* Re: [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend
2024-02-12 2:26 ` [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend Samuel Holland
@ 2024-02-12 7:59 ` Conor Dooley
2024-02-12 10:19 ` Andrew Jones
2024-02-12 13:21 ` Stefan O'Rear
2 siblings, 0 replies; 9+ messages in thread
From: Conor Dooley @ 2024-02-12 7:59 UTC (permalink / raw)
To: Samuel Holland
Cc: Andrew Jones, Palmer Dabbelt, linux-riscv, linux-kernel, stable
[-- Attachment #1: Type: text/plain, Size: 1454 bytes --]
On Sun, Feb 11, 2024 at 06:26:15PM -0800, Samuel Holland wrote:
> The value of the [ms]envcfg CSR is lost when entering a nonretentive
> idle state, so the CSR must be rewritten when resuming the CPU.
>
> Because the [ms]envcfg CSR is part of the base RISC-V privileged ISA
> specification, it cannot be detected from the ISA string. However, most
> existing hardware is too old to implement this CSR. As a result, it must
> be probed at runtime.
>
> Extend the logic for the Zicsr ISA extension to probe for the presence
> of specific CSRs. Since the CSR number is encoded as an immediate value
> within the csrr instruction, a switch case is necessary for any CSR that
> must be probed this way. Use the exception table to handle the illegal
> instruction exception raised when the CSR is not implemented.
>
> Cc: stable@kernel.org
> Fixes: 43c16d51a19b ("RISC-V: Enable cbo.zero in usermode")
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
On nommu:
/build/tmp.RcQwFfgWA4/arch/riscv/kernel/cpufeature.c:967:21: error: expected ':' or ')' before '_ASM_EXTABLE'
/build/tmp.RcQwFfgWA4/arch/riscv/kernel/cpufeature.c:967:34: error: invalid suffix "b" on integer constant
/build/tmp.RcQwFfgWA4/arch/riscv/kernel/cpufeature.c:967:38: error: invalid suffix "b" on integer constant
https://patchwork.kernel.org/project/linux-riscv/patch/20240212022642.1968739-2-samuel.holland@sifive.com/
Cheers,
Conor.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend
2024-02-12 2:26 ` [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend Samuel Holland
2024-02-12 7:59 ` Conor Dooley
@ 2024-02-12 10:19 ` Andrew Jones
2024-02-12 13:21 ` Stefan O'Rear
2 siblings, 0 replies; 9+ messages in thread
From: Andrew Jones @ 2024-02-12 10:19 UTC (permalink / raw)
To: Samuel Holland; +Cc: Palmer Dabbelt, linux-riscv, linux-kernel, stable
On Sun, Feb 11, 2024 at 06:26:15PM -0800, Samuel Holland wrote:
> The value of the [ms]envcfg CSR is lost when entering a nonretentive
> idle state, so the CSR must be rewritten when resuming the CPU.
>
> Because the [ms]envcfg CSR is part of the base RISC-V privileged ISA
> specification, it cannot be detected from the ISA string. However, most
> existing hardware is too old to implement this CSR. As a result, it must
> be probed at runtime.
>
> Extend the logic for the Zicsr ISA extension to probe for the presence
> of specific CSRs. Since the CSR number is encoded as an immediate value
> within the csrr instruction, a switch case is necessary for any CSR that
> must be probed this way. Use the exception table to handle the illegal
> instruction exception raised when the CSR is not implemented.
