Skip to contentSkip to conversation

target/ppc: Fix lqarx to set cpu_reserve

PATCH Superseded Nicholas Piggin 3 participants 15 messages Jun 4, 2023 – Jun 20, 2023

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/4 target/ppc: Fix lqarx to set cpu_reserve

lqarx does not set cpu_reserve, which causes stqcx. to never succeed. Fix this and slightly rearrange gen_load_locked so the two functions match more closely.

Fixes 94bf2658676 ("target/ppc: Use atomic load for LQ and LQARX")
Fixes 57b38ffd0c6 ("target/ppc: Use tcg_gen_qemu_{ld,st}_i128 for LQARX, LQ, STQ")

---
cpu_reserve got lost in the parallel part with the first patch, then from serial part when it was merged with the parallel by the second patch.

Thanks,
Nick

 target/ppc/translate.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
target/ppc/translate.c +2 -1
index 3650d2985d..e129cdcb8f 100644
@@ -3583,8 +3583,8 @@ static void gen_load_locked(DisasContext *ctx, MemOp memop)
 
     gen_set_access_type(ctx, ACCESS_RES);
     gen_addr_reg_index(ctx, t0);
-    tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
     tcg_gen_mov_tl(cpu_reserve, t0);
+    tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
     tcg_gen_mov_tl(cpu_reserve_val, gpr);
     tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
 }
@@ -3872,6 +3872,7 @@ static void gen_lqarx(DisasContext *ctx)
     gen_set_access_type(ctx, ACCESS_RES);
     EA = tcg_temp_new();
     gen_addr_reg_index(ctx, EA);
+    tcg_gen_mov_tl(cpu_reserve, EA);
 
     /* Note that the low part is always in RD+1, even in LE mode.  */
     lo = cpu_gpr[rd + 1];
Signature · 1 line
2.40.1
#1 lore raw

PATCH 2/4 target/ppc: Ensure stcx size matches larx

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

---
 target/ppc/cpu.h       | 1 +
 target/ppc/cpu_init.c  | 4 ++--
 target/ppc/translate.c | 8 ++++++++
 3 files changed, 11 insertions(+), 2 deletions(-)
target/ppc/cpu.h +1 -0
index 7959bfed0a..1d71f325d8 100644
@@ -1124,6 +1124,7 @@ struct CPUArchState {
     target_ulong ca32;
 
     target_ulong reserve_addr; /* Reservation address */
+    target_ulong reserve_size; /* Reservation size */
     target_ulong reserve_val;  /* Reservation value */
     target_ulong reserve_val2;
 
target/ppc/cpu_init.c +2 -2
index 944a74befe..082981b148 100644
@@ -7421,8 +7421,8 @@ void ppc_cpu_dump_state(CPUState *cs, FILE *f, int flags)
         }
         qemu_fprintf(f, " %c%c", a, env->crf[i] & 0x01 ? 'O' : ' ');
     }
-    qemu_fprintf(f, " ]             RES " TARGET_FMT_lx "\n",
-                 env->reserve_addr);
+    qemu_fprintf(f, " ]     RES %03x@" TARGET_FMT_lx "\n",
+                 (int)env->reserve_size, env->reserve_addr);
 
     if (flags & CPU_DUMP_FPU) {
         for (i = 0; i < 32; i++) {
target/ppc/translate.c +8 -0
index e129cdcb8f..5195047146 100644
@@ -71,6 +71,7 @@ static TCGv cpu_cfar;
 #endif
 static TCGv cpu_xer, cpu_so, cpu_ov, cpu_ca, cpu_ov32, cpu_ca32;
 static TCGv cpu_reserve;
+static TCGv cpu_reserve_size;
 static TCGv cpu_reserve_val;
 static TCGv cpu_reserve_val2;
 static TCGv cpu_fpscr;
@@ -141,6 +142,9 @@ void ppc_translate_init(void)
     cpu_reserve = tcg_global_mem_new(cpu_env,
                                      offsetof(CPUPPCState, reserve_addr),
                                      "reserve_addr");
+    cpu_reserve_size = tcg_global_mem_new(cpu_env,
+                                          offsetof(CPUPPCState, reserve_size),
+                                          "reserve_size");
     cpu_reserve_val = tcg_global_mem_new(cpu_env,
                                          offsetof(CPUPPCState, reserve_val),
                                          "reserve_val");
@@ -3584,6 +3588,7 @@ static void gen_load_locked(DisasContext *ctx, MemOp memop)
     gen_set_access_type(ctx, ACCESS_RES);
     gen_addr_reg_index(ctx, t0);
     tcg_gen_mov_tl(cpu_reserve, t0);
+    tcg_gen_movi_tl(cpu_reserve_size, memop_size(memop));
     tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
     tcg_gen_mov_tl(cpu_reserve_val, gpr);
     tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
@@ -3816,6 +3821,7 @@ static void gen_conditional_store(DisasContext *ctx, MemOp memop)
     gen_set_access_type(ctx, ACCESS_RES);
     gen_addr_reg_index(ctx, t0);
     tcg_gen_brcond_tl(TCG_COND_NE, t0, cpu_reserve, l1);
+    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, memop_size(memop), l1);
 
