* GNU General Public License.
*
* Written by Arvind R.
- * Copied from i82875p_edac.c source:
+ * Copied from i82875p_edac.c source,
+ *
+ * (c) 2012 Mauro Carvalho Chehab <mchehab@redhat.com>
+ * Driver re-written in order to fix lots of issues on it:
+ * - Fix Single Mode;
+ * - Add support for Asymetrical mode
+ * - Fix several issues with Interleaved mode
+ * - Fix memory label logic
+ *
+ * Intel datasheet: Intel(R) 975X Express Chipset Datasheet
+ * http://www.intel.com/assets/pdf/datasheet/310158.pdf
*/
#include <linux/module.h>
#define PCI_DEVICE_ID_INTEL_82975_0 0x277c
#endif /* PCI_DEVICE_ID_INTEL_82975_0 */
-#define I82975X_NR_DIMMS 8
-#define I82975X_NR_CSROWS(nr_chans) (I82975X_NR_DIMMS / (nr_chans))
+#define DIMMS_PER_CHANNEL 4
+#define NUM_CHANNELS 2
/* Intel 82975X register addresses - device 0 function 0 - DRAM Controller */
#define I82975X_EAP 0x58 /* Dram Error Address Pointer (32b)
/* NOTE: Following addresses have to indexed using MCHBAR offset (44h, 32b) */
/* Intel 82975x memory mapped register space */
-#define I82975X_DRB_SHIFT 25 /* fixed 32MiB grain */
+#define I82975X_DRB_SHIFT 25 /* fixed 2^25 = 32 MiB grain */
#define I82975X_DRB 0x100 /* DRAM Row Boundary (8b x 8)
*
#define I82975X_DRA_CH1R01 0x188
#define I82975X_DRA_CH1R23 0x189
+/* Channels 0/1 DRAM Timing Register 1 */
+#define I82975X_C0DRT1 0x114
+#define I82975X_C1DRT1 0x194
+
#define I82975X_BNKARC 0x10e /* Type of device in each rank - Bank Arch (16b)
*
I82975X = 0,
};
+struct mem_range {
+ u32 start, end;
+};
+
struct i82975x_pvt {
- void __iomem *mch_window;
+ void __iomem *mch_window;
+ int num_channels;
+ bool is_symetric;
+ u8 drb[DIMMS_PER_CHANNEL][NUM_CHANNELS];
+ struct mem_range page[DIMMS_PER_CHANNEL][NUM_CHANNELS];
};
struct i82975x_dev_info {
static int i82975x_process_error_info(struct mem_ctl_info *mci,
struct i82975x_error_info *info, int handle_errors)
{
- int row, chan;
- unsigned long offst, page;
+ struct i82975x_pvt *pvt = mci->pvt_info;
+ struct mem_range *range;
+ unsigned int row, chan, grain;
+ unsigned long offst, page;
if (!(info->errsts2 & 0x0003))
return 0;
info->errsts = info->errsts2;
}
+ /* Calculate page and offset of the error */
+
page = (unsigned long) info->eap;
page >>= 1;
+
if (info->xeap & 1)
page |= 0x80000000;
page >>= (PAGE_SHIFT - 1);
- row = edac_mc_find_csrow_by_page(mci, page);
-
- if (row == -1) {
- i82975x_mc_printk(mci, KERN_ERR, "error processing EAP:\n"
- "\tXEAP=%u\n"
- "\t EAP=0x%08x\n"
- "\tPAGE=0x%08x\n",
- (info->xeap & 1) ? 1 : 0, info->eap, (unsigned int) page);
- return 0;
+
+ if (pvt->is_symetric)
+ grain = 1 << 7;
+ else
+ grain = 1 << 6;
+
+ offst = info->eap & ((1 << PAGE_SHIFT) - (1 << grain));
+
+ /*
+ * Search for the DIMM chip that match the error page.
+ *
+ * On Symmetric mode, this will always return channel = 0, as
+ * both channel A and B ranges are identical.
