This patchset is a third one in the series created in the framework of my
Baikal-T1 DDRC-related work:
[1: In-progress] EDAC/mc/synopsys: Various fixes and cleanups
Link: https://lore.kernel.org/linux-edac/[email protected]
[2: In-progress] EDAC/synopsys: Add generic DDRC info and address mapping
Link: https://lore.kernel.org/linux-edac/[email protected]
[3: In-progress] EDAC/synopsys: Add generic resources and Baikal-T1 support
Link: ---you are looking at it---
Note the patchsets above must be merged in the same order as they are
placed in the list in order to prevent conflicts. Nothing prevents them
from being reviewed synchronously though. Any tests are very welcome.
Thanks in advance.
This is a final patchset in the framework of my Synopsys DW uMCTL2 DDRC
work, which completes the driver updates with the new functionality and
at the closure introduces the Baikal-T1 DDRC support.
The series starts from extending the Synopsys DW uMCTL2 DDRC DT-schema
with the controller specific IRQs, clocks and resets properties. In
addition to the Baikal-T1 DDRC is added to the DT-bindings since it's
based on the DW uMCTL2 DDRC v2.61a.
After that we suggest to finally inform the MCI core with the detected
SDRAM ranks and make sure the detected errors are reported to the
corresponding rank. Then we extend the DDRC capabilities with optional
Scrub functionality. It's indeed possible to have the DW uMCTL2 controller
with no HW-accelerated Scrub support (no RMW engine). In that case the MCI
core is supposed to perform the erroneous location ECC update by means of
the platform-specific scrub method.
Then we get to fix the error-injection functionality a bit. First since
the driver now has the Sys<->SDRAM address translation infrastructure we
can use it to convert the supplied poisonous system address to the SDRAM
one. Thus there is no longer need in preserving the address in the device
private data. Second we suggest to add a DebuFS node-based command to
disable the error-injection feature (no idea why it hasn't been done in
the first place).
Afterwards a series of the IRQ-related patches goes. First we introduce the
individual DDRC event IRQs support in accordance with what has been added
to the DT-bindings and what the native DW uMCTL2 DDR controller actually
provides. Then aside to the ECC CE/UE errors detection we suggest to the
DFI/SDRAM CRC/Parity errors report. It specifically useful for the DDR4
memory which has dedicated ALARM_n signal, but can be still utilized in
the framework of the older protocols if the device DFI-PHY calculates
the HIF-interface signals parity. Third after adding the platform
clock/resets request procedure we introduce the HW-accelerated Scrubber
support. Its performance can be tuned by means of the sdram_scrub_rate
SysFS node and the Core clock rate. Note it is possible to one-time-run
the Scrubber in the back-to-back mode so to perform a burst-like scan of
the whole SDRAM memory.
At the patchset closure we finally fix the DW uMCTL2 DDRC kernel config to
be available not only on the Xilinx, Intel and MXC platforms and add the
Baikal-T1 DDRC support which the whole work has been dedicated for in the
first place.
Link: https://lore.kernel.org/linux-edac/[email protected]/
Changelog v2:
- Replace "snps,ddrc-3.80a" compatible string with "snps,dw-umctl2-ddrc"
in the example.
- Move unrelated changes in to the dedicated patches. (@Krzysztof)
- Use the IRQ macros in the example. (@Krzysztof)
- Add a new patch:
[PATCH v2 01/15] dt-bindings: memory: snps: Replace opencoded numbers with macros
(@Krzysztof)
- Add a new patch:
[PATCH v2 03/15] dt-bindings: memory: snps: Convert the schema to being generic
(@Krzysztof)
- Drop the PHY CSR region. (@Rob)
- Move the Baikal-T1 DDRC bindings to the separate DT-schema.
Signed-off-by: Serge Semin <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Michail Ivanov <[email protected]>
Cc: Pavel Parkhomenko <[email protected]>
Cc: Punnaiah Choudary Kalluri <[email protected]>
Cc: Manish Narani <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: James Morse <[email protected]>
Cc: Robert Richter <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Krzysztof Kozlowski <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Serge Semin (15):
dt-bindings: memory: snps: Replace opencoded numbers with macros
dt-bindings: memory: snps: Extend schema with IRQs/resets/clocks props
dt-bindings: memory: snps: Convert the schema to being generic
dt-bindings: memory: Add Baikal-T1 DDRC DT-schema
EDAC/synopsys: Add multi-ranked memory support
EDAC/synopsys: Add optional ECC Scrub support
EDAC/synopsys: Drop ECC poison address from private data
EDAC/synopsys: Add data poisoning disable support
EDAC/synopsys: Split up ECC UE/CE IRQs handler
EDAC/synopsys: Add individual named ECC IRQs support
EDAC/synopsys: Add DFI alert_n IRQ support
EDAC/synopsys: Add reference clocks support
EDAC/synopsys: Add ECC Scrubber support
EDAC/synopsys: Drop vendor-specific arch dependency
EDAC/synopsys: Add Baikal-T1 DDRC support
.../memory-controllers/baikal,bt1-ddrc.yaml | 91 ++
.../snps,dw-umctl2-ddrc.yaml | 83 +-
drivers/edac/Kconfig | 1 -
drivers/edac/synopsys_edac.c | 952 ++++++++++++++----
4 files changed, 931 insertions(+), 196 deletions(-)
create mode 100644 Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml
--
2.37.2
At the current state the DW uMCTL2 DDRC DT-schema can't be used as the
common one for all the IP-core-based devices due to the compatible string
property constraining the list of the supported device names. In order to
fix that we suggest to update the compatible property constraints so one
would permit having any value aside with the generic device names. At the
same time the generic DT-schema selection must be restricted to the
denoted generic devices only so not to permit the generic fallback
compatibles. Finally since the generic schema will be referenced from the
vendor-specific DT-bindings with possibly non-standard properties defined
it must permit having additional properties specified.
Signed-off-by: Serge Semin <[email protected]>
---
Note alternatively we could drop the "additionalProperties" keyword
modification since currently there is no actual device available with the
properties not listed in the generic DT-schema.
Changelog v2:
- This is a new patch created on v2 cycle of the patchset. (@Krzysztof)
---
.../memory-controllers/snps,dw-umctl2-ddrc.yaml | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml b/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
index e68c4306025a..a3394b4600ef 100644
--- a/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
+++ b/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
@@ -21,9 +21,21 @@ description: |
controller. It has an optional SEC/DEC ECC support in 64- and 32-bits
bus width configurations.
+# Please create a separate DT-schema for your DW uMCTL2 DDR controller
+# and make sure it's assigned with the vendor-specific compatible string.
+select:
+ properties:
+ compatible:
+ enum:
+ - snps,ddrc-3.80a
+ - snps,dw-umctl2-ddrc
+ - xlnx,zynqmp-ddrc-2.40a
+ required:
+ - compatible
+
properties:
compatible:
- oneOf:
+ anyOf:
- deprecated: true
description: Synopsys DW uMCTL2 DDR controller v3.80a
const: snps,ddrc-3.80a
@@ -31,6 +43,7 @@ properties:
const: snps,dw-umctl2-ddrc
- description: Xilinx ZynqMP DDR controller v2.40a
const: xlnx,zynqmp-ddrc-2.40a
+ - {}
interrupts:
description:
@@ -87,7 +100,7 @@ required:
- reg
- interrupts
-additionalProperties: false
+additionalProperties: true
examples:
- |
--
2.37.2
DW uMCTL2 DDRC supports multi-rank memory attached to the controller. If
so the MSTR.active_ranks field will be set with the populated ranks
bitfield. It is permitted to have one, two or four ranks activated at a
time [1]. Since we now have the number of ranks determined in the
controller configuration detection procedure, it can be easily used for
accordingly extending the MCI chip-select layer. In case of the ECC errors
the affected rank will be read from the CE/UE address CSRs [2].
Note we need to drop the ranks from the total memory size calculation
since multiple ranks are taken into account by means of the layer[0]
definition.
[1] DesignWare® Cores Enhanced Universal DDR Memory Controller (uMCTL2)
Databook, Version 3.91a, October 2020, p.739
[2] DesignWare® Cores Enhanced Universal DDR Memory Controller (uMCTL2)
Databook, Version 3.91a, October 2020, p.821, p.832
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 14 ++++----------
1 file changed, 4 insertions(+), 10 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index e1e5468acbd2..d5d82531719b 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -23,9 +23,6 @@
#include "edac_module.h"
-/* Number of cs_rows needed per memory controller */
-#define SNPS_EDAC_NR_CSROWS 1
-
/* Number of channels per memory controller */
#define SNPS_EDAC_NR_CHANS 1
@@ -795,7 +792,7 @@ static void snps_handle_error(struct mem_ctl_info *mci, struct snps_ecc_status *
edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, p->ce_cnt,
PHYS_PFN(sys), offset_in_page(sys),
- pinf->syndrome, 0, 0, -1,
+ pinf->syndrome, pinf->sdram.rank, 0, -1,
priv->message, "");
}
@@ -812,7 +809,7 @@ static void snps_handle_error(struct mem_ctl_info *mci, struct snps_ecc_status *
edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, p->ue_cnt,
PHYS_PFN(sys), offset_in_page(sys),
- pinf->syndrome, 0, 0, -1,
+ pinf->syndrome, pinf->sdram.rank, 0, -1,
priv->message, "");
}
@@ -1411,10 +1408,7 @@ static u64 snps_get_sdram_size(struct snps_edac_priv *priv)
size++;
}
- for (i = 0; i < DDR_MAX_RANK_WIDTH; i++) {
- if (map->rank[i] != DDR_ADDRMAP_UNUSED)
- size++;
- }
+ /* Skip the ranks since the multi-rankness is determined by layer[0] */
return 1ULL << (size + priv->info.dq_width);
}
@@ -1468,7 +1462,7 @@ static struct mem_ctl_info *snps_mc_create(struct snps_edac_priv *priv)
struct mem_ctl_info *mci;
layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
- layers[0].size = SNPS_EDAC_NR_CSROWS;
+ layers[0].size = priv->info.ranks;
layers[0].is_virt_csrow = true;
layers[1].type = EDAC_MC_LAYER_CHANNEL;
layers[1].size = SNPS_EDAC_NR_CHANS;
--
2.37.2
Xilinx ZynqMP DDRC-based example contains the opencoded numerical literals
in the IRQ lines definition. It doesn't seem justified since the
corresponding platform has well defined ARM GIC interface. Let's replace
the numbers with the corresponding macros then.
Signed-off-by: Serge Semin <[email protected]>
---
Changelog v2:
- This is a new patch created on v2 based on the @Krzysztof' request not
to have the opencoded numbers in the bindings example. (@Krzysztof)
---
.../bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml b/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
index 9212dfe6e956..fb571d3d665d 100644
--- a/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
+++ b/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
@@ -47,10 +47,13 @@ additionalProperties: false
examples:
- |
+ #include <dt-bindings/interrupt-controller/arm-gic.h>
+
memory-controller@fd070000 {
compatible = "xlnx,zynqmp-ddrc-2.40a";
reg = <0xfd070000 0x30000>;
+
interrupt-parent = <&gic>;
- interrupts = <0 112 4>;
+ interrupts = <GIC_SPI 112 IRQ_TYPE_LEVEL_HIGH>;
};
...
--
2.37.2
Baikal-T1 DDR controller is based on the DW uMCTL2 DDRC IP-core v2.51a
with up to DDR3 protocol capability and 32-bit data bus + 8-bit ECC. There
are individual IRQs for each ECC and DFI events. The dedicated scrubber
clock source is absent since it's fully synchronous to the core clock.
In addition to that the DFI-DDR PHY CSRs can be accessed via a separate
registers space.
Signed-off-by: Serge Semin <[email protected]>
---
Changelog v2:
- Keep the alphabetically ordered compatible strings list. (@Krzysztof)
- Fix grammar nitpicks in the patch log. (@Krzysztof)
- Drop the PHY CSR region. (@Rob)
- Move the device bindings to the separate DT-schema.
---
.../memory-controllers/baikal,bt1-ddrc.yaml | 91 +++++++++++++++++++
1 file changed, 91 insertions(+)
create mode 100644 Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml
diff --git a/Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml b/Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml
new file mode 100644
index 000000000000..a778acdd6928
--- /dev/null
+++ b/Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml
@@ -0,0 +1,91 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/memory-controllers/baikal,bt1-ddrc.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Baikal-T1 DDR Controller
+
+maintainers:
+ - Serge Semin <[email protected]>
+
+description:
+ Baikal-T1 DDRC is based on the DW uMCTL2 DDRC IP-core v2.51a with DDR2
+ and DDR3 protocol capability, 32-bit data bus + 8-bit ECC + up to 2
+ SDRAM ranks. There are individual IRQs for each ECC and DFI events.
+ The dedicated scrubber clock source is absent since it's fully
+ synchronous to the core clock.