     t0 = tcg_temp_new();
     tcg_gen_atomic_cmpxchg_tl(t0, cpu_reserve, cpu_reserve_val,
@@ -3873,6 +3879,7 @@ static void gen_lqarx(DisasContext *ctx)
     EA = tcg_temp_new();
     gen_addr_reg_index(ctx, EA);
     tcg_gen_mov_tl(cpu_reserve, EA);
+    tcg_gen_movi_tl(cpu_reserve_size, 128);
 
     /* Note that the low part is always in RD+1, even in LE mode.  */
     lo = cpu_gpr[rd + 1];
@@ -3907,6 +3914,7 @@ static void gen_stqcx_(DisasContext *ctx)
     gen_addr_reg_index(ctx, EA);
 
     tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lab_fail);
+    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, 128, lab_fail);
 
     cmp = tcg_temp_new_i128();
     val = tcg_temp_new_i128();
Signature · 1 line
2.40.1
Reviewed-by on v1 2/4
#2 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

target/ppc/translate.c static void gen_load_locked(DisasContext *ctx, MemOp memop) · line 3591

Quoted 5 lines
target/ppc/translate.c +1 -0
@@ -3584,6 +3588,7 @@ static void gen_load_locked(DisasContext *ctx, MemOp memop)
      gen_set_access_type(ctx, ACCESS_RES);
      gen_addr_reg_index(ctx, t0);
      tcg_gen_mov_tl(cpu_reserve, t0);
+    tcg_gen_movi_tl(cpu_reserve_size, memop_size(memop));

Not that it really matters, this produces a byte value...

target/ppc/translate.c static void gen_lqarx(DisasContext *ctx) · line 3882

Quoted 5 lines
target/ppc/translate.c +1 -0
@@ -3873,6 +3879,7 @@ static void gen_lqarx(DisasContext *ctx)
      EA = tcg_temp_new();
      gen_addr_reg_index(ctx, EA);
      tcg_gen_mov_tl(cpu_reserve, EA);
+    tcg_gen_movi_tl(cpu_reserve_size, 128);

... so perhaps ideally this would be 16.

Perhaps name it reserve_length to exactly match the manual. Otherwise,

r~

#3 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

In reply to Nicholas Piggin
Quoted 3 lines · Nicholas Piggin

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

Hmm, question: reserve_addr is a VMSTATE field, but reserve_val is not (nor reserve_size after this patch).

Blue Swirl added that with commit a456d59c20f ("VM load/save support for PPC CPU"), and when reserve_val was added in commit 18b21a2f83a ("target-ppc: retain l{w,d}arx loaded value") it did not get migrated.

Could we end up with reserve_addr != -1, but with a bogus reserve_val, which could then permit a stcx. incorrectly? Not entirely outlandish if reserve_val starts out initialised to zero.

Could we just clear the reserve in cpu_post_load? It is permitted to be lost for an implementation-specific reason. Doesn't seem necessary to try keep it alive over a migration.

Thanks,
Nick

#4 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

Quoted 12 lines · Nicholas Piggin
Quoted 3 lines · Nicholas Piggin

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

Hmm, question: reserve_addr is a VMSTATE field, but reserve_val is not (nor reserve_size after this patch).

Blue Swirl added that with commit a456d59c20f ("VM load/save support for PPC CPU"), and when reserve_val was added in commit 18b21a2f83a ("target-ppc: retain l{w,d}arx loaded value") it did not get migrated.

