linux/drivers/edac/i3200_edac.c

546 lines
13 KiB
C
Raw Normal View History

/*
* Intel 3200/3210 Memory Controller kernel module
* Copyright (C) 2008-2009 Akamai Technologies, Inc.
* Portions by Hitoshi Mitake <h.mitake@gmail.com>.
*
* This file may be distributed under the terms of the
* GNU General Public License.
*/
#include <linux/module.h>
#include <linux/init.h>
#include <linux/pci.h>
#include <linux/pci_ids.h>
#include <linux/edac.h>
#include <linux/io.h>
#include "edac_module.h"
#include <linux/io-64-nonatomic-lo-hi.h>
asm-generic: architecture independent readq/writeq for 32bit environment This provides unified readq()/writeq() helper functions for 32-bit drivers. For some cases, readq/writeq without atomicity is harmful, and order of io access has to be specified explicitly. So in this patch, new two header files which contain non-atomic readq/writeq are added. - <asm-generic/io-64-nonatomic-lo-hi.h> provides non-atomic readq/ writeq with the order of lower address -> higher address - <asm-generic/io-64-nonatomic-hi-lo.h> provides non-atomic readq/ writeq with reversed order This allows us to remove some readq()s that were added drivers when the default non-atomic ones were removed in commit dbee8a0affd5 ("x86: remove 32-bit versions of readq()/writeq()") The drivers which need readq/writeq but can do with the non-atomic ones must add the line: #include <asm-generic/io-64-nonatomic-lo-hi.h> /* or hi-lo.h */ But this will be nop in 64-bit environments, and no other #ifdefs are required. So I believe that this patch can solve the problem of 1. driver-specific readq/writeq 2. atomicity and order of io access This patch is tested with building allyesconfig and allmodconfig as ARCH=x86 and ARCH=i386 on top of tip/master. Cc: Kashyap Desai <Kashyap.Desai@lsi.com> Cc: Len Brown <lenb@kernel.org> Cc: Ravi Anand <ravi.anand@qlogic.com> Cc: Vikas Chaudhary <vikas.chaudhary@qlogic.com> Cc: Matthew Garrett <mjg@redhat.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Roland Dreier <roland@purestorage.com> Cc: James Bottomley <jbottomley@parallels.com> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Hitoshi Mitake <h.mitake@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-07 02:45:33 +00:00
#define EDAC_MOD_STR "i3200_edac"
#define PCI_DEVICE_ID_INTEL_3200_HB 0x29f0
#define I3200_DIMMS 4
#define I3200_RANKS 8
#define I3200_RANKS_PER_CHANNEL 4
#define I3200_CHANNELS 2
/* Intel 3200 register addresses - device 0 function 0 - DRAM Controller */
#define I3200_MCHBAR_LOW 0x48 /* MCH Memory Mapped Register BAR */
#define I3200_MCHBAR_HIGH 0x4c
#define I3200_MCHBAR_MASK 0xfffffc000ULL /* bits 35:14 */
#define I3200_MMR_WINDOW_SIZE 16384
#define I3200_TOM 0xa0 /* Top of Memory (16b)
*
* 15:10 reserved
* 9:0 total populated physical memory
*/
#define I3200_TOM_MASK 0x3ff /* bits 9:0 */
#define I3200_TOM_SHIFT 26 /* 64MiB grain */
#define I3200_ERRSTS 0xc8 /* Error Status Register (16b)
*
* 15 reserved
* 14 Isochronous TBWRR Run Behind FIFO Full
* (ITCV)
* 13 Isochronous TBWRR Run Behind FIFO Put
* (ITSTV)
* 12 reserved
* 11 MCH Thermal Sensor Event
* for SMI/SCI/SERR (GTSE)
* 10 reserved
* 9 LOCK to non-DRAM Memory Flag (LCKF)
* 8 reserved
* 7 DRAM Throttle Flag (DTF)
