Skip to contentSkip to conversation

riscv: sifive_u: Improve the emulation fidelity of sifive_u machine

PATCH Superseded Bin Meng 3 participants 41 messages Aug 23, 2019 – Aug 26, 2019 1 message not in the archive

Subsystems (2)

v5 of 8 Superseded Latest is v8 →

lore
Share & tools
lore

Bin Meng

author
#0 lore raw

v5 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

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 [1]

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 [2] was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. The original plan was to get the drop patch applied after this QEMU series is merged. However after discussion in the OpenSBI mailing list, it seems the best option for us is to let OpenSBI continue shipping the special "qemu/sifive_u" platform support to work with QEMU version <= 4.1 and deprecate the support sometime in the future. A patch will need to be sent to OpenSBI mailing list to update its document.

v4 is now rebased on Palmer's QEMU RISC-V repo "for-master" branch. Dropped the following v3 patch that was already done by someone else.
- riscv: sifive_u: Generate an aliases node in the device tree
- riscv: sifive_u: Support loading initramfs

The following v3 patch was dropped too due to a different cluster approach suggested by Richard Henderson is used in v4:
- riscv: hart: Support heterogeneous harts population

[1]: https://patchwork.kernel.org/patch/11050003/
[2]: http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models
- new patch to remove the unnecessary include of target/riscv/cpu.h
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_otp
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_otp_write()
- move all register initialization to sifive_u_otp_reset() function
- drop sifive_u_otp_create()
- create sifive_u_otp block directly in the machine codes, instead of calling sifive_u_otp_create()
- add the missing "local-mac-address" property in the ethernet node

Changes in v4:
- remove 2 more "linux,phandle" instances in sifive_u.c and spike.c after rebasing on Palmer's QEMU RISC-V tree
- change create_fdt() to return void in sifive_u.c too, after rebasing on Palmer's QEMU RISC-V tree
- new patch to remove executable attribute of opensbi images
- prefix all macros/variables/functions with SIFIVE_E/sifive_e in the sifive_e_prci driver
- new patch to add a "hartid-base" property to RISC-V hart array
- changed to create clusters for each cpu type
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_otp driver
- new patch to remove handcrafted clock nodes for UART and ethernet

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:
- 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()"
- 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

Bin Meng (30):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: hw: Remove not needed PLIC properties in device tree
riscv: hw: Change create_fdt() to return void
riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead
riscv: hw: Remove the unnecessary include of target/riscv/cpu.h
riscv: roms: Remove executable attribute of opensbi images
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_e: Drop sifive_mmio_emulate()
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: Add a "hartid-base" property to RISC-V hart array
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: 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: Reference PRCI clocks in UART and ethernet nodes
riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Change UART node name in device tree
riscv: roms: Update default bios for sifive_u machine
riscv: sifive: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Remove handcrafted clock nodes for UART and ethernet
riscv: sifive_u: Update model and compatible strings in device tree

 hw/riscv/Makefile.objs                       |   4 +-
 hw/riscv/riscv_hart.c                        |  35 ++--
 hw/riscv/sifive_e.c                          |  27 +--
 hw/riscv/sifive_e_prci.c                     | 123 +++++++++++++
 hw/riscv/sifive_prci.c                       | 123 -------------
 hw/riscv/sifive_test.c                       |   6 +-
 hw/riscv/sifive_u.c                          | 262 ++++++++++++++++++---------
 hw/riscv/sifive_u_otp.c                      | 190 +++++++++++++++++++
 hw/riscv/sifive_u_prci.c                     | 171 +++++++++++++++++
 hw/riscv/sifive_uart.c                       |  10 +-
 hw/riscv/spike.c                             |   1 -
 hw/riscv/virt.c                              |  40 ++--
 include/hw/riscv/riscv_hart.h                |   1 +
 include/hw/riscv/sifive_cpu.h                |  31 ++++
 include/hw/riscv/sifive_e.h                  |   7 +-
 include/hw/riscv/sifive_e_prci.h             |  71 ++++++++
 include/hw/riscv/sifive_prci.h               |  69 -------
 include/hw/riscv/sifive_u.h                  |  31 ++--
 include/hw/riscv/sifive_u_otp.h              |  80 ++++++++
 include/hw/riscv/sifive_u_prci.h             |  91 ++++++++++
 pc-bios/opensbi-riscv32-virt-fw_jump.bin     | Bin
 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin | Bin 40968 -> 45064 bytes
 pc-bios/opensbi-riscv64-virt-fw_jump.bin     | Bin
 roms/Makefile                                |   4 +-
 24 files changed, 1018 insertions(+), 359 deletions(-)
 create mode 100644 hw/riscv/sifive_e_prci.c
 delete mode 100644 hw/riscv/sifive_prci.c
 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
 create mode 100644 include/hw/riscv/sifive_e_prci.h
 delete mode 100644 include/hw/riscv/sifive_prci.h
 create mode 100644 include/hw/riscv/sifive_u_otp.h
 create mode 100644 include/hw/riscv/sifive_u_prci.h
 mode change 100755 => 100644 pc-bios/opensbi-riscv32-virt-fw_jump.bin
 mode change 100755 => 100644 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin
 mode change 100755 => 100644 pc-bios/opensbi-riscv64-virt-fw_jump.bin
Signature · 1 line
2.7.4
author
#1 lore raw

PATCH v5 1/30 riscv: hw: Remove superfluous "linux, phandle" property

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

---

Changes in v5: None
Changes in v4:
- remove 2 more "linux,phandle" instances in sifive_u.c and spike.c after rebasing on Palmer's QEMU RISC-V tree

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 4 ----
 hw/riscv/spike.c    | 1 -
 hw/riscv/virt.c     | 3 ---
 3 files changed, 8 deletions(-)
hw/riscv/sifive_u.c +0 -4
index 64e233d..afe304f 100644
@@ -126,7 +126,6 @@ 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);
@@ -185,7 +184,6 @@ 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);
@@ -198,7 +196,6 @@ 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);
 
@@ -234,7 +231,6 @@ static void *create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "#clock-cells", 0x0);
     qemu_fdt_setprop_cell(fdt, nodename, "clock-frequency", 3686400);
     qemu_fdt_setprop_cell(fdt, nodename, "phandle", uartclk_phandle);
-    qemu_fdt_setprop_cell(fdt, nodename, "linux,phandle", uartclk_phandle);
     uartclk_phandle = qemu_fdt_get_phandle(fdt, nodename);
     g_free(nodename);
 
hw/riscv/spike.c +0 -1
index 2991b34..14acaef 100644
@@ -112,7 +112,6 @@ static void create_fdt(SpikeState *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", 1);
-        qemu_fdt_setprop_cell(fdt, intc, "linux,phandle", 1);
         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);
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 v5 2/30 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 v5: None
Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 18 +++++++++---------
 hw/riscv/virt.c     | 24 ++++++++++++------------
 2 files changed, 21 insertions(+), 21 deletions(-)
hw/riscv/sifive_u.c +9 -9
index afe304f..3f9284e 100644
@@ -183,7 +183,7 @@ 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);
@@ -208,20 +208,20 @@ 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, "clock-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);
 
     uartclk_phandle = phandle++;
@@ -241,9 +241,9 @@ 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_cells(fdt, nodename, "clocks", uartclk_phandle);
-    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, "clocks", uartclk_phandle);
+    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 v5 3/30 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 v5: None
Changes in v4: None
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 3f9284e..5fe0033 100644
@@ -180,8 +180,6 @@ 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
author
#4 lore raw

PATCH v5 4/30 riscv: hw: 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.

---

Changes in v5: None
Changes in v4:
- change create_fdt() to return void in sifive_u.c too, after rebasing on Palmer's QEMU RISC-V tree

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c | 11 ++++-------
 hw/riscv/virt.c     | 11 ++++-------
 2 files changed, 8 insertions(+), 14 deletions(-)
hw/riscv/sifive_u.c +4 -7
index 5fe0033..e22803b 100644
@@ -67,7 +67,7 @@ static const struct MemmapEntry {
 
 #define GEM_REVISION        0x10070109
 
-static void *create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
+static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     uint64_t mem_size, const char *cmdline)
 {
     void *fdt;
@@ -253,14 +253,11 @@ static void *create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_string(fdt, "/aliases", "serial0", nodename);
 
     g_free(nodename);
-
-    return fdt;
 }
 
 static void riscv_sifive_u_init(MachineState *machine)
 {
     const struct MemmapEntry *memmap = sifive_u_memmap;
-    void *fdt;
 
     SiFiveUState *s = g_new0(SiFiveUState, 1);
     MemoryRegion *system_memory = get_system_memory();
@@ -281,7 +278,7 @@ static void riscv_sifive_u_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);
 
     riscv_find_and_load_firmware(machine, BIOS_FILENAME,
                                  memmap[SIFIVE_U_DRAM].base);
@@ -294,9 +291,9 @@ static void riscv_sifive_u_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);
         }
     }
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
author
#5 lore raw

PATCH v5 5/30 riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR, ...) instead

In reply to Bin Meng

Replace the call to hw_error() with qemu_log_mask(LOG_GUEST_ERROR,...) in various sifive models.

---

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models

Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_prci.c | 8 +++++---
 hw/riscv/sifive_test.c | 5 +++--
 hw/riscv/sifive_uart.c | 9 +++++----
 3 files changed, 13 insertions(+), 9 deletions(-)
hw/riscv/sifive_prci.c +5 -3
index f406682..1ab98d4 100644
@@ -20,6 +20,7 @@
 
 #include "qemu/osdep.h"
 #include "hw/sysbus.h"
+#include "qemu/log.h"
 #include "qemu/module.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_prci.h"
@@ -37,7 +38,8 @@ static uint64_t sifive_prci_read(void *opaque, hwaddr addr, unsigned int size)
     case SIFIVE_PRCI_PLLOUTDIV:
         return s->plloutdiv;
     }
-    hw_error("%s: read: addr=0x%x\n", __func__, (int)addr);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: read: addr=0x%x\n",
+                  __func__, (int)addr);
     return 0;
 }
 
@@ -65,8 +67,8 @@ static void sifive_prci_write(void *opaque, hwaddr addr,
         s->plloutdiv = (uint32_t) val64;
         break;
     default:
-        hw_error("%s: bad write: addr=0x%x v=0x%x\n",
-                 __func__, (int)addr, (int)val64);
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: bad write: addr=0x%x v=0x%x\n",
+                      __func__, (int)addr, (int)val64);
     }
 }
 
hw/riscv/sifive_test.c +3 -2
index cd86831..655a3d7 100644
@@ -20,6 +20,7 @@
 
 #include "qemu/osdep.h"
 #include "hw/sysbus.h"
+#include "qemu/log.h"
 #include "qemu/module.h"
 #include "sysemu/sysemu.h"
 #include "target/riscv/cpu.h"
@@ -48,8 +49,8 @@ static void sifive_test_write(void *opaque, hwaddr addr,
             break;
         }
     }
-    hw_error("%s: write: addr=0x%x val=0x%016" PRIx64 "\n",
-        __func__, (int)addr, val64);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: write: addr=0x%x val=0x%016" PRIx64 "\n",
+                  __func__, (int)addr, val64);
 }
 
 static const MemoryRegionOps sifive_test_ops = {
hw/riscv/sifive_uart.c +5 -4
index 3b3f94f..cd74043 100644
@@ -18,6 +18,7 @@
 
 #include "qemu/osdep.h"
 #include "qapi/error.h"
+#include "qemu/log.h"
 #include "hw/sysbus.h"
 #include "chardev/char.h"
 #include "chardev/char-fe.h"
@@ -93,8 +94,8 @@ uart_read(void *opaque, hwaddr addr, unsigned int size)
         return s->div;
     }
 
-    hw_error("%s: bad read: addr=0x%x\n",
-        __func__, (int)addr);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: bad read: addr=0x%x\n",
+                  __func__, (int)addr);
     return 0;
 }
 
@@ -125,8 +126,8 @@ uart_write(void *opaque, hwaddr addr,
         s->div = val64;
         return;
     }
-    hw_error("%s: bad write: addr=0x%x v=0x%x\n",
-        __func__, (int)addr, (int)value);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: bad write: addr=0x%x v=0x%x\n",
+                  __func__, (int)addr, (int)value);
 }
 
 static const MemoryRegionOps uart_ops = {
Signature · 1 line
2.7.4
Reviewed-by on v5 5/30
#6 lore raw

Re: [Qemu-devel] [PATCH v5 05/30] riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR, ...) instead

Quoted 5 lines · Bin Meng

Replace the call to hw_error() with qemu_log_mask(LOG_GUEST_ERROR,...) in various sifive models.

