Skip to contentSkip to conversation

RISC-V SBI debug console extension support

PATCH Applied Anup Patel 6 participants 24 messages Oct 20, 2023 – Nov 17, 2023

Subsystems (6)

v3 of 6 Applied 5/9 Latest is v6 →

lore
Share & tools
lore

Anup Patel

author
#0 lore raw

v3 cover RISC-V SBI debug console extension support

The SBI v2.0 specification is now frozen. The SBI v2.0 specification defines SBI debug console (DBCN) extension which replaces the legacy SBI v0.1 functions sbi_console_putchar() and sbi_console_getchar(). (Refer v2.0-rc5 at https://github.com/riscv-non-isa/riscv-sbi-doc/releases)

This series adds support for SBI debug console (DBCN) extension in KVM RISC-V and Linux RISC-V.

To try these patches with KVM RISC-V, use KVMTOOL from riscv_sbi_dbcn_v1 branch at: https://github.com/avpatel/kvmtool.git

These patches can also be found in the riscv_sbi_dbcn_v3 branch at:
https://github.com/avpatel/linux.git

Changes since v2:
- Rebased on Linux-6.6-rc5
- Handled page-crossing in PATCH7 of v2 series
- Addressed Drew's comment in PATCH3 of v2 series
- Added new PATCH5 to make get-reg-list test aware of SBI DBCN extension

Changes since v1:
- Remove use of #ifdef from PATCH4 and PATCH5 of the v1 series
- Improved commit description of PATCH3 in v1 series
- Introduced new PATCH3 in this series to allow some SBI extensions (such as SBI DBCN) do to disabled by default so that older KVM user space work fine and newer KVM user space have to explicitly opt-in for emulating SBI DBCN.
- Introduced new PATCH5 in this series which adds inline version of sbi_console_getchar() and sbi_console_putchar() for the case where CONFIG_RISCV_SBI_V01 is disabled.

Anup Patel (8):
RISC-V: Add defines for SBI debug console extension
RISC-V: KVM: Change the SBI specification version to v2.0
RISC-V: KVM: Allow some SBI extensions to be disabled by default
RISC-V: KVM: Forward SBI DBCN extension to user-space
KVM: riscv: selftests: Add SBI DBCN extension to get-reg-list test
RISC-V: Add stubs for sbi_console_putchar/getchar() tty/serial: Add RISC-V SBI debug console based earlycon
RISC-V: Enable SBI based earlycon support

Atish Patra (1):
tty: Add SBI debug console support to HVC SBI driver

 arch/riscv/configs/defconfig                  |  1 +
 arch/riscv/configs/rv32_defconfig             |  1 +
 arch/riscv/include/asm/kvm_vcpu_sbi.h         |  7 +-
 arch/riscv/include/asm/sbi.h                  | 12 +++
 arch/riscv/include/uapi/asm/kvm.h             |  1 +
 arch/riscv/kvm/vcpu.c                         |  6 ++
 arch/riscv/kvm/vcpu_sbi.c                     | 61 +++++++-------
 arch/riscv/kvm/vcpu_sbi_replace.c             | 32 ++++++++
 drivers/tty/hvc/Kconfig                       |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c               | 82 +++++++++++++++++--
 drivers/tty/serial/Kconfig                    |  2 +-
 drivers/tty/serial/earlycon-riscv-sbi.c       | 32 +++++++-
 .../selftests/kvm/riscv/get-reg-list.c        |  2 +
 13 files changed, 198 insertions(+), 43 deletions(-)
Signature · 1 line
2.34.1
author
#1 lore raw

PATCH v3 1/9 RISC-V: Add defines for SBI debug console extension

We add SBI debug console extension related defines/enum to the asm/sbi.h header.

---
 arch/riscv/include/asm/sbi.h | 7 +++++++
 1 file changed, 7 insertions(+)
arch/riscv/include/asm/sbi.h +7 -0
index 5b4a1bf5f439..12dfda6bb924 100644
@@ -30,6 +30,7 @@ enum sbi_ext_id {
 	SBI_EXT_HSM = 0x48534D,
 	SBI_EXT_SRST = 0x53525354,
 	SBI_EXT_PMU = 0x504D55,
+	SBI_EXT_DBCN = 0x4442434E,
 
 	/* Experimentals extensions must lie within this range */
 	SBI_EXT_EXPERIMENTAL_START = 0x08000000,
@@ -236,6 +237,12 @@ enum sbi_pmu_ctr_type {
 /* Flags defined for counter stop function */
 #define SBI_PMU_STOP_FLAG_RESET (1 << 0)
 
+enum sbi_ext_dbcn_fid {
+	SBI_EXT_DBCN_CONSOLE_WRITE = 0,
+	SBI_EXT_DBCN_CONSOLE_READ = 1,
+	SBI_EXT_DBCN_CONSOLE_WRITE_BYTE = 2,
+};
+
 #define SBI_SPEC_VERSION_DEFAULT	0x1
 #define SBI_SPEC_VERSION_MAJOR_SHIFT	24
 #define SBI_SPEC_VERSION_MAJOR_MASK	0x7f
Signature · 1 line
2.34.1
author
#2 lore raw

PATCH v3 2/9 RISC-V: KVM: Change the SBI specification version to v2.0

In reply to Anup Patel

We will be implementing SBI DBCN extension for KVM RISC-V so let us change the KVM RISC-V SBI specification version to v2.0.

---
 arch/riscv/include/asm/kvm_vcpu_sbi.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/riscv/include/asm/kvm_vcpu_sbi.h +1 -1
index cdcf0ff07be7..8d6d4dce8a5e 100644
@@ -11,7 +11,7 @@
 
 #define KVM_SBI_IMPID 3
 
-#define KVM_SBI_VERSION_MAJOR 1
+#define KVM_SBI_VERSION_MAJOR 2
 #define KVM_SBI_VERSION_MINOR 0
 
 enum kvm_riscv_sbi_ext_status {
Signature · 1 line
2.34.1
author
#3 lore raw

PATCH v3 3/9 RISC-V: KVM: Allow some SBI extensions to be disabled by default

In reply to Anup Patel

Currently, all SBI extensions are enabled by default which is problematic for SBI extensions (such as DBCN) which are forwarded to the KVM user-space because we might have an older KVM user-space which is not aware/ready to handle newer SBI extensions. Ideally, the SBI extensions forwarded to the KVM user-space must be disabled by default.

To address above, we allow certain SBI extensions to be disabled by default so that KVM user-space must explicitly enable such SBI extensions to receive forwarded calls from Guest VCPU.