Quoted 7 lines

Could we end up with reserve_addr != -1, but with a bogus reserve_val, which could then permit a stcx. incorrectly? Not entirely outlandish if reserve_val starts out initialised to zero.

Could we just clear the reserve in cpu_post_load? It is permitted to be lost for an implementation-specific reason. Doesn't seem necessary to try keep it alive over a migration.

It's not a bad idea to flush the reservation over migrate. You can do

- VMSTATE_UINTTL(env.reserve_addr, PowerPCCPU),
+ VMSTATE_UNUSED(sizeof(target_long))

when making that change.

Peter, any thoughts on this? If we're going to do one guest, we might ought to do the same for all load-lock/store-conditional guests.

r~

#5 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

Quoted 18 lines · Richard Henderson
Quoted 16 lines · Nicholas Piggin
Quoted 3 lines · Nicholas Piggin

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

Hmm, question: reserve_addr is a VMSTATE field, but reserve_val is not (nor reserve_size after this patch).

Blue Swirl added that with commit a456d59c20f ("VM load/save support for PPC CPU"), and when reserve_val was added in commit 18b21a2f83a ("target-ppc: retain l{w,d}arx loaded value") it did not get migrated.

Could we end up with reserve_addr != -1, but with a bogus reserve_val, which could then permit a stcx. incorrectly? Not entirely outlandish if reserve_val starts out initialised to zero.

Quoted 5 lines
Quoted 3 lines

Could we just clear the reserve in cpu_post_load? It is permitted to be lost for an implementation-specific reason. Doesn't seem necessary to try keep it alive over a migration.

It's not a bad idea to flush the reservation over migrate.

Is there any particular reason to do so? The default simple thing is "if this is state that persists across instructions then migrate it"; we usually reserve "do something special in post-load" for oddball cases where "just copy the data" doesn't work.

target/arm migrates both the exclusive addr and value.

target/mips migrates lladdr but has forgotten llval (and perhaps llval_wp and llnewval_wp, depending on what those fields do).

thanks
-- PMM

#6 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

Quoted 26 lines · Peter Maydell
Quoted 18 lines · Richard Henderson
Quoted 16 lines · Nicholas Piggin
Quoted 3 lines · Nicholas Piggin

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

Hmm, question: reserve_addr is a VMSTATE field, but reserve_val is not (nor reserve_size after this patch).

Blue Swirl added that with commit a456d59c20f ("VM load/save support for PPC CPU"), and when reserve_val was added in commit 18b21a2f83a ("target-ppc: retain l{w,d}arx loaded value") it did not get migrated.

Could we end up with reserve_addr != -1, but with a bogus reserve_val, which could then permit a stcx. incorrectly? Not entirely outlandish if reserve_val starts out initialised to zero.

Quoted 5 lines
Quoted 3 lines

Could we just clear the reserve in cpu_post_load? It is permitted to be lost for an implementation-specific reason. Doesn't seem necessary to try keep it alive over a migration.

It's not a bad idea to flush the reservation over migrate.

Quoted 7 lines

Is there any particular reason to do so? The default simple thing is "if this is state that persists across instructions then migrate it"; we usually reserve "do something special in post-load" for oddball cases where "just copy the data" doesn't work.

target/arm migrates both the exclusive addr and value.

ppc is adding "size", which arm technically should have as well.

Quoted 3 lines

target/mips migrates lladdr but has forgotten llval (and perhaps llval_wp and llnewval_wp, depending on what those fields do).

So, similarly, would need to handle migration for which all of the required data is not present.

The thought is, rather than migrate this new data also, and handle compatibility, simply discard all reservations.

r~

#7 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

Quoted 34 lines · Richard Henderson
Quoted 32 lines · Peter Maydell
Quoted 18 lines · Richard Henderson
Quoted 16 lines · Nicholas Piggin
Quoted 3 lines · Nicholas Piggin

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

Hmm, question: reserve_addr is a VMSTATE field, but reserve_val is not (nor reserve_size after this patch).

Blue Swirl added that with commit a456d59c20f ("VM load/save support for PPC CPU"), and when reserve_val was added in commit 18b21a2f83a ("target-ppc: retain l{w,d}arx loaded value") it did not get migrated.

Could we end up with reserve_addr != -1, but with a bogus reserve_val, which could then permit a stcx. incorrectly? Not entirely outlandish if reserve_val starts out initialised to zero.

