Skip to contentSkip to conversation

riscv: sifive_u: Improve the emulation fidelity of sifive_u machine

PATCH Superseded Bin Meng 3 participants 40 messages Aug 11, 2019

Subsystems (2)

v3 of 8 Superseded Latest is v8 →

lore
Share & tools
lore

Bin Meng

author
#0 lore raw

v3 cover riscv: sifive_u: Improve the emulation fidelity of sifive_u machine

As of today, the QEMU 'sifive_u' machine is a special target that does not boot the upstream OpenSBI/U-Boot firmware images built for the real SiFive HiFive Unleashed board. Hence OpenSBI supports a special platform "qemu/sifive_u". For U-Boot, the sifive_fu540_defconfig is referenced in the OpenSBI doc as its payload, but that does not boot at all due to various issues in current QEMU 'sifive_u' machine codes.

This series aims to improve the emulation fidelity of sifive_u machine, so that the upstream OpenSBI, U-Boot and kernel images built for the SiFive HiFive Unleashed board can be used out of the box without any special hack.

The major changes include:
- Heterogeneous harts creation supported, so that we can create a CPU that exactly mirrors the real hardware: 1 E51 + 4 U54.
- Implemented a PRCI model for FU540
- Implemented an OTP model for FU540, primarily used for storing serial number of the board
- Fixed GEM support that was seriously broken on sifive_u
- Synced device tree with upstream Linux kernel on sifive_u
- Adding initramfs loading support on sifive_u

OpenSBI v0.4 image built for sifive/fu540 is included as the default bios image for 'sifive_u' machine.

The series is tested against OpenSBI v0.4 image for sifive/fu540 paltform, U-Boot v2019.10-rc1 image for sifive_fu540_defconfig, and Linux kernel v5.3-rc3 image with the following patch:

macb: Update compatibility string for SiFive FU540-C000
https://patchwork.kernel.org/patch/11050003/

OpenSBI + U-Boot, ping/tftpboot with U-Boot MACB driver works well. Boot Linux 64-bit defconfig image, verified that system console on the serial 0 and ping host work pretty well.

An OpenSBI patch was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. It will be applied after this QEMU series is merged.
http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v3:
- changed to use macros for management and compute cpu count
- use management cpu count + 1 for the min_cpus
- update IRQ numbers of both UARTs to match hardware as well

Changes in v2:
- fixed the "interrupts-extended" property size
- update the file header to indicate at least 2 harts are created
- use create_unimplemented_device() to create the GEM management block instead of sifive_mmio_emulate()
- add "phy-handle" property to the ethernet node
- keep the PLIC compatible string unchanged as OpenSBI uses that for DT fix up
- drop patch "riscv: sifive: Move sifive_mmio_emulate() to a common place"
- new patch "riscv: sifive_e: Drop sifive_mmio_emulate()"

Bin Meng (28):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: Add a sifive_cpu.h to include both E and U cpu type defines
riscv: hart: Extract hart realize to a separate routine
riscv: hart: Support heterogeneous harts population
riscv: sifive_u: Update hart configuration to reflect the real FU540
SoC
riscv: sifive_u: Set the minimum number of cpus to 2
riscv: sifive_u: Update PLIC hart topology configuration string
riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Remove the unnecessary include of prci header
riscv: sifive: Rename sifive_prci.{c,h} to sifive_e_prci.{c,h}
riscv: sifive_e: prci: Fix a typo of hfxosccfg register programming
riscv: sifive_e: prci: Update the PRCI register block size
riscv: sifive: Implement PRCI model for FU540
riscv: sifive_u: Generate hfclk and rtcclk nodes
riscv: sifive_u: Add PRCI block to the SoC
riscv: sifive_u: Change UART node name in device tree
riscv: hw: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: roms: Update default bios for sifive_u machine
riscv: sifive_u: Update UART and ethernet node clock properties
riscv: sifive_u: Generate an aliases node in the device tree
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Support loading initramfs
riscv: hw: Remove not needed PLIC properties in device tree
riscv: sifive_e: Drop sifive_mmio_emulate()
riscv: virt: Change create_fdt() to return void
riscv: sifive_u: Update model and compatible strings in device tree

 hw/riscv/Makefile.objs                             |   4 +-
 hw/riscv/riscv_hart.c                              |  75 ++++++--
 hw/riscv/sifive_e.c                                |  27 ++-
 hw/riscv/{sifive_prci.c => sifive_e_prci.c}        |  16 +-
 hw/riscv/sifive_u.c                                | 185 ++++++++++++++------
 hw/riscv/sifive_u_otp.c                            | 194 +++++++++++++++++++++
 hw/riscv/sifive_u_prci.c                           | 163 +++++++++++++++++
 hw/riscv/virt.c                                    |  40 ++---
 include/hw/riscv/sifive_cpu.h                      |  31 ++++
 include/hw/riscv/sifive_e.h                        |   7 +-
 .../hw/riscv/{sifive_prci.h => sifive_e_prci.h}    |  16 +-
 include/hw/riscv/sifive_u.h                        |  22 +--
 include/hw/riscv/sifive_u_otp.h                    |  90 ++++++++++
 include/hw/riscv/sifive_u_prci.h                   | 100 +++++++++++
 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin       | Bin 40968 -> 45064 bytes
 roms/Makefile                                      |   4 +-
 16 files changed, 830 insertions(+), 144 deletions(-)
 rename hw/riscv/{sifive_prci.c => sifive_e_prci.c} (88%)
 create mode 100644 hw/riscv/sifive_u_otp.c
 create mode 100644 hw/riscv/sifive_u_prci.c
 create mode 100644 include/hw/riscv/sifive_cpu.h
 rename include/hw/riscv/{sifive_prci.h => sifive_e_prci.h} (80%)
 create mode 100644 include/hw/riscv/sifive_u_otp.h
 create mode 100644 include/hw/riscv/sifive_u_prci.h
Signature · 1 line
2.7.4
author
#1 lore raw

PATCH v3 1/28 riscv: hw: Remove superfluous "linux, phandle" property

"linux,phandle" property is optional. Remove all instances in the sifive_u and virt machine device tree.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 3 ---
 hw/riscv/virt.c     | 3 ---
 2 files changed, 6 deletions(-)
hw/riscv/sifive_u.c +0 -3
index 71b8083..ef36948 100644
@@ -125,7 +125,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         qemu_fdt_setprop_string(fdt, nodename, "device_type", "cpu");
         qemu_fdt_add_subnode(fdt, intc);
         qemu_fdt_setprop_cell(fdt, intc, "phandle", cpu_phandle);
-        qemu_fdt_setprop_cell(fdt, intc, "linux,phandle", cpu_phandle);
         qemu_fdt_setprop_string(fdt, intc, "compatible", "riscv,cpu-intc");
         qemu_fdt_setprop(fdt, intc, "interrupt-controller", NULL, 0);
         qemu_fdt_setprop_cell(fdt, intc, "#interrupt-cells", 1);
@@ -184,7 +183,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", 0x35);
     qemu_fdt_setprop_cells(fdt, nodename, "phandle", plic_phandle);
-    qemu_fdt_setprop_cells(fdt, nodename, "linux,phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
     g_free(cells);
     g_free(nodename);
@@ -197,7 +195,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
         SIFIVE_U_GEM_CLOCK_FREQ);
     qemu_fdt_setprop_cell(fdt, nodename, "phandle", ethclk_phandle);
-    qemu_fdt_setprop_cell(fdt, nodename, "linux,phandle", ethclk_phandle);
     ethclk_phandle = qemu_fdt_get_phandle(fdt, nodename);
     g_free(nodename);
 
hw/riscv/virt.c +0 -3
index 25faf3b..00be05a 100644
@@ -170,11 +170,9 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
         qemu_fdt_setprop_cell(fdt, nodename, "reg", cpu);
         qemu_fdt_setprop_string(fdt, nodename, "device_type", "cpu");
         qemu_fdt_setprop_cell(fdt, nodename, "phandle", cpu_phandle);
-        qemu_fdt_setprop_cell(fdt, nodename, "linux,phandle", cpu_phandle);
         intc_phandle = phandle++;
         qemu_fdt_add_subnode(fdt, intc);
         qemu_fdt_setprop_cell(fdt, intc, "phandle", intc_phandle);
-        qemu_fdt_setprop_cell(fdt, intc, "linux,phandle", intc_phandle);
         qemu_fdt_setprop_string(fdt, intc, "compatible", "riscv,cpu-intc");
         qemu_fdt_setprop(fdt, intc, "interrupt-controller", NULL, 0);
         qemu_fdt_setprop_cell(fdt, intc, "#interrupt-cells", 1);
@@ -250,7 +248,6 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", VIRTIO_NDEV);
     qemu_fdt_setprop_cells(fdt, nodename, "phandle", plic_phandle);
-    qemu_fdt_setprop_cells(fdt, nodename, "linux,phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
     g_free(cells);
     g_free(nodename);
Signature · 1 line
2.7.4
author
#2 lore raw

PATCH v3 2/28 riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell

In reply to Bin Meng

Some of the properties only have 1 cell so we should use qemu_fdt_setprop_cell() instead of qemu_fdt_setprop_cells().

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 16 ++++++++--------
 hw/riscv/virt.c     | 24 ++++++++++++------------
 2 files changed, 20 insertions(+), 20 deletions(-)
hw/riscv/sifive_u.c +8 -8
index ef36948..623ee64 100644
@@ -182,7 +182,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", 0x35);
-    qemu_fdt_setprop_cells(fdt, nodename, "phandle", plic_phandle);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
     g_free(cells);
     g_free(nodename);
@@ -207,20 +207,20 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         0x0, memmap[SIFIVE_U_GEM].size);
     qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
     qemu_fdt_setprop_string(fdt, nodename, "phy-mode", "gmii");
-    qemu_fdt_setprop_cells(fdt, nodename, "interrupt-parent", plic_phandle);
-    qemu_fdt_setprop_cells(fdt, nodename, "interrupts", SIFIVE_U_GEM_IRQ);
+    qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
+    qemu_fdt_setprop_cell(fdt, nodename, "interrupts", SIFIVE_U_GEM_IRQ);
     qemu_fdt_setprop_cells(fdt, nodename, "clocks",
         ethclk_phandle, ethclk_phandle, ethclk_phandle);
     qemu_fdt_setprop(fdt, nodename, "clocks-names", ethclk_names,
         sizeof(ethclk_names));
-    qemu_fdt_setprop_cells(fdt, nodename, "#address-cells", 1);
-    qemu_fdt_setprop_cells(fdt, nodename, "#size-cells", 0);
+    qemu_fdt_setprop_cell(fdt, nodename, "#address-cells", 1);
+    qemu_fdt_setprop_cell(fdt, nodename, "#size-cells", 0);
     g_free(nodename);
 
     nodename = g_strdup_printf("/soc/ethernet@%lx/ethernet-phy@0",
         (long)memmap[SIFIVE_U_GEM].base);
     qemu_fdt_add_subnode(fdt, nodename);
-    qemu_fdt_setprop_cells(fdt, nodename, "reg", 0x0);
+    qemu_fdt_setprop_cell(fdt, nodename, "reg", 0x0);
     g_free(nodename);
 
     nodename = g_strdup_printf("/soc/uart@%lx",
@@ -232,8 +232,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         0x0, memmap[SIFIVE_U_UART0].size);
     qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
                           SIFIVE_U_CLOCK_FREQ / 2);
-    qemu_fdt_setprop_cells(fdt, nodename, "interrupt-parent", plic_phandle);
-    qemu_fdt_setprop_cells(fdt, nodename, "interrupts", SIFIVE_U_UART0_IRQ);
+    qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
+    qemu_fdt_setprop_cell(fdt, nodename, "interrupts", SIFIVE_U_UART0_IRQ);
 
     qemu_fdt_add_subnode(fdt, "/chosen");
     qemu_fdt_setprop_string(fdt, "/chosen", "stdout-path", nodename);
hw/riscv/virt.c +12 -12
index 00be05a..127f005 100644
@@ -233,8 +233,8 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     nodename = g_strdup_printf("/soc/interrupt-controller@%lx",
         (long)memmap[VIRT_PLIC].base);
     qemu_fdt_add_subnode(fdt, nodename);
-    qemu_fdt_setprop_cells(fdt, nodename, "#address-cells",
-                           FDT_PLIC_ADDR_CELLS);
+    qemu_fdt_setprop_cell(fdt, nodename, "#address-cells",
+                          FDT_PLIC_ADDR_CELLS);
     qemu_fdt_setprop_cell(fdt, nodename, "#interrupt-cells",
                           FDT_PLIC_INT_CELLS);
     qemu_fdt_setprop_string(fdt, nodename, "compatible", "riscv,plic0");
@@ -247,7 +247,7 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", VIRTIO_NDEV);
-    qemu_fdt_setprop_cells(fdt, nodename, "phandle", plic_phandle);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
     g_free(cells);
     g_free(nodename);
@@ -260,19 +260,19 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
         qemu_fdt_setprop_cells(fdt, nodename, "reg",
             0x0, memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
             0x0, memmap[VIRT_VIRTIO].size);
-        qemu_fdt_setprop_cells(fdt, nodename, "interrupt-parent", plic_phandle);
-        qemu_fdt_setprop_cells(fdt, nodename, "interrupts", VIRTIO_IRQ + i);
+        qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
+        qemu_fdt_setprop_cell(fdt, nodename, "interrupts", VIRTIO_IRQ + i);
         g_free(nodename);
     }
 
     nodename = g_strdup_printf("/soc/pci@%lx",
         (long) memmap[VIRT_PCIE_ECAM].base);
     qemu_fdt_add_subnode(fdt, nodename);
-    qemu_fdt_setprop_cells(fdt, nodename, "#address-cells",
-                           FDT_PCI_ADDR_CELLS);
-    qemu_fdt_setprop_cells(fdt, nodename, "#interrupt-cells",
-                           FDT_PCI_INT_CELLS);
-    qemu_fdt_setprop_cells(fdt, nodename, "#size-cells", 0x2);
+    qemu_fdt_setprop_cell(fdt, nodename, "#address-cells",
+                          FDT_PCI_ADDR_CELLS);
+    qemu_fdt_setprop_cell(fdt, nodename, "#interrupt-cells",
+                          FDT_PCI_INT_CELLS);
+    qemu_fdt_setprop_cell(fdt, nodename, "#size-cells", 0x2);
     qemu_fdt_setprop_string(fdt, nodename, "compatible",
                             "pci-host-ecam-generic");
     qemu_fdt_setprop_string(fdt, nodename, "device_type", "pci");
@@ -309,8 +309,8 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
         0x0, memmap[VIRT_UART0].base,
         0x0, memmap[VIRT_UART0].size);
     qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency", 3686400);
-        qemu_fdt_setprop_cells(fdt, nodename, "interrupt-parent", plic_phandle);
-        qemu_fdt_setprop_cells(fdt, nodename, "interrupts", UART0_IRQ);
+    qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
+    qemu_fdt_setprop_cell(fdt, nodename, "interrupts", UART0_IRQ);
 
     qemu_fdt_add_subnode(fdt, "/chosen");
     qemu_fdt_setprop_string(fdt, "/chosen", "stdout-path", nodename);
Signature · 1 line
2.7.4
author
#3 lore raw

PATCH v3 3/28 riscv: Add a sifive_cpu.h to include both E and U cpu type defines

In reply to Bin Meng

Group SiFive E and U cpu type defines into one header file.