+ * A latter logic will determinte the channel on symetric mode
+ *
+ * On asymetric mode or single mode, there will be just one match,
+ * that will point to the csrow with the error.
+ */
+ for (chan = 0; chan < pvt->num_channels; chan++) {
+ for (row = 0; row < DIMMS_PER_CHANNEL; row++) {
+ range = &pvt->page[row][chan];
+
+ if (page >= range->start && page <= range->end)
+ goto found;
+ }
}
- chan = (mci->csrows[row]->nr_channels == 1) ? 0 : info->eap & 1;
- offst = info->eap
- & ((1 << PAGE_SHIFT) -
- (1 << mci->csrows[row]->channels[chan]->dimm->grain));
+ chan = -1;
+ row = -1;
- if (info->errsts & 0x0002)
+found:
+ if (info->errsts & 0x0002) {
+ /*
+ * On uncorrected error, ECC doesn't allow do determine the
+ * channel where the error has occurred.
+ */
edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
page, offst, 0,
row, -1, -1,
"i82975x UE", "");
- else
- edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
- page, offst, info->derrsyn,
- row, chan ? chan : 0, -1,
- "i82975x CE", "");
+ return 1;
+ }
+
+ if (pvt->is_symetric && row >= 0) {
+ /*
+ * On Symetric mode, the memory switch happens after each
+ * cache line (64 byte boundary). Channel 0 goes first.
+ */
+ if (info->eap & (1 << 6))
+ chan = 1;
+ else
+ chan = 0;
+ }
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ page, offst, info->derrsyn,
+ row, chan, -1,
+ "i82975x CE", "");
return 1;
}
{
struct i82975x_error_info info;
- edac_dbg(1, "MC%d\n", mci->mc_idx);
+ edac_dbg(4, "MC%d\n", mci->mc_idx);
i82975x_get_error_info(mci, &info);
i82975x_process_error_info(mci, &info, 1);
}
-/* Return 1 if dual channel mode is active. Else return 0. */
-static int dual_channel_active(void __iomem *mch_window)
+/**
+ * detect_memory_style - Detect on what mode the memory controller is programmed
+ *
+ * @pvt: pointer to the private structure
+ *
+ * This function detects how many channels are in use, and if the memory
+ * controller is in symetric (interleaved) or asymetric mode. There's no
+ * need to distinguish between asymetric and single mode, as the routines
+ * that fill the csrows data and handle error are written in order to handle
+ * both at the same way.
+ */
+static void detect_memory_style(struct i82975x_pvt *pvt)
{
- /*
- * We treat interleaved-symmetric configuration as dual-channel - EAP's
- * bit-0 giving the channel of the error location.
- *
- * All other configurations are treated as single channel - the EAP's
- * bit-0 will resolve ok in symmetric area of mixed
- * (symmetric/asymmetric) configurations
- */
- u8 drb[4][2];
int row;
- int dualch;
+ bool has_chan_a = false;
+ bool has_chan_b = false;
- for (dualch = 1, row = 0; dualch && (row < 4); row++) {
- drb[row][0] = readb(mch_window + I82975X_DRB + row);
- drb[row][1] = readb(mch_window + I82975X_DRB + row + 0x80);
- dualch = dualch && (drb[row][0] == drb[row][1]);
+ pvt->is_symetric = true;
+ pvt->num_channels = 0;
+
+ for (row = 0; row < DIMMS_PER_CHANNEL; row++) {
+ pvt->drb[row][0] = readb(pvt->mch_window + I82975X_DRB + row);
+ pvt->drb[row][1] = readb(pvt->mch_window + I82975X_DRB + row + 0x80);
+
+ /* On symetric mode, both channels have the same boundaries */
+ if (pvt->drb[row][0] != pvt->drb[row][1])
+ pvt->is_symetric = false;
+
+ if (pvt->drb[row][0])
+ has_chan_a = true;
