Skip to contentSkip to conversation

s390x/tcg: Implement some z14 facilities

PATCH Superseded David Hildenbrand 4 participants 28 messages Sep 22, 2020 – Sep 28, 2020

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover s390x/tcg: Implement some z14 facilities

With this patchset, I can boot a kernel compiled for z14 (requiring Miscellaneous-Instruction-Extensions Facility 2) (using -cpu max).

I am currently working on "vector-enhancements facility 1", I have most stuff in place except:
- VECTOR FP (MAXIMUM|MINIMUM): needs some manual work to cover all types of comparisons.
- VECTOR FP (NEGATIVE) MULTIPLY AND (ADD|SUBTRACT): needs float128_muladd()

Once we have that in place, we can convert the QEMU machine to a stripped-down z14.

David Hildenbrand (8):
s390x/tcg: Implement ADD HALFWORD (AGH)
s390x/tcg: Implement SUBTRACT HALFWORD (SGH)
s390x/tcg: Implement MULTIPLY (MG, MGRK)
s390x/tcg: Implement MULTIPLY HALFWORD (MGH)
s390x/tcg: Implement BRANCH INDIRECT ON CONDITION (BIC) s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC) s390x/tcg: We support Miscellaneous-Instruction-Extensions Facility 2 s390x/tcg: Implement CIPHER MESSAGE WITH AUTHENTICATION (KMA)

 target/s390x/cc_helper.c    | 34 ++++++++++++++++++++++++++++
 target/s390x/gen-features.c |  2 ++
 target/s390x/helper.c       |  2 ++
 target/s390x/insn-data.def  | 12 ++++++++++
 target/s390x/internal.h     |  2 ++
 target/s390x/translate.c    | 45 +++++++++++++++++++++++++++++++++++++
 6 files changed, 97 insertions(+)
Signature · 1 line
2.26.2
#9 lore raw

PATCH 3/8 s390x/tcg: Implement MULTIPLY (MG, MGRK)

Multiply two signed 64bit values and store the 128bit result in r1 (0-63) and r1 + 1 (64-127).

---
 target/s390x/insn-data.def |  2 ++
 target/s390x/translate.c   | 13 +++++++++++++
 2 files changed, 15 insertions(+)
target/s390x/insn-data.def +2 -0
index e851e9df5e..2b4ad1530d 100644
@@ -650,8 +650,10 @@
 
 /* MULTIPLY */
     C(0x1c00, MR,      RR_a,  Z,   r1p1_32s, r2_32s, new, r1_D32, mul, 0)
+    C(0xb9ec, MGRK,    RRF_a, MIE2,r3_o, r2_o, r1_P, 0, muls128, 0)
     C(0x5c00, M,       RX_a,  Z,   r1p1_32s, m2_32s, new, r1_D32, mul, 0)
     C(0xe35c, MFY,     RXY_a, GIE, r1p1_32s, m2_32s, new, r1_D32, mul, 0)
+    C(0xe384, MG,      RXY_a, MIE2,r1p1_o, m2_64, r1_P, 0, muls128, 0)
     F(0xb317, MEEBR,   RRE,   Z,   e1, e2, new, e1, meeb, 0, IF_BFP)
     F(0xb31c, MDBR,    RRE,   Z,   f1, f2, new, f1, mdb, 0, IF_BFP)
     F(0xb34c, MXBR,    RRE,   Z,   x2h, x2l, x1, x1, mxb, 0, IF_BFP)
target/s390x/translate.c +13 -0
index 21d77b7e74..b536491892 100644
@@ -3518,6 +3518,12 @@ static DisasJumpType op_mul128(DisasContext *s, DisasOps *o)
     return DISAS_NEXT;
 }
 
+static DisasJumpType op_muls128(DisasContext *s, DisasOps *o)
+{
+    tcg_gen_muls2_i64(o->out2, o->out, o->in1, o->in2);
+    return DISAS_NEXT;
+}
+
 static DisasJumpType op_meeb(DisasContext *s, DisasOps *o)
 {
     gen_helper_meeb(o->out, cpu_env, o->in1, o->in2);
@@ -5542,6 +5548,13 @@ static void in1_r1p1(DisasContext *s, DisasOps *o)
 }
 #define SPEC_in1_r1p1 SPEC_r1_even
 
+static void in1_r1p1_o(DisasContext *s, DisasOps *o)
+{
+    o->in1 = regs[get_field(s, r1) + 1];
+    o->g_in1 = true;
+}
+#define SPEC_in1_r1p1_o SPEC_r1_even
+
 static void in1_r1p1_32s(DisasContext *s, DisasOps *o)
 {
     o->in1 = tcg_temp_new_i64();
Signature · 1 line
2.26.2
#6 lore raw

PATCH 2/8 s390x/tcg: Implement SUBTRACT HALFWORD (SGH)

In reply to David Hildenbrand

Easy to wire up.

---
 target/s390x/insn-data.def | 1 +
 1 file changed, 1 insertion(+)
target/s390x/insn-data.def +1 -0
index 8dbeaf8c49..e851e9df5e 100644
@@ -883,6 +883,7 @@
 /* SUBTRACT HALFWORD */
     C(0x4b00, SH,      RX_a,  Z,   r1, m2_16s, new, r1_32, sub, subs32)
     C(0xe37b, SHY,     RXY_a, LD,  r1, m2_16s, new, r1_32, sub, subs32)
+    C(0xe339, SGH,     RXY_a, MIE2,r1, m2_16s, r1, 0, sub, subs64)
 /* SUBTRACT HIGH */
     C(0xb9c9, SHHHR,   RRF_a, HW,  r2_sr32, r3_sr32, new, r1_32h, sub, subs32)
     C(0xb9d9, SHHLR,   RRF_a, HW,  r2_sr32, r3, new, r1_32h, sub, subs32)
Signature · 1 line
2.26.2
#1 lore raw

PATCH 1/8 s390x/tcg: Implement ADD HALFWORD (AGH)

In reply to David Hildenbrand

Easy, just like ADD HALFWORD IMMEDIATE (AGHI).

