lkml.org 
[lkml]   [2022]   [Dec]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [PATCH v2 01/13] riscv: fix jal offsets in patched alternatives
    On Mon, Dec 05, 2022 at 04:31:08PM +0100, Heiko Stübner wrote:
    > Am Sonntag, 4. Dezember 2022, 18:46:20 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.
    > >
    > > Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
    > > ---
    > > arch/riscv/include/asm/alternative.h | 2 ++
    > > arch/riscv/kernel/alternative.c | 38 ++++++++++++++++++++++++++++
    > > arch/riscv/kernel/cpufeature.c | 3 +++
    > > 3 files changed, 43 insertions(+)
    > >
    > > diff --git a/arch/riscv/include/asm/alternative.h b/arch/riscv/include/asm/alternative.h
    > > index c58ec3cc4bc3..33eae9541684 100644
    > > --- a/arch/riscv/include/asm/alternative.h
    > > +++ b/arch/riscv/include/asm/alternative.h
    > > @@ -29,6 +29,8 @@ void apply_module_alternatives(void *start, size_t length);
    > >
    > > void riscv_alternative_fix_auipc_jalr(void *alt_ptr, unsigned int len,
    > > int patch_offset);
    > > +void riscv_alternative_fix_jal(void *alt_ptr, unsigned int len,
    > > + int patch_offset);
    > >
    > > struct alt_entry {
    > > void *old_ptr; /* address of original instruciton or data */
    > > diff --git a/arch/riscv/kernel/alternative.c b/arch/riscv/kernel/alternative.c
    > > index 292cc42dc3be..9d88375624b5 100644
    > > --- a/arch/riscv/kernel/alternative.c
    > > +++ b/arch/riscv/kernel/alternative.c
    > > @@ -125,6 +125,44 @@ void riscv_alternative_fix_auipc_jalr(void *alt_ptr, unsigned int len,
    > > }
    > > }
    > >
    > > +#define to_jal_imm(value) \
    > > + (((value & (RV_J_IMM_10_1_MASK << RV_J_IMM_10_1_OFF)) << RV_I_IMM_11_0_OPOFF) | \
    > > + ((value & (RV_J_IMM_11_MASK << RV_J_IMM_11_OFF)) << RV_J_IMM_11_OPOFF) | \
    > > + ((value & (RV_J_IMM_19_12_OPOFF << RV_J_IMM_19_12_OFF)) << RV_J_IMM_19_12_OPOFF) | \
    > > + ((value & (1 << RV_J_IMM_SIGN_OFF)) << RV_J_IMM_SIGN_OPOFF))
    > > +
    > > +void riscv_alternative_fix_jal(void *alt_ptr, unsigned int len,
    > > + int patch_offset)
    > > +{
    >
    > I think we might want to unfiy this into a common function like
    >
    > riscv_alternative_fix_offsets(...)
    >
    > so that we only run through the code block once
    >
    > for (i = 0; i < num_instr; i++) {
    > if (riscv_insn_is_auipc_jalr(inst1, inst2)) {
    > riscv_alternative_fix_auipc_jalr(...)
    > continue;
    > }
    >
    > if (riscv_insn_is_jal(inst)) {
    > riscv_alternative_fix_jal(...)
    > continue;
    > }
    > }
    >
    > This would also remove the need from calling multiple functions
    > after patching alternatives.

    Yesterday, I also wanted to unify the two instruction fix into
    one. But that would need to roll back the
    riscv_alternative_fix_auipc_jalr() to your v1 version. And IMHO,
    it's better if you can split the Zbb string optimizations series
    into two: one for alternative improvements, another for Zbb. Then
    we may get the alternative improvements and this inst extension
    series merged in v6.2-rc1.

    >
    > Thoughts?
    >
    >
    > Heiko
    >
    > > + int num_instr = len / sizeof(u32);
    > > + unsigned int call;
    > > + int i;
    > > + int imm;
    > > +
    > > + for (i = 0; i < num_instr; i++) {
    > > + u32 inst = riscv_instruction_at(alt_ptr, i);
    > > +
    > > + if (!riscv_insn_is_jal(inst))
    > > + continue;
    > > +
    > > + /* get and adjust new target address */
    > > + imm = RV_EXTRACT_JTYPE_IMM(inst);
    > > + imm -= patch_offset;
    > > +
    > > + /* pick the original jal */
    > > + call = inst;
    > > +
    > > + /* drop the old IMMs, all jal imm bits sit at 31:12 */
    > > + call &= ~GENMASK(31, 12);
    > > +
    > > + /* add the adapted IMMs */
    > > + call |= to_jal_imm(imm);
    > > +
    > > + /* patch the call place again */
    > > + patch_text_nosync(alt_ptr + i * sizeof(u32), &call, 4);
    > > + }
    > > +}
    > > +
    > > /*
    > > * This is called very early in the boot process (directly after we run
    > > * a feature detect on the boot CPU). No need to worry about other CPUs
    > > diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
    > > index ba62a4ff5ccd..c743f0adc794 100644
    > > --- a/arch/riscv/kernel/cpufeature.c
    > > +++ b/arch/riscv/kernel/cpufeature.c
    > > @@ -324,6 +324,9 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
    > > riscv_alternative_fix_auipc_jalr(alt->old_ptr,
    > > alt->alt_len,
    > > alt->old_ptr - alt->alt_ptr);
    > > + riscv_alternative_fix_jal(alt->old_ptr,
    > > + alt->alt_len,
    > > + alt->old_ptr - alt->alt_ptr);
    > > }
    > > }
    > > }
    > >
    >
    >
    >
    >

    \
     
     \ /
      Last update: 2022-12-05 16:51    [W:4.090 / U:0.176 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site