Thanks for this cleanup!

Alistair

Quoted 110 lines

---

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models

Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_prci.c | 8 +++++---
 hw/riscv/sifive_test.c | 5 +++--
 hw/riscv/sifive_uart.c | 9 +++++----
 3 files changed, 13 insertions(+), 9 deletions(-)
hw/riscv/sifive_prci.c +5 -3
index f406682..1ab98d4 100644
@@ -20,6 +20,7 @@
 
 #include "qemu/osdep.h"
 #include "hw/sysbus.h"
+#include "qemu/log.h"
 #include "qemu/module.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_prci.h"
@@ -37,7 +38,8 @@ static uint64_t sifive_prci_read(void *opaque, hwaddr addr, unsigned int size)
     case SIFIVE_PRCI_PLLOUTDIV:
         return s->plloutdiv;
     }
-    hw_error("%s: read: addr=0x%x\n", __func__, (int)addr);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: read: addr=0x%x\n",
+                  __func__, (int)addr);
     return 0;
 }
 
@@ -65,8 +67,8 @@ static void sifive_prci_write(void *opaque, hwaddr addr,
         s->plloutdiv = (uint32_t) val64;
         break;
     default:
-        hw_error("%s: bad write: addr=0x%x v=0x%x\n",
-                 __func__, (int)addr, (int)val64);
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: bad write: addr=0x%x v=0x%x\n",
+                      __func__, (int)addr, (int)val64);
     }
 }
 
hw/riscv/sifive_test.c +3 -2
index cd86831..655a3d7 100644
@@ -20,6 +20,7 @@
 
 #include "qemu/osdep.h"
 #include "hw/sysbus.h"
+#include "qemu/log.h"
 #include "qemu/module.h"
 #include "sysemu/sysemu.h"
 #include "target/riscv/cpu.h"
@@ -48,8 +49,8 @@ static void sifive_test_write(void *opaque, hwaddr addr,
             break;
         }
     }
-    hw_error("%s: write: addr=0x%x val=0x%016" PRIx64 "\n",
-        __func__, (int)addr, val64);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: write: addr=0x%x val=0x%016" PRIx64 "\n",
+                  __func__, (int)addr, val64);
 }
 
 static const MemoryRegionOps sifive_test_ops = {
hw/riscv/sifive_uart.c +5 -5
index 3b3f94f..cd74043 100644
@@ -18,6 +18,7 @@
 
 #include "qemu/osdep.h"
 #include "qapi/error.h"
+#include "qemu/log.h"
 #include "hw/sysbus.h"
 #include "chardev/char.h"
 #include "chardev/char-fe.h"
@@ -93,8 +94,8 @@ uart_read(void *opaque, hwaddr addr, unsigned int size)
         return s->div;
     }
 
-    hw_error("%s: bad read: addr=0x%x\n",
-        __func__, (int)addr);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: bad read: addr=0x%x\n",
+                  __func__, (int)addr);
     return 0;
 }
 
@@ -125,8 +126,8 @@ uart_write(void *opaque, hwaddr addr,
         s->div = val64;
         return;
     }
-    hw_error("%s: bad write: addr=0x%x v=0x%x\n",
-        __func__, (int)addr, (int)value);
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: bad write: addr=0x%x v=0x%x\n",
+                  __func__, (int)addr, (int)value);
 }
 
 static const MemoryRegionOps uart_ops = {
--

2.7.4

author
#7 lore raw

PATCH v5 6/30 riscv: hw: Remove the unnecessary include of target/riscv/cpu.h

In reply to Bin Meng

The inclusion of "target/riscv/cpu.h" is unnecessary in various sifive model drivers.

---

Changes in v5:
- new patch to remove the unnecessary include of target/riscv/cpu.h

Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_prci.c | 1 -
 hw/riscv/sifive_test.c | 1 -
 hw/riscv/sifive_uart.c | 1 -
 3 files changed, 3 deletions(-)
hw/riscv/sifive_prci.c +0 -1
index 1ab98d4..1957dcd 100644
@@ -22,7 +22,6 @@
 #include "hw/sysbus.h"
 #include "qemu/log.h"
 #include "qemu/module.h"
-#include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_prci.h"
 
 static uint64_t sifive_prci_read(void *opaque, hwaddr addr, unsigned int size)
hw/riscv/sifive_test.c +0 -1
index 655a3d7..31cad9f 100644
@@ -23,7 +23,6 @@
 #include "qemu/log.h"
 #include "qemu/module.h"
 #include "sysemu/sysemu.h"
-#include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_test.h"
 
 static uint64_t sifive_test_read(void *opaque, hwaddr addr, unsigned int size)
hw/riscv/sifive_uart.c +0 -1
index cd74043..1601bd9 100644
@@ -22,7 +22,6 @@
 #include "hw/sysbus.h"
 #include "chardev/char.h"
 #include "chardev/char-fe.h"
-#include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_uart.h"
 
 /*
Signature · 1 line
2.7.4
Reviewed-by on v5 6/30
#8 lore raw

Re: [Qemu-devel] [PATCH v5 06/30] riscv: hw: Remove the unnecessary include of target/riscv/cpu.h

Quoted 5 lines · Bin Meng

The inclusion of "target/riscv/cpu.h" is unnecessary in various sifive model drivers.

Alistair

Quoted 55 lines

---

Changes in v5:
- new patch to remove the unnecessary include of target/riscv/cpu.h

Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_prci.c | 1 -
 hw/riscv/sifive_test.c | 1 -
 hw/riscv/sifive_uart.c | 1 -
 3 files changed, 3 deletions(-)
hw/riscv/sifive_prci.c +0 -1
index 1ab98d4..1957dcd 100644
@@ -22,7 +22,6 @@
 #include "hw/sysbus.h"
 #include "qemu/log.h"
 #include "qemu/module.h"
-#include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_prci.h"
 
 static uint64_t sifive_prci_read(void *opaque, hwaddr addr, unsigned int size)
hw/riscv/sifive_test.c +0 -1
index 655a3d7..31cad9f 100644
@@ -23,7 +23,6 @@
 #include "qemu/log.h"
 #include "qemu/module.h"
 #include "sysemu/sysemu.h"
-#include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_test.h"
 
 static uint64_t sifive_test_read(void *opaque, hwaddr addr, unsigned int size)
hw/riscv/sifive_uart.c +0 -2
index cd74043..1601bd9 100644
@@ -22,7 +22,6 @@
 #include "hw/sysbus.h"
 #include "chardev/char.h"
 #include "chardev/char-fe.h"
-#include "target/riscv/cpu.h"
 #include "hw/riscv/sifive_uart.h"
 
 /*
--

2.7.4

author
#9 lore raw

PATCH v5 7/30 riscv: roms: Remove executable attribute of opensbi images

In reply to Bin Meng

Like other binary files, the executable attribute of opensbi images should not be set.

---

Changes in v5: None
Changes in v4:
- new patch to remove executable attribute of opensbi images

Changes in v3: None
Changes in v2: None

 pc-bios/opensbi-riscv32-virt-fw_jump.bin     | Bin
 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin | Bin
 pc-bios/opensbi-riscv64-virt-fw_jump.bin     | Bin
 3 files changed, 0 insertions(+), 0 deletions(-)
 mode change 100755 => 100644 pc-bios/opensbi-riscv32-virt-fw_jump.bin
 mode change 100755 => 100644 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin
 mode change 100755 => 100644 pc-bios/opensbi-riscv64-virt-fw_jump.bin
pc-bios/opensbi-riscv32-virt-fw_jump.bin +0 -0
old mode 100755
new mode 100644
pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin +0 -0
old mode 100755
new mode 100644
pc-bios/opensbi-riscv64-virt-fw_jump.bin +0 -0
old mode 100755
new mode 100644
Signature · 1 line
2.7.4
author
#10 lore raw

PATCH v5 8/30 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 v5: None
Changes in v4: None
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 e22803b..3f58f61 100644
@@ -39,7 +39,6 @@
 #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
#11 lore raw

PATCH v5 9/30 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. This also prefix "sifive_e"/"SIFIVE_E" for all macros, variables and functions.

---

Changes in v5: None
Changes in v4:
- prefix all macros/variables/functions with SIFIVE_E/sifive_e in the sifive_e_prci driver

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} | 79 ++++++++++++++---------------
 include/hw/riscv/sifive_e_prci.h            | 69 +++++++++++++++++++++++++
 include/hw/riscv/sifive_prci.h              | 69 -------------------------
 5 files changed, 111 insertions(+), 112 deletions(-)
 rename hw/riscv/{sifive_prci.c => sifive_e_prci.c} (51%)
 create mode 100644 include/hw/riscv/sifive_e_prci.h
 delete mode 100644 include/hw/riscv/sifive_prci.h
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 +39 -40

142 lines not on this page · raw message

include/hw/riscv/sifive_e_prci.h new file +69 -0

69 lines not on this page · raw message

include/hw/riscv/sifive_prci.h deleted +0 -69

69 lines not on this page · raw message

Signature · 1 line
2.7.4
author
#12 lore raw

PATCH v5 10/30 riscv: sifive_e: prci: Fix a typo of hfxosccfg register programming

In reply to Bin Meng

For hfxosccfg register programming, SIFIVE_E_PRCI_HFXOSCCFG_RDY and SIFIVE_E_PRCI_HFXOSCCFG_EN should be used.