---
 target/s390x/insn-data.def | 1 +
 target/s390x/translate.c   | 1 +
 2 files changed, 2 insertions(+)
target/s390x/insn-data.def +1 -0
index d79ae9e3f1..8dbeaf8c49 100644
@@ -52,6 +52,7 @@
 /* ADD HALFWORD */
     C(0x4a00, AH,      RX_a,  Z,   r1, m2_16s, new, r1_32, add, adds32)
     C(0xe37a, AHY,     RXY_a, LD,  r1, m2_16s, new, r1_32, add, adds32)
+    C(0xe338, AGH,     RXY_a, MIE2,r1, m2_16s, r1, 0, add, adds64)
 /* ADD HALFWORD IMMEDIATE */
     C(0xa70a, AHI,     RI_a,  Z,   r1, i2, new, r1_32, add, adds32)
     C(0xa70b, AGHI,    RI_a,  Z,   r1, i2, r1, 0, add, adds64)
target/s390x/translate.c +1 -0
index a777343821..21d77b7e74 100644
@@ -6098,6 +6098,7 @@ enum DisasInsnEnum {
 #define FAC_AIS         S390_FEAT_ADAPTER_INT_SUPPRESSION
 #define FAC_V           S390_FEAT_VECTOR /* vector facility */
 #define FAC_VE          S390_FEAT_VECTOR_ENH /* vector enhancements facility 1 */
+#define FAC_MIE2        S390_FEAT_MISC_INSTRUCTION_EXT /* miscellaneous-instruction-extensions facility 2 */
 
 static const DisasInsn insn_info[] = {
 #include "insn-data.def"
Signature · 1 line
2.26.2
#11 lore raw

PATCH 4/8 s390x/tcg: Implement MULTIPLY HALFWORD (MGH)

In reply to David Hildenbrand

Just like MULTIPLY HALFWORD IMMEDIATE (MGHI), only the second operand (signed 16 bit) comes from memory.

---
 target/s390x/insn-data.def | 1 +
 1 file changed, 1 insertion(+)
target/s390x/insn-data.def +1 -0
index 2b4ad1530d..455efe73da 100644
@@ -666,6 +666,7 @@
 /* MULTIPLY HALFWORD */
     C(0x4c00, MH,      RX_a,  Z,   r1_o, m2_16s, new, r1_32, mul, 0)
     C(0xe37c, MHY,     RXY_a, GIE, r1_o, m2_16s, new, r1_32, mul, 0)
+    C(0xe33c, MGH,     RXY_a, MIE2,r1_o, m2_16s, r1, 0, mul, 0)
 /* MULTIPLY HALFWORD IMMEDIATE */
     C(0xa70c, MHI,     RI_a,  Z,   r1_o, i2, new, r1_32, mul, 0)
     C(0xa70d, MGHI,    RI_a,  Z,   r1_o, i2, r1, 0, mul, 0)
Signature · 1 line
2.26.2
#13 lore raw

PATCH 5/8 s390x/tcg: Implement BRANCH INDIRECT ON CONDITION (BIC)

In reply to David Hildenbrand

Just like BRANCH ON CONDITION - however the address is read from memory (always 8 bytes are read), we have to wrap the address manually. The address is read using current CPU DAT/address-space controls, just like ordinary data.

---
 target/s390x/insn-data.def | 2 ++
 target/s390x/translate.c   | 5 +++++
 2 files changed, 7 insertions(+)
target/s390x/insn-data.def +2 -0
index 455efe73da..dfb0ec067b 100644
@@ -115,6 +115,8 @@
 /* BRANCH RELATIVE AND SAVE */
     C(0xa705, BRAS,    RI_b,  Z,   0, 0, r1, 0, basi, 0)
     C(0xc005, BRASL,   RIL_b, Z,   0, 0, r1, 0, basi, 0)
+/* BRANCH INDIRECT ON CONDITION */
+    C(0xe347, BIC,     RXY_b, MIE2,0, m2_64, 0, 0, bc, 0)
 /* BRANCH ON CONDITION */
     C(0x0700, BCR,     RR_b,  Z,   0, r2_nz, 0, 0, bc, 0)
     C(0x4700, BC,      RX_b,  Z,   0, a2, 0, 0, bc, 0)
target/s390x/translate.c +5 -0
index b536491892..383edf7419 100644
@@ -1626,6 +1626,11 @@ static DisasJumpType op_bc(DisasContext *s, DisasOps *o)
         return DISAS_NEXT;
     }
 
+    /* For BIC the address came from memory, we need to wrap it again. */
+    if (s->fields.op2 == 0x47) {
+        gen_addi_and_wrap_i64(s, o->in2, o->in2, 0);
+    }
+
     disas_jcc(s, &c, m1);
     return help_branch(s, &c, is_imm, imm, o->in2);
 }
Signature · 1 line
2.26.2
#21 lore raw

PATCH 7/8 s390x/tcg: We support Miscellaneous-Instruction-Extensions Facility 2

In reply to David Hildenbrand

We implement all relevant instructions.

