| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20230920194656.30879-1-fancer.lancer@gmail.com>
Date: Wed, 20 Sep 2023 22:46:53 +0300
From: Serge Semin <fancer.lancer@...il.com>
To: Michal Simek <michal.simek@....com>,
Alexander Stein <alexander.stein@...tq-group.com>,
Borislav Petkov <bp@...en8.de>,
Tony Luck <tony.luck@...el.com>,
James Morse <james.morse@....com>,
Mauro Carvalho Chehab <mchehab@...nel.org>,
Robert Richter <rric@...nel.org>
Cc: Punnaiah Choudary Kalluri <punnaiah.choudary.kalluri@...inx.com>,
Dinh Nguyen <dinguyen@...nel.org>,
Arnd Bergmann <arnd@...db.de>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
linux-arm-kernel@...ts.infradead.org, linux-edac@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: [PATCH v4 16/18] EDAC/synopsys: Add erroneous page-frame/offset reporting
A full System/SDRAM address translation interface is now available. Use it
to determine the system address causing the ECC faults: add the System->
Application->HIF->SDRAM address translation procedures based on the DW
uMCTL2 DDRC DQ-bus config and HIF/SDRAM mapping table retrieved on the
device probe stage; for the sake of simplification convert the
snps_ecc_error_info structure to containing the snps_sdram_addr structure
instance, since the erroneous SDRAM address will now participate in the
address translation chain; issue the SDRAM->System address translation
before passing the later to the edac_mc_handle_error() method.
Note the ECC address rank needs to be retrieved now too in order to
determine a correct system address. But the rank won't be passed to the
MCI core for now since the MCI device is registered with a single ranked
layer 0.
Signed-off-by: Serge Semin <fancer.lancer@...il.com>
---
drivers/edac/synopsys_edac.c | 143 +++++++++++++++++++++++++++++------
1 file changed, 118 insertions(+), 25 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 6b8949c66eef..5384e93ec58c 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -12,6 +12,7 @@
#include <linux/fs.h>
#include <linux/log2.h>
#include <linux/module.h>
+#include <linux/pfn.h>
#include <linux/platform_device.h>
#include <linux/seq_file.h>
#include <linux/spinlock.h>
@@ -346,20 +347,14 @@ struct snps_sdram_addr {
/**
* struct snps_ecc_error_info - ECC error log information.
- * @row: Row number.
- * @col: Column number.
- * @bank: Bank number.
- * @bankgrp: Bank group number.
+ * @sdram: SDRAM address.
* @syndrome: Error syndrome.
* @bitpos: Bit position.
* @data: Data causing the error.
* @ecc: Data ECC.
*/
struct snps_ecc_error_info {
- u32 row;
- u32 col;
- u32 bank;
- u32 bankgrp;
+ struct snps_sdram_addr sdram;
u32 syndrome;
u32 bitpos;
u64 data;
@@ -420,6 +415,21 @@ static void snps_map_app_to_hif(struct snps_edac_priv *priv,
*hif = app >> priv->info.dq_width;
}
+/**
+ * snps_map_hif_to_app - Map HIF address to Application address.
+ * @priv: DDR memory controller private instance data.
+ * @hif: HIF address (source).
+ * @app: Application address (destination).
+ *
+ * Backward HIF-to-App translation is just the opposite DQ-width-based
+ * shift operation.
+ */
+static void snps_map_hif_to_app(struct snps_edac_priv *priv,
+ u64 hif, u64 *app)
+{
+ *app = hif << priv->info.dq_width;
+}
+
/**
* snps_map_hif_to_sdram - Map HIF address to SDRAM address.
* @priv: DDR memory controller private instance data.
@@ -471,6 +481,58 @@ static void snps_map_hif_to_sdram(struct snps_edac_priv *priv,
}
}
+/**
+ * snps_map_sdram_to_hif - Map SDRAM address to HIF address.
+ * @priv: DDR memory controller private instance data.
+ * @sdram: SDRAM address (source).
+ * @hif: HIF address (destination).
