2023-01-11 17:22:52

by Jisheng Zhang

[permalink] [raw]
Subject: [PATCH v3 01/13] riscv: fix jal offsets in patched alternatives

Alternatives live in a different section, so offsets used by jal
instruction will point to wrong locations after the patch got applied.

Similar to arm64, adjust the location to consider that offset.

Co-developed-by: Heiko Stuebner <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Signed-off-by: Jisheng Zhang <[email protected]>
---
arch/riscv/include/asm/insn.h | 27 +++++++++++++++++++++++++++
arch/riscv/kernel/alternative.c | 27 +++++++++++++++++++++++++++
2 files changed, 54 insertions(+)

diff --git a/arch/riscv/include/asm/insn.h b/arch/riscv/include/asm/insn.h
index 98453535324a..1d2df245d0bd 100644
--- a/arch/riscv/include/asm/insn.h
+++ b/arch/riscv/include/asm/insn.h
@@ -291,6 +291,33 @@ static __always_inline bool riscv_insn_is_branch(u32 code)
(RVC_X(x_, RVC_B_IMM_7_6_OPOFF, RVC_B_IMM_7_6_MASK) << RVC_B_IMM_7_6_OFF) | \
(RVC_IMM_SIGN(x_) << RVC_B_IMM_SIGN_OFF); })

+/*
+ * Get the immediate from a J-type instruction.
+ *
+ * @insn: instruction to process
+ * Return: immediate
+ */
+static inline s32 riscv_insn_extract_jtype_imm(u32 insn)
+{
+ return RV_EXTRACT_JTYPE_IMM(insn);
+}
+
+/*
+ * Update a J-type instruction with an immediate value.
+ *
+ * @insn: pointer to the jtype instruction
+ * @imm: the immediate to insert into the instruction
+ */
+static inline void riscv_insn_insert_jtype_imm(u32 *insn, s32 imm)
+{
+ /* drop the old IMMs, all jal IMM bits sit at 31:12 */
+ *insn &= ~GENMASK(31, 12);
+ *insn |= (RV_X(imm, RV_J_IMM_10_1_OFF, RV_J_IMM_10_1_MASK) << RV_J_IMM_10_1_OPOFF) |
+ (RV_X(imm, RV_J_IMM_11_OFF, RV_J_IMM_11_MASK) << RV_J_IMM_11_OPOFF) |
+ (RV_X(imm, RV_J_IMM_19_12_OFF, RV_J_IMM_19_12_MASK) << RV_J_IMM_19_12_OPOFF) |
+ (RV_X(imm, RV_J_IMM_SIGN_OFF, 1) << RV_J_IMM_SIGN_OPOFF);
+}
+
/*
* Put together one immediate from a U-type and I-type instruction pair.
*
diff --git a/arch/riscv/kernel/alternative.c b/arch/riscv/kernel/alternative.c
index 6212ea0eed72..3d4f1f32c7f6 100644
--- a/arch/riscv/kernel/alternative.c
+++ b/arch/riscv/kernel/alternative.c
@@ -79,6 +79,21 @@ static void riscv_alternative_fix_auipc_jalr(void *ptr, u32 auipc_insn,
patch_text_nosync(ptr, call, sizeof(u32) * 2);
}

+static void riscv_alternative_fix_jal(void *ptr, u32 jal_insn, int patch_offset)
+{
+ s32 imm;
+
+ /* get and adjust new target address */
+ imm = riscv_insn_extract_jtype_imm(jal_insn);
+ imm -= patch_offset;
+
+ /* update instruction */
+ riscv_insn_insert_jtype_imm(&jal_insn, imm);
+
+ /* patch the call place again */
+ patch_text_nosync(ptr, &jal_insn, sizeof(u32));
+}
+
void riscv_alternative_fix_offsets(void *alt_ptr, unsigned int len,
int patch_offset)
{
@@ -106,6 +121,18 @@ void riscv_alternative_fix_offsets(void *alt_ptr, unsigned int len,
riscv_alternative_fix_auipc_jalr(alt_ptr + i * sizeof(u32),
insn, insn2, patch_offset);
}
+
+ if (riscv_insn_is_jal(insn)) {
+ s32 imm = riscv_insn_extract_jtype_imm(insn);
+
+ /* Don't modify jumps inside the alternative block */
+ if ((alt_ptr + i * sizeof(u32) + imm) >= alt_ptr &&
+ (alt_ptr + i * sizeof(u32) + imm) < (alt_ptr + len))
+ continue;
+
+ riscv_alternative_fix_jal(alt_ptr + i * sizeof(u32),
+ insn, patch_offset);
+ }
}
}

