Skip to contentSkip to conversation

arm: add DCZID_EL0 to idregs array

PATCH Superseded Cornelia Huck 3 participants 7 messages Nov 19, 2025 – Nov 26, 2025

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm: add DCZID_EL0 to idregs array

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize via a wrapper working on DCZID_EL.BS.

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

---

This is another followup on the ID register rework. The last missing set of registers are CCSIDR* and friends, then we should be able to switch to autogenerated registers (probably with a different script than the last attempt.)

---
 target/arm/cpu-sysregs.h.inc   |  1 +
 target/arm/cpu.c               |  2 +-
 target/arm/cpu.h               | 15 +++++++++++++--
 target/arm/cpu64.c             |  4 ++--
 target/arm/helper.c            |  3 ++-
 target/arm/kvm.c               |  7 +++++++
 target/arm/tcg/cpu64.c         | 22 +++++++++++-----------
 target/arm/tcg/helper-a64.c    |  2 +-
 target/arm/tcg/mte_helper.c    |  4 ++--
 target/arm/tcg/translate-a64.c |  2 +-
 10 files changed, 41 insertions(+), 21 deletions(-)
target/arm/cpu-sysregs.h.inc +1 -0
index 2bb2861c6234..7f3aa8b991aa 100644
@@ -39,3 +39,4 @@ DEF(ID_MMFR5_EL1, 3, 0, 0, 3, 6)
 DEF(CLIDR_EL1, 3, 1, 0, 0, 1)
 DEF(ID_AA64ZFR0_EL1, 3, 0, 0, 4, 4)
 DEF(CTR_EL0, 3, 3, 0, 0, 1)
+DEF(DCZID_EL0, 3, 3, 0, 0, 7)
target/arm/cpu.c +1 -1
index 39292fb9bc1f..ad99233839c0 100644
@@ -2184,7 +2184,7 @@ static void arm_cpu_realizefn(DeviceState *dev, Error **errp)
 #endif
 
     if (tcg_enabled()) {
-        int dcz_blocklen = 4 << cpu->dcz_blocksize;
+        int dcz_blocklen = 4 << get_dcz_blocksize(cpu);
 
         /*
          * We only support DCZ blocklen that fits on one page.
target/arm/cpu.h +13 -2
index 39f2b2e54deb..e3fa6ebda3f2 100644
@@ -1111,8 +1111,6 @@ struct ArchCPU {
     bool prop_pauth_qarma5;
     bool prop_lpa2;
 
-    /* DCZ blocksize, in log_2(words), ie low 4 bits of DCZID_EL0 */
-    uint8_t dcz_blocksize;
     /* GM blocksize, in log_2(words), ie low 4 bits of GMID_EL0 */
     uint8_t gm_blocksize;
 
@@ -1178,6 +1176,19 @@ struct ARMCPUClass {
     ResettablePhases parent_phases;
 };
 
+static inline uint8_t get_dcz_blocksize(ARMCPU *cpu)
+{
+    return cpu->isar.idregs[DCZID_EL0_IDX] & 0xf;
+}
+
+static inline void set_dcz_blocksize(ARMCPU *cpu, uint8_t bs)
+{
+    uint64_t dczid = cpu->isar.idregs[DCZID_EL0_IDX];
+
+    /* keep dzp unchanged */
+    cpu->isar.idregs[DCZID_EL0_IDX] = (dczid & ~0xf) | bs;
+}
+
 /* Callback functions for the generic timer's timers. */
 void arm_gt_ptimer_cb(void *opaque);
 void arm_gt_vtimer_cb(void *opaque);
target/arm/cpu64.c +2 -2
index ae84d8e42050..23924f251020 100644
@@ -689,7 +689,7 @@ static void aarch64_a57_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 3, 64, 48 * KiB, 2);
     /* 2048KB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 64, 2 * MiB, 7);
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
@@ -751,7 +751,7 @@ static void aarch64_a53_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 1, 64, 32 * KiB, 2);
     /* 1024KB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 64, 1 * MiB, 7);
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
target/arm/helper.c +2 -1
index 27ebc6f29b82..8dfeaff25350 100644
@@ -3324,7 +3324,8 @@ static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri)
     if (aa64_zva_access(env, NULL, false) == CP_ACCESS_OK) {
         dzp_bit = 0;
     }
-    return cpu->dcz_blocksize | dzp_bit;
+
+    return cpu->isar.idregs[DCZID_EL0_IDX] | dzp_bit;
 }
 
 static CPAccessResult sp_el0_access(CPUARMState *env, const ARMCPRegInfo *ri,
target/arm/kvm.c +7 -0
index 0d57081e69fb..5d65f64addc6 100644
@@ -2020,6 +2020,13 @@ int kvm_arch_init_vcpu(CPUState *cs)
     }
     cpu->mp_affinity = mpidr & ARM64_AFFINITY_MASK;
 