---

Changes in v5: None
Changes in v4: None
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 c514032..71de089 100644
@@ -90,7 +90,7 @@ static void sifive_e_prci_init(Object *obj)
     sysbus_init_mmio(SYS_BUS_DEVICE(obj), &s->mmio);
 
     s->hfrosccfg = (SIFIVE_E_PRCI_HFROSCCFG_RDY | SIFIVE_E_PRCI_HFROSCCFG_EN);
-    s->hfxosccfg = (SIFIVE_E_PRCI_HFROSCCFG_RDY | SIFIVE_E_PRCI_HFROSCCFG_EN);
+    s->hfxosccfg = (SIFIVE_E_PRCI_HFXOSCCFG_RDY | SIFIVE_E_PRCI_HFXOSCCFG_EN);
     s->pllcfg = (SIFIVE_E_PRCI_PLLCFG_REFSEL | SIFIVE_E_PRCI_PLLCFG_BYPASS |
                  SIFIVE_E_PRCI_PLLCFG_LOCK);
     s->plloutdiv = SIFIVE_E_PRCI_PLLOUTDIV_DIV1;
Signature · 1 line
2.7.4
author
#13 lore raw

PATCH v5 11/30 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 v5: None
Changes in v4: None
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 71de089..ad6c624 100644
@@ -86,7 +86,7 @@ static void sifive_e_prci_init(Object *obj)
     SiFiveEPRCIState *s = SIFIVE_E_PRCI(obj);
 
     memory_region_init_io(&s->mmio, obj, &sifive_e_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_E_PRCI_HFROSCCFG_RDY | SIFIVE_E_PRCI_HFROSCCFG_EN);
include/hw/riscv/sifive_e_prci.h +2 -0
index c4b76aa..698b0b4 100644
@@ -47,6 +47,8 @@ enum {
     SIFIVE_E_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
author
#14 lore raw

PATCH v5 13/30 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 v5: None
Changes in v4: None
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
#15 lore raw

PATCH v5 14/30 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 homogeneous harts. Exact the hart realize to a separate routine in preparation for supporting multiple hart arrays.

Note the file header says the RISC-V hart array holds the state of a heterogeneous array of RISC-V harts, which is not true. Update the comment to mention homogeneous array of RISC-V harts.

---

Changes in v5: None
Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/riscv_hart.c | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)
hw/riscv/riscv_hart.c +20 -13
index ca69a1b..9deef869 100644
@@ -3,7 +3,7 @@
  *
  * Copyright (c) 2017 SiFive, Inc.
  *
- * Holds the state of a heterogenous array of RISC-V harts
+ * Holds the state of a homogeneous array of RISC-V 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,
@@ -37,26 +37,33 @@ static void riscv_harts_cpu_reset(void *opaque)
     cpu_reset(CPU(cpu));
 }
 
+static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
+                               char *cpu_type, Error **errp)
+{
+    Error *err = NULL;
+
+    object_initialize_child(OBJECT(s), "harts[*]", &s->harts[idx],
+                            sizeof(RISCVCPU), cpu_type,
+                            &error_abort, NULL);
+    s->harts[idx].env.mhartid = idx;
+    qemu_register_reset(riscv_harts_cpu_reset, &s->harts[idx]);
+    object_property_set_bool(OBJECT(&s->harts[idx]), 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
author
#16 lore raw

PATCH v5 15/30 riscv: hart: Add a "hartid-base" property to RISC-V hart array

In reply to Bin Meng

At present each hart's hartid in a RISC-V hart array is assigned the same value of its index in the hart array. But for a system that has multiple hart arrays, this is not the case any more.

Add a new "hartid-base" property so that hartid number can be assigned based on the property value.

---

Changes in v5: None
Changes in v4:
- new patch to add a "hartid-base" property to RISC-V hart array

Changes in v3: None
Changes in v2: None

 hw/riscv/riscv_hart.c         | 8 +++++---
 include/hw/riscv/riscv_hart.h | 1 +
 2 files changed, 6 insertions(+), 3 deletions(-)
hw/riscv/riscv_hart.c +5 -3
index 9deef869..52ab86a 100644
@@ -27,6 +27,7 @@
 
 static Property riscv_harts_props[] = {
     DEFINE_PROP_UINT32("num-harts", RISCVHartArrayState, num_harts, 1),
+    DEFINE_PROP_UINT32("hartid-base", RISCVHartArrayState, hartid_base, 0),
     DEFINE_PROP_STRING("cpu-type", RISCVHartArrayState, cpu_type),
     DEFINE_PROP_END_OF_LIST(),
 };
@@ -37,7 +38,7 @@ static void riscv_harts_cpu_reset(void *opaque)
     cpu_reset(CPU(cpu));
 }
 
-static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
+static void riscv_hart_realize(RISCVHartArrayState *s, int idx, uint32_t hartid,
                                char *cpu_type, Error **errp)
 {
     Error *err = NULL;
@@ -45,7 +46,7 @@ static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
     object_initialize_child(OBJECT(s), "harts[*]", &s->harts[idx],
                             sizeof(RISCVCPU), cpu_type,
                             &error_abort, NULL);
-    s->harts[idx].env.mhartid = idx;
+    s->harts[idx].env.mhartid = hartid;
     qemu_register_reset(riscv_harts_cpu_reset, &s->harts[idx]);
     object_property_set_bool(OBJECT(&s->harts[idx]), true,
                              "realized", &err);
@@ -58,12 +59,13 @@ static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
 static void riscv_harts_realize(DeviceState *dev, Error **errp)
 {
     RISCVHartArrayState *s = RISCV_HART_ARRAY(dev);
+    uint32_t hartid = s->hartid_base;
     int n;
 
     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);
+        riscv_hart_realize(s, n, hartid + n, s->cpu_type, errp);
     }
 }
 
include/hw/riscv/riscv_hart.h +1 -0
index 0671d88..1984e30 100644
@@ -32,6 +32,7 @@ typedef struct RISCVHartArrayState {
 
     /*< public >*/
     uint32_t num_harts;
+    uint32_t hartid_base;
     char *cpu_type;
     RISCVCPU *harts;
 } RISCVHartArrayState;
Signature · 1 line
2.7.4
#17 lore raw

Re: [Qemu-devel] [PATCH v5 15/30] riscv: hart: Add a "hartid-base" property to RISC-V hart array

Quoted 36 lines · Bin Meng

At present each hart's hartid in a RISC-V hart array is assigned the same value of its index in the hart array. But for a system that has multiple hart arrays, this is not the case any more.

Add a new "hartid-base" property so that hartid number can be assigned based on the property value.

---

Changes in v5: None
Changes in v4:
- new patch to add a "hartid-base" property to RISC-V hart array

Changes in v3: None
Changes in v2: None

 hw/riscv/riscv_hart.c         | 8 +++++---
 include/hw/riscv/riscv_hart.h | 1 +
 2 files changed, 6 insertions(+), 3 deletions(-)
hw/riscv/riscv_hart.c +1 -0
index 9deef869..52ab86a 100644
@@ -27,6 +27,7 @@
 
 static Property riscv_harts_props[] = {
     DEFINE_PROP_UINT32("num-harts", RISCVHartArrayState, num_harts, 1),
+    DEFINE_PROP_UINT32("hartid-base", RISCVHartArrayState, hartid_base, 0),
     DEFINE_PROP_STRING("cpu-type", RISCVHartArrayState, cpu_type),
     DEFINE_PROP_END_OF_LIST(),
 };

hw/riscv/riscv_hart.c static void riscv_harts_cpu_reset(void *opaque) · line 42

Quoted 6 lines
hw/riscv/riscv_hart.c +1 -1
@@ -37,7 +38,7 @@ static void riscv_harts_cpu_reset(void *opaque)
     cpu_reset(CPU(cpu));
 }
 
-static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
+static void riscv_hart_realize(RISCVHartArrayState *s, int idx, uint32_t hartid,
                                char *cpu_type, Error **errp)

Do we need to pass hartid, can't we just get it from s here?

Quoted 12 lines
hw/riscv/riscv_hart.c +1 -1
 {
     Error *err = NULL;
@@ -45,7 +46,7 @@ static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
     object_initialize_child(OBJECT(s), "harts[*]", &s->harts[idx],
                             sizeof(RISCVCPU), cpu_type,
                             &error_abort, NULL);
-    s->harts[idx].env.mhartid = idx;
+    s->harts[idx].env.mhartid = hartid;
     qemu_register_reset(riscv_harts_cpu_reset, &s->harts[idx]);
     object_property_set_bool(OBJECT(&s->harts[idx]), true,
                              "realized", &err);

hw/riscv/riscv_hart.c static void riscv_hart_realize(RISCVHartArrayState *s, int idx, · line 62

Quoted 4 lines
hw/riscv/riscv_hart.c +1 -0
@@ -58,12 +59,13 @@ static void riscv_hart_realize(RISCVHartArrayState *s, int idx,
 static void riscv_harts_realize(DeviceState *dev, Error **errp)
 {
     RISCVHartArrayState *s = RISCV_HART_ARRAY(dev);
+    uint32_t hartid = s->hartid_base;

I think s->hartid_base can just be used directly, instead of a extra variable.

Alistair

Quoted 26 lines
hw/riscv/riscv_hart.c +1 -1
     int n;
 
     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);
+        riscv_hart_realize(s, n, hartid + n, s->cpu_type, errp);
     }
 }
 
include/hw/riscv/riscv_hart.h +1 -1
index 0671d88..1984e30 100644
@@ -32,6 +32,7 @@ typedef struct RISCVHartArrayState {
 
     /*< public >*/
     uint32_t num_harts;
+    uint32_t hartid_base;
     char *cpu_type;
     RISCVCPU *harts;
 } RISCVHartArrayState;
--

2.7.4

author
#18 lore raw

PATCH v5 16/30 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, by creating 2 CPU clusters as containers for RISC-V hart arrays to populate heterogeneous harts.