* 6:2 reserved
* 1 Multi-bit DRAM ECC Error Flag (DMERR)
* 0 Single-bit DRAM ECC Error Flag (DSERR)
*/
#define I3200_ERRSTS_UE 0x0002
#define I3200_ERRSTS_CE 0x0001
#define I3200_ERRSTS_BITS (I3200_ERRSTS_UE | I3200_ERRSTS_CE)
/* Intel MMIO register space - device 0 function 0 - MMR space */
#define I3200_C0DRB 0x200 /* Channel 0 DRAM Rank Boundary (16b x 4)
*
* 15:10 reserved
* 9:0 Channel 0 DRAM Rank Boundary Address
*/
#define I3200_C1DRB 0x600 /* Channel 1 DRAM Rank Boundary (16b x 4) */
#define I3200_DRB_MASK 0x3ff /* bits 9:0 */
#define I3200_DRB_SHIFT 26 /* 64MiB grain */
#define I3200_C0ECCERRLOG 0x280 /* Channel 0 ECC Error Log (64b)
*
* 63:48 Error Column Address (ERRCOL)
* 47:32 Error Row Address (ERRROW)
* 31:29 Error Bank Address (ERRBANK)
* 28:27 Error Rank Address (ERRRANK)
* 26:24 reserved
* 23:16 Error Syndrome (ERRSYND)
* 15: 2 reserved
* 1 Multiple Bit Error Status (MERRSTS)
* 0 Correctable Error Status (CERRSTS)
*/
#define I3200_C1ECCERRLOG 0x680 /* Chan 1 ECC Error Log (64b) */
#define I3200_ECCERRLOG_CE 0x1
#define I3200_ECCERRLOG_UE 0x2
#define I3200_ECCERRLOG_RANK_BITS 0x18000000
#define I3200_ECCERRLOG_RANK_SHIFT 27
#define I3200_ECCERRLOG_SYNDROME_BITS 0xff0000
#define I3200_ECCERRLOG_SYNDROME_SHIFT 16
#define I3200_CAPID0 0xe0 /* P.95 of spec for details */
struct i3200_priv {
void __iomem *window;
};
static int nr_channels;
static int how_many_channels(struct pci_dev *pdev)
{
int n_channels;
unsigned char capid0_8b; /* 8th byte of CAPID0 */
pci_read_config_byte(pdev, I3200_CAPID0 + 8, &capid0_8b);
if (capid0_8b & 0x20) { /* check DCD: Dual Channel Disable */
edac_dbg(0, "In single channel mode\n");
n_channels = 1;
} else {
edac_dbg(0, "In dual channel mode\n");
n_channels = 2;
}
if (capid0_8b & 0x10) /* check if both channels are filled */
edac_dbg(0, "2 DIMMS per channel disabled\n");
else
edac_dbg(0, "2 DIMMS per channel enabled\n");
return n_channels;
}
static unsigned long eccerrlog_syndrome(u64 log)
{
return (log & I3200_ECCERRLOG_SYNDROME_BITS) >>
I3200_ECCERRLOG_SYNDROME_SHIFT;
}
static int eccerrlog_row(int channel, u64 log)
{
u64 rank = ((log & I3200_ECCERRLOG_RANK_BITS) >>
I3200_ECCERRLOG_RANK_SHIFT);
return rank | (channel * I3200_RANKS_PER_CHANNEL);
}
enum i3200_chips {
I3200 = 0,
};
struct i3200_dev_info {
const char *ctl_name;
};
struct i3200_error_info {
u16 errsts;
u16 errsts2;
u64 eccerrlog[I3200_CHANNELS];
};
static const struct i3200_dev_info i3200_devs[] = {
[I3200] = {
.ctl_name = "i3200"
},
};
static struct pci_dev *mci_pdev;
static int i3200_registered = 1;
static void i3200_clear_error_info(struct mem_ctl_info *mci)
{
struct pci_dev *pdev;
pdev = to_pci_dev(mci->pdev);
/*
* Clear any error bits.
* (Yes, we really clear bits by writing 1 to them.)
*/
pci_write_bits16(pdev, I3200_ERRSTS, I3200_ERRSTS_BITS,
I3200_ERRSTS_BITS);
}
static void i3200_get_and_clear_error_info(struct mem_ctl_info *mci,
struct i3200_error_info *info)
{
struct pci_dev *pdev;
struct i3200_priv *priv = mci->pvt_info;
void __iomem *window = priv->window;
pdev = to_pci_dev(mci->pdev);
/*
* This is a mess because there is no atomic way to read all the
* registers at once and the registers can transition from CE being
* overwritten by UE.