>
> Cc: stable@kernel.org
> Fixes: 43c16d51a19b ("RISC-V: Enable cbo.zero in usermode")
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> ---
>
> arch/riscv/include/asm/csr.h | 23 +++++++++++++++++++++++
> arch/riscv/include/asm/suspend.h | 1 +
> arch/riscv/kernel/cpufeature.c | 23 +++++++++++++++++++++++
> arch/riscv/kernel/suspend.c | 2 ++
> 4 files changed, 49 insertions(+)
>
> diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
> index 2468c55933cd..daff95feb817 100644
> --- a/arch/riscv/include/asm/csr.h
> +++ b/arch/riscv/include/asm/csr.h
> @@ -542,6 +542,29 @@
> : "memory"); \
> })
>
> +#define ALT_CSR_READ(csr) \
> +({ \
> + unsigned long __v; \
> + __asm__ __volatile__ ( \
> + ALTERNATIVE("li %[v], 0", "csrr %[v], %[r]", 0, \
> + csr << 16 | RISCV_ISA_EXT_ZICSR, 1) \
> + : [v] "=r" (__v) \
> + : [r] "i" (csr) \
> + : "memory"); \
> + __v; \
> +})
> +
> +#define ALT_CSR_WRITE(csr, val) \
> +({ \
> + unsigned long __v = (unsigned long)(val); \
> + __asm__ __volatile__ ( \
> + ALTERNATIVE("nop", "csrw %[r], %[v]", 0, \
> + csr << 16 | RISCV_ISA_EXT_ZICSR, 1) \
> + : : [r] "i" (csr), [v] "rK" (__v) \
> + : "memory"); \
> + __v; \
> +})
> +
> #endif /* __ASSEMBLY__ */
>
> #endif /* _ASM_RISCV_CSR_H */
> diff --git a/arch/riscv/include/asm/suspend.h b/arch/riscv/include/asm/suspend.h
> index 02f87867389a..491296a335d0 100644
> --- a/arch/riscv/include/asm/suspend.h
> +++ b/arch/riscv/include/asm/suspend.h
> @@ -14,6 +14,7 @@ struct suspend_context {
> struct pt_regs regs;
> /* Saved and restored by high-level functions */
> unsigned long scratch;
> + unsigned long envcfg;
> unsigned long tvec;
> unsigned long ie;
> #ifdef CONFIG_MMU
> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> index c5b13f7dd482..934090270ae5 100644
> --- a/arch/riscv/kernel/cpufeature.c
> +++ b/arch/riscv/kernel/cpufeature.c
> @@ -954,6 +954,27 @@ void riscv_user_isa_enable(void)
> }
>
> #ifdef CONFIG_RISCV_ALTERNATIVE
> +static bool riscv_cpufeature_probe_csr(u16 csr)
> +{
> + bool ret = false;
> +
> + switch (csr) {
> +#define PROBE_CSR_CASE(_csr) \
> + case _csr: \
> + asm("1: csrr zero, %[csr]\n" \
> + " li %[r], 1\n" \
> + "2:\n" \
> + _ASM_EXTABLE(1b, 2b) \
> + : [r] "+r" (ret) \
> + : [csr] "i" (_csr)); \
> + break
> + PROBE_CSR_CASE(CSR_ENVCFG);
> +#undef PROBE_CSR_CASE
> + }
> +
> + return ret;
> +}
> +
> /*
> * Alternative patch sites consider 48 bits when determining when to patch
> * the old instruction sequence with the new. These bits are broken into a
> @@ -974,6 +995,8 @@ static bool riscv_cpufeature_patch_check(u16 id, u16 value)
> return true;
>
> switch (id) {
> + case RISCV_ISA_EXT_ZICSR:
> + return riscv_cpufeature_probe_csr(value);
> case RISCV_ISA_EXT_ZICBOZ:
> /*
> * Zicboz alternative applications provide the maximum
> diff --git a/arch/riscv/kernel/suspend.c b/arch/riscv/kernel/suspend.c
> index 239509367e42..fe544f12a5c5 100644
> --- a/arch/riscv/kernel/suspend.c
> +++ b/arch/riscv/kernel/suspend.c
> @@ -15,6 +15,7 @@
> void suspend_save_csrs(struct suspend_context *context)
> {
> context->scratch = csr_read(CSR_SCRATCH);
> + context->envcfg = ALT_CSR_READ(CSR_ENVCFG);
> context->tvec = csr_read(CSR_TVEC);
> context->ie = csr_read(CSR_IE);
>
> @@ -36,6 +37,7 @@ void suspend_save_csrs(struct suspend_context *context)
> void suspend_restore_csrs(struct suspend_context *context)
> {
> csr_write(CSR_SCRATCH, context->scratch);
> + ALT_CSR_WRITE(CSR_ENVCFG, context->envcfg);
> csr_write(CSR_TVEC, context->tvec);
> csr_write(CSR_IE, context->ie);
>
> --
> 2.43.0
>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend
2024-02-12 2:26 ` [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend Samuel Holland
2024-02-12 7:59 ` Conor Dooley
2024-02-12 10:19 ` Andrew Jones
@ 2024-02-12 13:21 ` Stefan O'Rear
2024-02-13 3:25 ` Samuel Holland
2 siblings, 1 reply; 9+ messages in thread
From: Stefan O'Rear @ 2024-02-12 13:21 UTC (permalink / raw)
To: Samuel Holland, Andrew Jones, Palmer Dabbelt
Cc: linux-riscv, linux-kernel, stable
On Sun, Feb 11, 2024, at 9:26 PM, Samuel Holland wrote:
> The value of the [ms]envcfg CSR is lost when entering a nonretentive
> idle state, so the CSR must be rewritten when resuming the CPU.