---
 arch/riscv/include/asm/kvm_vcpu_sbi.h |  4 ++
 arch/riscv/kvm/vcpu.c                 |  6 +++
 arch/riscv/kvm/vcpu_sbi.c             | 57 +++++++++++++--------------
 3 files changed, 38 insertions(+), 29 deletions(-)
arch/riscv/include/asm/kvm_vcpu_sbi.h +4 -0
index 8d6d4dce8a5e..c02bda5559d7 100644
@@ -35,6 +35,9 @@ struct kvm_vcpu_sbi_return {
 struct kvm_vcpu_sbi_extension {
 	unsigned long extid_start;
 	unsigned long extid_end;
+
+	bool default_unavail;
+
 	/**
 	 * SBI extension handler. It can be defined for a given extension or group of
 	 * extension. But it should always return linux error codes rather than SBI
@@ -59,6 +62,7 @@ int kvm_riscv_vcpu_get_reg_sbi_ext(struct kvm_vcpu *vcpu,
 const struct kvm_vcpu_sbi_extension *kvm_vcpu_sbi_find_ext(
 				struct kvm_vcpu *vcpu, unsigned long extid);
 int kvm_riscv_vcpu_sbi_ecall(struct kvm_vcpu *vcpu, struct kvm_run *run);
+void kvm_riscv_vcpu_sbi_init(struct kvm_vcpu *vcpu);
 
 #ifdef CONFIG_RISCV_SBI_V01
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_v01;
arch/riscv/kvm/vcpu.c +6 -0
index c061a1c5fe98..e087c809073c 100644
@@ -141,6 +141,12 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
 	if (rc)
 		return rc;
 
+	/*
+	 * Setup SBI extensions
+	 * NOTE: This must be the last thing to be initialized.
+	 */
+	kvm_riscv_vcpu_sbi_init(vcpu);
+
 	/* Reset VCPU */
 	kvm_riscv_reset_vcpu(vcpu);
 
arch/riscv/kvm/vcpu_sbi.c +28 -29

79 lines not on this page · raw message

Signature · 1 line
2.34.1
author
#5 lore raw

PATCH v3 4/9 RISC-V: KVM: Forward SBI DBCN extension to user-space

In reply to Anup Patel

The frozen SBI v2.0 specification defines the SBI debug console (DBCN) extension which replaces the legacy SBI v0.1 console functions namely sbi_console_getchar() and sbi_console_putchar().

The SBI DBCN extension needs to be emulated in the KVM user-space (i.e. QEMU-KVM or KVMTOOL) so we forward SBI DBCN calls from KVM guest to the KVM user-space which can then redirect the console input/output to wherever it wants (e.g. telnet, file, stdio, etc).

The SBI debug console is simply a early console available to KVM guest for early prints and it does not intend to replace the proper console devices such as 8250, VirtIO console, etc.

---
 arch/riscv/include/asm/kvm_vcpu_sbi.h |  1 +
 arch/riscv/include/uapi/asm/kvm.h     |  1 +
 arch/riscv/kvm/vcpu_sbi.c             |  4 ++++
 arch/riscv/kvm/vcpu_sbi_replace.c     | 32 +++++++++++++++++++++++++++
 4 files changed, 38 insertions(+)
arch/riscv/include/asm/kvm_vcpu_sbi.h +1 -0
index c02bda5559d7..6a453f7f8b56 100644
@@ -73,6 +73,7 @@ extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_ipi;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_rfence;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_srst;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_hsm;
+extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_dbcn;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_experimental;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_vendor;
 
arch/riscv/include/uapi/asm/kvm.h +1 -0
index 917d8cc2489e..60d3b21dead7 100644
@@ -156,6 +156,7 @@ enum KVM_RISCV_SBI_EXT_ID {
 	KVM_RISCV_SBI_EXT_PMU,
 	KVM_RISCV_SBI_EXT_EXPERIMENTAL,
 	KVM_RISCV_SBI_EXT_VENDOR,
+	KVM_RISCV_SBI_EXT_DBCN,
 	KVM_RISCV_SBI_EXT_MAX,
 };
 
arch/riscv/kvm/vcpu_sbi.c +4 -0
index bda8b0b33343..a04ff98085d9 100644
@@ -66,6 +66,10 @@ static const struct kvm_riscv_sbi_extension_entry sbi_ext[] = {
 		.ext_idx = KVM_RISCV_SBI_EXT_PMU,
 		.ext_ptr = &vcpu_sbi_ext_pmu,
 	},
+	{
+		.ext_idx = KVM_RISCV_SBI_EXT_DBCN,
+		.ext_ptr = &vcpu_sbi_ext_dbcn,
+	},
 	{
 		.ext_idx = KVM_RISCV_SBI_EXT_EXPERIMENTAL,
 		.ext_ptr = &vcpu_sbi_ext_experimental,
arch/riscv/kvm/vcpu_sbi_replace.c +32 -0
index 7c4d5d38a339..23b57c931b15 100644
@@ -175,3 +175,35 @@ const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_srst = {
 	.extid_end = SBI_EXT_SRST,
 	.handler = kvm_sbi_ext_srst_handler,
 };
+
+static int kvm_sbi_ext_dbcn_handler(struct kvm_vcpu *vcpu,
+				    struct kvm_run *run,
+				    struct kvm_vcpu_sbi_return *retdata)
+{
+	struct kvm_cpu_context *cp = &vcpu->arch.guest_context;
+	unsigned long funcid = cp->a6;
+
+	switch (funcid) {
+	case SBI_EXT_DBCN_CONSOLE_WRITE:
+	case SBI_EXT_DBCN_CONSOLE_READ:
+	case SBI_EXT_DBCN_CONSOLE_WRITE_BYTE:
+		/*
+		 * The SBI debug console functions are unconditionally
+		 * forwarded to the userspace.
+		 */
+		kvm_riscv_vcpu_sbi_forward(vcpu, run);
+		retdata->uexit = true;
+		break;
+	default:
+		retdata->err_val = SBI_ERR_NOT_SUPPORTED;
+	}
+
+	return 0;
+}
+
+const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_dbcn = {
+	.extid_start = SBI_EXT_DBCN,
+	.extid_end = SBI_EXT_DBCN,
+	.default_unavail = true,
+	.handler = kvm_sbi_ext_dbcn_handler,
+};
Signature · 1 line
2.34.1
author
#6 lore raw

PATCH v3 5/9 KVM: riscv: selftests: Add SBI DBCN extension to get-reg-list test

In reply to Anup Patel

We have a new SBI debug console (DBCN) extension supported by in-kernel KVM so let us add this extension to get-reg-list test.

---
 tools/testing/selftests/kvm/riscv/get-reg-list.c | 2 ++
 1 file changed, 2 insertions(+)
tools/testing/selftests/kvm/riscv/get-reg-list.c +2 -0
index 234006d035c9..6bedaea95395 100644
@@ -394,6 +394,7 @@ static const char *sbi_ext_single_id_to_str(__u64 reg_off)
 		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_PMU),
 		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_EXPERIMENTAL),
 		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_VENDOR),
+		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_DBCN),
 	};
 
 	if (reg_off >= ARRAY_SIZE(kvm_sbi_ext_reg_name))
@@ -567,6 +568,7 @@ static __u64 base_regs[] = {
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_PMU,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_EXPERIMENTAL,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_VENDOR,
+	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_DBCN,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_MULTI_EN | 0,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_MULTI_DIS | 0,
 };
Signature · 1 line
2.34.1
author
#8 lore raw

PATCH v3 6/9 RISC-V: Add stubs for sbi_console_putchar/getchar()

In reply to Anup Patel

The functions sbi_console_putchar() and sbi_console_getchar() are not defined when CONFIG_RISCV_SBI_V01 is disabled so let us add stub of these functions to avoid "#ifdef" on user side.