--
2.38.1


2023-01-11 19:15:55

by Andrew Jones

[permalink] [raw]
Subject: Re: [PATCH v3 01/13] riscv: fix jal offsets in patched alternatives

On Thu, Jan 12, 2023 at 01:10:15AM +0800, Jisheng Zhang wrote:
> Alternatives live in a different section, so offsets used by jal
> instruction will point to wrong locations after the patch got applied.
>
> Similar to arm64, adjust the location to consider that offset.
>
> Co-developed-by: Heiko Stuebner <[email protected]>
> Signed-off-by: Heiko Stuebner <[email protected]>
> Signed-off-by: Jisheng Zhang <[email protected]>
> ---
> arch/riscv/include/asm/insn.h | 27 +++++++++++++++++++++++++++
> arch/riscv/kernel/alternative.c | 27 +++++++++++++++++++++++++++
> 2 files changed, 54 insertions(+)
>
> diff --git a/arch/riscv/include/asm/insn.h b/arch/riscv/include/asm/insn.h
> index 98453535324a..1d2df245d0bd 100644
> --- a/arch/riscv/include/asm/insn.h
> +++ b/arch/riscv/include/asm/insn.h
> @@ -291,6 +291,33 @@ static __always_inline bool riscv_insn_is_branch(u32 code)
> (RVC_X(x_, RVC_B_IMM_7_6_OPOFF, RVC_B_IMM_7_6_MASK) << RVC_B_IMM_7_6_OFF) | \
> (RVC_IMM_SIGN(x_) << RVC_B_IMM_SIGN_OFF); })
>
> +/*
> + * Get the immediate from a J-type instruction.
> + *
> + * @insn: instruction to process
> + * Return: immediate
> + */
> +static inline s32 riscv_insn_extract_jtype_imm(u32 insn)
> +{
> + return RV_EXTRACT_JTYPE_IMM(insn);
> +}
> +
> +/*
> + * Update a J-type instruction with an immediate value.
> + *
> + * @insn: pointer to the jtype instruction
> + * @imm: the immediate to insert into the instruction
> + */
> +static inline void riscv_insn_insert_jtype_imm(u32 *insn, s32 imm)
> +{
> + /* drop the old IMMs, all jal IMM bits sit at 31:12 */
> + *insn &= ~GENMASK(31, 12);
> + *insn |= (RV_X(imm, RV_J_IMM_10_1_OFF, RV_J_IMM_10_1_MASK) << RV_J_IMM_10_1_OPOFF) |
> + (RV_X(imm, RV_J_IMM_11_OFF, RV_J_IMM_11_MASK) << RV_J_IMM_11_OPOFF) |
> + (RV_X(imm, RV_J_IMM_19_12_OFF, RV_J_IMM_19_12_MASK) << RV_J_IMM_19_12_OPOFF) |
> + (RV_X(imm, RV_J_IMM_SIGN_OFF, 1) << RV_J_IMM_SIGN_OPOFF);
> +}
> +
> /*
> * Put together one immediate from a U-type and I-type instruction pair.
> *
> diff --git a/arch/riscv/kernel/alternative.c b/arch/riscv/kernel/alternative.c
> index 6212ea0eed72..3d4f1f32c7f6 100644
> --- a/arch/riscv/kernel/alternative.c
> +++ b/arch/riscv/kernel/alternative.c
> @@ -79,6 +79,21 @@ static void riscv_alternative_fix_auipc_jalr(void *ptr, u32 auipc_insn,
> patch_text_nosync(ptr, call, sizeof(u32) * 2);
> }
>
> +static void riscv_alternative_fix_jal(void *ptr, u32 jal_insn, int patch_offset)
> +{
> + s32 imm;
> +
> + /* get and adjust new target address */
> + imm = riscv_insn_extract_jtype_imm(jal_insn);
> + imm -= patch_offset;
> +
> + /* update instruction */
> + riscv_insn_insert_jtype_imm(&jal_insn, imm);
> +
> + /* patch the call place again */
> + patch_text_nosync(ptr, &jal_insn, sizeof(u32));
> +}
> +
> void riscv_alternative_fix_offsets(void *alt_ptr, unsigned int len,
> int patch_offset)
> {
> @@ -106,6 +121,18 @@ void riscv_alternative_fix_offsets(void *alt_ptr, unsigned int len,
> riscv_alternative_fix_auipc_jalr(alt_ptr + i * sizeof(u32),
> insn, insn2, patch_offset);
> }
> +
> + if (riscv_insn_is_jal(insn)) {
> + s32 imm = riscv_insn_extract_jtype_imm(insn);
> +
> + /* Don't modify jumps inside the alternative block */
> + if ((alt_ptr + i * sizeof(u32) + imm) >= alt_ptr &&
> + (alt_ptr + i * sizeof(u32) + imm) < (alt_ptr + len))
> + continue;
> +
> + riscv_alternative_fix_jal(alt_ptr + i * sizeof(u32),
> + insn, patch_offset);
> + }
> }
> }
>
> --
> 2.38.1
>