*/
pci_read_config_word(pdev, I3200_ERRSTS, &info->errsts);
if (!(info->errsts & I3200_ERRSTS_BITS))
return;
info->eccerrlog[0] = readq(window + I3200_C0ECCERRLOG);
if (nr_channels == 2)
info->eccerrlog[1] = readq(window + I3200_C1ECCERRLOG);
pci_read_config_word(pdev, I3200_ERRSTS, &info->errsts2);
/*
* If the error is the same for both reads then the first set
* of reads is valid. If there is a change then there is a CE
* with no info and the second set of reads is valid and
* should be UE info.
*/
if ((info->errsts ^ info->errsts2) & I3200_ERRSTS_BITS) {
info->eccerrlog[0] = readq(window + I3200_C0ECCERRLOG);
if (nr_channels == 2)
info->eccerrlog[1] = readq(window + I3200_C1ECCERRLOG);
}
i3200_clear_error_info(mci);
}
static void i3200_process_error_info(struct mem_ctl_info *mci,
struct i3200_error_info *info)
{
int channel;
u64 log;
if (!(info->errsts & I3200_ERRSTS_BITS))
return;
if ((info->errsts ^ info->errsts2) & I3200_ERRSTS_BITS) {
edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
-1, -1, -1, "UE overwrote CE", "");
info->errsts = info->errsts2;
}
for (channel = 0; channel < nr_channels; channel++) {
log = info->eccerrlog[channel];
if (log & I3200_ECCERRLOG_UE) {
edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
0, 0, 0,
eccerrlog_row(channel, log),
-1, -1,
"i3000 UE", "");
} else if (log & I3200_ECCERRLOG_CE) {
edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
0, 0, eccerrlog_syndrome(log),
eccerrlog_row(channel, log),
-1, -1,
"i3000 CE", "");
}
}
}
static void i3200_check(struct mem_ctl_info *mci)
{
struct i3200_error_info info;
i3200_get_and_clear_error_info(mci, &info);
i3200_process_error_info(mci, &info);
}
static void __iomem *i3200_map_mchbar(struct pci_dev *pdev)
{
union {
u64 mchbar;
struct {
u32 mchbar_low;
u32 mchbar_high;
};
} u;
void __iomem *window;
pci_read_config_dword(pdev, I3200_MCHBAR_LOW, &u.mchbar_low);
pci_read_config_dword(pdev, I3200_MCHBAR_HIGH, &u.mchbar_high);
u.mchbar &= I3200_MCHBAR_MASK;
if (u.mchbar != (resource_size_t)u.mchbar) {
printk(KERN_ERR
"i3200: mmio space beyond accessible range (0x%llx)\n",
(unsigned long long)u.mchbar);
return NULL;
}
window = ioremap(u.mchbar, I3200_MMR_WINDOW_SIZE);
if (!window)
printk(KERN_ERR "i3200: cannot map mmio space at 0x%llx\n",
(unsigned long long)u.mchbar);
return window;
}
static void i3200_get_drbs(void __iomem *window,
u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL])
{
int i;
for (i = 0; i < I3200_RANKS_PER_CHANNEL; i++) {
drbs[0][i] = readw(window + I3200_C0DRB + 2*i) & I3200_DRB_MASK;
drbs[1][i] = readw(window + I3200_C1DRB + 2*i) & I3200_DRB_MASK;
edac_dbg(0, "drb[0][%d] = %d, drb[1][%d] = %d\n", i, drbs[0][i], i, drbs[1][i]);
}
}
static bool i3200_is_stacked(struct pci_dev *pdev,
u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL])
{
u16 tom;
pci_read_config_word(pdev, I3200_TOM, &tom);
tom &= I3200_TOM_MASK;
return drbs[I3200_CHANNELS - 1][I3200_RANKS_PER_CHANNEL - 1] == tom;
}
static unsigned long drb_to_nr_pages(
u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL], bool stacked,
int channel, int rank)
{
int n;
n = drbs[channel][rank];