>
> Because the [ms]envcfg CSR is part of the base RISC-V privileged ISA
> specification, it cannot be detected from the ISA string. However, most
> existing hardware is too old to implement this CSR. As a result, it must
> be probed at runtime.
>
> Extend the logic for the Zicsr ISA extension to probe for the presence
> of specific CSRs. Since the CSR number is encoded as an immediate value
> within the csrr instruction, a switch case is necessary for any CSR that
> must be probed this way. Use the exception table to handle the illegal
> instruction exception raised when the CSR is not implemented.
We support non-conforming extensions, so we can't assume that if an
implementation does not provide the Ss1p12 extension which defines senvcfg,
the corresponding CSR number will not be used for other purposes.
-s
> Cc: stable@kernel.org
> Fixes: 43c16d51a19b ("RISC-V: Enable cbo.zero in usermode")
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> ---
>
> arch/riscv/include/asm/csr.h | 23 +++++++++++++++++++++++
> arch/riscv/include/asm/suspend.h | 1 +
> arch/riscv/kernel/cpufeature.c | 23 +++++++++++++++++++++++
> arch/riscv/kernel/suspend.c | 2 ++
> 4 files changed, 49 insertions(+)
>
> diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
> index 2468c55933cd..daff95feb817 100644
> --- a/arch/riscv/include/asm/csr.h
> +++ b/arch/riscv/include/asm/csr.h
> @@ -542,6 +542,29 @@
> : "memory"); \
> })
>
> +#define ALT_CSR_READ(csr) \
> +({ \
> + unsigned long __v; \
> + __asm__ __volatile__ ( \
> + ALTERNATIVE("li %[v], 0", "csrr %[v], %[r]", 0, \
> + csr << 16 | RISCV_ISA_EXT_ZICSR, 1) \
> + : [v] "=r" (__v) \
> + : [r] "i" (csr) \
> + : "memory"); \
> + __v; \
> +})
> +
> +#define ALT_CSR_WRITE(csr, val) \
> +({ \
> + unsigned long __v = (unsigned long)(val); \
> + __asm__ __volatile__ ( \
> + ALTERNATIVE("nop", "csrw %[r], %[v]", 0, \
> + csr << 16 | RISCV_ISA_EXT_ZICSR, 1) \
> + : : [r] "i" (csr), [v] "rK" (__v) \
> + : "memory"); \
> + __v; \
> +})
> +
> #endif /* __ASSEMBLY__ */
>
> #endif /* _ASM_RISCV_CSR_H */
> diff --git a/arch/riscv/include/asm/suspend.h b/arch/riscv/include/asm/suspend.h
> index 02f87867389a..491296a335d0 100644
> --- a/arch/riscv/include/asm/suspend.h
> +++ b/arch/riscv/include/asm/suspend.h
> @@ -14,6 +14,7 @@ struct suspend_context {
> struct pt_regs regs;
> /* Saved and restored by high-level functions */
> unsigned long scratch;
> + unsigned long envcfg;
> unsigned long tvec;
> unsigned long ie;
> #ifdef CONFIG_MMU
> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> index c5b13f7dd482..934090270ae5 100644
> --- a/arch/riscv/kernel/cpufeature.c
> +++ b/arch/riscv/kernel/cpufeature.c
> @@ -954,6 +954,27 @@ void riscv_user_isa_enable(void)
> }
>
> #ifdef CONFIG_RISCV_ALTERNATIVE
> +static bool riscv_cpufeature_probe_csr(u16 csr)
> +{
> + bool ret = false;
> +
> + switch (csr) {
> +#define PROBE_CSR_CASE(_csr) \
> + case _csr: \
> + asm("1: csrr zero, %[csr]\n" \
> + " li %[r], 1\n" \
> + "2:\n" \
> + _ASM_EXTABLE(1b, 2b) \
> + : [r] "+r" (ret) \