+
+allOf:
+ - $ref: /schemas/memory-controllers/snps,dw-umctl2-ddrc.yaml#
+
+properties:
+ compatible:
+ const: baikal,bt1-ddrc
+
+ reg:
+ maxItems: 1
+
+ interrupts:
+ maxItems: 4
+
+ interrupt-names:
+ items:
+ - const: dfi_e
+ - const: ecc_ce
+ - const: ecc_ue
+ - const: ecc_sbr
+
+ clocks:
+ maxItems: 3
+
+ clock-names:
+ items:
+ - const: pclk
+ - const: aclk
+ - const: core
+
+ resets:
+ maxItems: 2
+
+ reset-names:
+ items:
+ - const: arst
+ - const: core
+
+required:
+ - compatible
+ - reg
+ - clocks
+ - clock-names
+ - interrupts
+ - interrupt-names
+
+additionalProperties: false
+
+examples:
+ - |
+ #include <dt-bindings/interrupt-controller/mips-gic.h>
+ #include <dt-bindings/clock/bt1-ccu.h>
+ #include <dt-bindings/reset/bt1-ccu.h>
+
+ memory-controller@1f042000 {
+ compatible = "baikal,bt1-ddrc";
+ reg = <0x1f042000 0x1000>;
+
+ interrupts = <GIC_SHARED 96 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 97 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 98 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 99 IRQ_TYPE_LEVEL_HIGH>;
+ interrupt-names = "dfi_e", "ecc_ce", "ecc_ue", "ecc_sbr";
+
+ clocks = <&ccu_sys CCU_SYS_APB_CLK>,
+ <&ccu_axi CCU_AXI_DDR_CLK>,
+ <&ccu_pll CCU_DDR_PLL>;
+ clock-names = "pclk", "aclk", "core";
+
+ resets = <&ccu_axi CCU_AXI_DDR_RST>,
+ <&ccu_sys CCU_SYS_DDR_INIT_RST>;
+ reset-names = "arst", "core";
+ };
+...
--
2.37.2
Baikal-T1 SoC is equipped with the DW uMCTl2 DDRC of v2.61a with 32-bit
DQ-bus accepting DDR2/DDR3 SDRAMs of up to 2 ranks, 1:2 HIF/SDRAM clocks
rate ratio, HIF interface burst length of 8 Full DQ-bus words, 40-bit
System/Application address width and 128-bits data width, 3 System address
regions with block size 256MB. There is SEC/DED ECC capability with Scrub
(RMW) and Scrubber features.
Since the Baikal-T1 DDR controller is capable of the ECC let's add it to
the DW uMCTL2 DDRC EDAC driver. The most of the parameters above will be
autodetected except HIF burst length and SAR block size, which will be set
by means of the Baikal-T1-specific initialization method. The controller
compatible string "baikal,bt1-ddrc" will be used to attach the driver to
the kernel device. It's chosen in accordance with the just updated
DT-bindings.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 0a00e80ddeb9..24cd8f23242a 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -1342,6 +1342,20 @@ static int zynqmp_init_plat(struct snps_edac_priv *priv)
return 0;
}
+/*
+ * bt1_init_plat - Baikal-T1-specific platform initialization.
+ * @priv: DDR memory controller private data.
+ *
+ * Return: always zero.
+ */
+static int bt1_init_plat(struct snps_edac_priv *priv)
+{
+ priv->info.hif_burst_len = SNPS_DDR_BL8;
+ priv->sys_app_map.minsize = DDR_MIN_SARSIZE;
+
+ return 0;
+}
+
/**
* snps_get_dtype - Return the controller memory width.
* @mstr: Master CSR value.
@@ -2470,6 +2484,7 @@ static int snps_mc_remove(struct platform_device *pdev)
static const struct of_device_id snps_edac_match[] = {
{ .compatible = "xlnx,zynqmp-ddrc-2.40a", .data = zynqmp_init_plat },
+ { .compatible = "baikal,bt1-ddrc", .data = bt1_init_plat },
{ .compatible = "snps,ddrc-3.80a" },
{ }
};
--
2.37.2
DW uMCTL2 DDR controller IP-core can by synthesized with an embedded
Scrubber engine. The ECC Scrubber (SBR) is a block which initiates
periodic background burst read commands to the DDRC and further towards
the DDR memory in an attempt to trigger Correctable or Uncorrectable
errors. If a Correctable error is detected the ECC Scrub feature will
execute the Read-Modify-Write (RMW) procedure in order to fix the ECC. In
case of the Uncorrectable error it will be just reported as the
corresponding IRQ event. So it's definitely very useful feature. Let's add
it to the driver then especially seeing the MCI core already has some
infrastructure for it.
First of all even though the Core clock rate is only used for the Scrub
rate calculations we need to have the Scrubber clock enabled if one is
supplied otherwise the engine won't work.
Secondly the Scrubber engine support needs to be detected. Alas there is
no any special CSR indicating whether the DW uMCTL2 DDRC IP-core has been
synthesized with one embedded. Instead we suggest to implement the
detection procedure based on the Scrubber-specific CSRs writability. So if
the SBRWDATA0 CSR is writable then the CSR exists, which means the
Scrubber is available, otherwise the capability will be considered as
absent.
Thirdly the MCI core provides two callbacks utilized for the Scrubber
tuning: set the Scrubber bandwidth in bytes, which can also be used to
disable the periodic scrubbing, and get the Scrubber bandwidth (zero if
disabled). We can implement both of them by using the Scrubber CSRs the
controller provides. In particular aside with the back-to-back periodic
reads the Scrubber provides a way to delay the next read command for the
predefined set of 512's Core/Scrubber clock cycles. It can be used to
change the Scrubber bandwidth from the DDR maximal bandwidth (no delay) to
up to (0x1FFF * 512) Core/Scrubber clock cycles (see the inline comments
for details and utilized formulae). Note the Scrubber clock must be
synchronous to the Core clock by the controller design so here we get to
use the Core clock rate for the calculations. Pleas also note if no Core
clock specified the Scrubber will still be supported, but the bandwidth
will be used directly to calculate the Scrubber reads interval. The
back-to-back reads mode in this case will be indicated by the INT_MAX
bandwidth.
Fourthly the back-to-back scrubbing most likely will cause the significant
system performance drop. The manual says that it has been added to the
controller for the initial SDRAM initialization and the fast SDRAM
scrubbing after getting out of the low-power state. In anyway it is
supposed to be enabled only for a single SDRAM pass. We get to preserve
that semantic here so the back-to-back scrubbing will be disabled in the
Scrubber Done IRQ handler.
Finally the denoted scrub-rate callbacks and the SCRUB_FLAG_HW_PROG and
SCRUB_FLAG_HW_TUN flags will set to the MCI descriptor based on the
detected Scrubber capability. So no capability - no flags and no
callbacks.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 304 +++++++++++++++++++++++++++++++++++
1 file changed, 304 insertions(+)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 4b204b2050d4..0a00e80ddeb9 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -18,6 +18,7 @@
#include <linux/seq_file.h>
#include <linux/sizes.h>
#include <linux/spinlock.h>
+#include <linux/units.h>
#include <linux/interrupt.h>
#include <linux/of.h>
#include <linux/of_device.h>
@@ -34,6 +35,7 @@
/* DDR capabilities */
#define SNPS_CAP_ECC_SCRUB BIT(0)
+#define SNPS_CAP_ECC_SCRUBBER BIT(1)
#define SNPS_CAP_ZYNQMP BIT(31)
/* Synopsys uMCTL2 DDR controller registers that are relevant to ECC */
@@ -102,6 +104,12 @@
#define DDR_SARBASE0_OFST 0xF04
#define DDR_SARSIZE0_OFST 0xF08
+/* ECC Scrubber registers */
+#define ECC_SBRCTL_OFST 0xF24
+#define ECC_SBRSTAT_OFST 0xF28
+#define ECC_SBRWDATA0_OFST 0xF2C
+#define ECC_SBRWDATA1_OFST 0xF30
+
/* DDR Master Register 0 definitions */
#define DDR_MSTR_DEV_CFG_MASK GENMASK(31, 30)
#define DDR_MSTR_DEV_X4 0x0
@@ -244,6 +252,18 @@
#define DDR_MAX_NSAR 4
#define DDR_MIN_SARSIZE SZ_256M
+/* ECC Scrubber registers definitions */
+#define ECC_SBRCTL_SCRUB_INTERVAL GENMASK(20, 8)
+#define ECC_SBRCTL_INTERVAL_STEP 512
+#define ECC_SBRCTL_INTERVAL_MIN 0
+#define ECC_SBRCTL_INTERVAL_SAFE 1
+#define ECC_SBRCTL_INTERVAL_MAX (ECC_SBRCTL_SCRUB_INTERVAL >> 8)
+#define ECC_SBRCTL_SCRUB_BURST GENMASK(6, 4)
+#define ECC_SBRCTL_SCRUB_MODE_WR BIT(2)
+#define ECC_SBRCTL_SCRUB_EN BIT(0)
+#define ECC_SBRSTAT_SCRUB_DONE BIT(1)
+#define ECC_SBRSTAT_SCRUB_BUSY BIT(0)
+
/**
* enum snps_dq_width - SDRAM DQ bus width (ECC capable).
* SNPS_DQ_32: 32-bit memory data width.
@@ -906,6 +926,47 @@ static irqreturn_t snps_dfi_irq_handler(int irq, void *dev_id)
return IRQ_HANDLED;
}
+/**
+ * snps_sbr_irq_handler - Scrubber Done interrupt handler.
+ * @irq: IRQ number.
+ * @dev_id: Device ID.
+ *
+ * It just checks whether the IRQ has been caused by the Scrubber Done event
+ * and disables the back-to-back scrubbing by falling back to the smallest
+ * delay between the Scrubber read commands.
+ *
+ * Return: IRQ_NONE, if interrupt not set or IRQ_HANDLED otherwise.
+ */
+static irqreturn_t snps_sbr_irq_handler(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct snps_edac_priv *priv = mci->pvt_info;
+ unsigned long flags;
+ u32 regval, en;
+
+ /* Make sure IRQ is caused by the Scrubber Done event */
+ regval = readl(priv->baseaddr + ECC_SBRSTAT_OFST);
+ if (!(regval & ECC_SBRSTAT_SCRUB_DONE))
+ return IRQ_NONE;
+
+ spin_lock_irqsave(&priv->lock, flags);
+
+ regval = readl(priv->baseaddr + ECC_SBRCTL_OFST);
+ en = regval & ECC_SBRCTL_SCRUB_EN;
+ writel(regval & ~en, priv->baseaddr + ECC_SBRCTL_OFST);
+
+ regval = FIELD_PREP(ECC_SBRCTL_SCRUB_INTERVAL, ECC_SBRCTL_INTERVAL_SAFE);
+ writel(regval, priv->baseaddr + ECC_SBRCTL_OFST);
+
+ writel(regval | en, priv->baseaddr + ECC_SBRCTL_OFST);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ edac_mc_printk(mci, KERN_WARNING, "Back-to-back scrubbing disabled\n");
+
+ return IRQ_HANDLED;
+}
+
/**
* snps_com_irq_handler - Interrupt IRQ signal handler.
* @irq: IRQ number.
@@ -915,6 +976,8 @@ static irqreturn_t snps_dfi_irq_handler(int irq, void *dev_id)
*/
static irqreturn_t snps_com_irq_handler(int irq, void *dev_id)
{
+ struct mem_ctl_info *mci = dev_id;
+ struct snps_edac_priv *priv = mci->pvt_info;
irqreturn_t rc = IRQ_NONE;
rc |= snps_ce_irq_handler(irq, dev_id);
@@ -923,6 +986,9 @@ static irqreturn_t snps_com_irq_handler(int irq, void *dev_id)
rc |= snps_dfi_irq_handler(irq, dev_id);
+ if (priv->info.caps & SNPS_CAP_ECC_SCRUBBER)
+ rc |= snps_sbr_irq_handler(irq, dev_id);
+
return rc;
}
@@ -972,6 +1038,205 @@ static void snps_disable_irq(struct snps_edac_priv *priv)
spin_unlock_irqrestore(&priv->lock, flags);
}
+/**
+ * snps_get_sdram_bw - Get SDRAM bandwidth.
+ * @priv: DDR memory controller private instance data.
+ *
+ * The SDRAM interface bandwidth is calculated based on the DDRC Core clock rate
+ * and the DW uMCTL2 IP-core parameters like DQ-bus width and mode and
+ * Core/SDRAM clocks frequency ratio. Note it returns the theoretical bandwidth
+ * which in reality is hardly possible to reach.
+ *
+ * Return: SDRAM bandwidth or zero if no Core clock specified.
+ */
+static u64 snps_get_sdram_bw(struct snps_edac_priv *priv)
+{
+ unsigned long rate;
+
+ /*
+ * Depending on the ratio mode the SDRAM clock either matches the Core
+ * clock or runs with the twice its frequency.
+ */
+ rate = clk_get_rate(priv->clks[SNPS_CORE_CLK].clk);
+ rate *= priv->info.freq_ratio;
+
+ /*
+ * Scale up by 2 since it's DDR (Double Data Rate) and subtract the
+ * DQ-mode since in non-Full mode only a part of the DQ-bus is utilised
+ * on each SDRAM clock edge.
+ */
+ return (2U << (priv->info.dq_width - priv->info.dq_mode)) * (u64)rate;
+}
+
+/**
+ * snps_get_scrub_bw - Get Scrubber bandwidth.
+ * @priv: DDR memory controller private instance data.
+ * @interval: Scrub interval.
+ *
+ * DW uMCTL2 DDRC Scrubber performs periodical progressive burst reads (RMW if
+ * ECC CE is detected) commands from the whole memory space. The read commands
+ * can be delayed by means of the SBRCTL.scrub_interval field. The Scrubber
+ * cycles look as follows:
+ *
+ * |---HIF-burst-read---|-------delay-------|-HIF-burst-read-| etc
+ *
+ * Tb = Bl*[DQ]/Bw[RAM] Td = 512*interval/Fc - periods of the stages, where
+ * Bl - HIF burst length, [DQ] - Full DQ-bus width, Bw[RAM] - SDRAM bandwidth,
+ * Fc - Core clock frequency (Scrubber and Core clocks are synchronous).
+ *
+ * After some simple calculations the expressions above can be used to get the
+ * next Scrubber bandwidth formulae:
+ *
+ * Bw[Sbr] = Bw[RAM] / (1 + F * interval), where
+ * F = 2 * 512 * Fr * Fc * [DQ]e - interval scale factor with
+ * Fr - HIF/SDRAM clock frequency ratio (1 or 2), [DQ]e - DQ-bus width mode.
+ *
+ * Return: Scrubber bandwidth or zero if no Core clock specified.