---

Changes in v3: None
Changes in v2: None

 include/hw/riscv/sifive_cpu.h | 31 +++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_e.h   |  7 +------
 include/hw/riscv/sifive_u.h   |  7 +------
 3 files changed, 33 insertions(+), 12 deletions(-)
 create mode 100644 include/hw/riscv/sifive_cpu.h
include/hw/riscv/sifive_cpu.h new file +31 -0
new file mode 100644
index 0000000..1367996
@@ -0,0 +1,31 @@
+/*
+ * SiFive CPU types
+ *
+ * Copyright (c) 2017 SiFive, Inc.
+ * Copyright (c) 2019 Bin Meng <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2 or later, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef HW_SIFIVE_CPU_H
+#define HW_SIFIVE_CPU_H
+
+#if defined(TARGET_RISCV32)
+#define SIFIVE_E_CPU TYPE_RISCV_CPU_SIFIVE_E31
+#define SIFIVE_U_CPU TYPE_RISCV_CPU_SIFIVE_U34
+#elif defined(TARGET_RISCV64)
+#define SIFIVE_E_CPU TYPE_RISCV_CPU_SIFIVE_E51
+#define SIFIVE_U_CPU TYPE_RISCV_CPU_SIFIVE_U54
+#endif
+
+#endif /* HW_SIFIVE_CPU_H */
include/hw/riscv/sifive_e.h +1 -6
index d175b24..e17cdfd 100644
@@ -19,6 +19,7 @@
 #ifndef HW_SIFIVE_E_H
 #define HW_SIFIVE_E_H
 
+#include "hw/riscv/sifive_cpu.h"
 #include "hw/riscv/sifive_gpio.h"
 
 #define TYPE_RISCV_E_SOC "riscv.sifive.e.soc"
@@ -83,10 +84,4 @@ enum {
 #define SIFIVE_E_PLIC_CONTEXT_BASE 0x200000
 #define SIFIVE_E_PLIC_CONTEXT_STRIDE 0x1000
 
-#if defined(TARGET_RISCV32)
-#define SIFIVE_E_CPU TYPE_RISCV_CPU_SIFIVE_E31
-#elif defined(TARGET_RISCV64)
-#define SIFIVE_E_CPU TYPE_RISCV_CPU_SIFIVE_E51
-#endif
-
 #endif
include/hw/riscv/sifive_u.h +1 -6
index 892f0ee..4abc621 100644
@@ -20,6 +20,7 @@
 #define HW_SIFIVE_U_H
 
 #include "hw/net/cadence_gem.h"
+#include "hw/riscv/sifive_cpu.h"
 
 #define TYPE_RISCV_U_SOC "riscv.sifive.u.soc"
 #define RISCV_U_SOC(obj) \
@@ -77,10 +78,4 @@ enum {
 #define SIFIVE_U_PLIC_CONTEXT_BASE 0x200000
 #define SIFIVE_U_PLIC_CONTEXT_STRIDE 0x1000
 
-#if defined(TARGET_RISCV32)
-#define SIFIVE_U_CPU TYPE_RISCV_CPU_SIFIVE_U34
-#elif defined(TARGET_RISCV64)
-#define SIFIVE_U_CPU TYPE_RISCV_CPU_SIFIVE_U54
-#endif
-
 #endif
Signature · 1 line
2.7.4
author
#4 lore raw

PATCH v3 4/28 riscv: hart: Extract hart realize to a separate routine

In reply to Bin Meng

Currently riscv_harts_realize() creates all harts based on the same cpu type given in the hart array property. With current implementation it can only create symmetric harts. Exact the hart realize to a separate routine in preparation for supporting heterogeneous hart arrays.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/riscv_hart.c | 31 +++++++++++++++++++------------
 1 file changed, 19 insertions(+), 12 deletions(-)
hw/riscv/riscv_hart.c +19 -12
index ca69a1b..3dd1c6a 100644
@@ -37,26 +37,33 @@ static void riscv_harts_cpu_reset(void *opaque)
     cpu_reset(CPU(cpu));
 }
 
+static void riscv_hart_realize(RISCVHartArrayState *s, int hart,
+                               char *cpu_type, Error **errp)
+{
+    Error *err = NULL;
+
+    object_initialize_child(OBJECT(s), "harts[*]", &s->harts[hart],
+                            sizeof(RISCVCPU), cpu_type,
+                            &error_abort, NULL);
+    s->harts[hart].env.mhartid = hart;
+    qemu_register_reset(riscv_harts_cpu_reset, &s->harts[hart]);
+    object_property_set_bool(OBJECT(&s->harts[hart]), true,
+                             "realized", &err);
+    if (err) {
+        error_propagate(errp, err);
+        return;
+    }
+}
+
 static void riscv_harts_realize(DeviceState *dev, Error **errp)
 {
     RISCVHartArrayState *s = RISCV_HART_ARRAY(dev);
-    Error *err = NULL;
     int n;
 
     s->harts = g_new0(RISCVCPU, s->num_harts);
 
     for (n = 0; n < s->num_harts; n++) {
-        object_initialize_child(OBJECT(s), "harts[*]", &s->harts[n],
-                                sizeof(RISCVCPU), s->cpu_type,
-                                &error_abort, NULL);
-        s->harts[n].env.mhartid = n;
-        qemu_register_reset(riscv_harts_cpu_reset, &s->harts[n]);
-        object_property_set_bool(OBJECT(&s->harts[n]), true,
-                                 "realized", &err);
-        if (err) {
-            error_propagate(errp, err);
-            return;
-        }
+        riscv_hart_realize(s, n, s->cpu_type, errp);
     }
 }
 
Signature · 1 line
2.7.4
Reviewed-by on v3 4/28
#5 lore raw

Re: [Qemu-devel] [PATCH v3 04/28] riscv: hart: Extract hart realize to a separate routine

Quoted 8 lines · Bin Meng

Currently riscv_harts_realize() creates all harts based on the same cpu type given in the hart array property. With current implementation it can only create symmetric harts. Exact the hart realize to a separate routine in preparation for supporting heterogeneous hart arrays.

Alistair

Quoted 62 lines

---

Changes in v3: None
Changes in v2: None

 hw/riscv/riscv_hart.c | 31 +++++++++++++++++++------------
 1 file changed, 19 insertions(+), 12 deletions(-)
hw/riscv/riscv_hart.c +19 -13
index ca69a1b..3dd1c6a 100644
@@ -37,26 +37,33 @@ static void riscv_harts_cpu_reset(void *opaque)
     cpu_reset(CPU(cpu));
 }
 
+static void riscv_hart_realize(RISCVHartArrayState *s, int hart,
+                               char *cpu_type, Error **errp)
+{
+    Error *err = NULL;
+
+    object_initialize_child(OBJECT(s), "harts[*]", &s->harts[hart],
+                            sizeof(RISCVCPU), cpu_type,
+                            &error_abort, NULL);
+    s->harts[hart].env.mhartid = hart;
+    qemu_register_reset(riscv_harts_cpu_reset, &s->harts[hart]);
+    object_property_set_bool(OBJECT(&s->harts[hart]), true,
+                             "realized", &err);
+    if (err) {
+        error_propagate(errp, err);
+        return;
+    }
+}
+
 static void riscv_harts_realize(DeviceState *dev, Error **errp)
 {
     RISCVHartArrayState *s = RISCV_HART_ARRAY(dev);
-    Error *err = NULL;
     int n;
 
     s->harts = g_new0(RISCVCPU, s->num_harts);
 
     for (n = 0; n < s->num_harts; n++) {
-        object_initialize_child(OBJECT(s), "harts[*]", &s->harts[n],
-                                sizeof(RISCVCPU), s->cpu_type,
-                                &error_abort, NULL);
-        s->harts[n].env.mhartid = n;
-        qemu_register_reset(riscv_harts_cpu_reset, &s->harts[n]);
-        object_property_set_bool(OBJECT(&s->harts[n]), true,
-                                 "realized", &err);
-        if (err) {
-            error_propagate(errp, err);
-            return;
-        }
+        riscv_hart_realize(s, n, s->cpu_type, errp);
     }
 }
 
--

2.7.4

author
#6 lore raw

PATCH v3 5/28 riscv: hart: Support heterogeneous harts population

In reply to Bin Meng

At present we only allow symmetric harts to be created. In order to support heterogeneous harts like SiFive FU540, update hart array's "cpu-type" property to allow cpu type to be set per hart, separated by delimiter ",". The frist cpu type before the delimiter is assigned to hart 0, and the second cpu type before delimiter is assigned to hart 1, and so on.

If the total number of cpu types supplied in "cpu-type" property is less than number of maximum harts, the last cpu type in the property will be used to populate remaining harts.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/riscv_hart.c | 48 +++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 45 insertions(+), 3 deletions(-)
hw/riscv/riscv_hart.c +45 -3
index 3dd1c6a..27093e0 100644
@@ -58,13 +58,55 @@ static void riscv_hart_realize(RISCVHartArrayState *s, int hart,
 static void riscv_harts_realize(DeviceState *dev, Error **errp)
 {
     RISCVHartArrayState *s = RISCV_HART_ARRAY(dev);
-    int n;
+    char *cpu_types;
+    char *first_type, *last_type, *tmp_type;
+    int n = 0;
 
     s->harts = g_new0(RISCVCPU, s->num_harts);
 
-    for (n = 0; n < s->num_harts; n++) {
-        riscv_hart_realize(s, n, s->cpu_type, errp);
+    /* we should not touch the original s->cpu_type */
+    cpu_types = g_strdup(s->cpu_type);
+
+    /*
+     * Expect s->cpu_type property was initialized this way:
+     *
+     * "cpu-type-a": symmetric harts
+     * "cpu-type-a,cpu-type-b,cpu-type-c": heterogeneous harts
+     *
+     * For heterogeneous harts, hart cpu types are separated by delimiter ",".
+     * The frist cpu type before the delimiter is assigned to hart 0, and the
+     * second cpu type before delimiter is assigned to hart 1, and so on.
+     *
+     * If the total number of cpu types is less than s->num_harts, the last
+     * cpu type in s->cpu_type will be used to populate remaining harts.
+     */
+
+    first_type = strtok(cpu_types, ",");
+    riscv_hart_realize(s, n++, first_type, errp);
+    tmp_type = strtok(NULL, ",");
+    if (!tmp_type) {
+        /* symmetric harts */
+        for (; n < s->num_harts; n++) {
+            riscv_hart_realize(s, n, first_type, errp);
+       }
+    } else {
+        /* heterogeneous harts */
+        while (tmp_type) {
+            if (n >= s->num_harts) {
+                break;
+            }
+            riscv_hart_realize(s, n++, tmp_type, errp);
+            last_type = tmp_type;
+            tmp_type = strtok(NULL, ",");
+        }
+
+        /* populate remaining harts using the last cpu type in s->cpu_type */
+        for (; n < s->num_harts; n++) {
+            riscv_hart_realize(s, n, last_type, errp);
+        }
     }
+
+    g_free(cpu_types);
 }
 
 static void riscv_harts_class_init(ObjectClass *klass, void *data)
Signature · 1 line
2.7.4
#7 lore raw

Re: [Qemu-devel] [PATCH v3 05/28] riscv: hart: Support heterogeneous harts population

hw/riscv/riscv_hart.c static void riscv_hart_realize(RISCVHartArrayState *s, int hart, · line 101

Quoted 9 lines · Bin Meng
hw/riscv/riscv_hart.c +9 -0
+        /* heterogeneous harts */
+        while (tmp_type) {
+            if (n >= s->num_harts) {
+                break;
+            }
+            riscv_hart_realize(s, n++, tmp_type, errp);
+            last_type = tmp_type;
+            tmp_type = strtok(NULL, ",");
+        }

You need to create clusters for each cpu type.

In particular, E51 does not have the fpu that U54 does, and so we need to generate different code for each cluster.

Because each cluster generates different TBs, you'll minimize code generation if you do more than just create one per cpu, e.g. pop these cpu type names into a GHashTable as you create each cluster.

The only other example of clusters in tree is hw/arm/armsse.c. Note that board has one cpu per cluster, as each cpu may be configured differently via other property settings.

r~

author
#8 lore raw

PATCH v3 6/28 riscv: sifive_u: Update hart configuration to reflect the real FU540 SoC

In reply to Bin Meng

The FU540-C000 includes a 64-bit E51 RISC-V core and four 64-bit U54 RISC-V cores. Currently the sifive_u machine only populates 4 U54 cores. Update the max cpu number to 5 to reflect the real hardware, and pass "cpu-type" to populate heterogeneous harts.

The cpu nodes in the generated DTS have been updated as well.

---

Changes in v3:
- changed to use macros for management and compute cpu count

Changes in v2:
- fixed the "interrupts-extended" property size

 hw/riscv/sifive_u.c         | 40 +++++++++++++++++++++++++++-------------
 include/hw/riscv/sifive_u.h |  3 +++
 2 files changed, 30 insertions(+), 13 deletions(-)
hw/riscv/sifive_u.c +27 -13
index 623ee64..295ca77 100644
@@ -10,7 +10,8 @@
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
  *
- * This board currently uses a hardcoded devicetree that indicates one hart.
+ * This board currently generates devicetree dynamically that indicates at most
+ * five harts.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms and conditions of the GNU General Public License,
@@ -26,6 +27,7 @@
  */
 
 #include "qemu/osdep.h"
+#include "qemu/cutils.h"
 #include "qemu/log.h"
 #include "qemu/error-report.h"
 #include "qapi/error.h"
@@ -117,7 +119,10 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         qemu_fdt_add_subnode(fdt, nodename);
         qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
                               SIFIVE_U_CLOCK_FREQ);
-        qemu_fdt_setprop_string(fdt, nodename, "mmu-type", "riscv,sv48");
+        /* cpu 0 is the management hart that does not have mmu */
+        if (cpu != 0) {
+            qemu_fdt_setprop_string(fdt, nodename, "mmu-type", "riscv,sv48");
+        }
         qemu_fdt_setprop_string(fdt, nodename, "riscv,isa", isa);
         qemu_fdt_setprop_string(fdt, nodename, "compatible", "riscv");
         qemu_fdt_setprop_string(fdt, nodename, "status", "okay");
@@ -157,15 +162,21 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     g_free(nodename);
 
     plic_phandle = phandle++;
-    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4);
+    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4 - 2);
     for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
         nodename =
             g_strdup_printf("/cpus/cpu@%d/interrupt-controller", cpu);
         uint32_t intc_phandle = qemu_fdt_get_phandle(fdt, nodename);
-        cells[cpu * 4 + 0] = cpu_to_be32(intc_phandle);
-        cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_EXT);
-        cells[cpu * 4 + 2] = cpu_to_be32(intc_phandle);
-        cells[cpu * 4 + 3] = cpu_to_be32(IRQ_S_EXT);
+        /* cpu 0 is the management hart that does not have S-mode */
+        if (cpu == 0) {
+            cells[0] = cpu_to_be32(intc_phandle);
+            cells[1] = cpu_to_be32(IRQ_M_EXT);
+        } else {
+            cells[cpu * 4 - 2] = cpu_to_be32(intc_phandle);
+            cells[cpu * 4 - 1] = cpu_to_be32(IRQ_M_EXT);
+            cells[cpu * 4 + 0] = cpu_to_be32(intc_phandle);
+            cells[cpu * 4 + 1] = cpu_to_be32(IRQ_S_EXT);
+        }
         g_free(nodename);
     }
     nodename = g_strdup_printf("/soc/interrupt-controller@%lx",
@@ -175,7 +186,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_string(fdt, nodename, "compatible", "riscv,plic0");
     qemu_fdt_setprop(fdt, nodename, "interrupt-controller", NULL, 0);
     qemu_fdt_setprop(fdt, nodename, "interrupts-extended",
-        cells, s->soc.cpus.num_harts * sizeof(uint32_t) * 4);
+        cells, (s->soc.cpus.num_harts * 4 - 2) * sizeof(uint32_t));
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_PLIC].base,
         0x0, memmap[SIFIVE_U_PLIC].size);
@@ -315,10 +326,16 @@ static void riscv_sifive_u_soc_init(Object *obj)
 {
     MachineState *ms = MACHINE(qdev_get_machine());
     SiFiveUSoCState *s = RISCV_U_SOC(obj);
+    char cpu_type[64];
+
+    /* create cpu type representing SiFive FU540 SoC */
+    pstrcpy(cpu_type, sizeof(cpu_type), SIFIVE_E_CPU);
+    pstrcat(cpu_type, sizeof(cpu_type), ",");
+    pstrcat(cpu_type, sizeof(cpu_type), SIFIVE_U_CPU);
 
     object_initialize_child(obj, "cpus", &s->cpus, sizeof(s->cpus),
                             TYPE_RISCV_HART_ARRAY, &error_abort, NULL);
-    object_property_set_str(OBJECT(&s->cpus), SIFIVE_U_CPU, "cpu-type",
+    object_property_set_str(OBJECT(&s->cpus), cpu_type, "cpu-type",
                             &error_abort);
     object_property_set_int(OBJECT(&s->cpus), ms->smp.cpus, "num-harts",
                             &error_abort);
@@ -407,10 +424,7 @@ static void riscv_sifive_u_machine_init(MachineClass *mc)
 {
     mc->desc = "RISC-V Board compatible with SiFive U SDK";
     mc->init = riscv_sifive_u_init;
-    /* The real hardware has 5 CPUs, but one of them is a small embedded power
-     * management CPU.
-     */
-    mc->max_cpus = 4;
+    mc->max_cpus = SIFIVE_U_MANAGEMENT_CPU_COUNT + SIFIVE_U_COMPUTE_CPU_COUNT;
 }
 
