = kzalloc(sizeof(*grp_kobj), GFP_KERNEL); | ||
| 810 | if (!grp_kobj) | ||
| 811 | return -ENOMEM; | ||
| 812 | |||
| 813 | list_add_tail(&grp_kobj->list, &mci->grp_kobj_list); | ||
| 814 | |||
| 815 | grp_kobj->grp = sysfs_attrib->grp; | ||
| 816 | grp_kobj->mci = mci; | ||
| 817 | |||
| 818 | debugf0("%s() grp %s, mci %p\n", __func__, | ||
| 819 | sysfs_attrib->grp->name, mci); | ||
| 820 | |||
| 821 | err = kobject_init_and_add(&grp_kobj->kobj, | ||
| 822 | &ktype_inst_grp, | ||
| 823 | &mci->edac_mci_kobj, | ||
| 824 | sysfs_attrib->grp->name); | ||
| 825 | if (err) | ||
| 826 | return err; | ||
| 827 | |||
| 828 | err = edac_create_mci_instance_attributes(mci, | ||
| 829 | grp_kobj->grp->mcidev_attr, | ||
| 830 | &grp_kobj->kobj); | ||
| 831 | |||
| 832 | if (err) | ||
| 833 | return err; | ||
| 834 | } else if (sysfs_attrib->attr.name) { | ||
| 835 | debugf0("%s() file %s\n", __func__, | ||
| 836 | sysfs_attrib->attr.name); | ||
| 837 | |||
| 838 | err = sysfs_create_file(kobj, &sysfs_attrib->attr); | ||
| 839 | } else | ||
| 840 | break; | ||
| 743 | 841 | ||
| 744 | while (sysfs_attrib && sysfs_attrib->attr.name) { | ||
| 745 | err = sysfs_create_file(&mci->edac_mci_kobj, | ||
| 746 | (struct attribute*) sysfs_attrib); | ||
| 747 | if (err) { | 842 | if (err) { |
| 748 | return err; | 843 | return err; |
| 749 | } | 844 | } |
| 750 | |||
| 751 | sysfs_attrib++; | 845 | sysfs_attrib++; |
| 752 | } | 846 | } |
| 753 | 847 | ||
| @@ -759,21 +853,44 @@ static int edac_create_mci_instance_attributes(struct mem_ctl_info *mci) | |||
| 759 | * remove MC driver specific attributes at the topmost level | 853 | * remove MC driver specific attributes at the topmost level |
| 760 | * directory of this mci instance. | 854 | * directory of this mci instance. |
| 761 | */ | 855 | */ |
| 762 | static void edac_remove_mci_instance_attributes(struct mem_ctl_info *mci) | 856 | static void edac_remove_mci_instance_attributes(struct mem_ctl_info *mci, |
| 857 | struct mcidev_sysfs_attribute *sysfs_attrib, | ||
| 858 | struct kobject *kobj, int count) | ||
| 763 | { | 859 | { |
| 764 | struct mcidev_sysfs_attribute *sysfs_attrib; | 860 | struct mcidev_sysfs_group_kobj *grp_kobj, *tmp; |
| 765 | 861 | ||
| 766 | /* point to the start of the array and iterate over it | 862 | debugf1("%s()\n", __func__); |
| 767 | * adding each attribute listed to this mci instance's kobject | ||
| 768 | */ | ||
| 769 | sysfs_attrib = mci->mc_driver_sysfs_attributes; | ||
| 770 | 863 | ||
| 771 | /* loop if there are attributes and until we hit a NULL entry */ | 864 | /* |
| 772 | while (sysfs_attrib && sysfs_attrib->attr.name) { | 865 | * loop if there are attributes and until we hit a NULL entry |
| 773 | sysfs_remove_file(&mci->edac_mci_kobj, | 866 | * Remove first all the atributes |
| 774 | (struct attribute *) sysfs_attrib); | 867 | */ |
| 868 | while (sysfs_attrib) { | ||
| 869 | if (sysfs_attrib->grp) { | ||
| 870 | list_for_each_entry(grp_kobj, &mci->grp_kobj_list, | ||
| 871 | list) | ||
| 872 | if (grp_kobj->grp == sysfs_attrib->grp) | ||
| 873 | edac_remove_mci_instance_attributes(mci, | ||
| 874 | grp_kobj->grp->mcidev_attr, | ||
| 875 | &grp_kobj->kobj, count + 1); | ||
| 876 | } else if (sysfs_attrib->attr.name) { | ||
| 877 | debugf0("%s() file %s\n", __func__, | ||
| 878 | sysfs_attrib->attr.name); | ||
| 879 | sysfs_remove_file(kobj, &sysfs_attrib->attr); | ||
| 880 | } else | ||
| 881 | break; | ||
| 775 | sysfs_attrib++; | 882 | sysfs_attrib++; |
| 776 | } | 883 | } |
| 884 | |||
| 885 | /* | ||
| 886 | * Now that all attributes got removed, it is save to remove all groups | ||
| 887 | */ | ||
| 888 | if (!count) | ||
| 889 | list_for_each_entry_safe(grp_kobj, tmp, &mci->grp_kobj_list, | ||
| 890 | list) { | ||
| 891 | debugf0("%s() grp %s\n", __func__, grp_kobj->grp->name); | ||
| 892 | kobject_put(&grp_kobj->kobj); | ||
| 893 | } | ||
| 777 | } | 894 | } |
| 778 | 895 | ||
| 779 | 896 | ||
| @@ -794,6 +911,8 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci) | |||
| 794 | 911 | ||
| 795 | debugf0("%s() idx=%d\n", __func__, mci->mc_idx); | 912 | debugf0("%s() idx=%d\n", __func__, mci->mc_idx); |
| 796 | 913 | ||
| 914 | INIT_LIST_HEAD(&mci->grp_kobj_list); | ||
| 915 | |||
| 797 | /* create a symlink for the device */ | 916 | /* create a symlink for the device */ |
| 798 | err = sysfs_create_link(kobj_mci, &mci->dev->kobj, | 917 | err = sysfs_create_link(kobj_mci, &mci->dev->kobj, |
| 799 | EDAC_DEVICE_SYMLINK); | 918 | EDAC_DEVICE_SYMLINK); |
| @@ -806,7 +925,9 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci) | |||
| 806 | * then create them now for the driver. | 925 | * then create them now for the driver. |
| 807 | */ | 926 | */ |
| 808 | if (mci->mc_driver_sysfs_attributes) { | 927 | if (mci->mc_driver_sysfs_attributes) { |
| 809 | err = edac_create_mci_instance_attributes(mci); | 928 | err = edac_create_mci_instance_attributes(mci, |
| 929 | mci->mc_driver_sysfs_attributes, | ||
| 930 | &mci->edac_mci_kobj); | ||
| 810 | if (err) { | 931 | if (err) { |
| 811 | debugf1("%s() failure to create mci attributes\n", | 932 | debugf1("%s() failure to create mci attributes\n", |
| 812 | __func__); | 933 | __func__); |
| @@ -841,7 +962,8 @@ fail1: | |||
| 841 | } | 962 | } |
| 842 | 963 | ||
| 843 | /* remove the mci instance's attributes, if any */ | 964 | /* remove the mci instance's attributes, if any */ |
| 844 | edac_remove_mci_instance_attributes(mci); | 965 | edac_remove_mci_instance_attributes(mci, |
| 966 | mci->mc_driver_sysfs_attributes, &mci->edac_mci_kobj, 0); | ||
| 845 | 967 | ||
| 846 | /* remove the symlink */ | 968 | /* remove the symlink */ |
| 847 | sysfs_remove_link(kobj_mci, EDAC_DEVICE_SYMLINK); | 969 | sysfs_remove_link(kobj_mci, EDAC_DEVICE_SYMLINK); |
| @@ -875,8 +997,9 @@ void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci) | |||
| 875 | debugf0("%s() remove_mci_instance\n", __func__); | 997 | debugf0("%s() remove_mci_instance\n", __func__); |
| 876 | 998 | ||
| 877 | /* remove this mci instance's attribtes */ | 999 | /* remove this mci instance's attribtes */ |
| 878 | edac_remove_mci_instance_attributes(mci); | 1000 | edac_remove_mci_instance_attributes(mci, |
| 879 | 1001 | mci->mc_driver_sysfs_attributes, | |
| 1002 | &mci->edac_mci_kobj, 0); | ||
| 880 | debugf0("%s() unregister this mci kobj\n", __func__); | 1003 | debugf0("%s() unregister this mci kobj\n", __func__); |
| 881 | 1004 | ||
| 882 | /* unregister this instance's kobject */ | 1005 | /* unregister this instance's kobject */ |
diff --git a/drivers/edac/edac_mce.c b/drivers/edac/edac_mce.c new file mode 100644 index 000000000000..9ccdc5b140e7 --- /dev/null +++ b/drivers/edac/edac_mce.c | |||
| @@ -0,0 +1,61 @@ | |||
| 1 | /* Provides edac interface to mcelog events | ||
| 2 | * | ||
| 3 | * This file may be distributed under the terms of the | ||
| 4 | * GNU General Public License version 2. | ||
| 5 | * | ||
| 6 | * Copyright (c) 2009 by: | ||
| 7 | * Mauro Carvalho Chehab <mchehab@redhat.com> | ||
| 8 | * | ||
| 9 | * Red Hat Inc. http://www.redhat.com | ||
| 10 | */ | ||
| 11 | |||
| 12 | #include <linux/module.h> | ||
| 13 | #include <linux/edac_mce.h> | ||
| 14 | #include <asm/mce.h> | ||
| 15 | |||
| 16 | int edac_mce_enabled; | ||
| 17 | EXPORT_SYMBOL_GPL(edac_mce_enabled); | ||
| 18 | |||
| 19 | |||
| 20 | /* | ||
| 21 | * Extension interface | ||
| 22 | */ | ||
| 23 | |||
| 24 | static LIST_HEAD(edac_mce_list); | ||
| 25 | static DEFINE_MUTEX(edac_mce_lock); | ||
| 26 | |||
| 27 | int edac_mce_register(struct edac_mce *edac_mce) | ||
| 28 | { | ||
| 29 | mutex_lock(&edac_mce_lock); | ||
| 30 | list_add_tail(&edac_mce->list, &edac_mce_list); | ||
| 31 | mutex_unlock(&edac_mce_lock); | ||
| 32 | return 0; | ||
| 33 | } | ||
| 34 | EXPORT_SYMBOL(edac_mce_register); | ||
| 35 | |||
| 36 | void edac_mce_unregister(struct edac_mce *edac_mce) | ||
| 37 | { | ||
| 38 | mutex_lock(&edac_mce_lock); | ||
| 39 | list_del(&edac_mce->list); | ||
| 40 | mutex_unlock(&edac_mce_lock); | ||
| 41 | } | ||
| 42 | EXPORT_SYMBOL(edac_mce_unregister); | ||
| 43 | |||
| 44 | int edac_mce_parse(struct mce *mce) | ||
| 45 | { | ||
| 46 | struct edac_mce *edac_mce; | ||
| 47 | |||
| 48 | list_for_each_entry(edac_mce, &edac_mce_list, list) { | ||
| 49 | if (edac_mce->check_error(edac_mce->priv, mce)) | ||
| 50 | return 1; | ||
| 51 | } | ||
| 52 | |||
| 53 | /* Nobody queued the error */ | ||
| 54 | return 0; | ||
| 55 | } | ||
| 56 | EXPORT_SYMBOL_GPL(edac_mce_parse); | ||
| 57 | |||
| 58 | MODULE_LICENSE("GPL"); | ||
| 59 | MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>"); | ||
| 60 | MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)"); | ||
| 61 | MODULE_DESCRIPTION("EDAC Driver for mcelog captured errors"); | ||
diff --git a/drivers/edac/i7core_edac.c b/drivers/edac/i7core_edac.c new file mode 100644 index 000000000000..6b8b7b41ec5f --- /dev/null +++ b/drivers/edac/i7core_edac.c | |||
| @@ -0,0 +1,2078 @@ | |||
| 1 | /* Intel i7 core/Nehalem Memory Controller kernel module | ||
| 2 | * | ||
| 3 | * This driver supports yhe memory controllers found on the Intel | ||
| 4 | * processor families i7core, i7core 7xx/8xx, i5core, Xeon 35xx, | ||
| 5 | * Xeon 55xx and Xeon 56xx also known as Nehalem, Nehalem-EP, Lynnfield | ||
| 6 | * and Westmere-EP. | ||
| 7 | * | ||
| 8 | * This file may be distributed under the terms of the | ||
| 9 | * GNU General Public License version 2 only. | ||
| 10 | * | ||
| 11 | * Copyright (c) 2009-2010 by: | ||
| 12 | * Mauro Carvalho Chehab <mchehab@redhat.com> | ||
| 13 | * | ||
| 14 | * Red Hat Inc. http://www.redhat.com | ||
| 15 | * | ||
| 16 | * Forked and adapted from the i5400_edac driver | ||
| 17 | * | ||
| 18 | * Based on the following public Intel datasheets: | ||
| 19 | * Intel Core i7 Processor Extreme Edition and Intel Core i7 Processor | ||
| 20 | * Datasheet, Volume 2: | ||
| 21 | * http://download.intel.com/design/processor/datashts/320835.pdf | ||
| 22 | * Intel Xeon Processor 5500 Series Datasheet Volume 2 | ||
| 23 | * http://www.intel.com/Assets/PDF/datasheet/321322.pdf | ||
| 24 | * also available at: | ||
| 25 | * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf | ||
| 26 | */ | ||
| 27 | |||
| 28 | #include <linux/module.h> | ||
| 29 | #include <linux/init.h> | ||
| 30 | #include <linux/pci.h> | ||
| 31 | #include <linux/pci_ids.h> | ||
| 32 | #include <linux/slab.h> | ||
| 33 | #include <linux/delay.h> | ||
| 34 | #include <linux/edac.h> | ||
| 35 | #include <linux/mmzone.h> | ||
| 36 | #include <linux/edac_mce.h> | ||
| 37 | #include <linux/smp.h> | ||
| 38 | #include <asm/processor.h> | ||
| 39 | |||
| 40 | #include "edac_core.h" | ||
| 41 | |||
| 42 | /* | ||
| 43 | * This is used for Nehalem-EP and Nehalem-EX devices, where the non-core | ||
| 44 | * registers start at bus 255, and are not reported by BIOS. | ||
| 45 | * We currently find devices with only 2 sockets. In order to support more QPI | ||
| 46 | * Quick Path Interconnect, just increment this number. | ||
| 47 | */ | ||
| 48 | #define MAX_SOCKET_BUSES 2 | ||
| 49 | |||
| 50 | |||
| 51 | /* | ||
| 52 | * Alter this version for the module when modifications are made | ||
| 53 | */ | ||
| 54 | #define I7CORE_REVISION " Ver: 1.0.0 " __DATE__ | ||
| 55 | #define EDAC_MOD_STR "i7core_edac" | ||
| 56 | |||
| 57 | /* | ||
| 58 | * Debug macros | ||
| 59 | */ | ||
| 60 | #define i7core_printk(level, fmt, arg...) \ | ||
| 61 | edac_printk(level, "i7core", fmt, ##arg) | ||
| 62 | |||
| 63 | #define i7core_mc_printk(mci, level, fmt, arg...) \ | ||
| 64 | edac_mc_chipset_printk(mci, level, "i7core", fmt, ##arg) | ||
| 65 | |||
| 66 | /* | ||
| 67 | * i7core Memory Controller Registers | ||
| 68 | */ | ||
| 69 | |||
| 70 | /* OFFSETS for Device 0 Function 0 */ | ||
| 71 | |||
| 72 | #define MC_CFG_CONTROL 0x90 | ||
| 73 | |||
| 74 | /* OFFSETS for Device 3 Function 0 */ | ||
| 75 | |||
| 76 | #define MC_CONTROL 0x48 | ||
| 77 | #define MC_STATUS 0x4c | ||