Quoted 5 lines
Quoted 3 lines

Could we just clear the reserve in cpu_post_load? It is permitted to be lost for an implementation-specific reason. Doesn't seem necessary to try keep it alive over a migration.

It's not a bad idea to flush the reservation over migrate.

Is there any particular reason to do so? The default simple thing is "if this is state that persists across instructions then migrate it"; we usually reserve "do something special in post-load" for oddball cases where "just copy the data" doesn't work.

Quoted 3 lines
Quoted 1 line

target/arm migrates both the exclusive addr and value.

ppc is adding "size", which arm technically should have as well.

Arm allows an implementation to require the transaction size to match on loadexcl and storexcl, but doesn't mandate it, fwiw. (Also, our implementation is miles away from the architectural requirements anyway because we operate on virtual addresses, not physical addresses.)

Quoted 9 lines
Quoted 3 lines

target/mips migrates lladdr but has forgotten llval (and perhaps llval_wp and llnewval_wp, depending on what those fields do).

So, similarly, would need to handle migration for which all of the required data is not present.

The thought is, rather than migrate this new data also, and handle compatibility, simply discard all reservations.

I don't see a problem for normal migration and snapshotting. I do wonder whether this would have a bad interaction with record-and-replay's use of snapshots. Does that expect "execution from the loaded snapshot" to match "execution continues from point of snapshot save" ?

-- PMM

#8 lore raw

Re: [PATCH 2/4] target/ppc: Ensure stcx size matches larx

Quoted 38 lines · Peter Maydell
Quoted 36 lines · Richard Henderson
Quoted 33 lines · Peter Maydell
Quoted 18 lines · Richard Henderson
Quoted 16 lines · Nicholas Piggin
Quoted 3 lines · Nicholas Piggin

Differently-sized larx/stcx. pairs can succeed if the starting address matches. Add a size check to require stcx. exactly match the larx that established the reservation.

Hmm, question: reserve_addr is a VMSTATE field, but reserve_val is not (nor reserve_size after this patch).

Blue Swirl added that with commit a456d59c20f ("VM load/save support for PPC CPU"), and when reserve_val was added in commit 18b21a2f83a ("target-ppc: retain l{w,d}arx loaded value") it did not get migrated.

Could we end up with reserve_addr != -1, but with a bogus reserve_val, which could then permit a stcx. incorrectly? Not entirely outlandish if reserve_val starts out initialised to zero.

Quoted 5 lines
Quoted 3 lines

Could we just clear the reserve in cpu_post_load? It is permitted to be lost for an implementation-specific reason. Doesn't seem necessary to try keep it alive over a migration.

It's not a bad idea to flush the reservation over migrate.

Is there any particular reason to do so? The default simple thing is "if this is state that persists across instructions then migrate it"; we usually reserve "do something special in post-load" for oddball cases where "just copy the data" doesn't work.

target/arm migrates both the exclusive addr and value.

Quoted 7 lines
Quoted 1 line

ppc is adding "size", which arm technically should have as well.

Arm allows an implementation to require the transaction size to match on loadexcl and storexcl, but doesn't mandate it, fwiw. (Also, our implementation is miles away from the architectural requirements anyway because we operate on virtual addresses, not physical addresses.)

The same as powerpc. Size *and* address within reserve granule does not have to match the larx which established the reserve, but the latter we always enforced and in practice no open source software seems to hit it (or AIX).

My thinking is that it is good to tighten it because very likely software that gets it wrong is deviating from ISA unintentionally. Linux provides no HWCAP bit to allow code to test such implementation details, for example.

Quoted 7 lines
Quoted 7 lines
Quoted 3 lines

target/mips migrates lladdr but has forgotten llval (and perhaps llval_wp and llnewval_wp, depending on what those fields do).

So, similarly, would need to handle migration for which all of the required data is not present.

Quoted 8 lines
Quoted 2 lines

The thought is, rather than migrate this new data also, and handle compatibility, simply discard all reservations.

I don't see a problem for normal migration and snapshotting. I do wonder whether this would have a bad interaction with record-and-replay's use of snapshots. Does that expect "execution from the loaded snapshot" to match "execution continues from point of snapshot save" ?

I don't mind the idea of moving the new state across, I wondered if clearing the reserve would be easier for compatibility and backporting.

I don't know the rr code but if the snapshots use this vmstate and the replay from that is expected to match exactly the recording, then I think you must be right.