---
 arch/riscv/include/asm/sbi.h | 5 +++++
 1 file changed, 5 insertions(+)
arch/riscv/include/asm/sbi.h +5 -0
index 12dfda6bb924..cbcefa344417 100644
@@ -271,8 +271,13 @@ struct sbiret sbi_ecall(int ext, int fid, unsigned long arg0,
 			unsigned long arg3, unsigned long arg4,
 			unsigned long arg5);
 
+#ifdef CONFIG_RISCV_SBI_V01
 void sbi_console_putchar(int ch);
 int sbi_console_getchar(void);
+#else
+static inline void sbi_console_putchar(int ch) { }
+static inline int sbi_console_getchar(void) { return -1; }
+#endif
 long sbi_get_mvendorid(void);
 long sbi_get_marchid(void);
 long sbi_get_mimpid(void);
Signature · 1 line
2.34.1
author
#11 lore raw

PATCH v3 7/9 tty/serial: Add RISC-V SBI debug console based earlycon

In reply to Anup Patel

We extend the existing RISC-V SBI earlycon support to use the new RISC-V SBI debug console extension.

---
 drivers/tty/serial/Kconfig              |  2 +-
 drivers/tty/serial/earlycon-riscv-sbi.c | 32 +++++++++++++++++++++----
 2 files changed, 29 insertions(+), 5 deletions(-)
drivers/tty/serial/Kconfig +1 -1
index bdc568a4ab66..cec46091a716 100644
@@ -87,7 +87,7 @@ config SERIAL_EARLYCON_SEMIHOST
 
 config SERIAL_EARLYCON_RISCV_SBI
 	bool "Early console using RISC-V SBI"
-	depends on RISCV_SBI_V01
+	depends on RISCV_SBI
 	select SERIAL_CORE
 	select SERIAL_CORE_CONSOLE
 	select SERIAL_EARLYCON
drivers/tty/serial/earlycon-riscv-sbi.c +28 -4
index 27afb0b74ea7..c21cdef254e7 100644
@@ -15,17 +15,41 @@ static void sbi_putc(struct uart_port *port, unsigned char c)
 	sbi_console_putchar(c);
 }
 
-static void sbi_console_write(struct console *con,
-			      const char *s, unsigned n)
+static void sbi_0_1_console_write(struct console *con,
+				  const char *s, unsigned int n)
 {
 	struct earlycon_device *dev = con->data;
 	uart_console_write(&dev->port, s, n, sbi_putc);
 }
 
+static void sbi_dbcn_console_write(struct console *con,
+				   const char *s, unsigned int n)
+{
+	phys_addr_t pa = __pa(s);
+
+	if (IS_ENABLED(CONFIG_32BIT))
+		sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+			  n, lower_32_bits(pa), upper_32_bits(pa), 0, 0, 0);
+	else
+		sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+			  n, pa, 0, 0, 0, 0);
+}
+
 static int __init early_sbi_setup(struct earlycon_device *device,
 				  const char *opt)
 {
-	device->con->write = sbi_console_write;
-	return 0;
+	int ret = 0;
+
+	if ((sbi_spec_version >= sbi_mk_version(2, 0)) &&
+	    (sbi_probe_extension(SBI_EXT_DBCN) > 0)) {
+		device->con->write = sbi_dbcn_console_write;
+	} else {
+		if (IS_ENABLED(CONFIG_RISCV_SBI_V01))
+			device->con->write = sbi_0_1_console_write;
+		else
+			ret = -ENODEV;
+	}
+
+	return ret;
 }
 EARLYCON_DECLARE(sbi, early_sbi_setup);
Signature · 1 line
2.34.1
author
#14 lore raw

PATCH v3 8/9 tty: Add SBI debug console support to HVC SBI driver

In reply to Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
 	bool "RISC-V SBI console support"
-	depends on RISCV_SBI_V01
+	depends on RISCV_SBI
 	select HVC_DRIVER
 	help
 	  This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +75 -7

96 lines not on this page · raw message

Signature · 1 line
2.34.1
author
#21 lore raw

PATCH v3 9/9 RISC-V: Enable SBI based earlycon support

In reply to Anup Patel

Let us enable SBI based earlycon support in defconfigs for both RV32 and RV64 so that "earlycon=sbi" can be used again.

---
 arch/riscv/configs/defconfig      | 1 +
 arch/riscv/configs/rv32_defconfig | 1 +
 2 files changed, 2 insertions(+)
arch/riscv/configs/defconfig +1 -0
index ab86ec3b9eab..f82700da0056 100644
@@ -132,6 +132,7 @@ CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_SERIAL_8250_DW=y
 CONFIG_SERIAL_OF_PLATFORM=y
 CONFIG_SERIAL_SH_SCI=y
+CONFIG_SERIAL_EARLYCON_RISCV_SBI=y
 CONFIG_VIRTIO_CONSOLE=y
 CONFIG_HW_RANDOM=y
 CONFIG_HW_RANDOM_VIRTIO=y
arch/riscv/configs/rv32_defconfig +1 -0
index 89b601e253a6..5721af39afd1 100644
@@ -66,6 +66,7 @@ CONFIG_INPUT_MOUSEDEV=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_SERIAL_OF_PLATFORM=y
+CONFIG_SERIAL_EARLYCON_RISCV_SBI=y
 CONFIG_VIRTIO_CONSOLE=y
 CONFIG_HW_RANDOM=y
 CONFIG_HW_RANDOM_VIRTIO=y
Signature · 1 line
2.34.1
Reviewed-by on v3 3/9
#4 lore raw

Re: [PATCH v3 3/9] RISC-V: KVM: Allow some SBI extensions to be disabled by default

In reply to Anup Patel
Quoted 11 lines · Anup Patel

Currently, all SBI extensions are enabled by default which is problematic for SBI extensions (such as DBCN) which are forwarded to the KVM user-space because we might have an older KVM user-space which is not aware/ready to handle newer SBI extensions. Ideally, the SBI extensions forwarded to the KVM user-space must be disabled by default.

To address above, we allow certain SBI extensions to be disabled by default so that KVM user-space must explicitly enable such SBI extensions to receive forwarded calls from Guest VCPU.

Quoted 7 lines
---
 arch/riscv/include/asm/kvm_vcpu_sbi.h |  4 ++
 arch/riscv/kvm/vcpu.c                 |  6 +++
 arch/riscv/kvm/vcpu_sbi.c             | 57 +++++++++++++--------------
 3 files changed, 38 insertions(+), 29 deletions(-)
Reviewed-by on v3 5/9
#7 lore raw

Re: [PATCH v3 5/9] KVM: riscv: selftests: Add SBI DBCN extension to get-reg-list test

In reply to Anup Patel
Quoted 31 lines · Anup Patel

We have a new SBI debug console (DBCN) extension supported by in-kernel KVM so let us add this extension to get-reg-list test.