| 78 | #define MC_MAX_DOD 0x64 | ||
| 79 | |||
| 80 | /* | ||
| 81 | * OFFSETS for Device 3 Function 4, as inicated on Xeon 5500 datasheet: | ||
| 82 | * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf | ||
| 83 | */ | ||
| 84 | |||
| 85 | #define MC_TEST_ERR_RCV1 0x60 | ||
| 86 | #define DIMM2_COR_ERR(r) ((r) & 0x7fff) | ||
| 87 | |||
| 88 | #define MC_TEST_ERR_RCV0 0x64 | ||
| 89 | #define DIMM1_COR_ERR(r) (((r) >> 16) & 0x7fff) | ||
| 90 | #define DIMM0_COR_ERR(r) ((r) & 0x7fff) | ||
| 91 | |||
| 92 | /* OFFSETS for Device 3 Function 2, as inicated on Xeon 5500 datasheet */ | ||
| 93 | #define MC_COR_ECC_CNT_0 0x80 | ||
| 94 | #define MC_COR_ECC_CNT_1 0x84 | ||
| 95 | #define MC_COR_ECC_CNT_2 0x88 | ||
| 96 | #define MC_COR_ECC_CNT_3 0x8c | ||
| 97 | #define MC_COR_ECC_CNT_4 0x90 | ||
| 98 | #define MC_COR_ECC_CNT_5 0x94 | ||
| 99 | |||
| 100 | #define DIMM_TOP_COR_ERR(r) (((r) >> 16) & 0x7fff) | ||
| 101 | #define DIMM_BOT_COR_ERR(r) ((r) & 0x7fff) | ||
| 102 | |||
| 103 | |||
| 104 | /* OFFSETS for Devices 4,5 and 6 Function 0 */ | ||
| 105 | |||
| 106 | #define MC_CHANNEL_DIMM_INIT_PARAMS 0x58 | ||
| 107 | #define THREE_DIMMS_PRESENT (1 << 24) | ||
| 108 | #define SINGLE_QUAD_RANK_PRESENT (1 << 23) | ||
| 109 | #define QUAD_RANK_PRESENT (1 << 22) | ||
| 110 | #define REGISTERED_DIMM (1 << 15) | ||
| 111 | |||
| 112 | #define MC_CHANNEL_MAPPER 0x60 | ||
| 113 | #define RDLCH(r, ch) ((((r) >> (3 + (ch * 6))) & 0x07) - 1) | ||
| 114 | #define WRLCH(r, ch) ((((r) >> (ch * 6)) & 0x07) - 1) | ||
| 115 | |||
| 116 | #define MC_CHANNEL_RANK_PRESENT 0x7c | ||
| 117 | #define RANK_PRESENT_MASK 0xffff | ||
| 118 | |||
| 119 | #define MC_CHANNEL_ADDR_MATCH 0xf0 | ||
| 120 | #define MC_CHANNEL_ERROR_MASK 0xf8 | ||
| 121 | #define MC_CHANNEL_ERROR_INJECT 0xfc | ||
| 122 | #define INJECT_ADDR_PARITY 0x10 | ||
| 123 | #define INJECT_ECC 0x08 | ||
| 124 | #define MASK_CACHELINE 0x06 | ||
| 125 | #define MASK_FULL_CACHELINE 0x06 | ||
| 126 | #define MASK_MSB32_CACHELINE 0x04 | ||
| 127 | #define MASK_LSB32_CACHELINE 0x02 | ||
| 128 | #define NO_MASK_CACHELINE 0x00 | ||
| 129 | #define REPEAT_EN 0x01 | ||
| 130 | |||
| 131 | /* OFFSETS for Devices 4,5 and 6 Function 1 */ | ||
| 132 | |||
| 133 | #define MC_DOD_CH_DIMM0 0x48 | ||
| 134 | #define MC_DOD_CH_DIMM1 0x4c | ||
| 135 | #define MC_DOD_CH_DIMM2 0x50 | ||
| 136 | #define RANKOFFSET_MASK ((1 << 12) | (1 << 11) | (1 << 10)) | ||
| 137 | #define RANKOFFSET(x) ((x & RANKOFFSET_MASK) >> 10) | ||
| 138 | #define DIMM_PRESENT_MASK (1 << 9) | ||
| 139 | #define DIMM_PRESENT(x) (((x) & DIMM_PRESENT_MASK) >> 9) | ||
| 140 | #define MC_DOD_NUMBANK_MASK ((1 << 8) | (1 << 7)) | ||
| 141 | #define MC_DOD_NUMBANK(x) (((x) & MC_DOD_NUMBANK_MASK) >> 7) | ||
| 142 | #define MC_DOD_NUMRANK_MASK ((1 << 6) | (1 << 5)) | ||
| 143 | #define MC_DOD_NUMRANK(x) (((x) & MC_DOD_NUMRANK_MASK) >> 5) | ||
| 144 | #define MC_DOD_NUMROW_MASK ((1 << 4) | (1 << 3) | (1 << 2)) | ||
| 145 | #define MC_DOD_NUMROW(x) (((x) & MC_DOD_NUMROW_MASK) >> 2) | ||
| 146 | #define MC_DOD_NUMCOL_MASK 3 | ||
| 147 | #define MC_DOD_NUMCOL(x) ((x) & MC_DOD_NUMCOL_MASK) | ||
| 148 | |||
| 149 | #define MC_RANK_PRESENT 0x7c | ||
| 150 | |||
| 151 | #define MC_SAG_CH_0 0x80 | ||
| 152 | #define MC_SAG_CH_1 0x84 | ||
| 153 | #define MC_SAG_CH_2 0x88 | ||
| 154 | #define MC_SAG_CH_3 0x8c | ||
| 155 | #define MC_SAG_CH_4 0x90 | ||
| 156 | #define MC_SAG_CH_5 0x94 | ||
| 157 | #define MC_SAG_CH_6 0x98 | ||
| 158 | #define MC_SAG_CH_7 0x9c | ||
| 159 | |||
| 160 | #define MC_RIR_LIMIT_CH_0 0x40 | ||
| 161 | #define MC_RIR_LIMIT_CH_1 0x44 | ||
| 162 | #define MC_RIR_LIMIT_CH_2 0x48 | ||
| 163 | #define MC_RIR_LIMIT_CH_3 0x4C | ||
| 164 | #define MC_RIR_LIMIT_CH_4 0x50 | ||
| 165 | #define MC_RIR_LIMIT_CH_5 0x54 | ||
| 166 | #define MC_RIR_LIMIT_CH_6 0x58 | ||
| 167 | #define MC_RIR_LIMIT_CH_7 0x5C | ||
| 168 | #define MC_RIR_LIMIT_MASK ((1 << 10) - 1) | ||
| 169 | |||
| 170 | #define MC_RIR_WAY_CH 0x80 | ||
| 171 | #define MC_RIR_WAY_OFFSET_MASK (((1 << 14) - 1) & ~0x7) | ||
| 172 | #define MC_RIR_WAY_RANK_MASK 0x7 | ||
| 173 | |||
| 174 | /* | ||
| 175 | * i7core structs | ||
| 176 | */ | ||
| 177 | |||
| 178 | #define NUM_CHANS 3 | ||
| 179 | #define MAX_DIMMS 3 /* Max DIMMS per channel */ | ||
| 180 | #define MAX_MCR_FUNC 4 | ||
| 181 | #define MAX_CHAN_FUNC 3 | ||
| 182 | |||
| 183 | struct i7core_info { | ||
| 184 | u32 mc_control; | ||
| 185 | u32 mc_status; | ||
| 186 | u32 max_dod; | ||
| 187 | u32 ch_map; | ||
| 188 | }; | ||
| 189 | |||
| 190 | |||
| 191 | struct i7core_inject { | ||
| 192 | int enable; | ||
| 193 | |||
| 194 | u32 section; | ||
| 195 | u32 type; | ||
| 196 | u32 eccmask; | ||
| 197 | |||
| 198 | /* Error address mask */ | ||
| 199 | int channel, dimm, rank, bank, page, col; | ||
| 200 | }; | ||
| 201 | |||
| 202 | struct i7core_channel { | ||
| 203 | u32 ranks; | ||
| 204 | u32 dimms; | ||
| 205 | }; | ||
| 206 | |||
| 207 | struct pci_id_descr { | ||
| 208 | int dev; | ||
| 209 | int func; | ||
| 210 | int dev_id; | ||
| 211 | int optional; | ||
| 212 | }; | ||
| 213 | |||
| 214 | struct pci_id_table { | ||
| 215 | struct pci_id_descr *descr; | ||
| 216 | int n_devs; | ||
| 217 | }; | ||
| 218 | |||
| 219 | struct i7core_dev { | ||
| 220 | struct list_head list; | ||
| 221 | u8 socket; | ||
| 222 | struct pci_dev **pdev; | ||
| 223 | int n_devs; | ||
| 224 | struct mem_ctl_info *mci; | ||
| 225 | }; | ||
| 226 | |||
| 227 | struct i7core_pvt { | ||
| 228 | struct pci_dev *pci_noncore; | ||
| 229 | struct pci_dev *pci_mcr[MAX_MCR_FUNC + 1]; | ||
| 230 | struct pci_dev *pci_ch[NUM_CHANS][MAX_CHAN_FUNC + 1]; | ||
| 231 | |||
| 232 | struct i7core_dev *i7core_dev; | ||
| 233 | |||
| 234 | struct i7core_info info; | ||
| 235 | struct i7core_inject inject; | ||
| 236 | struct i7core_channel channel[NUM_CHANS]; | ||
| 237 | |||
| 238 | int channels; /* Number of active channels */ | ||
| 239 | |||
| 240 | int ce_count_available; | ||
| 241 | int csrow_map[NUM_CHANS][MAX_DIMMS]; | ||
| 242 | |||
| 243 | /* ECC corrected errors counts per udimm */ | ||
| 244 | unsigned long udimm_ce_count[MAX_DIMMS]; | ||
| 245 | int udimm_last_ce_count[MAX_DIMMS]; | ||
| 246 | /* ECC corrected errors counts per rdimm */ | ||
| 247 | unsigned long rdimm_ce_count[NUM_CHANS][MAX_DIMMS]; | ||
| 248 | int rdimm_last_ce_count[NUM_CHANS][MAX_DIMMS]; | ||
| 249 | |||
| 250 | unsigned int is_registered; | ||
| 251 | |||
| 252 | /* mcelog glue */ | ||
| 253 | struct edac_mce edac_mce; | ||
| 254 | |||
| 255 | /* Fifo double buffers */ | ||
| 256 | struct mce mce_entry[MCE_LOG_LEN]; | ||
| 257 | struct mce mce_outentry[MCE_LOG_LEN]; | ||
| 258 | |||
| 259 | /* Fifo in/out counters */ | ||
| 260 | unsigned mce_in, mce_out; | ||
| 261 | |||
| 262 | /* Count indicator to show errors not got */ | ||
| 263 | unsigned mce_overrun; | ||
| 264 | }; | ||
| 265 | |||
| 266 | /* Static vars */ | ||
| 267 | static LIST_HEAD(i7core_edac_list); | ||
| 268 | static DEFINE_MUTEX(i7core_edac_lock); | ||
| 269 | |||
| 270 | #define PCI_DESCR(device, function, device_id) \ | ||
| 271 | .dev = (device), \ | ||
| 272 | .func = (function), \ | ||
| 273 | .dev_id = (device_id) | ||
| 274 | |||
| 275 | struct pci_id_descr pci_dev_descr_i7core_nehalem[] = { | ||
| 276 | /* Memory controller */ | ||
| 277 | { PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_I7_MCR) }, | ||
| 278 | { PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_I7_MC_TAD) }, | ||
| 279 | /* Exists only for RDIMM */ | ||
| 280 | { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_I7_MC_RAS), .optional = 1 }, | ||
| 281 | { PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_I7_MC_TEST) }, | ||
| 282 | |||
| 283 | /* Channel 0 */ | ||
| 284 | { PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH0_CTRL) }, | ||
| 285 | { PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH0_ADDR) }, | ||
| 286 | { PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH0_RANK) }, | ||
| 287 | { PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH0_TC) }, | ||
| 288 | |||
| 289 | /* Channel 1 */ | ||
| 290 | { PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH1_CTRL) }, | ||
| 291 | { PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH1_ADDR) }, | ||
| 292 | { PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH1_RANK) }, | ||
| 293 | { PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH1_TC) }, | ||
| 294 | |||
| 295 | /* Channel 2 */ | ||
| 296 | { PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH2_CTRL) }, | ||
| 297 | { PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH2_ADDR) }, | ||
| 298 | { PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH2_RANK) }, | ||
| 299 | { PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH2_TC) }, | ||
| 300 | |||
| 301 | /* Generic Non-core registers */ | ||
| 302 | /* | ||
| 303 | * This is the PCI device on i7core and on Xeon 35xx (8086:2c41) | ||
| 304 | * On Xeon 55xx, however, it has a different id (8086:2c40). So, | ||
| 305 | * the probing code needs to test for the other address in case of | ||
| 306 | * failure of this one | ||
| 307 | */ | ||
| 308 | { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_I7_NONCORE) }, | ||
| 309 | |||
| 310 | }; | ||
| 311 | |||
| 312 | struct pci_id_descr pci_dev_descr_lynnfield[] = { | ||
| 313 | { PCI_DESCR( 3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR) }, | ||
| 314 | { PCI_DESCR( 3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD) }, | ||
| 315 | { PCI_DESCR( 3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST) }, | ||
| 316 | |||
| 317 | { PCI_DESCR( 4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL) }, | ||
| 318 | { PCI_DESCR( 4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR) }, | ||
| 319 | { PCI_DESCR( 4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK) }, | ||
| 320 | { PCI_DESCR( 4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC) }, | ||
| 321 | |||
| 322 | { PCI_DESCR( 5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL) }, | ||
| 323 | { PCI_DESCR( 5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR) }, | ||
| 324 | { PCI_DESCR( 5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK) }, | ||
| 325 | { PCI_DESCR( 5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC) }, | ||
| 326 | |||
| 327 | /* | ||
| 328 | * This is the PCI device has an alternate address on some | ||
| 329 | * processors like Core i7 860 | ||
| 330 | */ | ||
| 331 | { PCI_DESCR( 0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE) }, | ||
| 332 | }; | ||
| 333 | |||
| 334 | struct pci_id_descr pci_dev_descr_i7core_westmere[] = { | ||
| 335 | /* Memory controller */ | ||
| 336 | { PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR_REV2) }, | ||
| 337 | { PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD_REV2) }, | ||
| 338 | /* Exists only for RDIMM */ | ||
| 339 | { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_RAS_REV2), .optional = 1 }, | ||
| 340 | { PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST_REV2) }, | ||
| 341 | |||
| 342 | /* Channel 0 */ | ||
| 343 | { PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL_REV2) }, | ||
| 344 | { PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR_REV2) }, | ||
| 345 | { PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK_REV2) }, | ||
| 346 | { PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC_REV2) }, | ||
| 347 | |||
| 348 | /* Channel 1 */ | ||
| 349 | { PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL_REV2) }, | ||
| 350 | { PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR_REV2) }, | ||
| 351 | { PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK_REV2) }, | ||
| 352 | { PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC_REV2) }, | ||
| 353 | |||
| 354 | /* Channel 2 */ | ||
| 355 | { PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_CTRL_REV2) }, | ||
| 356 | { PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_ADDR_REV2) }, | ||
