summaryrefslogtreecommitdiff
path: root/drivers/edac/sb_edac.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/edac/sb_edac.c')
-rw-r--r--drivers/edac/sb_edac.c208
1 files changed, 170 insertions, 38 deletions
diff --git a/drivers/edac/sb_edac.c b/drivers/edac/sb_edac.c
index 4c626fcd4dcb..d5f12219598a 100644
--- a/drivers/edac/sb_edac.c
+++ b/drivers/edac/sb_edac.c
@@ -29,6 +29,8 @@
/* Static vars */
static LIST_HEAD(sbridge_edac_list);
+static char sb_msg[256];
+static char sb_msg_full[512];
/*
* Alter this version for the module when modifications are made
@@ -109,8 +111,8 @@ static const u32 knl_interleave_list[] = {
0x104, 0x10c, 0x114, 0x11c, /* 20-23 */
};
#define MAX_INTERLEAVE \
- (max_t(unsigned int, ARRAY_SIZE(sbridge_interleave_list), \
- max_t(unsigned int, ARRAY_SIZE(ibridge_interleave_list), \
+ (MAX_T(unsigned int, ARRAY_SIZE(sbridge_interleave_list), \
+ MAX_T(unsigned int, ARRAY_SIZE(ibridge_interleave_list), \
ARRAY_SIZE(knl_interleave_list))))
struct interleave_pkg {
@@ -335,6 +337,12 @@ struct sbridge_info {
struct sbridge_channel {
u32 ranks;
u32 dimms;
+ struct dimm {
+ u32 rowbits;
+ u32 colbits;
+ u32 bank_xor_enable;
+ u32 amap_fine;
+ } dimm[MAX_DIMMS];
};
struct pci_id_descr {
@@ -433,7 +441,7 @@ static const struct pci_id_descr pci_dev_descr_sbridge[] = {
static const struct pci_id_table pci_dev_descr_sbridge_table[] = {
PCI_ID_TABLE_ENTRY(pci_dev_descr_sbridge, ARRAY_SIZE(pci_dev_descr_sbridge), 1, SANDY_BRIDGE),
- {0,} /* 0 terminated list. */
+ { NULL, }
};
/* This changes depending if 1HA or 2HA:
@@ -499,7 +507,7 @@ static const struct pci_id_descr pci_dev_descr_ibridge[] = {
static const struct pci_id_table pci_dev_descr_ibridge_table[] = {
PCI_ID_TABLE_ENTRY(pci_dev_descr_ibridge, 12, 2, IVY_BRIDGE),
- {0,} /* 0 terminated list. */
+ { NULL, }
};
/* Haswell support */
@@ -570,7 +578,7 @@ static const struct pci_id_descr pci_dev_descr_haswell[] = {
static const struct pci_id_table pci_dev_descr_haswell_table[] = {
PCI_ID_TABLE_ENTRY(pci_dev_descr_haswell, 13, 2, HASWELL),
- {0,} /* 0 terminated list. */
+ { NULL, }
};
/* Knight's Landing Support */
@@ -614,7 +622,7 @@ static const struct pci_id_descr pci_dev_descr_knl[] = {
static const struct pci_id_table pci_dev_descr_knl_table[] = {
PCI_ID_TABLE_ENTRY(pci_dev_descr_knl, ARRAY_SIZE(pci_dev_descr_knl), 1, KNIGHTS_LANDING),
- {0,}
+ { NULL, }
};
/*
@@ -680,7 +688,7 @@ static const struct pci_id_descr pci_dev_descr_broadwell[] = {
static const struct pci_id_table pci_dev_descr_broadwell_table[] = {
PCI_ID_TABLE_ENTRY(pci_dev_descr_broadwell, 10, 2, BROADWELL),
- {0,} /* 0 terminated list. */
+ { NULL, }
};
@@ -1052,7 +1060,7 @@ static u64 haswell_get_tohm(struct sbridge_pvt *pvt)
pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOHM_1, &reg);
rc = ((reg << 6) | rc) << 26;
- return rc | 0x1ffffff;
+ return rc | 0x3ffffff;
}
static u64 knl_get_tolm(struct sbridge_pvt *pvt)
@@ -1603,7 +1611,7 @@ static int __populate_dimms(struct mem_ctl_info *mci,
