Skip to contentSkip to conversation

Add proper alignment check and pending 'C' extension for riscv

Qemu-devel In review [email protected] 3 participants 10 messages Feb 22, 2019 – Feb 26, 2019 RISC-V

lore
Share & tools
lore
#0 lore raw

Dear All,

I'm submiting this patch to properly check the next instruction alignment and scheduale compression extenstion enable upon 'MISA' register writes to later aligned instruction through exporting next instruction 'pc' to riscv cpu state

Thanks,
Ahmed

#1 lore raw

PATCH 1/1 riscv: Add proper alignment check and pending 'C' extension upon misa writes

From: ahmed_magdy <[email protected]>

---
 target/riscv/cpu.h       |  2 ++
 target/riscv/csr.c       |  5 +++--
 target/riscv/translate.c | 14 ++++++++++----
 3 files changed, 15 insertions(+), 6 deletions(-)
target/riscv/cpu.h +2 -0
index a269c07..b49bdb3 100644
@@ -109,6 +109,8 @@ struct CPURISCVState {
     target_ulong gpr[32];
     uint64_t fpr[32]; /* assume both F and D extensions */
     target_ulong pc;
+    target_ulong pc_next; /* next target pc */
+    int pending_rvc;
     target_ulong load_res;
     target_ulong load_val;
 
target/riscv/csr.c +3 -2
index 90f6866..b0886f0 100644
@@ -373,9 +373,10 @@ static int write_misa(CPURISCVState *env, int csrno, target_ulong val)
     }
 
     /* Suppress 'C' if next instruction is not aligned
-       TODO: this should check next_pc */
-    if ((val & RVC) && (GETPC() & ~3) != 0) {
+       check next target pc */
+    if ((val & RVC) && (env->pc_next & 3) != 0) {
         val &= ~RVC;
+        env->pending_rvc = 1;
     }
 
     /* misa.MXL writes are not supported by QEMU */
target/riscv/translate.c +10 -4
index 2321bba..c9d84ea 100644
@@ -1999,20 +1999,26 @@ static void decode_RV32_64G(DisasContext *ctx)
     }
 }
 
-static void decode_opc(DisasContext *ctx)
+static void decode_opc(DisasContext *ctx, CPUState *cpu)
 {
+    CPURISCVState *env = cpu->env_ptr;
     /* check for compressed insn */
     if (extract32(ctx->opcode, 0, 2) != 3) {
         if (!has_ext(ctx, RVC)) {
             gen_exception_illegal(ctx);
         } else {
-            ctx->pc_succ_insn = ctx->base.pc_next + 2;
+            env->pc_next = ctx->pc_succ_insn = ctx->base.pc_next + 2;
             decode_RV32_64C(ctx);
         }
     } else {
-        ctx->pc_succ_insn = ctx->base.pc_next + 4;
+        env->pc_next = ctx->pc_succ_insn = ctx->base.pc_next + 4;
         decode_RV32_64G(ctx);
     }
+    /* check pending RVC */
+    if (env->pending_rvc && ((env->pc_next & 3) != 0)) {
+        env->misa |= RVC;
+        env->pending_rvc = 0;
+    }
 }
 
 static void riscv_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cpu)
@@ -2061,7 +2067,7 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
     CPURISCVState *env = cpu->env_ptr;
 
     ctx->opcode = cpu_ldl_code(env, ctx->base.pc_next);
-    decode_opc(ctx);
+    decode_opc(ctx, cpu);
     ctx->base.pc_next = ctx->pc_succ_insn;
 
     if (ctx->base.is_jmp == DISAS_NEXT) {
Signature · 1 line
1.9.1
#2 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Quoted 33 lines · On 2/22/19 8:25 AM, [email protected] wrote:
target/riscv/csr.c +3 -2
@@ -373,9 +373,10 @@ static int write_misa(CPURISCVState *env, int csrno, target_ulong val)
     }
 