+ */
+static u64 snps_get_scrub_bw(struct snps_edac_priv *priv, u32 interval)
+{
+ unsigned long fac;
+ u64 bw_ram;
+
+ fac = (2 * ECC_SBRCTL_INTERVAL_STEP * priv->info.freq_ratio) /
+ (priv->info.hif_burst_len * (1UL << priv->info.dq_mode));
+
+ bw_ram = snps_get_sdram_bw(priv);
+
+ do_div(bw_ram, 1 + fac * interval);
+
+ return bw_ram;
+}
+
+/**
+ * snps_get_scrub_interval - Get Scrubber delay interval.
+ * @priv: DDR memory controller private instance data.
+ * @bw: Scrubber bandwidth.
+ *
+ * Similarly to the Scrubber bandwidth the interval formulae can be inferred
+ * from the same expressions:
+ *
+ * interval = (Bw[RAM] - Bw[Sbr]) / (F * Bw[Sbr])
+ *
+ * Return: Scrubber delay interval or zero if no Core clock specified.
+ */
+static u32 snps_get_scrub_interval(struct snps_edac_priv *priv, u32 bw)
+{
+ unsigned long fac;
+ u64 bw_ram;
+
+ fac = (2 * priv->info.freq_ratio * ECC_SBRCTL_INTERVAL_STEP) /
+ (priv->info.hif_burst_len * (1UL << priv->info.dq_mode));
+
+ bw_ram = snps_get_sdram_bw(priv);
+
+ /* Divide twice so not to cause the integer overflow in (fac * bw) */
+ bw_ram -= bw;
+ do_div(bw_ram, bw);
+ do_div(bw_ram, fac);
+
+ return bw_ram;
+}
+
+/**
+ * snps_set_sdram_scrub_rate - Set the Scrubber bandwidth.
+ * @mci: EDAC memory controller instance.
+ * @bw: Bandwidth.
+ *
+ * It calculates the delay between the Scrubber read commands based on the
+ * specified bandwidth and the Core clock rate. If the Core clock is unavailable
+ * the passed bandwidth will be directly used as the interval value.
+ *
+ * Note the method warns about the back-to-back scrubbing since it may
+ * significantly degrade the system performance. This mode is supposed to be
+ * used for a single SDRAM scrubbing pass only. So it will be turned off in the
+ * Scrubber Done IRQ handler.
+ *
+ * Return: Actually set bandwidth (interval-based approximated bandwidth if the
+ * Core clock is unavailable) or zero if the Scrubber was disabled.
+ */
+static int snps_set_sdram_scrub_rate(struct mem_ctl_info *mci, u32 bw)
+{
+ struct snps_edac_priv *priv = mci->pvt_info;
+ u32 regval, interval;
+ unsigned long flags;
+ u64 bw_min, bw_max;
+
+ /* Don't bother with the calculations just disable and return. */
+ if (!bw) {
+ spin_lock_irqsave(&priv->lock, flags);
+
+ regval = readl(priv->baseaddr + ECC_SBRCTL_OFST);
+ regval &= ~ECC_SBRCTL_SCRUB_EN;
+ writel(regval, priv->baseaddr + ECC_SBRCTL_OFST);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ return 0;
+ }
+
+ /* If no Core clock specified fallback to the direct interval setup. */
+ bw_max = snps_get_scrub_bw(priv, ECC_SBRCTL_INTERVAL_MIN);
+ if (bw_max) {
+ bw_min = snps_get_scrub_bw(priv, ECC_SBRCTL_INTERVAL_MAX);
+ bw = clamp_t(u64, bw, bw_min, bw_max);
+
+ interval = snps_get_scrub_interval(priv, bw);
+ } else {
+ bw = clamp_val(bw, ECC_SBRCTL_INTERVAL_MIN, ECC_SBRCTL_INTERVAL_MAX);
+
+ interval = ECC_SBRCTL_INTERVAL_MAX - bw;
+ }
+
+ /*
+ * SBRCTL.scrub_en bitfield must be accessed separately from the other
+ * CSR bitfields. It means the flag must be set/clear with no updates
+ * to the rest of the fields.
+ */
+ spin_lock_irqsave(&priv->lock, flags);
+
+ regval = FIELD_PREP(ECC_SBRCTL_SCRUB_INTERVAL, interval);
+ writel(regval, priv->baseaddr + ECC_SBRCTL_OFST);
+
+ writel(regval | ECC_SBRCTL_SCRUB_EN, priv->baseaddr + ECC_SBRCTL_OFST);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ if (!interval)
+ edac_mc_printk(mci, KERN_WARNING, "Back-to-back scrubbing enabled\n");
+
+ if (!bw_max)
+ return interval ? bw : INT_MAX;
+
+ return snps_get_scrub_bw(priv, interval);
+}
+
+/**
+ * snps_get_sdram_scrub_rate - Get the Scrubber bandwidth.
+ * @mci: EDAC memory controller instance.
+ *
+ * Return: Scrubber bandwidth (interval-based approximated bandwidth if the
+ * Core clock is unavailable) or zero if the Scrubber was disabled.
+ */
+static int snps_get_sdram_scrub_rate(struct mem_ctl_info *mci)
+{
+ struct snps_edac_priv *priv = mci->pvt_info;
+ u32 regval;
+ u64 bw;
+
+ regval = readl(priv->baseaddr + ECC_SBRCTL_OFST);
+ if (!(regval & ECC_SBRCTL_SCRUB_EN))
+ return 0;
+
+ regval = FIELD_GET(ECC_SBRCTL_SCRUB_INTERVAL, regval);
+
+ bw = snps_get_scrub_bw(priv, regval);
+ if (!bw)
+ return regval ? ECC_SBRCTL_INTERVAL_MAX - regval : INT_MAX;
+
+ return bw;
+}
+
/**
* snps_create_data - Create private data.
* @pdev: platform device.
@@ -1038,7 +1303,18 @@ static int snps_get_res(struct snps_edac_priv *priv)
return rc;
}
+ rc = clk_prepare_enable(priv->clks[SNPS_SBR_CLK].clk);
+ if (rc) {
+ edac_printk(KERN_INFO, EDAC_MC, "Couldn't enable Scrubber clock\n");
+ goto err_disable_pclk;
+ }
+
return 0;
+
+err_disable_pclk:
+ clk_disable_unprepare(priv->clks[SNPS_CSR_CLK].clk);
+
+ return rc;
}
/**
@@ -1047,6 +1323,8 @@ static int snps_get_res(struct snps_edac_priv *priv)
*/
static void snps_put_res(struct snps_edac_priv *priv)
{
+ clk_disable_unprepare(priv->clks[SNPS_SBR_CLK].clk);
+
clk_disable_unprepare(priv->clks[SNPS_CSR_CLK].clk);
}
@@ -1147,6 +1425,14 @@ static int snps_get_ddrc_info(struct snps_edac_priv *priv)
if (!(regval & ECC_CFG0_DIS_SCRUB))
priv->info.caps |= SNPS_CAP_ECC_SCRUB;
+ /* Auto-detect the scrubber by writing to the SBRWDATA0 CSR */
+ regval = readl(priv->baseaddr + ECC_SBRWDATA0_OFST);
+ writel(~regval, priv->baseaddr + ECC_SBRWDATA0_OFST);
+ if (regval != readl(priv->baseaddr + ECC_SBRWDATA0_OFST)) {
+ priv->info.caps |= SNPS_CAP_ECC_SCRUBBER;
+ writel(regval, priv->baseaddr + ECC_SBRWDATA0_OFST);
+ }
+
/* Auto-detect the basic HIF/SDRAM bus parameters */
regval = readl(priv->baseaddr + DDR_MSTR_OFST);
@@ -1633,6 +1919,12 @@ static struct mem_ctl_info *snps_mc_create(struct snps_edac_priv *priv)
mci->scrub_cap = SCRUB_FLAG_SW_SRC;
}
+ if (priv->info.caps & SNPS_CAP_ECC_SCRUBBER) {
+ mci->scrub_cap |= SCRUB_FLAG_HW_PROG | SCRUB_FLAG_HW_TUN;
+ mci->set_sdram_scrub_rate = snps_set_sdram_scrub_rate;
+ mci->get_sdram_scrub_rate = snps_get_sdram_scrub_rate;
+ }
+
mci->ctl_name = "snps_umctl2_ddrc";
mci->dev_name = SNPS_EDAC_MOD_STRING;
mci->mod_name = SNPS_EDAC_MOD_VER;
@@ -1707,6 +1999,16 @@ static int snps_request_ind_irq(struct mem_ctl_info *mci)
}
}
+ irq = platform_get_irq_byname_optional(priv->pdev, "ecc_sbr");
+ if (irq > 0) {
+ rc = devm_request_irq(dev, irq, snps_sbr_irq_handler, 0, "ecc_sbr", mci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MC, "Failed to request Sbr IRQ\n");
+ return rc;
+ }
+ }
+
+
return 0;
}
@@ -1813,6 +2115,8 @@ static int snps_ddrc_info_show(struct seq_file *s, void *data)
if (priv->info.caps) {
if (priv->info.caps & SNPS_CAP_ECC_SCRUB)
seq_puts(s, " +Scrub");
+ if (priv->info.caps & SNPS_CAP_ECC_SCRUBBER)
+ seq_puts(s, " +Scrubber");
if (priv->info.caps & SNPS_CAP_ZYNQMP)
seq_puts(s, " +ZynqMP");
} else {
--
2.37.2
DW uMCTL2 DDRC ECC has a so called ECC Scrub feature in case if an
single-bit error is detected. The scrub is executed as a new RMW operation
to the location that resulted in a single-bit error thus fixing the ECC
code preserved in the SDRAM. But that feature not only optional, but also
runtime switchable. So there can be platforms with DW uMCTL2 DDRC not
supporting hardware-base scrub. In those cases the single-bit errors will
still be detected but won't be fixed until the next SDRAM write commands
to the erroneous location. Since the ECC Scrub feature availability is
detectable by means of the ECCCFG0.dis_scrub flag state we can use it to
tune the MCI core up so one would automatically execute the
platform-specific the platform-specific scrubbing to the affected SDRAM
location. It's now possible to be done since the DW uMCTL2 DDRC driver
supports the actual system address reported to the MCI core. The only
thing left to do is to auto-detect the ECC Scrub feature availability and
set the mem_ctl.info.scrub_mode mode with SCRUB_SW_SRC if the feature is
unavailable. The rest will be done by the MCI core when the single-bit
errors happen.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 18 ++++++++++++++++--
1 file changed, 16 insertions(+), 2 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index d5d82531719b..b36f7ec4a79e 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -32,6 +32,7 @@
#define SNPS_EDAC_MOD_VER "1"
/* DDR capabilities */
+#define SNPS_CAP_ECC_SCRUB BIT(0)
#define SNPS_CAP_ZYNQMP BIT(31)
/* Synopsys uMCTL2 DDR controller registers that are relevant to ECC */
@@ -114,6 +115,7 @@
#define DDR_MSTR_MEM_LPDDR4 BIT(5)
/* ECC CFG0 register definitions */
+#define ECC_CFG0_DIS_SCRUB BIT(4)
#define ECC_CFG0_MODE_MASK GENMASK(2, 0)
/* ECC status register definitions */
@@ -1008,6 +1010,10 @@ static int snps_get_ddrc_info(struct snps_edac_priv *priv)
return -ENXIO;
}
+ /* Assume HW-src scrub is always available if it isn't disabled */
+ if (!(regval & ECC_CFG0_DIS_SCRUB))
+ priv->info.caps |= SNPS_CAP_ECC_SCRUB;
+
/* Auto-detect the basic HIF/SDRAM bus parameters */
regval = readl(priv->baseaddr + DDR_MSTR_OFST);
@@ -1484,8 +1490,14 @@ static struct mem_ctl_info *snps_mc_create(struct snps_edac_priv *priv)
MEM_FLAG_DDR3 | MEM_FLAG_LPDDR3 |
MEM_FLAG_DDR4 | MEM_FLAG_LPDDR4;
mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
- mci->scrub_cap = SCRUB_FLAG_HW_SRC;
- mci->scrub_mode = SCRUB_NONE;
+
+ if (priv->info.caps & SNPS_CAP_ECC_SCRUB) {
+ mci->scrub_mode = SCRUB_HW_SRC;
+ mci->scrub_cap = SCRUB_FLAG_HW_SRC;
+ } else {
+ mci->scrub_mode = SCRUB_SW_SRC;
+ mci->scrub_cap = SCRUB_FLAG_SW_SRC;
+ }
mci->edac_cap = EDAC_FLAG_SECDED;
mci->ctl_name = "snps_umctl2_ddrc";
@@ -1578,6 +1590,8 @@ static int snps_ddrc_info_show(struct seq_file *s, void *data)
seq_puts(s, "Caps:");
if (priv->info.caps) {
+ if (priv->info.caps & SNPS_CAP_ECC_SCRUB)
+ seq_puts(s, " +Scrub");
if (priv->info.caps & SNPS_CAP_ZYNQMP)
seq_puts(s, " +ZynqMP");
} else {
--
2.37.2
DW uMCTL2 DDRC EDAC driver is no longer specific to particular DDRC
versions. It's generic in the most of the aspects now. So set its kernel
config independently from the ZynqMP/IntelFPAG/MXC platforms.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/Kconfig | 1 -
1 file changed, 1 deletion(-)
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 98bcdadf4143..6aa59a0bacf1 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -486,7 +486,6 @@ config EDAC_ARMADA_XP
config EDAC_SYNOPSYS
tristate "Synopsys DDR Memory Controller"
- depends on ARCH_ZYNQMP || ARCH_INTEL_SOCFPGA || ARCH_MXC
help
Support for error detection and correction on the Synopsys DDR
memory controller.