> + : [csr] "i" (_csr)); \
> + break
> + PROBE_CSR_CASE(CSR_ENVCFG);
> +#undef PROBE_CSR_CASE
> + }
> +
> + return ret;
> +}
> +
> /*
> * Alternative patch sites consider 48 bits when determining when to patch
> * the old instruction sequence with the new. These bits are broken into a
> @@ -974,6 +995,8 @@ static bool riscv_cpufeature_patch_check(u16 id, u16 value)
> return true;
>
> switch (id) {
> + case RISCV_ISA_EXT_ZICSR:
> + return riscv_cpufeature_probe_csr(value);
> case RISCV_ISA_EXT_ZICBOZ:
> /*
> * Zicboz alternative applications provide the maximum
> diff --git a/arch/riscv/kernel/suspend.c b/arch/riscv/kernel/suspend.c
> index 239509367e42..fe544f12a5c5 100644
> --- a/arch/riscv/kernel/suspend.c
> +++ b/arch/riscv/kernel/suspend.c
> @@ -15,6 +15,7 @@
> void suspend_save_csrs(struct suspend_context *context)
> {
> context->scratch = csr_read(CSR_SCRATCH);
> + context->envcfg = ALT_CSR_READ(CSR_ENVCFG);
> context->tvec = csr_read(CSR_TVEC);
> context->ie = csr_read(CSR_IE);
>
> @@ -36,6 +37,7 @@ void suspend_save_csrs(struct suspend_context *context)
> void suspend_restore_csrs(struct suspend_context *context)
> {
> csr_write(CSR_SCRATCH, context->scratch);
> + ALT_CSR_WRITE(CSR_ENVCFG, context->envcfg);
> csr_write(CSR_TVEC, context->tvec);
> csr_write(CSR_IE, context->ie);
>
> --
> 2.43.0
>
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH -fixes 2/2] riscv: Save/restore envcfg CSR during CPU suspend
2024-02-12 13:21 ` Stefan O'Rear
@ 2024-02-13 3:25 ` Samuel Holland
0 siblings, 0 replies; 9+ messages in thread
From: Samuel Holland @ 2024-02-13 3:25 UTC (permalink / raw)
To: Stefan O'Rear, Andrew Jones, Palmer Dabbelt
Cc: linux-riscv, linux-kernel, stable
Hi Stefan,
On 2024-02-12 7:21 AM, Stefan O'Rear wrote:
> On Sun, Feb 11, 2024, at 9:26 PM, Samuel Holland wrote:
>> The value of the [ms]envcfg CSR is lost when entering a nonretentive
>> idle state, so the CSR must be rewritten when resuming the CPU.
>>
>> Because the [ms]envcfg CSR is part of the base RISC-V privileged ISA
>> specification, it cannot be detected from the ISA string. However, most
>> existing hardware is too old to implement this CSR. As a result, it must
>> be probed at runtime.
>>
>> Extend the logic for the Zicsr ISA extension to probe for the presence
>> of specific CSRs. Since the CSR number is encoded as an immediate value
>> within the csrr instruction, a switch case is necessary for any CSR that
>> must be probed this way. Use the exception table to handle the illegal
>> instruction exception raised when the CSR is not implemented.
>
> We support non-conforming extensions, so we can't assume that if an
> implementation does not provide the Ss1p12 extension which defines senvcfg,
> the corresponding CSR number will not be used for other purposes.
Yes, you're right, I'll do this for v2. Though this does prevent us from
supporting hardware which implements senvcfg but not all of the rest of Ss1p12
(e.g. Ss1p11 + Zicboz).
Regards,
Samuel
^ permalink raw reply [flat|nested] 9+ messages in thread