Thanks,
Nick

#9 lore raw

PATCH 3/4 target/ppc: Remove larx/stcx. memory barrier semantics

In reply to Nicholas Piggin

larx and stcx. are not defined to order any memory operations. Remove the barriers.

---
 target/ppc/translate.c | 11 -----------
 1 file changed, 11 deletions(-)
target/ppc/translate.c +0 -11
index 5195047146..77e1c5abb6 100644
@@ -3591,7 +3591,6 @@ static void gen_load_locked(DisasContext *ctx, MemOp memop)
     tcg_gen_movi_tl(cpu_reserve_size, memop_size(memop));
     tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
     tcg_gen_mov_tl(cpu_reserve_val, gpr);
-    tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
 }
 
 #define LARX(name, memop)                  \
@@ -3835,11 +3834,6 @@ static void gen_conditional_store(DisasContext *ctx, MemOp memop)
 
     gen_set_label(l1);
 
-    /*
-     * Address mismatch implies failure.  But we still need to provide
-     * the memory barrier semantics of the instruction.
-     */
-    tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
     tcg_gen_trunc_tl_i32(cpu_crf[0], cpu_so);
 
     gen_set_label(l2);
@@ -3943,11 +3937,6 @@ static void gen_stqcx_(DisasContext *ctx)
     tcg_gen_br(lab_over);
     gen_set_label(lab_fail);
 
-    /*
-     * Address mismatch implies failure.  But we still need to provide
-     * the memory barrier semantics of the instruction.
-     */
-    tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
     tcg_gen_trunc_tl_i32(cpu_crf[0], cpu_so);
 
     gen_set_label(lab_over);
Signature · 1 line
2.40.1
#11 lore raw

PATCH 4/4 target/ppc: Rework store conditional to avoid branch

In reply to Nicholas Piggin

Rework store conditional to avoid a branch in the success case. Change some of the variable names and layout while here so gen_conditional_store more closely matches gen_stqcx_.

---
 target/ppc/translate.c | 65 ++++++++++++++++++++----------------------
 1 file changed, 31 insertions(+), 34 deletions(-)
target/ppc/translate.c +31 -34
index 77e1c5abb6..cf99e961f7 100644
@@ -3812,31 +3812,32 @@ static void gen_stdat(DisasContext *ctx)
 
 static void gen_conditional_store(DisasContext *ctx, MemOp memop)
 {
-    TCGLabel *l1 = gen_new_label();
-    TCGLabel *l2 = gen_new_label();
-    TCGv t0 = tcg_temp_new();
-    int reg = rS(ctx->opcode);
+    TCGLabel *lfail;
+    TCGv EA;
+    TCGv cr0;
+    TCGv t0;
+    int rs = rS(ctx->opcode);
 
+    lfail = gen_new_label();
+    EA = tcg_temp_new();
+    cr0 = tcg_temp_new();
+    t0 = tcg_temp_new();
+
+    tcg_gen_mov_tl(cr0, cpu_so);
     gen_set_access_type(ctx, ACCESS_RES);
-    gen_addr_reg_index(ctx, t0);
-    tcg_gen_brcond_tl(TCG_COND_NE, t0, cpu_reserve, l1);
-    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, memop_size(memop), l1);
+    gen_addr_reg_index(ctx, EA);
+    tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lfail);
+    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, memop_size(memop), lfail);
 
-    t0 = tcg_temp_new();
     tcg_gen_atomic_cmpxchg_tl(t0, cpu_reserve, cpu_reserve_val,
-                              cpu_gpr[reg], ctx->mem_idx,
-                              DEF_MEMOP(memop) | MO_ALIGN);
+                              cpu_gpr[rs], ctx->mem_idx,
+                              memop | MO_ALIGN);
     tcg_gen_setcond_tl(TCG_COND_EQ, t0, t0, cpu_reserve_val);
     tcg_gen_shli_tl(t0, t0, CRF_EQ_BIT);
-    tcg_gen_or_tl(t0, t0, cpu_so);
-    tcg_gen_trunc_tl_i32(cpu_crf[0], t0);
-    tcg_gen_br(l2);
+    tcg_gen_or_tl(cr0, cr0, t0);
 