---
 target/s390x/gen-features.c | 1 +
 1 file changed, 1 insertion(+)
target/s390x/gen-features.c +1 -0
index 8ddeebc544..1736b85ab0 100644
@@ -716,6 +716,7 @@ static uint16_t qemu_MAX[] = {
     S390_FEAT_MSA_EXT_5,
     /* features introduced after the z13 */
     S390_FEAT_INSTRUCTION_EXEC_PROT,
+    S390_FEAT_MISC_INSTRUCTION_EXT,
 };
 
 /****** END FEATURE DEFS ******/
Signature · 1 line
2.26.2
#16 lore raw

PATCH 6/8 s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC)

In reply to David Hildenbrand

We need new CC handling, determining the CC based on the intermediate result (64bit for MSC and MSRKC, 128bit for MSGC and MSGRKC).

We want to store out2 ("low") after muls128 to r1, so add "wout_out2_r1".

---
 target/s390x/cc_helper.c   | 34 ++++++++++++++++++++++++++++++++++
 target/s390x/helper.c      |  2 ++
 target/s390x/insn-data.def |  4 ++++
 target/s390x/internal.h    |  2 ++
 target/s390x/translate.c   | 19 +++++++++++++++++++
 5 files changed, 61 insertions(+)
target/s390x/cc_helper.c +34 -0
index 44731e4a85..62074648e6 100644
@@ -417,6 +417,35 @@ static uint32_t cc_calc_vc(uint64_t low, uint64_t high)
     }
 }
 
+static uint32_t cc_calc_muls_32(int64_t res)
+{
+    /* Arithmetic shift with sign extension so we can compare against -1ull. */
+    const uint64_t tmp = res >> 31;
+
+    if (!res) {
+        return 0;
+    } else if (!(!tmp || tmp == -1ull)) {
+        return 3;
+    } else if (res < 0) {
+        return 1;
+    }
+    return 2;
+}
+
+static uint64_t cc_calc_muls_64(int64_t res_high, uint64_t res_low)
+{
+    const uint8_t tmp = res_low >> 63;
+
+    if (!res_high && !res_low) {
+        return 0;
+    } else if (!(!res_high && !tmp) || !(res_high == -1ull && tmp)) {
+        return 3;
+    } else if (res_high < 0) {
+        return 1;
+    }
+    return 2;
+}
+
 static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op,
                                   uint64_t src, uint64_t dst, uint64_t vr)
 {
@@ -484,6 +513,9 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op,
     case CC_OP_COMP_64:
         r =  cc_calc_comp_64(dst);
         break;
+    case CC_OP_MULS_64:
+        r = cc_calc_muls_64(src, dst);
+        break;
 
     case CC_OP_ADD_32:
         r =  cc_calc_add_32(src, dst, vr);
@@ -512,6 +544,8 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op,
     case CC_OP_COMP_32:
         r =  cc_calc_comp_32(dst);
         break;
+    case CC_OP_MULS_32:
+        r = cc_calc_muls_32(dst);
 
     case CC_OP_ICM:
         r =  cc_calc_icm(src, dst);
target/s390x/helper.c +2 -0
index 9257d388ba..b877690845 100644
@@ -430,6 +430,8 @@ const char *cc_name(enum cc_op cc_op)
         [CC_OP_FLOGR]     = "CC_OP_FLOGR",
         [CC_OP_LCBB]      = "CC_OP_LCBB",
         [CC_OP_VC]        = "CC_OP_VC",
+        [CC_OP_MULS_32]   = "CC_OP_MULS_32",
+        [CC_OP_MULS_64]   = "CC_OP_MULS_64",
     };
 
     return cc_names[cc_op];
target/s390x/insn-data.def +4 -0
index dfb0ec067b..bcd424e9ae 100644
@@ -679,11 +679,15 @@
     C(0xe386, MLG,     RXY_a, Z,   r1p1, m2_64, r1_P, 0, mul128, 0)
 /* MULTIPLY SINGLE */
     C(0xb252, MSR,     RRE,   Z,   r1_o, r2_o, new, r1_32, mul, 0)
+    C(0xb9fd, MSRKC,   RRF_a, MIE2,r3_32s, r2_32s, new, r1_32, mul, muls32)
     C(0x7100, MS,      RX_a,  Z,   r1_o, m2_32s, new, r1_32, mul, 0)
     C(0xe351, MSY,     RXY_a, LD,  r1_o, m2_32s, new, r1_32, mul, 0)
+    C(0xe353, MSC,     RXY_a, MIE2,r1_32s, m2_32s, new, r1_32, mul, muls32)
     C(0xb90c, MSGR,    RRE,   Z,   r1_o, r2_o, r1, 0, mul, 0)
+    C(0xb9ed, MSGRKC,  RRF_a, MIE2,r3_o, r2_o, new_P, out2_r1, muls128, muls64)
     C(0xb91c, MSGFR,   RRE,   Z,   r1_o, r2_32s, r1, 0, mul, 0)
     C(0xe30c, MSG,     RXY_a, Z,   r1_o, m2_64, r1, 0, mul, 0)
+    C(0xe383, MSGC,    RXY_a, MIE2,r1_o, m2_64, new_P, out2_r1, muls128, muls64)
     C(0xe31c, MSGF,    RXY_a, Z,   r1_o, m2_32s, r1, 0, mul, 0)
 /* MULTIPLY SINGLE IMMEDIATE */
     C(0xc201, MSFI,    RIL_a, GIE, r1_o, i2, new, r1_32, mul, 0)
target/s390x/internal.h +2 -0
index b1e0ebf67f..c5d32237ea 100644
@@ -175,6 +175,7 @@ enum cc_op {
     CC_OP_SUBB_64,              /* overflow on unsigned sub-borrow (64bit) */
     CC_OP_ABS_64,               /* sign eval on abs (64bit) */
     CC_OP_NABS_64,              /* sign eval on nabs (64bit) */
+    CC_OP_MULS_64,              /* overflow on signed multiply (64bit) */
 