i3200_edac: Fix the logic that detects filled memories After running a series of tests on an HP DL320, filled with different memory sizes, it was noticed that, when filled with just one DIMM on such hardware, the driver wrongly detects twice the memory, and thinks that both channels 0 and 1 are filled. It seems to be partially caused by the BIOS and partially by the driver. The i3200_edac current logic would be working fine if the BIOS were disabling the unused second channel when just one DIMM is connected, in order to do power-saving, as recommended on this chipset's datasheet. However, the BIOS on this particular machine doesn't do it: [ 16.741421] EDAC DEBUG: how_many_channels: In dual channel mode [ 16.741424] EDAC DEBUG: how_many_channels: 2 DIMMS per channel enabled So, the driver were assuming that 2 channels are enabled (well, they are, but the second is unused). Combined with that, I found two issues at the logic that creates the EDAC data, that were failing when the two channels are not equally filled (AFAICT, that happens only when just 1 DIMM is plugged). The first one is that a 0 at DRB means that nothing is filled. The driver's logic, however, do some calculation with that. The second one is that the logic that fills the DIMM data currently assumes that both channels are equally filled. I tested the system already with the current configuration and my patch and it is now working fine. So, for a 2R single DIMM 2Gb memory at dimm slot 01 (channel 0), it is now displaying: [ 16.741406] EDAC DEBUG: i3200_get_drbs: drb[0][0] = 16, drb[1][0] = 0 [ 16.741410] EDAC DEBUG: i3200_get_drbs: drb[0][1] = 32, drb[1][1] = 0 [ 16.741413] EDAC DEBUG: i3200_get_drbs: drb[0][2] = 32, drb[1][2] = 0 [ 16.741416] EDAC DEBUG: i3200_get_drbs: drb[0][3] = 32, drb[1][3] = 0 ... [ 16.741896] EDAC DEBUG: i3200_probe1: csrow 0, channel 0, size = 1024 Mb [ 16.741899] EDAC DEBUG: i3200_probe1: csrow 1, channel 0, size = 1024 Mb and the corresponding sysfs nodes are now properly filled. Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2013-01-10 16:31:47 +00:00
if (!n)
return 0;
if (rank > 0)
n -= drbs[channel][rank - 1];
if (stacked && (channel == 1) &&
drbs[channel][rank] == drbs[channel][I3200_RANKS_PER_CHANNEL - 1])
n -= drbs[0][I3200_RANKS_PER_CHANNEL - 1];
n <<= (I3200_DRB_SHIFT - PAGE_SHIFT);
return n;
}
static int i3200_probe1(struct pci_dev *pdev, int dev_idx)
{
int rc;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-27 21:38:08 +00:00
int i, j;
struct mem_ctl_info *mci = NULL;
struct edac_mc_layer layers[2];
u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL];
bool stacked;
void __iomem *window;
struct i3200_priv *priv;
edac_dbg(0, "MC:\n");
window = i3200_map_mchbar(pdev);
if (!window)
return -ENODEV;
i3200_get_drbs(window, drbs);
nr_channels = how_many_channels(pdev);
layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
layers[0].size = I3200_DIMMS;
layers[0].is_virt_csrow = true;
layers[1].type = EDAC_MC_LAYER_CHANNEL;
layers[1].size = nr_channels;
layers[1].is_virt_csrow = false;
mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
sizeof(struct i3200_priv));
if (!mci)
return -ENOMEM;
edac_dbg(3, "MC: init mci\n");
mci->pdev = &pdev->dev;
mci->mtype_cap = MEM_FLAG_DDR2;
mci->edac_ctl_cap = EDAC_FLAG_SECDED;
mci->edac_cap = EDAC_FLAG_SECDED;
mci->mod_name = EDAC_MOD_STR;
mci->ctl_name = i3200_devs[dev_idx].ctl_name;
mci->dev_name = pci_name(pdev);
mci->edac_check = i3200_check;
mci->ctl_page_to_phys = NULL;
priv = mci->pvt_info;
priv->window = window;
stacked = i3200_is_stacked(pdev, drbs);
/*
* The dram rank boundary (DRB) reg values are boundary addresses
* for each DRAM rank with a granularity of 64MB. DRB regs are
* cumulative; the last one will contain the total memory
* contained in all ranks.