| 357 | { PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_RANK_REV2) }, | ||
| 358 | { PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_TC_REV2) }, | ||
| 359 | |||
| 360 | /* Generic Non-core registers */ | ||
| 361 | { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2) }, | ||
| 362 | |||
| 363 | }; | ||
| 364 | |||
| 365 | #define PCI_ID_TABLE_ENTRY(A) { A, ARRAY_SIZE(A) } | ||
| 366 | struct pci_id_table pci_dev_table[] = { | ||
| 367 | PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_nehalem), | ||
| 368 | PCI_ID_TABLE_ENTRY(pci_dev_descr_lynnfield), | ||
| 369 | PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_westmere), | ||
| 370 | }; | ||
| 371 | |||
| 372 | /* | ||
| 373 | * pci_device_id table for which devices we are looking for | ||
| 374 | */ | ||
| 375 | static const struct pci_device_id i7core_pci_tbl[] __devinitdata = { | ||
| 376 | {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_X58_HUB_MGMT)}, | ||
| 377 | {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_LINK0)}, | ||
| 378 | {0,} /* 0 terminated list. */ | ||
| 379 | }; | ||
| 380 | |||
| 381 | static struct edac_pci_ctl_info *i7core_pci; | ||
| 382 | |||
| 383 | /**************************************************************************** | ||
| 384 | Anciliary status routines | ||
| 385 | ****************************************************************************/ | ||
| 386 | |||
| 387 | /* MC_CONTROL bits */ | ||
| 388 | #define CH_ACTIVE(pvt, ch) ((pvt)->info.mc_control & (1 << (8 + ch))) | ||
| 389 | #define ECCx8(pvt) ((pvt)->info.mc_control & (1 << 1)) | ||
| 390 | |||
| 391 | /* MC_STATUS bits */ | ||
| 392 | #define ECC_ENABLED(pvt) ((pvt)->info.mc_status & (1 << 4)) | ||
| 393 | #define CH_DISABLED(pvt, ch) ((pvt)->info.mc_status & (1 << ch)) | ||
| 394 | |||
| 395 | /* MC_MAX_DOD read functions */ | ||
| 396 | static inline int numdimms(u32 dimms) | ||
| 397 | { | ||
| 398 | return (dimms & 0x3) + 1; | ||
| 399 | } | ||
| 400 | |||
| 401 | static inline int numrank(u32 rank) | ||
| 402 | { | ||
| 403 | static int ranks[4] = { 1, 2, 4, -EINVAL }; | ||
| 404 | |||
| 405 | return ranks[rank & 0x3]; | ||
| 406 | } | ||
| 407 | |||
| 408 | static inline int numbank(u32 bank) | ||
| 409 | { | ||
| 410 | static int banks[4] = { 4, 8, 16, -EINVAL }; | ||
| 411 | |||
| 412 | return banks[bank & 0x3]; | ||
| 413 | } | ||
| 414 | |||
| 415 | static inline int numrow(u32 row) | ||
| 416 | { | ||
| 417 | static int rows[8] = { | ||
| 418 | 1 << 12, 1 << 13, 1 << 14, 1 << 15, | ||
| 419 | 1 << 16, -EINVAL, -EINVAL, -EINVAL, | ||
| 420 | }; | ||
| 421 | |||
| 422 | return rows[row & 0x7]; | ||
| 423 | } | ||
| 424 | |||
| 425 | static inline int numcol(u32 col) | ||
| 426 | { | ||
| 427 | static int cols[8] = { | ||
| 428 | 1 << 10, 1 << 11, 1 << 12, -EINVAL, | ||
| 429 | }; | ||
| 430 | return cols[col & 0x3]; | ||
| 431 | } | ||
| 432 | |||
| 433 | static struct i7core_dev *get_i7core_dev(u8 socket) | ||
| 434 | { | ||
| 435 | struct i7core_dev *i7core_dev; | ||
| 436 | |||
| 437 | list_for_each_entry(i7core_dev, &i7core_edac_list, list) { | ||
| 438 | if (i7core_dev->socket == socket) | ||
| 439 | return i7core_dev; | ||
| 440 | } | ||
| 441 | |||
| 442 | return NULL; | ||
| 443 | } | ||
| 444 | |||
| 445 | /**************************************************************************** | ||
| 446 | Memory check routines | ||
| 447 | ****************************************************************************/ | ||
| 448 | static struct pci_dev *get_pdev_slot_func(u8 socket, unsigned slot, | ||
| 449 | unsigned func) | ||
| 450 | { | ||
| 451 | struct i7core_dev *i7core_dev = get_i7core_dev(socket); | ||
| 452 | int i; | ||
| 453 | |||
| 454 | if (!i7core_dev) | ||
| 455 | return NULL; | ||
| 456 | |||
| 457 | for (i = 0; i < i7core_dev->n_devs; i++) { | ||
| 458 | if (!i7core_dev->pdev[i]) | ||
| 459 | continue; | ||
| 460 | |||
| 461 | if (PCI_SLOT(i7core_dev->pdev[i]->devfn) == slot && | ||
| 462 | PCI_FUNC(i7core_dev->pdev[i]->devfn) == func) { | ||
| 463 | return i7core_dev->pdev[i]; | ||
| 464 | } | ||
| 465 | } | ||
| 466 | |||
| 467 | return NULL; | ||
| 468 | } | ||
| 469 | |||
| 470 | /** | ||
| 471 | * i7core_get_active_channels() - gets the number of channels and csrows | ||
| 472 | * @socket: Quick Path Interconnect socket | ||
| 473 | * @channels: Number of channels that will be returned | ||
| 474 | * @csrows: Number of csrows found | ||
| 475 | * | ||
| 476 | * Since EDAC core needs to know in advance the number of available channels | ||
| 477 | * and csrows, in order to allocate memory for csrows/channels, it is needed | ||
| 478 | * to run two similar steps. At the first step, implemented on this function, | ||
| 479 | * it checks the number of csrows/channels present at one socket. | ||
| 480 | * this is used in order to properly allocate the size of mci components. | ||
| 481 | * | ||
| 482 | * It should be noticed that none of the current available datasheets explain | ||
| 483 | * or even mention how csrows are seen by the memory controller. So, we need | ||
| 484 | * to add a fake description for csrows. | ||
| 485 | * So, this driver is attributing one DIMM memory for one csrow. | ||
| 486 | */ | ||
| 487 | static int i7core_get_active_channels(u8 socket, unsigned *channels, | ||
| 488 | unsigned *csrows) | ||
| 489 | { | ||
| 490 | struct pci_dev *pdev = NULL; | ||
| 491 | int i, j; | ||
| 492 | u32 status, control; | ||
| 493 | |||
| 494 | *channels = 0; | ||
| 495 | *csrows = 0; | ||
| 496 | |||
| 497 | pdev = get_pdev_slot_func(socket, 3, 0); | ||
| 498 | if (!pdev) { | ||
| 499 | i7core_printk(KERN_ERR, "Couldn't find socket %d fn 3.0!!!\n", | ||
| 500 | socket); | ||
| 501 | return -ENODEV; | ||
| 502 | } | ||
| 503 | |||
| 504 | /* Device 3 function 0 reads */ | ||
| 505 | pci_read_config_dword(pdev, MC_STATUS, &status); | ||
| 506 | pci_read_config_dword(pdev, MC_CONTROL, &control); | ||
| 507 | |||
| 508 | for (i = 0; i < NUM_CHANS; i++) { | ||
| 509 | u32 dimm_dod[3]; | ||
| 510 | /* Check if the channel is active */ | ||
| 511 | if (!(control & (1 << (8 + i)))) | ||
| 512 | continue; | ||
| 513 | |||
| 514 | /* Check if the channel is disabled */ | ||
| 515 | if (status & (1 << i)) | ||
| 516 | continue; | ||
| 517 | |||
| 518 | pdev = get_pdev_slot_func(socket, i + 4, 1); | ||
| 519 | if (!pdev) { | ||
| 520 | i7core_printk(KERN_ERR, "Couldn't find socket %d " | ||
| 521 | "fn %d.%d!!!\n", | ||
| 522 | socket, i + 4, 1); | ||
| 523 | return -ENODEV; | ||
| 524 | } | ||
| 525 | /* Devices 4-6 function 1 */ | ||
| 526 | pci_read_config_dword(pdev, | ||
| 527 | MC_DOD_CH_DIMM0, &dimm_dod[0]); | ||
| 528 | pci_read_config_dword(pdev, | ||
| 529 | MC_DOD_CH_DIMM1, &dimm_dod[1]); | ||
| 530 | pci_read_config_dword(pdev, | ||
| 531 | MC_DOD_CH_DIMM2, &dimm_dod[2]); | ||
| 532 | |||
| 533 | (*channels)++; | ||
| 534 | |||
| 535 | for (j = 0; j < 3; j++) { | ||
| 536 | if (!DIMM_PRESENT(dimm_dod[j])) | ||
| 537 | continue; | ||
| 538 | (*csrows)++; | ||
| 539 | } | ||
| 540 | } | ||
| 541 | |||
| 542 | debugf0("Number of active channels on socket %d: %d\n", | ||
| 543 | socket, *channels); | ||
| 544 | |||
| 545 | return 0; | ||
| 546 | } | ||
| 547 | |||
| 548 | static int get_dimm_config(struct mem_ctl_info *mci, int *csrow) | ||
| 549 | { | ||
| 550 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 551 | struct csrow_info *csr; | ||
| 552 | struct pci_dev *pdev; | ||
| 553 | int i, j; | ||
| 554 | unsigned long last_page = 0; | ||
| 555 | enum edac_type mode; | ||
| 556 | enum mem_type mtype; | ||
| 557 | |||
| 558 | /* Get data from the MC register, function 0 */ | ||
| 559 | pdev = pvt->pci_mcr[0]; | ||
| 560 | if (!pdev) | ||
| 561 | return -ENODEV; | ||
| 562 | |||
| 563 | /* Device 3 function 0 reads */ | ||
| 564 | pci_read_config_dword(pdev, MC_CONTROL, &pvt->info.mc_control); | ||
| 565 | pci_read_config_dword(pdev, MC_STATUS, &pvt->info.mc_status); | ||
| 566 | pci_read_config_dword(pdev, MC_MAX_DOD, &pvt->info.max_dod); | ||
| 567 | pci_read_config_dword(pdev, MC_CHANNEL_MAPPER, &pvt->info.ch_map); | ||
| 568 | |||
| 569 | debugf0("QPI %d control=0x%08x status=0x%08x dod=0x%08x map=0x%08x\n", | ||
| 570 | pvt->i7core_dev->socket, pvt->info.mc_control, pvt->info.mc_status, | ||
| 571 | pvt->info.max_dod, pvt->info.ch_map); | ||
| 572 | |||
| 573 | if (ECC_ENABLED(pvt)) { | ||
| 574 | debugf0("ECC enabled with x%d SDCC\n", ECCx8(pvt) ? 8 : 4); | ||
| 575 | if (ECCx8(pvt)) | ||
| 576 | mode = EDAC_S8ECD8ED; | ||
| 577 | else | ||
| 578 | mode = EDAC_S4ECD4ED; | ||
| 579 | } else { | ||
| 580 | debugf0("ECC disabled\n"); | ||
| 581 | mode = EDAC_NONE; | ||
| 582 | } | ||
| 583 | |||
| 584 | /* FIXME: need to handle the error codes */ | ||
| 585 | debugf0("DOD Max limits: DIMMS: %d, %d-ranked, %d-banked " | ||
| 586 | "x%x x 0x%x\n", | ||
| 587 | numdimms(pvt->info.max_dod), | ||
| 588 | numrank(pvt->info.max_dod >> 2), | ||
| 589 | numbank(pvt->info.max_dod >> 4), | ||
| 590 | numrow(pvt->info.max_dod >> 6), | ||
| 591 | numcol(pvt->info.max_dod >> 9)); | ||
| 592 | |||
| 593 | for (i = 0; i < NUM_CHANS; i++) { | ||
| 594 | u32 data, dimm_dod[3], value[8]; | ||
| 595 | |||
| 596 | if (!pvt->pci_ch[i][0]) | ||
| 597 | continue; | ||
| 598 | |||
| 599 | if (!CH_ACTIVE(pvt, i)) { | ||
| 600 | debugf0("Channel %i is not active\n", i); | ||
| 601 | continue; | ||
| 602 | } | ||
| 603 | if (CH_DISABLED(pvt, i)) { | ||
| 604 | debugf0("Channel %i is disabled\n", i); | ||
| 605 | continue; | ||
| 606 | } | ||
| 607 | |||
| 608 | /* Devices 4-6 function 0 */ | ||
| 609 | pci_read_config_dword(pvt->pci_ch[i][0], | ||
| 610 | MC_CHANNEL_DIMM_INIT_PARAMS, &data); | ||
| 611 | |||
| 612 | pvt->channel[i].ranks = (data & QUAD_RANK_PRESENT) ? | ||
| 613 | 4 : 2; | ||
| 614 | |||
| 615 | if (data & REGISTERED_DIMM) | ||
| 616 | mtype = MEM_RDDR3; | ||
| 617 | else | ||
| 618 | mtype = MEM_DDR3; | ||
| 619 | #if 0 | ||
| 620 | if (data & THREE_DIMMS_PRESENT) | ||
| 621 | pvt->channel[i].dimms = 3; | ||
| 622 | else if (data & SINGLE_QUAD_RANK_PRESENT) | ||
| 623 | pvt->channel[i].dimms = 1; | ||
| 624 | else | ||
| 625 | pvt->channel[i].dimms = 2; | ||
| 626 | #endif | ||
| 627 | |||
| 628 | /* Devices 4-6 function 1 */ | ||
| 629 | pci_read_config_dword(pvt->pci_ch[i][1], | ||
| 630 | MC_DOD_CH_DIMM0, &dimm_dod[0]); | ||
| 631 | pci_read_config_dword(pvt->pci_ch[i][1], | ||
| 632 | MC_DOD_CH_DIMM1, &dimm_dod[1]); | ||
| 633 | pci_read_config_dword(pvt->pci_ch[i][1], | ||
| 634 | MC_DOD_CH_DIMM2, &dimm_dod[2]); | ||
| 635 | |||
| 636 | debugf0("Ch%d phy rd%d, wr%d (0x%08x): " | ||
| 637 | "%d ranks, %cDIMMs\n", | ||
| 638 | i, | ||
| 639 | RDLCH(pvt->info.ch_map, i), WRLCH(pvt->info.ch_map, i), | ||
| 640 | data, | ||
| 641 | pvt->channel[i].ranks, | ||
| 642 | (data & REGISTERED_DIMM) ? 'R' : 'U'); | ||
| 643 | |||
| 644 | for (j = 0; j < 3; j++) { | ||
| 645 | u32 banks, ranks, rows, cols; | ||
| 646 | u32 size, npages; | ||
| 647 | |||
| 648 | if (!DIMM_PRESENT(dimm_dod[j])) | ||
| 649 | continue; | ||
| 650 | |||
| 651 | banks = numbank(MC_DOD_NUMBANK(dimm_dod[j])); | ||
| 652 | ranks = numrank(MC_DOD_NUMRANK(dimm_dod[j])); | ||
| 653 | rows = numrow(MC_DOD_NUMROW(dimm_dod[j])); | ||
| 654 | cols = numcol(MC_DOD_NUMCOL(dimm_dod[j])); | ||
| 655 | |||
| 656 | /* DDR3 has 8 I/O banks */ | ||
| 657 | size = (rows * cols * banks * ranks) >> (20 - 3); | ||
| 658 | |||
| 659 | pvt->channel[i].dimms++; | ||
| 660 | |||
| 661 | debugf0("\tdimm %d %d Mb offset: %x, " | ||
| 662 | "bank: %d, rank: %d, row: %#x, col: %#x\n", | ||
| 663 | j, size, | ||
| 664 | RANKOFFSET(dimm_dod[j]), | ||
| 665 | banks, ranks, rows, cols); | ||
| 666 | |||
| 667 | #if PAGE_SHIFT > 20 | ||
| 668 | npages = size >> (PAGE_SHIFT - 20); | ||
| 669 | #else | ||
| 670 | npages = size << (20 - PAGE_SHIFT); | ||
| 671 | #endif | ||
| 672 | |||
| 673 | csr = &mci->csrows[*csrow]; | ||
| 674 | csr->first_page = last_page + 1; | ||
| 675 | last_page += npages; | ||
| 676 | csr->last_page = last_page; | ||