banks = 8;
for (i = 0; i < channels; i++) {
- u32 mtr;
+ u32 mtr, amap = 0;
int max_dimms_per_channel;
@@ -1615,6 +1623,7 @@ static int __populate_dimms(struct mem_ctl_info *mci,
max_dimms_per_channel = ARRAY_SIZE(mtr_regs);
if (!pvt->pci_tad[i])
continue;
+ pci_read_config_dword(pvt->pci_tad[i], 0x8c, &amap);
}
for (j = 0; j < max_dimms_per_channel; j++) {
@@ -1627,6 +1636,7 @@ static int __populate_dimms(struct mem_ctl_info *mci,
mtr_regs[j], &mtr);
}
edac_dbg(4, "Channel #%d MTR%d = %x\n", i, j, mtr);
+
if (IS_DIMM_PRESENT(mtr)) {
if (!IS_ECC_ENABLED(pvt->info.mcmtr)) {
sbridge_printk(KERN_ERR, "CPU SrcID #%d, Ha #%d, Channel #%d has DIMMs, but ECC is disabled\n",
@@ -1661,6 +1671,11 @@ static int __populate_dimms(struct mem_ctl_info *mci,
dimm->dtype = pvt->info.get_width(pvt, mtr);
dimm->mtype = mtype;
dimm->edac_mode = mode;
+ pvt->channel[i].dimm[j].rowbits = order_base_2(rows);
+ pvt->channel[i].dimm[j].colbits = order_base_2(cols);
+ pvt->channel[i].dimm[j].bank_xor_enable =
+ GET_BITFIELD(pvt->info.mcmtr, 9, 9);
+ pvt->channel[i].dimm[j].amap_fine = GET_BITFIELD(amap, 0, 0);
snprintf(dimm->label, sizeof(dimm->label),
"CPU_SrcID#%u_Ha#%u_Chan#%u_DIMM#%u",
pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom, i, j);
@@ -1922,6 +1937,99 @@ static struct mem_ctl_info *get_mci_for_node_id(u8 node_id, u8 ha)
return NULL;
}
+static u8 sb_close_row[] = {
+ 15, 16, 17, 18, 20, 21, 22, 28, 10, 11, 12, 13, 29, 30, 31, 32, 33
+};
+
+static u8 sb_close_column[] = {
+ 3, 4, 5, 14, 19, 23, 24, 25, 26, 27
+};
+
+static u8 sb_open_row[] = {
+ 14, 15, 16, 20, 28, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33
+};
+
+static u8 sb_open_column[] = {
+ 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
+};
+
+static u8 sb_open_fine_column[] = {
+ 3, 4, 5, 7, 8, 9, 10, 11, 12, 13
+};
+
+static int sb_bits(u64 addr, int nbits, u8 *bits)
+{
+ int i, res = 0;
+
+ for (i = 0; i < nbits; i++)
+ res |= ((addr >> bits[i]) & 1) << i;
+ return res;
+}
+
+static int sb_bank_bits(u64 addr, int b0, int b1, int do_xor, int x0, int x1)
+{
+ int ret = GET_BITFIELD(addr, b0, b0) | (GET_BITFIELD(addr, b1, b1) << 1);
+
+ if (do_xor)
+ ret ^= GET_BITFIELD(addr, x0, x0) | (GET_BITFIELD(addr, x1, x1) << 1);
+
+ return ret;
+}
+
+static bool sb_decode_ddr4(struct mem_ctl_info *mci, int ch, u8 rank,
+ u64 rank_addr, char *msg)
+{
+ int dimmno = 0;
+ int row, col, bank_address, bank_group;
+ struct sbridge_pvt *pvt;
+ u32 bg0 = 0, rowbits = 0, colbits = 0;
+ u32 amap_fine = 0, bank_xor_enable = 0;
+
+ dimmno = (rank < 12) ? rank / 4 : 2;
+ pvt = mci->pvt_info;
+ amap_fine = pvt->channel[ch].dimm[dimmno].amap_fine;
+ bg0 = amap_fine ? 6 : 13;
+ rowbits = pvt->channel[ch].dimm[dimmno].rowbits;
+ colbits = pvt->channel[ch].dimm[dimmno].colbits;
+ bank_xor_enable = pvt->channel[ch].dimm[dimmno].bank_xor_enable;
+
+ if (pvt->is_lockstep) {
+ pr_warn_once("LockStep row/column decode is not supported yet!\n");