---
 tools/testing/selftests/kvm/riscv/get-reg-list.c | 2 ++
 1 file changed, 2 insertions(+)
tools/testing/selftests/kvm/riscv/get-reg-list.c +2 -0
index 234006d035c9..6bedaea95395 100644
@@ -394,6 +394,7 @@ static const char *sbi_ext_single_id_to_str(__u64 reg_off)
 		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_PMU),
 		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_EXPERIMENTAL),
 		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_VENDOR),
+		KVM_SBI_EXT_ARR(KVM_RISCV_SBI_EXT_DBCN),
 	};
 
 	if (reg_off >= ARRAY_SIZE(kvm_sbi_ext_reg_name))
@@ -567,6 +568,7 @@ static __u64 base_regs[] = {
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_PMU,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_EXPERIMENTAL,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_VENDOR,
+	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_SINGLE | KVM_RISCV_SBI_EXT_DBCN,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_MULTI_EN | 0,
 	KVM_REG_RISCV | KVM_REG_SIZE_ULONG | KVM_REG_RISCV_SBI_EXT | KVM_REG_RISCV_SBI_MULTI_DIS | 0,
 };
Signature · 1 line
2.34.1
#15 lore raw

Re: [PATCH v3 8/9] tty: Add SBI debug console support to HVC SBI driver

In reply to Anup Patel
Quoted 42 lines · Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
 	bool "RISC-V SBI console support"
-	depends on RISCV_SBI_V01
+	depends on RISCV_SBI
 	select HVC_DRIVER
 	help
 	  This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +1 -2
index 31f53fa77e4a..56da1a4b5aca 100644
@@ -39,21 +39,89 @@ static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count)
 	return i;
 }
 
-static const struct hv_ops hvc_sbi_ops = {
+static const struct hv_ops hvc_sbi_v01_ops = {
 	.get_chars = hvc_sbi_tty_get,
 	.put_chars = hvc_sbi_tty_put,
 };
 
-static int __init hvc_sbi_init(void)

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 55

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +8 -1
+static int hvc_sbi_dbcn_tty_put(uint32_t vtermno, const char *buf, int count)
 {
-	return PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_ops, 16));
+	phys_addr_t pa;
+	struct sbiret ret;
+
+	if (is_vmalloc_addr(buf)) {
+		pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+		if (PAGE_SIZE < (offset_in_page(buf) + count))
+			count = PAGE_SIZE - offset_in_page(buf);

Thanks for fixing the cross-page issue. Now you're cutting the buffer off. What about doing two SBI calls instead? (Dito on the get side)

Bj?rn

author
#16 lore raw

Re: [PATCH v3 8/9] tty: Add SBI debug console support to HVC SBI driver

Quoted 44 lines · On Fri, Oct 20, 2023 at 3:25?PM Bj?rn T?pel <[email protected]> wrote:
Quoted 41 lines · Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
      bool "RISC-V SBI console support"
-     depends on RISCV_SBI_V01
+     depends on RISCV_SBI
      select HVC_DRIVER
      help
        This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +1 -1
index 31f53fa77e4a..56da1a4b5aca 100644
@@ -39,21 +39,89 @@ static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count)
      return i;
 }
 
-static const struct hv_ops hvc_sbi_ops = {
+static const struct hv_ops hvc_sbi_v01_ops = {
      .get_chars = hvc_sbi_tty_get,
      .put_chars = hvc_sbi_tty_put,
 };
 
Quoted 14 lines

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 55

Quoted 11 lines
drivers/tty/hvc/hvc_riscv_sbi.c +8 -2
-static int __init hvc_sbi_init(void)
+static int hvc_sbi_dbcn_tty_put(uint32_t vtermno, const char *buf, int count)
 {
-     return PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_ops, 16));
+     phys_addr_t pa;
+     struct sbiret ret;
+
+     if (is_vmalloc_addr(buf)) {
+             pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+             if (PAGE_SIZE < (offset_in_page(buf) + count))
+                     count = PAGE_SIZE - offset_in_page(buf);

Thanks for fixing the cross-page issue. Now you're cutting the buffer off. What about doing two SBI calls instead? (Dito on the get side)

We don't need to handle that because the hvc_console framework will ensure remaining characters are sent-out. Same applies to get side as well.

Regards,
Anup

#17 lore raw

Re: [PATCH v3 8/9] tty: Add SBI debug console support to HVC SBI driver

In reply to Anup Patel
Quoted 41 lines · Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
 	bool "RISC-V SBI console support"
-	depends on RISCV_SBI_V01
+	depends on RISCV_SBI
 	select HVC_DRIVER
 	help
 	  This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +1 -1
index 31f53fa77e4a..56da1a4b5aca 100644
@@ -39,21 +39,89 @@ static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count)
 	return i;
 }
 
-static const struct hv_ops hvc_sbi_ops = {
+static const struct hv_ops hvc_sbi_v01_ops = {
 	.get_chars = hvc_sbi_tty_get,
 	.put_chars = hvc_sbi_tty_put,
 };
 

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 54

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +7 -2
-static int __init hvc_sbi_init(void)
+static int hvc_sbi_dbcn_tty_put(uint32_t vtermno, const char *buf, int count)
 {
-	return PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_ops, 16));
+	phys_addr_t pa;
+	struct sbiret ret;
+
+	if (is_vmalloc_addr(buf)) {
+		pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+		if (PAGE_SIZE < (offset_in_page(buf) + count))

I thought checkpatch complained about uppercase constants being on the left in comparisons.

Quoted 6 lines
drivers/tty/hvc/hvc_riscv_sbi.c +6 -0
+			count = PAGE_SIZE - offset_in_page(buf);
+	} else {
+		pa = __pa(buf);
+	}
+
+	if (IS_ENABLED(CONFIG_32BIT))

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 70

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +10 -0
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+				count, lower_32_bits(pa), upper_32_bits(pa),
+				0, 0, 0);
+	else
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+				count, pa, 0, 0, 0, 0);
+	if (ret.error)
+		return 0;
+
+	return count;

Shouldn't we return ret.value here in case it's less than count? I see we already do that below in get().

Quoted 33 lines
drivers/tty/hvc/hvc_riscv_sbi.c +27 -3
 }
-device_initcall(hvc_sbi_init);
 
-static int __init hvc_sbi_console_init(void)
+static int hvc_sbi_dbcn_tty_get(uint32_t vtermno, char *buf, int count)
 {
-	hvc_instantiate(0, 0, &hvc_sbi_ops);
+	phys_addr_t pa;
+	struct sbiret ret;
+
+	if (is_vmalloc_addr(buf)) {
+		pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+		if (PAGE_SIZE < (offset_in_page(buf) + count))
+			count = PAGE_SIZE - offset_in_page(buf);
+	} else {
+		pa = __pa(buf);
+	}
+
+	if (IS_ENABLED(CONFIG_32BIT))
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+				count, lower_32_bits(pa), upper_32_bits(pa),
+				0, 0, 0);
+	else
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+				count, pa, 0, 0, 0, 0);
+	if (ret.error)
+		return 0;
+
+	return ret.value;
+}
+
+static const struct hv_ops hvc_sbi_dbcn_ops = {
+	.put_chars = hvc_sbi_dbcn_tty_put,

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 110

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +10 -0
+	.get_chars = hvc_sbi_dbcn_tty_get,
+};
+
+static int __init hvc_sbi_init(void)
+{
+	int err;
+
+	if ((sbi_spec_version >= sbi_mk_version(2, 0)) &&
+	    (sbi_probe_extension(SBI_EXT_DBCN) > 0)) {
+		err = PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_dbcn_ops, 16));

Why an outbuf size of only 16?