| 677 | csr->nr_pages = npages; | ||
| 678 | |||
| 679 | csr->page_mask = 0; | ||
| 680 | csr->grain = 8; | ||
| 681 | csr->csrow_idx = *csrow; | ||
| 682 | csr->nr_channels = 1; | ||
| 683 | |||
| 684 | csr->channels[0].chan_idx = i; | ||
| 685 | csr->channels[0].ce_count = 0; | ||
| 686 | |||
| 687 | pvt->csrow_map[i][j] = *csrow; | ||
| 688 | |||
| 689 | switch (banks) { | ||
| 690 | case 4: | ||
| 691 | csr->dtype = DEV_X4; | ||
| 692 | break; | ||
| 693 | case 8: | ||
| 694 | csr->dtype = DEV_X8; | ||
| 695 | break; | ||
| 696 | case 16: | ||
| 697 | csr->dtype = DEV_X16; | ||
| 698 | break; | ||
| 699 | default: | ||
| 700 | csr->dtype = DEV_UNKNOWN; | ||
| 701 | } | ||
| 702 | |||
| 703 | csr->edac_mode = mode; | ||
| 704 | csr->mtype = mtype; | ||
| 705 | |||
| 706 | (*csrow)++; | ||
| 707 | } | ||
| 708 | |||
| 709 | pci_read_config_dword(pdev, MC_SAG_CH_0, &value[0]); | ||
| 710 | pci_read_config_dword(pdev, MC_SAG_CH_1, &value[1]); | ||
| 711 | pci_read_config_dword(pdev, MC_SAG_CH_2, &value[2]); | ||
| 712 | pci_read_config_dword(pdev, MC_SAG_CH_3, &value[3]); | ||
| 713 | pci_read_config_dword(pdev, MC_SAG_CH_4, &value[4]); | ||
| 714 | pci_read_config_dword(pdev, MC_SAG_CH_5, &value[5]); | ||
| 715 | pci_read_config_dword(pdev, MC_SAG_CH_6, &value[6]); | ||
| 716 | pci_read_config_dword(pdev, MC_SAG_CH_7, &value[7]); | ||
| 717 | debugf1("\t[%i] DIVBY3\tREMOVED\tOFFSET\n", i); | ||
| 718 | for (j = 0; j < 8; j++) | ||
| 719 | debugf1("\t\t%#x\t%#x\t%#x\n", | ||
| 720 | (value[j] >> 27) & 0x1, | ||
| 721 | (value[j] >> 24) & 0x7, | ||
| 722 | (value[j] && ((1 << 24) - 1))); | ||
| 723 | } | ||
| 724 | |||
| 725 | return 0; | ||
| 726 | } | ||
| 727 | |||
| 728 | /**************************************************************************** | ||
| 729 | Error insertion routines | ||
| 730 | ****************************************************************************/ | ||
| 731 | |||
| 732 | /* The i7core has independent error injection features per channel. | ||
| 733 | However, to have a simpler code, we don't allow enabling error injection | ||
| 734 | on more than one channel. | ||
| 735 | Also, since a change at an inject parameter will be applied only at enable, | ||
| 736 | we're disabling error injection on all write calls to the sysfs nodes that | ||
| 737 | controls the error code injection. | ||
| 738 | */ | ||
| 739 | static int disable_inject(struct mem_ctl_info *mci) | ||
| 740 | { | ||
| 741 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 742 | |||
| 743 | pvt->inject.enable = 0; | ||
| 744 | |||
| 745 | if (!pvt->pci_ch[pvt->inject.channel][0]) | ||
| 746 | return -ENODEV; | ||
| 747 | |||
| 748 | pci_write_config_dword(pvt->pci_ch[pvt->inject.channel][0], | ||
| 749 | MC_CHANNEL_ERROR_INJECT, 0); | ||
| 750 | |||
| 751 | return 0; | ||
| 752 | } | ||
| 753 | |||
| 754 | /* | ||
| 755 | * i7core inject inject.section | ||
| 756 | * | ||
| 757 | * accept and store error injection inject.section value | ||
| 758 | * bit 0 - refers to the lower 32-byte half cacheline | ||
| 759 | * bit 1 - refers to the upper 32-byte half cacheline | ||
| 760 | */ | ||
| 761 | static ssize_t i7core_inject_section_store(struct mem_ctl_info *mci, | ||
| 762 | const char *data, size_t count) | ||
| 763 | { | ||
| 764 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 765 | unsigned long value; | ||
| 766 | int rc; | ||
| 767 | |||
| 768 | if (pvt->inject.enable) | ||
| 769 | disable_inject(mci); | ||
| 770 | |||
| 771 | rc = strict_strtoul(data, 10, &value); | ||
| 772 | if ((rc < 0) || (value > 3)) | ||
| 773 | return -EIO; | ||
| 774 | |||
| 775 | pvt->inject.section = (u32) value; | ||
| 776 | return count; | ||
| 777 | } | ||
| 778 | |||
| 779 | static ssize_t i7core_inject_section_show(struct mem_ctl_info *mci, | ||
| 780 | char *data) | ||
| 781 | { | ||
| 782 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 783 | return sprintf(data, "0x%08x\n", pvt->inject.section); | ||
| 784 | } | ||
| 785 | |||
| 786 | /* | ||
| 787 | * i7core inject.type | ||
| 788 | * | ||
| 789 | * accept and store error injection inject.section value | ||
| 790 | * bit 0 - repeat enable - Enable error repetition | ||
| 791 | * bit 1 - inject ECC error | ||
| 792 | * bit 2 - inject parity error | ||
| 793 | */ | ||
| 794 | static ssize_t i7core_inject_type_store(struct mem_ctl_info *mci, | ||
| 795 | const char *data, size_t count) | ||
| 796 | { | ||
| 797 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 798 | unsigned long value; | ||
| 799 | int rc; | ||
| 800 | |||
| 801 | if (pvt->inject.enable) | ||
| 802 | disable_inject(mci); | ||
| 803 | |||
| 804 | rc = strict_strtoul(data, 10, &value); | ||
| 805 | if ((rc < 0) || (value > 7)) | ||
| 806 | return -EIO; | ||
| 807 | |||
| 808 | pvt->inject.type = (u32) value; | ||
| 809 | return count; | ||
| 810 | } | ||
| 811 | |||
| 812 | static ssize_t i7core_inject_type_show(struct mem_ctl_info *mci, | ||
| 813 | char *data) | ||
| 814 | { | ||
| 815 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 816 | return sprintf(data, "0x%08x\n", pvt->inject.type); | ||
| 817 | } | ||
| 818 | |||
| 819 | /* | ||
| 820 | * i7core_inject_inject.eccmask_store | ||
| 821 | * | ||
| 822 | * The type of error (UE/CE) will depend on the inject.eccmask value: | ||
| 823 | * Any bits set to a 1 will flip the corresponding ECC bit | ||
| 824 | * Correctable errors can be injected by flipping 1 bit or the bits within | ||
| 825 | * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or | ||
| 826 | * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an | ||
| 827 | * uncorrectable error to be injected. | ||
| 828 | */ | ||
| 829 | static ssize_t i7core_inject_eccmask_store(struct mem_ctl_info *mci, | ||
| 830 | const char *data, size_t count) | ||
| 831 | { | ||
| 832 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 833 | unsigned long value; | ||
| 834 | int rc; | ||
| 835 | |||
| 836 | if (pvt->inject.enable) | ||
| 837 | disable_inject(mci); | ||
| 838 | |||
| 839 | rc = strict_strtoul(data, 10, &value); | ||
| 840 | if (rc < 0) | ||
| 841 | return -EIO; | ||
| 842 | |||
| 843 | pvt->inject.eccmask = (u32) value; | ||
| 844 | return count; | ||
| 845 | } | ||
| 846 | |||
| 847 | static ssize_t i7core_inject_eccmask_show(struct mem_ctl_info *mci, | ||
| 848 | char *data) | ||
| 849 | { | ||
| 850 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 851 | return sprintf(data, "0x%08x\n", pvt->inject.eccmask); | ||
| 852 | } | ||
| 853 | |||
| 854 | /* | ||
| 855 | * i7core_addrmatch | ||
| 856 | * | ||
| 857 | * The type of error (UE/CE) will depend on the inject.eccmask value: | ||
| 858 | * Any bits set to a 1 will flip the corresponding ECC bit | ||
| 859 | * Correctable errors can be injected by flipping 1 bit or the bits within | ||
| 860 | * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or | ||
| 861 | * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an | ||
| 862 | * uncorrectable error to be injected. | ||
| 863 | */ | ||
| 864 | |||
| 865 | #define DECLARE_ADDR_MATCH(param, limit) \ | ||
| 866 | static ssize_t i7core_inject_store_##param( \ | ||
| 867 | struct mem_ctl_info *mci, \ | ||
| 868 | const char *data, size_t count) \ | ||
| 869 | { \ | ||
| 870 | struct i7core_pvt *pvt; \ | ||
| 871 | long value; \ | ||
| 872 | int rc; \ | ||
| 873 | \ | ||
| 874 | debugf1("%s()\n", __func__); \ | ||
| 875 | pvt = mci->pvt_info; \ | ||
| 876 | \ | ||
| 877 | if (pvt->inject.enable) \ | ||
| 878 | disable_inject(mci); \ | ||
| 879 | \ | ||
| 880 | if (!strcasecmp(data, "any") || !strcasecmp(data, "any\n"))\ | ||
| 881 | value = -1; \ | ||
| 882 | else { \ | ||
| 883 | rc = strict_strtoul(data, 10, &value); \ | ||
| 884 | if ((rc < 0) || (value >= limit)) \ | ||
| 885 | return -EIO; \ | ||
| 886 | } \ | ||
| 887 | \ | ||
| 888 | pvt->inject.param = value; \ | ||
| 889 | \ | ||
| 890 | return count; \ | ||
| 891 | } \ | ||
| 892 | \ | ||
| 893 | static ssize_t i7core_inject_show_##param( \ | ||
| 894 | struct mem_ctl_info *mci, \ | ||
| 895 | char *data) \ | ||
| 896 | { \ | ||
| 897 | struct i7core_pvt *pvt; \ | ||
| 898 | \ | ||
| 899 | pvt = mci->pvt_info; \ | ||
| 900 | debugf1("%s() pvt=%p\n", __func__, pvt); \ | ||
| 901 | if (pvt->inject.param < 0) \ | ||
| 902 | return sprintf(data, "any\n"); \ | ||
| 903 | else \ | ||
| 904 | return sprintf(data, "%d\n", pvt->inject.param);\ | ||
| 905 | } | ||
| 906 | |||
| 907 | #define ATTR_ADDR_MATCH(param) \ | ||
| 908 | { \ | ||
| 909 | .attr = { \ | ||
| 910 | .name = #param, \ | ||
| 911 | .mode = (S_IRUGO | S_IWUSR) \ | ||
| 912 | }, \ | ||
| 913 | .show = i7core_inject_show_##param, \ | ||
| 914 | .store = i7core_inject_store_##param, \ | ||
| 915 | } | ||
| 916 | |||
| 917 | DECLARE_ADDR_MATCH(channel, 3); | ||
| 918 | DECLARE_ADDR_MATCH(dimm, 3); | ||
| 919 | DECLARE_ADDR_MATCH(rank, 4); | ||
| 920 | DECLARE_ADDR_MATCH(bank, 32); | ||
| 921 | DECLARE_ADDR_MATCH(page, 0x10000); | ||
| 922 | DECLARE_ADDR_MATCH(col, 0x4000); | ||
| 923 | |||
| 924 | static int write_and_test(struct pci_dev *dev, int where, u32 val) | ||
| 925 | { | ||
| 926 | u32 read; | ||
| 927 | int count; | ||
| 928 | |||
| 929 | debugf0("setting pci %02x:%02x.%x reg=%02x value=%08x\n", | ||
| 930 | dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn), | ||
| 931 | where, val); | ||
| 932 | |||
| 933 | for (count = 0; count < 10; count++) { | ||
| 934 | if (count) | ||
| 935 | msleep(100); | ||
| 936 | pci_write_config_dword(dev, where, val); | ||
| 937 | pci_read_config_dword(dev, where, &read); | ||
| 938 | |||
| 939 | if (read == val) | ||
| 940 | return 0; | ||
| 941 | } | ||
| 942 | |||
| 943 | i7core_printk(KERN_ERR, "Error during set pci %02x:%02x.%x reg=%02x " | ||
| 944 | "write=%08x. Read=%08x\n", | ||
| 945 | dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn), | ||
| 946 | where, val, read); | ||
| 947 | |||
| 948 | return -EINVAL; | ||
| 949 | } | ||
| 950 | |||
| 951 | /* | ||
| 952 | * This routine prepares the Memory Controller for error injection. | ||
| 953 | * The error will be injected when some process tries to write to the | ||
| 954 | * memory that matches the given criteria. | ||
| 955 | * The criteria can be set in terms of a mask where dimm, rank, bank, page | ||
| 956 | * and col can be specified. | ||
| 957 | * A -1 value for any of the mask items will make the MCU to ignore | ||
| 958 | * that matching criteria for error injection. | ||
| 959 | * | ||
| 960 | * It should be noticed that the error will only happen after a write operation | ||
| 961 | * on a memory that matches the condition. if REPEAT_EN is not enabled at | ||
| 962 | * inject mask, then it will produce just one error. Otherwise, it will repeat | ||
| 963 | * until the injectmask would be cleaned. | ||
| 964 | * | ||
| 965 | * FIXME: This routine assumes that MAXNUMDIMMS value of MC_MAX_DOD | ||
| 966 | * is reliable enough to check if the MC is using the | ||
| 967 | * three channels. However, this is not clear at the datasheet. | ||
| 968 | */ | ||
| 969 | static ssize_t i7core_inject_enable_store(struct mem_ctl_info *mci, | ||
| 970 | const char *data, size_t count) | ||
| 971 | { | ||
| 972 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 973 | u32 injectmask; | ||
| 974 | u64 mask = 0; | ||
| 975 | int rc; | ||
| 976 | long enable; | ||
| 977 | |||
| 978 | if (!pvt->pci_ch[pvt->inject.channel][0]) | ||
| 979 | return 0; | ||
| 980 | |||
| 981 | rc = strict_strtoul(data, 10, &enable); | ||
| 982 | if ((rc < 0)) | ||
| 983 | return 0; | ||
| 984 | |||
| 985 | if (enable) { | ||
| 986 | pvt->inject.enable = 1; | ||
| 987 | } else { | ||
| 988 | disable_inject(mci); | ||
| 989 | return count; | ||
| 990 | } | ||
| 991 | |||
| 992 | /* Sets pvt->inject.dimm mask */ | ||
| 993 | if (pvt->inject.dimm < 0) | ||
| 994 | mask |= 1LL << 41; | ||
| 995 | else { | ||
| 996 | if (pvt->channel[pvt->inject.channel].dimms > 2) | ||
| 997 | mask |= (pvt->inject.dimm & 0x3LL) << 35; | ||