     CC_OP_ADD_32,               /* overflow on add (32bit) */
     CC_OP_ADDU_32,              /* overflow on unsigned add (32bit) */
@@ -184,6 +185,7 @@ enum cc_op {
     CC_OP_SUBB_32,              /* overflow on unsigned sub-borrow (32bit) */
     CC_OP_ABS_32,               /* sign eval on abs (64bit) */
     CC_OP_NABS_32,              /* sign eval on nabs (64bit) */
+    CC_OP_MULS_32,              /* overflow on signed multiply (32bit) */
 
     CC_OP_COMP_32,              /* complement */
     CC_OP_COMP_64,              /* complement */
target/s390x/translate.c +19 -0
index 383edf7419..c90cb37aae 100644
@@ -646,6 +646,7 @@ static void gen_op_calc_cc(DisasContext *s)
     case CC_OP_NZ_F64:
     case CC_OP_FLOGR:
     case CC_OP_LCBB:
+    case CC_OP_MULS_32:
         /* 1 argument */
         gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, dummy, cc_dst, dummy);
         break;
@@ -660,6 +661,7 @@ static void gen_op_calc_cc(DisasContext *s)
     case CC_OP_SLA_64:
     case CC_OP_NZ_F128:
     case CC_OP_VC:
+    case CC_OP_MULS_64:
         /* 2 arguments */
         gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, cc_src, cc_dst, dummy);
         break;
@@ -5294,6 +5296,17 @@ static void cout_tm64(DisasContext *s, DisasOps *o)
     gen_op_update2_cc_i64(s, CC_OP_TM_64, o->in1, o->in2);
 }
 
+static void cout_muls32(DisasContext *s, DisasOps *o)
+{
+    gen_op_update1_cc_i64(s, CC_OP_MULS_32, o->out);
+}
+
+static void cout_muls64(DisasContext *s, DisasOps *o)
+{
+    /* out contains "high" part, out2 contains "low" part of 128 bit result */
+    gen_op_update2_cc_i64(s, CC_OP_MULS_64, o->out, o->out2);
+}
+
 /* ====================================================================== */
 /* The "PREParation" generators.  These initialize the DisasOps.OUT fields
    with the TCG register to which we will write.  Used in combination with
@@ -5349,6 +5362,12 @@ static void wout_r1(DisasContext *s, DisasOps *o)
 }
 #define SPEC_wout_r1 0
 
+static void wout_out2_r1(DisasContext *s, DisasOps *o)
+{
+    store_reg(get_field(s, r1), o->out2);
+}
+#define SPEC_wout_out2_r1 0
+
 static void wout_r1_8(DisasContext *s, DisasOps *o)
 {
     int r1 = get_field(s, r1);
Signature · 1 line
2.26.2
#23 lore raw

PATCH 8/8 s390x/tcg: Implement CIPHER MESSAGE WITH AUTHENTICATION (KMA)

In reply to David Hildenbrand

As with the other crypto functions, we only implement subcode 0 (query) and no actual encryption/decryption. We now implement S390_FEAT_MSA_EXT_8.

---
 target/s390x/gen-features.c | 1 +
 target/s390x/insn-data.def  | 1 +
 target/s390x/translate.c    | 7 +++++++
 3 files changed, 9 insertions(+)
target/s390x/gen-features.c +1 -0
index 1736b85ab0..8de18f7f41 100644
@@ -717,6 +717,7 @@ static uint16_t qemu_MAX[] = {
     /* features introduced after the z13 */
     S390_FEAT_INSTRUCTION_EXEC_PROT,
     S390_FEAT_MISC_INSTRUCTION_EXT,
+    S390_FEAT_MSA_EXT_8,
 };
 
 /****** END FEATURE DEFS ******/
target/s390x/insn-data.def +1 -0
index bcd424e9ae..1ffdc20d59 100644
@@ -979,6 +979,7 @@
     D(0xb92d, KMCTR,   RRF_b, MSA4, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KMCTR)
     D(0xb92e, KM,      RRE,   MSA,  0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KM)
     D(0xb92f, KMC,     RRE,   MSA,  0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KMC)
+    D(0xb929, KMA,     RRF_b, MSA8, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KMA)
     D(0xb93c, PPNO,    RRE,   MSA5, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_PPNO)
     D(0xb93e, KIMD,    RRE,   MSA,  0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KIMD)
     D(0xb93f, KLMD,    RRE,   MSA,  0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KLMD)
target/s390x/translate.c +7 -0
index c90cb37aae..df549e59e9 100644
@@ -2715,6 +2715,12 @@ static DisasJumpType op_msa(DisasContext *s, DisasOps *o)
     TCGv_i32 t_r1, t_r2, t_r3, type;
 