 DEFINE_MACHINE("sifive_u", riscv_sifive_u_machine_init)
include/hw/riscv/sifive_u.h +3 -0
index 4abc621..650bc4c 100644
@@ -68,6 +68,9 @@ enum {
     SIFIVE_U_GEM_CLOCK_FREQ = 125000000
 };
 
+#define SIFIVE_U_MANAGEMENT_CPU_COUNT   1
+#define SIFIVE_U_COMPUTE_CPU_COUNT      4
+
 #define SIFIVE_U_PLIC_HART_CONFIG "MS"
 #define SIFIVE_U_PLIC_NUM_SOURCES 54
 #define SIFIVE_U_PLIC_NUM_PRIORITIES 7
Signature · 1 line
2.7.4
Reviewed-by on v3 6/28
#9 lore raw

Re: [Qemu-devel] [PATCH v3 06/28] riscv: sifive_u: Update hart configuration to reflect the real FU540 SoC

Quoted 9 lines · Bin Meng

The FU540-C000 includes a 64-bit E51 RISC-V core and four 64-bit U54 RISC-V cores. Currently the sifive_u machine only populates 4 U54 cores. Update the max cpu number to 5 to reflect the real hardware, and pass "cpu-type" to populate heterogeneous harts.

The cpu nodes in the generated DTS have been updated as well.

Alistair

Quoted 131 lines

---

Changes in v3:
- changed to use macros for management and compute cpu count

Changes in v2:
- fixed the "interrupts-extended" property size

 hw/riscv/sifive_u.c         | 40 +++++++++++++++++++++++++++-------------
 include/hw/riscv/sifive_u.h |  3 +++
 2 files changed, 30 insertions(+), 13 deletions(-)
hw/riscv/sifive_u.c +27 -13
index 623ee64..295ca77 100644
@@ -10,7 +10,8 @@
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
  *
- * This board currently uses a hardcoded devicetree that indicates one hart.
+ * This board currently generates devicetree dynamically that indicates at most
+ * five harts.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms and conditions of the GNU General Public License,
@@ -26,6 +27,7 @@
  */
 
 #include "qemu/osdep.h"
+#include "qemu/cutils.h"
 #include "qemu/log.h"
 #include "qemu/error-report.h"
 #include "qapi/error.h"
@@ -117,7 +119,10 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         qemu_fdt_add_subnode(fdt, nodename);
         qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
                               SIFIVE_U_CLOCK_FREQ);
-        qemu_fdt_setprop_string(fdt, nodename, "mmu-type", "riscv,sv48");
+        /* cpu 0 is the management hart that does not have mmu */
+        if (cpu != 0) {
+            qemu_fdt_setprop_string(fdt, nodename, "mmu-type", "riscv,sv48");
+        }
         qemu_fdt_setprop_string(fdt, nodename, "riscv,isa", isa);
         qemu_fdt_setprop_string(fdt, nodename, "compatible", "riscv");
         qemu_fdt_setprop_string(fdt, nodename, "status", "okay");
@@ -157,15 +162,21 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     g_free(nodename);
 
     plic_phandle = phandle++;
-    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4);
+    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4 - 2);
     for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
         nodename =
             g_strdup_printf("/cpus/cpu@%d/interrupt-controller", cpu);
         uint32_t intc_phandle = qemu_fdt_get_phandle(fdt, nodename);
-        cells[cpu * 4 + 0] = cpu_to_be32(intc_phandle);
-        cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_EXT);
-        cells[cpu * 4 + 2] = cpu_to_be32(intc_phandle);
-        cells[cpu * 4 + 3] = cpu_to_be32(IRQ_S_EXT);
+        /* cpu 0 is the management hart that does not have S-mode */
+        if (cpu == 0) {
+            cells[0] = cpu_to_be32(intc_phandle);
+            cells[1] = cpu_to_be32(IRQ_M_EXT);
+        } else {
+            cells[cpu * 4 - 2] = cpu_to_be32(intc_phandle);
+            cells[cpu * 4 - 1] = cpu_to_be32(IRQ_M_EXT);
+            cells[cpu * 4 + 0] = cpu_to_be32(intc_phandle);
+            cells[cpu * 4 + 1] = cpu_to_be32(IRQ_S_EXT);
+        }
         g_free(nodename);
     }
     nodename = g_strdup_printf("/soc/interrupt-controller@%lx",
@@ -175,7 +186,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_string(fdt, nodename, "compatible", "riscv,plic0");
     qemu_fdt_setprop(fdt, nodename, "interrupt-controller", NULL, 0);
     qemu_fdt_setprop(fdt, nodename, "interrupts-extended",
-        cells, s->soc.cpus.num_harts * sizeof(uint32_t) * 4);
+        cells, (s->soc.cpus.num_harts * 4 - 2) * sizeof(uint32_t));
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_PLIC].base,
         0x0, memmap[SIFIVE_U_PLIC].size);
@@ -315,10 +326,16 @@ static void riscv_sifive_u_soc_init(Object *obj)
 {
     MachineState *ms = MACHINE(qdev_get_machine());
     SiFiveUSoCState *s = RISCV_U_SOC(obj);
+    char cpu_type[64];
+
+    /* create cpu type representing SiFive FU540 SoC */
+    pstrcpy(cpu_type, sizeof(cpu_type), SIFIVE_E_CPU);
+    pstrcat(cpu_type, sizeof(cpu_type), ",");
+    pstrcat(cpu_type, sizeof(cpu_type), SIFIVE_U_CPU);
 
     object_initialize_child(obj, "cpus", &s->cpus, sizeof(s->cpus),
                             TYPE_RISCV_HART_ARRAY, &error_abort, NULL);
-    object_property_set_str(OBJECT(&s->cpus), SIFIVE_U_CPU, "cpu-type",
+    object_property_set_str(OBJECT(&s->cpus), cpu_type, "cpu-type",
                             &error_abort);
     object_property_set_int(OBJECT(&s->cpus), ms->smp.cpus, "num-harts",
                             &error_abort);
@@ -407,10 +424,7 @@ static void riscv_sifive_u_machine_init(MachineClass *mc)
 {
     mc->desc = "RISC-V Board compatible with SiFive U SDK";
     mc->init = riscv_sifive_u_init;
-    /* The real hardware has 5 CPUs, but one of them is a small embedded power
-     * management CPU.
-     */
-    mc->max_cpus = 4;
+    mc->max_cpus = SIFIVE_U_MANAGEMENT_CPU_COUNT + SIFIVE_U_COMPUTE_CPU_COUNT;
 }
 
 DEFINE_MACHINE("sifive_u", riscv_sifive_u_machine_init)
include/hw/riscv/sifive_u.h +3 -1
index 4abc621..650bc4c 100644
@@ -68,6 +68,9 @@ enum {
     SIFIVE_U_GEM_CLOCK_FREQ = 125000000
 };
 
+#define SIFIVE_U_MANAGEMENT_CPU_COUNT   1
+#define SIFIVE_U_COMPUTE_CPU_COUNT      4
+
 #define SIFIVE_U_PLIC_HART_CONFIG "MS"
 #define SIFIVE_U_PLIC_NUM_SOURCES 54
 #define SIFIVE_U_PLIC_NUM_PRIORITIES 7
--

2.7.4

author
#10 lore raw

PATCH v3 7/28 riscv: sifive_u: Set the minimum number of cpus to 2

In reply to Bin Meng

It is not useful if we only have one management CPU.

---

Changes in v3:
- use management cpu count + 1 for the min_cpus

Changes in v2:
- update the file header to indicate at least 2 harts are created

 hw/riscv/sifive_u.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/riscv/sifive_u.c +3 -2
index 295ca77..f8ffc0b 100644
@@ -10,8 +10,8 @@
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
  *
- * This board currently generates devicetree dynamically that indicates at most
- * five harts.
+ * This board currently generates devicetree dynamically that indicates at least
+ * two harts and up to five harts.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms and conditions of the GNU General Public License,
@@ -425,6 +425,7 @@ static void riscv_sifive_u_machine_init(MachineClass *mc)
     mc->desc = "RISC-V Board compatible with SiFive U SDK";
     mc->init = riscv_sifive_u_init;
     mc->max_cpus = SIFIVE_U_MANAGEMENT_CPU_COUNT + SIFIVE_U_COMPUTE_CPU_COUNT;
+    mc->min_cpus = SIFIVE_U_MANAGEMENT_CPU_COUNT + 1;
 }
 
 DEFINE_MACHINE("sifive_u", riscv_sifive_u_machine_init)
Signature · 1 line
2.7.4
Reviewed-by on v3 7/28
#11 lore raw

Re: [Qemu-devel] [PATCH v3 07/28] riscv: sifive_u: Set the minimum number of cpus to 2

Quoted 4 lines · Bin Meng

It is not useful if we only have one management CPU.

Alistair

Quoted 39 lines

---

Changes in v3:
- use management cpu count + 1 for the min_cpus

Changes in v2:
- update the file header to indicate at least 2 harts are created

 hw/riscv/sifive_u.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/riscv/sifive_u.c +3 -3
index 295ca77..f8ffc0b 100644
@@ -10,8 +10,8 @@
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
  *
- * This board currently generates devicetree dynamically that indicates at most
- * five harts.
+ * This board currently generates devicetree dynamically that indicates at least
+ * two harts and up to five harts.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms and conditions of the GNU General Public License,
@@ -425,6 +425,7 @@ static void riscv_sifive_u_machine_init(MachineClass *mc)
     mc->desc = "RISC-V Board compatible with SiFive U SDK";
     mc->init = riscv_sifive_u_init;
     mc->max_cpus = SIFIVE_U_MANAGEMENT_CPU_COUNT + SIFIVE_U_COMPUTE_CPU_COUNT;
+    mc->min_cpus = SIFIVE_U_MANAGEMENT_CPU_COUNT + 1;
 }
 
 DEFINE_MACHINE("sifive_u", riscv_sifive_u_machine_init)
--

2.7.4

author
#12 lore raw

PATCH v3 8/28 riscv: sifive_u: Update PLIC hart topology configuration string

In reply to Bin Meng

With heterogeneous harts config, the PLIC hart topology configuration string are "M,MS,.." because of the monitor hart #0.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
hw/riscv/sifive_u.c +4 -3
index f8ffc0b..16ab95c 100644
@@ -373,10 +373,11 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
     plic_hart_config = g_malloc0(plic_hart_config_len);
     for (i = 0; i < ms->smp.cpus; i++) {
         if (i != 0) {
-            strncat(plic_hart_config, ",", plic_hart_config_len);
+            strncat(plic_hart_config, "," SIFIVE_U_PLIC_HART_CONFIG,
+                    plic_hart_config_len);
+        } else {
+            strncat(plic_hart_config, "M", plic_hart_config_len);
         }
-        strncat(plic_hart_config, SIFIVE_U_PLIC_HART_CONFIG,
-                plic_hart_config_len);
         plic_hart_config_len -= (strlen(SIFIVE_U_PLIC_HART_CONFIG) + 1);
     }
 
Signature · 1 line
2.7.4
author
#13 lore raw

PATCH v3 9/28 riscv: sifive_u: Update UART base addresses and IRQs

In reply to Bin Meng

This updates the UART base address and IRQs to match the hardware.

---

Changes in v3:
- update IRQ numbers of both UARTs to match hardware as well

Changes in v2: None

 hw/riscv/sifive_u.c         | 4 ++--
 include/hw/riscv/sifive_u.h | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
hw/riscv/sifive_u.c +2 -2
index 16ab95c..f24ec2e 100644
@@ -61,8 +61,8 @@ static const struct MemmapEntry {
     [SIFIVE_U_MROM] =     {     0x1000,    0x11000 },
     [SIFIVE_U_CLINT] =    {  0x2000000,    0x10000 },
     [SIFIVE_U_PLIC] =     {  0xc000000,  0x4000000 },
-    [SIFIVE_U_UART0] =    { 0x10013000,     0x1000 },
-    [SIFIVE_U_UART1] =    { 0x10023000,     0x1000 },
+    [SIFIVE_U_UART0] =    { 0x10010000,     0x1000 },
+    [SIFIVE_U_UART1] =    { 0x10011000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
     [SIFIVE_U_GEM] =      { 0x100900FC,     0x2000 },
 };
include/hw/riscv/sifive_u.h +2 -2
index 650bc4c..d0d8528 100644
@@ -58,8 +58,8 @@ enum {
 };
 
 enum {
-    SIFIVE_U_UART0_IRQ = 3,
-    SIFIVE_U_UART1_IRQ = 4,
+    SIFIVE_U_UART0_IRQ = 4,
+    SIFIVE_U_UART1_IRQ = 5,
     SIFIVE_U_GEM_IRQ = 0x35
 };
 
Signature · 1 line
2.7.4
author
#14 lore raw

PATCH v3 10/28 riscv: sifive_u: Remove the unnecessary include of prci header

In reply to Bin Meng

sifive_u machine does not use PRCI as of today. Remove the prci header inclusion.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 1 -
 1 file changed, 1 deletion(-)
hw/riscv/sifive_u.c +0 -1
index f24ec2e..e071838 100644
@@ -41,7 +41,6 @@
 #include "hw/riscv/sifive_plic.h"
 #include "hw/riscv/sifive_clint.h"
 #include "hw/riscv/sifive_uart.h"
-#include "hw/riscv/sifive_prci.h"
 #include "hw/riscv/sifive_u.h"
 #include "hw/riscv/boot.h"
 #include "chardev/char.h"
Signature · 1 line
2.7.4
author
#15 lore raw

PATCH v3 11/28 riscv: sifive: Rename sifive_prci.{c, h} to sifive_e_prci.{c, h}

In reply to Bin Meng

Current SiFive PRCI model only works with sifive_e machine, as it only emulates registers or PRCI block in the FE310 SoC.