| 998 | else | ||
| 999 | mask |= (pvt->inject.dimm & 0x1LL) << 36; | ||
| 1000 | } | ||
| 1001 | |||
| 1002 | /* Sets pvt->inject.rank mask */ | ||
| 1003 | if (pvt->inject.rank < 0) | ||
| 1004 | mask |= 1LL << 40; | ||
| 1005 | else { | ||
| 1006 | if (pvt->channel[pvt->inject.channel].dimms > 2) | ||
| 1007 | mask |= (pvt->inject.rank & 0x1LL) << 34; | ||
| 1008 | else | ||
| 1009 | mask |= (pvt->inject.rank & 0x3LL) << 34; | ||
| 1010 | } | ||
| 1011 | |||
| 1012 | /* Sets pvt->inject.bank mask */ | ||
| 1013 | if (pvt->inject.bank < 0) | ||
| 1014 | mask |= 1LL << 39; | ||
| 1015 | else | ||
| 1016 | mask |= (pvt->inject.bank & 0x15LL) << 30; | ||
| 1017 | |||
| 1018 | /* Sets pvt->inject.page mask */ | ||
| 1019 | if (pvt->inject.page < 0) | ||
| 1020 | mask |= 1LL << 38; | ||
| 1021 | else | ||
| 1022 | mask |= (pvt->inject.page & 0xffff) << 14; | ||
| 1023 | |||
| 1024 | /* Sets pvt->inject.column mask */ | ||
| 1025 | if (pvt->inject.col < 0) | ||
| 1026 | mask |= 1LL << 37; | ||
| 1027 | else | ||
| 1028 | mask |= (pvt->inject.col & 0x3fff); | ||
| 1029 | |||
| 1030 | /* | ||
| 1031 | * bit 0: REPEAT_EN | ||
| 1032 | * bits 1-2: MASK_HALF_CACHELINE | ||
| 1033 | * bit 3: INJECT_ECC | ||
| 1034 | * bit 4: INJECT_ADDR_PARITY | ||
| 1035 | */ | ||
| 1036 | |||
| 1037 | injectmask = (pvt->inject.type & 1) | | ||
| 1038 | (pvt->inject.section & 0x3) << 1 | | ||
| 1039 | (pvt->inject.type & 0x6) << (3 - 1); | ||
| 1040 | |||
| 1041 | /* Unlock writes to registers - this register is write only */ | ||
| 1042 | pci_write_config_dword(pvt->pci_noncore, | ||
| 1043 | MC_CFG_CONTROL, 0x2); | ||
| 1044 | |||
| 1045 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], | ||
| 1046 | MC_CHANNEL_ADDR_MATCH, mask); | ||
| 1047 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], | ||
| 1048 | MC_CHANNEL_ADDR_MATCH + 4, mask >> 32L); | ||
| 1049 | |||
| 1050 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], | ||
| 1051 | MC_CHANNEL_ERROR_MASK, pvt->inject.eccmask); | ||
| 1052 | |||
| 1053 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], | ||
| 1054 | MC_CHANNEL_ERROR_INJECT, injectmask); | ||
| 1055 | |||
| 1056 | /* | ||
| 1057 | * This is something undocumented, based on my tests | ||
| 1058 | * Without writing 8 to this register, errors aren't injected. Not sure | ||
| 1059 | * why. | ||
| 1060 | */ | ||
| 1061 | pci_write_config_dword(pvt->pci_noncore, | ||
| 1062 | MC_CFG_CONTROL, 8); | ||
| 1063 | |||
| 1064 | debugf0("Error inject addr match 0x%016llx, ecc 0x%08x," | ||
| 1065 | " inject 0x%08x\n", | ||
| 1066 | mask, pvt->inject.eccmask, injectmask); | ||
| 1067 | |||
| 1068 | |||
| 1069 | return count; | ||
| 1070 | } | ||
| 1071 | |||
| 1072 | static ssize_t i7core_inject_enable_show(struct mem_ctl_info *mci, | ||
| 1073 | char *data) | ||
| 1074 | { | ||
| 1075 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1076 | u32 injectmask; | ||
| 1077 | |||
| 1078 | if (!pvt->pci_ch[pvt->inject.channel][0]) | ||
| 1079 | return 0; | ||
| 1080 | |||
| 1081 | pci_read_config_dword(pvt->pci_ch[pvt->inject.channel][0], | ||
| 1082 | MC_CHANNEL_ERROR_INJECT, &injectmask); | ||
| 1083 | |||
| 1084 | debugf0("Inject error read: 0x%018x\n", injectmask); | ||
| 1085 | |||
| 1086 | if (injectmask & 0x0c) | ||
| 1087 | pvt->inject.enable = 1; | ||
| 1088 | |||
| 1089 | return sprintf(data, "%d\n", pvt->inject.enable); | ||
| 1090 | } | ||
| 1091 | |||
| 1092 | #define DECLARE_COUNTER(param) \ | ||
| 1093 | static ssize_t i7core_show_counter_##param( \ | ||
| 1094 | struct mem_ctl_info *mci, \ | ||
| 1095 | char *data) \ | ||
| 1096 | { \ | ||
| 1097 | struct i7core_pvt *pvt = mci->pvt_info; \ | ||
| 1098 | \ | ||
| 1099 | debugf1("%s() \n", __func__); \ | ||
| 1100 | if (!pvt->ce_count_available || (pvt->is_registered)) \ | ||
| 1101 | return sprintf(data, "data unavailable\n"); \ | ||
| 1102 | return sprintf(data, "%lu\n", \ | ||
| 1103 | pvt->udimm_ce_count[param]); \ | ||
| 1104 | } | ||
| 1105 | |||
| 1106 | #define ATTR_COUNTER(param) \ | ||
| 1107 | { \ | ||
| 1108 | .attr = { \ | ||
| 1109 | .name = __stringify(udimm##param), \ | ||
| 1110 | .mode = (S_IRUGO | S_IWUSR) \ | ||
| 1111 | }, \ | ||
| 1112 | .show = i7core_show_counter_##param \ | ||
| 1113 | } | ||
| 1114 | |||
| 1115 | DECLARE_COUNTER(0); | ||
| 1116 | DECLARE_COUNTER(1); | ||
| 1117 | DECLARE_COUNTER(2); | ||
| 1118 | |||
| 1119 | /* | ||
| 1120 | * Sysfs struct | ||
| 1121 | */ | ||
| 1122 | |||
| 1123 | |||
| 1124 | static struct mcidev_sysfs_attribute i7core_addrmatch_attrs[] = { | ||
| 1125 | ATTR_ADDR_MATCH(channel), | ||
| 1126 | ATTR_ADDR_MATCH(dimm), | ||
| 1127 | ATTR_ADDR_MATCH(rank), | ||
| 1128 | ATTR_ADDR_MATCH(bank), | ||
| 1129 | ATTR_ADDR_MATCH(page), | ||
| 1130 | ATTR_ADDR_MATCH(col), | ||
| 1131 | { .attr = { .name = NULL } } | ||
| 1132 | }; | ||
| 1133 | |||
| 1134 | static struct mcidev_sysfs_group i7core_inject_addrmatch = { | ||
| 1135 | .name = "inject_addrmatch", | ||
| 1136 | .mcidev_attr = i7core_addrmatch_attrs, | ||
| 1137 | }; | ||
| 1138 | |||
| 1139 | static struct mcidev_sysfs_attribute i7core_udimm_counters_attrs[] = { | ||
| 1140 | ATTR_COUNTER(0), | ||
| 1141 | ATTR_COUNTER(1), | ||
| 1142 | ATTR_COUNTER(2), | ||
| 1143 | }; | ||
| 1144 | |||
| 1145 | static struct mcidev_sysfs_group i7core_udimm_counters = { | ||
| 1146 | .name = "all_channel_counts", | ||
| 1147 | .mcidev_attr = i7core_udimm_counters_attrs, | ||
| 1148 | }; | ||
| 1149 | |||
| 1150 | static struct mcidev_sysfs_attribute i7core_sysfs_attrs[] = { | ||
| 1151 | { | ||
| 1152 | .attr = { | ||
| 1153 | .name = "inject_section", | ||
| 1154 | .mode = (S_IRUGO | S_IWUSR) | ||
| 1155 | }, | ||
| 1156 | .show = i7core_inject_section_show, | ||
| 1157 | .store = i7core_inject_section_store, | ||
| 1158 | }, { | ||
| 1159 | .attr = { | ||
| 1160 | .name = "inject_type", | ||
| 1161 | .mode = (S_IRUGO | S_IWUSR) | ||
| 1162 | }, | ||
| 1163 | .show = i7core_inject_type_show, | ||
| 1164 | .store = i7core_inject_type_store, | ||
| 1165 | }, { | ||
| 1166 | .attr = { | ||
| 1167 | .name = "inject_eccmask", | ||
| 1168 | .mode = (S_IRUGO | S_IWUSR) | ||
| 1169 | }, | ||
| 1170 | .show = i7core_inject_eccmask_show, | ||
| 1171 | .store = i7core_inject_eccmask_store, | ||
| 1172 | }, { | ||
| 1173 | .grp = &i7core_inject_addrmatch, | ||
| 1174 | }, { | ||
| 1175 | .attr = { | ||
| 1176 | .name = "inject_enable", | ||
| 1177 | .mode = (S_IRUGO | S_IWUSR) | ||
| 1178 | }, | ||
| 1179 | .show = i7core_inject_enable_show, | ||
| 1180 | .store = i7core_inject_enable_store, | ||
| 1181 | }, | ||
| 1182 | { .attr = { .name = NULL } }, /* Reserved for udimm counters */ | ||
| 1183 | { .attr = { .name = NULL } } | ||
| 1184 | }; | ||
| 1185 | |||
| 1186 | /**************************************************************************** | ||
| 1187 | Device initialization routines: put/get, init/exit | ||
| 1188 | ****************************************************************************/ | ||
| 1189 | |||
| 1190 | /* | ||
| 1191 | * i7core_put_devices 'put' all the devices that we have | ||
| 1192 | * reserved via 'get' | ||
| 1193 | */ | ||
| 1194 | static void i7core_put_devices(struct i7core_dev *i7core_dev) | ||
| 1195 | { | ||
| 1196 | int i; | ||
| 1197 | |||
| 1198 | debugf0(__FILE__ ": %s()\n", __func__); | ||
| 1199 | for (i = 0; i < i7core_dev->n_devs; i++) { | ||
| 1200 | struct pci_dev *pdev = i7core_dev->pdev[i]; | ||
| 1201 | if (!pdev) | ||
| 1202 | continue; | ||
| 1203 | debugf0("Removing dev %02x:%02x.%d\n", | ||
| 1204 | pdev->bus->number, | ||
| 1205 | PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn)); | ||
| 1206 | pci_dev_put(pdev); | ||
| 1207 | } | ||
| 1208 | kfree(i7core_dev->pdev); | ||
| 1209 | list_del(&i7core_dev->list); | ||
| 1210 | kfree(i7core_dev); | ||
| 1211 | } | ||
| 1212 | |||
| 1213 | static void i7core_put_all_devices(void) | ||
| 1214 | { | ||
| 1215 | struct i7core_dev *i7core_dev, *tmp; | ||
| 1216 | |||
| 1217 | list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list) | ||
| 1218 | i7core_put_devices(i7core_dev); | ||
| 1219 | } | ||
| 1220 | |||
| 1221 | static void __init i7core_xeon_pci_fixup(struct pci_id_table *table) | ||
| 1222 | { | ||
| 1223 | struct pci_dev *pdev = NULL; | ||
| 1224 | int i; | ||
| 1225 | /* | ||
| 1226 | * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core pci buses | ||
| 1227 | * aren't announced by acpi. So, we need to use a legacy scan probing | ||
| 1228 | * to detect them | ||
| 1229 | */ | ||
| 1230 | while (table && table->descr) { | ||
| 1231 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, table->descr[0].dev_id, NULL); | ||
| 1232 | if (unlikely(!pdev)) { | ||
| 1233 | for (i = 0; i < MAX_SOCKET_BUSES; i++) | ||
| 1234 | pcibios_scan_specific_bus(255-i); | ||
| 1235 | } | ||
| 1236 | table++; | ||
| 1237 | } | ||
| 1238 | } | ||
| 1239 | |||
| 1240 | /* | ||
| 1241 | * i7core_get_devices Find and perform 'get' operation on the MCH's | ||
| 1242 | * device/functions we want to reference for this driver | ||
| 1243 | * | ||
| 1244 | * Need to 'get' device 16 func 1 and func 2 | ||
| 1245 | */ | ||
| 1246 | int i7core_get_onedevice(struct pci_dev **prev, int devno, | ||
| 1247 | struct pci_id_descr *dev_descr, unsigned n_devs) | ||
| 1248 | { | ||
| 1249 | struct i7core_dev *i7core_dev; | ||
| 1250 | |||
| 1251 | struct pci_dev *pdev = NULL; | ||
| 1252 | u8 bus = 0; | ||
| 1253 | u8 socket = 0; | ||
| 1254 | |||
| 1255 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, | ||
| 1256 | dev_descr->dev_id, *prev); | ||
| 1257 | |||
| 1258 | /* | ||
| 1259 | * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core regs | ||
| 1260 | * is at addr 8086:2c40, instead of 8086:2c41. So, we need | ||
| 1261 | * to probe for the alternate address in case of failure | ||
| 1262 | */ | ||
| 1263 | if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_I7_NONCORE && !pdev) | ||
| 1264 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, | ||
| 1265 | PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT, *prev); | ||
| 1266 | |||
| 1267 | if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE && !pdev) | ||
| 1268 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, | ||
| 1269 | PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT, | ||
| 1270 | *prev); | ||
| 1271 | |||
| 1272 | if (!pdev) { | ||
| 1273 | if (*prev) { | ||
| 1274 | *prev = pdev; | ||
| 1275 | return 0; | ||
| 1276 | } | ||
| 1277 | |||
| 1278 | if (dev_descr->optional) | ||
| 1279 | return 0; | ||
| 1280 | |||
| 1281 | if (devno == 0) | ||
| 1282 | return -ENODEV; | ||
| 1283 | |||
| 1284 | i7core_printk(KERN_ERR, | ||
| 1285 | "Device not found: dev %02x.%d PCI ID %04x:%04x\n", | ||
| 1286 | dev_descr->dev, dev_descr->func, | ||
| 1287 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | ||
| 1288 | |||
| 1289 | /* End of list, leave */ | ||
| 1290 | return -ENODEV; | ||
| 1291 | } | ||
| 1292 | bus = pdev->bus->number; | ||
| 1293 | |||
| 1294 | if (bus == 0x3f) | ||
| 1295 | socket = 0; | ||
| 1296 | else | ||
| 1297 | socket = 255 - bus; | ||
| 1298 | |||
| 1299 | i7core_dev = get_i7core_dev(socket); | ||
| 1300 | if (!i7core_dev) { | ||
| 1301 | i7core_dev = kzalloc(sizeof(*i7core_dev), GFP_KERNEL); | ||
| 1302 | if (!i7core_dev) | ||
| 1303 | return -ENOMEM; | ||
| 1304 | i7core_dev->pdev = kzalloc(sizeof(*i7core_dev->pdev) * n_devs, | ||
| 1305 | GFP_KERNEL); | ||
| 1306 | if (!i7core_dev->pdev) { | ||
| 1307 | kfree(i7core_dev); | ||
| 1308 | return -ENOMEM; | ||
| 1309 | } | ||
| 1310 | i7core_dev->socket = socket; | ||
| 1311 | i7core_dev->n_devs = n_devs; | ||
| 1312 | list_add_tail(&i7core_dev->list, &i7core_edac_list); | ||
| 1313 | } | ||
| 1314 | |||
| 1315 | if (i7core_dev->pdev[devno]) { | ||
| 1316 | i7core_printk(KERN_ERR, | ||
| 1317 | "Duplicated device for " | ||
| 1318 | "dev %02x:%02x.%d PCI ID %04x:%04x\n", | ||
| 1319 | bus, dev_descr->dev, dev_descr->func, | ||