     switch (s->insn->data) {
+    case S390_FEAT_TYPE_KMA:
+        if (r3 == r1 || r3 == r2) {
+            gen_program_exception(s, PGM_SPECIFICATION);
+            return DISAS_NORETURN;
+        }
+        /* FALL THROUGH */
     case S390_FEAT_TYPE_KMCTR:
         if (r3 & 1 || !r3) {
             gen_program_exception(s, PGM_SPECIFICATION);
@@ -6130,6 +6136,7 @@ enum DisasInsnEnum {
 #define FAC_MSA3        S390_FEAT_MSA_EXT_3 /* msa-extension-3 facility */
 #define FAC_MSA4        S390_FEAT_MSA_EXT_4 /* msa-extension-4 facility */
 #define FAC_MSA5        S390_FEAT_MSA_EXT_5 /* msa-extension-5 facility */
+#define FAC_MSA8        S390_FEAT_MSA_EXT_8 /* msa-extension-8 facility */
 #define FAC_ECT         S390_FEAT_EXTRACT_CPU_TIME
 #define FAC_PCI         S390_FEAT_ZPCI /* z/PCI facility */
 #define FAC_AIS         S390_FEAT_ADAPTER_INT_SUPPRESSION
Signature · 1 line
2.26.2
#25 lore raw
In reply to David Hildenbrand
Patchew URL: https://patchew.org/QEMU/[email protected]/

Patchew URL: https://patchew.org/QEMU/[email protected]/

Hi,

This series seems to have some coding style problems. See output below for more information:

Type: series
Message-id: [email protected]
Subject: [PATCH v1 0/8] s390x/tcg: Implement some z14 facilities

=== TEST SCRIPT BEGIN ===
#!/bin/bash
git rev-parse base > /dev/null || exit 0
git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram
./scripts/checkpatch.pl --mailback base..
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
Switched to a new branch 'test'
4a710cb s390x/tcg: Implement CIPHER MESSAGE WITH AUTHENTICATION (KMA) 67c8e40 s390x/tcg: We support Miscellaneous-Instruction-Extensions Facility 2 e867c7b s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC) 2af9b83 s390x/tcg: Implement BRANCH INDIRECT ON CONDITION (BIC) 656cf89 s390x/tcg: Implement MULTIPLY HALFWORD (MGH)
6262d2e s390x/tcg: Implement MULTIPLY (MG, MGRK)
3aba4e2 s390x/tcg: Implement SUBTRACT HALFWORD (SGH)
c69a385 s390x/tcg: Implement ADD HALFWORD (AGH)

=== OUTPUT BEGIN ===
1/8 Checking commit c69a385d838c (s390x/tcg: Implement ADD HALFWORD (AGH))
ERROR: line over 90 characters
#32: FILE: target/s390x/translate.c:6101:
+#define FAC_MIE2 S390_FEAT_MISC_INSTRUCTION_EXT /* miscellaneous-instruction-extensions facility 2 */

total: 1 errors, 0 warnings, 14 lines checked

Patch 1/8 has style problems, please review. If any of these errors are false positives report them to the maintainer, see CHECKPATCH in MAINTAINERS.

2/8 Checking commit 3aba4e25e456 (s390x/tcg: Implement SUBTRACT HALFWORD (SGH))
3/8 Checking commit 6262d2edc76d (s390x/tcg: Implement MULTIPLY (MG, MGRK))
4/8 Checking commit 656cf89d2355 (s390x/tcg: Implement MULTIPLY HALFWORD (MGH))
5/8 Checking commit 2af9b83b61be (s390x/tcg: Implement BRANCH INDIRECT ON CONDITION (BIC)) 6/8 Checking commit e867c7b42400 (s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC)) 7/8 Checking commit 67c8e4024fd0 (s390x/tcg: We support Miscellaneous-Instruction-Extensions Facility 2) 8/8 Checking commit 4a710cb58f52 (s390x/tcg: Implement CIPHER MESSAGE WITH AUTHENTICATION (KMA)) === OUTPUT END ===

Test command exited with code: 1

The full log is available at
http://patchew.org/logs/[email protected]/testing.checkpatch/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]

#18 lore raw

Re: [PATCH v1 6/8] s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC)

In reply to Richard Henderson
Quoted 11 lines · Richard Henderson

target/s390x/cc_helper.c static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, · line 550

Quoted 8 lines · David Hildenbrand
target/s390x/cc_helper.c +2 -0
@@ -512,6 +544,8 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op,
     case CC_OP_COMP_32:
         r =  cc_calc_comp_32(dst);
         break;
+    case CC_OP_MULS_32:
+        r = cc_calc_muls_32(dst);
 
     case CC_OP_ICM:

Missing break.

Argh, thanks!

Signature · 3 lines
Thanks,

David / dhildenb
#17 lore raw

Re: [PATCH v1 6/8] s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC)

In reply to David Hildenbrand

target/s390x/cc_helper.c static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, · line 550

Quoted 8 lines · David Hildenbrand
target/s390x/cc_helper.c +2 -0
@@ -512,6 +544,8 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op,
     case CC_OP_COMP_32:
         r =  cc_calc_comp_32(dst);
         break;
+    case CC_OP_MULS_32:
+        r = cc_calc_muls_32(dst);
 
     case CC_OP_ICM:

Missing break.

r~

Reviewed-by on v1 1/8
#2 lore raw

Re: [PATCH v1 1/8] s390x/tcg: Implement ADD HALFWORD (AGH)

In reply to David Hildenbrand
Quoted 26 lines · David Hildenbrand

Easy, just like ADD HALFWORD IMMEDIATE (AGHI).