--
2.37.2
Currently the DW uMCTL2 DDRC EDAC driver supports a common unnamed IRQ
only. It isn't suitable for our platform which has the individual IRQ
lines for each DDRC event (ECC UE, ECC CE, DFI parity error, Scrubber
done, etc). Moreover the DW uMCTL2 DDRC IP-core doesn't have an option to
be configured with a common interrupts output line. So in order to have
the generic DW uMCTL2 DDR controller and our platform supported by the
driver we need to add the individual, per DDRC event, IRQs request
support. There is not much to do really since the common IRQs handler has
already been split up into the sub-handlers. So the only thing we need to
do is first try to request the individual IRQs, if failed then fallback to
the common IRQ. The IRQ names are used in accordance with the DW uMCTL2
DDRC DT-bindings.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 91 ++++++++++++++++++++++++++++++++----
1 file changed, 81 insertions(+), 10 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index e5359ff2ed25..6a47e53deab5 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -1524,25 +1524,96 @@ static void snps_mc_free(struct mem_ctl_info *mci)
edac_mc_free(mci);
}
-static int snps_setup_irq(struct mem_ctl_info *mci)
+/**
+ * snps_request_ind_irq - Request individual DDRC IRQs.
+ * @mci: EDAC memory controller instance.
+ *
+ * Return: 0 if the IRQs were successfully requested, 1 if the individual IRQs
+ * are unavailable, otherwise negative errno.
+ */
+static int snps_request_ind_irq(struct mem_ctl_info *mci)
{
struct snps_edac_priv *priv = mci->pvt_info;
- int ret, irq;
+ struct device *dev = &priv->pdev->dev;
+ int rc, irq;
- irq = platform_get_irq(priv->pdev, 0);
- if (irq < 0) {
- edac_printk(KERN_ERR, EDAC_MC,
- "No IRQ %d in DT\n", irq);
+ irq = platform_get_irq_byname_optional(priv->pdev, "ecc_ce");
+ if (irq == -ENXIO)
+ return 1;
+ if (irq < 0)
+ return irq;
+
+ rc = devm_request_irq(dev, irq, snps_ce_irq_handler, 0, "ecc_ce", mci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MC, "Failed to request ECC CE IRQ\n");
+ return rc;
+ }
+
+ irq = platform_get_irq_byname(priv->pdev, "ecc_ue");
+ if (irq < 0)
return irq;
+
+ rc = devm_request_irq(dev, irq, snps_ue_irq_handler, 0, "ecc_ue", mci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MC, "Failed to request ECC UE IRQ\n");
+ return rc;
+ }
+
+ return 0;
+}
+
+/**
+ * snps_request_com_irq - Request common DDRC IRQ.
+ * @mci: EDAC memory controller instance.
+ *
+ * It first attempts to get the named IRQ. If failed the method fallbacks
+ * to first available one.
+ *
+ * Return: 0 if the IRQ was successfully requested otherwise negative errno.
+ */
+static int snps_request_com_irq(struct mem_ctl_info *mci)
+{
+ struct snps_edac_priv *priv = mci->pvt_info;
+ struct device *dev = &priv->pdev->dev;
+ int rc, irq;
+
+ irq = platform_get_irq_byname_optional(priv->pdev, "ecc");
+ if (irq < 0) {
+ irq = platform_get_irq(priv->pdev, 0);
+ if (irq < 0)
+ return irq;
}
- ret = devm_request_irq(&priv->pdev->dev, irq, snps_com_irq_handler,
- 0, dev_name(&priv->pdev->dev), mci);
- if (ret < 0) {
+ rc = devm_request_irq(dev, irq, snps_com_irq_handler, 0, "ecc", mci);
+ if (rc) {
edac_printk(KERN_ERR, EDAC_MC, "Failed to request IRQ\n");
- return ret;
+ return rc;
}
+ return 0;
+}
+
+/**
+ * snps_setup_irq - Request and enable DDRC IRQs.
+ * @mci: EDAC memory controller instance.
+ *
+ * It first tries to get and request individual IRQs. If failed the method
+ * fallbacks to the common IRQ line case. The IRQs will be enabled only if
+ * some of these requests have been successful.
+ *
+ * Return: 0 if IRQs were successfully setup otherwise negative errno.
+ */
+static int snps_setup_irq(struct mem_ctl_info *mci)
+{
+ struct snps_edac_priv *priv = mci->pvt_info;
+ int rc;
+
+ rc = snps_request_ind_irq(mci);
+ if (rc > 0)
+ rc = snps_request_com_irq(mci);
+ if (rc)
+ return rc;
+
snps_enable_irq(priv);
return 0;
--
2.37.2
Since the driver now has the generic Sys/SDRAM address translation
interface there is no need in preserving the poisonous address in the
driver private data especially seeing it is used in the framework of the
DebugFS node anyway. So let's drop the snps_edac_priv.poison_addr field
and just perform Sys/SDRAM back and forth address translation right in
place of the "inject_data_error" node accessors.
It causes a bit more modifications than a simple field removal. Since the
poisonous address is not preserved now there is no point in having the
snps_data_poison_setup() method so its content can be moved right into the
"inject_data_error" write operation. For the same reason there is no point
in printing the ECCPOISONADDR{0,1} registers content in the
"inject_data_error" read operation. Since the CSRs content is now parsed
anyway let's print the SDRAM address instead.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 68 +++++++++++++++++-------------------
1 file changed, 32 insertions(+), 36 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index b36f7ec4a79e..8771d69f6c5a 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -407,7 +407,6 @@ struct snps_ecc_status {
* @lock: Concurrent CSRs access lock.
* @message: Buffer for framing the event specific info.
* @stat: ECC status information.
- * @poison_addr: Data poison address.
*/
struct snps_edac_priv {
struct snps_ddrc_info info;
@@ -418,9 +417,6 @@ struct snps_edac_priv {
spinlock_t lock;
char message[SNPS_EDAC_MSG_SIZE];
struct snps_ecc_status stat;
-#ifdef CONFIG_EDAC_DEBUG
- ulong poison_addr;
-#endif
};
/**
@@ -1713,44 +1709,32 @@ static int snps_hif_sdram_map_show(struct seq_file *s, void *data)
DEFINE_SHOW_ATTRIBUTE(snps_hif_sdram_map);
-/**
- * snps_data_poison_setup - Update poison registers.
- * @priv: DDR memory controller private instance data.
- *
- * Update poison registers as per DDR mapping.
- * Return: none.
- */
-static void snps_data_poison_setup(struct snps_edac_priv *priv)
-{
- struct snps_sdram_addr sdram;
- u32 regval;
-
- snps_map_sys_to_sdram(priv, priv->poison_addr, &sdram);
-
- regval = FIELD_PREP(ECC_POISON0_RANK_MASK, sdram.rank) |
- FIELD_PREP(ECC_POISON0_COL_MASK, sdram.col);
- writel(regval, priv->baseaddr + ECC_POISON0_OFST);
-
- regval = FIELD_PREP(ECC_POISON1_BANKGRP_MASK, sdram.bankgrp) |
- FIELD_PREP(ECC_POISON1_BANK_MASK, sdram.bank) |
- FIELD_PREP(ECC_POISON1_ROW_MASK, sdram.row);
- writel(regval, priv->baseaddr + ECC_POISON1_OFST);
-}
-
static ssize_t snps_inject_data_error_read(struct file *filep, char __user *ubuf,
size_t size, loff_t *offp)
{
struct mem_ctl_info *mci = filep->private_data;
struct snps_edac_priv *priv = mci->pvt_info;
+ struct snps_sdram_addr sdram;
char buf[SNPS_DBGFS_BUF_LEN];
+ dma_addr_t sys;
+ u32 regval;
int pos;
- pos = scnprintf(buf, sizeof(buf), "Poison0 Addr: 0x%08x\n\r",
- readl(priv->baseaddr + ECC_POISON0_OFST));
- pos += scnprintf(buf + pos, sizeof(buf) - pos, "Poison1 Addr: 0x%08x\n\r",
- readl(priv->baseaddr + ECC_POISON1_OFST));
- pos += scnprintf(buf + pos, sizeof(buf) - pos, "Error injection Address: 0x%lx\n\r",
- priv->poison_addr);
+ regval = readl(priv->baseaddr + ECC_POISON0_OFST);
+ sdram.rank = FIELD_GET(ECC_POISON0_RANK_MASK, regval);
+ sdram.col = FIELD_GET(ECC_POISON0_COL_MASK, regval);
+
+ regval = readl(priv->baseaddr + ECC_POISON1_OFST);
+ sdram.bankgrp = FIELD_PREP(ECC_POISON1_BANKGRP_MASK, regval);
+ sdram.bank = FIELD_PREP(ECC_POISON1_BANK_MASK, regval);
+ sdram.row = FIELD_PREP(ECC_POISON1_ROW_MASK, regval);
+
+ snps_map_sdram_to_sys(priv, &sdram, &sys);
+
+ pos = scnprintf(buf, sizeof(buf),
+ "%pad: Row %hu Rank %hu Bank %hhu Bank Group %hhu Rank %hhu\n",
+ &sys, sdram.row, sdram.col, sdram.bank, sdram.bankgrp,
+ sdram.rank);
return simple_read_from_buffer(ubuf, size, offp, buf, pos);
}
@@ -1760,13 +1744,25 @@ static ssize_t snps_inject_data_error_write(struct file *filep, const char __use
{
struct mem_ctl_info *mci = filep->private_data;
struct snps_edac_priv *priv = mci->pvt_info;
+ struct snps_sdram_addr sdram;
+ u32 regval;
+ u64 sys;
int rc;
- rc = kstrtoul_from_user(ubuf, size, 0, &priv->poison_addr);
+ rc = kstrtou64_from_user(ubuf, size, 0, &sys);
if (rc)
return rc;
- snps_data_poison_setup(priv);
+ snps_map_sys_to_sdram(priv, sys, &sdram);
+
+ regval = FIELD_PREP(ECC_POISON0_RANK_MASK, sdram.rank) |
+ FIELD_PREP(ECC_POISON0_COL_MASK, sdram.col);
+ writel(regval, priv->baseaddr + ECC_POISON0_OFST);
+
+ regval = FIELD_PREP(ECC_POISON1_BANKGRP_MASK, sdram.bankgrp) |
+ FIELD_PREP(ECC_POISON1_BANK_MASK, sdram.bank) |
+ FIELD_PREP(ECC_POISON1_ROW_MASK, sdram.row);
+ writel(regval, priv->baseaddr + ECC_POISON1_OFST);
return size;
}
--
2.37.2
First of all the DW uMCTL2 DDRC IP-core supports the individual IRQ lines
for each standard event: ECC Corrected Error, ECC Uncorrected Error, ECC
Address Protection, Scrubber-Done signal, DFI Parity/CRC Error. It's
possible that the platform engineers merge them up in the IRQ controller
level. So let's add both configuration support to the DT-schema.
Secondly the DW uMCTL2 DDRC IP-core can have clock sources like APB
reference clock, AXI-ports clock, main DDRC core reference clock and
Scrubber low-power clock. In addition to that each clock domain can have a
dedicated reset signal. Let's add the properties for at least the denoted
clock sources and the corresponding reset controls.
Note the IRQs and the phandles order is deliberately not fixed since some
of the sources may be absent depending on the IP-core synthesize
parameters and the particular platform setups.
Signed-off-by: Serge Semin <[email protected]>
---
Changelog v2:
- Replace "snps,ddrc-3.80a" compatible string with "snps,dw-umctl2-ddrc"
in the example.
- Move unrelated changes in to the dedicated patches. (@Krzysztof)
- Use the IRQ macros in the example. (@Krzysztof)
---
.../snps,dw-umctl2-ddrc.yaml | 61 ++++++++++++++++++-
1 file changed, 60 insertions(+), 1 deletion(-)
diff --git a/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml b/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
index fb571d3d665d..e68c4306025a 100644
--- a/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
+++ b/Documentation/devicetree/bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml
@@ -33,11 +33,55 @@ properties:
const: xlnx,zynqmp-ddrc-2.40a
interrupts:
- maxItems: 1
+ description:
+ DW uMCTL2 DDRC IP-core provides individual IRQ signal for each event":"
+ ECC Corrected Error, ECC Uncorrected Error, ECC Address Protection,
+ Scrubber-Done signal, DFI Parity/CRC Error. Some platforms may have the
+ signals merged before they reach the IRQ controller or have some of them
+ absent in case if the corresponding feature is unavailable/disabled.
+ minItems: 1
+ maxItems: 5
+
+ interrupt-names:
+ minItems: 1
+ maxItems: 5
+ oneOf:
+ - description: Common ECC CE/UE/Scrubber/DFI Errors IRQ
+ items:
+ - const: ecc
+ - description: Individual ECC CE/UE/Scrubber/DFI Errors IRQs
+ items:
+ enum: [ ecc_ce, ecc_ue, ecc_ap, ecc_sbr, dfi_e ]
reg:
maxItems: 1
+ clocks:
+ description:
+ A standard set of the clock sources contains CSRs bus clock, AXI-ports
+ reference clock, DDRC core clock, Scrubber standalone clock
+ (synchronous to the DDRC clock).
+ minItems: 1
+ maxItems: 4
+
+ clock-names:
+ minItems: 1
+ maxItems: 4
+ items:
+ enum: [ pclk, aclk, core, sbr ]
+
+ resets:
+ description:
+ Each clock domain can have separate reset signal.