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

---

Changes in v5: None
Changes in v4:
- changed to create clusters for each cpu type

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         | 102 +++++++++++++++++++++++++++++++++-----------
 include/hw/riscv/sifive_u.h |   8 +++-
 2 files changed, 84 insertions(+), 26 deletions(-)
hw/riscv/sifive_u.c +77 -25
index 3f58f61..0e5bbe7 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"
@@ -34,6 +36,7 @@
 #include "hw/loader.h"
 #include "hw/sysbus.h"
 #include "hw/char/serial.h"
+#include "hw/cpu/cluster.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
@@ -69,6 +72,7 @@ static const struct MemmapEntry {
 static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     uint64_t mem_size, const char *cmdline)
 {
+    MachineState *ms = MACHINE(qdev_get_machine());
     void *fdt;
     int cpu;
     uint32_t *cells;
@@ -109,15 +113,21 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, "/cpus", "#size-cells", 0x0);
     qemu_fdt_setprop_cell(fdt, "/cpus", "#address-cells", 0x1);
 
-    for (cpu = s->soc.cpus.num_harts - 1; cpu >= 0; cpu--) {
+    for (cpu = ms->smp.cpus - 1; cpu >= 0; cpu--) {
         int cpu_phandle = phandle++;
         nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
         char *intc = g_strdup_printf("/cpus/cpu@%d/interrupt-controller", cpu);
-        char *isa = riscv_isa_string(&s->soc.cpus.harts[cpu]);
+        char *isa;
         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");
+            isa = riscv_isa_string(&s->soc.u_cpus.harts[cpu - 1]);
+        } else {
+            isa = riscv_isa_string(&s->soc.e_cpus.harts[0]);
+        }
         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");
@@ -133,8 +143,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         g_free(nodename);
     }
 
-    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4);
-    for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
+    cells =  g_new0(uint32_t, ms->smp.cpus * 4);
+    for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
         nodename =
             g_strdup_printf("/cpus/cpu@%d/interrupt-controller", cpu);
         uint32_t intc_phandle = qemu_fdt_get_phandle(fdt, nodename);
@@ -152,20 +162,26 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         0x0, memmap[SIFIVE_U_CLINT].base,
         0x0, memmap[SIFIVE_U_CLINT].size);
     qemu_fdt_setprop(fdt, nodename, "interrupts-extended",
-        cells, s->soc.cpus.num_harts * sizeof(uint32_t) * 4);
+        cells, ms->smp.cpus * sizeof(uint32_t) * 4);
     g_free(cells);
     g_free(nodename);
 
     plic_phandle = phandle++;
-    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4);
-    for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
+    cells =  g_new0(uint32_t, ms->smp.cpus * 4 - 2);
+    for (cpu = 0; cpu < ms->smp.cpus; 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 +191,7 @@ 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, (ms->smp.cpus * 4 - 2) * sizeof(uint32_t));
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_PLIC].base,
         0x0, memmap[SIFIVE_U_PLIC].size);
@@ -338,12 +354,39 @@ static void riscv_sifive_u_soc_init(Object *obj)
     MachineState *ms = MACHINE(qdev_get_machine());
     SiFiveUSoCState *s = RISCV_U_SOC(obj);
 
-    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",
-                            &error_abort);
-    object_property_set_int(OBJECT(&s->cpus), ms->smp.cpus, "num-harts",
-                            &error_abort);
+    object_initialize_child(obj, "e-cluster", &s->e_cluster,
+                            sizeof(s->e_cluster), TYPE_CPU_CLUSTER,
+                            &error_abort, NULL);
+    qdev_prop_set_uint32(DEVICE(&s->e_cluster), "cluster-id", 0);
+
+    object_initialize_child(OBJECT(&s->e_cluster), "e-cpus",
+                            &s->e_cpus, sizeof(s->e_cpus),
+                            TYPE_RISCV_HART_ARRAY, &error_abort,
+                            NULL);
+    object_property_set_uint(OBJECT(&s->e_cpus), 1,
+                             "num-harts", &error_abort);
+    object_property_set_uint(OBJECT(&s->e_cpus), 0,
+                             "hartid-base", &error_abort);
+    object_property_set_str(OBJECT(&s->e_cpus), SIFIVE_E_CPU,
+                            "cpu-type", &error_abort);
+
+    if (ms->smp.cpus > 1) {
+        object_initialize_child(obj, "u-cluster", &s->u_cluster,
+                                sizeof(s->u_cluster), TYPE_CPU_CLUSTER,
+                                &error_abort, NULL);
+        qdev_prop_set_uint32(DEVICE(&s->u_cluster), "cluster-id", 1);
+
+        object_initialize_child(OBJECT(&s->u_cluster), "u-cpus",
+                                &s->u_cpus, sizeof(s->u_cpus),
+                                TYPE_RISCV_HART_ARRAY, &error_abort,
+                                NULL);
+        object_property_set_uint(OBJECT(&s->u_cpus), ms->smp.cpus - 1,
+                                 "num-harts", &error_abort);
+        object_property_set_uint(OBJECT(&s->u_cpus), 1,
+                                 "hartid-base", &error_abort);
+        object_property_set_str(OBJECT(&s->u_cpus), SIFIVE_U_CPU,
+                                "cpu-type", &error_abort);
+    }
 
     sysbus_init_child_obj(obj, "gem", &s->gem, sizeof(s->gem),
                           TYPE_CADENCE_GEM);
@@ -363,7 +406,19 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
     Error *err = NULL;
     NICInfo *nd = &nd_table[0];
 
-    object_property_set_bool(OBJECT(&s->cpus), true, "realized",
+    object_property_set_bool(OBJECT(&s->e_cpus), true, "realized",
+                             &error_abort);
+    object_property_set_bool(OBJECT(&s->u_cpus), true, "realized",
+                             &error_abort);
+    /*
+     * The cluster must be realized after the RISC-V hart array container,
+     * as the container's CPU object is only created on realize, and the
+     * CPU must exist and have been parented into the cluster before the
+     * cluster is realized.
+     */
+    object_property_set_bool(OBJECT(&s->e_cluster), true, "realized",
+                             &error_abort);
+    object_property_set_bool(OBJECT(&s->u_cluster), true, "realized",
                              &error_abort);
 
     /* boot rom */
@@ -429,10 +484,7 @@ 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 +7 -1
index 4abc621..7a1a4f3 100644
@@ -31,7 +31,10 @@ typedef struct SiFiveUSoCState {
     SysBusDevice parent_obj;
 
     /*< public >*/
-    RISCVHartArrayState cpus;
+    CPUClusterState e_cluster;
+    CPUClusterState u_cluster;
+    RISCVHartArrayState e_cpus;
+    RISCVHartArrayState u_cpus;
     DeviceState *plic;
     CadenceGEMState gem;
 } SiFiveUSoCState;
@@ -68,6 +71,9 @@ 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
#19 lore raw

Re: [Qemu-devel] [PATCH v5 16/30] riscv: sifive_u: Update hart configuration to reflect the real FU540 SoC

Quoted 162 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, by creating 2 CPU clusters as containers for RISC-V hart arrays to populate heterogeneous harts.

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

---

Changes in v5: None
Changes in v4:
- changed to create clusters for each cpu type

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         | 102 +++++++++++++++++++++++++++++++++-----------
 include/hw/riscv/sifive_u.h |   8 +++-
 2 files changed, 84 insertions(+), 26 deletions(-)
hw/riscv/sifive_u.c +37 -20
index 3f58f61..0e5bbe7 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"
@@ -34,6 +36,7 @@
 #include "hw/loader.h"
 #include "hw/sysbus.h"
 #include "hw/char/serial.h"
+#include "hw/cpu/cluster.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
@@ -69,6 +72,7 @@ static const struct MemmapEntry {
 static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     uint64_t mem_size, const char *cmdline)
 {
+    MachineState *ms = MACHINE(qdev_get_machine());
     void *fdt;
     int cpu;
     uint32_t *cells;
@@ -109,15 +113,21 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, "/cpus", "#size-cells", 0x0);
     qemu_fdt_setprop_cell(fdt, "/cpus", "#address-cells", 0x1);
 
-    for (cpu = s->soc.cpus.num_harts - 1; cpu >= 0; cpu--) {
+    for (cpu = ms->smp.cpus - 1; cpu >= 0; cpu--) {
         int cpu_phandle = phandle++;
         nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
         char *intc = g_strdup_printf("/cpus/cpu@%d/interrupt-controller", cpu);
-        char *isa = riscv_isa_string(&s->soc.cpus.harts[cpu]);
+        char *isa;
         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");
+            isa = riscv_isa_string(&s->soc.u_cpus.harts[cpu - 1]);
+        } else {
+            isa = riscv_isa_string(&s->soc.e_cpus.harts[0]);
+        }
         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");
@@ -133,8 +143,8 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         g_free(nodename);
     }
 
-    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4);
-    for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
+    cells =  g_new0(uint32_t, ms->smp.cpus * 4);
+    for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
         nodename =
             g_strdup_printf("/cpus/cpu@%d/interrupt-controller", cpu);
         uint32_t intc_phandle = qemu_fdt_get_phandle(fdt, nodename);
@@ -152,20 +162,26 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
         0x0, memmap[SIFIVE_U_CLINT].base,
         0x0, memmap[SIFIVE_U_CLINT].size);
     qemu_fdt_setprop(fdt, nodename, "interrupts-extended",
-        cells, s->soc.cpus.num_harts * sizeof(uint32_t) * 4);
+        cells, ms->smp.cpus * sizeof(uint32_t) * 4);
     g_free(cells);
     g_free(nodename);
 
     plic_phandle = phandle++;
-    cells =  g_new0(uint32_t, s->soc.cpus.num_harts * 4);
-    for (cpu = 0; cpu < s->soc.cpus.num_harts; cpu++) {
+    cells =  g_new0(uint32_t, ms->smp.cpus * 4 - 2);
+    for (cpu = 0; cpu < ms->smp.cpus; 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 +191,7 @@ 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, (ms->smp.cpus * 4 - 2) * sizeof(uint32_t));
     qemu_fdt_setprop_cells(fdt, nodename, "reg",
         0x0, memmap[SIFIVE_U_PLIC].base,
         0x0, memmap[SIFIVE_U_PLIC].size);
@@ -338,12 +354,39 @@ static void riscv_sifive_u_soc_init(Object *obj)
     MachineState *ms = MACHINE(qdev_get_machine());
     SiFiveUSoCState *s = RISCV_U_SOC(obj);
 
-    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",
-                            &error_abort);
-    object_property_set_int(OBJECT(&s->cpus), ms->smp.cpus, "num-harts",
-                            &error_abort);
+    object_initialize_child(obj, "e-cluster", &s->e_cluster,
+                            sizeof(s->e_cluster), TYPE_CPU_CLUSTER,
+                            &error_abort, NULL);
+    qdev_prop_set_uint32(DEVICE(&s->e_cluster), "cluster-id", 0);
+
+    object_initialize_child(OBJECT(&s->e_cluster), "e-cpus",
+                            &s->e_cpus, sizeof(s->e_cpus),

hw/riscv/sifive_u.c static void riscv_sifive_u_soc_init(Object *obj) · line 373

Quoted 10 lines
hw/riscv/sifive_u.c +10 -0
+                            TYPE_RISCV_HART_ARRAY, &error_abort,
+                            NULL);
+    object_property_set_uint(OBJECT(&s->e_cpus), 1,
+                             "num-harts", &error_abort);
+    object_property_set_uint(OBJECT(&s->e_cpus), 0,
+                             "hartid-base", &error_abort);
+    object_property_set_str(OBJECT(&s->e_cpus), SIFIVE_E_CPU,
+                            "cpu-type", &error_abort);
+
+    if (ms->smp.cpus > 1) {

I don't think this is right.

If a user specifies -smp 1 for the SiFive U machine, it seems unlikey they just want the power management CPU and don't care about the U cores.

The fix to this is to add a minimum CPU count of 2 which we can do. From memory you had this originally and I said to remove it.

The problem with that though is then we force everyone to use at least 2 CPUs, which I don't like. It can be very useful to run -smp 1 sometimes.