---
 target/s390x/insn-data.def | 1 +
 target/s390x/translate.c   | 1 +
 2 files changed, 2 insertions(+)
target/s390x/insn-data.def +1 -0
index d79ae9e3f1..8dbeaf8c49 100644
@@ -52,6 +52,7 @@
 /* ADD HALFWORD */
     C(0x4a00, AH,      RX_a,  Z,   r1, m2_16s, new, r1_32, add, adds32)
     C(0xe37a, AHY,     RXY_a, LD,  r1, m2_16s, new, r1_32, add, adds32)
+    C(0xe338, AGH,     RXY_a, MIE2,r1, m2_16s, r1, 0, add, adds64)
 /* ADD HALFWORD IMMEDIATE */
     C(0xa70a, AHI,     RI_a,  Z,   r1, i2, new, r1_32, add, adds32)
     C(0xa70b, AGHI,    RI_a,  Z,   r1, i2, r1, 0, add, adds64)
target/s390x/translate.c +0 -0
index a777343821..21d77b7e74 100644

target/s390x/translate.c enum DisasInsnEnum { · line 6104

Quoted 7 lines
target/s390x/translate.c +1 -0
@@ -6098,6 +6098,7 @@ enum DisasInsnEnum {
 #define FAC_AIS         S390_FEAT_ADAPTER_INT_SUPPRESSION
 #define FAC_V           S390_FEAT_VECTOR /* vector facility */
 #define FAC_VE          S390_FEAT_VECTOR_ENH /* vector enhancements facility 1 */
+#define FAC_MIE2        S390_FEAT_MISC_INSTRUCTION_EXT /* miscellaneous-instruction-extensions facility 2 */
 
 static const DisasInsn insn_info[] = {
 #include "insn-data.def"

Looks right to me.

Reviewed-by on v1 2/8
#7 lore raw

Re: [PATCH v1 2/8] s390x/tcg: Implement SUBTRACT HALFWORD (SGH)

In reply to David Hildenbrand
Quoted 13 lines · David Hildenbrand

Easy to wire up.

---
 target/s390x/insn-data.def | 1 +
 1 file changed, 1 insertion(+)
target/s390x/insn-data.def +0 -0
index 8dbeaf8c49..e851e9df5e 100644

target/s390x/insn-data.def line 889

Quoted 7 lines
target/s390x/insn-data.def +1 -0
@@ -883,6 +883,7 @@
 /* SUBTRACT HALFWORD */
     C(0x4b00, SH,      RX_a,  Z,   r1, m2_16s, new, r1_32, sub, subs32)
     C(0xe37b, SHY,     RXY_a, LD,  r1, m2_16s, new, r1_32, sub, subs32)
+    C(0xe339, SGH,     RXY_a, MIE2,r1, m2_16s, r1, 0, sub, subs64)
 /* SUBTRACT HIGH */
     C(0xb9c9, SHHHR,   RRF_a, HW,  r2_sr32, r3_sr32, new, r1_32h, sub, subs32)
     C(0xb9d9, SHHLR,   RRF_a, HW,  r2_sr32, r3, new, r1_32h, sub, subs32)
#3 lore raw

Re: [PATCH v1 1/8] s390x/tcg: Implement ADD HALFWORD (AGH)

In reply to David Hildenbrand
Quoted 25 lines · David Hildenbrand

Easy, just like ADD HALFWORD IMMEDIATE (AGHI).

---
 target/s390x/insn-data.def | 1 +
 target/s390x/translate.c   | 1 +
 2 files changed, 2 insertions(+)
target/s390x/insn-data.def +1 -0
index d79ae9e3f1..8dbeaf8c49 100644
@@ -52,6 +52,7 @@
 /* ADD HALFWORD */
     C(0x4a00, AH,      RX_a,  Z,   r1, m2_16s, new, r1_32, add, adds32)
     C(0xe37a, AHY,     RXY_a, LD,  r1, m2_16s, new, r1_32, add, adds32)
+    C(0xe338, AGH,     RXY_a, MIE2,r1, m2_16s, r1, 0, add, adds64)
 /* ADD HALFWORD IMMEDIATE */
     C(0xa70a, AHI,     RI_a,  Z,   r1, i2, new, r1_32, add, adds32)
     C(0xa70b, AGHI,    RI_a,  Z,   r1, i2, r1, 0, add, adds64)
target/s390x/translate.c +0 -0
index a777343821..21d77b7e74 100644

target/s390x/translate.c enum DisasInsnEnum { · line 6101

Quoted 4 lines
target/s390x/translate.c +1 -0
@@ -6098,6 +6098,7 @@ enum DisasInsnEnum {
 #define FAC_AIS         S390_FEAT_ADAPTER_INT_SUPPRESSION
 #define FAC_V           S390_FEAT_VECTOR /* vector facility */
 #define FAC_VE          S390_FEAT_VECTOR_ENH /* vector enhancements facility 1 */
+#define FAC_MIE2        S390_FEAT_MISC_INSTRUCTION_EXT /* miscellaneous-instruction-extensions facility 2 */

Maybe you should use "S390_FEAT_MISC_INSTRUCTION_EXT2" (i.e. with 2 at the end) to avoid that it gets confused with the first MIE ?

Thomas

#4 lore raw

Re: [PATCH v1 1/8] s390x/tcg: Implement ADD HALFWORD (AGH)

Quoted 25 lines · Thomas Huth
Quoted 24 lines · David Hildenbrand

Easy, just like ADD HALFWORD IMMEDIATE (AGHI).