+ if (pvt->drb[row][1])
+ has_chan_b = true;
}
- return dualch;
-}
-static enum dev_type i82975x_dram_type(void __iomem *mch_window, int rank)
-{
- /*
- * ECC is possible on i92975x ONLY with DEV_X8
- */
- return DEV_X8;
+ if (has_chan_a)
+ pvt->num_channels++;
+
+ if (has_chan_b)
+ pvt->num_channels++;
}
static void i82975x_init_csrows(struct mem_ctl_info *mci,
- struct pci_dev *pdev, void __iomem *mch_window)
+ struct i82975x_pvt *pvt,
+ struct pci_dev *pdev)
{
- static const char *labels[4] = {
- "DIMM A1", "DIMM A2",
- "DIMM B1", "DIMM B2"
- };
- struct csrow_info *csrow;
- unsigned long last_cumul_size;
- u8 value;
- u32 cumul_size, nr_pages;
- int index, chan;
- struct dimm_info *dimm;
- enum dev_type dtype;
-
- last_cumul_size = 0;
+ struct dimm_info *dimm;
+ struct mem_range *range;
+ u8 boundary;
+ u32 initial_page = 0, last_page;
+ int row, chan;
/*
- * 82875 comment:
- * The dram row boundary (DRB) reg values are boundary address
- * for each DRAM row with a granularity of 32 or 64MB (single/dual
- * channel operation). DRB regs are cumulative; therefore DRB7 will
- * contain the total memory contained in all rows.
- *
+ * This chipset provides 3 address modes:
+ * Single channel - either Channel A or channel B is filled
+ * Dual channel, interleaved: Memory is organized in pairs,
+ * where channel A gets the lower address for each pair
+ * Dual channel, asymmetric: Channel A memory goes first.
+ * In order to cover all modes, we need to start describing
+ * memories considering the dual channel, asymmetric one.
*/
- for (index = 0; index < mci->nr_csrows; index++) {
- csrow = mci->csrows[index];
-
- value = readb(mch_window + I82975X_DRB + index +
- ((index >= 4) ? 0x80 : 0));
- cumul_size = value;
- cumul_size <<= (I82975X_DRB_SHIFT - PAGE_SHIFT);
- /*
- * Adjust cumul_size w.r.t number of channels
- *
- */
- if (csrow->nr_channels > 1)
- cumul_size <<= 1;
- edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
-
- nr_pages = cumul_size - last_cumul_size;
- if (!nr_pages)
- continue;
-
+ for (chan = 0; chan < pvt->num_channels; chan++) {
/*
- * Initialise dram labels
- * index values:
- * [0-7] for single-channel; i.e. csrow->nr_channels = 1
- * [0-3] for dual-channel; i.e. csrow->nr_channels = 2
+ * On symetric mode, both channels start from address 0
*/
- dtype = i82975x_dram_type(mch_window, index);
- for (chan = 0; chan < csrow->nr_channels; chan++) {
- dimm = mci->csrows[index]->channels[chan]->dimm;
-
- dimm->nr_pages = nr_pages / csrow->nr_channels;
- strncpy(csrow->channels[chan]->dimm->label,
- labels[(index >> 1) + (chan * 2)],
- EDAC_MC_LABEL_LEN);
- dimm->grain = 1 << 7; /* 128Byte cache-line resolution */
- dimm->dtype = i82975x_dram_type(mch_window, index);
+ if (pvt->is_symetric)
+ initial_page = 0;
+
+ for (row = 0; row < DIMMS_PER_CHANNEL; row++) {
+ boundary = pvt->drb[row][chan];
+ dimm = mci->csrows[row]->channels[chan]->dimm;
+
+ last_page = boundary << (I82975X_DRB_SHIFT - PAGE_SHIFT);
+ dimm->nr_pages = last_page - initial_page;
+ if (!dimm->nr_pages)
+ continue;
+
+ range = &pvt->page[row][chan];
+ range->start = initial_page;
+ range->end = range->start + dimm->nr_pages - 1;
+
+ /*
+ * Grain is one cache-line:
+ * On dual symetric mode, it is 128 Bytes;
+ * On single mode or asymetric, it is 64 bytes.