+    /*
+     * We currently do not need this, except for tcg. Should KVM gain support
+     * for accessing DCZID_EL0 via ONE_REG, we'll overwrite this below. Just
+     * set a dummy value that corresponds to the minimum value for FEAT_MTE2.
+     */
+    set_dcz_blocksize(cpu, 2);
+
     return kvm_arm_init_cpreg_list(cpu);
 }
 
target/arm/tcg/cpu64.c +11 -11
index 6871956382f7..d86fc38e7157 100644
@@ -72,7 +72,7 @@ static void aarch64_a35_initfn(Object *obj)
     SET_IDREG(isar, ID_AA64MMFR0, 0x00101122);
     SET_IDREG(isar, ID_AA64MMFR1, 0);
     SET_IDREG(isar, CLIDR, 0x0a200023);
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
 
     /* From B2.4 AArch64 Virtual Memory control registers */
     cpu->reset_sctlr = 0x00c50838;
@@ -219,7 +219,7 @@ static void aarch64_a55_initfn(Object *obj)
     /* Ordered by B2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x84448004; /* L1Ip = VIPT */
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000010305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0000100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000000100001ull);
@@ -325,7 +325,7 @@ static void aarch64_a72_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 3, 64, 48 * KiB, 2);
     /* 1MB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 64, 1 * MiB, 7);
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
@@ -352,7 +352,7 @@ static void aarch64_a76_initfn(Object *obj)
     /* Ordered by B2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x8444C004;
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000010305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0000100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000000100001ull);
@@ -424,7 +424,7 @@ static void aarch64_a78ae_initfn(Object *obj)
     /* Ordered by 3.2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x9444c004;
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000110305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0010100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000001200031ull);
@@ -517,7 +517,7 @@ static void aarch64_a64fx_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 4, 256, 64 * KiB, 2);
     /* 8MB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 256, 8 * MiB, 7);
-    cpu->dcz_blocksize = 6; /* 256 bytes */
+    set_dcz_blocksize(cpu, 6); /* 256 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
@@ -673,7 +673,7 @@ static void aarch64_neoverse_n1_initfn(Object *obj)
     /* Ordered by B2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x8444c004;
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000110305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0000100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000000100001ull);
@@ -749,7 +749,7 @@ static void aarch64_neoverse_v1_initfn(Object *obj)
     /* Ordered by 3.2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0xb444c004; /* With DIC and IDC set */
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64AFR0, 0x00000000);
     SET_IDREG(isar, ID_AA64AFR1, 0x00000000);
     SET_IDREG(isar, ID_AA64DFR0, 0x000001f210305519ull);
@@ -1011,7 +1011,7 @@ static void aarch64_a710_initfn(Object *obj)
     SET_IDREG(isar, CLIDR, 0x0000001482000023ull);
     cpu->gm_blocksize      = 4;
     cpu->ctr               = 0x000000049444c004ull;
-    cpu->dcz_blocksize     = 4;
+    set_dcz_blocksize(cpu, 4);
     /* TODO FEAT_MPAM: mpamidr_el1 = 0x0000_0001_0006_003f */
 
     /* Section B.5.2: PMCR_EL0 */
@@ -1113,7 +1113,7 @@ static void aarch64_neoverse_n2_initfn(Object *obj)
     SET_IDREG(isar, CLIDR, 0x0000001482000023ull);
     cpu->gm_blocksize      = 4;
     cpu->ctr               = 0x00000004b444c004ull;
-    cpu->dcz_blocksize     = 4;
+    set_dcz_blocksize(cpu, 4);
     /* TODO FEAT_MPAM: mpamidr_el1 = 0x0000_0001_001e_01ff */
 
     /* Section B.7.2: PMCR_EL0 */
@@ -1377,7 +1377,7 @@ void aarch64_max_tcg_initfn(Object *obj)
      * blocksize since we don't have to follow what the hardware does.
      */
     cpu->ctr = 0x80038003; /* 32 byte I and D cacheline size, VIPT icache */
-    cpu->dcz_blocksize = 7; /*  512 bytes */
+    set_dcz_blocksize(cpu, 7); /*  512 bytes */
 #endif
     cpu->gm_blocksize = 6;  /*  256 bytes */
 