Quoted 21 lines
drivers/tty/hvc/hvc_riscv_sbi.c +14 -1
+		if (err)
+			return err;
+		hvc_instantiate(0, 0, &hvc_sbi_dbcn_ops);
+	} else {
+		if (IS_ENABLED(CONFIG_RISCV_SBI_V01)) {
+			err = PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_v01_ops, 16));
+			if (err)
+				return err;
+			hvc_instantiate(0, 0, &hvc_sbi_v01_ops);
+		} else {
+			return -ENODEV;
+		}
+	}
 
 	return 0;
 }
-console_initcall(hvc_sbi_console_init);
+device_initcall(hvc_sbi_init);
Signature · 1 line
2.34.1

Thanks,
drew

maintainer Applied
#22 lore raw
In reply to Anup Patel
Quoted 42 lines · On Fri, Oct 20, 2023 at 12:51?PM Anup Patel <[email protected]> wrote:

The SBI v2.0 specification is now frozen. The SBI v2.0 specification defines SBI debug console (DBCN) extension which replaces the legacy SBI v0.1 functions sbi_console_putchar() and sbi_console_getchar(). (Refer v2.0-rc5 at https://github.com/riscv-non-isa/riscv-sbi-doc/releases)

This series adds support for SBI debug console (DBCN) extension in KVM RISC-V and Linux RISC-V.

To try these patches with KVM RISC-V, use KVMTOOL from riscv_sbi_dbcn_v1 branch at: https://github.com/avpatel/kvmtool.git

These patches can also be found in the riscv_sbi_dbcn_v3 branch at:
https://github.com/avpatel/linux.git

Changes since v2:
- Rebased on Linux-6.6-rc5
- Handled page-crossing in PATCH7 of v2 series
- Addressed Drew's comment in PATCH3 of v2 series
- Added new PATCH5 to make get-reg-list test aware of SBI DBCN extension

Changes since v1:
- Remove use of #ifdef from PATCH4 and PATCH5 of the v1 series
- Improved commit description of PATCH3 in v1 series
- Introduced new PATCH3 in this series to allow some SBI extensions (such as SBI DBCN) do to disabled by default so that older KVM user space work fine and newer KVM user space have to explicitly opt-in for emulating SBI DBCN.
- Introduced new PATCH5 in this series which adds inline version of sbi_console_getchar() and sbi_console_putchar() for the case where CONFIG_RISCV_SBI_V01 is disabled.

Anup Patel (8):
RISC-V: Add defines for SBI debug console extension
RISC-V: KVM: Change the SBI specification version to v2.0
RISC-V: KVM: Allow some SBI extensions to be disabled by default
RISC-V: KVM: Forward SBI DBCN extension to user-space
KVM: riscv: selftests: Add SBI DBCN extension to get-reg-list test
RISC-V: Add stubs for sbi_console_putchar/getchar() tty/serial: Add RISC-V SBI debug console based earlycon
RISC-V: Enable SBI based earlycon support

Quoted 2 lines

Atish Patra (1):
tty: Add SBI debug console support to HVC SBI driver

Queued PATCH1 to PATCH5 for Linux-6.7

Remaining PATCH6 to PATCH9 are still under review.

Thanks,
Anup

Quoted 24 lines
 arch/riscv/configs/defconfig                  |  1 +
 arch/riscv/configs/rv32_defconfig             |  1 +
 arch/riscv/include/asm/kvm_vcpu_sbi.h         |  7 +-
 arch/riscv/include/asm/sbi.h                  | 12 +++
 arch/riscv/include/uapi/asm/kvm.h             |  1 +
 arch/riscv/kvm/vcpu.c                         |  6 ++
 arch/riscv/kvm/vcpu_sbi.c                     | 61 +++++++-------
 arch/riscv/kvm/vcpu_sbi_replace.c             | 32 ++++++++
 drivers/tty/hvc/Kconfig                       |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c               | 82 +++++++++++++++++--
 drivers/tty/serial/Kconfig                    |  2 +-
 drivers/tty/serial/earlycon-riscv-sbi.c       | 32 +++++++-
 .../selftests/kvm/riscv/get-reg-list.c        |  2 +
 13 files changed, 198 insertions(+), 43 deletions(-)

--
2.34.1

--
kvm-riscv mailing list
kvm-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kvm-riscv

author
#18 lore raw

Re: [PATCH v3 8/9] tty: Add SBI debug console support to HVC SBI driver

In reply to Andrew Jones
Quoted 43 lines · On Fri, Oct 20, 2023 at 4:16?PM Andrew Jones <[email protected]> wrote:
Quoted 41 lines · Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
      bool "RISC-V SBI console support"
-     depends on RISCV_SBI_V01
+     depends on RISCV_SBI
      select HVC_DRIVER
      help
        This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +1 -1
index 31f53fa77e4a..56da1a4b5aca 100644
@@ -39,21 +39,89 @@ static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count)
      return i;
 }
 
-static const struct hv_ops hvc_sbi_ops = {
+static const struct hv_ops hvc_sbi_v01_ops = {
      .get_chars = hvc_sbi_tty_get,
      .put_chars = hvc_sbi_tty_put,
 };
 
Quoted 13 lines

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 54

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +7 -2
-static int __init hvc_sbi_init(void)
+static int hvc_sbi_dbcn_tty_put(uint32_t vtermno, const char *buf, int count)
 {
-     return PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_ops, 16));
+     phys_addr_t pa;
+     struct sbiret ret;
+
+     if (is_vmalloc_addr(buf)) {
+             pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+             if (PAGE_SIZE < (offset_in_page(buf) + count))

I thought checkpatch complained about uppercase constants being on the left in comparisons.

Nope checkpatch does not complain about this.

Quoted 12 lines
Quoted 12 lines
drivers/tty/hvc/hvc_riscv_sbi.c +11 -0
+                     count = PAGE_SIZE - offset_in_page(buf);
+     } else {
+             pa = __pa(buf);
+     }
+
+     if (IS_ENABLED(CONFIG_32BIT))
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+                             count, lower_32_bits(pa), upper_32_bits(pa),
+                             0, 0, 0);
+     else
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
Quoted 8 lines

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 70

Quoted 5 lines
drivers/tty/hvc/hvc_riscv_sbi.c +5 -0
+                             count, pa, 0, 0, 0, 0);
+     if (ret.error)
+             return 0;
+
+     return count;

Shouldn't we return ret.value here in case it's less than count? I see we already do that below in get().

Ahh, yes. Good catch, I will update.