---
 target/s390x/insn-data.def | 1 +
 target/s390x/translate.c   | 1 +
 2 files changed, 2 insertions(+)
target/s390x/insn-data.def +1 -0
index d79ae9e3f1..8dbeaf8c49 100644
@@ -52,6 +52,7 @@
 /* ADD HALFWORD */
     C(0x4a00, AH,      RX_a,  Z,   r1, m2_16s, new, r1_32, add, adds32)
     C(0xe37a, AHY,     RXY_a, LD,  r1, m2_16s, new, r1_32, add, adds32)
+    C(0xe338, AGH,     RXY_a, MIE2,r1, m2_16s, r1, 0, add, adds64)
 /* ADD HALFWORD IMMEDIATE */
     C(0xa70a, AHI,     RI_a,  Z,   r1, i2, new, r1_32, add, adds32)
     C(0xa70b, AGHI,    RI_a,  Z,   r1, i2, r1, 0, add, adds64)
target/s390x/translate.c +0 -0
index a777343821..21d77b7e74 100644
Quoted 8 lines

target/s390x/translate.c enum DisasInsnEnum { · line 6101

Quoted 5 lines
target/s390x/translate.c +1 -0
@@ -6098,6 +6098,7 @@ enum DisasInsnEnum {
 #define FAC_AIS         S390_FEAT_ADAPTER_INT_SUPPRESSION
 #define FAC_V           S390_FEAT_VECTOR /* vector facility */
 #define FAC_VE          S390_FEAT_VECTOR_ENH /* vector enhancements facility 1 */
+#define FAC_MIE2        S390_FEAT_MISC_INSTRUCTION_EXT /* miscellaneous-instruction-extensions facility 2 */

Maybe you should use "S390_FEAT_MISC_INSTRUCTION_EXT2" (i.e. with 2 at the end) to avoid that it gets confused with the first MIE ?

Yes, renaming S390_FEAT_MISC_INSTRUCTION_EXT ->
S390_FEAT_MISC_INSTRUCTION_EXT2 makes sense.

Quoted 3 lines

Thomas

Signature · 3 lines
Thanks,

David / dhildenb
Reviewed-by on v1 3/8
#10 lore raw

Re: [PATCH v1 3/8] s390x/tcg: Implement MULTIPLY (MG, MGRK)

In reply to David Hildenbrand
Quoted 8 lines · David Hildenbrand

Multiply two signed 64bit values and store the 128bit result in r1 (0-63) and r1 + 1 (64-127).

---
 target/s390x/insn-data.def |  2 ++
 target/s390x/translate.c   | 13 +++++++++++++
 2 files changed, 15 insertions(+)

r~

#14 lore raw

Re: [PATCH v1 5/8] s390x/tcg: Implement BRANCH INDIRECT ON CONDITION (BIC)

In reply to David Hildenbrand
Quoted 10 lines · David Hildenbrand
target/s390x/insn-data.def +2 -0
+/* BRANCH INDIRECT ON CONDITION */
+    C(0xe347, BIC,     RXY_b, MIE2,0, m2_64, 0, 0, bc, 0)
 /* BRANCH ON CONDITION */
     C(0x0700, BCR,     RR_b,  Z,   0, r2_nz, 0, 0, bc, 0)
     C(0x4700, BC,      RX_b,  Z,   0, a2, 0, 0, bc, 0)
target/s390x/translate.c +0 -0
index b536491892..383edf7419 100644

target/s390x/translate.c static DisasJumpType op_bc(DisasContext *s, DisasOps *o) · line 1632

Quoted 7 lines
target/s390x/translate.c +4 -0
@@ -1626,6 +1626,11 @@ static DisasJumpType op_bc(DisasContext *s, DisasOps *o)
         return DISAS_NEXT;
     }
 
+    /* For BIC the address came from memory, we need to wrap it again. */
+    if (s->fields.op2 == 0x47) {
+        gen_addi_and_wrap_i64(s, o->in2, o->in2, 0);
+    }

I'm not keen on this sort of per-opcode checks.

I'd prefer to add an in2_m2_64w() helper that performs the load and then wraps.

r~

#19 lore raw

Re: [PATCH v1 6/8] s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC)

In reply to David Hildenbrand

target/s390x/cc_helper.c static uint32_t cc_calc_vc(uint64_t low, uint64_t high) · line 427

Quoted 8 lines · David Hildenbrand
target/s390x/cc_helper.c +8 -0
+static uint32_t cc_calc_muls_32(int64_t res)
+{
+    /* Arithmetic shift with sign extension so we can compare against -1ull. */
+    const uint64_t tmp = res >> 31;
+
+    if (!res) {
+        return 0;
+    } else if (!(!tmp || tmp == -1ull)) {

Comparing signed vs unsigned. Use -1 without suffix.

target/s390x/cc_helper.c static uint32_t cc_calc_vc(uint64_t low, uint64_t high) · line 441

Quoted 7 lines
target/s390x/cc_helper.c +7 -0
+static uint64_t cc_calc_muls_64(int64_t res_high, uint64_t res_low)
+{
+    const uint8_t tmp = res_low >> 63;
+
+    if (!res_high && !res_low) {
+        return 0;
+    } else if (!(!res_high && !tmp) || !(res_high == -1ull && tmp)) {

This simplifies to res_high + tmp != 0.

Probably better to keep tmp as uint64_t; otherwise we're likely to have an unnecessary zero-extension from uint8_t to uint64_t.
Or, drop 'tmp' altogether and use

if (res_high + (res_low >> 63) != 0)

Otherwise, looks good.

r~

Reviewed-by on v1 8/8
#24 lore raw

Re: [PATCH v1 8/8] s390x/tcg: Implement CIPHER MESSAGE WITH AUTHENTICATION (KMA)

In reply to David Hildenbrand
Quoted 9 lines · David Hildenbrand

As with the other crypto functions, we only implement subcode 0 (query) and no actual encryption/decryption. We now implement S390_FEAT_MSA_EXT_8.