target/arm/tcg/helper-a64.c +1 -1
index ba1d775d818e..045a00b43373 100644
@@ -792,7 +792,7 @@ void HELPER(dc_zva)(CPUARMState *env, uint64_t vaddr_in)
      * (which matches the usual QEMU behaviour of not implementing either
      * alignment faults or any memory attribute handling).
      */
-    int blocklen = 4 << env_archcpu(env)->dcz_blocksize;
+    int blocklen = 4 << get_dcz_blocksize(env_archcpu(env));
     uint64_t vaddr = vaddr_in & ~(blocklen - 1);
     int mmu_idx = arm_env_mmu_index(env);
     void *mem;
target/arm/tcg/mte_helper.c +2 -2
index bb48fe359b8c..20f8351d8676 100644
@@ -545,7 +545,7 @@ void HELPER(stzgm_tags)(CPUARMState *env, uint64_t ptr, uint64_t val)
      * i.e. 32 bytes, which is an unreasonably small dcz anyway,
      * to make sure that we can access one complete tag byte here.
      */
-    log2_dcz_bytes = env_archcpu(env)->dcz_blocksize + 2;
+    log2_dcz_bytes = get_dcz_blocksize(env_archcpu(env)) + 2;
     log2_tag_bytes = log2_dcz_bytes - (LOG2_TAG_GRANULE + 1);
     dcz_bytes = (intptr_t)1 << log2_dcz_bytes;
     tag_bytes = (intptr_t)1 << log2_tag_bytes;
@@ -945,7 +945,7 @@ uint64_t HELPER(mte_check_zva)(CPUARMState *env, uint32_t desc, uint64_t ptr)
      * i.e. 32 bytes, which is an unreasonably small dcz anyway, to make
      * sure that we can access one complete tag byte here.
      */
-    log2_dcz_bytes = env_archcpu(env)->dcz_blocksize + 2;
+    log2_dcz_bytes = get_dcz_blocksize(env_archcpu(env)) + 2;
     log2_tag_bytes = log2_dcz_bytes - (LOG2_TAG_GRANULE + 1);
     dcz_bytes = (intptr_t)1 << log2_dcz_bytes;
     tag_bytes = (intptr_t)1 << log2_tag_bytes;
target/arm/tcg/translate-a64.c +1 -1
index 08b21d7dbfa6..c72611e55e45 100644
@@ -10712,7 +10712,7 @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase,
     dc->vec_stride = 0;
     dc->cp_regs = arm_cpu->cp_regs;
     dc->features = env->features;
-    dc->dcz_blocksize = arm_cpu->dcz_blocksize;
+    dc->dcz_blocksize = get_dcz_blocksize(arm_cpu);
     dc->gm_blocksize = arm_cpu->gm_blocksize;
 
 #ifdef CONFIG_USER_ONLY
Signature · 1 line
2.51.1
#1 lore raw

Hi Connie,

Quoted 2 lines · Cornelia Huck

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize

what do you mean by EL specific part?

you may simply say that at the moment only bs field is handled as part of cpu.dcz_blocklen
so you replace all the users of that field by accessors to the isar.idreg[] storage

Quoted 62 lines

via a wrapper working on DCZID_EL.BS.

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

---

This is another followup on the ID register rework. The last missing set of registers are CCSIDR* and friends, then we should be able to switch to autogenerated registers (probably with a different script than the last attempt.)

---
 target/arm/cpu-sysregs.h.inc   |  1 +
 target/arm/cpu.c               |  2 +-
 target/arm/cpu.h               | 15 +++++++++++++--
 target/arm/cpu64.c             |  4 ++--
 target/arm/helper.c            |  3 ++-
 target/arm/kvm.c               |  7 +++++++
 target/arm/tcg/cpu64.c         | 22 +++++++++++-----------
 target/arm/tcg/helper-a64.c    |  2 +-
 target/arm/tcg/mte_helper.c    |  4 ++--
 target/arm/tcg/translate-a64.c |  2 +-
 10 files changed, 41 insertions(+), 21 deletions(-)
target/arm/cpu-sysregs.h.inc +1 -0
index 2bb2861c6234..7f3aa8b991aa 100644
@@ -39,3 +39,4 @@ DEF(ID_MMFR5_EL1, 3, 0, 0, 3, 6)
 DEF(CLIDR_EL1, 3, 1, 0, 0, 1)
 DEF(ID_AA64ZFR0_EL1, 3, 0, 0, 4, 4)
 DEF(CTR_EL0, 3, 3, 0, 0, 1)
+DEF(DCZID_EL0, 3, 3, 0, 0, 7)
target/arm/cpu.c +1 -1
index 39292fb9bc1f..ad99233839c0 100644
@@ -2184,7 +2184,7 @@ static void arm_cpu_realizefn(DeviceState *dev, Error **errp)
 #endif
 
     if (tcg_enabled()) {
-        int dcz_blocklen = 4 << cpu->dcz_blocksize;
+        int dcz_blocklen = 4 << get_dcz_blocksize(cpu);
 