+ */
+ if (pvt->is_symetric) {
+ dimm->grain = 1 << 7;
+
+ /*
+ * In dual interleaved mode, the addresses
+ * need to be multiplied by 2, as both
+ * channels are interlaced, and the boundary
+ * limit there actually match each DIMM size
+ */
+ range->start <<= 1;
+ range->end <<= 1;
+ } else {
+ dimm->grain = 1 << 6;
+ }
+
+ snprintf(dimm->label,
+ EDAC_MC_LABEL_LEN, "DIMM %c%d",
+ (chan == 0) ? 'A' : 'B', row);
dimm->mtype = MEM_DDR2; /* I82975x supports only DDR2 */
dimm->edac_mode = EDAC_SECDED; /* only supported */
- }
- csrow->first_page = last_cumul_size;
- csrow->last_page = cumul_size - 1;
- last_cumul_size = cumul_size;
+ /*
+ * This chipset supports both x8 and x16 memories,
+ * but datasheet doesn't describe how to distinguish
+ * between them.
+ *
+ * Also, the "Rank" comment at initial_page 17 says that
+ * ECC is only available with x8 memories. As this
+ * driver doesn't even initialize without ECC, better
+ * to assume that everything is x8. This is not
+ * actually true, on a mixed ECC/non-ECC scenario.
+ */
+ dimm->dtype = DEV_X8;
+
+ edac_dbg(1,
+ "%s: from page 0x%08x to 0x%08x (size: 0x%08x pages)\n",
+ dimm->label,
+ range->start, range->end,
+ dimm->nr_pages);
+ initial_page = last_page;
+ }
}
}
-/* #define i82975x_DEBUG_IOMEM */
-
-#ifdef i82975x_DEBUG_IOMEM
-static void i82975x_print_dram_timings(void __iomem *mch_window)
+static void i82975x_print_dram_config(struct i82975x_pvt *pvt,
+ u32 mchbar, u32 *drc)
{
+#ifdef CONFIG_EDAC_DEBUG
/*
* The register meanings are from Intel specs;
* (shows 13-5-5-5 for 800-DDR2)
* Asus P5W Bios reports 15-5-4-4
* What's your religion?
*/
- static const int caslats[4] = { 5, 4, 3, 6 };
- u32 dtreg[2];
-
- dtreg[0] = readl(mch_window + 0x114);
- dtreg[1] = readl(mch_window + 0x194);
- i82975x_printk(KERN_INFO, "DRAM Timings : Ch0 Ch1\n"
- " RAS Active Min = %d %d\n"
- " CAS latency = %d %d\n"
- " RAS to CAS = %d %d\n"
- " RAS precharge = %d %d\n",
- (dtreg[0] >> 19 ) & 0x0f,
- (dtreg[1] >> 19) & 0x0f,
- caslats[(dtreg[0] >> 8) & 0x03],
- caslats[(dtreg[1] >> 8) & 0x03],
- ((dtreg[0] >> 4) & 0x07) + 2,
- ((dtreg[1] >> 4) & 0x07) + 2,
- (dtreg[0] & 0x07) + 2,
- (dtreg[1] & 0x07) + 2
- );
+ static const int caslats[4] = { 5, 4, 3, 6 };
+ u32 dtreg[2];
+ int row;
-}
+ /* Show memory config if debug level is 1 or upper */
+ if (!edac_debug_level)
+ return;
+
+ i82975x_printk(KERN_INFO, "MCHBAR real = %0x, remapped = %p\n",
+ mchbar, pvt->mch_window);
+
+ for (row = 0; row < DIMMS_PER_CHANNEL; row++) {
+ if (row)
+ /* Only show if at least one bank is filled */
+ if ((pvt->drb[row][0] == pvt->drb[row-1][0]) &&
+ (pvt->drb[row][1] == pvt->drb[row-1][1]))
+ continue;
+
+ i82975x_printk(KERN_INFO,
+ "DRAM%i Rank Boundary Address: Channel A: 0x%08x; Channel B: 0x%08x\n",
+ row,
+ pvt->drb[row][0],
+ pvt->drb[row][1]);
+ }
+
+ i82975x_printk(KERN_INFO, "DRAM Controller mode Channel A: = 0x%08x (%s); Channel B: 0x%08x (%s)\n",
+ drc[0],
+ ((drc[0] >> 21) & 3) == 1 ?