So, I see two options:
1. Just set the minimum and default CPUs to 2 and just live with 2
CPUs all the time
2. Have some setup where -smp < x means no E CPU and -smp > x means
with the E cpu. From memory this is what the Xilinx ZynqMP boards do. x would probably be 4 in this case, but it could also be 2

I'm ok with both options, I'm leaning towards option 2 though unless it ends up being too confusing for users. In saying that I think either is acceptable so it's up to you Bin.

Alistair

Quoted 82 lines
hw/riscv/sifive_u.c +30 -5
+        object_initialize_child(obj, "u-cluster", &s->u_cluster,
+                                sizeof(s->u_cluster), TYPE_CPU_CLUSTER,
+                                &error_abort, NULL);
+        qdev_prop_set_uint32(DEVICE(&s->u_cluster), "cluster-id", 1);
+
+        object_initialize_child(OBJECT(&s->u_cluster), "u-cpus",
+                                &s->u_cpus, sizeof(s->u_cpus),
+                                TYPE_RISCV_HART_ARRAY, &error_abort,
+                                NULL);
+        object_property_set_uint(OBJECT(&s->u_cpus), ms->smp.cpus - 1,
+                                 "num-harts", &error_abort);
+        object_property_set_uint(OBJECT(&s->u_cpus), 1,
+                                 "hartid-base", &error_abort);
+        object_property_set_str(OBJECT(&s->u_cpus), SIFIVE_U_CPU,
+                                "cpu-type", &error_abort);
+    }
 
     sysbus_init_child_obj(obj, "gem", &s->gem, sizeof(s->gem),
                           TYPE_CADENCE_GEM);
@@ -363,7 +406,19 @@ static void riscv_sifive_u_soc_realize(DeviceState *dev, Error **errp)
     Error *err = NULL;
     NICInfo *nd = &nd_table[0];
 
-    object_property_set_bool(OBJECT(&s->cpus), true, "realized",
+    object_property_set_bool(OBJECT(&s->e_cpus), true, "realized",
+                             &error_abort);
+    object_property_set_bool(OBJECT(&s->u_cpus), true, "realized",
+                             &error_abort);
+    /*
+     * The cluster must be realized after the RISC-V hart array container,
+     * as the container's CPU object is only created on realize, and the
+     * CPU must exist and have been parented into the cluster before the
+     * cluster is realized.
+     */
+    object_property_set_bool(OBJECT(&s->e_cluster), true, "realized",
+                             &error_abort);
+    object_property_set_bool(OBJECT(&s->u_cluster), true, "realized",
                              &error_abort);
 
     /* boot rom */
@@ -429,10 +484,7 @@ 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 +7 -2
index 4abc621..7a1a4f3 100644
@@ -31,7 +31,10 @@ typedef struct SiFiveUSoCState {
     SysBusDevice parent_obj;
 
     /*< public >*/
-    RISCVHartArrayState cpus;
+    CPUClusterState e_cluster;
+    CPUClusterState u_cluster;
+    RISCVHartArrayState e_cpus;
+    RISCVHartArrayState u_cpus;
     DeviceState *plic;
     CadenceGEMState gem;
 } SiFiveUSoCState;
@@ -68,6 +71,9 @@ 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
#20 lore raw

PATCH v5 17/30 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 v5: None
Changes in v4: None
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 0e5bbe7..a36cd77 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,
@@ -485,6 +485,7 @@ 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
#21 lore raw

Re: [Qemu-devel] [PATCH v5 17/30] riscv: sifive_u: Set the minimum number of cpus to 2

Quoted 5 lines · Bin Meng

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

Ah, I jumped the gun with patch 16. This should be before patch 16, otherwise patch 16 is fine.

Alistair

Quoted 41 lines

---

Changes in v5: None
Changes in v4: None
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 0e5bbe7..a36cd77 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,
@@ -485,6 +485,7 @@ 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
#22 lore raw

PATCH v5 18/30 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 v5: None
Changes in v4: None
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 a36cd77..284f7a5 100644
@@ -433,10 +433,11 @@ 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
#23 lore raw

PATCH v5 19/30 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 v5:
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g

Changes in v4:
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs           |   1 +
 hw/riscv/sifive_u_prci.c         | 171 +++++++++++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_u_prci.h |  81 +++++++++++++++++++
 3 files changed, 253 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 +171 -0

171 lines not on this page · raw message

include/hw/riscv/sifive_u_prci.h new file +81 -0

81 lines not on this page · raw message

Signature · 1 line
2.7.4
Reviewed-by on v5 19/30
#24 lore raw

Re: [Qemu-devel] [PATCH v5 19/30] riscv: sifive: Implement PRCI model for FU540

Quoted 130 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 v5:
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g

Changes in v4:
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs           |   1 +
 hw/riscv/sifive_u_prci.c         | 171 +++++++++++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_u_prci.h |  81 +++++++++++++++++++
 3 files changed, 253 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 +81 -0
new file mode 100644
index 0000000..c6438fb
@@ -0,0 +1,171 @@
+/*
+ * 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/log.h"
+#include "qemu/module.h"
+#include "hw/riscv/sifive_u_prci.h"
+
+static uint64_t sifive_u_prci_read(void *opaque, hwaddr addr, unsigned int size)
+{
+    SiFiveUPRCIState *s = opaque;
+
+    switch (addr) {
+    case SIFIVE_U_PRCI_HFXOSCCFG:
+        return s->hfxosccfg;
+    case SIFIVE_U_PRCI_COREPLLCFG0:
+        return s->corepllcfg0;
+    case SIFIVE_U_PRCI_DDRPLLCFG0:
+        return s->ddrpllcfg0;
+    case SIFIVE_U_PRCI_DDRPLLCFG1:
+        return s->ddrpllcfg1;
+    case SIFIVE_U_PRCI_GEMGXLPLLCFG0:
+        return s->gemgxlpllcfg0;
+    case SIFIVE_U_PRCI_GEMGXLPLLCFG1:
+        return s->gemgxlpllcfg1;
+    case SIFIVE_U_PRCI_CORECLKSEL:
+        return s->coreclksel;
+    case SIFIVE_U_PRCI_DEVICESRESET:
+        return s->devicesreset;
+    case SIFIVE_U_PRCI_CLKMUXSTATUS:
+        return s->clkmuxstatus;
+    }
+
+    qemu_log_mask(LOG_GUEST_ERROR, "%s: read: addr=0x%x\n",
+                  __func__, (int)addr);
+
+    return 0;
+}
+
+static void sifive_u_prci_write(void *opaque, hwaddr addr,
+                                uint64_t val64, unsigned int size)
+{
+    SiFiveUPRCIState *s = opaque;
+    uint32_t val32 = (uint32_t)val64;
+
+    switch (addr) {
+    case SIFIVE_U_PRCI_HFXOSCCFG:
+        s->hfxosccfg = val32;
+        /* OSC stays ready */
+        s->hfxosccfg |= SIFIVE_U_PRCI_HFXOSCCFG_RDY;
+        break;
+    case SIFIVE_U_PRCI_COREPLLCFG0:
+        s->corepllcfg0 = val32;
+        /* internal feedback */
+        s->corepllcfg0 |= SIFIVE_U_PRCI_PLLCFG0_FSE;
+        /* PLL stays locked */
+        s->corepllcfg0 |= SIFIVE_U_PRCI_PLLCFG0_LOCK;
+        break;
+    case SIFIVE_U_PRCI_DDRPLLCFG0:
+        s->ddrpllcfg0 = val32;
+        /* internal feedback */
+        s->ddrpllcfg0 |= SIFIVE_U_PRCI_PLLCFG0_FSE;
+        /* PLL stays locked */

hw/riscv/sifive_u_prci.c line 91

Quoted 10 lines
hw/riscv/sifive_u_prci.c new file +10 -0
+        s->ddrpllcfg0 |= SIFIVE_U_PRCI_PLLCFG0_LOCK;
+        break;
+    case SIFIVE_U_PRCI_DDRPLLCFG1:
+        s->ddrpllcfg1 = val32;
+        break;
+    case SIFIVE_U_PRCI_GEMGXLPLLCFG0:
+        s->gemgxlpllcfg0 = val32;
+         /* internal feedback */
+        s->gemgxlpllcfg0 |= SIFIVE_U_PRCI_PLLCFG0_FSE;
+       /* PLL stays locked */

This line seems to be incorrectly indented.

Quoted 46 lines
hw/riscv/sifive_u_prci.c new file +46 -0
+        s->gemgxlpllcfg0 |= SIFIVE_U_PRCI_PLLCFG0_LOCK;
+        break;
+    case SIFIVE_U_PRCI_GEMGXLPLLCFG1:
+        s->gemgxlpllcfg1 = val32;
+        break;
+    case SIFIVE_U_PRCI_CORECLKSEL:
+        s->coreclksel = val32;
+        break;
+    case SIFIVE_U_PRCI_DEVICESRESET:
+        s->devicesreset = val32;
+        break;
+    case SIFIVE_U_PRCI_CLKMUXSTATUS:
+        s->clkmuxstatus = val32;
+        break;
+    default:
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: bad write: addr=0x%x v=0x%x\n",
+                      __func__, (int)addr, (int)val64);
+    }
+}
+
+static const MemoryRegionOps sifive_u_prci_ops = {
+    .read = sifive_u_prci_read,
+    .write = sifive_u_prci_write,
+    .endianness = DEVICE_NATIVE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 4
+    }
+};
+
+static void sifive_u_prci_realize(DeviceState *dev, Error **errp)
+{
+    SiFiveUPRCIState *s = SIFIVE_U_PRCI(dev);
+
+    memory_region_init_io(&s->mmio, OBJECT(dev), &sifive_u_prci_ops, s,
+                          TYPE_SIFIVE_U_PRCI, SIFIVE_U_PRCI_REG_SIZE);
+    sysbus_init_mmio(SYS_BUS_DEVICE(dev), &s->mmio);
+}
+
+static void sifive_u_prci_reset(DeviceState *dev)
+{
+    SiFiveUPRCIState *s = SIFIVE_U_PRCI(dev);
+
+    /* Initialize register to power-on-reset values */
+    s->hfxosccfg = (SIFIVE_U_PRCI_HFXOSCCFG_RDY | SIFIVE_U_PRCI_HFXOSCCFG_EN);
+    s->corepllcfg0 = (SIFIVE_U_PRCI_PLLCFG0_DIVR | SIFIVE_U_PRCI_PLLCFG0_DIVF |

hw/riscv/sifive_u_prci.c line 147

Quoted 10 lines
hw/riscv/sifive_u_prci.c new file +10 -0
+                      SIFIVE_U_PRCI_PLLCFG0_DIVQ | SIFIVE_U_PRCI_PLLCFG0_FSE |
+                      SIFIVE_U_PRCI_PLLCFG0_LOCK);
+    s->ddrpllcfg0 = (SIFIVE_U_PRCI_PLLCFG0_DIVR | SIFIVE_U_PRCI_PLLCFG0_DIVF |
+                     SIFIVE_U_PRCI_PLLCFG0_DIVQ | SIFIVE_U_PRCI_PLLCFG0_FSE |
+                     SIFIVE_U_PRCI_PLLCFG0_LOCK);
+    s->gemgxlpllcfg0 = (SIFIVE_U_PRCI_PLLCFG0_DIVR |
+                        SIFIVE_U_PRCI_PLLCFG0_DIVF |
+                        SIFIVE_U_PRCI_PLLCFG0_DIVQ |
+                        SIFIVE_U_PRCI_PLLCFG0_FSE |
+                        SIFIVE_U_PRCI_PLLCFG0_LOCK);

You don't need brackets around these.