| 1320 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | ||
| 1321 | pci_dev_put(pdev); | ||
| 1322 | return -ENODEV; | ||
| 1323 | } | ||
| 1324 | |||
| 1325 | i7core_dev->pdev[devno] = pdev; | ||
| 1326 | |||
| 1327 | /* Sanity check */ | ||
| 1328 | if (unlikely(PCI_SLOT(pdev->devfn) != dev_descr->dev || | ||
| 1329 | PCI_FUNC(pdev->devfn) != dev_descr->func)) { | ||
| 1330 | i7core_printk(KERN_ERR, | ||
| 1331 | "Device PCI ID %04x:%04x " | ||
| 1332 | "has dev %02x:%02x.%d instead of dev %02x:%02x.%d\n", | ||
| 1333 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id, | ||
| 1334 | bus, PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn), | ||
| 1335 | bus, dev_descr->dev, dev_descr->func); | ||
| 1336 | return -ENODEV; | ||
| 1337 | } | ||
| 1338 | |||
| 1339 | /* Be sure that the device is enabled */ | ||
| 1340 | if (unlikely(pci_enable_device(pdev) < 0)) { | ||
| 1341 | i7core_printk(KERN_ERR, | ||
| 1342 | "Couldn't enable " | ||
| 1343 | "dev %02x:%02x.%d PCI ID %04x:%04x\n", | ||
| 1344 | bus, dev_descr->dev, dev_descr->func, | ||
| 1345 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | ||
| 1346 | return -ENODEV; | ||
| 1347 | } | ||
| 1348 | |||
| 1349 | debugf0("Detected socket %d dev %02x:%02x.%d PCI ID %04x:%04x\n", | ||
| 1350 | socket, bus, dev_descr->dev, | ||
| 1351 | dev_descr->func, | ||
| 1352 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | ||
| 1353 | |||
| 1354 | *prev = pdev; | ||
| 1355 | |||
| 1356 | return 0; | ||
| 1357 | } | ||
| 1358 | |||
| 1359 | static int i7core_get_devices(struct pci_id_table *table) | ||
| 1360 | { | ||
| 1361 | int i, rc; | ||
| 1362 | struct pci_dev *pdev = NULL; | ||
| 1363 | struct pci_id_descr *dev_descr; | ||
| 1364 | |||
| 1365 | while (table && table->descr) { | ||
| 1366 | dev_descr = table->descr; | ||
| 1367 | for (i = 0; i < table->n_devs; i++) { | ||
| 1368 | pdev = NULL; | ||
| 1369 | do { | ||
| 1370 | rc = i7core_get_onedevice(&pdev, i, &dev_descr[i], | ||
| 1371 | table->n_devs); | ||
| 1372 | if (rc < 0) { | ||
| 1373 | if (i == 0) { | ||
| 1374 | i = table->n_devs; | ||
| 1375 | break; | ||
| 1376 | } | ||
| 1377 | i7core_put_all_devices(); | ||
| 1378 | return -ENODEV; | ||
| 1379 | } | ||
| 1380 | } while (pdev); | ||
| 1381 | } | ||
| 1382 | table++; | ||
| 1383 | } | ||
| 1384 | |||
| 1385 | return 0; | ||
| 1386 | return 0; | ||
| 1387 | } | ||
| 1388 | |||
| 1389 | static int mci_bind_devs(struct mem_ctl_info *mci, | ||
| 1390 | struct i7core_dev *i7core_dev) | ||
| 1391 | { | ||
| 1392 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1393 | struct pci_dev *pdev; | ||
| 1394 | int i, func, slot; | ||
| 1395 | |||
| 1396 | /* Associates i7core_dev and mci for future usage */ | ||
| 1397 | pvt->i7core_dev = i7core_dev; | ||
| 1398 | i7core_dev->mci = mci; | ||
| 1399 | |||
| 1400 | pvt->is_registered = 0; | ||
| 1401 | for (i = 0; i < i7core_dev->n_devs; i++) { | ||
| 1402 | pdev = i7core_dev->pdev[i]; | ||
| 1403 | if (!pdev) | ||
| 1404 | continue; | ||
| 1405 | |||
| 1406 | func = PCI_FUNC(pdev->devfn); | ||
| 1407 | slot = PCI_SLOT(pdev->devfn); | ||
| 1408 | if (slot == 3) { | ||
| 1409 | if (unlikely(func > MAX_MCR_FUNC)) | ||
| 1410 | goto error; | ||
| 1411 | pvt->pci_mcr[func] = pdev; | ||
| 1412 | } else if (likely(slot >= 4 && slot < 4 + NUM_CHANS)) { | ||
| 1413 | if (unlikely(func > MAX_CHAN_FUNC)) | ||
| 1414 | goto error; | ||
| 1415 | pvt->pci_ch[slot - 4][func] = pdev; | ||
| 1416 | } else if (!slot && !func) | ||
| 1417 | pvt->pci_noncore = pdev; | ||
| 1418 | else | ||
| 1419 | goto error; | ||
| 1420 | |||
| 1421 | debugf0("Associated fn %d.%d, dev = %p, socket %d\n", | ||
| 1422 | PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn), | ||
| 1423 | pdev, i7core_dev->socket); | ||
| 1424 | |||
| 1425 | if (PCI_SLOT(pdev->devfn) == 3 && | ||
| 1426 | PCI_FUNC(pdev->devfn) == 2) | ||
| 1427 | pvt->is_registered = 1; | ||
| 1428 | } | ||
| 1429 | |||
| 1430 | /* | ||
| 1431 | * Add extra nodes to count errors on udimm | ||
| 1432 | * For registered memory, this is not needed, since the counters | ||
| 1433 | * are already displayed at the standard locations | ||
| 1434 | */ | ||
| 1435 | if (!pvt->is_registered) | ||
| 1436 | i7core_sysfs_attrs[ARRAY_SIZE(i7core_sysfs_attrs)-2].grp = | ||
| 1437 | &i7core_udimm_counters; | ||
| 1438 | |||
| 1439 | return 0; | ||
| 1440 | |||
| 1441 | error: | ||
| 1442 | i7core_printk(KERN_ERR, "Device %d, function %d " | ||
| 1443 | "is out of the expected range\n", | ||
| 1444 | slot, func); | ||
| 1445 | return -EINVAL; | ||
| 1446 | } | ||
| 1447 | |||
| 1448 | /**************************************************************************** | ||
| 1449 | Error check routines | ||
| 1450 | ****************************************************************************/ | ||
| 1451 | static void i7core_rdimm_update_csrow(struct mem_ctl_info *mci, | ||
| 1452 | int chan, int dimm, int add) | ||
| 1453 | { | ||
| 1454 | char *msg; | ||
| 1455 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1456 | int row = pvt->csrow_map[chan][dimm], i; | ||
| 1457 | |||
| 1458 | for (i = 0; i < add; i++) { | ||
| 1459 | msg = kasprintf(GFP_KERNEL, "Corrected error " | ||
| 1460 | "(Socket=%d channel=%d dimm=%d)", | ||
| 1461 | pvt->i7core_dev->socket, chan, dimm); | ||
| 1462 | |||
| 1463 | edac_mc_handle_fbd_ce(mci, row, 0, msg); | ||
| 1464 | kfree (msg); | ||
| 1465 | } | ||
| 1466 | } | ||
| 1467 | |||
| 1468 | static void i7core_rdimm_update_ce_count(struct mem_ctl_info *mci, | ||
| 1469 | int chan, int new0, int new1, int new2) | ||
| 1470 | { | ||
| 1471 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1472 | int add0 = 0, add1 = 0, add2 = 0; | ||
| 1473 | /* Updates CE counters if it is not the first time here */ | ||
| 1474 | if (pvt->ce_count_available) { | ||
| 1475 | /* Updates CE counters */ | ||
| 1476 | |||
| 1477 | add2 = new2 - pvt->rdimm_last_ce_count[chan][2]; | ||
| 1478 | add1 = new1 - pvt->rdimm_last_ce_count[chan][1]; | ||
| 1479 | add0 = new0 - pvt->rdimm_last_ce_count[chan][0]; | ||
| 1480 | |||
| 1481 | if (add2 < 0) | ||
| 1482 | add2 += 0x7fff; | ||
| 1483 | pvt->rdimm_ce_count[chan][2] += add2; | ||
| 1484 | |||
| 1485 | if (add1 < 0) | ||
| 1486 | add1 += 0x7fff; | ||
| 1487 | pvt->rdimm_ce_count[chan][1] += add1; | ||
| 1488 | |||
| 1489 | if (add0 < 0) | ||
| 1490 | add0 += 0x7fff; | ||
| 1491 | pvt->rdimm_ce_count[chan][0] += add0; | ||
| 1492 | } else | ||
| 1493 | pvt->ce_count_available = 1; | ||
| 1494 | |||
| 1495 | /* Store the new values */ | ||
| 1496 | pvt->rdimm_last_ce_count[chan][2] = new2; | ||
| 1497 | pvt->rdimm_last_ce_count[chan][1] = new1; | ||
| 1498 | pvt->rdimm_last_ce_count[chan][0] = new0; | ||
| 1499 | |||
| 1500 | /*updated the edac core */ | ||
| 1501 | if (add0 != 0) | ||
| 1502 | i7core_rdimm_update_csrow(mci, chan, 0, add0); | ||
| 1503 | if (add1 != 0) | ||
| 1504 | i7core_rdimm_update_csrow(mci, chan, 1, add1); | ||
| 1505 | if (add2 != 0) | ||
| 1506 | i7core_rdimm_update_csrow(mci, chan, 2, add2); | ||
| 1507 | |||
| 1508 | } | ||
| 1509 | |||
| 1510 | static void i7core_rdimm_check_mc_ecc_err(struct mem_ctl_info *mci) | ||
| 1511 | { | ||
| 1512 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1513 | u32 rcv[3][2]; | ||
| 1514 | int i, new0, new1, new2; | ||
| 1515 | |||
| 1516 | /*Read DEV 3: FUN 2: MC_COR_ECC_CNT regs directly*/ | ||
| 1517 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_0, | ||
| 1518 | &rcv[0][0]); | ||
| 1519 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_1, | ||
| 1520 | &rcv[0][1]); | ||
| 1521 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_2, | ||
| 1522 | &rcv[1][0]); | ||
| 1523 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_3, | ||
| 1524 | &rcv[1][1]); | ||
| 1525 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_4, | ||
| 1526 | &rcv[2][0]); | ||
| 1527 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_5, | ||
| 1528 | &rcv[2][1]); | ||
| 1529 | for (i = 0 ; i < 3; i++) { | ||
| 1530 | debugf3("MC_COR_ECC_CNT%d = 0x%x; MC_COR_ECC_CNT%d = 0x%x\n", | ||
| 1531 | (i * 2), rcv[i][0], (i * 2) + 1, rcv[i][1]); | ||
| 1532 | /*if the channel has 3 dimms*/ | ||
| 1533 | if (pvt->channel[i].dimms > 2) { | ||
| 1534 | new0 = DIMM_BOT_COR_ERR(rcv[i][0]); | ||
| 1535 | new1 = DIMM_TOP_COR_ERR(rcv[i][0]); | ||
| 1536 | new2 = DIMM_BOT_COR_ERR(rcv[i][1]); | ||
| 1537 | } else { | ||
| 1538 | new0 = DIMM_TOP_COR_ERR(rcv[i][0]) + | ||
| 1539 | DIMM_BOT_COR_ERR(rcv[i][0]); | ||
| 1540 | new1 = DIMM_TOP_COR_ERR(rcv[i][1]) + | ||
| 1541 | DIMM_BOT_COR_ERR(rcv[i][1]); | ||
| 1542 | new2 = 0; | ||
| 1543 | } | ||
| 1544 | |||
| 1545 | i7core_rdimm_update_ce_count(mci, i, new0, new1, new2); | ||
| 1546 | } | ||
| 1547 | } | ||
| 1548 | |||
| 1549 | /* This function is based on the device 3 function 4 registers as described on: | ||
| 1550 | * Intel Xeon Processor 5500 Series Datasheet Volume 2 | ||
| 1551 | * http://www.intel.com/Assets/PDF/datasheet/321322.pdf | ||
| 1552 | * also available at: | ||
| 1553 | * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf | ||
| 1554 | */ | ||
| 1555 | static void i7core_udimm_check_mc_ecc_err(struct mem_ctl_info *mci) | ||
| 1556 | { | ||
| 1557 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1558 | u32 rcv1, rcv0; | ||
| 1559 | int new0, new1, new2; | ||
| 1560 | |||
| 1561 | if (!pvt->pci_mcr[4]) { | ||
| 1562 | debugf0("%s MCR registers not found\n", __func__); | ||
| 1563 | return; | ||
| 1564 | } | ||
| 1565 | |||
| 1566 | /* Corrected test errors */ | ||
| 1567 | pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV1, &rcv1); | ||
| 1568 | pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV0, &rcv0); | ||
| 1569 | |||
| 1570 | /* Store the new values */ | ||
| 1571 | new2 = DIMM2_COR_ERR(rcv1); | ||
| 1572 | new1 = DIMM1_COR_ERR(rcv0); | ||
| 1573 | new0 = DIMM0_COR_ERR(rcv0); | ||
| 1574 | |||
| 1575 | /* Updates CE counters if it is not the first time here */ | ||
| 1576 | if (pvt->ce_count_available) { | ||
| 1577 | /* Updates CE counters */ | ||
| 1578 | int add0, add1, add2; | ||
| 1579 | |||
| 1580 | add2 = new2 - pvt->udimm_last_ce_count[2]; | ||
| 1581 | add1 = new1 - pvt->udimm_last_ce_count[1]; | ||
| 1582 | add0 = new0 - pvt->udimm_last_ce_count[0]; | ||
| 1583 | |||
| 1584 | if (add2 < 0) | ||
| 1585 | add2 += 0x7fff; | ||
| 1586 | pvt->udimm_ce_count[2] += add2; | ||
| 1587 | |||
| 1588 | if (add1 < 0) | ||
| 1589 | add1 += 0x7fff; | ||
| 1590 | pvt->udimm_ce_count[1] += add1; | ||
| 1591 | |||
| 1592 | if (add0 < 0) | ||
| 1593 | add0 += 0x7fff; | ||
| 1594 | pvt->udimm_ce_count[0] += add0; | ||
| 1595 | |||
| 1596 | if (add0 | add1 | add2) | ||
| 1597 | i7core_printk(KERN_ERR, "New Corrected error(s): " | ||
| 1598 | "dimm0: +%d, dimm1: +%d, dimm2 +%d\n", | ||
| 1599 | add0, add1, add2); | ||
| 1600 | } else | ||
| 1601 | pvt->ce_count_available = 1; | ||
| 1602 | |||
| 1603 | /* Store the new values */ | ||
| 1604 | pvt->udimm_last_ce_count[2] = new2; | ||
| 1605 | pvt->udimm_last_ce_count[1] = new1; | ||
| 1606 | pvt->udimm_last_ce_count[0] = new0; | ||
| 1607 | } | ||
| 1608 | |||
| 1609 | /* | ||
| 1610 | * According with tables E-11 and E-12 of chapter E.3.3 of Intel 64 and IA-32 | ||
| 1611 | * Architectures Software Developer’s Manual Volume 3B. | ||
| 1612 | * Nehalem are defined as family 0x06, model 0x1a | ||
| 1613 | * | ||
| 1614 | * The MCA registers used here are the following ones: | ||
| 1615 | * struct mce field MCA Register | ||
| 1616 | * m->status MSR_IA32_MC8_STATUS | ||
| 1617 | * m->addr MSR_IA32_MC8_ADDR | ||
| 1618 | * m->misc MSR_IA32_MC8_MISC | ||
| 1619 | * In the case of Nehalem, the error information is masked at .status and .misc | ||
| 1620 | * fields | ||
| 1621 | */ | ||
| 1622 | static void i7core_mce_output_error(struct mem_ctl_info *mci, | ||
| 1623 | struct mce *m) | ||
| 1624 | { | ||
| 1625 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1626 | char *type, *optype, *err, *msg; | ||
| 1627 | unsigned long error = m->status & 0x1ff0000l; | ||