     /* Suppress 'C' if next instruction is not aligned
-       TODO: this should check next_pc */
-    if ((val & RVC) && (GETPC() & ~3) != 0) {
+       check next target pc */
+    if ((val & RVC) && (env->pc_next & 3) != 0) {
         val &= ~RVC;
+        env->pending_rvc = 1;
     }
 
     /* misa.MXL writes are not supported by QEMU */
target/riscv/translate.c +3 -2
index 2321bba..c9d84ea 100644
@@ -1999,20 +1999,26 @@ static void decode_RV32_64G(DisasContext *ctx)
     }
 }
 
-static void decode_opc(DisasContext *ctx)
+static void decode_opc(DisasContext *ctx, CPUState *cpu)
 {
+    CPURISCVState *env = cpu->env_ptr;
     /* check for compressed insn */
     if (extract32(ctx->opcode, 0, 2) != 3) {
         if (!has_ext(ctx, RVC)) {
             gen_exception_illegal(ctx);
         } else {
-            ctx->pc_succ_insn = ctx->base.pc_next + 2;
+            env->pc_next = ctx->pc_succ_insn = ctx->base.pc_next + 2;
             decode_RV32_64C(ctx);

target/riscv/translate.c static void decode_RV32_64G(DisasContext *ctx) · line 2020

Quoted 10 lines
target/riscv/translate.c +5 -1
         }
     } else {
-        ctx->pc_succ_insn = ctx->base.pc_next + 4;
+        env->pc_next = ctx->pc_succ_insn = ctx->base.pc_next + 4;
         decode_RV32_64G(ctx);
     }
+    /* check pending RVC */
+    if (env->pending_rvc && ((env->pc_next & 3) != 0)) {
+        env->misa |= RVC;
+        env->pending_rvc = 0;

You cannot manipulate env like this during translation.

Neither the write to env->pc_next nor the read from env->pending_rvc here will be in any synchronization with the execution of write_misa.

What semantics are you attempting to implement wrt setting/clearing RVC from MISA?

target/riscv/translate.c static void riscv_tr_translate_insn · line 2070

Quoted 6 lines
target/riscv/translate.c +1 -1
@@ -2061,7 +2067,7 @@ static void riscv_tr_translate_insn
     CPURISCVState *env = cpu->env_ptr;
 
     ctx->opcode = cpu_ldl_code(env, ctx->base.pc_next);
-    decode_opc(ctx);
+    decode_opc(ctx, cpu);

This is exactly the reason why cpu is *not* passed down to decode_opc, so that you cannot make this kind of mistake.

r~

author
#3 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Thank you for your review and feedback, Richard.
As Eric mentioned, this is the first time contribution. I have been exploring Qemu for some time and try to understand main flow, internals, ..etc.

Quoted 3 lines

You cannot manipulate env like this during translation.

Neither the write to env->pc_next nor the read from env->pending_rvc

here will

Quoted 1 line

be in any synchronization with the execution of write_misa.

Does this applies for translated code in a single translation block only or for different TBs also ?

So should I manipulate "env" from translation context through helpers only ? for example:

TCGv temp;
tcg_gen_movi_tl(temp, ctx->pc_succ_insn);
gen_helper_next_pc(cpu_env, temp);

while the helper function definition like that:
void helper_next_pc(CPURISCVState *env, target_ulong pc_next)

{
    env->pc_next = pc_next;
}

and the same to read "env->pending_rvc"

I'm thinking of a way to add 'C' extension at run time through waiting the correct aligned instruction, which I believe it might be after branch something quite similar to switching between ARM and THUMB states in ARM, for misa 'RVC' enable to take effect since it will be no possibility to check alignment with 'C' extension.