+ "ECC enabled" : "ECC disabled",
+ drc[1],
+ ((drc[1] >> 21) & 3) == 1 ?
+ "ECC enabled" : "ECC disabled");
+
+ i82975x_printk(KERN_INFO, "Bank Architecture Channel A: 0x%08x, Channel B: 0x%08x\n",
+ readw(pvt->mch_window + I82975X_C0BNKARC),
+ readw(pvt->mch_window + I82975X_C1BNKARC));
+
+ dtreg[0] = readl(pvt->mch_window + I82975X_C0DRT1);
+ dtreg[1] = readl(pvt->mch_window + I82975X_C1DRT1);
+ i82975x_printk(KERN_INFO, "DRAM Timings : ChA ChB\n");
+ i82975x_printk(KERN_INFO, " RAS Active Min = %2d %2d\n",
+ (dtreg[0] >> 19 ) & 0x0f,(dtreg[1] >> 19) & 0x0f);
+ i82975x_printk(KERN_INFO, " CAS latency = %2d %2d\n",
+ caslats[(dtreg[0] >> 8) & 0x03],
+ caslats[(dtreg[1] >> 8) & 0x03]);
+ i82975x_printk(KERN_INFO, " RAS to CAS = %2d %2d\n",
+ ((dtreg[0] >> 4) & 0x07) + 2,
+ ((dtreg[1] >> 4) & 0x07) + 2);
+ i82975x_printk(KERN_INFO, " RAS precharge = %2d %2d\n",
+ (dtreg[0] & 0x07) + 2,
+ (dtreg[1] & 0x07) + 2);
#endif
+}
static int i82975x_probe1(struct pci_dev *pdev, int dev_idx)
{
int rc = -ENODEV;
struct mem_ctl_info *mci;
struct edac_mc_layer layers[2];
- struct i82975x_pvt *pvt;
- void __iomem *mch_window;
+ struct i82975x_pvt tmp_pvt, *pvt;
u32 mchbar;
u32 drc[2];
struct i82975x_error_info discard;
- int chans;
-#ifdef i82975x_DEBUG_IOMEM
- u8 c0drb[4];
- u8 c1drb[4];
-#endif
edac_dbg(0, "\n");
goto fail0;
}
mchbar &= 0xffffc000; /* bits 31:14 used for 16K window */
- mch_window = ioremap_nocache(mchbar, 0x1000);
+ tmp_pvt.mch_window = ioremap_nocache(mchbar, 0x1000);
+ if (!tmp_pvt.mch_window) {
+ i82975x_printk(KERN_ERR, "Couldn't map MCHBAR registers.\n");
+ rc = -ENOMEM;
+ goto fail0;
+ }
-#ifdef i82975x_DEBUG_IOMEM
- i82975x_printk(KERN_INFO, "MCHBAR real = %0x, remapped = %p\n",
- mchbar, mch_window);
-
- c0drb[0] = readb(mch_window + I82975X_DRB_CH0R0);
- c0drb[1] = readb(mch_window + I82975X_DRB_CH0R1);
- c0drb[2] = readb(mch_window + I82975X_DRB_CH0R2);
- c0drb[3] = readb(mch_window + I82975X_DRB_CH0R3);
- c1drb[0] = readb(mch_window + I82975X_DRB_CH1R0);
- c1drb[1] = readb(mch_window + I82975X_DRB_CH1R1);
- c1drb[2] = readb(mch_window + I82975X_DRB_CH1R2);
- c1drb[3] = readb(mch_window + I82975X_DRB_CH1R3);
- i82975x_printk(KERN_INFO, "DRBCH0R0 = 0x%02x\n", c0drb[0]);
- i82975x_printk(KERN_INFO, "DRBCH0R1 = 0x%02x\n", c0drb[1]);
- i82975x_printk(KERN_INFO, "DRBCH0R2 = 0x%02x\n", c0drb[2]);
- i82975x_printk(KERN_INFO, "DRBCH0R3 = 0x%02x\n", c0drb[3]);