*/
i3200_edac: Fix the logic that detects filled memories After running a series of tests on an HP DL320, filled with different memory sizes, it was noticed that, when filled with just one DIMM on such hardware, the driver wrongly detects twice the memory, and thinks that both channels 0 and 1 are filled. It seems to be partially caused by the BIOS and partially by the driver. The i3200_edac current logic would be working fine if the BIOS were disabling the unused second channel when just one DIMM is connected, in order to do power-saving, as recommended on this chipset's datasheet. However, the BIOS on this particular machine doesn't do it: [ 16.741421] EDAC DEBUG: how_many_channels: In dual channel mode [ 16.741424] EDAC DEBUG: how_many_channels: 2 DIMMS per channel enabled So, the driver were assuming that 2 channels are enabled (well, they are, but the second is unused). Combined with that, I found two issues at the logic that creates the EDAC data, that were failing when the two channels are not equally filled (AFAICT, that happens only when just 1 DIMM is plugged). The first one is that a 0 at DRB means that nothing is filled. The driver's logic, however, do some calculation with that. The second one is that the logic that fills the DIMM data currently assumes that both channels are equally filled. I tested the system already with the current configuration and my patch and it is now working fine. So, for a 2R single DIMM 2Gb memory at dimm slot 01 (channel 0), it is now displaying: [ 16.741406] EDAC DEBUG: i3200_get_drbs: drb[0][0] = 16, drb[1][0] = 0 [ 16.741410] EDAC DEBUG: i3200_get_drbs: drb[0][1] = 32, drb[1][1] = 0 [ 16.741413] EDAC DEBUG: i3200_get_drbs: drb[0][2] = 32, drb[1][2] = 0 [ 16.741416] EDAC DEBUG: i3200_get_drbs: drb[0][3] = 32, drb[1][3] = 0 ... [ 16.741896] EDAC DEBUG: i3200_probe1: csrow 0, channel 0, size = 1024 Mb [ 16.741899] EDAC DEBUG: i3200_probe1: csrow 1, channel 0, size = 1024 Mb and the corresponding sysfs nodes are now properly filled. Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2013-01-10 16:31:47 +00:00
for (i = 0; i < I3200_DIMMS; i++) {
unsigned long nr_pages;
i3200_edac: Fix the logic that detects filled memories After running a series of tests on an HP DL320, filled with different memory sizes, it was noticed that, when filled with just one DIMM on such hardware, the driver wrongly detects twice the memory, and thinks that both channels 0 and 1 are filled. It seems to be partially caused by the BIOS and partially by the driver. The i3200_edac current logic would be working fine if the BIOS were disabling the unused second channel when just one DIMM is connected, in order to do power-saving, as recommended on this chipset's datasheet. However, the BIOS on this particular machine doesn't do it: [ 16.741421] EDAC DEBUG: how_many_channels: In dual channel mode [ 16.741424] EDAC DEBUG: how_many_channels: 2 DIMMS per channel enabled So, the driver were assuming that 2 channels are enabled (well, they are, but the second is unused). Combined with that, I found two issues at the logic that creates the EDAC data, that were failing when the two channels are not equally filled (AFAICT, that happens only when just 1 DIMM is plugged). The first one is that a 0 at DRB means that nothing is filled. The driver's logic, however, do some calculation with that. The second one is that the logic that fills the DIMM data currently assumes that both channels are equally filled. I tested the system already with the current configuration and my patch and it is now working fine. So, for a 2R single DIMM 2Gb memory at dimm slot 01 (channel 0), it is now displaying: [ 16.741406] EDAC DEBUG: i3200_get_drbs: drb[0][0] = 16, drb[1][0] = 0 [ 16.741410] EDAC DEBUG: i3200_get_drbs: drb[0][1] = 32, drb[1][1] = 0 [ 16.741413] EDAC DEBUG: i3200_get_drbs: drb[0][2] = 32, drb[1][2] = 0 [ 16.741416] EDAC DEBUG: i3200_get_drbs: drb[0][3] = 32, drb[1][3] = 0 ... [ 16.741896] EDAC DEBUG: i3200_probe1: csrow 0, channel 0, size = 1024 Mb [ 16.741899] EDAC DEBUG: i3200_probe1: csrow 1, channel 0, size = 1024 Mb and the corresponding sysfs nodes are now properly filled. Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2013-01-10 16:31:47 +00:00