Thanks,
Ahmed

Quoted 69 lines · Richard Henderson

Review static int write_misa(CPURISCVState *env, int

Quoted 1 line · On 2/22/19 8:25 AM, [email protected] wrote:
(quoted diff) +0 -0
@@ -373,9 +373,10 @@ static int write_misa(CPURISCVState *env, int

csrno, target_ulong val)

Quoted 32 lines
target/riscv/csr.c +3 -2
     }
 
     /* Suppress 'C' if next instruction is not aligned
-       TODO: this should check next_pc */
-    if ((val & RVC) && (GETPC() & ~3) != 0) {
+       check next target pc */
+    if ((val & RVC) && (env->pc_next & 3) != 0) {
         val &= ~RVC;
+        env->pending_rvc = 1;
     }
 
     /* misa.MXL writes are not supported by QEMU */
target/riscv/translate.c +3 -2
index 2321bba..c9d84ea 100644
@@ -1999,20 +1999,26 @@ static void decode_RV32_64G(DisasContext *ctx)
     }
 }
 
-static void decode_opc(DisasContext *ctx)
+static void decode_opc(DisasContext *ctx, CPUState *cpu)
 {
+    CPURISCVState *env = cpu->env_ptr;
     /* check for compressed insn */
     if (extract32(ctx->opcode, 0, 2) != 3) {
         if (!has_ext(ctx, RVC)) {
             gen_exception_illegal(ctx);
         } else {
-            ctx->pc_succ_insn = ctx->base.pc_next + 2;
+            env->pc_next = ctx->pc_succ_insn = ctx->base.pc_next + 2;
             decode_RV32_64C(ctx);

target/riscv/translate.c static void decode_RV32_64G(DisasContext *ctx) · line 2020

Quoted 10 lines
target/riscv/translate.c +5 -1
         }
     } else {
-        ctx->pc_succ_insn = ctx->base.pc_next + 4;
+        env->pc_next = ctx->pc_succ_insn = ctx->base.pc_next + 4;
         decode_RV32_64G(ctx);
     }
+    /* check pending RVC */
+    if (env->pending_rvc && ((env->pc_next & 3) != 0)) {
+        env->misa |= RVC;
+        env->pending_rvc = 0;

You cannot manipulate env like this during translation.

Neither the write to env->pc_next nor the read from env->pending_rvc here will
be in any synchronization with the execution of write_misa.

What semantics are you attempting to implement wrt setting/clearing RVC from MISA?

target/riscv/translate.c static void riscv_tr_translate_insn · line 2070

Quoted 6 lines
target/riscv/translate.c +1 -1
@@ -2061,7 +2067,7 @@ static void riscv_tr_translate_insn
     CPURISCVState *env = cpu->env_ptr;
 
     ctx->opcode = cpu_ldl_code(env, ctx->base.pc_next);
-    decode_opc(ctx);
+    decode_opc(ctx, cpu);

This is exactly the reason why cpu is *not* passed down to decode_opc, so that
you cannot make this kind of mistake.

r~

#4 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Quoted 11 lines · Amed Magdy

Thank you for your review and feedback, Richard.
As Eric mentioned, this is the first time contribution. I have been exploring Qemu for some time and try to understand main flow, internals, ..etc.

Quoted 4 lines

  You cannot manipulate env like this during translation.

  Neither the write to env->pc_next nor the read from env->pending_rvc here will be in any synchronization with the execution of write_misa.


Does this applies for translated code in a single translation block only or for different TBs also ?

All of the time this will not work.

Quoted 7 lines

So should I manipulate "env" from translation context through helpers only ? for example:

TCGv temp;
tcg_gen_movi_tl(temp, ctx->pc_succ_insn);
gen_helper_next_pc(cpu_env, temp);

Quoted 5 lines

while the helper function definition like that:
void helper_next_pc(CPURISCVState *env, target_ulong pc_next) {
    env->pc_next = pc_next;
}

This is a correct way to write pc_next. (Although I still don't understand why you need it.)

Quoted 1 line

and the same to read "env->pending_rvc"

No, you cannot read pending_rvc this way. You would need to incorporate pending_rvc into the flags set by cpu_get_tb_cpu_state. (Although I still don't understand why you need it.)