| 1628 | u32 optypenum = (m->status >> 4) & 0x07; | ||
| 1629 | u32 core_err_cnt = (m->status >> 38) && 0x7fff; | ||
| 1630 | u32 dimm = (m->misc >> 16) & 0x3; | ||
| 1631 | u32 channel = (m->misc >> 18) & 0x3; | ||
| 1632 | u32 syndrome = m->misc >> 32; | ||
| 1633 | u32 errnum = find_first_bit(&error, 32); | ||
| 1634 | int csrow; | ||
| 1635 | |||
| 1636 | if (m->mcgstatus & 1) | ||
| 1637 | type = "FATAL"; | ||
| 1638 | else | ||
| 1639 | type = "NON_FATAL"; | ||
| 1640 | |||
| 1641 | switch (optypenum) { | ||
| 1642 | case 0: | ||
| 1643 | optype = "generic undef request"; | ||
| 1644 | break; | ||
| 1645 | case 1: | ||
| 1646 | optype = "read error"; | ||
| 1647 | break; | ||
| 1648 | case 2: | ||
| 1649 | optype = "write error"; | ||
| 1650 | break; | ||
| 1651 | case 3: | ||
| 1652 | optype = "addr/cmd error"; | ||
| 1653 | break; | ||
| 1654 | case 4: | ||
| 1655 | optype = "scrubbing error"; | ||
| 1656 | break; | ||
| 1657 | default: | ||
| 1658 | optype = "reserved"; | ||
| 1659 | break; | ||
| 1660 | } | ||
| 1661 | |||
| 1662 | switch (errnum) { | ||
| 1663 | case 16: | ||
| 1664 | err = "read ECC error"; | ||
| 1665 | break; | ||
| 1666 | case 17: | ||
| 1667 | err = "RAS ECC error"; | ||
| 1668 | break; | ||
| 1669 | case 18: | ||
| 1670 | err = "write parity error"; | ||
| 1671 | break; | ||
| 1672 | case 19: | ||
| 1673 | err = "redundacy loss"; | ||
| 1674 | break; | ||
| 1675 | case 20: | ||
| 1676 | err = "reserved"; | ||
| 1677 | break; | ||
| 1678 | case 21: | ||
| 1679 | err = "memory range error"; | ||
| 1680 | break; | ||
| 1681 | case 22: | ||
| 1682 | err = "RTID out of range"; | ||
| 1683 | break; | ||
| 1684 | case 23: | ||
| 1685 | err = "address parity error"; | ||
| 1686 | break; | ||
| 1687 | case 24: | ||
| 1688 | err = "byte enable parity error"; | ||
| 1689 | break; | ||
| 1690 | default: | ||
| 1691 | err = "unknown"; | ||
| 1692 | } | ||
| 1693 | |||
| 1694 | /* FIXME: should convert addr into bank and rank information */ | ||
| 1695 | msg = kasprintf(GFP_ATOMIC, | ||
| 1696 | "%s (addr = 0x%08llx, cpu=%d, Dimm=%d, Channel=%d, " | ||
| 1697 | "syndrome=0x%08x, count=%d, Err=%08llx:%08llx (%s: %s))\n", | ||
| 1698 | type, (long long) m->addr, m->cpu, dimm, channel, | ||
| 1699 | syndrome, core_err_cnt, (long long)m->status, | ||
| 1700 | (long long)m->misc, optype, err); | ||
| 1701 | |||
| 1702 | debugf0("%s", msg); | ||
| 1703 | |||
| 1704 | csrow = pvt->csrow_map[channel][dimm]; | ||
| 1705 | |||
| 1706 | /* Call the helper to output message */ | ||
| 1707 | if (m->mcgstatus & 1) | ||
| 1708 | edac_mc_handle_fbd_ue(mci, csrow, 0, | ||
| 1709 | 0 /* FIXME: should be channel here */, msg); | ||
| 1710 | else if (!pvt->is_registered) | ||
| 1711 | edac_mc_handle_fbd_ce(mci, csrow, | ||
| 1712 | 0 /* FIXME: should be channel here */, msg); | ||
| 1713 | |||
| 1714 | kfree(msg); | ||
| 1715 | } | ||
| 1716 | |||
| 1717 | /* | ||
| 1718 | * i7core_check_error Retrieve and process errors reported by the | ||
| 1719 | * hardware. Called by the Core module. | ||
| 1720 | */ | ||
| 1721 | static void i7core_check_error(struct mem_ctl_info *mci) | ||
| 1722 | { | ||
| 1723 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1724 | int i; | ||
| 1725 | unsigned count = 0; | ||
| 1726 | struct mce *m; | ||
| 1727 | |||
| 1728 | /* | ||
| 1729 | * MCE first step: Copy all mce errors into a temporary buffer | ||
| 1730 | * We use a double buffering here, to reduce the risk of | ||
| 1731 | * loosing an error. | ||
| 1732 | */ | ||
| 1733 | smp_rmb(); | ||
| 1734 | count = (pvt->mce_out + MCE_LOG_LEN - pvt->mce_in) | ||
| 1735 | % MCE_LOG_LEN; | ||
| 1736 | if (!count) | ||
| 1737 | goto check_ce_error; | ||
| 1738 | |||
| 1739 | m = pvt->mce_outentry; | ||
| 1740 | if (pvt->mce_in + count > MCE_LOG_LEN) { | ||
| 1741 | unsigned l = MCE_LOG_LEN - pvt->mce_in; | ||
| 1742 | |||
| 1743 | memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * l); | ||
| 1744 | smp_wmb(); | ||
| 1745 | pvt->mce_in = 0; | ||
| 1746 | count -= l; | ||
| 1747 | m += l; | ||
| 1748 | } | ||
| 1749 | memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * count); | ||
| 1750 | smp_wmb(); | ||
| 1751 | pvt->mce_in += count; | ||
| 1752 | |||
| 1753 | smp_rmb(); | ||
| 1754 | if (pvt->mce_overrun) { | ||
| 1755 | i7core_printk(KERN_ERR, "Lost %d memory errors\n", | ||
| 1756 | pvt->mce_overrun); | ||
| 1757 | smp_wmb(); | ||
| 1758 | pvt->mce_overrun = 0; | ||
| 1759 | } | ||
| 1760 | |||
| 1761 | /* | ||
| 1762 | * MCE second step: parse errors and display | ||
| 1763 | */ | ||
| 1764 | for (i = 0; i < count; i++) | ||
| 1765 | i7core_mce_output_error(mci, &pvt->mce_outentry[i]); | ||
| 1766 | |||
| 1767 | /* | ||
| 1768 | * Now, let's increment CE error counts | ||
| 1769 | */ | ||
| 1770 | check_ce_error: | ||
| 1771 | if (!pvt->is_registered) | ||
| 1772 | i7core_udimm_check_mc_ecc_err(mci); | ||
| 1773 | else | ||
| 1774 | i7core_rdimm_check_mc_ecc_err(mci); | ||
| 1775 | } | ||
| 1776 | |||
| 1777 | /* | ||
| 1778 | * i7core_mce_check_error Replicates mcelog routine to get errors | ||
| 1779 | * This routine simply queues mcelog errors, and | ||
| 1780 | * return. The error itself should be handled later | ||
| 1781 | * by i7core_check_error. | ||
| 1782 | * WARNING: As this routine should be called at NMI time, extra care should | ||
| 1783 | * be taken to avoid deadlocks, and to be as fast as possible. | ||
| 1784 | */ | ||
| 1785 | static int i7core_mce_check_error(void *priv, struct mce *mce) | ||
| 1786 | { | ||
| 1787 | struct mem_ctl_info *mci = priv; | ||
| 1788 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 1789 | |||
| 1790 | /* | ||
| 1791 | * Just let mcelog handle it if the error is | ||
| 1792 | * outside the memory controller | ||
| 1793 | */ | ||
| 1794 | if (((mce->status & 0xffff) >> 7) != 1) | ||
| 1795 | return 0; | ||
| 1796 | |||
| 1797 | /* Bank 8 registers are the only ones that we know how to handle */ | ||
| 1798 | if (mce->bank != 8) | ||
| 1799 | return 0; | ||
| 1800 | |||
| 1801 | #ifdef CONFIG_SMP | ||
| 1802 | /* Only handle if it is the right mc controller */ | ||
| 1803 | if (cpu_data(mce->cpu).phys_proc_id != pvt->i7core_dev->socket) | ||
| 1804 | return 0; | ||
| 1805 | #endif | ||
| 1806 | |||
| 1807 | smp_rmb(); | ||
| 1808 | if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) { | ||
| 1809 | smp_wmb(); | ||
| 1810 | pvt->mce_overrun++; | ||
| 1811 | return 0; | ||
| 1812 | } | ||
| 1813 | |||
| 1814 | /* Copy memory error at the ringbuffer */ | ||
| 1815 | memcpy(&pvt->mce_entry[pvt->mce_out], mce, sizeof(*mce)); | ||
| 1816 | smp_wmb(); | ||
| 1817 | pvt->mce_out = (pvt->mce_out + 1) % MCE_LOG_LEN; | ||
| 1818 | |||
| 1819 | /* Handle fatal errors immediately */ | ||
| 1820 | if (mce->mcgstatus & 1) | ||
| 1821 | i7core_check_error(mci); | ||
| 1822 | |||
| 1823 | /* Advice mcelog that the error were handled */ | ||
| 1824 | return 1; | ||
| 1825 | } | ||
| 1826 | |||
| 1827 | static int i7core_register_mci(struct i7core_dev *i7core_dev, | ||
| 1828 | int num_channels, int num_csrows) | ||
| 1829 | { | ||
| 1830 | struct mem_ctl_info *mci; | ||
| 1831 | struct i7core_pvt *pvt; | ||
| 1832 | int csrow = 0; | ||
| 1833 | int rc; | ||
| 1834 | |||
| 1835 | /* allocate a new MC control structure */ | ||
| 1836 | mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels, | ||
| 1837 | i7core_dev->socket); | ||
| 1838 | if (unlikely(!mci)) | ||
| 1839 | return -ENOMEM; | ||
| 1840 | |||
| 1841 | debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci); | ||
| 1842 | |||
| 1843 | /* record ptr to the generic device */ | ||
| 1844 | mci->dev = &i7core_dev->pdev[0]->dev; | ||
| 1845 | |||
| 1846 | pvt = mci->pvt_info; | ||
| 1847 | memset(pvt, 0, sizeof(*pvt)); | ||
| 1848 | |||
| 1849 | /* | ||
| 1850 | * FIXME: how to handle RDDR3 at MCI level? It is possible to have | ||
| 1851 | * Mixed RDDR3/UDDR3 with Nehalem, provided that they are on different | ||
| 1852 | * memory channels | ||
| 1853 | */ | ||
| 1854 | mci->mtype_cap = MEM_FLAG_DDR3; | ||
| 1855 | mci->edac_ctl_cap = EDAC_FLAG_NONE; | ||
| 1856 | mci->edac_cap = EDAC_FLAG_NONE; | ||
| 1857 | mci->mod_name = "i7core_edac.c"; | ||
| 1858 | mci->mod_ver = I7CORE_REVISION; | ||
| 1859 | mci->ctl_name = kasprintf(GFP_KERNEL, "i7 core #%d", | ||
| 1860 | i7core_dev->socket); | ||
| 1861 | mci->dev_name = pci_name(i7core_dev->pdev[0]); | ||
| 1862 | mci->ctl_page_to_phys = NULL; | ||
| 1863 | mci->mc_driver_sysfs_attributes = i7core_sysfs_attrs; | ||
| 1864 | /* Set the function pointer to an actual operation function */ | ||
| 1865 | mci->edac_check = i7core_check_error; | ||
| 1866 | |||
| 1867 | /* Store pci devices at mci for faster access */ | ||
| 1868 | rc = mci_bind_devs(mci, i7core_dev); | ||
| 1869 | if (unlikely(rc < 0)) | ||
| 1870 | goto fail; | ||
| 1871 | |||
| 1872 | /* Get dimm basic config */ | ||
| 1873 | get_dimm_config(mci, &csrow); | ||
| 1874 | |||
| 1875 | /* add this new MC control structure to EDAC's list of MCs */ | ||
| 1876 | if (unlikely(edac_mc_add_mc(mci))) { | ||
| 1877 | debugf0("MC: " __FILE__ | ||
| 1878 | ": %s(): failed edac_mc_add_mc()\n", __func__); | ||
| 1879 | /* FIXME: perhaps some code should go here that disables error | ||
| 1880 | * reporting if we just enabled it | ||
| 1881 | */ | ||
| 1882 | |||
| 1883 | rc = -EINVAL; | ||
| 1884 | goto fail; | ||
| 1885 | } | ||
| 1886 | |||
| 1887 | /* allocating generic PCI control info */ | ||
| 1888 | i7core_pci = edac_pci_create_generic_ctl(&i7core_dev->pdev[0]->dev, | ||
| 1889 | EDAC_MOD_STR); | ||
| 1890 | if (unlikely(!i7core_pci)) { | ||
| 1891 | printk(KERN_WARNING | ||
| 1892 | "%s(): Unable to create PCI control\n", | ||
| 1893 | __func__); | ||
| 1894 | printk(KERN_WARNING | ||
| 1895 | "%s(): PCI error report via EDAC not setup\n", | ||
| 1896 | __func__); | ||
| 1897 | } | ||
| 1898 | |||
| 1899 | /* Default error mask is any memory */ | ||
| 1900 | pvt->inject.channel = 0; | ||
| 1901 | pvt->inject.dimm = -1; | ||
| 1902 | pvt->inject.rank = -1; | ||
| 1903 | pvt->inject.bank = -1; | ||
| 1904 | pvt->inject.page = -1; | ||
| 1905 | pvt->inject.col = -1; | ||
| 1906 | |||
| 1907 | /* Registers on edac_mce in order to receive memory errors */ | ||
| 1908 | pvt->edac_mce.priv = mci; | ||
| 1909 | pvt->edac_mce.check_error = i7core_mce_check_error; | ||
| 1910 | |||
| 1911 | rc = edac_mce_register(&pvt->edac_mce); | ||
| 1912 | if (unlikely(rc < 0)) { | ||
| 1913 | debugf0("MC: " __FILE__ | ||
| 1914 | ": %s(): failed edac_mce_register()\n", __func__); | ||
| 1915 | } | ||
| 1916 | |||
| 1917 | fail: | ||
| 1918 | if (rc < 0) | ||
| 1919 | edac_mc_free(mci); | ||
| 1920 | return rc; | ||
| 1921 | } | ||
| 1922 | |||
| 1923 | /* | ||
| 1924 | * i7core_probe Probe for ONE instance of device to see if it is | ||
| 1925 | * present. | ||
| 1926 | * return: | ||
| 1927 | * 0 for FOUND a device | ||
| 1928 | * < 0 for error code | ||
| 1929 | */ | ||
| 1930 | static int __devinit i7core_probe(struct pci_dev *pdev, | ||
| 1931 | const struct pci_device_id *id) | ||
| 1932 | { | ||
| 1933 | int dev_idx = id->driver_data; | ||
| 1934 | int rc; | ||
| 1935 | struct i7core_dev *i7core_dev; | ||
| 1936 | |||
| 1937 | /* | ||
| 1938 | * All memory controllers are allocated at the first pass. | ||
| 1939 | */ | ||
| 1940 | if (unlikely(dev_idx >= 1)) | ||
| 1941 | return -EINVAL; | ||
| 1942 | |||
| 1943 | /* get the pci devices we want to reserve for our use */ | ||
| 1944 | mutex_lock(&i7core_edac_lock); | ||
| 1945 | |||
| 1946 | rc = i7core_get_devices(pci_dev_table); | ||
| 1947 | if (unlikely(rc < 0)) | ||
| 1948 | goto fail0; | ||
| 1949 | |||
| 1950 | list_for_each_entry(i7core_dev, &i7core_edac_list, list) { | ||
| 1951 | int channels; | ||
| 1952 | int csrows; | ||
| 1953 | |||
| 1954 | /* Check the number of active and not disabled channels */ | ||
| 1955 | rc = i7core_get_active_channels(i7core_dev->socket, | ||
| 1956 | &channels, &csrows); | ||