Quoted 38 lines
Quoted 38 lines
drivers/tty/hvc/hvc_riscv_sbi.c +31 -3
 }
-device_initcall(hvc_sbi_init);
 
-static int __init hvc_sbi_console_init(void)
+static int hvc_sbi_dbcn_tty_get(uint32_t vtermno, char *buf, int count)
 {
-     hvc_instantiate(0, 0, &hvc_sbi_ops);
+     phys_addr_t pa;
+     struct sbiret ret;
+
+     if (is_vmalloc_addr(buf)) {
+             pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+             if (PAGE_SIZE < (offset_in_page(buf) + count))
+                     count = PAGE_SIZE - offset_in_page(buf);
+     } else {
+             pa = __pa(buf);
+     }
+
+     if (IS_ENABLED(CONFIG_32BIT))
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+                             count, lower_32_bits(pa), upper_32_bits(pa),
+                             0, 0, 0);
+     else
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+                             count, pa, 0, 0, 0, 0);
+     if (ret.error)
+             return 0;
+
+     return ret.value;
+}
+
+static const struct hv_ops hvc_sbi_dbcn_ops = {
+     .put_chars = hvc_sbi_dbcn_tty_put,
+     .get_chars = hvc_sbi_dbcn_tty_get,
+};
+
+static int __init hvc_sbi_init(void)
Quoted 8 lines

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 110

Quoted 6 lines
drivers/tty/hvc/hvc_riscv_sbi.c +6 -0
+{
+     int err;
+
+     if ((sbi_spec_version >= sbi_mk_version(2, 0)) &&
+         (sbi_probe_extension(SBI_EXT_DBCN) > 0)) {
+             err = PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_dbcn_ops, 16));

Why an outbuf size of only 16?

The output buffer size of 16 is a very common choice across HVC drivers. The next best choice is 256.

I guess 256 is better so I will go with that.

Quoted 15 lines
Quoted 15 lines
drivers/tty/hvc/hvc_riscv_sbi.c +13 -0
+             if (err)
+                     return err;
+             hvc_instantiate(0, 0, &hvc_sbi_dbcn_ops);
+     } else {
+             if (IS_ENABLED(CONFIG_RISCV_SBI_V01)) {
+                     err = PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_v01_ops, 16));
+                     if (err)
+                             return err;
+                     hvc_instantiate(0, 0, &hvc_sbi_v01_ops);
+             } else {
+                     return -ENODEV;
+             }
+     }
 
Quoted 10 lines
Quoted 7 lines
drivers/tty/hvc/hvc_riscv_sbi.c +1 -2
      return 0;
 }
-console_initcall(hvc_sbi_console_init);
+device_initcall(hvc_sbi_init);
--

2.34.1

Thanks,
drew

Regards,
Anup

#9 lore raw

Re: [PATCH v3 6/9] RISC-V: Add stubs for sbi_console_putchar/getchar()

In reply to Anup Patel
Quoted 15 lines · Anup Patel

The functions sbi_console_putchar() and sbi_console_getchar() are not defined when CONFIG_RISCV_SBI_V01 is disabled so let us add stub of these functions to avoid "#ifdef" on user side.

---
 arch/riscv/include/asm/sbi.h | 5 +++++
 1 file changed, 5 insertions(+)
arch/riscv/include/asm/sbi.h +0 -0
index 12dfda6bb924..cbcefa344417 100644

arch/riscv/include/asm/sbi.h struct sbiret sbi_ecall(int ext, int fid, unsigned long arg0, · line 279

Quoted 9 lines
arch/riscv/include/asm/sbi.h +4 -0
@@ -271,8 +271,13 @@ struct sbiret sbi_ecall(int ext, int fid, unsigned long arg0,
 			unsigned long arg3, unsigned long arg4,
 			unsigned long arg5);
 
+#ifdef CONFIG_RISCV_SBI_V01
 void sbi_console_putchar(int ch);
 int sbi_console_getchar(void);
+#else
+static inline void sbi_console_putchar(int ch) { }
+static inline int sbi_console_getchar(void) { return -1; }

Why not return a real error, "-1" isn't that :)

thanks,

greg k-h

#12 lore raw

Re: [PATCH v3 7/9] tty/serial: Add RISC-V SBI debug console based earlycon

In reply to Anup Patel
Quoted 41 lines · Anup Patel

We extend the existing RISC-V SBI earlycon support to use the new RISC-V SBI debug console extension.

---
 drivers/tty/serial/Kconfig              |  2 +-
 drivers/tty/serial/earlycon-riscv-sbi.c | 32 +++++++++++++++++++++----
 2 files changed, 29 insertions(+), 5 deletions(-)
drivers/tty/serial/Kconfig +1 -1
index bdc568a4ab66..cec46091a716 100644
@@ -87,7 +87,7 @@ config SERIAL_EARLYCON_SEMIHOST
 
 config SERIAL_EARLYCON_RISCV_SBI
 	bool "Early console using RISC-V SBI"
-	depends on RISCV_SBI_V01
+	depends on RISCV_SBI
 	select SERIAL_CORE
 	select SERIAL_CORE_CONSOLE
 	select SERIAL_EARLYCON
drivers/tty/serial/earlycon-riscv-sbi.c +3 -2
index 27afb0b74ea7..c21cdef254e7 100644
@@ -15,17 +15,41 @@ static void sbi_putc(struct uart_port *port, unsigned char c)
 	sbi_console_putchar(c);
 }
 
-static void sbi_console_write(struct console *con,
-			      const char *s, unsigned n)
+static void sbi_0_1_console_write(struct console *con,
+				  const char *s, unsigned int n)
 {
 	struct earlycon_device *dev = con->data;
 	uart_console_write(&dev->port, s, n, sbi_putc);
 }
 
+static void sbi_dbcn_console_write(struct console *con,

drivers/tty/serial/earlycon-riscv-sbi.c static void sbi_putc(struct uart_port *port, unsigned char c) · line 35

Quoted 10 lines
drivers/tty/serial/earlycon-riscv-sbi.c +10 -0
+				   const char *s, unsigned int n)
+{
+	phys_addr_t pa = __pa(s);
+
+	if (IS_ENABLED(CONFIG_32BIT))
+		sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+			  n, lower_32_bits(pa), upper_32_bits(pa), 0, 0, 0);
+	else
+		sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+			  n, pa, 0, 0, 0, 0);

This is still a bit hard to follow, and I guarantee it will be a pain to maintain over time, trying to keep both calls in sync, right?

Why not fix up sbi_ecall() to get this correct instead? It should be handling phys_addr_t values, not forcing you to do odd bit masking every single time you call it, right? That would make things much easier overall, and this patch simpler, as well as the next one.