+ msg[0] = '\0';
+ return false;
+ }
+
+ if (pvt->is_close_pg) {
+ row = sb_bits(rank_addr, rowbits, sb_close_row);
+ col = sb_bits(rank_addr, colbits, sb_close_column);
+ col |= 0x400; /* C10 is autoprecharge, always set */
+ bank_address = sb_bank_bits(rank_addr, 8, 9, bank_xor_enable, 22, 28);
+ bank_group = sb_bank_bits(rank_addr, 6, 7, bank_xor_enable, 20, 21);
+ } else {
+ row = sb_bits(rank_addr, rowbits, sb_open_row);
+ if (amap_fine)
+ col = sb_bits(rank_addr, colbits, sb_open_fine_column);
+ else
+ col = sb_bits(rank_addr, colbits, sb_open_column);
+ bank_address = sb_bank_bits(rank_addr, 18, 19, bank_xor_enable, 22, 23);
+ bank_group = sb_bank_bits(rank_addr, bg0, 17, bank_xor_enable, 20, 21);
+ }
+
+ row &= (1u << rowbits) - 1;
+
+ sprintf(msg, "row:0x%x col:0x%x bank_addr:%d bank_group:%d",
+ row, col, bank_address, bank_group);
+ return true;
+}
+
+static bool sb_decode_ddr3(struct mem_ctl_info *mci, int ch, u8 rank,
+ u64 rank_addr, char *msg)
+{
+ pr_warn_once("DDR3 row/column decode not support yet!\n");
+ msg[0] = '\0';
+ return false;
+}
+
static int get_memory_error_data(struct mem_ctl_info *mci,
u64 addr,
u8 *socket, u8 *ha,
@@ -1937,12 +2045,13 @@ static int get_memory_error_data(struct mem_ctl_info *mci,
int interleave_mode, shiftup = 0;
unsigned int sad_interleave[MAX_INTERLEAVE];
u32 reg, dram_rule;
- u8 ch_way, sck_way, pkg, sad_ha = 0;
+ u8 ch_way, sck_way, pkg, sad_ha = 0, rankid = 0;
u32 tad_offset;
u32 rir_way;
u32 mb, gb;
u64 ch_addr, offset, limit = 0, prv = 0;
-
+ u64 rank_addr;
+ enum mem_type mtype;
/*
* Step 0) Check if the address is at special memory ranges
@@ -2226,6 +2335,28 @@ static int get_memory_error_data(struct mem_ctl_info *mci,
pci_read_config_dword(pvt->pci_tad[base_ch], rir_offset[n_rir][idx], &reg);
*rank = RIR_RNK_TGT(pvt->info.type, reg);
+ if (pvt->info.type == BROADWELL) {
+ if (pvt->is_close_pg)
+ shiftup = 6;
+ else
+ shiftup = 13;
+
+ rank_addr = ch_addr >> shiftup;
+ rank_addr /= (1 << rir_way);
+ rank_addr <<= shiftup;
+ rank_addr |= ch_addr & GENMASK_ULL(shiftup - 1, 0);
+ rank_addr -= RIR_OFFSET(pvt->info.type, reg);
+
+ mtype = pvt->info.get_memory_type(pvt);
+ rankid = *rank;
+ if (mtype == MEM_DDR4 || mtype == MEM_RDDR4)
+ sb_decode_ddr4(mci, base_ch, rankid, rank_addr, msg);
+ else
+ sb_decode_ddr3(mci, base_ch, rankid, rank_addr, msg);
+ } else {
+ msg[0] = '\0';
+ }
+
edac_dbg(0, "RIR#%d: channel address 0x%08Lx < 0x%08Lx, RIR interleave %d, index %d\n",
n_rir,
ch_addr,
@@ -2950,7 +3081,6 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
struct mem_ctl_info *new_mci;
struct sbridge_pvt *pvt = mci->pvt_info;
enum hw_event_mc_err_type tp_event;
- char *optype, msg[256];
bool ripv = GET_BITFIELD(m->mcgstatus, 0, 0);
bool overflow = GET_BITFIELD(m->status, 62, 62);
bool uncorrected_error = GET_BITFIELD(m->status, 61, 61);
@@ -2966,10 +3096,10 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
* aligned address reported by patrol scrubber.