Rename the file name to make it clear that it is for sifive_e.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs                              |  2 +-
 hw/riscv/sifive_e.c                                 |  4 ++--
 hw/riscv/{sifive_prci.c => sifive_e_prci.c}         | 14 +++++++-------
 include/hw/riscv/{sifive_prci.h => sifive_e_prci.h} | 14 +++++++-------
 4 files changed, 17 insertions(+), 17 deletions(-)
 rename hw/riscv/{sifive_prci.c => sifive_e_prci.c} (90%)
 rename include/hw/riscv/{sifive_prci.h => sifive_e_prci.h} (82%)
hw/riscv/Makefile.objs +1 -1
index eb9d4f9..c859697 100644
@@ -2,9 +2,9 @@ obj-y += boot.o
 obj-$(CONFIG_SPIKE) += riscv_htif.o
 obj-$(CONFIG_HART) += riscv_hart.o
 obj-$(CONFIG_SIFIVE_E) += sifive_e.o
+obj-$(CONFIG_SIFIVE_E) += sifive_e_prci.o
 obj-$(CONFIG_SIFIVE) += sifive_clint.o
 obj-$(CONFIG_SIFIVE) += sifive_gpio.o
-obj-$(CONFIG_SIFIVE) += sifive_prci.o
 obj-$(CONFIG_SIFIVE) += sifive_plic.o
 obj-$(CONFIG_SIFIVE) += sifive_test.o
 obj-$(CONFIG_SIFIVE_U) += sifive_u.o
hw/riscv/sifive_e.c +2 -2
index 2a499d8..2d67670 100644
@@ -41,9 +41,9 @@
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
 #include "hw/riscv/sifive_clint.h"
-#include "hw/riscv/sifive_prci.h"
 #include "hw/riscv/sifive_uart.h"
 #include "hw/riscv/sifive_e.h"
+#include "hw/riscv/sifive_e_prci.h"
 #include "hw/riscv/boot.h"
 #include "chardev/char.h"
 #include "sysemu/arch_init.h"
@@ -174,7 +174,7 @@ static void riscv_sifive_e_soc_realize(DeviceState *dev, Error **errp)
         SIFIVE_SIP_BASE, SIFIVE_TIMECMP_BASE, SIFIVE_TIME_BASE);
     sifive_mmio_emulate(sys_mem, "riscv.sifive.e.aon",
         memmap[SIFIVE_E_AON].base, memmap[SIFIVE_E_AON].size);
-    sifive_prci_create(memmap[SIFIVE_E_PRCI].base);
+    sifive_e_prci_create(memmap[SIFIVE_E_PRCI].base);
 
     /* GPIO */
 
hw/riscv/sifive_e_prci.c renamed from hw/riscv/sifive_prci.c +7 -7
similarity index 90%
rename from hw/riscv/sifive_prci.c
rename to hw/riscv/sifive_e_prci.c
index f406682..acb914d 100644
@@ -1,5 +1,5 @@
 /*
- * QEMU SiFive PRCI (Power, Reset, Clock, Interrupt)
+ * QEMU SiFive E PRCI (Power, Reset, Clock, Interrupt)
  *
  * Copyright (c) 2017 SiFive, Inc.
  *
@@ -22,7 +22,7 @@
 #include "hw/sysbus.h"
 #include "qemu/module.h"
 #include "target/riscv/cpu.h"
-#include "hw/riscv/sifive_prci.h"
+#include "hw/riscv/sifive_e_prci.h"
 
 static uint64_t sifive_prci_read(void *opaque, hwaddr addr, unsigned int size)
 {
@@ -82,10 +82,10 @@ static const MemoryRegionOps sifive_prci_ops = {
 
 static void sifive_prci_init(Object *obj)
 {
-    SiFivePRCIState *s = SIFIVE_PRCI(obj);
+    SiFivePRCIState *s = SIFIVE_E_PRCI(obj);
 
     memory_region_init_io(&s->mmio, obj, &sifive_prci_ops, s,
-                          TYPE_SIFIVE_PRCI, 0x8000);
+                          TYPE_SIFIVE_E_PRCI, 0x8000);
     sysbus_init_mmio(SYS_BUS_DEVICE(obj), &s->mmio);
 
     s->hfrosccfg = (SIFIVE_PRCI_HFROSCCFG_RDY | SIFIVE_PRCI_HFROSCCFG_EN);
@@ -97,7 +97,7 @@ static void sifive_prci_init(Object *obj)
 }
 
 static const TypeInfo sifive_prci_info = {
-    .name          = TYPE_SIFIVE_PRCI,
+    .name          = TYPE_SIFIVE_E_PRCI,
     .parent        = TYPE_SYS_BUS_DEVICE,
     .instance_size = sizeof(SiFivePRCIState),
     .instance_init = sifive_prci_init,
@@ -114,9 +114,9 @@ type_init(sifive_prci_register_types)
 /*
  * Create PRCI device.
  */
-DeviceState *sifive_prci_create(hwaddr addr)
+DeviceState *sifive_e_prci_create(hwaddr addr)
 {
-    DeviceState *dev = qdev_create(NULL, TYPE_SIFIVE_PRCI);
+    DeviceState *dev = qdev_create(NULL, TYPE_SIFIVE_E_PRCI);
     qdev_init_nofail(dev);
     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, addr);
     return dev;
include/hw/riscv/sifive_e_prci.h renamed from include/hw/riscv/sifive_prci.h +7 -7
similarity index 82%
rename from include/hw/riscv/sifive_prci.h
rename to include/hw/riscv/sifive_e_prci.h
index bd51c4a..7932fe7 100644
@@ -1,5 +1,5 @@
 /*
- * QEMU SiFive PRCI (Power, Reset, Clock, Interrupt) interface
+ * QEMU SiFive E PRCI (Power, Reset, Clock, Interrupt) interface
  *
  * Copyright (c) 2017 SiFive, Inc.
  *
@@ -16,8 +16,8 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef HW_SIFIVE_PRCI_H
-#define HW_SIFIVE_PRCI_H
+#ifndef HW_SIFIVE_E_PRCI_H
+#define HW_SIFIVE_E_PRCI_H
 
 enum {
     SIFIVE_PRCI_HFROSCCFG   = 0x0,
@@ -47,10 +47,10 @@ enum {
     SIFIVE_PRCI_PLLOUTDIV_DIV1  = (1 << 8)
 };
 
-#define TYPE_SIFIVE_PRCI "riscv.sifive.prci"
+#define TYPE_SIFIVE_E_PRCI      "riscv.sifive.e.prci"
 
-#define SIFIVE_PRCI(obj) \
-    OBJECT_CHECK(SiFivePRCIState, (obj), TYPE_SIFIVE_PRCI)
+#define SIFIVE_E_PRCI(obj) \
+    OBJECT_CHECK(SiFivePRCIState, (obj), TYPE_SIFIVE_E_PRCI)
 
 typedef struct SiFivePRCIState {
     /*< private >*/
@@ -64,6 +64,6 @@ typedef struct SiFivePRCIState {
     uint32_t plloutdiv;
 } SiFivePRCIState;
 
-DeviceState *sifive_prci_create(hwaddr addr);
+DeviceState *sifive_e_prci_create(hwaddr addr);
 
 #endif
Signature · 1 line
2.7.4
author
#16 lore raw

PATCH v3 12/28 riscv: sifive_e: prci: Fix a typo of hfxosccfg register programming

In reply to Bin Meng

It should use SIFIVE_PRCI_HFXOSCCFG_RDY and SIFIVE_PRCI_HFXOSCCFG_EN for hfxosccfg register programming.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_e_prci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/riscv/sifive_e_prci.c +1 -1
index acb914d..c906f11 100644
@@ -89,7 +89,7 @@ static void sifive_prci_init(Object *obj)
     sysbus_init_mmio(SYS_BUS_DEVICE(obj), &s->mmio);
 
     s->hfrosccfg = (SIFIVE_PRCI_HFROSCCFG_RDY | SIFIVE_PRCI_HFROSCCFG_EN);
-    s->hfxosccfg = (SIFIVE_PRCI_HFROSCCFG_RDY | SIFIVE_PRCI_HFROSCCFG_EN);
+    s->hfxosccfg = (SIFIVE_PRCI_HFXOSCCFG_RDY | SIFIVE_PRCI_HFXOSCCFG_EN);
     s->pllcfg = (SIFIVE_PRCI_PLLCFG_REFSEL | SIFIVE_PRCI_PLLCFG_BYPASS |
                 SIFIVE_PRCI_PLLCFG_LOCK);
     s->plloutdiv = SIFIVE_PRCI_PLLOUTDIV_DIV1;
Signature · 1 line
2.7.4
author
#17 lore raw

PATCH v3 13/28 riscv: sifive_e: prci: Update the PRCI register block size

In reply to Bin Meng

Currently the PRCI register block size is set to 0x8000, but in fact 0x1000 is enough, which is also what the manual says.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_e_prci.c         | 2 +-
 include/hw/riscv/sifive_e_prci.h | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)
hw/riscv/sifive_e_prci.c +1 -1
index c906f11..4cbce48 100644
@@ -85,7 +85,7 @@ static void sifive_prci_init(Object *obj)
     SiFivePRCIState *s = SIFIVE_E_PRCI(obj);
 
     memory_region_init_io(&s->mmio, obj, &sifive_prci_ops, s,
-                          TYPE_SIFIVE_E_PRCI, 0x8000);
+                          TYPE_SIFIVE_E_PRCI, SIFIVE_E_PRCI_REG_SIZE);
     sysbus_init_mmio(SYS_BUS_DEVICE(obj), &s->mmio);
 
     s->hfrosccfg = (SIFIVE_PRCI_HFROSCCFG_RDY | SIFIVE_PRCI_HFROSCCFG_EN);
include/hw/riscv/sifive_e_prci.h +2 -0
index 7932fe7..81e506b 100644
@@ -47,6 +47,8 @@ enum {
     SIFIVE_PRCI_PLLOUTDIV_DIV1  = (1 << 8)
 };
 
+#define SIFIVE_E_PRCI_REG_SIZE  0x1000
+
 #define TYPE_SIFIVE_E_PRCI      "riscv.sifive.e.prci"
 
 #define SIFIVE_E_PRCI(obj) \
Signature · 1 line
2.7.4
Reviewed-by on v3 13/28
#18 lore raw

Re: [Qemu-devel] [PATCH v3 13/28] riscv: sifive_e: prci: Update the PRCI register block size

Quoted 6 lines · Bin Meng

Currently the PRCI register block size is set to 0x8000, but in fact 0x1000 is enough, which is also what the manual says.

Alistair

Quoted 39 lines

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_e_prci.c         | 2 +-
 include/hw/riscv/sifive_e_prci.h | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)
hw/riscv/sifive_e_prci.c +1 -1
index c906f11..4cbce48 100644
@@ -85,7 +85,7 @@ static void sifive_prci_init(Object *obj)
     SiFivePRCIState *s = SIFIVE_E_PRCI(obj);
 
     memory_region_init_io(&s->mmio, obj, &sifive_prci_ops, s,
-                          TYPE_SIFIVE_E_PRCI, 0x8000);
+                          TYPE_SIFIVE_E_PRCI, SIFIVE_E_PRCI_REG_SIZE);
     sysbus_init_mmio(SYS_BUS_DEVICE(obj), &s->mmio);
 
     s->hfrosccfg = (SIFIVE_PRCI_HFROSCCFG_RDY | SIFIVE_PRCI_HFROSCCFG_EN);
include/hw/riscv/sifive_e_prci.h +2 -1
index 7932fe7..81e506b 100644
@@ -47,6 +47,8 @@ enum {
     SIFIVE_PRCI_PLLOUTDIV_DIV1  = (1 << 8)
 };
 
+#define SIFIVE_E_PRCI_REG_SIZE  0x1000
+
 #define TYPE_SIFIVE_E_PRCI      "riscv.sifive.e.prci"
 
 #define SIFIVE_E_PRCI(obj) \
--

2.7.4

author
#19 lore raw

PATCH v3 14/28 riscv: sifive: Implement PRCI model for FU540

In reply to Bin Meng

This adds a simple PRCI model for FU540 (sifive_u). It has different register layout from the existing PRCI model for FE310 (sifive_e).

---

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs           |   1 +
 hw/riscv/sifive_u_prci.c         | 163 +++++++++++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_u_prci.h |  90 +++++++++++++++++++++
 3 files changed, 254 insertions(+)
 create mode 100644 hw/riscv/sifive_u_prci.c
 create mode 100644 include/hw/riscv/sifive_u_prci.h
hw/riscv/Makefile.objs +1 -0
index c859697..b95bbd5 100644
@@ -8,6 +8,7 @@ obj-$(CONFIG_SIFIVE) += sifive_gpio.o
 obj-$(CONFIG_SIFIVE) += sifive_plic.o
 obj-$(CONFIG_SIFIVE) += sifive_test.o
 obj-$(CONFIG_SIFIVE_U) += sifive_u.o
+obj-$(CONFIG_SIFIVE_U) += sifive_u_prci.o
 obj-$(CONFIG_SIFIVE) += sifive_uart.o
 obj-$(CONFIG_SPIKE) += spike.o
 obj-$(CONFIG_RISCV_VIRT) += virt.o
hw/riscv/sifive_u_prci.c new file +163 -0

163 lines not on this page · raw message

include/hw/riscv/sifive_u_prci.h new file +90 -0

90 lines not on this page · raw message

Signature · 1 line
2.7.4
#20 lore raw

Re: [Qemu-devel] [PATCH v3 14/28] riscv: sifive: Implement PRCI model for FU540

Quoted 265 lines · Bin Meng

This adds a simple PRCI model for FU540 (sifive_u). It has different register layout from the existing PRCI model for FE310 (sifive_e).