Quoted 5 lines

I'm thinking of a way to add 'C' extension at run time through waiting the correct aligned instruction, which I believe it might be after branch something quite similar to switching between ARM and THUMB states in ARM, for misa 'RVC' enable to take effect since it will be no possibility to check alignment with 'C' extension.

It seems to me that the C extension can be enabled at any point, since if C is off, you know that the next insn is aligned modulo 4.

It is only if the C extension is enabled, and you want to disable it, that is when we must check to see if the next insn is aligned mod 4. It is trivial to arrange for a particular instruction to be aligned, via assembler directives. So it seems silly to make the definition of the csr write to misa any more complicated than it is.

You are right that the existing code is broken, in that it is checking the alignment of the host PC and not the guest PC. However, I see no reason to introduce a new "pc_next" field when we already update the pc field ...

Quoted 11 lines

static void gen_system(DisasContext *ctx, uint32_t opc, int rd, int rs1,
int csr)

{
    TCGv source1, csr_store, dest, rs1_pass, imm_rs1;
    source1 = tcg_temp_new();
    csr_store = tcg_temp_new();
    dest = tcg_temp_new();
    rs1_pass = tcg_temp_new();
    imm_rs1 = tcg_temp_new();
    gen_get_gpr(source1, rs1);
    tcg_gen_movi_tl(cpu_pc, ctx->base.pc_next);

... here.

So as far as I can see, the only thing that needs fixing is GETPC.

r~

target/riscv/csr.c +0 -0
index 960d2b0aa9..8726ef802e 100644
@@ -370,10 +370,11 @@ static int write_misa(CPURISCVState *env, int csrno,

target_ulong val)

         val &= ~RVD;
     }

- /* Suppress 'C' if next instruction is not aligned
- * TODO: this should check next_pc
+ /*
+ * Suppress 'C' if next instruction is not aligned.
+ * We updated env->pc to the next insn in the translator.

      */
-    if ((val & RVC) && (GETPC() & ~3) != 0) {
+    if ((val & RVC) && (env->pc & ~3) != 0) {
         val &= ~RVC;
     }
author
#5 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Quoted 5 lines
Quoted 1 line

It seems to me that the C extension can be enabled at any point, since

if C is

Quoted 2 lines

off, you know that the next insn is aligned modulo 4.

Ok, This is mostly right. When C extension is enabled 32-bit base
instructions can be aligned on 2 bytes boundaries instead of 4 bytes only. So multiple enables and disables of C bit at different code areas theoretically may require this check on C extension enable. I'm not really sure, may be this is might not be a practical use scenario.

Quoted 11 lines

It is only if the C extension is enabled, and you want to disable it, that is

Quoted 1 line

when we must check to see if the next insn is aligned mod 4. It is

trivial to

Quoted 1 line

arrange for a particular instruction to be aligned, via assembler

directives.

Quoted 1 line

So it seems silly to make the definition of the csr write to misa any

more

Quoted 1 line

complicated than it is.

I completely agree with you that C extension disable should have

alignment check.

Quoted 21 lines

target/riscv/csr.c static int write_misa(CPURISCVState *env, int

Quoted 6 lines
target/riscv/csr.c +0 -0
index 960d2b0aa9..8726ef802e 100644
@@ -370,10 +370,11 @@ static int write_misa(CPURISCVState *env, int

csrno,

Quoted 14 lines
target/riscv/csr.c +4 -3
 target_ulong val)
         val &= ~RVD;
   }
 
-    /* Suppress 'C' if next instruction is not aligned
-     * TODO: this should check next_pc
+    /*
+     * Suppress 'C' if next instruction is not aligned.
+     * We updated env->pc to the next insn in the translator.
     */
-    if ((val & RVC) && (GETPC() & ~3) != 0) {
+    if ((val & RVC) && (env->pc & ~3) != 0) {
        val &= ~RVC;
    }

Just a hint, (env->pc & 3) instead of (env->pc & ~3) , right ?