Besides the nits:

Alistair

Quoted 115 lines
hw/riscv/sifive_u_prci.c new file +24 -0
+    s->coreclksel = SIFIVE_U_PRCI_CORECLKSEL_HFCLK;
+}
+
+static void sifive_u_prci_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+
+    dc->realize = sifive_u_prci_realize;
+    dc->reset = sifive_u_prci_reset;
+}
+
+static const TypeInfo sifive_u_prci_info = {
+    .name          = TYPE_SIFIVE_U_PRCI,
+    .parent        = TYPE_SYS_BUS_DEVICE,
+    .instance_size = sizeof(SiFiveUPRCIState),
+    .class_init    = sifive_u_prci_class_init,
+};
+
+static void sifive_u_prci_register_types(void)
+{
+    type_register_static(&sifive_u_prci_info);
+}
+
+type_init(sifive_u_prci_register_types)
include/hw/riscv/sifive_u_prci.h new file +81 -1
new file mode 100644
index 0000000..60a2eab
@@ -0,0 +1,81 @@
+/*
+ * 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
+
+#define SIFIVE_U_PRCI_HFXOSCCFG     0x00
+#define SIFIVE_U_PRCI_COREPLLCFG0   0x04
+#define SIFIVE_U_PRCI_DDRPLLCFG0    0x0C
+#define SIFIVE_U_PRCI_DDRPLLCFG1    0x10
+#define SIFIVE_U_PRCI_GEMGXLPLLCFG0 0x1C
+#define SIFIVE_U_PRCI_GEMGXLPLLCFG1 0x20
+#define SIFIVE_U_PRCI_CORECLKSEL    0x24
+#define SIFIVE_U_PRCI_DEVICESRESET  0x28
+#define SIFIVE_U_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 code that works.
+ *
+ * see https://github.com/sifive/freedom-u540-c000-bootloader
+ */
+
+#define SIFIVE_U_PRCI_HFXOSCCFG_EN  (1 << 30)
+#define SIFIVE_U_PRCI_HFXOSCCFG_RDY (1 << 31)
+
+/* xxxPLLCFG0 register bits */
+#define SIFIVE_U_PRCI_PLLCFG0_DIVR  (1 << 0)
+#define SIFIVE_U_PRCI_PLLCFG0_DIVF  (31 << 6)
+#define SIFIVE_U_PRCI_PLLCFG0_DIVQ  (3 << 15)
+#define SIFIVE_U_PRCI_PLLCFG0_FSE   (1 << 25)
+#define SIFIVE_U_PRCI_PLLCFG0_LOCK  (1 << 31)
+
+/* xxxPLLCFG1 register bits */
+#define SIFIVE_U_PRCI_PLLCFG1_CKE   (1 << 24)
+
+/* coreclksel register bits */
+#define SIFIVE_U_PRCI_CORECLKSEL_HFCLK  (1 << 0)
+
+
+#define SIFIVE_U_PRCI_REG_SIZE  0x1000
+
+#define TYPE_SIFIVE_U_PRCI      "riscv.sifive.u.prci"
+
+#define SIFIVE_U_PRCI(obj) \
+    OBJECT_CHECK(SiFiveUPRCIState, (obj), TYPE_SIFIVE_U_PRCI)
+
+typedef struct SiFiveUPRCIState {
+    /*< 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;
+} SiFiveUPRCIState;
+
+#endif /* HW_SIFIVE_U_PRCI_H */
--

2.7.4

author
#25 lore raw

PATCH v5 20/30 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 v5: None
Changes in v4: None
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 284f7a5..08db741 100644
@@ -80,6 +80,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     char ethclk_names[] = "pclk\0hclk\0tx_clk";
     uint32_t plic_phandle, ethclk_phandle, phandle = 1;
     uint32_t uartclk_phandle;
+    uint32_t hfclk_phandle, rtcclk_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
     if (!fdt) {
@@ -98,6 +99,28 @@ 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 7a1a4f3..debbf28 100644
@@ -68,6 +68,8 @@ 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
#26 lore raw

PATCH v5 21/30 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 v5:
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()

Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 24 +++++++++++++++++++++++-
 include/hw/riscv/sifive_u.h |  3 +++
 2 files changed, 26 insertions(+), 1 deletion(-)
hw/riscv/sifive_u.c +23 -1
index 08db741..c777d41 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.
@@ -61,6 +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] =    { 0x10013000,     0x1000 },
     [SIFIVE_U_UART1] =    { 0x10023000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
@@ -78,7 +80,7 @@ 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 uartclk_phandle;
     uint32_t hfclk_phandle, rtcclk_phandle;
 
@@ -189,6 +191,21 @@ 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, ms->smp.cpus * 4 - 2);
     for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
@@ -411,6 +428,8 @@ static void riscv_sifive_u_soc_init(Object *obj)
                                 "cpu-type", &error_abort);
     }
 
+    sysbus_init_child_obj(obj, "prci", &s->prci, sizeof(s->prci),
+                          TYPE_SIFIVE_U_PRCI);
     sysbus_init_child_obj(obj, "gem", &s->gem, sizeof(s->gem),
                           TYPE_CADENCE_GEM);
 }
@@ -484,6 +503,9 @@ 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);
 
+    object_property_set_bool(OBJECT(&s->prci), true, "realized", &err);
+    sysbus_mmio_map(SYS_BUS_DEVICE(&s->prci), 0, 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 +3 -0
index debbf28..2a023be 100644
@@ -21,6 +21,7 @@
 
 #include "hw/net/cadence_gem.h"
 #include "hw/riscv/sifive_cpu.h"
+#include "hw/riscv/sifive_u_prci.h"
 
 #define TYPE_RISCV_U_SOC "riscv.sifive.u.soc"
 #define RISCV_U_SOC(obj) \
@@ -36,6 +37,7 @@ typedef struct SiFiveUSoCState {
     RISCVHartArrayState e_cpus;
     RISCVHartArrayState u_cpus;
     DeviceState *plic;
+    SiFiveUPRCIState prci;
     CadenceGEMState gem;
 } SiFiveUSoCState;
 
@@ -54,6 +56,7 @@ 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
Reviewed-by on v5 21/30
#27 lore raw

Re: [Qemu-devel] [PATCH v5 21/30] riscv: sifive_u: Add PRCI block to the SoC

Quoted 5 lines · Bin Meng

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

Alistair

Quoted 117 lines

---

Changes in v5:
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()

Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 24 +++++++++++++++++++++++-
 include/hw/riscv/sifive_u.h |  3 +++
 2 files changed, 26 insertions(+), 1 deletion(-)
hw/riscv/sifive_u.c +23 -1
index 08db741..c777d41 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.
@@ -61,6 +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] =    { 0x10013000,     0x1000 },
     [SIFIVE_U_UART1] =    { 0x10023000,     0x1000 },
     [SIFIVE_U_DRAM] =     { 0x80000000,        0x0 },
@@ -78,7 +80,7 @@ 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 uartclk_phandle;
     uint32_t hfclk_phandle, rtcclk_phandle;
 
@@ -189,6 +191,21 @@ 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, ms->smp.cpus * 4 - 2);
     for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
@@ -411,6 +428,8 @@ static void riscv_sifive_u_soc_init(Object *obj)
                                 "cpu-type", &error_abort);
     }
 
+    sysbus_init_child_obj(obj, "prci", &s->prci, sizeof(s->prci),
+                          TYPE_SIFIVE_U_PRCI);
     sysbus_init_child_obj(obj, "gem", &s->gem, sizeof(s->gem),
                           TYPE_CADENCE_GEM);
 }
@@ -484,6 +503,9 @@ 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);
 
+    object_property_set_bool(OBJECT(&s->prci), true, "realized", &err);
+    sysbus_mmio_map(SYS_BUS_DEVICE(&s->prci), 0, 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 +3 -1
index debbf28..2a023be 100644
@@ -21,6 +21,7 @@
 
 #include "hw/net/cadence_gem.h"
 #include "hw/riscv/sifive_cpu.h"
+#include "hw/riscv/sifive_u_prci.h"
 
 #define TYPE_RISCV_U_SOC "riscv.sifive.u.soc"
 #define RISCV_U_SOC(obj) \
@@ -36,6 +37,7 @@ typedef struct SiFiveUSoCState {
     RISCVHartArrayState e_cpus;
     RISCVHartArrayState u_cpus;
     DeviceState *plic;
+    SiFiveUPRCIState prci;
     CadenceGEMState gem;
 } SiFiveUSoCState;
 
@@ -54,6 +56,7 @@ enum {
     SIFIVE_U_MROM,
     SIFIVE_U_CLINT,
     SIFIVE_U_PLIC,
+    SIFIVE_U_PRCI,
     SIFIVE_U_UART0,
     SIFIVE_U_UART1,
     SIFIVE_U_DRAM,
--

2.7.4

author
#28 lore raw

PATCH v5 22/30 riscv: sifive_u: Reference PRCI clocks in UART and ethernet nodes

In reply to Bin Meng

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

---

Changes in v5: None
Changes in v4: None
Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c              |  7 ++++---
 include/hw/riscv/sifive_u_prci.h | 10 ++++++++++
 2 files changed, 14 insertions(+), 3 deletions(-)
hw/riscv/sifive_u.c +4 -3
index c777d41..e0842ad 100644
@@ -79,7 +79,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     int cpu;
     uint32_t *cells;
     char *nodename;
-    char ethclk_names[] = "pclk\0hclk\0tx_clk";
+    char ethclk_names[] = "pclk\0hclk";
     uint32_t plic_phandle, prci_phandle, ethclk_phandle, phandle = 1;
     uint32_t uartclk_phandle;
     uint32_t hfclk_phandle, rtcclk_phandle;
@@ -264,7 +264,7 @@ 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, "clock-names", ethclk_names,
         sizeof(ethclk_names));
     qemu_fdt_setprop_cell(fdt, nodename, "#address-cells", 1);
@@ -294,7 +294,8 @@ 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, "clocks", uartclk_phandle);
+    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_prci.h +10 -0
index 60a2eab..0a531fd 100644
@@ -78,4 +78,14 @@ typedef struct SiFiveUPRCIState {
     uint32_t clkmuxstatus;
 } SiFiveUPRCIState;
 
+/*
+ * 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
#29 lore raw

PATCH v5 23/30 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 v5: None
Changes in v4: None
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 e0842ad..1a178dc 100644
@@ -63,8 +63,8 @@ static const struct MemmapEntry {
     [SIFIVE_U_CLINT] =    {  0x2000000,    0x10000 },
     [SIFIVE_U_PLIC] =     {  0xc000000,  0x4000000 },
     [SIFIVE_U_PRCI] =     { 0x10000000,     0x1000 },
-    [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 2a023be..b41e730 100644
@@ -64,8 +64,8 @@ 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
#30 lore raw

PATCH v5 24/30 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 v5: None
Changes in v4: None
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 1a178dc..6cf669c 100644
@@ -287,7 +287,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     uartclk_phandle = qemu_fdt_get_phandle(fdt, nodename);
     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
#31 lore raw

PATCH v5 25/30 riscv: roms: Update default bios for sifive_u machine

In reply to Bin Meng

With the support of heterogeneous harts and PRCI model, it's now possible to use the OpenSBI image (PLATFORM=sifive/fu540) built for the real hardware.