---

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs           |   1 +
 hw/riscv/sifive_u_prci.c         | 163 +++++++++++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_u_prci.h |  90 +++++++++++++++++++++
 3 files changed, 254 insertions(+)
 create mode 100644 hw/riscv/sifive_u_prci.c
 create mode 100644 include/hw/riscv/sifive_u_prci.h
hw/riscv/Makefile.objs +1 -0
index c859697..b95bbd5 100644
@@ -8,6 +8,7 @@ obj-$(CONFIG_SIFIVE) += sifive_gpio.o
 obj-$(CONFIG_SIFIVE) += sifive_plic.o
 obj-$(CONFIG_SIFIVE) += sifive_test.o
 obj-$(CONFIG_SIFIVE_U) += sifive_u.o
+obj-$(CONFIG_SIFIVE_U) += sifive_u_prci.o
 obj-$(CONFIG_SIFIVE) += sifive_uart.o
 obj-$(CONFIG_SPIKE) += spike.o
 obj-$(CONFIG_RISCV_VIRT) += virt.o
hw/riscv/sifive_u_prci.c new file +163 -0
new file mode 100644
index 0000000..35e5962
@@ -0,0 +1,163 @@
+/*
+ * QEMU SiFive U PRCI (Power, Reset, Clock, Interrupt)
+ *
+ * Copyright (c) 2019 Bin Meng <[email protected]>
+ *
+ * Simple model of the PRCI to emulate register reads made by the SDK BSP
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2 or later, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "qemu/osdep.h"
+#include "hw/sysbus.h"
+#include "qemu/module.h"
+#include "target/riscv/cpu.h"
+#include "hw/riscv/sifive_u_prci.h"
+
+static uint64_t sifive_prci_read(void *opaque, hwaddr addr, unsigned int size)
+{
+    SiFivePRCIState *s = opaque;
+
+    switch (addr) {
+    case SIFIVE_PRCI_HFXOSCCFG:
+        return s->hfxosccfg;
+    case SIFIVE_PRCI_COREPLLCFG0:
+        return s->corepllcfg0;
+    case SIFIVE_PRCI_DDRPLLCFG0:
+        return s->ddrpllcfg0;
+    case SIFIVE_PRCI_DDRPLLCFG1:
+        return s->ddrpllcfg1;
+    case SIFIVE_PRCI_GEMGXLPLLCFG0:
+        return s->gemgxlpllcfg0;
+    case SIFIVE_PRCI_GEMGXLPLLCFG1:
+        return s->gemgxlpllcfg1;
+    case SIFIVE_PRCI_CORECLKSEL:
+        return s->coreclksel;
+    case SIFIVE_PRCI_DEVICESRESET:
+        return s->devicesreset;
+    case SIFIVE_PRCI_CLKMUXSTATUS:
+        return s->clkmuxstatus;
+    }
+
+    hw_error("%s: read: addr=0x%x\n", __func__, (int)addr);
+    return 0;
+}
+
+static void sifive_prci_write(void *opaque, hwaddr addr,
+                              uint64_t val64, unsigned int size)
+{
+    SiFivePRCIState *s = opaque;
+
+    switch (addr) {
+    case SIFIVE_PRCI_HFXOSCCFG:
+        s->hfxosccfg = (uint32_t) val64;
+        /* OSC stays ready */
+        s->hfxosccfg |= SIFIVE_PRCI_HFXOSCCFG_RDY;
+        break;
+    case SIFIVE_PRCI_COREPLLCFG0:
+        s->corepllcfg0 = (uint32_t) val64;
+        /* internal feedback */
+        s->corepllcfg0 |= SIFIVE_PRCI_PLLCFG0_FSE;
+        /* PLL stays locked */
+        s->corepllcfg0 |= SIFIVE_PRCI_PLLCFG0_LOCK;
+        break;
+    case SIFIVE_PRCI_DDRPLLCFG0:
+        s->ddrpllcfg0 = (uint32_t) val64;
+        /* internal feedback */
+        s->ddrpllcfg0 |= SIFIVE_PRCI_PLLCFG0_FSE;
+        /* PLL stays locked */
+        s->ddrpllcfg0 |= SIFIVE_PRCI_PLLCFG0_LOCK;
+        break;
+    case SIFIVE_PRCI_DDRPLLCFG1:
+        s->ddrpllcfg1 = (uint32_t) val64;
+        break;
+    case SIFIVE_PRCI_GEMGXLPLLCFG0:
+        s->gemgxlpllcfg0 = (uint32_t) val64;
+         /* internal feedback */
+        s->gemgxlpllcfg0 |= SIFIVE_PRCI_PLLCFG0_FSE;
+       /* PLL stays locked */
+        s->gemgxlpllcfg0 |= SIFIVE_PRCI_PLLCFG0_LOCK;
+        break;
+    case SIFIVE_PRCI_GEMGXLPLLCFG1:
+        s->gemgxlpllcfg1 = (uint32_t) val64;
+        break;
+    case SIFIVE_PRCI_CORECLKSEL:
+        s->coreclksel = (uint32_t) val64;
+        break;
+    case SIFIVE_PRCI_DEVICESRESET:
+        s->devicesreset = (uint32_t) val64;
+        break;
+    case SIFIVE_PRCI_CLKMUXSTATUS:
+        s->clkmuxstatus = (uint32_t) val64;
+        break;
+    default:
+        hw_error("%s: bad write: addr=0x%x v=0x%x\n",
+                 __func__, (int)addr, (int)val64);
+    }
+}
+
+static const MemoryRegionOps sifive_prci_ops = {
+    .read = sifive_prci_read,
+    .write = sifive_prci_write,
+    .endianness = DEVICE_NATIVE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 4
+    }
+};
+
+static void sifive_prci_init(Object *obj)
+{
+    SiFivePRCIState *s = SIFIVE_U_PRCI(obj);
+
+    memory_region_init_io(&s->mmio, obj, &sifive_prci_ops, s,
+                          TYPE_SIFIVE_U_PRCI, SIFIVE_U_PRCI_REG_SIZE);
+    sysbus_init_mmio(SYS_BUS_DEVICE(obj), &s->mmio);
+
+    /* Initialize register to power-on-reset values */
+    s->hfxosccfg = (SIFIVE_PRCI_HFXOSCCFG_RDY | SIFIVE_PRCI_HFXOSCCFG_EN);
+    s->corepllcfg0 = (SIFIVE_PRCI_PLLCFG0_DIVR | SIFIVE_PRCI_PLLCFG0_DIVF |
+                      SIFIVE_PRCI_PLLCFG0_DIVQ | SIFIVE_PRCI_PLLCFG0_FSE |
+                      SIFIVE_PRCI_PLLCFG0_LOCK);
+    s->ddrpllcfg0 = (SIFIVE_PRCI_PLLCFG0_DIVR | SIFIVE_PRCI_PLLCFG0_DIVF |
+                     SIFIVE_PRCI_PLLCFG0_DIVQ | SIFIVE_PRCI_PLLCFG0_FSE |
+                     SIFIVE_PRCI_PLLCFG0_LOCK);
+    s->gemgxlpllcfg0 = (SIFIVE_PRCI_PLLCFG0_DIVR | SIFIVE_PRCI_PLLCFG0_DIVF |
+                        SIFIVE_PRCI_PLLCFG0_DIVQ | SIFIVE_PRCI_PLLCFG0_FSE |
+                        SIFIVE_PRCI_PLLCFG0_LOCK);
+    s->coreclksel = SIFIVE_PRCI_CORECLKSEL_HFCLK;
+}
+
+static const TypeInfo sifive_prci_info = {
+    .name          = TYPE_SIFIVE_U_PRCI,
+    .parent        = TYPE_SYS_BUS_DEVICE,
+    .instance_size = sizeof(SiFivePRCIState),
+    .instance_init = sifive_prci_init,
+};
+
+static void sifive_prci_register_types(void)
+{
+    type_register_static(&sifive_prci_info);
+}
+
+type_init(sifive_prci_register_types)
+
+
+/* Create PRCI device */
+DeviceState *sifive_u_prci_create(hwaddr addr)
+{
+    DeviceState *dev = qdev_create(NULL, TYPE_SIFIVE_U_PRCI);
+    qdev_init_nofail(dev);
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, addr);
+    return dev;
+}
include/hw/riscv/sifive_u_prci.h new file +61 -0
new file mode 100644
index 0000000..f3a4656
@@ -0,0 +1,90 @@
+/*
+ * QEMU SiFive U PRCI (Power, Reset, Clock, Interrupt) interface
+ *
+ * Copyright (c) 2019 Bin Meng <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2 or later, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef HW_SIFIVE_U_PRCI_H
+#define HW_SIFIVE_U_PRCI_H
+
+enum {
+    SIFIVE_PRCI_HFXOSCCFG       = 0x00,
+    SIFIVE_PRCI_COREPLLCFG0     = 0x04,
+    SIFIVE_PRCI_DDRPLLCFG0      = 0x0C,
+    SIFIVE_PRCI_DDRPLLCFG1      = 0x10,
+    SIFIVE_PRCI_GEMGXLPLLCFG0   = 0x1C,
+    SIFIVE_PRCI_GEMGXLPLLCFG1   = 0x20,
+    SIFIVE_PRCI_CORECLKSEL      = 0x24,
+    SIFIVE_PRCI_DEVICESRESET    = 0x28,
+    SIFIVE_PRCI_CLKMUXSTATUS    = 0x2C
+};
+
+/*
+ * Current FU540-C000 manual says ready bit is at bit 29, but
+ * freedom-u540-c000-bootloader codes (ux00prci.h) says it is at bit 31.
+ * We have to trust the actual codes that worked.
+ *
+ * see https://github.com/sifive/freedom-u540-c000-bootloader
+ */
+enum {
+    SIFIVE_PRCI_HFXOSCCFG_EN    = (1 << 30),
+    SIFIVE_PRCI_HFXOSCCFG_RDY   = (1 << 31),
+};
+
+/* xxxPLLCFG0 register bits */
+enum {
+    SIFIVE_PRCI_PLLCFG0_DIVR    = (1 << 0),
+    SIFIVE_PRCI_PLLCFG0_DIVF    = (31 << 6),
+    SIFIVE_PRCI_PLLCFG0_DIVQ    = (3 << 15),
+    SIFIVE_PRCI_PLLCFG0_FSE     = (1 << 25),
+    SIFIVE_PRCI_PLLCFG0_LOCK    = (1 << 31)
+};
+
+/* xxxPLLCFG1 register bits */
+enum {
+    SIFIVE_PRCI_PLLCFG1_CKE     = (1 << 24)
+};
+
+enum {
+    SIFIVE_PRCI_CORECLKSEL_HFCLK = (1 << 0)

include/hw/riscv/sifive_u_prci.h line 71

Quoted 10 lines
include/hw/riscv/sifive_u_prci.h new file +10 -0
+};
+
+#define SIFIVE_U_PRCI_REG_SIZE  0x1000
+
+#define TYPE_SIFIVE_U_PRCI      "riscv.sifive.u.prci"
+
+#define SIFIVE_U_PRCI(obj) \
+    OBJECT_CHECK(SiFivePRCIState, (obj), TYPE_SIFIVE_U_PRCI)
+
+typedef struct SiFivePRCIState {

This is a public struct and should be SiFive U.

Alistair

Quoted 23 lines
include/hw/riscv/sifive_u_prci.h new file +19 -1
+    /*< private >*/
+    SysBusDevice parent_obj;
+
+    /*< public >*/
+    MemoryRegion mmio;
+    uint32_t hfxosccfg;
+    uint32_t corepllcfg0;
+    uint32_t ddrpllcfg0;
+    uint32_t ddrpllcfg1;
+    uint32_t gemgxlpllcfg0;
+    uint32_t gemgxlpllcfg1;
+    uint32_t coreclksel;
+    uint32_t devicesreset;
+    uint32_t clkmuxstatus;
+} SiFivePRCIState;
+
+DeviceState *sifive_u_prci_create(hwaddr addr);
+
+#endif /* HW_SIFIVE_U_PRCI_H */
--

2.7.4

author
#21 lore raw

PATCH v3 15/28 riscv: sifive_u: Generate hfclk and rtcclk nodes

In reply to Bin Meng

To keep in sync with Linux kernel device tree, generate hfclk and rtcclk nodes in the device tree, to be referenced by PRCI node.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 23 +++++++++++++++++++++++
 include/hw/riscv/sifive_u.h |  2 ++
 2 files changed, 25 insertions(+)
hw/riscv/sifive_u.c +23 -0
index e071838..f2b711a 100644
@@ -77,6 +77,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     char *nodename;
     char ethclk_names[] = "pclk\0hclk\0tx_clk";
     uint32_t plic_phandle, ethclk_phandle, phandle = 1;
+    uint32_t hfclk_phandle, rtcclk_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
     if (!fdt) {
@@ -95,6 +96,28 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, "/soc", "#size-cells", 0x2);
     qemu_fdt_setprop_cell(fdt, "/soc", "#address-cells", 0x2);
 
+    hfclk_phandle = phandle++;
+    nodename = g_strdup_printf("/hfclk");
+    qemu_fdt_add_subnode(fdt, nodename);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", hfclk_phandle);
+    qemu_fdt_setprop_string(fdt, nodename, "clock-output-names", "hfclk");
+    qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
+        SIFIVE_U_HFCLK_FREQ);
+    qemu_fdt_setprop_string(fdt, nodename, "compatible", "fixed-clock");
+    qemu_fdt_setprop_cell(fdt, nodename, "#clock-cells", 0x0);
+    g_free(nodename);
+
+    rtcclk_phandle = phandle++;
+    nodename = g_strdup_printf("/rtcclk");
+    qemu_fdt_add_subnode(fdt, nodename);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", rtcclk_phandle);
+    qemu_fdt_setprop_string(fdt, nodename, "clock-output-names", "rtcclk");
+    qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
+        SIFIVE_U_RTCCLK_FREQ);
+    qemu_fdt_setprop_string(fdt, nodename, "compatible", "fixed-clock");
+    qemu_fdt_setprop_cell(fdt, nodename, "#clock-cells", 0x0);
+    g_free(nodename);
+
     nodename = g_strdup_printf("/memory@%lx",
         (long)memmap[SIFIVE_U_DRAM].base);
     qemu_fdt_add_subnode(fdt, nodename);
include/hw/riscv/sifive_u.h +2 -0
index d0d8528..2b57ffc 100644
@@ -65,6 +65,8 @@ enum {
 
 enum {
     SIFIVE_U_CLOCK_FREQ = 1000000000,
+    SIFIVE_U_HFCLK_FREQ = 33333333,
+    SIFIVE_U_RTCCLK_FREQ = 1000000,
     SIFIVE_U_GEM_CLOCK_FREQ = 125000000
 };
 
Signature · 1 line
2.7.4
author
#22 lore raw

PATCH v3 16/28 riscv: sifive_u: Add PRCI block to the SoC

In reply to Bin Meng

Add PRCI mmio base address and size mappings to sifive_u machine, and generate the corresponding device tree node.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 21 ++++++++++++++++++++-
 include/hw/riscv/sifive_u.h |  1 +
 2 files changed, 21 insertions(+), 1 deletion(-)
hw/riscv/sifive_u.c +20 -1
index f2b711a..c0b7498 100644
@@ -9,6 +9,7 @@
  * 0) UART
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
+ * 3) PRCI (Power, Reset, Clock, Interrupt)
  *
  * This board currently generates devicetree dynamically that indicates at least
  * two harts and up to five harts.
@@ -42,6 +43,7 @@
 #include "hw/riscv/sifive_clint.h"
 #include "hw/riscv/sifive_uart.h"
 #include "hw/riscv/sifive_u.h"
+#include "hw/riscv/sifive_u_prci.h"
 #include "hw/riscv/boot.h"
 #include "chardev/char.h"
 #include "sysemu/arch_init.h"
@@ -60,6 +62,7 @@ static const struct MemmapEntry {
     [SIFIVE_U_MROM] =     {     0x1000,    0x11000 },
     [SIFIVE_U_CLINT] =    {  0x2000000,    0x10000 },
     [SIFIVE_U_PLIC] =     {  0xc000000,  0x4000000 },
+    [SIFIVE_U_PRCI] =     { 0x10000000,     0x1000 },
     [SIFIVE_U_UART0] =    { 0x10010000,     0x1000 },
     [SIFIVE_U_UART1] =    { 0x10011000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
@@ -76,7 +79,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     uint32_t *cells;
     char *nodename;
     char ethclk_names[] = "pclk\0hclk\0tx_clk";
-    uint32_t plic_phandle, ethclk_phandle, phandle = 1;
+    uint32_t plic_phandle, prci_phandle, ethclk_phandle, phandle = 1;
     uint32_t hfclk_phandle, rtcclk_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
@@ -183,6 +186,21 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     g_free(cells);
     g_free(nodename);
 
+    prci_phandle = phandle++;
+    nodename = g_strdup_printf("/soc/clock-controller@%lx",
+        (long)memmap[SIFIVE_U_PRCI].base);
+    qemu_fdt_add_subnode(fdt, nodename);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", prci_phandle);
+    qemu_fdt_setprop_cell(fdt, nodename, "#clock-cells", 0x1);
+    qemu_fdt_setprop_cells(fdt, nodename, "clocks",
+        hfclk_phandle, rtcclk_phandle);
+    qemu_fdt_setprop_cells(fdt, nodename, "reg",
+        0x0, memmap[SIFIVE_U_PRCI].base,
+        0x0, memmap[SIFIVE_U_PRCI].size);
+    qemu_fdt_setprop_string(fdt, nodename, "compatible",
+        "sifive,fu540-c000-prci");
+    g_free(nodename);
+
     plic_phandle = phandle++;
     cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4 - 2);
     for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
@@ -422,6 +440,7 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
     sifive_clint_create(memmap[SIFIVE_U_CLINT].base,
         memmap[SIFIVE_U_CLINT].size, ms->smp.cpus,
         SIFIVE_SIP_BASE, SIFIVE_TIMECMP_BASE, SIFIVE_TIME_BASE);
+    sifive_u_prci_create(memmap[SIFIVE_U_PRCI].base);
 
     for (i = 0; i < SIFIVE_U_PLIC_NUM_SOURCES; i++) {
         plic_gpios[i] = qdev_get_gpio_in(DEVICE(s->plic), i);
include/hw/riscv/sifive_u.h +1 -0
index 2b57ffc..e318ecb 100644
@@ -51,6 +51,7 @@ enum {
     SIFIVE_U_MROM,
     SIFIVE_U_CLINT,
     SIFIVE_U_PLIC,
+    SIFIVE_U_PRCI,
     SIFIVE_U_UART0,
     SIFIVE_U_UART1,
     SIFIVE_U_DRAM,
Signature · 1 line
2.7.4
author
#23 lore raw