+ *
+ * SDRAM-HIF address mapping is similar to the HIF-SDRAM mapping procedure, but
+ * we'll traverse each SDRAM rank/bank/column/row bit.
+ *
+ * Note the unmapped bits of the SDRAM address components will be just
+ * ignored. So make sure the source address is valid.
+ */
+static void snps_map_sdram_to_hif(struct snps_edac_priv *priv,
+ struct snps_sdram_addr *sdram, u64 *hif)
+{
+ struct snps_hif_sdram_map *map = &priv->hif_sdram_map;
+ unsigned long addr;
+ int i;
+
+ *hif = 0;
+
+ addr = sdram->row;
+ for_each_set_bit(i, &addr, DDR_MAX_ROW_WIDTH) {
+ if (map->row[i] != DDR_ADDRMAP_UNUSED)
+ *hif |= BIT_ULL(map->row[i]);
+ }
+
+ addr = sdram->col;
+ for_each_set_bit(i, &addr, DDR_MAX_COL_WIDTH) {
+ if (map->col[i] != DDR_ADDRMAP_UNUSED)
+ *hif |= BIT_ULL(map->col[i]);
+ }
+
+ addr = sdram->bank;
+ for_each_set_bit(i, &addr, DDR_MAX_BANK_WIDTH) {
+ if (map->bank[i] != DDR_ADDRMAP_UNUSED)
+ *hif |= BIT_ULL(map->bank[i]);
+ }
+
+ addr = sdram->bankgrp;
+ for_each_set_bit(i, &addr, DDR_MAX_BANKGRP_WIDTH) {
+ if (map->bankgrp[i] != DDR_ADDRMAP_UNUSED)
+ *hif |= BIT_ULL(map->bankgrp[i]);
+ }
+
+ addr = sdram->rank;
+ for_each_set_bit(i, &addr, DDR_MAX_RANK_WIDTH) {
+ if (map->rank[i] != DDR_ADDRMAP_UNUSED)
+ *hif |= BIT_ULL(map->rank[i]);
+ }
+}
+
/**
* snps_map_sys_to_sdram - Map System address to SDRAM address.
* @priv: DDR memory controller private instance data.
@@ -492,6 +554,27 @@ static void snps_map_sys_to_sdram(struct snps_edac_priv *priv,
snps_map_hif_to_sdram(priv, hif, sdram);
}
+/**
+ * snps_map_sdram_to_sys - Map SDRAM address to SDRAM address.
+ * @priv: DDR memory controller private instance data.
+ * @sys: System address (source).
+ * @sdram: SDRAM address (destination).
+ *
+ * Perform a full mapping of the SDRAM address (row/column/bank/etc) to
+ * the system address specific to the controller system bus ports.
+ */
+static void snps_map_sdram_to_sys(struct snps_edac_priv *priv,
+ struct snps_sdram_addr *sdram, dma_addr_t *sys)
+{
+ u64 app, hif;
+
+ snps_map_sdram_to_hif(priv, sdram, &hif);
+
+ snps_map_hif_to_app(priv, hif, &app);
+
+ *sys = app;
+}
+
/**
* snps_get_bitpos - Get DQ-bus corrected bit position.
* @syndrome: Error syndrome.