for (j = 0; j < nr_channels; j++) {
struct dimm_info *dimm = edac_get_dimm(mci, i, j, 0);
i3200_edac: Fix the logic that detects filled memories After running a series of tests on an HP DL320, filled with different memory sizes, it was noticed that, when filled with just one DIMM on such hardware, the driver wrongly detects twice the memory, and thinks that both channels 0 and 1 are filled. It seems to be partially caused by the BIOS and partially by the driver. The i3200_edac current logic would be working fine if the BIOS were disabling the unused second channel when just one DIMM is connected, in order to do power-saving, as recommended on this chipset's datasheet. However, the BIOS on this particular machine doesn't do it: [ 16.741421] EDAC DEBUG: how_many_channels: In dual channel mode [ 16.741424] EDAC DEBUG: how_many_channels: 2 DIMMS per channel enabled So, the driver were assuming that 2 channels are enabled (well, they are, but the second is unused). Combined with that, I found two issues at the logic that creates the EDAC data, that were failing when the two channels are not equally filled (AFAICT, that happens only when just 1 DIMM is plugged). The first one is that a 0 at DRB means that nothing is filled. The driver's logic, however, do some calculation with that. The second one is that the logic that fills the DIMM data currently assumes that both channels are equally filled. I tested the system already with the current configuration and my patch and it is now working fine. So, for a 2R single DIMM 2Gb memory at dimm slot 01 (channel 0), it is now displaying: [ 16.741406] EDAC DEBUG: i3200_get_drbs: drb[0][0] = 16, drb[1][0] = 0 [ 16.741410] EDAC DEBUG: i3200_get_drbs: drb[0][1] = 32, drb[1][1] = 0 [ 16.741413] EDAC DEBUG: i3200_get_drbs: drb[0][2] = 32, drb[1][2] = 0 [ 16.741416] EDAC DEBUG: i3200_get_drbs: drb[0][3] = 32, drb[1][3] = 0 ... [ 16.741896] EDAC DEBUG: i3200_probe1: csrow 0, channel 0, size = 1024 Mb [ 16.741899] EDAC DEBUG: i3200_probe1: csrow 1, channel 0, size = 1024 Mb and the corresponding sysfs nodes are now properly filled. Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2013-01-10 16:31:47 +00:00
nr_pages = drb_to_nr_pages(drbs, stacked, j, i);
if (nr_pages == 0)
continue;
edac_dbg(0, "csrow %d, channel %d%s, size = %ld MiB\n", i, j,
i3200_edac: Fix the logic that detects filled memories After running a series of tests on an HP DL320, filled with different memory sizes, it was noticed that, when filled with just one DIMM on such hardware, the driver wrongly detects twice the memory, and thinks that both channels 0 and 1 are filled. It seems to be partially caused by the BIOS and partially by the driver. The i3200_edac current logic would be working fine if the BIOS were disabling the unused second channel when just one DIMM is connected, in order to do power-saving, as recommended on this chipset's datasheet. However, the BIOS on this particular machine doesn't do it: [ 16.741421] EDAC DEBUG: how_many_channels: In dual channel mode [ 16.741424] EDAC DEBUG: how_many_channels: 2 DIMMS per channel enabled So, the driver were assuming that 2 channels are enabled (well, they are, but the second is unused). Combined with that, I found two issues at the logic that creates the EDAC data, that were failing when