-    gen_set_label(l1);
-
-    tcg_gen_trunc_tl_i32(cpu_crf[0], cpu_so);
-
-    gen_set_label(l2);
+    gen_set_label(lfail);
+    tcg_gen_trunc_tl_i32(cpu_crf[0], cr0);
     tcg_gen_movi_tl(cpu_reserve, -1);
 }
 
@@ -3890,25 +3891,26 @@ static void gen_lqarx(DisasContext *ctx)
 /* stqcx. */
 static void gen_stqcx_(DisasContext *ctx)
 {
-    TCGLabel *lab_fail, *lab_over;
-    int rs = rS(ctx->opcode);
+    TCGLabel *lfail;
     TCGv EA, t0, t1;
+    TCGv cr0;
     TCGv_i128 cmp, val;
+    int rs = rS(ctx->opcode);
 
     if (unlikely(rs & 1)) {
         gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
         return;
     }
 
-    lab_fail = gen_new_label();
-    lab_over = gen_new_label();
+    lfail = gen_new_label();
+    EA = tcg_temp_new();
+    cr0 = tcg_temp_new();
 
+    tcg_gen_mov_tl(cr0, cpu_so);
     gen_set_access_type(ctx, ACCESS_RES);
-    EA = tcg_temp_new();
     gen_addr_reg_index(ctx, EA);
-
-    tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lab_fail);
-    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, 128, lab_fail);
+    tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lfail);
+    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, 128, lfail);
 
     cmp = tcg_temp_new_i128();
     val = tcg_temp_new_i128();
@@ -3931,15 +3933,10 @@ static void gen_stqcx_(DisasContext *ctx)
 
     tcg_gen_setcondi_tl(TCG_COND_EQ, t0, t0, 0);
     tcg_gen_shli_tl(t0, t0, CRF_EQ_BIT);
-    tcg_gen_or_tl(t0, t0, cpu_so);
-    tcg_gen_trunc_tl_i32(cpu_crf[0], t0);
-
-    tcg_gen_br(lab_over);
-    gen_set_label(lab_fail);
-
-    tcg_gen_trunc_tl_i32(cpu_crf[0], cpu_so);
+    tcg_gen_or_tl(cr0, cr0, t0);
 
-    gen_set_label(lab_over);
+    gen_set_label(lfail);
+    tcg_gen_trunc_tl_i32(cpu_crf[0], cr0);
     tcg_gen_movi_tl(cpu_reserve, -1);
 }
 #endif /* defined(TARGET_PPC64) */
Signature · 1 line
2.40.1
Reviewed-by on v1 4/4
#12 lore raw

Re: [PATCH 4/4] target/ppc: Rework store conditional to avoid branch

Quoted 37 lines · Nicholas Piggin

Rework store conditional to avoid a branch in the success case. Change some of the variable names and layout while here so gen_conditional_store more closely matches gen_stqcx_.

---
  target/ppc/translate.c | 65 ++++++++++++++++++++----------------------
  1 file changed, 31 insertions(+), 34 deletions(-)
target/ppc/translate.c +11 -7
index 77e1c5abb6..cf99e961f7 100644
@@ -3812,31 +3812,32 @@ static void gen_stdat(DisasContext *ctx)
 
  static void gen_conditional_store(DisasContext *ctx, MemOp memop)
  {
-    TCGLabel *l1 = gen_new_label();
-    TCGLabel *l2 = gen_new_label();
-    TCGv t0 = tcg_temp_new();
-    int reg = rS(ctx->opcode);
+    TCGLabel *lfail;
+    TCGv EA;
+    TCGv cr0;
+    TCGv t0;
+    int rs = rS(ctx->opcode);
 
+    lfail = gen_new_label();
+    EA = tcg_temp_new();
+    cr0 = tcg_temp_new();
+    t0 = tcg_temp_new();
+
+    tcg_gen_mov_tl(cr0, cpu_so);
      gen_set_access_type(ctx, ACCESS_RES);
-    gen_addr_reg_index(ctx, t0);
-    tcg_gen_brcond_tl(TCG_COND_NE, t0, cpu_reserve, l1);
-    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, memop_size(memop), l1);

target/ppc/translate.c static void gen_stdat(DisasContext *ctx) · line 3834

Quoted 10 lines
target/ppc/translate.c +5 -3
+    gen_addr_reg_index(ctx, EA);
+    tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lfail);
+    tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_size, memop_size(memop), lfail);
 
-    t0 = tcg_temp_new();
      tcg_gen_atomic_cmpxchg_tl(t0, cpu_reserve, cpu_reserve_val,
-                              cpu_gpr[reg], ctx->mem_idx,
-                              DEF_MEMOP(memop) | MO_ALIGN);
+                              cpu_gpr[rs], ctx->mem_idx,
+                              memop | MO_ALIGN);

Lost DEF_MEMOP here. Otherwise,

r~

#13 lore raw
In reply to Nicholas Piggin
Quoted 4 lines · Nicholas Piggin

lqarx does not set cpu_reserve, which causes stqcx. to never succeed. Fix this and slightly rearrange gen_load_locked so the two functions match more closely.