Quoted 1 line

Thanks,
Ahmed

author
#6 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Quoted 4 lines
Quoted 4 lines
Quoted 3 lines

It seems to me that the C extension can be enabled at any point, since

if C is

Quoted 10 lines
Quoted 5 lines
Quoted 2 lines

off, you know that the next insn is aligned modulo 4.

Ok, This is mostly right. When C extension is enabled 32-bit base

instructions can be aligned on 2 bytes boundaries instead of 4 bytes only.

Quoted 1 line

So multiple enables and disables of C bit at different code areas

theoretically may require this check on C extension enable. I'm not really

Quoted 1 line

sure, may be this is might not be a practical use scenario.

This paragraph considers the current implementation of csr write for misa
register.

#7 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

In reply to [email protected]
Quoted 3 lines · On 2/22/19 10:25 AM, [email protected] wrote:

From: ahmed_magdy <[email protected]>

This appears to be your first contribution to qemu. Welcome to the community!

Typically, a Signed-off-by designation should be a proper name (what you would sign a legal document with, as it has a legal significance on your right to contribute the code). Using all lowercase and _ instead of space looks like a username, and while I am not one to tell you it can't be a legal name, it is unusual enough to at least raise my suspicions.

Furthermore, your commit message doesn't give any details beyond the "what" in the subject line. The body of the commit message should explain the "why" (what bug are you fixing, how to reproduce it), so that a reviewer stands a chance of determining if the code matches the description you gave, and if the issue you describe really does warrant the inclusion of your patch. You gave a brief "why" in your cover letter:

"I'm submiting this patch to properly check the next instruction alignment and scheduale compression extenstion enable upon 'MISA' register writes to later aligned instruction through exporting next instruction 'pc' to riscv cpu state"

where it would be wise to include an improved version of that text with this commit proper (since the cover letter does not get applied to git).
For that matter, when sending a single patch, a cover letter is
optional (it is only mandatory when sending a multi-patch series).

For more patch submission hints, see:
https://wiki.qemu.org/Contribute/SubmitAPatch

Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org
author
#8 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Thank you so much, Eric.

Sorry about this unclear description.
I forgot to fix user name in git configuration before submitting the patch.

Sorry about any inconvenience.
Thanks,
Ahmed

Quoted 39 lines · Eric Blake
Quoted 3 lines · On 2/22/19 10:25 AM, [email protected] wrote:

From: ahmed_magdy <[email protected]>

This appears to be your first contribution to qemu. Welcome to the community!

Typically, a Signed-off-by designation should be a proper name (what you would sign a legal document with, as it has a legal significance on your right to contribute the code). Using all lowercase and _ instead of space looks like a username, and while I am not one to tell you it can't be a legal name, it is unusual enough to at least raise my suspicions.

Furthermore, your commit message doesn't give any details beyond the "what" in the subject line. The body of the commit message should explain the "why" (what bug are you fixing, how to reproduce it), so that a reviewer stands a chance of determining if the code matches the description you gave, and if the issue you describe really does warrant the inclusion of your patch. You gave a brief "why" in your cover letter:

"I'm submiting this patch to properly check the next instruction alignment and scheduale compression extenstion enable upon 'MISA' register writes to later aligned instruction through exporting next instruction 'pc' to riscv cpu state"

where it would be wise to include an improved version of that text with this commit proper (since the cover letter does not get applied to git).
For that matter, when sending a single patch, a cover letter is
optional (it is only mandatory when sending a multi-patch series).

For more patch submission hints, see:
https://wiki.qemu.org/Contribute/SubmitAPatch

--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3226
Virtualization: qemu.org | libvirt.org

#9 lore raw

Re: [Qemu-devel] [PATCH] riscv: Add proper alignment check and pending 'C' extension upon misa writes

Quoted 6 lines · Amed Magdy

Thank you so much, Eric.

Sorry about this unclear description.
I forgot to fix user name in git configuration before submitting the patch.

Sorry about any inconvenience.

Don't worry about it. We were all once first-time contributors, and it can be a large learning curve.

Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org