*/
u32 lsb = GET_BITFIELD(m->misc, 0, 5);
+ char *optype, *area_type = "DRAM";
long channel_mask, first_channel;
u8 rank = 0xff, socket, ha;
int rc, dimm;
- char *area_type = "DRAM";
if (pvt->info.type != SANDY_BRIDGE)
recoverable = true;
@@ -3039,32 +3169,32 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
channel = knl_channel_remap(m->bank == 16, channel);
channel_mask = 1 << channel;
- snprintf(msg, sizeof(msg),
- "%s%s err_code:%04x:%04x channel:%d (DIMM_%c)",
- overflow ? " OVERFLOW" : "",
- (uncorrected_error && recoverable)
- ? " recoverable" : " ",
- mscod, errcode, channel, A + channel);
+ snprintf(sb_msg, sizeof(sb_msg),
+ "%s%s err_code:%04x:%04x channel:%d (DIMM_%c)",
+ overflow ? " OVERFLOW" : "",
+ (uncorrected_error && recoverable)
+ ? " recoverable" : " ",
+ mscod, errcode, channel, A + channel);
edac_mc_handle_error(tp_event, mci, core_err_cnt,
m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
channel, 0, -1,
- optype, msg);
+ optype, sb_msg);
}
return;
} else if (lsb < 12) {
rc = get_memory_error_data(mci, m->addr, &socket, &ha,
&channel_mask, &rank,
- &area_type, msg);
+ &area_type, sb_msg);
} else {
rc = get_memory_error_data_from_mce(mci, m, &socket, &ha,
- &channel_mask, msg);
+ &channel_mask, sb_msg);
}
if (rc < 0)
goto err_parsing;
new_mci = get_mci_for_node_id(socket, ha);
if (!new_mci) {
- strcpy(msg, "Error: socket got corrupted!");
+ strscpy(sb_msg, "Error: socket got corrupted!");
goto err_parsing;
}
mci = new_mci;
@@ -3089,18 +3219,17 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
*/
if (!pvt->is_lockstep && !pvt->is_cur_addr_mirrored && !pvt->is_close_pg)
channel = first_channel;
-
- snprintf(msg, sizeof(msg),
- "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d",
+ snprintf(sb_msg_full, sizeof(sb_msg_full),
+ "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d %s",
overflow ? " OVERFLOW" : "",
(uncorrected_error && recoverable) ? " recoverable" : "",
area_type,
mscod, errcode,
socket, ha,
channel_mask,
- rank);
+ rank, sb_msg);
- edac_dbg(0, "%s\n", msg);
+ edac_dbg(0, "%s\n", sb_msg_full);
/* FIXME: need support for channel mask */
@@ -3111,12 +3240,12 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
edac_mc_handle_error(tp_event, mci, core_err_cnt,
m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
channel, dimm, -1,
- optype, msg);
+ optype, sb_msg_full);
return;
err_parsing:
edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0,
-1, -1, -1,
- msg, "");
+ sb_msg, "");
}
@@ -3418,13 +3547,13 @@ fail0:
}
static const struct x86_cpu_id sbridge_cpuids[] = {
- X86_MATCH_INTEL_FAM6_MODEL(SANDYBRIDGE_X, &pci_dev_descr_sbridge_table),
- X86_MATCH_INTEL_FAM6_MODEL(IVYBRIDGE_X, &pci_dev_descr_ibridge_table),
- X86_MATCH_INTEL_FAM6_MODEL(HASWELL_X, &pci_dev_descr_haswell_table),
- X86_MATCH_INTEL_FAM6_MODEL(BROADWELL_X, &pci_dev_descr_broadwell_table),
- X86_MATCH_INTEL_FAM6_MODEL(BROADWELL_D, &pci_dev_descr_broadwell_table),
- X86_MATCH_INTEL_FAM6_MODEL(XEON_PHI_KNL, &pci_dev_descr_knl_table),
- X86_MATCH_INTEL_FAM6_MODEL(XEON_PHI_KNM, &pci_dev_descr_knl_table),
+ X86_MATCH_VFM(INTEL_SANDYBRIDGE_X, &pci_dev_descr_sbridge_table),
+ X86_MATCH_VFM(INTEL_IVYBRIDGE_X, &pci_dev_descr_ibridge_table),
+ X86_MATCH_VFM(INTEL_HASWELL_X, &pci_dev_descr_haswell_table),
+ X86_MATCH_VFM(INTEL_BROADWELL_X, &pci_dev_descr_broadwell_table),
+ X86_MATCH_VFM(INTEL_BROADWELL_D, &pci_dev_descr_broadwell_table),
+ X86_MATCH_VFM(INTEL_XEON_PHI_KNL, &pci_dev_descr_knl_table),
+ X86_MATCH_VFM(INTEL_XEON_PHI_KNM, &pci_dev_descr_knl_table),
{ }
};
MODULE_DEVICE_TABLE(x86cpu, sbridge_cpuids);
@@ -3439,7 +3568,7 @@ MODULE_DEVICE_TABLE(x86cpu, sbridge_cpuids);
static int sbridge_probe(const struct x86_cpu_id *id)
{
- int rc = -ENODEV;
+ int rc;
u8 mc, num_mc = 0;
struct sbridge_dev *sbridge_dev;
struct pci_id_table *ptable = (struct pci_id_table *)id->driver_data;
@@ -3506,6 +3635,9 @@ static int __init sbridge_init(void)
edac_dbg(2, "\n");
+ if (ghes_get_devices())
+ return -EBUSY;
+
owner = edac_get_owner();
if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
return -EBUSY;
generated by cgit (git 2.34.1) at 2025-12-25 06:07:32 +0000