@@ -544,12 +627,13 @@ static int snps_get_error_info(struct snps_edac_priv *priv)
p->ceinfo.bitpos = snps_get_bitpos(p->ceinfo.syndrome, priv->info.dq_width);
regval = readl(base + ECC_CEADDR0_OFST);
- p->ceinfo.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
+ p->ceinfo.sdram.rank = FIELD_GET(ECC_CEADDR0_RANK_MASK, regval);
+ p->ceinfo.sdram.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
regval = readl(base + ECC_CEADDR1_OFST);
- p->ceinfo.bank = FIELD_GET(ECC_CEADDR1_BANK_MASK, regval);
- p->ceinfo.bankgrp = FIELD_GET(ECC_CEADDR1_BANKGRP_MASK, regval);
- p->ceinfo.col = FIELD_GET(ECC_CEADDR1_COL_MASK, regval);
+ p->ceinfo.sdram.bankgrp = FIELD_GET(ECC_CEADDR1_BANKGRP_MASK, regval);
+ p->ceinfo.sdram.bank = FIELD_GET(ECC_CEADDR1_BANK_MASK, regval);
+ p->ceinfo.sdram.col = FIELD_GET(ECC_CEADDR1_COL_MASK, regval);
p->ceinfo.data = readl(base + ECC_CSYND0_OFST);
if (priv->info.dq_width == SNPS_DQ_64)
@@ -562,12 +646,13 @@ static int snps_get_error_info(struct snps_edac_priv *priv)
goto out;
regval = readl(base + ECC_UEADDR0_OFST);
- p->ueinfo.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
+ p->ueinfo.sdram.rank = FIELD_GET(ECC_CEADDR0_RANK_MASK, regval);
+ p->ueinfo.sdram.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
regval = readl(base + ECC_UEADDR1_OFST);
- p->ueinfo.bankgrp = FIELD_GET(ECC_CEADDR1_BANKGRP_MASK, regval);
- p->ueinfo.bank = FIELD_GET(ECC_CEADDR1_BANK_MASK, regval);
- p->ueinfo.col = FIELD_GET(ECC_CEADDR1_COL_MASK, regval);
+ p->ueinfo.sdram.bankgrp = FIELD_GET(ECC_CEADDR1_BANKGRP_MASK, regval);
+ p->ueinfo.sdram.bank = FIELD_GET(ECC_CEADDR1_BANK_MASK, regval);
+ p->ueinfo.sdram.col = FIELD_GET(ECC_CEADDR1_COL_MASK, regval);
p->ueinfo.data = readl(base + ECC_UESYND0_OFST);
if (priv->info.dq_width == SNPS_DQ_64)
@@ -599,31 +684,39 @@ static void snps_handle_error(struct mem_ctl_info *mci, struct snps_ecc_status *
{
struct snps_edac_priv *priv = mci->pvt_info;
struct snps_ecc_error_info *pinf;
+ dma_addr_t sys;
if (p->ce_cnt) {
pinf = &p->ceinfo;
+ snps_map_sdram_to_sys(priv, &pinf->sdram, &sys);
+
snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
- "Row %d Col %d Bank %d Bank Group %d Bit %d Data 0x%08llx:0x%02x",
- pinf->row, pinf->col, pinf->bank, pinf->bankgrp,
+ "Row %hu Col %hu Bank %hhu Bank Group %hhu Rank %hhu Bit %d Data 0x%08llx:0x%02x",
+ pinf->sdram.row, pinf->sdram.col, pinf->sdram.bank,
+ pinf->sdram.bankgrp, pinf->sdram.rank,
pinf->bitpos, pinf->data, pinf->ecc);
- edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
- p->ce_cnt, 0, 0, pinf->syndrome, 0, 0, -1,
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, p->ce_cnt,
+ PHYS_PFN(sys), offset_in_page(sys),
+ pinf->syndrome, 0, 0, -1,
priv->message, "");
}
if (p->ue_cnt) {
pinf = &p->ueinfo;
+ snps_map_sdram_to_sys(priv, &pinf->sdram, &sys);
+
snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
- "Row %d Col %d Bank %d Bank Group %d Data 0x%08llx:0x%02x",
- pinf->row, pinf->col, pinf->bank, pinf->bankgrp,
+ "Row %hu Col %hu Bank %hhu Bank Group %hhu Rank %hhu Data 0x%08llx:0x%02x",
+ pinf->sdram.row, pinf->sdram.col, pinf->sdram.bank,
+ pinf->sdram.bankgrp, pinf->sdram.rank,
pinf->data, pinf->ecc);
- edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
- p->ue_cnt, 0, 0, 0, 0, 0, -1,
- priv->message, "");
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, p->ue_cnt,
+ PHYS_PFN(sys), offset_in_page(sys),
+ 0, 0, 0, -1, priv->message, "");
}
memset(p, 0, sizeof(*p));
--
2.41.0
Powered by blists - more mailing lists