- i82975x_printk(KERN_INFO, "DRBCH1R0 = 0x%02x\n", c1drb[0]);
- i82975x_printk(KERN_INFO, "DRBCH1R1 = 0x%02x\n", c1drb[1]);
- i82975x_printk(KERN_INFO, "DRBCH1R2 = 0x%02x\n", c1drb[2]);
- i82975x_printk(KERN_INFO, "DRBCH1R3 = 0x%02x\n", c1drb[3]);
-#endif
+ drc[0] = readl(tmp_pvt.mch_window + I82975X_DRC_CH0M0);
+ drc[1] = readl(tmp_pvt.mch_window + I82975X_DRC_CH1M0);
+
+ detect_memory_style(&tmp_pvt);
+ if (!tmp_pvt.num_channels) {
+ edac_dbg(3, "No memories installed? This shouldn't be running!\n");
+ goto fail0;
+ }
+
+ i82975x_print_dram_config(&tmp_pvt, mchbar, drc);
- drc[0] = readl(mch_window + I82975X_DRC_CH0M0);
- drc[1] = readl(mch_window + I82975X_DRC_CH1M0);
-#ifdef i82975x_DEBUG_IOMEM
- i82975x_printk(KERN_INFO, "DRC_CH0 = %0x, %s\n", drc[0],
- ((drc[0] >> 21) & 3) == 1 ?
- "ECC enabled" : "ECC disabled");
- i82975x_printk(KERN_INFO, "DRC_CH1 = %0x, %s\n", drc[1],
- ((drc[1] >> 21) & 3) == 1 ?
- "ECC enabled" : "ECC disabled");
-
- i82975x_printk(KERN_INFO, "C0 BNKARC = %0x\n",
- readw(mch_window + I82975X_C0BNKARC));
- i82975x_printk(KERN_INFO, "C1 BNKARC = %0x\n",
- readw(mch_window + I82975X_C1BNKARC));
- i82975x_print_dram_timings(mch_window);
- goto fail1;
-#endif
if (!(((drc[0] >> 21) & 3) == 1 || ((drc[1] >> 21) & 3) == 1)) {
i82975x_printk(KERN_INFO, "ECC disabled on both channels.\n");
goto fail1;
}
- chans = dual_channel_active(mch_window) + 1;
-
/* assuming only one controller, index thus is 0 */
layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
- layers[0].size = I82975X_NR_DIMMS;
+ layers[0].size = DIMMS_PER_CHANNEL;
layers[0].is_virt_csrow = true;
layers[1].type = EDAC_MC_LAYER_CHANNEL;
- layers[1].size = I82975X_NR_CSROWS(chans);
+ layers[1].size = tmp_pvt.num_channels;
layers[1].is_virt_csrow = false;
mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
if (!mci) {
mci->dev_name = pci_name(pdev);
mci->edac_check = i82975x_check;
mci->ctl_page_to_phys = NULL;
+
edac_dbg(3, "init pvt\n");
pvt = (struct i82975x_pvt *) mci->pvt_info;
- pvt->mch_window = mch_window;
- i82975x_init_csrows(mci, pdev, mch_window);
+ *pvt = tmp_pvt;
+
+ i82975x_init_csrows(mci, pvt, pdev);
mci->scrub_mode = SCRUB_HW_SRC;
i82975x_get_error_info(mci, &discard); /* clear counters */
edac_mc_free(mci);
fail1:
- iounmap(mch_window);
+ iounmap(tmp_pvt.mch_window);
fail0:
return rc;
}
projects / karo-tx-linux.git / commitdiff