Reviewed-by: Andrew Jones <[email protected]>

2023-01-12 00:05:25

by Heiko Stuebner

[permalink] [raw]
Subject: Re: [PATCH v3 01/13] riscv: fix jal offsets in patched alternatives

Am Mittwoch, 11. Januar 2023, 18:10:15 CET schrieb Jisheng Zhang:
> Alternatives live in a different section, so offsets used by jal
> instruction will point to wrong locations after the patch got applied.
>
> Similar to arm64, adjust the location to consider that offset.
>
> Co-developed-by: Heiko Stuebner <[email protected]>
> Signed-off-by: Heiko Stuebner <[email protected]>
> Signed-off-by: Jisheng Zhang <[email protected]>

looks good, thanks for fixing the issues Andrew and Conor pointed
out in the variant in my zbb series. I've now switched over to this one.

I guess as you said, we really should separate this out into a single patch
[or if Palmer is fine with that, just pick this one patch to apply until the
rest is ready]


Heiko


2023-01-12 21:31:14

by Conor Dooley

[permalink] [raw]
Subject: Re: [PATCH v3 01/13] riscv: fix jal offsets in patched alternatives

On Thu, Jan 12, 2023 at 12:31:59AM +0100, Heiko St?bner wrote:
> Am Mittwoch, 11. Januar 2023, 18:10:15 CET schrieb Jisheng Zhang:
> > Alternatives live in a different section, so offsets used by jal
> > instruction will point to wrong locations after the patch got applied.
> >
> > Similar to arm64, adjust the location to consider that offset.
> >
> > Co-developed-by: Heiko Stuebner <[email protected]>
> > Signed-off-by: Heiko Stuebner <[email protected]>
> > Signed-off-by: Jisheng Zhang <[email protected]>
>
> looks good, thanks for fixing the issues Andrew and Conor pointed
> out in the variant in my zbb series. I've now switched over to this one.
>
> I guess as you said, we really should separate this out into a single patch
> [or if Palmer is fine with that, just pick this one patch to apply until the
> rest is ready]

Splitting it out may make it easier to flag for him during the pw sync
next week? Either way, I'm fine w/ it..

Reviewed-by: Conor Dooley <[email protected]>


Attachments:
(No filename) (1.02 kB)
signature.asc (235.00 B)
Download all attachments