Quoted 8 lines
Fixes 94bf2658676 ("target/ppc: Use atomic load for LQ and LQARX")
Fixes 57b38ffd0c6 ("target/ppc: Use tcg_gen_qemu_{ld,st}_i128 for LQARX, LQ, STQ")

---
cpu_reserve got lost in the parallel part with the first patch, then from serial part when it was merged with the parallel by the second patch.

Oops, sorry about that.

Quoted 12 lines

Thanks,
Nick

  target/ppc/translate.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
target/ppc/translate.c +0 -0
index 3650d2985d..e129cdcb8f 100644

target/ppc/translate.c static void gen_load_locked(DisasContext *ctx, MemOp memop) · line 3588

Quoted 7 lines
target/ppc/translate.c +1 -1
@@ -3583,8 +3583,8 @@ static void gen_load_locked(DisasContext *ctx, MemOp memop)
 
      gen_set_access_type(ctx, ACCESS_RES);
      gen_addr_reg_index(ctx, t0);
-    tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
      tcg_gen_mov_tl(cpu_reserve, t0);
+    tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
      tcg_gen_mov_tl(cpu_reserve_val, gpr);

This change is wrong. Reserve should not be set if the load faults.

target/ppc/translate.c static void gen_lqarx(DisasContext *ctx) · line 3878

Quoted 10 lines
target/ppc/translate.c +1 -0
      tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
  }
@@ -3872,6 +3872,7 @@ static void gen_lqarx(DisasContext *ctx)
      gen_set_access_type(ctx, ACCESS_RES);
      EA = tcg_temp_new();
      gen_addr_reg_index(ctx, EA);
+    tcg_gen_mov_tl(cpu_reserve, EA);
 
      /* Note that the low part is always in RD+1, even in LE mode.  */
      lo = cpu_gpr[rd + 1];

This needs to go lower with the sets of reserve_val*.

r~

#14 lore raw
Quoted 14 lines · Richard Henderson
Quoted 4 lines · Nicholas Piggin

lqarx does not set cpu_reserve, which causes stqcx. to never succeed. Fix this and slightly rearrange gen_load_locked so the two functions match more closely.

Quoted 8 lines
Fixes 94bf2658676 ("target/ppc: Use atomic load for LQ and LQARX")
Fixes 57b38ffd0c6 ("target/ppc: Use tcg_gen_qemu_{ld,st}_i128 for LQARX, LQ, STQ")

---
cpu_reserve got lost in the parallel part with the first patch, then from serial part when it was merged with the parallel by the second patch.

Quoted 1 line

Oops, sorry about that.

No problem, I really appreciate your work on ppc, ppc just should have more unit tests particularly for non-trivial instructions like lqarx which would have caught it. That's the real problem.

Quoted 14 lines
Quoted 14 lines

Thanks,
Nick

  target/ppc/translate.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
target/ppc/translate.c +0 -0
index 3650d2985d..e129cdcb8f 100644
@@ -3583,8 +3583,8 @@ static void gen_load_locked(DisasContext *ctx, MemOp memop)
 
Quoted 8 lines

target/ppc/translate.c static void gen_load_locked(DisasContext *ctx, MemOp memop) · line 3588

Quoted 6 lines
target/ppc/translate.c +1 -1
      gen_set_access_type(ctx, ACCESS_RES);
      gen_addr_reg_index(ctx, t0);
-    tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
      tcg_gen_mov_tl(cpu_reserve, t0);
+    tcg_gen_qemu_ld_tl(gpr, t0, ctx->mem_idx, memop | MO_ALIGN);
      tcg_gen_mov_tl(cpu_reserve_val, gpr);

This change is wrong. Reserve should not be set if the load faults.

Oh yeah, good catch.

Thanks
Nick