PATCH v3 17/28 riscv: sifive_u: Change UART node name in device tree

In reply to Bin Meng

OpenSBI for fu540 does DT fix up (see fu540_modify_dt()) by updating chosen "stdout-path" to point to "/soc/serial@...", and U-Boot will use this information to locate the serial node and probe its driver. However currently we generate the UART node name as "/soc/uart@...", causing U-Boot fail to find the serial node in DT.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/riscv/sifive_u.c +1 -1
index c0b7498..5022b8f 100644
@@ -274,7 +274,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "reg", 0x0);
     g_free(nodename);
 
-    nodename = g_strdup_printf("/soc/uart@%lx",
+    nodename = g_strdup_printf("/soc/serial@%lx",
         (long)memmap[SIFIVE_U_UART0].base);
     qemu_fdt_add_subnode(fdt, nodename);
     qemu_fdt_setprop_string(fdt, nodename, "compatible", "sifive,uart0");
Signature · 1 line
2.7.4
author
#24 lore raw

PATCH v3 18/28 riscv: hw: Implement a model for SiFive FU540 OTP

In reply to Bin Meng

This implements a simple model for SiFive FU540 OTP (One-Time Programmable) Memory interface, primarily for reading out the stored serial number from the first 1 KiB of the 16 KiB OTP memory reserved by SiFive for internal use.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs          |   1 +
 hw/riscv/sifive_u_otp.c         | 194 ++++++++++++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_u_otp.h |  90 +++++++++++++++++++
 3 files changed, 285 insertions(+)
 create mode 100644 hw/riscv/sifive_u_otp.c
 create mode 100644 include/hw/riscv/sifive_u_otp.h
hw/riscv/Makefile.objs +1 -0
index b95bbd5..fc3c6dd 100644
@@ -8,6 +8,7 @@ obj-$(CONFIG_SIFIVE) += sifive_gpio.o
 obj-$(CONFIG_SIFIVE) += sifive_plic.o
 obj-$(CONFIG_SIFIVE) += sifive_test.o
 obj-$(CONFIG_SIFIVE_U) += sifive_u.o
+obj-$(CONFIG_SIFIVE_U) += sifive_u_otp.o
 obj-$(CONFIG_SIFIVE_U) += sifive_u_prci.o
 obj-$(CONFIG_SIFIVE) += sifive_uart.o
 obj-$(CONFIG_SPIKE) += spike.o
hw/riscv/sifive_u_otp.c new file +194 -0

194 lines not on this page · raw message

include/hw/riscv/sifive_u_otp.h new file +90 -0

90 lines not on this page · raw message

Signature · 1 line
2.7.4
author
#25 lore raw

PATCH v3 19/28 riscv: sifive_u: Instantiate OTP memory with a serial number

In reply to Bin Meng

This adds an OTP memory with a given serial number to the sifive_u machine. With such support, the upstream U-Boot for sifive_fu540 boots out of the box on the sifive_u machine.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 5 +++++
 include/hw/riscv/sifive_u.h | 1 +
 2 files changed, 6 insertions(+)
hw/riscv/sifive_u.c +5 -0
index 5022b8f..486b247 100644
@@ -10,6 +10,7 @@
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
  * 3) PRCI (Power, Reset, Clock, Interrupt)
+ * 4) OTP (One-Time Programmable) memory with stored serial number
  *
  * This board currently generates devicetree dynamically that indicates at least
  * two harts and up to five harts.
@@ -43,6 +44,7 @@
 #include "hw/riscv/sifive_clint.h"
 #include "hw/riscv/sifive_uart.h"
 #include "hw/riscv/sifive_u.h"
+#include "hw/riscv/sifive_u_otp.h"
 #include "hw/riscv/sifive_u_prci.h"
 #include "hw/riscv/boot.h"
 #include "chardev/char.h"
@@ -65,10 +67,12 @@ static const struct MemmapEntry {
     [SIFIVE_U_PRCI] =     { 0x10000000,     0x1000 },
     [SIFIVE_U_UART0] =    { 0x10010000,     0x1000 },
     [SIFIVE_U_UART1] =    { 0x10011000,     0x1000 },
+    [SIFIVE_U_OTP] =      { 0x10070000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
     [SIFIVE_U_GEM] =      { 0x100900FC,     0x2000 },
 };
 
+#define SIFIVE_OTP_SERIAL   1
 #define GEM_REVISION        0x10070109
 
 static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
@@ -441,6 +445,7 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
         memmap[SIFIVE_U_CLINT].size, ms->smp.cpus,
         SIFIVE_SIP_BASE, SIFIVE_TIMECMP_BASE, SIFIVE_TIME_BASE);
     sifive_u_prci_create(memmap[SIFIVE_U_PRCI].base);
+    sifive_u_otp_create(memmap[SIFIVE_U_OTP].base, SIFIVE_OTP_SERIAL);
 
     for (i = 0; i < SIFIVE_U_PLIC_NUM_SOURCES; i++) {
         plic_gpios[i] = qdev_get_gpio_in(DEVICE(s->plic), i);
include/hw/riscv/sifive_u.h +1 -0
index e318ecb..3ae75b5 100644
@@ -54,6 +54,7 @@ enum {
     SIFIVE_U_PRCI,
     SIFIVE_U_UART0,
     SIFIVE_U_UART1,
+    SIFIVE_U_OTP,
     SIFIVE_U_DRAM,
     SIFIVE_U_GEM
 };
Signature · 1 line
2.7.4
Reviewed-by on v3 19/28
#26 lore raw

Re: [Qemu-devel] [PATCH v3 19/28] riscv: sifive_u: Instantiate OTP memory with a serial number

Quoted 6 lines · Bin Meng

This adds an OTP memory with a given serial number to the sifive_u machine. With such support, the upstream U-Boot for sifive_fu540 boots out of the box on the sifive_u machine.

Alistair

Quoted 66 lines

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 5 +++++
 include/hw/riscv/sifive_u.h | 1 +
 2 files changed, 6 insertions(+)
hw/riscv/sifive_u.c +5 -0
index 5022b8f..486b247 100644
@@ -10,6 +10,7 @@
  * 1) CLINT (Core Level Interruptor)
  * 2) PLIC (Platform Level Interrupt Controller)
  * 3) PRCI (Power, Reset, Clock, Interrupt)
+ * 4) OTP (One-Time Programmable) memory with stored serial number
  *
  * This board currently generates devicetree dynamically that indicates at least
  * two harts and up to five harts.
@@ -43,6 +44,7 @@
 #include "hw/riscv/sifive_clint.h"
 #include "hw/riscv/sifive_uart.h"
 #include "hw/riscv/sifive_u.h"
+#include "hw/riscv/sifive_u_otp.h"
 #include "hw/riscv/sifive_u_prci.h"
 #include "hw/riscv/boot.h"
 #include "chardev/char.h"
@@ -65,10 +67,12 @@ static const struct MemmapEntry {
     [SIFIVE_U_PRCI] =     { 0x10000000,     0x1000 },
     [SIFIVE_U_UART0] =    { 0x10010000,     0x1000 },
     [SIFIVE_U_UART1] =    { 0x10011000,     0x1000 },
+    [SIFIVE_U_OTP] =      { 0x10070000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
     [SIFIVE_U_GEM] =      { 0x100900FC,     0x2000 },
 };
 
+#define SIFIVE_OTP_SERIAL   1
 #define GEM_REVISION        0x10070109
 
 static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
@@ -441,6 +445,7 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
         memmap[SIFIVE_U_CLINT].size, ms->smp.cpus,
         SIFIVE_SIP_BASE, SIFIVE_TIMECMP_BASE, SIFIVE_TIME_BASE);
     sifive_u_prci_create(memmap[SIFIVE_U_PRCI].base);
+    sifive_u_otp_create(memmap[SIFIVE_U_OTP].base, SIFIVE_OTP_SERIAL);
 
     for (i = 0; i < SIFIVE_U_PLIC_NUM_SOURCES; i++) {
         plic_gpios[i] = qdev_get_gpio_in(DEVICE(s->plic), i);
include/hw/riscv/sifive_u.h +1 -1
index e318ecb..3ae75b5 100644
@@ -54,6 +54,7 @@ enum {
     SIFIVE_U_PRCI,
     SIFIVE_U_UART0,
     SIFIVE_U_UART1,
+    SIFIVE_U_OTP,
     SIFIVE_U_DRAM,
     SIFIVE_U_GEM
 };
--

2.7.4

author
#27 lore raw

PATCH v3 21/28 riscv: sifive_u: Update UART and ethernet node clock properties

In reply to Bin Meng

Now that we have added PRCI nodes, update existing UART and ethernet nodes to use PRCI as their clock sources, to keep in sync with the Linux kernel device tree.

With above changes, the previously handcrafted "/soc/ethclk" node is no longer needed. Remove it.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c              | 21 +++++----------------
 include/hw/riscv/sifive_u.h      |  3 +--
 include/hw/riscv/sifive_u_prci.h | 10 ++++++++++
 3 files changed, 16 insertions(+), 18 deletions(-)
hw/riscv/sifive_u.c +5 -16
index 486b247..7eb2b7e 100644
@@ -82,8 +82,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     int cpu;
     uint32_t *cells;
     char *nodename;
-    char ethclk_names[] = "pclk\0hclk\0tx_clk";
-    uint32_t plic_phandle, prci_phandle, ethclk_phandle, phandle = 1;
+    char ethclk_names[] = "pclk\0hclk";
+    uint32_t plic_phandle, prci_phandle, phandle = 1;
     uint32_t hfclk_phandle, rtcclk_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
@@ -242,17 +242,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     g_free(cells);
     g_free(nodename);
 
-    ethclk_phandle = phandle++;
-    nodename = g_strdup_printf("/soc/ethclk");
-    qemu_fdt_add_subnode(fdt, nodename);
-    qemu_fdt_setprop_string(fdt, nodename, "compatible", "fixed-clock");
-    qemu_fdt_setprop_cell(fdt, nodename, "#clock-cells", 0x0);
-    qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
-        SIFIVE_U_GEM_CLOCK_FREQ);
-    qemu_fdt_setprop_cell(fdt, nodename, "phandle", ethclk_phandle);
-    ethclk_phandle = qemu_fdt_get_phandle(fdt, nodename);
-    g_free(nodename);
-
     nodename = g_strdup_printf("/soc/ethernet@%lx",
         (long)memmap[SIFIVE_U_GEM].base);
     qemu_fdt_add_subnode(fdt, nodename);
@@ -265,7 +254,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupts", SIFIVE_U_GEM_IRQ);
     qemu_fdt_setprop_cells(fdt, nodename, "clocks",
-        ethclk_phandle, ethclk_phandle, ethclk_phandle);
+        prci_phandle, PRCI_CLK_GEMGXLPLL, prci_phandle, PRCI_CLK_GEMGXLPLL);
     qemu_fdt_setprop(fdt, nodename, "clocks-names", ethclk_names,
         sizeof(ethclk_names));
     qemu_fdt_setprop_cell(fdt, nodename, "#address-cells", 1);
@@ -285,8 +274,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_UART0].base,
         0x0, memmap[SIFIVE_U_UART0].size);
-    qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency",
-                          SIFIVE_U_CLOCK_FREQ / 2);
+    qemu_fdt_setprop_cells(fdt, nodename, "clocks",
+        prci_phandle, PRCI_CLK_TLCLK);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupts", SIFIVE_U_UART0_IRQ);
 
include/hw/riscv/sifive_u.h +1 -2
index 3ae75b5..2a7877e 100644
@@ -68,8 +68,7 @@ enum {
 enum {
     SIFIVE_U_CLOCK_FREQ = 1000000000,
     SIFIVE_U_HFCLK_FREQ = 33333333,
-    SIFIVE_U_RTCCLK_FREQ = 1000000,
-    SIFIVE_U_GEM_CLOCK_FREQ = 125000000
+    SIFIVE_U_RTCCLK_FREQ = 1000000
 };
 
 #define SIFIVE_U_MANAGEMENT_CPU_COUNT   1
include/hw/riscv/sifive_u_prci.h +10 -0
index f3a4656..640c641 100644
@@ -87,4 +87,14 @@ typedef struct SiFivePRCIState {
 
 DeviceState *sifive_u_prci_create(hwaddr addr);
 
+/*
+ * Clock indexes for use by Device Tree data and the PRCI driver.
+ *
+ * These values are from sifive-fu540-prci.h in the Linux kernel.
+ */
+#define PRCI_CLK_COREPLL        0
+#define PRCI_CLK_DDRPLL         1
+#define PRCI_CLK_GEMGXLPLL      2
+#define PRCI_CLK_TLCLK          3
+
 #endif /* HW_SIFIVE_U_PRCI_H */
Signature · 1 line
2.7.4
author
#28 lore raw

PATCH v3 22/28 riscv: sifive_u: Generate an aliases node in the device tree

In reply to Bin Meng

The Linux kernel SiFive UART driver expects an aliases node to be present in the device tree, from which the driver extracts the port number from "serial#" in the aliases node.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 2 ++
 1 file changed, 2 insertions(+)
hw/riscv/sifive_u.c +2 -0
index 7eb2b7e..0c1a89f 100644
@@ -284,6 +284,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     if (cmdline) {
         qemu_fdt_setprop_string(fdt, "/chosen", "bootargs", cmdline);
     }
+    qemu_fdt_add_subnode(fdt, "/aliases");
+    qemu_fdt_setprop_string(fdt, "/aliases", "serial0", nodename);
     g_free(nodename);
 }
 
Signature · 1 line
2.7.4
Reviewed-by on v3 22/28
#29 lore raw

Re: [Qemu-devel] [PATCH v3 22/28] riscv: sifive_u: Generate an aliases node in the device tree

Quoted 6 lines · Bin Meng

The Linux kernel SiFive UART driver expects an aliases node to be present in the device tree, from which the driver extracts the port number from "serial#" in the aliases node.

Alistair

Quoted 25 lines

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 2 ++
 1 file changed, 2 insertions(+)
hw/riscv/sifive_u.c +2 -1
index 7eb2b7e..0c1a89f 100644
@@ -284,6 +284,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     if (cmdline) {
         qemu_fdt_setprop_string(fdt, "/chosen", "bootargs", cmdline);
     }
+    qemu_fdt_add_subnode(fdt, "/aliases");
+    qemu_fdt_setprop_string(fdt, "/aliases", "serial0", nodename);
     g_free(nodename);
 }
 
--

2.7.4

author
#30 lore raw

PATCH v3 23/28 riscv: sifive_u: Fix broken GEM support

In reply to Bin Meng

At present the GEM support in sifive_u machine is seriously broken.

- The GEM block register base was set to a weird number (0x100900FC), which for no way could work with the cadence_gem model in QEMU.
- The generated DT node for GEM has a "clocks-names" which is an invalid property name.

Not like other GEM variants, the FU540-specific GEM has a management block to control 10/100/1000Mbps link speed changes, that is mapped to 0x100a0000. We can simply map it into MMIO space without special handling using create_unimplemented_device().

Update the GEM node compatible string to use the official name used by the upstream Linux kernel, and add the management block reg base & size to the <reg> property encoding.

Tested with upstream U-Boot and Linux kernel MACB drivers.