Oh wait, sbi_ecall() is crazy, and just a pass-through, so that's not going to work, you need a wrapper function for this mess to do that bit twiddeling for you instead of forcing you to do it each time, I guess that's what you are trying to do here, but ick, is it correct?

thanks,

greg k-h

#19 lore raw

Re: [PATCH v3 8/9] tty: Add SBI debug console support to HVC SBI driver

In reply to Anup Patel
Quoted 53 lines · Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
 	bool "RISC-V SBI console support"
-	depends on RISCV_SBI_V01
+	depends on RISCV_SBI
 	select HVC_DRIVER
 	help
 	  This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +10 -3
index 31f53fa77e4a..56da1a4b5aca 100644
@@ -39,21 +39,89 @@ static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count)
 	return i;
 }
 
-static const struct hv_ops hvc_sbi_ops = {
+static const struct hv_ops hvc_sbi_v01_ops = {
 	.get_chars = hvc_sbi_tty_get,
 	.put_chars = hvc_sbi_tty_put,
 };
 
-static int __init hvc_sbi_init(void)
+static int hvc_sbi_dbcn_tty_put(uint32_t vtermno, const char *buf, int count)
 {
-	return PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_ops, 16));
+	phys_addr_t pa;
+	struct sbiret ret;
+
+	if (is_vmalloc_addr(buf)) {
+		pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+		if (PAGE_SIZE < (offset_in_page(buf) + count))
+			count = PAGE_SIZE - offset_in_page(buf);
+	} else {

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 66

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +10 -0
+		pa = __pa(buf);
+	}
+
+	if (IS_ENABLED(CONFIG_32BIT))
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+				count, lower_32_bits(pa), upper_32_bits(pa),
+				0, 0, 0);
+	else
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+				count, pa, 0, 0, 0, 0);

Again, you need a helper function here to keep you from having to keep this all in sync.

Quoted 19 lines
drivers/tty/hvc/hvc_riscv_sbi.c +13 -3
+	if (ret.error)
+		return 0;
+
+	return count;
 }
-device_initcall(hvc_sbi_init);
 
-static int __init hvc_sbi_console_init(void)
+static int hvc_sbi_dbcn_tty_get(uint32_t vtermno, char *buf, int count)
 {
-	hvc_instantiate(0, 0, &hvc_sbi_ops);
+	phys_addr_t pa;
+	struct sbiret ret;
+
+	if (is_vmalloc_addr(buf)) {
+		pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+		if (PAGE_SIZE < (offset_in_page(buf) + count))
+			count = PAGE_SIZE - offset_in_page(buf);
+	} else {

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 92

Quoted 10 lines
drivers/tty/hvc/hvc_riscv_sbi.c +10 -0
+		pa = __pa(buf);
+	}
+
+	if (IS_ENABLED(CONFIG_32BIT))
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+				count, lower_32_bits(pa), upper_32_bits(pa),
+				0, 0, 0);
+	else
+		ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+				count, pa, 0, 0, 0, 0);

And here too.

thanks,

greg k-h

#23 lore raw
In reply to Anup Patel
This series was applied to riscv/linux.git (fixes) by Anup Patel <[email protected]>:

Hello:

This series was applied to riscv/linux.git (fixes) by Anup Patel <[email protected]>:

Quoted 9 lines · On Fri, 20 Oct 2023 12:51:31 +0530 you wrote:

The SBI v2.0 specification is now frozen. The SBI v2.0 specification defines SBI debug console (DBCN) extension which replaces the legacy SBI v0.1 functions sbi_console_putchar() and sbi_console_getchar(). (Refer v2.0-rc5 at https://github.com/riscv-non-isa/riscv-sbi-doc/releases)

This series adds support for SBI debug console (DBCN) extension in KVM RISC-V and Linux RISC-V.

[...]

Here is the summary with links:
- [v3,1/9] RISC-V: Add defines for SBI debug console extension

    https://git.kernel.org/riscv/c/dadf7886993c
  - [v3,2/9] RISC-V: KVM: Change the SBI specification version to v2.0
    https://git.kernel.org/riscv/c/b88e87a0a1ec
  - [v3,3/9] RISC-V: KVM: Allow some SBI extensions to be disabled by default
    https://git.kernel.org/riscv/c/56d8a385b605
  - [v3,4/9] RISC-V: KVM: Forward SBI DBCN extension to user-space
    https://git.kernel.org/riscv/c/c667ad229d13
  - [v3,5/9] KVM: riscv: selftests: Add SBI DBCN extension to get-reg-list test
    https://git.kernel.org/riscv/c/d9c00f44e5de

- [v3,6/9] RISC-V: Add stubs for sbi_console_putchar/getchar() (no matching commit)
- [v3,7/9] tty/serial: Add RISC-V SBI debug console based earlycon (no matching commit)
- [v3,8/9] tty: Add SBI debug console support to HVC SBI driver (no matching commit)
- [v3,9/9] RISC-V: Enable SBI based earlycon support (no matching commit)

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
maintainer
#10 lore raw

Re: [PATCH v3 6/9] RISC-V: Add stubs for sbi_console_putchar/getchar()

In reply to Greg Kroah-Hartman
Quoted 20 lines · On Sat, Oct 21, 2023 at 10:05?PM Greg Kroah-Hartman…
Quoted 18 lines · Anup Patel

The functions sbi_console_putchar() and sbi_console_getchar() are not defined when CONFIG_RISCV_SBI_V01 is disabled so let us add stub of these functions to avoid "#ifdef" on user side.

---
 arch/riscv/include/asm/sbi.h | 5 +++++
 1 file changed, 5 insertions(+)
arch/riscv/include/asm/sbi.h +0 -0
index 12dfda6bb924..cbcefa344417 100644
@@ -271,8 +271,13 @@ struct sbiret sbi_ecall(int ext, int fid, unsigned long arg0,
                      unsigned long arg3, unsigned long arg4,
                      unsigned long arg5);
 
Quoted 8 lines

arch/riscv/include/asm/sbi.h struct sbiret sbi_ecall(int ext, int fid, unsigned long arg0, · line 279

Quoted 6 lines
arch/riscv/include/asm/sbi.h +4 -0
+#ifdef CONFIG_RISCV_SBI_V01
 void sbi_console_putchar(int ch);
 int sbi_console_getchar(void);
+#else
+static inline void sbi_console_putchar(int ch) { }
+static inline int sbi_console_getchar(void) { return -1; }

Why not return a real error, "-1" isn't that :)

As-per SBI spec, the legacy sbi_console_getchar() returns -1 upon failure hence the code.

Refer, section 5.3 of the latest SBI spec
https://github.com/riscv-non-isa/riscv-sbi-doc/releases/download/commit-fe4562532a9cc57e5743b6466946c5e5c98c73ca/riscv-sbi.pdf

Although, the users of this function only expect a negative value upon failure so better to return proper error code here.

I will update.

Quoted 9 lines

thanks,

greg k-h

--
kvm-riscv mailing list
kvm-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kvm-riscv

Regards,
Anup

maintainer
#13 lore raw

Re: [PATCH v3 7/9] tty/serial: Add RISC-V SBI debug console based earlycon

In reply to Greg Kroah-Hartman
Quoted 57 lines · On Sat, Oct 21, 2023 at 10:16?PM Greg Kroah-Hartman…
Quoted 41 lines · Anup Patel

We extend the existing RISC-V SBI earlycon support to use the new RISC-V SBI debug console extension.