---
 target/s390x/gen-features.c | 1 +
 target/s390x/insn-data.def  | 1 +
 target/s390x/translate.c    | 7 +++++++
 3 files changed, 9 insertions(+)

r~

#26 lore raw
In reply to David Hildenbrand
Quoted 2 lines · David Hildenbrand

With this patchset, I can boot a kernel compiled for z14 (requiring Miscellaneous-Instruction-Extensions Facility 2) (using -cpu max).

Excellent.

Looks like it wouldn't take too much additional work for z15, if you're of a mind while we're at it.

Message-Security-Assist 9:
KDSA to ignore, like KMA.

Miscellaneous-Instruction-Extensions 3:
A bunch of logical operations, all of which already have tcg primitives.

SELECT, which can re-use op_loc with different inputs.

MOVE RIGHT TO LEFT, which can't reuse do_helper_mvc directly, but could be a copy with trivial changes.

Trivial changes to op_popcnt.

Vector-Enhancements 2:
Mostly load/store byte-reversed.
New vector shift insns.
FP/Int conversions.
Vector string search.

r~

#15 lore raw

Re: [PATCH v1 5/8] s390x/tcg: Implement BRANCH INDIRECT ON CONDITION (BIC)

In reply to Richard Henderson
Quoted 14 lines · Richard Henderson
Quoted 13 lines · David Hildenbrand
target/s390x/insn-data.def +2 -0
+/* BRANCH INDIRECT ON CONDITION */
+    C(0xe347, BIC,     RXY_b, MIE2,0, m2_64, 0, 0, bc, 0)
 /* BRANCH ON CONDITION */
     C(0x0700, BCR,     RR_b,  Z,   0, r2_nz, 0, 0, bc, 0)
     C(0x4700, BC,      RX_b,  Z,   0, a2, 0, 0, bc, 0)
target/s390x/translate.c +0 -0
index b536491892..383edf7419 100644
@@ -1626,6 +1626,11 @@ static DisasJumpType op_bc(DisasContext *s, DisasOps *o)
         return DISAS_NEXT;
     }
 
Quoted 8 lines

target/s390x/translate.c static DisasJumpType op_bc(DisasContext *s, DisasOps *o) · line 1632

Quoted 4 lines
target/s390x/translate.c +4 -0
+    /* For BIC the address came from memory, we need to wrap it again. */
+    if (s->fields.op2 == 0x47) {
+        gen_addi_and_wrap_i64(s, o->in2, o->in2, 0);
+    }

I'm not keen on this sort of per-opcode checks.

I'd prefer to add an in2_m2_64w() helper that performs the load and then wraps.

Makes sense, thanks!

Signature · 3 lines
Thanks,

David / dhildenb
#27 lore raw
In reply to Richard Henderson
Quoted 8 lines · Richard Henderson
Quoted 2 lines · David Hildenbrand

With this patchset, I can boot a kernel compiled for z14 (requiring Miscellaneous-Instruction-Extensions Facility 2) (using -cpu max).

Excellent.

Looks like it wouldn't take too much additional work for z15, if you're of a mind while we're at it.

Yes, on my todo list (however might take some time until I get to it). I'm focusing on z14 right now because next-gen distributions will most probably require at least z14 hw, so we're prepared for that.

I'm still wrapping my head around how to best implement the crazy VECTOR FP MINIMUM / MAXIMUM stuff in the least ugly way (expressing all cases via if/else/switch is just horribly error-prone).

Signature · 3 lines
Thanks,

David / dhildenb
#20 lore raw

Re: [PATCH v1 6/8] s390x/tcg: Implement MULTIPLY SINGLE (MSC, MSGC, MSGRKC, MSRKC)

In reply to Richard Henderson
Quoted 11 lines · Richard Henderson

target/s390x/cc_helper.c static uint32_t cc_calc_vc(uint64_t low, uint64_t high) · line 427

Quoted 8 lines · David Hildenbrand
target/s390x/cc_helper.c +8 -0
+static uint32_t cc_calc_muls_32(int64_t res)
+{
+    /* Arithmetic shift with sign extension so we can compare against -1ull. */
+    const uint64_t tmp = res >> 31;
+
+    if (!res) {
+        return 0;
+    } else if (!(!tmp || tmp == -1ull)) {

Comparing signed vs unsigned. Use -1 without suffix.

tmp is also uint64_t - but I can change that to int64_t.

(and condense to "tmp && tmp != -1")

Quoted 10 lines

target/s390x/cc_helper.c static uint32_t cc_calc_vc(uint64_t low, uint64_t high) · line 441

Quoted 8 lines
target/s390x/cc_helper.c +7 -0
+static uint64_t cc_calc_muls_64(int64_t res_high, uint64_t res_low)
+{
+    const uint8_t tmp = res_low >> 63;
+
+    if (!res_high && !res_low) {
+        return 0;
+    } else if (!(!res_high && !tmp) || !(res_high == -1ull && tmp)) {

This simplifies to res_high + tmp != 0.

Yeah, after messing up one time I decided to phrase it just as stated in the PoP.

Quoted 6 lines

Probably better to keep tmp as uint64_t; otherwise we're likely to have an unnecessary zero-extension from uint8_t to uint64_t.
Or, drop 'tmp' altogether and use

if (res_high + (res_low >> 63) != 0)

Thanks, I'll go with that.

Quoted 6 lines

Otherwise, looks good.

r~

Signature · 3 lines
Thanks,

David / dhildenb