---

Changes in v5: None
Changes in v4: None
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..eb22aefdfb468cfe2804cb4b0bc422d8ebcae93b 100644
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
author
#32 lore raw

PATCH v5 26/30 riscv: sifive: 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 v5:
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_otp
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_otp_write()
- move all register initialization to sifive_u_otp_reset() function
- drop sifive_u_otp_create()

Changes in v4:
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_otp driver

Changes in v3: None
Changes in v2: None

 hw/riscv/Makefile.objs          |   1 +
 hw/riscv/sifive_u_otp.c         | 190 ++++++++++++++++++++++++++++++++++++++++
 include/hw/riscv/sifive_u_otp.h |  80 +++++++++++++++++
 3 files changed, 271 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 +190 -0

190 lines not on this page · raw message

include/hw/riscv/sifive_u_otp.h new file +80 -0

80 lines not on this page · raw message

Signature · 1 line
2.7.4
author
#33 lore raw

PATCH v5 28/30 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.

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 v5:
- add the missing "local-mac-address" property in the ethernet node

Changes in v4: None
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         | 24 ++++++++++++++++++++----
 include/hw/riscv/sifive_u.h |  3 ++-
 2 files changed, 22 insertions(+), 5 deletions(-)
hw/riscv/sifive_u.c +20 -4
index b6ddf5d..503db4b 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.
@@ -39,6 +41,7 @@
 #include "hw/sysbus.h"
 #include "hw/char/serial.h"
 #include "hw/cpu/cluster.h"
+#include "hw/misc/unimp.h"
 #include "target/riscv/cpu.h"
 #include "hw/riscv/riscv_hart.h"
 #include "hw/riscv/sifive_plic.h"
@@ -47,6 +50,7 @@
 #include "hw/riscv/sifive_u.h"
 #include "hw/riscv/boot.h"
 #include "chardev/char.h"
+#include "net/eth.h"
 #include "sysemu/arch_init.h"
 #include "sysemu/device_tree.h"
 #include "exec/address-spaces.h"
@@ -68,7 +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 OTP_SERIAL          1
@@ -85,7 +90,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     char ethclk_names[] = "pclk\0hclk";
     uint32_t plic_phandle, prci_phandle, ethclk_phandle, phandle = 1;
     uint32_t uartclk_phandle;
-    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) {
@@ -255,21 +260,28 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     ethclk_phandle = qemu_fdt_get_phandle(fdt, nodename);
     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, "clock-names", ethclk_names,
         sizeof(ethclk_names));
+    qemu_fdt_setprop(fdt, nodename, "local-mac-address",
+        s->soc.gem.conf.macaddr.a, ETH_ALEN);
     qemu_fdt_setprop_cell(fdt, nodename, "#address-cells", 1);
     qemu_fdt_setprop_cell(fdt, nodename, "#size-cells", 0);
     g_free(nodename);
@@ -277,6 +289,7 @@ 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);
 
@@ -534,6 +547,9 @@ 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 7d9d901..d2b9d99 100644
@@ -63,7 +63,8 @@ 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
author
#34 lore raw

PATCH v5 29/30 riscv: sifive_u: Remove handcrafted clock nodes for UART and ethernet

In reply to Bin Meng

In the past we did not have a model for PRCI, hence two handcrafted clock nodes ("/soc/ethclk" and "/soc/uartclk") were created for the purpose of supplying hard-coded clock frequencies. But now since we have added the PRCI support in QEMU, we don't need them any more.

---

Changes in v5: None
Changes in v4:
- new patch to remove handcrafted clock nodes for UART and ethernet

Changes in v3: None
Changes in v2: None

 hw/riscv/sifive_u.c         | 24 +-----------------------
 include/hw/riscv/sifive_u.h |  3 +--
 2 files changed, 2 insertions(+), 25 deletions(-)
hw/riscv/sifive_u.c +1 -23
index 503db4b..1140c38 100644
@@ -88,8 +88,7 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     uint32_t *cells;
     char *nodename;
     char ethclk_names[] = "pclk\0hclk";
-    uint32_t plic_phandle, prci_phandle, ethclk_phandle, phandle = 1;
-    uint32_t uartclk_phandle;
+    uint32_t plic_phandle, prci_phandle, phandle = 1;
     uint32_t hfclk_phandle, rtcclk_phandle, phy_phandle;
 
     fdt = s->fdt = create_device_tree(&s->fdt_size);
@@ -249,17 +248,6 @@ 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);
-
     phy_phandle = phandle++;
     nodename = g_strdup_printf("/soc/ethernet@%lx",
         (long)memmap[SIFIVE_U_GEM].base);
@@ -293,16 +281,6 @@ static void create_fdt(SiFiveUState *s, const struct MemmapEntry *memmap,
     qemu_fdt_setprop_cell(fdt, nodename, "reg", 0x0);
     g_free(nodename);
 
-    uartclk_phandle = phandle++;
-    nodename = g_strdup_printf("/soc/uartclk");
-    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", 3686400);
-    qemu_fdt_setprop_cell(fdt, nodename, "phandle", uartclk_phandle);
-    uartclk_phandle = qemu_fdt_get_phandle(fdt, nodename);
-    g_free(nodename);
-
     nodename = g_strdup_printf("/soc/serial@%lx",
         (long)memmap[SIFIVE_U_UART0].base);
     qemu_fdt_add_subnode(fdt, nodename);
include/hw/riscv/sifive_u.h +1 -2
index d2b9d99..3bb87cb 100644
@@ -76,8 +76,7 @@ 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
Signature · 1 line
2.7.4
author
#35 lore raw

PATCH v5 30/30 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 v5: None
Changes in v4: None
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 1140c38..fae19fe 100644
@@ -97,8 +97,9 @@ 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
#36 lore raw
In reply to Bin Meng
Quoted 134 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

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 [1]

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 [2] was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. The original plan was to get the drop patch applied after this QEMU series is merged. However after discussion in the OpenSBI mailing list, it seems the best option for us is to let OpenSBI continue shipping the special "qemu/sifive_u" platform support to work with QEMU version <= 4.1 and deprecate the support sometime in the future. A patch will need to be sent to OpenSBI mailing list to update its document.

v4 is now rebased on Palmer's QEMU RISC-V repo "for-master" branch. Dropped the following v3 patch that was already done by someone else.
- riscv: sifive_u: Generate an aliases node in the device tree
- riscv: sifive_u: Support loading initramfs

The following v3 patch was dropped too due to a different cluster approach suggested by Richard Henderson is used in v4:
- riscv: hart: Support heterogeneous harts population

[1]: https://patchwork.kernel.org/patch/11050003/
[2]: http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models
- new patch to remove the unnecessary include of target/riscv/cpu.h
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_otp
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_otp_write()
- move all register initialization to sifive_u_otp_reset() function
- drop sifive_u_otp_create()
- create sifive_u_otp block directly in the machine codes, instead of calling sifive_u_otp_create()
- add the missing "local-mac-address" property in the ethernet node

Changes in v4:
- remove 2 more "linux,phandle" instances in sifive_u.c and spike.c after rebasing on Palmer's QEMU RISC-V tree
- change create_fdt() to return void in sifive_u.c too, after rebasing on Palmer's QEMU RISC-V tree
- new patch to remove executable attribute of opensbi images
- prefix all macros/variables/functions with SIFIVE_E/sifive_e in the sifive_e_prci driver
- new patch to add a "hartid-base" property to RISC-V hart array
- changed to create clusters for each cpu type
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_otp driver
- new patch to remove handcrafted clock nodes for UART and ethernet

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:
- 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()"
- 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

Bin Meng (30):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: hw: Remove not needed PLIC properties in device tree
riscv: hw: Change create_fdt() to return void
riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead
riscv: hw: Remove the unnecessary include of target/riscv/cpu.h
riscv: roms: Remove executable attribute of opensbi images
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_e: Drop sifive_mmio_emulate()
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: Add a "hartid-base" property to RISC-V hart array
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: 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: Reference PRCI clocks in UART and ethernet nodes

Quoted 8 lines

riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Change UART node name in device tree
riscv: roms: Update default bios for sifive_u machine
riscv: sifive: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Remove handcrafted clock nodes for UART and ethernet
riscv: sifive_u: Update model and compatible strings in device tree

I can't see patch 12 and it looks like it didn't make it to the list either:

https://patchew.org/QEMU/[email protected]/

Alistair

Quoted 43 lines
 hw/riscv/Makefile.objs                       |   4 +-
 hw/riscv/riscv_hart.c                        |  35 ++--
 hw/riscv/sifive_e.c                          |  27 +--
 hw/riscv/sifive_e_prci.c                     | 123 +++++++++++++
 hw/riscv/sifive_prci.c                       | 123 -------------
 hw/riscv/sifive_test.c                       |   6 +-
 hw/riscv/sifive_u.c                          | 262 ++++++++++++++++++---------
 hw/riscv/sifive_u_otp.c                      | 190 +++++++++++++++++++
 hw/riscv/sifive_u_prci.c                     | 171 +++++++++++++++++
 hw/riscv/sifive_uart.c                       |  10 +-
 hw/riscv/spike.c                             |   1 -
 hw/riscv/virt.c                              |  40 ++--
 include/hw/riscv/riscv_hart.h                |   1 +
 include/hw/riscv/sifive_cpu.h                |  31 ++++
 include/hw/riscv/sifive_e.h                  |   7 +-
 include/hw/riscv/sifive_e_prci.h             |  71 ++++++++
 include/hw/riscv/sifive_prci.h               |  69 -------
 include/hw/riscv/sifive_u.h                  |  31 ++--
 include/hw/riscv/sifive_u_otp.h              |  80 ++++++++
 include/hw/riscv/sifive_u_prci.h             |  91 ++++++++++
 pc-bios/opensbi-riscv32-virt-fw_jump.bin     | Bin
 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin | Bin 40968 -> 45064 bytes
 pc-bios/opensbi-riscv64-virt-fw_jump.bin     | Bin
 roms/Makefile                                |   4 +-
 24 files changed, 1018 insertions(+), 359 deletions(-)
 create mode 100644 hw/riscv/sifive_e_prci.c
 delete mode 100644 hw/riscv/sifive_prci.c
 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
 create mode 100644 include/hw/riscv/sifive_e_prci.h
 delete mode 100644 include/hw/riscv/sifive_prci.h
 create mode 100644 include/hw/riscv/sifive_u_otp.h
 create mode 100644 include/hw/riscv/sifive_u_prci.h
 mode change 100755 => 100644 pc-bios/opensbi-riscv32-virt-fw_jump.bin
 mode change 100755 => 100644 pc-bios/opensbi-riscv64-sifive_u-fw_jump.bin
 mode change 100755 => 100644 pc-bios/opensbi-riscv64-virt-fw_jump.bin