+ minItems: 1
+ maxItems: 4
+
+ reset-names:
+ minItems: 1
+ maxItems: 4
+ items:
+ enum: [ prst, arst, core, sbr ]
+
required:
- compatible
- reg
@@ -55,5 +99,20 @@ examples:
interrupt-parent = <&gic>;
interrupts = <GIC_SPI 112 IRQ_TYPE_LEVEL_HIGH>;
+ interrupt-names = "ecc";
+ };
+ - |
+ #include <dt-bindings/interrupt-controller/irq.h>
+
+ memory-controller@3d400000 {
+ compatible = "snps,dw-umctl2-ddrc";
+ reg = <0x3d400000 0x400000>;
+
+ interrupts = <147 IRQ_TYPE_LEVEL_HIGH>, <148 IRQ_TYPE_LEVEL_HIGH>,
+ <149 IRQ_TYPE_LEVEL_HIGH>, <150 IRQ_TYPE_LEVEL_HIGH>;
+ interrupt-names = "ecc_ce", "ecc_ue", "ecc_sbr", "dfi_e";
+
+ clocks = <&pclk>, <&aclk>, <&core_clk>, <&sbr_clk>;
+ clock-names = "pclk", "aclk", "core", "sbr";
};
...
--
2.37.2
Currently the driver doesn't support any clock-related resources request
and handling, fairly assuming that all of them are supposed to be enabled
anyway in order for the system to work correctly. It's true for the Core
and AXI Ports reference clocks, but the CSR (APB) and Scrubber clocks
might still be disabled in case if the system firmware doesn't imply any
other software touching the DDR controller internals. Since the DW uMCTL2
DDRC driver does access the controller registers at the very least we need
to make sure the APB clock is enabled. Let's add the reference clocks
support then. First of all the driver will request all the clocks possibly
defined for the controller (Core, AXI, APB and Scrubber). Secondly the APB
clock will be enabled/disabled only since the Scrubber is currently
unsupported by the driver, and the Core and AXI clocks feed the critical
system parts so we need to avoid touching them with a risk to de-stabilize
the system memory. Please note the clocks connection IDs have been chosen
in accordance with the DT-bindings.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 101 +++++++++++++++++++++++++++++++++--
1 file changed, 98 insertions(+), 3 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 8d8952826bce..4b204b2050d4 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -8,6 +8,7 @@
#include <linux/bitfield.h>
#include <linux/bits.h>
+#include <linux/clk.h>
#include <linux/edac.h>
#include <linux/fs.h>
#include <linux/log2.h>
@@ -301,6 +302,25 @@ enum snps_ecc_mode {
SNPS_ECC_ADVX4X8 = 5,
};
+/**
+ * enum snps_ref_clk - DW uMCTL2 DDR controller clocks.
+ * @SNPS_CSR_CLK: CSR/APB interface clock.
+ * @SNPS_AXI_CLK: AXI (AHB) Port reference clock.
+ * @SNPS_CORE_CLK: DDR controller (including DFI) clock. SDRAM clock
+ * matches runs with this freq in 1:1 ratio mode and
+ * with twice of this freq in case of 1:2 ratio mode.
+ * @SNPS_SBR_CLK: Scrubber port reference clock (synchronous to
+ * the core clock).
+ * @SNPS_MAX_NCLK: Total number of clocks.
+ */
+enum snps_ref_clk {
+ SNPS_CSR_CLK,
+ SNPS_AXI_CLK,
+ SNPS_CORE_CLK,
+ SNPS_SBR_CLK,
+ SNPS_MAX_NCLK
+};
+
/**
* struct snps_ddrc_info - DDR controller platform parameters.
* @caps: DDR controller capabilities.
@@ -405,6 +425,7 @@ struct snps_ecc_error_info {
* @pdev: Platform device.
* @baseaddr: Base address of the DDR controller.
* @lock: Concurrent CSRs access lock.
+ * @clks: Controller reference clocks.
* @message: Buffer for framing the event specific info.
*/
struct snps_edac_priv {
@@ -414,6 +435,7 @@ struct snps_edac_priv {
struct platform_device *pdev;
void __iomem *baseaddr;
spinlock_t lock;
+ struct clk_bulk_data clks[SNPS_MAX_NCLK];
char message[SNPS_EDAC_MSG_SIZE];
};
@@ -974,6 +996,60 @@ static struct snps_edac_priv *snps_create_data(struct platform_device *pdev)
return priv;
}
+/**
+ * snps_get_res - Get platform device resources.
+ * @priv: DDR memory controller private instance data.
+ *
+ * It's supposed to request all the controller resources available for the
+ * particular platform and enable all the required for the driver normal
+ * work. Note only the CSR and Scrubber clocks are supposed to be switched
+ * on/off by the driver.
+ *
+ * Return: negative errno if failed to get the resources, otherwise - zero.
+ */
+static int snps_get_res(struct snps_edac_priv *priv)
+{
+ const char * const ids[] = {
+ [SNPS_CSR_CLK] = "pclk",
+ [SNPS_AXI_CLK] = "aclk",
+ [SNPS_CORE_CLK] = "core",
+ [SNPS_SBR_CLK] = "sbr",
+ };
+ int i, rc;
+
+ for (i = 0; i < SNPS_MAX_NCLK; i++)
+ priv->clks[i].id = ids[i];
+
+ rc = devm_clk_bulk_get_optional(&priv->pdev->dev, SNPS_MAX_NCLK,
+ priv->clks);
+ if (rc) {
+ edac_printk(KERN_INFO, EDAC_MC, "Failed to get ref clocks\n");
+ return rc;
+ }
+
+ /*
+ * Don't touch the Core and AXI clocks since they are critical for the
+ * stable system functioning and are supposed to have been enabled
+ * anyway.
+ */
+ rc = clk_prepare_enable(priv->clks[SNPS_CSR_CLK].clk);
+ if (rc) {
+ edac_printk(KERN_INFO, EDAC_MC, "Couldn't enable CSR clock\n");
+ return rc;
+ }
+
+ return 0;
+}
+
+/**
+ * snps_put_res - Put platform device resources.
+ * @priv: DDR memory controller private instance data.
+ */
+static void snps_put_res(struct snps_edac_priv *priv)
+{
+ clk_disable_unprepare(priv->clks[SNPS_CSR_CLK].clk);
+}
+
/*
* zynqmp_init_plat - ZynqMP-specific platform initialization.
* @priv: DDR memory controller private data.
@@ -1707,9 +1783,17 @@ static int snps_ddrc_info_show(struct seq_file *s, void *data)
{
struct mem_ctl_info *mci = s->private;
struct snps_edac_priv *priv = mci->pvt_info;
+ unsigned long rate;
seq_printf(s, "SDRAM: %s\n", edac_mem_types[priv->info.sdram_mode]);
+ rate = clk_get_rate(priv->clks[SNPS_CORE_CLK].clk);
+ if (rate) {
+ rate = rate / HZ_PER_MHZ;
+ seq_printf(s, "Clock: Core %luMHz SDRAM %luMHz\n",
+ rate, priv->info.freq_ratio * rate);
+ }
+
seq_printf(s, "DQ bus: %u/%s\n", (BITS_PER_BYTE << priv->info.dq_width),
priv->info.dq_mode == SNPS_DQ_FULL ? "Full" :
priv->info.dq_mode == SNPS_DQ_HALF ? "Half" :
@@ -2018,15 +2102,21 @@ static int snps_mc_probe(struct platform_device *pdev)
if (IS_ERR(priv))
return PTR_ERR(priv);
- rc = snps_get_ddrc_info(priv);
+ rc = snps_get_res(priv);
if (rc)
return rc;
+ rc = snps_get_ddrc_info(priv);
+ if (rc)
+ goto put_res;
+
snps_get_addr_map(priv);
mci = snps_mc_create(priv);
- if (IS_ERR(mci))
- return PTR_ERR(mci);
+ if (IS_ERR(mci)) {
+ rc = PTR_ERR(mci);
+ goto put_res;
+ }
rc = snps_setup_irq(mci);
if (rc)
@@ -2046,6 +2136,9 @@ static int snps_mc_probe(struct platform_device *pdev)
free_edac_mc:
snps_mc_free(mci);
+put_res:
+ snps_put_res(priv);
+
return rc;
}
@@ -2066,6 +2159,8 @@ static int snps_mc_remove(struct platform_device *pdev)
snps_mc_free(mci);
+ snps_put_res(priv);
+
return 0;
}
--
2.37.2
In accordance with [1] DW uMCTL2 DDR controller can generate an IRQ in
case if an attached SDRAM detects a CRC/Parity error. That capability is
mainly applicable for the DDR4 memory which has an additional signals
PARITY/ALERT_n indicating the even SDRAM address/command parity signal and
alert if the parity turns to be not even. But in accordance with [1] at
least the SDRAM address/command parity is calculated irrespective of the
memory protocol and then sent out by means of the dfi_parity_n signal
further to the DDR PHY. So depending on the DDR protocol and the DDR PHY
implementation the CRC/Parity error can be checked at some point
independently from the DDR devices type and then signaled via the
dfi_alert_n line. In anycase it would be very much useful to catch the
event and at least warn the user about problems with the DFI/SDRAM signals
integrity.
So here we suggest to add the DFI CRC/Parity IRQs handling support. First
the IRQ line is requested by the name "dfi_e" (defined in the DT-bindings)
and register its handler in case of the platform with the individual DW
uMCTL2 DDRC IRQs. If individual IRQs are unavailable the common IRQ
handler will call the DFI CRC/Parity event handler. Note the handler
doesn't do much. It just checks the IRQ status, reads the number of
errors, reports the fatal error to the MCI core and clears the IRQ status.
Alas neither the erroneous SDRAM address nor the executed command are
available in this case. Secondly the DFI CRC/Parity IRQ is
enabled/disabled together with the ECC CE/UE interrupts in the controller
probe procedure. Finally the CRC/Parity capability is advertised by the
EDAC controller capabilities flags.
[1] DesignWare® Cores Enhanced Universal DDR Memory Controller (uMCTL2)
Databook, Version 3.91a, October 2020, p.131-132
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 78 ++++++++++++++++++++++++++++++++++--
1 file changed, 75 insertions(+), 3 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 6a47e53deab5..8d8952826bce 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -80,6 +80,12 @@
#define ECC_POISON0_OFST 0xB8
#define ECC_POISON1_OFST 0xBC
+/* DDR CRC/Parity register */
+#define DDR_CRCPARCTL0_OFST 0xC0
+#define DDR_CRCPARCTL1_OFST 0xC4
+#define DDR_CRCPARCTL2_OFST 0xC8
+#define DDR_CRCPARSTAT_OFST 0xCC
+
/* DDR Address map0 Registers */
#define DDR_ADDRMAP0_OFST 0x200
@@ -153,6 +159,13 @@
#define ECC_CEADDR1_BANK_MASK GENMASK(23, 16)
#define ECC_CEADDR1_COL_MASK GENMASK(11, 0)
+/* DDR CRC/Parity register definitions */
+#define DDR_CRCPARCTL0_CLR_ALRT_ERRCNT BIT(2)
+#define DDR_CRCPARCTL0_CLR_ALRT_ERR BIT(1)
+#define DDR_CRCPARCTL0_EN_ALRT_IRQ BIT(0)
+#define DDR_CRCPARSTAT_ALRT_ERR BIT(16)
+#define DDR_CRCPARSTAT_ALRT_CNT_MASK GENMASK(15, 0)
+
/* ECC Poison register shifts */
#define ECC_POISON0_RANK_MASK GENMASK(27, 24)
#define ECC_POISON0_COL_MASK GENMASK(11, 0)
@@ -829,6 +842,48 @@ static irqreturn_t snps_ue_irq_handler(int irq, void *dev_id)
return IRQ_HANDLED;
}
+/**
+ * snps_dfi_irq_handler - DFI CRC/Parity error interrupt handler.
+ * @irq: IRQ number.
+ * @dev_id: Device ID.
+ *
+ * Return: IRQ_NONE, if interrupt not set or IRQ_HANDLED otherwise.
+ */
+static irqreturn_t snps_dfi_irq_handler(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct snps_edac_priv *priv = mci->pvt_info;
+ unsigned long flags;
+ u32 regval;
+ u16 ecnt;
+
+ /* Make sure IRQ is caused by an DFI alert error */
+ regval = readl(priv->baseaddr + DDR_CRCPARSTAT_OFST);
+ if (!(regval & DDR_CRCPARSTAT_ALRT_ERR))
+ return IRQ_NONE;
+
+ /* Just a number of CRC/Parity errors is available */
+ ecnt = FIELD_GET(DDR_CRCPARSTAT_ALRT_CNT_MASK, regval);
+
+ /* Report the detected errors with just the custom message */
+ snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
+ "DFI CRC/Parity error detected on dfi_alert_n");
+
+ edac_mc_handle_error(HW_EVENT_ERR_FATAL, mci, ecnt,
+ 0, 0, 0, 0, 0, -1, priv->message, "");
+
+ /* Make sure the DFI alert IRQ status is cleared */
+ spin_lock_irqsave(&priv->lock, flags);
+
+ regval = readl(priv->baseaddr + DDR_CRCPARCTL0_OFST) |
+ DDR_CRCPARCTL0_CLR_ALRT_ERR | DDR_CRCPARCTL0_CLR_ALRT_ERRCNT;
+ writel(regval, priv->baseaddr + DDR_CRCPARCTL0_OFST);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ return IRQ_HANDLED;
+}
+
/**
* snps_com_irq_handler - Interrupt IRQ signal handler.
* @irq: IRQ number.
@@ -844,6 +899,8 @@ static irqreturn_t snps_com_irq_handler(int irq, void *dev_id)
rc |= snps_ue_irq_handler(irq, dev_id);
+ rc |= snps_dfi_irq_handler(irq, dev_id);
+
return rc;
}
@@ -859,11 +916,16 @@ static void snps_enable_irq(struct snps_edac_priv *priv)
return;
}
- /* IRQs Enable/Disable feature has been available since v3.10a */
+ /*
+ * ECC IRQs Enable/Disable feature has been available since v3.10a,
+ * while CRC/Parity interrupts control - since v2.10a.