| 1957 | if (unlikely(rc < 0)) | ||
| 1958 | goto fail1; | ||
| 1959 | |||
| 1960 | rc = i7core_register_mci(i7core_dev, channels, csrows); | ||
| 1961 | if (unlikely(rc < 0)) | ||
| 1962 | goto fail1; | ||
| 1963 | } | ||
| 1964 | |||
| 1965 | i7core_printk(KERN_INFO, "Driver loaded.\n"); | ||
| 1966 | |||
| 1967 | mutex_unlock(&i7core_edac_lock); | ||
| 1968 | return 0; | ||
| 1969 | |||
| 1970 | fail1: | ||
| 1971 | i7core_put_all_devices(); | ||
| 1972 | fail0: | ||
| 1973 | mutex_unlock(&i7core_edac_lock); | ||
| 1974 | return rc; | ||
| 1975 | } | ||
| 1976 | |||
| 1977 | /* | ||
| 1978 | * i7core_remove destructor for one instance of device | ||
| 1979 | * | ||
| 1980 | */ | ||
| 1981 | static void __devexit i7core_remove(struct pci_dev *pdev) | ||
| 1982 | { | ||
| 1983 | struct mem_ctl_info *mci; | ||
| 1984 | struct i7core_dev *i7core_dev, *tmp; | ||
| 1985 | |||
| 1986 | debugf0(__FILE__ ": %s()\n", __func__); | ||
| 1987 | |||
| 1988 | if (i7core_pci) | ||
| 1989 | edac_pci_release_generic_ctl(i7core_pci); | ||
| 1990 | |||
| 1991 | /* | ||
| 1992 | * we have a trouble here: pdev value for removal will be wrong, since | ||
| 1993 | * it will point to the X58 register used to detect that the machine | ||
| 1994 | * is a Nehalem or upper design. However, due to the way several PCI | ||
| 1995 | * devices are grouped together to provide MC functionality, we need | ||
| 1996 | * to use a different method for releasing the devices | ||
| 1997 | */ | ||
| 1998 | |||
| 1999 | mutex_lock(&i7core_edac_lock); | ||
| 2000 | list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list) { | ||
| 2001 | mci = edac_mc_del_mc(&i7core_dev->pdev[0]->dev); | ||
| 2002 | if (mci) { | ||
| 2003 | struct i7core_pvt *pvt = mci->pvt_info; | ||
| 2004 | |||
| 2005 | i7core_dev = pvt->i7core_dev; | ||
| 2006 | edac_mce_unregister(&pvt->edac_mce); | ||
| 2007 | kfree(mci->ctl_name); | ||
| 2008 | edac_mc_free(mci); | ||
| 2009 | i7core_put_devices(i7core_dev); | ||
| 2010 | } else { | ||
| 2011 | i7core_printk(KERN_ERR, | ||
| 2012 | "Couldn't find mci for socket %d\n", | ||
| 2013 | i7core_dev->socket); | ||
| 2014 | } | ||
| 2015 | } | ||
| 2016 | mutex_unlock(&i7core_edac_lock); | ||
| 2017 | } | ||
| 2018 | |||
| 2019 | MODULE_DEVICE_TABLE(pci, i7core_pci_tbl); | ||
| 2020 | |||
| 2021 | /* | ||
| 2022 | * i7core_driver pci_driver structure for this module | ||
| 2023 | * | ||
| 2024 | */ | ||
| 2025 | static struct pci_driver i7core_driver = { | ||
| 2026 | .name = "i7core_edac", | ||
| 2027 | .probe = i7core_probe, | ||
| 2028 | .remove = __devexit_p(i7core_remove), | ||
| 2029 | .id_table = i7core_pci_tbl, | ||
| 2030 | }; | ||
| 2031 | |||
| 2032 | /* | ||
| 2033 | * i7core_init Module entry function | ||
| 2034 | * Try to initialize this module for its devices | ||
| 2035 | */ | ||
| 2036 | static int __init i7core_init(void) | ||
| 2037 | { | ||
| 2038 | int pci_rc; | ||
| 2039 | |||
| 2040 | debugf2("MC: " __FILE__ ": %s()\n", __func__); | ||
| 2041 | |||
| 2042 | /* Ensure that the OPSTATE is set correctly for POLL or NMI */ | ||
| 2043 | opstate_init(); | ||
| 2044 | |||
| 2045 | i7core_xeon_pci_fixup(pci_dev_table); | ||
| 2046 | |||
| 2047 | pci_rc = pci_register_driver(&i7core_driver); | ||
| 2048 | |||
| 2049 | if (pci_rc >= 0) | ||
| 2050 | return 0; | ||
| 2051 | |||
| 2052 | i7core_printk(KERN_ERR, "Failed to register device with error %d.\n", | ||
| 2053 | pci_rc); | ||
| 2054 | |||
| 2055 | return pci_rc; | ||
| 2056 | } | ||
| 2057 | |||
| 2058 | /* | ||
| 2059 | * i7core_exit() Module exit function | ||
| 2060 | * Unregister the driver | ||
| 2061 | */ | ||
| 2062 | static void __exit i7core_exit(void) | ||
| 2063 | { | ||
| 2064 | debugf2("MC: " __FILE__ ": %s()\n", __func__); | ||
| 2065 | pci_unregister_driver(&i7core_driver); | ||
| 2066 | } | ||
| 2067 | |||
| 2068 | module_init(i7core_init); | ||
| 2069 | module_exit(i7core_exit); | ||
| 2070 | |||
| 2071 | MODULE_LICENSE("GPL"); | ||
| 2072 | MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>"); | ||
| 2073 | MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)"); | ||
| 2074 | MODULE_DESCRIPTION("MC Driver for Intel i7 Core memory controllers - " | ||
| 2075 | I7CORE_REVISION); | ||
| 2076 | |||
| 2077 | module_param(edac_op_state, int, 0444); | ||
| 2078 | MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI"); | ||
diff --git a/include/linux/edac_mce.h b/include/linux/edac_mce.h new file mode 100644 index 000000000000..f974fc035363 --- /dev/null +++ b/include/linux/edac_mce.h | |||
| @@ -0,0 +1,31 @@ | |||
| 1 | /* Provides edac interface to mcelog events | ||
| 2 | * | ||
| 3 | * This file may be distributed under the terms of the | ||
| 4 | * GNU General Public License version 2. | ||
| 5 | * | ||
| 6 | * Copyright (c) 2009 by: | ||
| 7 | * Mauro Carvalho Chehab <mchehab@redhat.com> | ||
| 8 | * | ||
| 9 | * Red Hat Inc. http://www.redhat.com | ||
| 10 | */ | ||
| 11 | |||
| 12 | #if defined(CONFIG_EDAC_MCE) || \ | ||
| 13 | (defined(CONFIG_EDAC_MCE_MODULE) && defined(MODULE)) | ||
| 14 | |||
| 15 | #include <asm/mce.h> | ||
| 16 | #include <linux/list.h> | ||
| 17 | |||
| 18 | struct edac_mce { | ||
| 19 | struct list_head list; | ||
| 20 | |||
| 21 | void *priv; | ||
| 22 | int (*check_error)(void *priv, struct mce *mce); | ||
| 23 | }; | ||
| 24 | |||
| 25 | int edac_mce_register(struct edac_mce *edac_mce); | ||
| 26 | void edac_mce_unregister(struct edac_mce *edac_mce); | ||
| 27 | int edac_mce_parse(struct mce *mce); | ||
| 28 | |||
| 29 | #else | ||
| 30 | #define edac_mce_parse(mce) (0) | ||
| 31 | #endif | ||
diff --git a/include/linux/pci.h b/include/linux/pci.h index 6a471aba3b07..7cb00845f150 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h | |||
| @@ -632,6 +632,7 @@ void pci_fixup_cardbus(struct pci_bus *); | |||
| 632 | 632 | ||
| 633 | /* Generic PCI functions used internally */ | 633 | /* Generic PCI functions used internally */ |
| 634 | 634 | ||
| 635 | void pcibios_scan_specific_bus(int busn); | ||
| 635 | extern struct pci_bus *pci_find_bus(int domain, int busnr); | 636 | extern struct pci_bus *pci_find_bus(int domain, int busnr); |
| 636 | void pci_bus_add_devices(const struct pci_bus *bus); | 637 | void pci_bus_add_devices(const struct pci_bus *bus); |
| 637 | struct pci_bus *pci_scan_bus_parented(struct device *parent, int bus, | 638 | struct pci_bus *pci_scan_bus_parented(struct device *parent, int bus, |
diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h index ae66851870be..f149dd10908b 100644 --- a/include/linux/pci_ids.h +++ b/include/linux/pci_ids.h | |||
| @@ -2532,11 +2532,63 @@ | |||
| 2532 | #define PCI_DEVICE_ID_INTEL_ICH9_6 0x2930 | 2532 | #define PCI_DEVICE_ID_INTEL_ICH9_6 0x2930 |
| 2533 | #define PCI_DEVICE_ID_INTEL_ICH9_7 0x2916 | 2533 | #define PCI_DEVICE_ID_INTEL_ICH9_7 0x2916 |
| 2534 | #define PCI_DEVICE_ID_INTEL_ICH9_8 0x2918 | 2534 | #define PCI_DEVICE_ID_INTEL_ICH9_8 0x2918 |
| 2535 | #define PCI_DEVICE_ID_INTEL_I7_MCR 0x2c18 | ||
| 2536 | #define PCI_DEVICE_ID_INTEL_I7_MC_TAD 0x2c19 | ||
| 2537 | #define PCI_DEVICE_ID_INTEL_I7_MC_RAS 0x2c1a | ||
| 2538 | #define PCI_DEVICE_ID_INTEL_I7_MC_TEST 0x2c1c | ||
| 2539 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH0_CTRL 0x2c20 | ||
| 2540 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH0_ADDR 0x2c21 | ||
| 2541 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH0_RANK 0x2c22 | ||
| 2542 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH0_TC 0x2c23 | ||
| 2543 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH1_CTRL 0x2c28 | ||
| 2544 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH1_ADDR 0x2c29 | ||
| 2545 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH1_RANK 0x2c2a | ||
| 2546 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH1_TC 0x2c2b | ||
| 2547 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH2_CTRL 0x2c30 | ||
| 2548 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH2_ADDR 0x2c31 | ||
| 2549 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH2_RANK 0x2c32 | ||
| 2550 | #define PCI_DEVICE_ID_INTEL_I7_MC_CH2_TC 0x2c33 | ||
| 2551 | #define PCI_DEVICE_ID_INTEL_I7_NONCORE 0x2c41 | ||
| 2552 | #define PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT 0x2c40 | ||
| 2553 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE 0x2c50 | ||
| 2554 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT 0x2c51 | ||
| 2555 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2 0x2c70 | ||
| 2556 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_SAD 0x2c81 | ||
| 2557 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_LINK0 0x2c90 | ||
| 2558 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_PHY0 0x2c91 | ||
| 2559 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR 0x2c98 | ||
| 2560 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD 0x2c99 | ||
| 2561 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST 0x2c9C | ||
| 2562 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL 0x2ca0 | ||
| 2563 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR 0x2ca1 | ||
| 2564 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK 0x2ca2 | ||
| 2565 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC 0x2ca3 | ||
| 2566 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL 0x2ca8 | ||
| 2567 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR 0x2ca9 | ||
| 2568 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK 0x2caa | ||
| 2569 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC 0x2cab | ||
| 2570 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR_REV2 0x2d98 | ||
| 2571 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD_REV2 0x2d99 | ||
| 2572 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_RAS_REV2 0x2d9a | ||
| 2573 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST_REV2 0x2d9c | ||
| 2574 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL_REV2 0x2da0 | ||
| 2575 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR_REV2 0x2da1 | ||
| 2576 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK_REV2 0x2da2 | ||
| 2577 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC_REV2 0x2da3 | ||
| 2578 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL_REV2 0x2da8 | ||
| 2579 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR_REV2 0x2da9 | ||
| 2580 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK_REV2 0x2daa | ||
| 2581 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC_REV2 0x2dab | ||
| 2582 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_CTRL_REV2 0x2db0 | ||
| 2583 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_ADDR_REV2 0x2db1 | ||
| 2584 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_RANK_REV2 0x2db2 | ||
| 2585 | #define PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_TC_REV2 0x2db3 | ||
| 2535 | #define PCI_DEVICE_ID_INTEL_82855PM_HB 0x3340 | 2586 | #define PCI_DEVICE_ID_INTEL_82855PM_HB 0x3340 |
| 2536 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG4 0x3429 | 2587 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG4 0x3429 |
| 2537 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG5 0x342a | 2588 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG5 0x342a |
| 2538 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG6 0x342b | 2589 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG6 0x342b |
| 2539 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG7 0x342c | 2590 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG7 0x342c |
| 2591 | #define PCI_DEVICE_ID_INTEL_X58_HUB_MGMT 0x342e | ||
| 2540 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG0 0x3430 | 2592 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG0 0x3430 |
| 2541 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG1 0x3431 | 2593 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG1 0x3431 |
| 2542 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG2 0x3432 | 2594 | #define PCI_DEVICE_ID_INTEL_IOAT_TBG2 0x3432 |