--
2.7.4

author
#37 lore raw

Hi Alistair,

Quoted 145 lines · Alistair Francis
Quoted 134 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

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 [1]

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 [2] was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. The original plan was to get the drop patch applied after this QEMU series is merged. However after discussion in the OpenSBI mailing list, it seems the best option for us is to let OpenSBI continue shipping the special "qemu/sifive_u" platform support to work with QEMU version <= 4.1 and deprecate the support sometime in the future. A patch will need to be sent to OpenSBI mailing list to update its document.

v4 is now rebased on Palmer's QEMU RISC-V repo "for-master" branch. Dropped the following v3 patch that was already done by someone else.
- riscv: sifive_u: Generate an aliases node in the device tree
- riscv: sifive_u: Support loading initramfs

The following v3 patch was dropped too due to a different cluster approach suggested by Richard Henderson is used in v4:
- riscv: hart: Support heterogeneous harts population

[1]: https://patchwork.kernel.org/patch/11050003/
[2]: http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models
- new patch to remove the unnecessary include of target/riscv/cpu.h
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_otp
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_otp_write()
- move all register initialization to sifive_u_otp_reset() function
- drop sifive_u_otp_create()
- create sifive_u_otp block directly in the machine codes, instead of calling sifive_u_otp_create()
- add the missing "local-mac-address" property in the ethernet node

Changes in v4:
- remove 2 more "linux,phandle" instances in sifive_u.c and spike.c after rebasing on Palmer's QEMU RISC-V tree
- change create_fdt() to return void in sifive_u.c too, after rebasing on Palmer's QEMU RISC-V tree
- new patch to remove executable attribute of opensbi images
- prefix all macros/variables/functions with SIFIVE_E/sifive_e in the sifive_e_prci driver
- new patch to add a "hartid-base" property to RISC-V hart array
- changed to create clusters for each cpu type
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_otp driver
- new patch to remove handcrafted clock nodes for UART and ethernet

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:
- 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()"
- 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

Bin Meng (30):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: hw: Remove not needed PLIC properties in device tree
riscv: hw: Change create_fdt() to return void
riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead
riscv: hw: Remove the unnecessary include of target/riscv/cpu.h
riscv: roms: Remove executable attribute of opensbi images
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_e: Drop sifive_mmio_emulate()
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: Add a "hartid-base" property to RISC-V hart array
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: 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: Reference PRCI clocks in UART and ethernet nodes

Quoted 8 lines

riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Change UART node name in device tree
riscv: roms: Update default bios for sifive_u machine
riscv: sifive: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Remove handcrafted clock nodes for UART and ethernet
riscv: sifive_u: Update model and compatible strings in device tree

Quoted 4 lines

I can't see patch 12 and it looks like it didn't make it to the list either:

https://patchew.org/QEMU/[email protected]/

Yes, it did not arrive at the mailing list for some weird unknown reason. This was seen in previous version too. Suspect there is something wrong with the QEMU mailing list.

But I think you should be able to find that in your inbox. I "bcc" another email address of mine when sending patches and confirmed all the patches in the series were received.

Regards,
Bin

#38 lore raw
Quoted 154 lines · Bin Meng

Hi Alistair,

Quoted 145 lines · Alistair Francis
Quoted 134 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

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 [1]

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 [2] was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. The original plan was to get the drop patch applied after this QEMU series is merged. However after discussion in the OpenSBI mailing list, it seems the best option for us is to let OpenSBI continue shipping the special "qemu/sifive_u" platform support to work with QEMU version <= 4.1 and deprecate the support sometime in the future. A patch will need to be sent to OpenSBI mailing list to update its document.

v4 is now rebased on Palmer's QEMU RISC-V repo "for-master" branch. Dropped the following v3 patch that was already done by someone else.
- riscv: sifive_u: Generate an aliases node in the device tree
- riscv: sifive_u: Support loading initramfs

The following v3 patch was dropped too due to a different cluster approach suggested by Richard Henderson is used in v4:
- riscv: hart: Support heterogeneous harts population

[1]: https://patchwork.kernel.org/patch/11050003/
[2]: http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models
- new patch to remove the unnecessary include of target/riscv/cpu.h
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_otp
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_otp_write()
- move all register initialization to sifive_u_otp_reset() function
- drop sifive_u_otp_create()
- create sifive_u_otp block directly in the machine codes, instead of calling sifive_u_otp_create()
- add the missing "local-mac-address" property in the ethernet node

Changes in v4:
- remove 2 more "linux,phandle" instances in sifive_u.c and spike.c after rebasing on Palmer's QEMU RISC-V tree
- change create_fdt() to return void in sifive_u.c too, after rebasing on Palmer's QEMU RISC-V tree
- new patch to remove executable attribute of opensbi images
- prefix all macros/variables/functions with SIFIVE_E/sifive_e in the sifive_e_prci driver
- new patch to add a "hartid-base" property to RISC-V hart array
- changed to create clusters for each cpu type
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_otp driver
- new patch to remove handcrafted clock nodes for UART and ethernet

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:
- 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()"
- 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

Bin Meng (30):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: hw: Remove not needed PLIC properties in device tree
riscv: hw: Change create_fdt() to return void
riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead
riscv: hw: Remove the unnecessary include of target/riscv/cpu.h
riscv: roms: Remove executable attribute of opensbi images
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_e: Drop sifive_mmio_emulate()
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: Add a "hartid-base" property to RISC-V hart array
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: 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: Reference PRCI clocks in UART and ethernet nodes

Quoted 8 lines

riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Change UART node name in device tree
riscv: roms: Update default bios for sifive_u machine
riscv: sifive: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Remove handcrafted clock nodes for UART and ethernet
riscv: sifive_u: Update model and compatible strings in device tree

Quoted 4 lines

I can't see patch 12 and it looks like it didn't make it to the list either:

https://patchew.org/QEMU/[email protected]/

Quoted 7 lines

Yes, it did not arrive at the mailing list for some weird unknown reason. This was seen in previous version too. Suspect there is something wrong with the QEMU mailing list.

But I think you should be able to find that in your inbox. I "bcc" another email address of mine when sending patches and confirmed all the patches in the series were received.

Ah, you are right. I didn't check my work email for the patch.

I have reviewed the patch.

Alistair

Quoted 3 lines

Regards,
Bin

#39 lore raw
In reply to Bin Meng
Quoted 134 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

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 [1]

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 [2] was sent to drop the special "qemu/sifive_u" platform support in OpenSBI. The original plan was to get the drop patch applied after this QEMU series is merged. However after discussion in the OpenSBI mailing list, it seems the best option for us is to let OpenSBI continue shipping the special "qemu/sifive_u" platform support to work with QEMU version <= 4.1 and deprecate the support sometime in the future. A patch will need to be sent to OpenSBI mailing list to update its document.

v4 is now rebased on Palmer's QEMU RISC-V repo "for-master" branch. Dropped the following v3 patch that was already done by someone else.
- riscv: sifive_u: Generate an aliases node in the device tree
- riscv: sifive_u: Support loading initramfs

The following v3 patch was dropped too due to a different cluster approach suggested by Richard Henderson is used in v4:
- riscv: hart: Support heterogeneous harts population

[1]: https://patchwork.kernel.org/patch/11050003/
[2]: http://lists.infradead.org/pipermail/opensbi/2019-August/000335.html

Changes in v5:
- new patch to change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead in various sifive models
- new patch to remove the unnecessary include of target/riscv/cpu.h
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_prci
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_prci_write()
- move all register initialization to sifive_u_prci_reset() function
- drop sifive_u_prci_create()
- s/codes that worked/code that works/g
- create sifive_u_prci block directly in the machine codes, instead of calling sifive_u_prci_create()
- change to use defines instead of enums
- change to use qemu_log_mask(LOG_GUEST_ERROR,...) in sifive_u_otp
- creating a 32-bit val variable and using that instead of casting everywhere in sifive_u_otp_write()
- move all register initialization to sifive_u_otp_reset() function
- drop sifive_u_otp_create()
- create sifive_u_otp block directly in the machine codes, instead of calling sifive_u_otp_create()
- add the missing "local-mac-address" property in the ethernet node

Changes in v4:
- remove 2 more "linux,phandle" instances in sifive_u.c and spike.c after rebasing on Palmer's QEMU RISC-V tree
- change create_fdt() to return void in sifive_u.c too, after rebasing on Palmer's QEMU RISC-V tree
- new patch to remove executable attribute of opensbi images
- prefix all macros/variables/functions with SIFIVE_E/sifive_e in the sifive_e_prci driver
- new patch to add a "hartid-base" property to RISC-V hart array
- changed to create clusters for each cpu type
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_prci driver
- prefix all macros/variables/functions with SIFIVE_U/sifive_u in the sifive_u_otp driver
- new patch to remove handcrafted clock nodes for UART and ethernet

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:
- 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()"
- 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

Bin Meng (30):
riscv: hw: Remove superfluous "linux,phandle" property
riscv: hw: Use qemu_fdt_setprop_cell() for property with only 1 cell
riscv: hw: Remove not needed PLIC properties in device tree
riscv: hw: Change create_fdt() to return void
riscv: hw: Change to use qemu_log_mask(LOG_GUEST_ERROR,...) instead
riscv: hw: Remove the unnecessary include of target/riscv/cpu.h
riscv: roms: Remove executable attribute of opensbi images
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_e: Drop sifive_mmio_emulate()
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: Add a "hartid-base" property to RISC-V hart array
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: 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: Reference PRCI clocks in UART and ethernet nodes

Quoted 8 lines

riscv: sifive_u: Update UART base addresses and IRQs
riscv: sifive_u: Change UART node name in device tree
riscv: roms: Update default bios for sifive_u machine
riscv: sifive: Implement a model for SiFive FU540 OTP
riscv: sifive_u: Instantiate OTP memory with a serial number
riscv: sifive_u: Fix broken GEM support
riscv: sifive_u: Remove handcrafted clock nodes for UART and ethernet
riscv: sifive_u: Update model and compatible strings in device tree

@Palmer, patches 1 - 11 are all reviewed and are good to go in. Can you take them in your next PR?

That'll save Bin from continuing to send a 30 patch series :)

Alistair

A message that is not in the archive (sent off-list or not yet received): [email protected]

Reviewed-by
#41 lore raw

Re: [Qemu-devel] [PATCH v5 12/30] riscv: sifive_e: Drop sifive_mmio_emulate()

Quoted 3 lines · Bin Meng

Use create_unimplemented_device() instead.

Alistair

Quoted 82 lines

---

Changes in v5: None
Changes in v4: None
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 +1 -8
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;
hw/riscv/sifive_e.c +7 -7
@@ -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 */