---

Changes in v3: None
Changes in v2:
- use create_unimplemented_device() to create the GEM management block instead of sifive_mmio_emulate()
- add "phy-handle" property to the ethernet node

 hw/riscv/sifive_u.c         | 23 ++++++++++++++++++-----
 include/hw/riscv/sifive_u.h |  3 ++-
 2 files changed, 20 insertions(+), 6 deletions(-)
hw/riscv/sifive_u.c +18 -5
index 0c1a89f..e8bef44 100644
@@ -3,6 +3,7 @@
  *
  * Copyright (c) 2016-2017 Sagar Karandikar, [email protected]
  * Copyright (c) 2017 SiFive, Inc.
+ * Copyright (c) 2019 Bin Meng <[email protected]>
  *
  * Provides a board compatible with the SiFive Freedom U SDK:
  *
@@ -11,6 +12,7 @@
  * 2) PLIC (Platform Level Interrupt Controller)
  * 3) PRCI (Power, Reset, Clock, Interrupt)
  * 4) OTP (One-Time Programmable) memory with stored serial number
+ * 5) GEM (Gigabit Ethernet Controller) and management block
  *
  * This board currently generates devicetree dynamically that indicates at least
  * two harts and up to five harts.
@@ -38,6 +40,7 @@
 #include "hw/loader.h"
 #include "hw/sysbus.h"
 #include "hw/char/serial.h"
+#include "hw/misc/unimp.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
@@ -69,7 +72,8 @@ static const struct MemmapEntry {
     [SIFIVE_U_UART1] =    { 0x10011000,     0x1000 },
     [SIFIVE_U_OTP] =      { 0x10070000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
-    [SIFIVE_U_GEM] =      { 0x100900FC,     0x2000 },
+    [SIFIVE_U_GEM] =      { 0x10090000,     0x2000 },
+    [SIFIVE_U_GEM_MGMT] = { 0x100a0000,     0x1000 },
 };
 
 #define SIFIVE_OTP_SERIAL   1
@@ -84,7 +88,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     char *nodename;
     char ethclk_names[] = "pclk\0hclk";
     uint32_t plic_phandle, prci_phandle, phandle = 1;
-    uint32_t hfclk_phandle, rtcclk_phandle;
+    uint32_t hfclk_phandle, rtcclk_phandle, phy_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
     if (!fdt) {
@@ -242,20 +246,25 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     g_free(cells);
     g_free(nodename);
 
+    phy_phandle = phandle++;
     nodename = g_strdup_printf("/soc/ethernet@%lx",
         (long)memmap[SIFIVE_U_GEM].base);
     qemu_fdt_add_subnode(fdt, nodename);
-    qemu_fdt_setprop_string(fdt, nodename, "compatible", "cdns,macb");
+    qemu_fdt_setprop_string(fdt, nodename, "compatible",
+        "sifive,fu540-c000-gem");
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_GEM].base,
-        0x0, memmap[SIFIVE_U_GEM].size);
+        0x0, memmap[SIFIVE_U_GEM].size,
+        0x0, memmap[SIFIVE_U_GEM_MGMT].base,
+        0x0, memmap[SIFIVE_U_GEM_MGMT].size);
     qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
     qemu_fdt_setprop_string(fdt, nodename, "phy-mode", "gmii");
+    qemu_fdt_setprop_cell(fdt, nodename, "phy-handle", phy_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupts", SIFIVE_U_GEM_IRQ);
     qemu_fdt_setprop_cells(fdt, nodename, "clocks",
         prci_phandle, PRCI_CLK_GEMGXLPLL, prci_phandle, PRCI_CLK_GEMGXLPLL);
-    qemu_fdt_setprop(fdt, nodename, "clocks-names", ethclk_names,
+    qemu_fdt_setprop(fdt, nodename, "clock-names", ethclk_names,
         sizeof(ethclk_names));
     qemu_fdt_setprop_cell(fdt, nodename, "#address-cells", 1);
     qemu_fdt_setprop_cell(fdt, nodename, "#size-cells", 0);
@@ -264,6 +273,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     nodename = g_strdup_printf("/soc/ethernet@%lx/ethernet-phy@0",
         (long)memmap[SIFIVE_U_GEM].base);
     qemu_fdt_add_subnode(fdt, nodename);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", phy_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "reg", 0x0);
     g_free(nodename);
 
@@ -456,6 +466,9 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
     sysbus_mmio_map(SYS_BUS_DEVICE(&s->gem), 0, memmap[SIFIVE_U_GEM].base);
     sysbus_connect_irq(SYS_BUS_DEVICE(&s->gem), 0,
                        plic_gpios[SIFIVE_U_GEM_IRQ]);
+
+    create_unimplemented_device("riscv.sifive.u.gem-mgmt",
+        memmap[SIFIVE_U_GEM_MGMT].base, memmap[SIFIVE_U_GEM_MGMT].size);
 }
 
 static void riscv_sifive_u_machine_init(MachineClass *mc)
include/hw/riscv/sifive_u.h +2 -1
index 2a7877e..27b87cf 100644
@@ -56,7 +56,8 @@ enum {
     SIFIVE_U_UART1,
     SIFIVE_U_OTP,
     SIFIVE_U_DRAM,
-    SIFIVE_U_GEM
+    SIFIVE_U_GEM,
+    SIFIVE_U_GEM_MGMT
 };
 
 enum {
Signature · 1 line
2.7.4
Reviewed-by on v3 23/28
#31 lore raw

Re: [Qemu-devel] [PATCH v3 23/28] riscv: sifive_u: Fix broken GEM support

Quoted 13 lines · Bin Meng

At present the GEM support in sifive_u machine is seriously broken.

- The GEM block register base was set to a weird number (0x100900FC), which for no way could work with the cadence_gem model in QEMU.
- The generated DT node for GEM has a "clocks-names" which is an invalid property name.

Not like other GEM variants, the FU540-specific GEM has a management block to control 10/100/1000Mbps link speed changes, that is mapped to 0x100a0000. We can simply map it into MMIO space without special handling using create_unimplemented_device().

Quoted 7 lines

Update the GEM node compatible string to use the official name used by the upstream Linux kernel, and add the management block reg base & size to the <reg> property encoding.

Tested with upstream U-Boot and Linux kernel MACB drivers.

Alistair

Quoted 125 lines

---

Changes in v3: None
Changes in v2:
- use create_unimplemented_device() to create the GEM management block instead of sifive_mmio_emulate()
- add "phy-handle" property to the ethernet node

 hw/riscv/sifive_u.c         | 23 ++++++++++++++++++-----
 include/hw/riscv/sifive_u.h |  3 ++-
 2 files changed, 20 insertions(+), 6 deletions(-)
hw/riscv/sifive_u.c +18 -5
index 0c1a89f..e8bef44 100644
@@ -3,6 +3,7 @@
  *
  * Copyright (c) 2016-2017 Sagar Karandikar, [email protected]
  * Copyright (c) 2017 SiFive, Inc.
+ * Copyright (c) 2019 Bin Meng <[email protected]>
  *
  * Provides a board compatible with the SiFive Freedom U SDK:
  *
@@ -11,6 +12,7 @@
  * 2) PLIC (Platform Level Interrupt Controller)
  * 3) PRCI (Power, Reset, Clock, Interrupt)
  * 4) OTP (One-Time Programmable) memory with stored serial number
+ * 5) GEM (Gigabit Ethernet Controller) and management block
  *
  * This board currently generates devicetree dynamically that indicates at least
  * two harts and up to five harts.
@@ -38,6 +40,7 @@
 #include "hw/loader.h"
 #include "hw/sysbus.h"
 #include "hw/char/serial.h"
+#include "hw/misc/unimp.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
@@ -69,7 +72,8 @@ static const struct MemmapEntry {
     [SIFIVE_U_UART1] =    { 0x10011000,     0x1000 },
     [SIFIVE_U_OTP] =      { 0x10070000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
-    [SIFIVE_U_GEM] =      { 0x100900FC,     0x2000 },
+    [SIFIVE_U_GEM] =      { 0x10090000,     0x2000 },
+    [SIFIVE_U_GEM_MGMT] = { 0x100a0000,     0x1000 },
 };
 
 #define SIFIVE_OTP_SERIAL   1
@@ -84,7 +88,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     char *nodename;
     char ethclk_names[] = "pclk\0hclk";
     uint32_t plic_phandle, prci_phandle, phandle = 1;
-    uint32_t hfclk_phandle, rtcclk_phandle;
+    uint32_t hfclk_phandle, rtcclk_phandle, phy_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
     if (!fdt) {
@@ -242,20 +246,25 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     g_free(cells);
     g_free(nodename);
 
+    phy_phandle = phandle++;
     nodename = g_strdup_printf("/soc/ethernet@%lx",
         (long)memmap[SIFIVE_U_GEM].base);
     qemu_fdt_add_subnode(fdt, nodename);
-    qemu_fdt_setprop_string(fdt, nodename, "compatible", "cdns,macb");
+    qemu_fdt_setprop_string(fdt, nodename, "compatible",
+        "sifive,fu540-c000-gem");
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_GEM].base,
-        0x0, memmap[SIFIVE_U_GEM].size);
+        0x0, memmap[SIFIVE_U_GEM].size,
+        0x0, memmap[SIFIVE_U_GEM_MGMT].base,
+        0x0, memmap[SIFIVE_U_GEM_MGMT].size);
     qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
     qemu_fdt_setprop_string(fdt, nodename, "phy-mode", "gmii");
+    qemu_fdt_setprop_cell(fdt, nodename, "phy-handle", phy_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupt-parent", plic_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "interrupts", SIFIVE_U_GEM_IRQ);
     qemu_fdt_setprop_cells(fdt, nodename, "clocks",
         prci_phandle, PRCI_CLK_GEMGXLPLL, prci_phandle, PRCI_CLK_GEMGXLPLL);
-    qemu_fdt_setprop(fdt, nodename, "clocks-names", ethclk_names,
+    qemu_fdt_setprop(fdt, nodename, "clock-names", ethclk_names,
         sizeof(ethclk_names));
     qemu_fdt_setprop_cell(fdt, nodename, "#address-cells", 1);
     qemu_fdt_setprop_cell(fdt, nodename, "#size-cells", 0);
@@ -264,6 +273,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     nodename = g_strdup_printf("/soc/ethernet@%lx/ethernet-phy@0",
         (long)memmap[SIFIVE_U_GEM].base);
     qemu_fdt_add_subnode(fdt, nodename);
+    qemu_fdt_setprop_cell(fdt, nodename, "phandle", phy_phandle);
     qemu_fdt_setprop_cell(fdt, nodename, "reg", 0x0);
     g_free(nodename);
 
@@ -456,6 +466,9 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
     sysbus_mmio_map(SYS_BUS_DEVICE(&s->gem), 0, memmap[SIFIVE_U_GEM].base);
     sysbus_connect_irq(SYS_BUS_DEVICE(&s->gem), 0,
                        plic_gpios[SIFIVE_U_GEM_IRQ]);
+
+    create_unimplemented_device("riscv.sifive.u.gem-mgmt",
+        memmap[SIFIVE_U_GEM_MGMT].base, memmap[SIFIVE_U_GEM_MGMT].size);
 }
 
 static void riscv_sifive_u_machine_init(MachineClass *mc)
include/hw/riscv/sifive_u.h +2 -2
index 2a7877e..27b87cf 100644
@@ -56,7 +56,8 @@ enum {
     SIFIVE_U_UART1,
     SIFIVE_U_OTP,
     SIFIVE_U_DRAM,
-    SIFIVE_U_GEM
+    SIFIVE_U_GEM,
+    SIFIVE_U_GEM_MGMT
 };
 
 enum {
--

2.7.4

author
#32 lore raw

PATCH v3 24/28 riscv: sifive_u: Support loading initramfs

In reply to Bin Meng

The loading of initramfs is currently not supported on 'sifive_u'. Add the support to make '-initrd' command line parameter useful.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
hw/riscv/sifive_u.c +12 -1
index e8bef44..b0026aa 100644
@@ -328,7 +328,18 @@ static void riscv_sifive_u_init(MachineState *machine)
                                  memmap[SIFIVE_U_DRAM].base);
 
     if (machine->kernel_filename) {
-        riscv_load_kernel(machine->kernel_filename);
+        uint64_t kernel_entry = riscv_load_kernel(machine->kernel_filename);
+
+        if (machine->initrd_filename) {
+            hwaddr start;
+            hwaddr end = riscv_load_initrd(machine->initrd_filename,
+                                           machine->ram_size, kernel_entry,
+                                           &start);
+            qemu_fdt_setprop_cell(s->fdt, "/chosen",
+                                  "linux,initrd-start", start);
+            qemu_fdt_setprop_cell(s->fdt, "/chosen", "linux,initrd-end",
+                                  end);
+        }
     }
 
     /* reset vector */
Signature · 1 line
2.7.4
Reviewed-by on v3 24/28
#33 lore raw

Re: [Qemu-devel] [PATCH v3 24/28] riscv: sifive_u: Support loading initramfs

Quoted 6 lines · Bin Meng

The loading of initramfs is currently not supported on 'sifive_u'. Add the support to make '-initrd' command line parameter useful.

Alistair

Quoted 36 lines

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
hw/riscv/sifive_u.c +12 -2
index e8bef44..b0026aa 100644
@@ -328,7 +328,18 @@ static void riscv_sifive_u_init(MachineState *machine)
                                  memmap[SIFIVE_U_DRAM].base);
 
     if (machine->kernel_filename) {
-        riscv_load_kernel(machine->kernel_filename);
+        uint64_t kernel_entry = riscv_load_kernel(machine->kernel_filename);
+
+        if (machine->initrd_filename) {
+            hwaddr start;
+            hwaddr end = riscv_load_initrd(machine->initrd_filename,
+                                           machine->ram_size, kernel_entry,
+                                           &start);
+            qemu_fdt_setprop_cell(s->fdt, "/chosen",
+                                  "linux,initrd-start", start);
+            qemu_fdt_setprop_cell(s->fdt, "/chosen", "linux,initrd-end",
+                                  end);
+        }
     }
 
     /* reset vector */
--

2.7.4

author
#34 lore raw

PATCH v3 25/28 riscv: hw: Remove not needed PLIC properties in device tree

In reply to Bin Meng

This removes "reg-names" and "riscv,max-priority" properties of the PLIC node from device tree.

---

Changes in v3: None
Changes in v2:
- keep the PLIC compatible string unchanged as OpenSBI uses that for DT fix up

 hw/riscv/sifive_u.c | 2 --
 hw/riscv/virt.c     | 2 --
 2 files changed, 4 deletions(-)
hw/riscv/sifive_u.c +0 -2
index b0026aa..8801ee2 100644
@@ -238,8 +238,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_PLIC].base,
         0x0, memmap[SIFIVE_U_PLIC].size);
-    qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
-    qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", 0x35);
     qemu_fdt_setprop_cell(fdt, nodename, "phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
hw/riscv/virt.c +0 -2
index 127f005..2f75195 100644
@@ -244,8 +244,6 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[VIRT_PLIC].base,
         0x0, memmap[VIRT_PLIC].size);
-    qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
-    qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", VIRTIO_NDEV);
     qemu_fdt_setprop_cell(fdt, nodename, "phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
Signature · 1 line
2.7.4
Reviewed-by on v3 25/28
#35 lore raw

Re: [Qemu-devel] [PATCH v3 25/28] riscv: hw: Remove not needed PLIC properties in device tree

Quoted 6 lines · Bin Meng

This removes "reg-names" and "riscv,max-priority" properties of the PLIC node from device tree.

Alistair

Quoted 41 lines

---

Changes in v3: None
Changes in v2:
- keep the PLIC compatible string unchanged as OpenSBI uses that for DT fix up

 hw/riscv/sifive_u.c | 2 --
 hw/riscv/virt.c     | 2 --
 2 files changed, 4 deletions(-)
hw/riscv/sifive_u.c +0 -2
index b0026aa..8801ee2 100644
@@ -238,8 +238,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_PLIC].base,
         0x0, memmap[SIFIVE_U_PLIC].size);
-    qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
-    qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", 0x35);
     qemu_fdt_setprop_cell(fdt, nodename, "phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
hw/riscv/virt.c +0 -3
index 127f005..2f75195 100644
@@ -244,8 +244,6 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[VIRT_PLIC].base,
         0x0, memmap[VIRT_PLIC].size);
-    qemu_fdt_setprop_string(fdt, nodename, "reg-names", "control");
-    qemu_fdt_setprop_cell(fdt, nodename, "riscv,max-priority", 7);
     qemu_fdt_setprop_cell(fdt, nodename, "riscv,ndev", VIRTIO_NDEV);
     qemu_fdt_setprop_cell(fdt, nodename, "phandle", plic_phandle);
     plic_phandle = qemu_fdt_get_phandle(fdt, nodename);
--

2.7.4

author
#36 lore raw

PATCH v3 27/28 riscv: virt: Change create_fdt() to return void

In reply to Bin Meng

There is no need to return fdt at the end of create_fdt() because it's already saved in s->fdt. Other machines (sifive_u, spike) don't do it neither.