+ */
spin_lock_irqsave(&priv->lock, flags);
writel(ECC_CTRL_EN_CE_IRQ | ECC_CTRL_EN_UE_IRQ,
priv->baseaddr + ECC_CLR_OFST);
+ writel(DDR_CRCPARCTL0_EN_ALRT_IRQ,
+ priv->baseaddr + DDR_CRCPARCTL0_OFST);
spin_unlock_irqrestore(&priv->lock, flags);
}
@@ -883,6 +945,7 @@ static void snps_disable_irq(struct snps_edac_priv *priv)
spin_lock_irqsave(&priv->lock, flags);
writel(0, priv->baseaddr + ECC_CLR_OFST);
+ writel(0, priv->baseaddr + DDR_CRCPARCTL0_OFST);
spin_unlock_irqrestore(&priv->lock, flags);
}
@@ -1483,7 +1546,8 @@ static struct mem_ctl_info *snps_mc_create(struct snps_edac_priv *priv)
mci->mtype_cap = MEM_FLAG_LPDDR | MEM_FLAG_DDR2 | MEM_FLAG_LPDDR2 |
MEM_FLAG_DDR3 | MEM_FLAG_LPDDR3 |
MEM_FLAG_DDR4 | MEM_FLAG_LPDDR4;
- mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED | EDAC_FLAG_PARITY;
+ mci->edac_cap = mci->edac_ctl_cap;
if (priv->info.caps & SNPS_CAP_ECC_SCRUB) {
mci->scrub_mode = SCRUB_HW_SRC;
@@ -1493,7 +1557,6 @@ static struct mem_ctl_info *snps_mc_create(struct snps_edac_priv *priv)
mci->scrub_cap = SCRUB_FLAG_SW_SRC;
}
- mci->edac_cap = EDAC_FLAG_SECDED;
mci->ctl_name = "snps_umctl2_ddrc";
mci->dev_name = SNPS_EDAC_MOD_STRING;
mci->mod_name = SNPS_EDAC_MOD_VER;
@@ -1559,6 +1622,15 @@ static int snps_request_ind_irq(struct mem_ctl_info *mci)
return rc;
}
+ irq = platform_get_irq_byname_optional(priv->pdev, "dfi_e");
+ if (irq > 0) {
+ rc = devm_request_irq(dev, irq, snps_dfi_irq_handler, 0, "dfi_e", mci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MC, "Failed to request DFI IRQ\n");
+ return rc;
+ }
+ }
+
return 0;
}
--
2.37.2
Even though being a pure-debug feature currently the data poison can't be
disabled once it has been initialized and enabled. Irrespective to the way
the feature has been implemented it doesn't seem right since the system
may print false ECC errors in case if the poisoned address is accessed by
the kernel or by the user-space applications. It's possible since the
poisoned address isn't reserved in any kernel mm subsystems. Even though
that doesn't seem right either at least it's tolerable since the ECC data
poison is supposed to be utilized in the framework of the EDAC driver
debugging, but having the feature unswitchable can't be justified that
easy especially seeing it's not that hard to implement.
So in order to have the ECC data poison switchable we suggest to define
three possible values acceptable by the "inject_data_poison" DebugFS node:
1. "CE" - emit correctable error (as before).
2. "UE" - emit uncorrectable error (used to be any non-"CE" value).
3. Any other value - disable data poison feature.
Note we have to redefine the macros describing the data poison-related
fields of the ECC_CFG0 register in a way so they would be used to
separately switch the feature on/off and to select the type of the ECC
error. As a result the suggest solution turns into a proper ECC_CFG0 CSRs
fields setup based on the value written to the "inject_data_poison"
DebugFS node.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 29 ++++++++++++++++++++---------
1 file changed, 20 insertions(+), 9 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 8771d69f6c5a..1df5be2af1de 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -118,6 +118,10 @@
#define ECC_CFG0_DIS_SCRUB BIT(4)
#define ECC_CFG0_MODE_MASK GENMASK(2, 0)
+/* ECC CFG1 register definitions */
+#define ECC_CFG1_POISON_BIT BIT(1)
+#define ECC_CFG1_POISON_EN BIT(0)
+
/* ECC status register definitions */
#define ECC_STAT_UE_MASK GENMASK(23, 16)
#define ECC_STAT_CE_MASK GENMASK(15, 8)
@@ -157,10 +161,6 @@
#define ECC_POISON1_BANK_MASK GENMASK(26, 24)
#define ECC_POISON1_ROW_MASK GENMASK(17, 0)
-/* DDRC ECC CE & UE poison mask */
-#define ECC_CEPOISON_MASK GENMASK(1, 0)
-#define ECC_UEPOISON_MASK BIT(0)
-
/* DDRC address mapping parameters */
#define DDR_ADDRMAP_NREGS 12
@@ -1781,10 +1781,14 @@ static ssize_t snps_inject_data_poison_read(struct file *filep, char __user *ubu
int pos;
regval = readl(priv->baseaddr + ECC_CFG1_OFST);
- errstr = FIELD_GET(ECC_CEPOISON_MASK, regval) == ECC_CEPOISON_MASK ?
- "Correctable Error" : "UnCorrectable Error";
+ if (!(regval & ECC_CFG1_POISON_EN))
+ errstr = "Off";
+ else if (regval & ECC_CFG1_POISON_BIT)
+ errstr = "CE";
+ else
+ errstr = "UE";
- pos = scnprintf(buf, sizeof(buf), "Data Poisoning: %s\n\r", errstr);
+ pos = scnprintf(buf, sizeof(buf), "%s\n", errstr);
return simple_read_from_buffer(ubuf, size, offp, buf, pos);
}
@@ -1795,6 +1799,7 @@ static ssize_t snps_inject_data_poison_write(struct file *filep, const char __us
struct mem_ctl_info *mci = filep->private_data;
struct snps_edac_priv *priv = mci->pvt_info;
char buf[SNPS_DBGFS_BUF_LEN];
+ u32 regval;
int rc;
rc = simple_write_to_buffer(buf, sizeof(buf), offp, ubuf, size);
@@ -1802,10 +1807,16 @@ static ssize_t snps_inject_data_poison_write(struct file *filep, const char __us
return rc;
writel(0, priv->baseaddr + DDR_SWCTL);
+
+ regval = readl(priv->baseaddr + ECC_CFG1_OFST);
if (strncmp(buf, "CE", 2) == 0)
- writel(ECC_CEPOISON_MASK, priv->baseaddr + ECC_CFG1_OFST);
+ regval |= ECC_CFG1_POISON_BIT | ECC_CFG1_POISON_EN;
+ else if (strncmp(buf, "UE", 2) == 0)
+ regval = (regval & ~ECC_CFG1_POISON_BIT) | ECC_CFG1_POISON_EN;
else
- writel(ECC_UEPOISON_MASK, priv->baseaddr + ECC_CFG1_OFST);
+ regval &= ~ECC_CFG1_POISON_EN;
+ writel(regval, priv->baseaddr + ECC_CFG1_OFST);
+
writel(1, priv->baseaddr + DDR_SWCTL);
return size;
--
2.37.2
DW uMCTL2 DDRC IP-core doesn't have common IRQ line. Instead it provides
individual IRQ output signals for each controller event like: corrected
error, uncorrected error, DFI parity error, address protection, scrubber
done, and so on. So the common IRQ handler implemented in the Synopsys
EDAC driver isn't device-specific but is a particular platform specific.
Obviously it won't be suitable for the generic devices which are added to
the platforms with the original individual IRQs as it has happened in our
case.
So let's split up the common IRQ handler into two ones handling ECC
corrected and uncorrected errors. It won't be that hard since both
sub-methods it calls are already logically divided into two CE/UE parts.
What we need to do is to move these parts into the dedicated methods and
redefine the local variables a bit. The new methods will be simply called
from the common IRQs handler if one is utilized on the particular
platform. Otherwise each new IRQ handler will be called on particular
interrupt request (the IRQ handlers registration will be added a bit
later). Note we now can discard the snps_ecc_status structure as unneeded
since the error data is collected and reported now within a single method.
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 272 +++++++++++++++++------------------
1 file changed, 135 insertions(+), 137 deletions(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index 1df5be2af1de..e5359ff2ed25 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -143,7 +143,6 @@
#define DDR_QOS_IRQ_STAT_OFST 0x20200
#define DDR_QOSUE_MASK BIT(2)
#define DDR_QOSCE_MASK BIT(1)
-#define ECC_CE_UE_INTR_MASK (DDR_QOSUE_MASK | DDR_QOSCE_MASK)
#define DDR_QOS_IRQ_EN_OFST 0x20208
#define DDR_QOS_IRQ_DB_OFST 0x2020C
@@ -372,31 +371,19 @@ struct snps_sdram_addr {
/**
* struct snps_ecc_error_info - ECC error log information.
* @sdram: SDRAM address.
+ * @ecnt: Number of detected errors.
* @bitpos: Bit position.
* @data: Data causing the error.
* @syndrome: Erroneous data syndrome.
*/
struct snps_ecc_error_info {
struct snps_sdram_addr sdram;
+ u16 ecnt;
u32 bitpos;
u64 data;
u32 syndrome;
};
-/**
- * struct snps_ecc_status - ECC status information to report.
- * @ce_cnt: Correctable error count.
- * @ue_cnt: Uncorrectable error count.
- * @ceinfo: Correctable error log information.
- * @ueinfo: Uncorrectable error log information.
- */
-struct snps_ecc_status {
- u32 ce_cnt;
- u32 ue_cnt;
- struct snps_ecc_error_info ceinfo;
- struct snps_ecc_error_info ueinfo;
-};
-
/**
* struct snps_edac_priv - DDR memory controller private data.
* @info: DDR controller config info.
@@ -406,7 +393,6 @@ struct snps_ecc_status {
* @baseaddr: Base address of the DDR controller.
* @lock: Concurrent CSRs access lock.
* @message: Buffer for framing the event specific info.
- * @stat: ECC status information.
*/
struct snps_edac_priv {
struct snps_ddrc_info info;
@@ -416,7 +402,6 @@ struct snps_edac_priv {
void __iomem *baseaddr;
spinlock_t lock;
char message[SNPS_EDAC_MSG_SIZE];
- struct snps_ecc_status stat;
};
/**
@@ -688,130 +673,178 @@ static inline u32 snps_get_bitpos(u32 bitnum, enum snps_dq_width dq_width)
}
/**
- * snps_get_error_info - Get the current ECC error info.
- * @priv: DDR memory controller private instance data.
+ * snps_ce_irq_handler - Corrected error interrupt handler.
+ * @irq: IRQ number.
+ * @dev_id: Device ID.
*
- * Return: one if there is no error otherwise returns zero.
+ * Return: IRQ_NONE, if interrupt not set or IRQ_HANDLED otherwise.
*/
-static int snps_get_error_info(struct snps_edac_priv *priv)
+static irqreturn_t snps_ce_irq_handler(int irq, void *dev_id)
{
- struct snps_ecc_status *p;
- u32 regval, clearval;
+ struct mem_ctl_info *mci = dev_id;
+ struct snps_edac_priv *priv = mci->pvt_info;
+ struct snps_ecc_error_info einfo;
unsigned long flags;
- void __iomem *base;
+ u32 qosval, regval;
+ dma_addr_t sys;
- base = priv->baseaddr;
- p = &priv->stat;
+ /* Make sure IRQ is caused by a corrected ECC error */
+ if (priv->info.caps & SNPS_CAP_ZYNQMP) {
+ qosval = readl(priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
+ if (!(qosval & DDR_QOSCE_MASK))
+ return IRQ_NONE;
- regval = readl(base + ECC_STAT_OFST);
- if (!regval)
- return 1;
+ qosval &= DDR_QOSCE_MASK;
+ }
- p->ceinfo.bitpos = FIELD_GET(ECC_STAT_BITNUM_MASK, regval);
+ regval = readl(priv->baseaddr + ECC_STAT_OFST);
+ if (!FIELD_GET(ECC_STAT_CE_MASK, regval))
+ return IRQ_NONE;
- regval = readl(base + ECC_ERRCNT_OFST);
- p->ce_cnt = FIELD_GET(ECC_ERRCNT_CECNT_MASK, regval);
- p->ue_cnt = FIELD_GET(ECC_ERRCNT_UECNT_MASK, regval);
- if (!p->ce_cnt)
- goto ue_err;
+ /* Read error info like bit position, SDRAM address, data, syndrome */
+ einfo.bitpos = FIELD_GET(ECC_STAT_BITNUM_MASK, regval);
+ einfo.bitpos = snps_get_bitpos(einfo.bitpos, priv->info.dq_width);
- p->ceinfo.bitpos = snps_get_bitpos(p->ceinfo.bitpos, priv->info.dq_width);
+ regval = readl(priv->baseaddr + ECC_ERRCNT_OFST);
+ einfo.ecnt = FIELD_GET(ECC_ERRCNT_CECNT_MASK, regval);
- regval = readl(base + ECC_CEADDR0_OFST);
- p->ceinfo.sdram.rank = FIELD_GET(ECC_CEADDR0_RANK_MASK, regval);
- p->ceinfo.sdram.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
+ regval = readl(priv->baseaddr + ECC_CEADDR0_OFST);
+ einfo.sdram.rank = FIELD_GET(ECC_CEADDR0_RANK_MASK, regval);
+ einfo.sdram.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
- regval = readl(base + ECC_CEADDR1_OFST);
- 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);
+ regval = readl(priv->baseaddr + ECC_CEADDR1_OFST);
+ einfo.sdram.bankgrp = FIELD_GET(ECC_CEADDR1_BANKGRP_MASK, regval);
+ einfo.sdram.bank = FIELD_GET(ECC_CEADDR1_BANK_MASK, regval);
+ einfo.sdram.col = FIELD_GET(ECC_CEADDR1_COL_MASK, regval);
- p->ceinfo.data = readl(base + ECC_CSYND0_OFST);
+ einfo.data = readl(priv->baseaddr + ECC_CSYND0_OFST);
if (priv->info.dq_width == SNPS_DQ_64)
- p->ceinfo.data |= (u64)readl(base + ECC_CSYND1_OFST) << 32;
-
- p->ceinfo.syndrome = readl(base + ECC_CSYND2_OFST);
+ einfo.data |= (u64)readl(priv->baseaddr + ECC_CSYND1_OFST) << 32;
-ue_err:
- if (!p->ue_cnt)
- goto out;
+ einfo.syndrome = readl(priv->baseaddr + ECC_CSYND2_OFST);
- regval = readl(base + ECC_UEADDR0_OFST);
- p->ueinfo.sdram.rank = FIELD_GET(ECC_CEADDR0_RANK_MASK, regval);
- p->ueinfo.sdram.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
+ /* Report the detected errors with the corresponding sys address */
+ snps_map_sdram_to_sys(priv, &einfo.sdram, &sys);
- regval = readl(base + ECC_UEADDR1_OFST);
- 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);
+ snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
+ "Row %hu Col %hu Bank %hhu Bank Group %hhu Rank %hhu Bit %d Data 0x%08llx",
+ einfo.sdram.row, einfo.sdram.col, einfo.sdram.bank,
+ einfo.sdram.bankgrp, einfo.sdram.rank,
+ einfo.bitpos, einfo.data);
- p->ueinfo.data = readl(base + ECC_UESYND0_OFST);
- if (priv->info.dq_width == SNPS_DQ_64)
- p->ueinfo.data |= (u64)readl(base + ECC_UESYND1_OFST) << 32;
-
- p->ueinfo.syndrome = readl(base + ECC_UESYND2_OFST);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, einfo.ecnt,
+ PHYS_PFN(sys), offset_in_page(sys),
+ einfo.syndrome, einfo.sdram.rank, 0, -1,
+ priv->message, "");
-out:
+ /* Make sure the CE IRQ status is cleared */
spin_lock_irqsave(&priv->lock, flags);
- clearval = readl(base + ECC_CLR_OFST) |
- ECC_CTRL_CLR_CE_ERR | ECC_CTRL_CLR_CE_ERRCNT |
- ECC_CTRL_CLR_UE_ERR | ECC_CTRL_CLR_UE_ERRCNT;
- writel(clearval, base + ECC_CLR_OFST);
+ regval = readl(priv->baseaddr + ECC_CLR_OFST) |
+ ECC_CTRL_CLR_CE_ERR | ECC_CTRL_CLR_CE_ERRCNT;
+ writel(regval, priv->baseaddr + ECC_CLR_OFST);
spin_unlock_irqrestore(&priv->lock, flags);
- return 0;
+ if (priv->info.caps & SNPS_CAP_ZYNQMP)
+ writel(qosval, priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
+
+ return IRQ_HANDLED;
}
/**
- * snps_handle_error - Handle Correctable and Uncorrectable errors.