the two channels are not equally filled (AFAICT, that happens only when just 1 DIMM is plugged). The first one is that a 0 at DRB means that nothing is filled. The driver's logic, however, do some calculation with that. The second one is that the logic that fills the DIMM data currently assumes that both channels are equally filled. I tested the system already with the current configuration and my patch and it is now working fine. So, for a 2R single DIMM 2Gb memory at dimm slot 01 (channel 0), it is now displaying: [ 16.741406] EDAC DEBUG: i3200_get_drbs: drb[0][0] = 16, drb[1][0] = 0 [ 16.741410] EDAC DEBUG: i3200_get_drbs: drb[0][1] = 32, drb[1][1] = 0 [ 16.741413] EDAC DEBUG: i3200_get_drbs: drb[0][2] = 32, drb[1][2] = 0 [ 16.741416] EDAC DEBUG: i3200_get_drbs: drb[0][3] = 32, drb[1][3] = 0 ... [ 16.741896] EDAC DEBUG: i3200_probe1: csrow 0, channel 0, size = 1024 Mb [ 16.741899] EDAC DEBUG: i3200_probe1: csrow 1, channel 0, size = 1024 Mb and the corresponding sysfs nodes are now properly filled. Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2013-01-10 16:31:47 +00:00
stacked ? " (stacked)" : "", PAGES_TO_MiB(nr_pages));
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-27 21:38:08 +00:00
dimm->nr_pages = nr_pages;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-27 21:38:08 +00:00
dimm->grain = nr_pages << PAGE_SHIFT;
dimm->mtype = MEM_DDR2;
dimm->dtype = DEV_UNKNOWN;
dimm->edac_mode = EDAC_UNKNOWN;
}
}
i3200_clear_error_info(mci);
rc = -ENODEV;
if (edac_mc_add_mc(mci)) {
edac_dbg(3, "MC: failed edac_mc_add_mc()\n");
goto fail;
}
/* get this far and it's successful */
edac_dbg(3, "MC: success\n");
return 0;
fail:
iounmap(window);
if (mci)
edac_mc_free(mci);
return rc;
}
static int i3200_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
{
int rc;
edac_dbg(0, "MC:\n");
if (pci_enable_device(pdev) < 0)
return -EIO;
rc = i3200_probe1(pdev, ent->driver_data);
if (!mci_pdev)
mci_pdev = pci_dev_get(pdev);
return rc;
}
static void i3200_remove_one(struct pci_dev *pdev)
{
struct mem_ctl_info *mci;
struct i3200_priv *priv;
edac_dbg(0, "\n");
mci = edac_mc_del_mc(&pdev->dev);
if (!mci)
return;
priv = mci->pvt_info;
iounmap(priv->window);
edac_mc_free(mci);
pci_disable_device(pdev);
}
static const struct pci_device_id i3200_pci_tbl[] = {
{
PCI_VEND_DEV(INTEL, 3200_HB), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
I3200},
{
0,
} /* 0 terminated list. */
};
MODULE_DEVICE_TABLE(pci, i3200_pci_tbl);
static struct pci_driver i3200_driver = {
.name = EDAC_MOD_STR,
.probe = i3200_init_one,
.remove = i3200_remove_one,
.id_table = i3200_pci_tbl,
};
static int __init i3200_init(void)
{
int pci_rc;
edac_dbg(3, "MC:\n");
/* Ensure that the OPSTATE is set correctly for POLL or NMI */
opstate_init();
pci_rc = pci_register_driver(&i3200_driver);
if (pci_rc < 0)
goto fail0;
if (!mci_pdev) {
i3200_registered = 0;
mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
PCI_DEVICE_ID_INTEL_3200_HB, NULL);
if (!mci_pdev) {
edac_dbg(0, "i3200 pci_get_device fail\n");
pci_rc = -ENODEV;
goto fail1;
}
pci_rc = i3200_init_one(mci_pdev, i3200_pci_tbl);
if (pci_rc < 0) {
edac_dbg(0, "i3200 init fail\n");
pci_rc = -ENODEV;
goto fail1;
}
}
return 0;
fail1:
pci_unregister_driver(&i3200_driver);
fail0:
pci_dev_put(mci_pdev);
return pci_rc;
}
static void __exit i3200_exit(void)
{
edac_dbg(3, "MC:\n");
pci_unregister_driver(&i3200_driver);
if (!i3200_registered) {
i3200_remove_one(mci_pdev);
pci_dev_put(mci_pdev);
}
}
module_init(i3200_init);
module_exit(i3200_exit);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Akamai Technologies, Inc.");
MODULE_DESCRIPTION("MC support for Intel 3200 memory hub controllers");
module_param(edac_op_state, int, 0444);
MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");