---

Changes in v3: None
Changes in v2: None

 hw/riscv/virt.c | 11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)
hw/riscv/virt.c +4 -7
index 2f75195..6bfa721 100644
@@ -112,7 +112,7 @@ static void create_pcie_irq_map(void *fdt, char *nodename,
                            0x1800, 0, 0, 0x7);
 }
 
-static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
+static void create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     uint64_t mem_size, const char *cmdline)
 {
     void *fdt;
@@ -316,8 +316,6 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
         qemu_fdt_setprop_string(fdt, "/chosen", "bootargs", cmdline);
     }
     g_free(nodename);
-
-    return fdt;
 }
 
 
@@ -373,7 +371,6 @@ static void riscv_virt_board_init(MachineState *machine)
     size_t plic_hart_config_len;
     int i;
     unsigned int smp_cpus = machine->smp.cpus;
-    void *fdt;
 
     /* Initialize SOC */
     object_initialize_child(OBJECT(machine), "soc", &s->soc, sizeof(s->soc),
@@ -392,7 +389,7 @@ static void riscv_virt_board_init(MachineState *machine)
         main_mem);
 
     /* create device tree */
-    fdt = create_fdt(s, memmap, machine->ram_size, machine->kernel_cmdline);
+    create_fdt(s, memmap, machine->ram_size, machine->kernel_cmdline);
 
     /* boot rom */
     memory_region_init_rom(mask_rom, NULL, "riscv_virt_board.mrom",
@@ -411,9 +408,9 @@ static void riscv_virt_board_init(MachineState *machine)
             hwaddr end = riscv_load_initrd(machine->initrd_filename,
                                            machine->ram_size, kernel_entry,
                                            &start);
-            qemu_fdt_setprop_cell(fdt, "/chosen",
+            qemu_fdt_setprop_cell(s->fdt, "/chosen",
                                   "linux,initrd-start", start);
-            qemu_fdt_setprop_cell(fdt, "/chosen", "linux,initrd-end",
+            qemu_fdt_setprop_cell(s->fdt, "/chosen", "linux,initrd-end",
                                   end);
         }
     }
Signature · 1 line
2.7.4
Reviewed-by on v3 27/28
#37 lore raw

Re: [Qemu-devel] [PATCH v3 27/28] riscv: virt: Change create_fdt() to return void

Quoted 8 lines · Bin Meng

There is no need to return fdt at the end of create_fdt() because it's already saved in s->fdt. Other machines (sifive_u, spike) don't do it neither.

Alistair

Quoted 63 lines

---

Changes in v3: None
Changes in v2: None

 hw/riscv/virt.c | 11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)
hw/riscv/virt.c +4 -8
index 2f75195..6bfa721 100644
@@ -112,7 +112,7 @@ static void create_pcie_irq_map(void *fdt, char *nodename,
                            0x1800, 0, 0, 0x7);
 }
 
-static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
+static void create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
     uint64_t mem_size, const char *cmdline)
 {
     void *fdt;
@@ -316,8 +316,6 @@ static void *create_fdt(RISCVVirtState *s, const struct MemmapEntry *memmap,
         qemu_fdt_setprop_string(fdt, "/chosen", "bootargs", cmdline);
     }
     g_free(nodename);
-
-    return fdt;
 }
 
 
@@ -373,7 +371,6 @@ static void riscv_virt_board_init(MachineState *machine)
     size_t plic_hart_config_len;
     int i;
     unsigned int smp_cpus = machine->smp.cpus;
-    void *fdt;
 
     /* Initialize SOC */
     object_initialize_child(OBJECT(machine), "soc", &s->soc, sizeof(s->soc),
@@ -392,7 +389,7 @@ static void riscv_virt_board_init(MachineState *machine)
         main_mem);
 
     /* create device tree */
-    fdt = create_fdt(s, memmap, machine->ram_size, machine->kernel_cmdline);
+    create_fdt(s, memmap, machine->ram_size, machine->kernel_cmdline);
 
     /* boot rom */
     memory_region_init_rom(mask_rom, NULL, "riscv_virt_board.mrom",
@@ -411,9 +408,9 @@ static void riscv_virt_board_init(MachineState *machine)
             hwaddr end = riscv_load_initrd(machine->initrd_filename,
                                            machine->ram_size, kernel_entry,
                                            &start);
-            qemu_fdt_setprop_cell(fdt, "/chosen",
+            qemu_fdt_setprop_cell(s->fdt, "/chosen",
                                   "linux,initrd-start", start);
-            qemu_fdt_setprop_cell(fdt, "/chosen", "linux,initrd-end",
+            qemu_fdt_setprop_cell(s->fdt, "/chosen", "linux,initrd-end",
                                   end);
         }
     }
--

2.7.4

author
#38 lore raw

PATCH v3 28/28 riscv: sifive_u: Update model and compatible strings in device tree

In reply to Bin Meng

This updates model and compatible strings to use the same strings as used in the Linux kernel device tree (hifive-unleashed-a00.dts).

---

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/riscv/sifive_u.c +3 -2
index 8801ee2..98fefbeb 100644
@@ -96,8 +96,9 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         exit(1);
     }
 
-    qemu_fdt_setprop_string(fdt, "/", "model", "ucbbar,spike-bare,qemu");
-    qemu_fdt_setprop_string(fdt, "/", "compatible", "ucbbar,spike-bare-dev");
+    qemu_fdt_setprop_string(fdt, "/", "model", "SiFive HiFive Unleashed A00");
+    qemu_fdt_setprop_string(fdt, "/", "compatible",
+                            "sifive,hifive-unleashed-a00");
     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
 
Signature · 1 line
2.7.4
author
#39 lore raw
In reply to Bin Meng
Quoted 79 lines · Bin Meng

As of today, the QEMU 'sifive_u' machine is a special target that does not boot the upstream OpenSBI/U-Boot firmware images built for the real SiFive HiFive Unleashed board. Hence OpenSBI supports a special platform "qemu/sifive_u". For U-Boot, the sifive_fu540_defconfig is referenced in the OpenSBI doc as its payload, but that does not boot at all due to various issues in current QEMU 'sifive_u' machine codes.

This series aims to improve the emulation fidelity of sifive_u machine, so that the upstream OpenSBI, U-Boot and kernel images built for the SiFive HiFive Unleashed board can be used out of the box without any special hack.

The major changes include:
- Heterogeneous harts creation supported, so that we can create a CPU that exactly mirrors the real hardware: 1 E51 + 4 U54.
- Implemented a PRCI model for FU540
- Implemented an OTP model for FU540, primarily used for storing serial number of the board
- Fixed GEM support that was seriously broken on sifive_u
- Synced device tree with upstream Linux kernel on sifive_u
- Adding initramfs loading support on sifive_u

OpenSBI v0.4 image built for sifive/fu540 is included as the default bios image for 'sifive_u' machine.

The series is tested against OpenSBI v0.4 image for sifive/fu540 paltform, U-Boot v2019.10-rc1 image for sifive_fu540_defconfig, and Linux kernel v5.3-rc3 image with the following patch:

macb: Update compatibility string for SiFive FU540-C000
https://patchwork.kernel.org/patch/11050003/

OpenSBI + U-Boot, ping/tftpboot with U-Boot MACB driver works well. Boot Linux 64-bit defconfig image, verified that system console on the serial 0 and ping host work pretty well.

An OpenSBI patch was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. It will be applied after this QEMU series is merged.
http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v3:
- changed to use macros for management and compute cpu count
- use management cpu count + 1 for the min_cpus
- update IRQ numbers of both UARTs to match hardware as well

Changes in v2:
- fixed the "interrupts-extended" property size
- update the file header to indicate at least 2 harts are created
- use create_unimplemented_device() to create the GEM management block instead of sifive_mmio_emulate()
- add "phy-handle" property to the ethernet node
- keep the PLIC compatible string unchanged as OpenSBI uses that for DT fix up
- drop patch "riscv: sifive: Move sifive_mmio_emulate() to a common place"
- new patch "riscv: sifive_e: Drop sifive_mmio_emulate()"

Bin Meng (28):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: Add a sifive_cpu.h to include both E and U cpu type defines
riscv: hart: Extract hart realize to a separate routine
riscv: hart: Support heterogeneous harts population
riscv: sifive_u: Update hart configuration to reflect the real FU540
SoC
riscv: sifive_u: Set the minimum number of cpus to 2
riscv: sifive_u: Update PLIC hart topology configuration string
riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Remove the unnecessary include of prci header
riscv: sifive: Rename sifive_prci.{c,h} to sifive_e_prci.{c,h}
riscv: sifive_e: prci: Fix a typo of hfxosccfg register programming
riscv: sifive_e: prci: Update the PRCI register block size
riscv: sifive: Implement PRCI model for FU540
riscv: sifive_u: Generate hfclk and rtcclk nodes
riscv: sifive_u: Add PRCI block to the SoC
riscv: sifive_u: Change UART node name in device tree
riscv: hw: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: roms: Update default bios for sifive_u machine

Quoted 9 lines

riscv: sifive_u: Update UART and ethernet node clock properties
riscv: sifive_u: Generate an aliases node in the device tree
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Support loading initramfs
riscv: hw: Remove not needed PLIC properties in device tree
riscv: sifive_e: Drop sifive_mmio_emulate()
riscv: virt: Change create_fdt() to return void
riscv: sifive_u: Update model and compatible strings in device tree

I have no idea on why patch [20/28] and [26/28] failed to arrive on the mailing list. The "git send-email" logs said OK when these two patches were sent. In v2, these 2 patches are missing on patchwork, and unfortunately this is the same situation for v3.

Attached these 2 patches.

Regards,
Bin

From e8bb57b3423b096f1efc851e0361232fde205f09 Mon Sep 17 00:00:00 2001
From: Bin Meng <[email protected]>
Date: Sat, 3 Aug 2019 13:48:24 +0800
Subject: [PATCH v3 20/28] riscv: roms: Update default bios for sifive_u
machine

With the support of heterogeneous harts and PRCI model, it's now possible to use the opensbi image built for the real hardware.

---
This patch should be applied after http://patchwork.ozlabs.org/patch/1141443/

Changes in v3: None
Changes in v2: None

 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin | Bin 40968 -> 45064 bytes
 roms/Makefile                                |   4 ++--
 2 files changed, 2 insertions(+), 2 deletions(-)
pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin binary +0 -0
index 5d7a1ef6818994bac4a36818ad36043b592ce309..f29b3bed6fa8b57a735436d9c684740aa0a6bb16 100755
GIT binary patch
roms/Makefile +2 -2
index 775c963..6cf07d3 100644
@@ -182,8 +182,8 @@ opensbi64-virt:
 opensbi64-sifive_u:
 	$(MAKE) -C opensbi \
 		CROSS_COMPILE=$(riscv64_cross_prefix) \
-		PLATFORM="qemu/sifive_u"
-	cp opensbi/build/platform/qemu/sifive_u/firmware/fw_jump.bin ../pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin
+		PLATFORM="sifive/fu540"
+	cp opensbi/build/platform/sifive/fu540/firmware/fw_jump.bin ../pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin
 
 clean:
 	rm -rf seabios/.config seabios/out seabios/builds
Signature · 1 line
2.7.4

From b611512239ea23629d89bbcdf12fe8ee6a0459c1 Mon Sep 17 00:00:00 2001
From: Bin Meng <[email protected]>
Date: Wed, 7 Aug 2019 14:56:09 +0800
Subject: [PATCH v3 26/28] riscv: sifive_e: Drop sifive_mmio_emulate()

Use create_unimplemented_device() instead.

---

Changes in v3: None
Changes in v2:
- drop patch "riscv: sifive: Move sifive_mmio_emulate() to a common place"
- new patch "riscv: sifive_e: Drop sifive_mmio_emulate()"

 hw/riscv/sifive_e.c | 23 ++++++++---------------
 1 file changed, 8 insertions(+), 15 deletions(-)
hw/riscv/sifive_e.c +8 -15
index 2d67670..040d59f 100644
@@ -37,6 +37,7 @@
 #include "hw/loader.h"
 #include "hw/sysbus.h"
 #include "hw/char/serial.h"
+#include "hw/misc/unimp.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
@@ -74,14 +75,6 @@ static const struct MemmapEntry {
     [SIFIVE_E_DTIM] =     { 0x80000000,     0x4000 }
 };
 
-static void sifive_mmio_emulate(MemoryRegion *parent, const char *name,
-                             uintptr_t offset, uintptr_t length)
-{
-    MemoryRegion *mock_mmio = g_new(MemoryRegion, 1);
-    memory_region_init_ram(mock_mmio, NULL, name, length, &error_fatal);
-    memory_region_add_subregion(parent, offset, mock_mmio);
-}
-
 static void riscv_sifive_e_init(MachineState *machine)
 {
     const struct MemmapEntry *memmap = sifive_e_memmap;
@@ -172,7 +165,7 @@ static void riscv_sifive_e_soc_realize(DeviceState *dev, Error **errp)
     sifive_clint_create(memmap[SIFIVE_E_CLINT].base,
         memmap[SIFIVE_E_CLINT].size, ms->smp.cpus,
         SIFIVE_SIP_BASE, SIFIVE_TIMECMP_BASE, SIFIVE_TIME_BASE);
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.aon",
+    create_unimplemented_device("riscv.sifive.e.aon",
         memmap[SIFIVE_E_AON].base, memmap[SIFIVE_E_AON].size);
     sifive_e_prci_create(memmap[SIFIVE_E_PRCI].base);
 
@@ -199,19 +192,19 @@ static void riscv_sifive_e_soc_realize(DeviceState *dev, Error **errp)
 
     sifive_uart_create(sys_mem, memmap[SIFIVE_E_UART0].base,
         serial_hd(0), qdev_get_gpio_in(DEVICE(s->plic), SIFIVE_E_UART0_IRQ));
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.qspi0",
+    create_unimplemented_device("riscv.sifive.e.qspi0",
         memmap[SIFIVE_E_QSPI0].base, memmap[SIFIVE_E_QSPI0].size);
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.pwm0",
+    create_unimplemented_device("riscv.sifive.e.pwm0",
         memmap[SIFIVE_E_PWM0].base, memmap[SIFIVE_E_PWM0].size);
     sifive_uart_create(sys_mem, memmap[SIFIVE_E_UART1].base,
         serial_hd(1), qdev_get_gpio_in(DEVICE(s->plic), SIFIVE_E_UART1_IRQ));
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.qspi1",
+    create_unimplemented_device("riscv.sifive.e.qspi1",
         memmap[SIFIVE_E_QSPI1].base, memmap[SIFIVE_E_QSPI1].size);
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.pwm1",
+    create_unimplemented_device("riscv.sifive.e.pwm1",
         memmap[SIFIVE_E_PWM1].base, memmap[SIFIVE_E_PWM1].size);
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.qspi2",
+    create_unimplemented_device("riscv.sifive.e.qspi2",
         memmap[SIFIVE_E_QSPI2].base, memmap[SIFIVE_E_QSPI2].size);
-    sifive_mmio_emulate(sys_mem, "riscv.sifive.e.pwm2",
+    create_unimplemented_device("riscv.sifive.e.pwm2",
         memmap[SIFIVE_E_PWM2].base, memmap[SIFIVE_E_PWM2].size);
 
     /* Flash memory */
Signature · 1 line
2.7.4