---
 drivers/tty/serial/Kconfig              |  2 +-
 drivers/tty/serial/earlycon-riscv-sbi.c | 32 +++++++++++++++++++++----
 2 files changed, 29 insertions(+), 5 deletions(-)
drivers/tty/serial/Kconfig +1 -1
index bdc568a4ab66..cec46091a716 100644
@@ -87,7 +87,7 @@ config SERIAL_EARLYCON_SEMIHOST
 
 config SERIAL_EARLYCON_RISCV_SBI
      bool "Early console using RISC-V SBI"
-     depends on RISCV_SBI_V01
+     depends on RISCV_SBI
      select SERIAL_CORE
      select SERIAL_CORE_CONSOLE
      select SERIAL_EARLYCON
drivers/tty/serial/earlycon-riscv-sbi.c +3 -2
index 27afb0b74ea7..c21cdef254e7 100644
@@ -15,17 +15,41 @@ static void sbi_putc(struct uart_port *port, unsigned char c)
      sbi_console_putchar(c);
 }
 
-static void sbi_console_write(struct console *con,
-                           const char *s, unsigned n)
+static void sbi_0_1_console_write(struct console *con,
+                               const char *s, unsigned int n)
 {
      struct earlycon_device *dev = con->data;
      uart_console_write(&dev->port, s, n, sbi_putc);
 }
 
+static void sbi_dbcn_console_write(struct console *con,

drivers/tty/serial/earlycon-riscv-sbi.c static void sbi_putc(struct uart_port *port, unsigned char c) · line 35

Quoted 10 lines
drivers/tty/serial/earlycon-riscv-sbi.c +10 -0
+                                const char *s, unsigned int n)
+{
+     phys_addr_t pa = __pa(s);
+
+     if (IS_ENABLED(CONFIG_32BIT))
+             sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+                       n, lower_32_bits(pa), upper_32_bits(pa), 0, 0, 0);
+     else
+             sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+                       n, pa, 0, 0, 0, 0);

This is still a bit hard to follow, and I guarantee it will be a pain to maintain over time, trying to keep both calls in sync, right?

Quoted 4 lines

Why not fix up sbi_ecall() to get this correct instead? It should be handling phys_addr_t values, not forcing you to do odd bit masking every single time you call it, right? That would make things much easier overall, and this patch simpler, as well as the next one.

On RV32 systems, the physical address can be 34bits wide hence the on RV32 we have to pass physical address as two parameters whereas on RV64 entier physical address can be passed as single parameter.

Quoted 5 lines

Oh wait, sbi_ecall() is crazy, and just a pass-through, so that's not going to work, you need a wrapper function for this mess to do that bit twiddeling for you instead of forcing you to do it each time, I guess that's what you are trying to do here, but ick, is it correct?

Yes, it is better to have a wrapper function to hide the differences of RV32 and RV64 systems. I will update.

Quoted 9 lines

thanks,

greg k-h

--
kvm-riscv mailing list
kvm-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kvm-riscv

Regards,
Anup

maintainer
#20 lore raw

Re: [PATCH v3 8/9] tty: Add SBI debug console support to HVC SBI driver

In reply to Greg Kroah-Hartman
Quoted 60 lines · On Sat, Oct 21, 2023 at 10:16?PM Greg Kroah-Hartman…
Quoted 58 lines · Anup Patel

From: Atish Patra <[email protected]>

RISC-V SBI specification supports advanced debug console support via SBI DBCN extension.

Extend the HVC SBI driver to support it.

---
 drivers/tty/hvc/Kconfig         |  2 +-
 drivers/tty/hvc/hvc_riscv_sbi.c | 82 ++++++++++++++++++++++++++++++---
 2 files changed, 76 insertions(+), 8 deletions(-)
drivers/tty/hvc/Kconfig +1 -1
index 4f9264d005c0..6e05c5c7bca1 100644
@@ -108,7 +108,7 @@ config HVC_DCC_SERIALIZE_SMP
 
 config HVC_RISCV_SBI
      bool "RISC-V SBI console support"
-     depends on RISCV_SBI_V01
+     depends on RISCV_SBI
      select HVC_DRIVER
      help
        This enables support for console output via RISC-V SBI calls, which
drivers/tty/hvc/hvc_riscv_sbi.c +15 -3
index 31f53fa77e4a..56da1a4b5aca 100644
@@ -39,21 +39,89 @@ static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count)
      return i;
 }
 
-static const struct hv_ops hvc_sbi_ops = {
+static const struct hv_ops hvc_sbi_v01_ops = {
      .get_chars = hvc_sbi_tty_get,
      .put_chars = hvc_sbi_tty_put,
 };
 
-static int __init hvc_sbi_init(void)
+static int hvc_sbi_dbcn_tty_put(uint32_t vtermno, const char *buf, int count)
 {
-     return PTR_ERR_OR_ZERO(hvc_alloc(0, 0, &hvc_sbi_ops, 16));
+     phys_addr_t pa;
+     struct sbiret ret;
+
+     if (is_vmalloc_addr(buf)) {
+             pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+             if (PAGE_SIZE < (offset_in_page(buf) + count))
+                     count = PAGE_SIZE - offset_in_page(buf);
+     } else {
+             pa = __pa(buf);
+     }
+
+     if (IS_ENABLED(CONFIG_32BIT))
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
Quoted 8 lines

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 66

Quoted 5 lines
drivers/tty/hvc/hvc_riscv_sbi.c +5 -0
+                             count, lower_32_bits(pa), upper_32_bits(pa),
+                             0, 0, 0);
+     else
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_WRITE,
+                             count, pa, 0, 0, 0, 0);

Again, you need a helper function here to keep you from having to keep this all in sync.

Sure, I will update.

Quoted 24 lines
Quoted 24 lines
drivers/tty/hvc/hvc_riscv_sbi.c +17 -3
+     if (ret.error)
+             return 0;
+
+     return count;
 }
-device_initcall(hvc_sbi_init);
 
-static int __init hvc_sbi_console_init(void)
+static int hvc_sbi_dbcn_tty_get(uint32_t vtermno, char *buf, int count)
 {
-     hvc_instantiate(0, 0, &hvc_sbi_ops);
+     phys_addr_t pa;
+     struct sbiret ret;
+
+     if (is_vmalloc_addr(buf)) {
+             pa = page_to_phys(vmalloc_to_page(buf)) + offset_in_page(buf);
+             if (PAGE_SIZE < (offset_in_page(buf) + count))
+                     count = PAGE_SIZE - offset_in_page(buf);
+     } else {
+             pa = __pa(buf);
+     }
+
+     if (IS_ENABLED(CONFIG_32BIT))
Quoted 8 lines

drivers/tty/hvc/hvc_riscv_sbi.c static int hvc_sbi_tty_get(uint32_t vtermno, char *buf, int count) · line 92

Quoted 6 lines
drivers/tty/hvc/hvc_riscv_sbi.c +6 -0
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+                             count, lower_32_bits(pa), upper_32_bits(pa),
+                             0, 0, 0);
+     else
+             ret = sbi_ecall(SBI_EXT_DBCN, SBI_EXT_DBCN_CONSOLE_READ,
+                             count, pa, 0, 0, 0, 0);

And here too.

Okay.

Quoted 4 lines

thanks,

greg k-h

Regards,
Anup