- * @mci: EDAC memory controller instance.
- * @p: Synopsys ECC status structure.
+ * snps_ue_irq_handler - Uncorrected error interrupt handler.
+ * @irq: IRQ number.
+ * @dev_id: Device ID.
*
- * Handles ECC correctable and uncorrectable errors.
+ * Return: IRQ_NONE, if interrupt not set or IRQ_HANDLED otherwise.
*/
-static void snps_handle_error(struct mem_ctl_info *mci, struct snps_ecc_status *p)
+static irqreturn_t snps_ue_irq_handler(int irq, void *dev_id)
{
+ struct mem_ctl_info *mci = dev_id;
struct snps_edac_priv *priv = mci->pvt_info;
- struct snps_ecc_error_info *pinf;
+ struct snps_ecc_error_info einfo;
+ unsigned long flags;
+ u32 qosval, regval;
dma_addr_t sys;
- if (p->ce_cnt) {
- pinf = &p->ceinfo;
+ /* Make sure IRQ is caused by an uncorrected ECC error */
+ if (priv->info.caps & SNPS_CAP_ZYNQMP) {
+ qosval = readl(priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
+ if (!(regval & DDR_QOSUE_MASK))
+ return IRQ_NONE;
+
+ qosval &= DDR_QOSUE_MASK;
+ }
- snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
- "Row %hu Col %hu Bank %hhu Bank Group %hhu Rank %hhu Bit %d Data 0x%08llx",
- pinf->sdram.row, pinf->sdram.col, pinf->sdram.bank,
- pinf->sdram.bankgrp, pinf->sdram.rank,
- pinf->bitpos, pinf->data);
+ regval = readl(priv->baseaddr + ECC_STAT_OFST);
+ if (!FIELD_GET(ECC_STAT_UE_MASK, regval))
+ return IRQ_NONE;
- snps_map_sdram_to_sys(priv, &pinf->sdram, &sys);
+ /* Read error info like SDRAM address, data and syndrome */
+ regval = readl(priv->baseaddr + ECC_ERRCNT_OFST);
+ einfo.ecnt = FIELD_GET(ECC_ERRCNT_UECNT_MASK, regval);
- edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, p->ce_cnt,
- PHYS_PFN(sys), offset_in_page(sys),
- pinf->syndrome, pinf->sdram.rank, 0, -1,
- priv->message, "");
- }
+ regval = readl(priv->baseaddr + ECC_UEADDR0_OFST);
+ einfo.sdram.rank = FIELD_GET(ECC_CEADDR0_RANK_MASK, regval);
+ einfo.sdram.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
- if (p->ue_cnt) {
- pinf = &p->ueinfo;
+ regval = readl(priv->baseaddr + ECC_UEADDR1_OFST);
+ einfo.sdram.bankgrp = FIELD_GET(ECC_CEADDR1_BANKGRP_MASK, regval);
+ einfo.sdram.bank = FIELD_GET(ECC_CEADDR1_BANK_MASK, regval);
+ einfo.sdram.col = FIELD_GET(ECC_CEADDR1_COL_MASK, regval);
- snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
- "Row %hu Col %hu Bank %hhu Bank Group %hhu Rank %hhu Data 0x%08llx",
- pinf->sdram.row, pinf->sdram.col, pinf->sdram.bank,
- pinf->sdram.bankgrp, pinf->sdram.rank,
- pinf->data);
+ einfo.data = readl(priv->baseaddr + ECC_UESYND0_OFST);
+ if (priv->info.dq_width == SNPS_DQ_64)
+ einfo.data |= (u64)readl(priv->baseaddr + ECC_UESYND1_OFST) << 32;
- snps_map_sdram_to_sys(priv, &pinf->sdram, &sys);
+ einfo.syndrome = readl(priv->baseaddr + ECC_UESYND2_OFST);
- edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, p->ue_cnt,
- PHYS_PFN(sys), offset_in_page(sys),
- pinf->syndrome, pinf->sdram.rank, 0, -1,
- priv->message, "");
- }
+ /* Report the detected errors with the corresponding sys address */
+ snps_map_sdram_to_sys(priv, &einfo.sdram, &sys);
+
+ snprintf(priv->message, SNPS_EDAC_MSG_SIZE,
+ "Row %hu Col %hu Bank %hhu Bank Group %hhu Rank %hhu Data 0x%08llx",
+ einfo.sdram.row, einfo.sdram.col, einfo.sdram.bank,
+ einfo.sdram.bankgrp, einfo.sdram.rank,
+ einfo.data);
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, einfo.ecnt,
+ PHYS_PFN(sys), offset_in_page(sys),
+ einfo.syndrome, einfo.sdram.rank, 0, -1,
+ priv->message, "");
+
+ /* Make sure the UE IRQ status is cleared */
+ spin_lock_irqsave(&priv->lock, flags);
+
+ regval = readl(priv->baseaddr + ECC_CLR_OFST) |
+ ECC_CTRL_CLR_UE_ERR | ECC_CTRL_CLR_UE_ERRCNT;
+ writel(regval, priv->baseaddr + ECC_CLR_OFST);
- memset(p, 0, sizeof(*p));
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ if (priv->info.caps & SNPS_CAP_ZYNQMP)
+ writel(qosval, priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * snps_com_irq_handler - Interrupt IRQ signal handler.
+ * @irq: IRQ number.
+ * @dev_id: Device ID.
+ *
+ * Return: IRQ_NONE, if interrupts not set or IRQ_HANDLED otherwise.
+ */
+static irqreturn_t snps_com_irq_handler(int irq, void *dev_id)
+{
+ irqreturn_t rc = IRQ_NONE;
+
+ rc |= snps_ce_irq_handler(irq, dev_id);
+
+ rc |= snps_ue_irq_handler(irq, dev_id);
+
+ return rc;
}
static void snps_enable_irq(struct snps_edac_priv *priv)
@@ -854,41 +887,6 @@ static void snps_disable_irq(struct snps_edac_priv *priv)
spin_unlock_irqrestore(&priv->lock, flags);
}
-/**
- * snps_irq_handler - Interrupt Handler for ECC interrupts.
- * @irq: IRQ number.
- * @dev_id: Device ID.
- *
- * Return: IRQ_NONE, if interrupt not set or IRQ_HANDLED otherwise.
- */
-static irqreturn_t snps_irq_handler(int irq, void *dev_id)
-{
- struct mem_ctl_info *mci = dev_id;
- struct snps_edac_priv *priv;
- int status, regval;
-
- priv = mci->pvt_info;
-
- if (priv->info.caps & SNPS_CAP_ZYNQMP) {
- regval = readl(priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
- regval &= (DDR_QOSCE_MASK | DDR_QOSUE_MASK);
- if (!(regval & ECC_CE_UE_INTR_MASK))
- return IRQ_NONE;
- }
-
- status = snps_get_error_info(priv);
- if (status)
- return IRQ_NONE;
-
- snps_handle_error(mci, &priv->stat);
-
-
- if (priv->info.caps & SNPS_CAP_ZYNQMP)
- writel(regval, priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
-
- return IRQ_HANDLED;
-}
-
/**
* snps_create_data - Create private data.
* @pdev: platform device.
@@ -1538,7 +1536,7 @@ static int snps_setup_irq(struct mem_ctl_info *mci)
return irq;
}
- ret = devm_request_irq(&priv->pdev->dev, irq, snps_irq_handler,
+ ret = devm_request_irq(&priv->pdev->dev, irq, snps_com_irq_handler,
0, dev_name(&priv->pdev->dev), mci);
if (ret < 0) {
edac_printk(KERN_ERR, EDAC_MC, "Failed to request IRQ\n");
--
2.37.2
On Sat, 10 Sep 2022 22:56:48 +0300, Serge Semin wrote:
> Baikal-T1 DDR controller is based on the DW uMCTL2 DDRC IP-core v2.51a
> with up to DDR3 protocol capability and 32-bit data bus + 8-bit ECC. There
> are individual IRQs for each ECC and DFI events. The dedicated scrubber
> clock source is absent since it's fully synchronous to the core clock.
> In addition to that the DFI-DDR PHY CSRs can be accessed via a separate
> registers space.
>
> Signed-off-by: Serge Semin <[email protected]>
>
> ---
>
> Changelog v2:
> - Keep the alphabetically ordered compatible strings list. (@Krzysztof)
> - Fix grammar nitpicks in the patch log. (@Krzysztof)
> - Drop the PHY CSR region. (@Rob)
> - Move the device bindings to the separate DT-schema.
> ---
> .../memory-controllers/baikal,bt1-ddrc.yaml | 91 +++++++++++++++++++
> 1 file changed, 91 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml
>
My bot found errors running 'make DT_CHECKER_FLAGS=-m dt_binding_check'
on your patch (DT_CHECKER_FLAGS is new in v5.13):
yamllint warnings/errors:
dtschema/dtc warnings/errors:
./Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml: Unable to find schema file matching $id: http://devicetree.org/schemas/memory-controllers/snps,dw-umctl2-ddrc.yaml
Error: Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.example.dts:41.30-31 syntax error
FATAL ERROR: Unable to parse input tree
make[1]: *** [scripts/Makefile.lib:384: Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.example.dtb] Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [Makefile:1420: dt_binding_check] Error 2
doc reference errors (make refcheckdocs):
See https://patchwork.ozlabs.org/patch/
This check can fail if there are any dependencies. The base for a patch
series is generally the most recent rc1.
If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure 'yamllint' is installed and dt-schema is up to
date:
pip3 install dtschema --upgrade
Please check and re-submit.
On Sat, Sep 10, 2022 at 10:56:45PM +0300, Serge Semin wrote:
> Xilinx ZynqMP DDRC-based example contains the opencoded numerical literals
> in the IRQ lines definition. It doesn't seem justified since the
> corresponding platform has well defined ARM GIC interface. Let's replace
> the numbers with the corresponding macros then.
>
> Signed-off-by: Serge Semin <[email protected]>
>
> ---
>
> Changelog v2:
> - This is a new patch created on v2 based on the @Krzysztof' request not
> to have the opencoded numbers in the bindings example. (@Krzysztof)
> ---
> .../bindings/memory-controllers/snps,dw-umctl2-ddrc.yaml | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
Acked-by: Rob Herring <[email protected]>
On Sat, Sep 10, 2022 at 10:56:47PM +0300, Serge Semin wrote:
> At the current state the DW uMCTL2 DDRC DT-schema can't be used as the
> common one for all the IP-core-based devices due to the compatible string
> property constraining the list of the supported device names. In order to
> fix that we suggest to update the compatible property constraints so one
> would permit having any value aside with the generic device names. At the
> same time the generic DT-schema selection must be restricted to the
> denoted generic devices only so not to permit the generic fallback
> compatibles. Finally since the generic schema will be referenced from the
> vendor-specific DT-bindings with possibly non-standard properties defined
> it must permit having additional properties specified.
>
> Signed-off-by: Serge Semin <[email protected]>
>
> ---
>
> Note alternatively we could drop the "additionalProperties" keyword
> modification since currently there is no actual device available with the
> properties not listed in the generic DT-schema.
Normally, this has required 2 schema files. However, I think you can
do something like this:
if:
compatible:
enum:
- snps,ddrc-3.80a
- snps,dw-umctl2-ddrc
- xlnx,zynqmp-ddrc-2.40a
then:
unevaluatedProperties: false
But please make sure that actually catches undocumented properties
because unevaluateProperties under 'then' is not something I've tried.
Rob
On Sat, Sep 10, 2022 at 10:56:46PM +0300, Serge Semin wrote:
> First of all the DW uMCTL2 DDRC IP-core supports the individual IRQ lines
> for each standard event: ECC Corrected Error, ECC Uncorrected Error, ECC
> Address Protection, Scrubber-Done signal, DFI Parity/CRC Error. It's
> possible that the platform engineers merge them up in the IRQ controller
> level. So let's add both configuration support to the DT-schema.
>
> Secondly the DW uMCTL2 DDRC IP-core can have clock sources like APB
> reference clock, AXI-ports clock, main DDRC core reference clock and
> Scrubber low-power clock. In addition to that each clock domain can have a
> dedicated reset signal. Let's add the properties for at least the denoted
> clock sources and the corresponding reset controls.
>
> Note the IRQs and the phandles order is deliberately not fixed since some
> of the sources may be absent depending on the IP-core synthesize
> parameters and the particular platform setups.
>
> Signed-off-by: Serge Semin <[email protected]>
>
> ---
>
> Changelog v2:
> - Replace "snps,ddrc-3.80a" compatible string with "snps,dw-umctl2-ddrc"
> in the example.
> - Move unrelated changes in to the dedicated patches. (@Krzysztof)
> - Use the IRQ macros in the example. (@Krzysztof)
> ---
> .../snps,dw-umctl2-ddrc.yaml | 61 ++++++++++++++++++-
> 1 file changed, 60 insertions(+), 1 deletion(-)
Reviewed-by: Rob Herring <[email protected]>
On Sat, 10 Sep 2022 22:56:48 +0300, Serge Semin wrote:
> Baikal-T1 DDR controller is based on the DW uMCTL2 DDRC IP-core v2.51a
> with up to DDR3 protocol capability and 32-bit data bus + 8-bit ECC. There
> are individual IRQs for each ECC and DFI events. The dedicated scrubber
> clock source is absent since it's fully synchronous to the core clock.
> In addition to that the DFI-DDR PHY CSRs can be accessed via a separate
> registers space.
>
> Signed-off-by: Serge Semin <[email protected]>
>
> ---
>
> Changelog v2:
> - Keep the alphabetically ordered compatible strings list. (@Krzysztof)
> - Fix grammar nitpicks in the patch log. (@Krzysztof)
> - Drop the PHY CSR region. (@Rob)
> - Move the device bindings to the separate DT-schema.
> ---
> .../memory-controllers/baikal,bt1-ddrc.yaml | 91 +++++++++++++++++++
> 1 file changed, 91 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/memory-controllers/baikal,bt1-ddrc.yaml
>
Reviewed-by: Rob Herring <[email protected]>
On Sat, 10 Sep 2022 22:56:46 +0300, Serge Semin wrote:
> First of all the DW uMCTL2 DDRC IP-core supports the individual IRQ lines
> for each standard event: ECC Corrected Error, ECC Uncorrected Error, ECC
> Address Protection, Scrubber-Done signal, DFI Parity/CRC Error. It's
> possible that the platform engineers merge them up in the IRQ controller
> level. So let's add both configuration support to the DT-schema.
>
> Secondly the DW uMCTL2 DDRC IP-core can have clock sources like APB
> reference clock, AXI-ports clock, main DDRC core reference clock and
> Scrubber low-power clock. In addition to that each clock domain can have a
> dedicated reset signal. Let's add the properties for at least the denoted
> clock sources and the corresponding reset controls.
>
> [...]
Applied, thanks!
[02/15] dt-bindings: memory: snps: Extend schema with IRQs/resets/clocks props
https://git.kernel.org/krzk/linux-mem-ctrl/c/5514acb0dd030356e628cdd88b266efaa0a22315
Best regards,
--
Krzysztof Kozlowski <[email protected]>
On Sat, 10 Sep 2022 22:56:45 +0300, Serge Semin wrote:
> Xilinx ZynqMP DDRC-based example contains the opencoded numerical literals
> in the IRQ lines definition. It doesn't seem justified since the
> corresponding platform has well defined ARM GIC interface. Let's replace
> the numbers with the corresponding macros then.
>
>
Applied, thanks!
[01/15] dt-bindings: memory: snps: Replace opencoded numbers with macros
https://git.kernel.org/krzk/linux-mem-ctrl/c/fc436e55a1abdac503e5b06ef57862a1bc944275
Best regards,
--
Krzysztof Kozlowski <[email protected]>
On Mon, Sep 12, 2022 at 09:32:19AM -0500, Rob Herring wrote:
> On Sat, Sep 10, 2022 at 10:56:47PM +0300, Serge Semin wrote:
> > At the current state the DW uMCTL2 DDRC DT-schema can't be used as the
> > common one for all the IP-core-based devices due to the compatible string
> > property constraining the list of the supported device names. In order to
> > fix that we suggest to update the compatible property constraints so one
> > would permit having any value aside with the generic device names. At the
> > same time the generic DT-schema selection must be restricted to the
> > denoted generic devices only so not to permit the generic fallback
> > compatibles. Finally since the generic schema will be referenced from the
> > vendor-specific DT-bindings with possibly non-standard properties defined
> > it must permit having additional properties specified.
> >
> > Signed-off-by: Serge Semin <[email protected]>
> >
> > ---
> >
> > Note alternatively we could drop the "additionalProperties" keyword
> > modification since currently there is no actual device available with the
> > properties not listed in the generic DT-schema.
>
> Normally, this has required 2 schema files. However, I think you can
> do something like this:
>
> if:
> compatible:
> enum:
> - snps,ddrc-3.80a
> - snps,dw-umctl2-ddrc
> - xlnx,zynqmp-ddrc-2.40a
> then:
> unevaluatedProperties: false
>
>
> But please make sure that actually catches undocumented properties
> because unevaluateProperties under 'then' is not something I've tried.
Oh, I wish this would work! Alas it doesn't. AFAIU the schemas under
the "then" and "else" keywords are considered as separate schemas
and are independently applied to the DT node. As soon as I added the
construction suggested by you the schema evaluation started failing
with error as none of the DT-node properties in the examples are valid:
< ... /snps,dw-umctl2-ddrc.example.dtb: memory-controller@fd070000:
< Unevaluated properties are not allowed ('compatible', 'reg', interrupts', 'interrupt-names', '$nodename' were unexpected)
< ... /snps,dw-umctl2-ddrc.example.dtb: memory-controller@3d400000:
< Unevaluated properties are not allowed ('compatible', 'reg', 'interrupts', 'interrupt-names', 'clocks', 'clock-names', '$nodename' were unexpected)
Any suggestion of how this could be fixed? Perhaps updating the
dtschema tool anyhow? (I failed to find a quick-fix for it) Creating
an additional separate schema with the common properties seems a bit
overkill in this case. On the other hand is there a decent
alternative?
What about accepting what I suggested in this patch? It does permit
additional properties, but we won't need to have a separate schema
with just several common properties.
-Sergey
>
> Rob
On Mon, Sep 26, 2022 at 5:56 AM Serge Semin <[email protected]> wrote:
>
> On Mon, Sep 12, 2022 at 09:32:19AM -0500, Rob Herring wrote:
> > On Sat, Sep 10, 2022 at 10:56:47PM +0300, Serge Semin wrote:
> > > At the current state the DW uMCTL2 DDRC DT-schema can't be used as the
> > > common one for all the IP-core-based devices due to the compatible string
> > > property constraining the list of the supported device names. In order to
> > > fix that we suggest to update the compatible property constraints so one
> > > would permit having any value aside with the generic device names. At the
> > > same time the generic DT-schema selection must be restricted to the
> > > denoted generic devices only so not to permit the generic fallback
> > > compatibles. Finally since the generic schema will be referenced from the
> > > vendor-specific DT-bindings with possibly non-standard properties defined
> > > it must permit having additional properties specified.
> > >
> > > Signed-off-by: Serge Semin <[email protected]>
> > >
> > > ---
> > >
> > > Note alternatively we could drop the "additionalProperties" keyword
> > > modification since currently there is no actual device available with the
> > > properties not listed in the generic DT-schema.
> >
>
> > Normally, this has required 2 schema files. However, I think you can
> > do something like this:
> >
> > if:
> > compatible:
> > enum:
> > - snps,ddrc-3.80a
> > - snps,dw-umctl2-ddrc
> > - xlnx,zynqmp-ddrc-2.40a
> > then:
> > unevaluatedProperties: false
> >
> >
> > But please make sure that actually catches undocumented properties
> > because unevaluateProperties under 'then' is not something I've tried.
>
> Oh, I wish this would work! Alas it doesn't. AFAIU the schemas under
> the "then" and "else" keywords are considered as separate schemas
> and are independently applied to the DT node. As soon as I added the
> construction suggested by you the schema evaluation started failing
> with error as none of the DT-node properties in the examples are valid:
>
> < ... /snps,dw-umctl2-ddrc.example.dtb: memory-controller@fd070000:
> < Unevaluated properties are not allowed ('compatible', 'reg', interrupts', 'interrupt-names', '$nodename' were unexpected)
>
> < ... /snps,dw-umctl2-ddrc.example.dtb: memory-controller@3d400000:
> < Unevaluated properties are not allowed ('compatible', 'reg', 'interrupts', 'interrupt-names', 'clocks', 'clock-names', '$nodename' were unexpected)
Indeed. While unevaluatedProperties takes if/then/else into account,
flipping it around doesn't.
> Any suggestion of how this could be fixed? Perhaps updating the
> dtschema tool anyhow? (I failed to find a quick-fix for it) Creating
> an additional separate schema with the common properties seems a bit
> overkill in this case. On the other hand is there a decent
> alternative?
I don't think there is any other fix.
> What about accepting what I suggested in this patch? It does permit
> additional properties, but we won't need to have a separate schema
> with just several common properties.
No. You can't have it both ways. Either it is a common schema or a
specific device schema.
Rob
On Tue, Sep 27, 2022 at 05:02:40PM -0500, Rob Herring wrote:
> On Mon, Sep 26, 2022 at 5:56 AM Serge Semin <[email protected]> wrote:
> >
> > On Mon, Sep 12, 2022 at 09:32:19AM -0500, Rob Herring wrote:
> > > On Sat, Sep 10, 2022 at 10:56:47PM +0300, Serge Semin wrote:
> > > > At the current state the DW uMCTL2 DDRC DT-schema can't be used as the
> > > > common one for all the IP-core-based devices due to the compatible string
> > > > property constraining the list of the supported device names. In order to
> > > > fix that we suggest to update the compatible property constraints so one
> > > > would permit having any value aside with the generic device names. At the
> > > > same time the generic DT-schema selection must be restricted to the
> > > > denoted generic devices only so not to permit the generic fallback
> > > > compatibles. Finally since the generic schema will be referenced from the
> > > > vendor-specific DT-bindings with possibly non-standard properties defined
> > > > it must permit having additional properties specified.
> > > >
> > > > Signed-off-by: Serge Semin <[email protected]>
> > > >
> > > > ---
> > > >
> > > > Note alternatively we could drop the "additionalProperties" keyword
> > > > modification since currently there is no actual device available with the
> > > > properties not listed in the generic DT-schema.
> > >
> >
> > > Normally, this has required 2 schema files. However, I think you can
> > > do something like this:
> > >
> > > if:
> > > compatible:
> > > enum:
> > > - snps,ddrc-3.80a
> > > - snps,dw-umctl2-ddrc
> > > - xlnx,zynqmp-ddrc-2.40a
> > > then:
> > > unevaluatedProperties: false
> > >
> > >
> > > But please make sure that actually catches undocumented properties
> > > because unevaluateProperties under 'then' is not something I've tried.
> >
> > Oh, I wish this would work! Alas it doesn't. AFAIU the schemas under
> > the "then" and "else" keywords are considered as separate schemas
> > and are independently applied to the DT node. As soon as I added the
> > construction suggested by you the schema evaluation started failing
> > with error as none of the DT-node properties in the examples are valid:
> >
> > < ... /snps,dw-umctl2-ddrc.example.dtb: memory-controller@fd070000:
> > < Unevaluated properties are not allowed ('compatible', 'reg', interrupts', 'interrupt-names', '$nodename' were unexpected)
> >
> > < ... /snps,dw-umctl2-ddrc.example.dtb: memory-controller@3d400000:
> > < Unevaluated properties are not allowed ('compatible', 'reg', 'interrupts', 'interrupt-names', 'clocks', 'clock-names', '$nodename' were unexpected)
>
> Indeed. While unevaluatedProperties takes if/then/else into account,
> flipping it around doesn't.
>
> > Any suggestion of how this could be fixed? Perhaps updating the
> > dtschema tool anyhow? (I failed to find a quick-fix for it) Creating
> > an additional separate schema with the common properties seems a bit
> > overkill in this case. On the other hand is there a decent
> > alternative?
>
> I don't think there is any other fix.
>
> > What about accepting what I suggested in this patch? It does permit
> > additional properties, but we won't need to have a separate schema
> > with just several common properties.
>
> No. You can't have it both ways. Either it is a common schema or a
> specific device schema.
Sigh... I see. Will fix it in the next patchset round.
-Sergey
>
> Rob