         /*
          * We only support DCZ blocklen that fits on one page.
target/arm/cpu.h +0 -2
index 39f2b2e54deb..e3fa6ebda3f2 100644
@@ -1111,8 +1111,6 @@ struct ArchCPU {
     bool prop_pauth_qarma5;
     bool prop_lpa2;
 
-    /* DCZ blocksize, in log_2(words), ie low 4 bits of DCZID_EL0 */
-    uint8_t dcz_blocksize;
     /* GM blocksize, in log_2(words), ie low 4 bits of GMID_EL0 */
     uint8_t gm_blocksize;
 

target/arm/cpu.h struct ARMCPUClass { · line 1179

Quoted 4 lines
target/arm/cpu.h +1 -0
@@ -1178,6 +1176,19 @@ struct ARMCPUClass {
     ResettablePhases parent_phases;
 };
 
+static inline uint8_t get_dcz_blocksize(ARMCPU *cpu)

While at it I would replace dcz_blocksize by dczid_el0_bs to be more explicit

target/arm/cpu.h struct ARMCPUClass { · line 1181

Quoted 2 lines
target/arm/cpu.h +2 -0
+{
+    return cpu->isar.idregs[DCZID_EL0_IDX] & 0xf;

extract64?

target/arm/cpu.h struct ARMCPUClass { · line 1189

Quoted 8 lines
target/arm/cpu.h +8 -0
+}
+
+static inline void set_dcz_blocksize(ARMCPU *cpu, uint8_t bs)
+{
+    uint64_t dczid = cpu->isar.idregs[DCZID_EL0_IDX];
+
+    /* keep dzp unchanged */
+    cpu->isar.idregs[DCZID_EL0_IDX] = (dczid & ~0xf) | bs;

deposit64?

target/arm/helper.c static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri)

Quoted 32 lines
target/arm/cpu.h +2 -0
+}
+
 /* Callback functions for the generic timer's timers. */
 void arm_gt_ptimer_cb(void *opaque);
 void arm_gt_vtimer_cb(void *opaque);
target/arm/cpu64.c +2 -2
index ae84d8e42050..23924f251020 100644
@@ -689,7 +689,7 @@ static void aarch64_a57_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 3, 64, 48 * KiB, 2);
     /* 2048KB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 64, 2 * MiB, 7);
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
@@ -751,7 +751,7 @@ static void aarch64_a53_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 1, 64, 32 * KiB, 2);
     /* 1024KB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 64, 1 * MiB, 7);
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
target/arm/helper.c +0 -0
index 27ebc6f29b82..8dfeaff25350 100644
@@ -3324,7 +3324,8 @@ static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri)

the name of the function seems to indicate you read the whole DCZID but I see the DZP bit is tweaked below. Do you get why we can't return the raw id reg?

target/arm/helper.c static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri) · line 3327

Quoted 5 lines
target/arm/helper.c +1 -1
     if (aa64_zva_access(env, NULL, false) == CP_ACCESS_OK) {
         dzp_bit = 0;
     }
-    return cpu->dcz_blocksize | dzp_bit;
+

nit spurious NL

Quoted 9 lines
target/arm/helper.c +1 -0
+    return cpu->isar.idregs[DCZID_EL0_IDX] | dzp_bit;
 }
 
 static CPAccessResult sp_el0_access(CPUARMState *env, const ARMCPRegInfo *ri,
target/arm/kvm.c +0 -0
index 0d57081e69fb..5d65f64addc6 100644

target/arm/kvm.c int kvm_arch_init_vcpu(CPUState *cs) · line 2028

Quoted 9 lines
target/arm/kvm.c +6 -0
@@ -2020,6 +2020,13 @@ int kvm_arch_init_vcpu(CPUState *cs)
     }
     cpu->mp_affinity = mpidr & ARM64_AFFINITY_MASK;
 
+    /*
+     * We currently do not need this, except for tcg. Should KVM gain support
+     * for accessing DCZID_EL0 via ONE_REG, we'll overwrite this below. Just
+     * set a dummy value that corresponds to the minimum value for FEAT_MTE2.
+     */
+    set_dcz_blocksize(cpu, 2);

it is not clear to me why we need that?

Quoted 147 lines
target/arm/kvm.c +1 -0
+
     return kvm_arm_init_cpreg_list(cpu);
 }
 
target/arm/tcg/cpu64.c +11 -11
index 6871956382f7..d86fc38e7157 100644
@@ -72,7 +72,7 @@ static void aarch64_a35_initfn(Object *obj)
     SET_IDREG(isar, ID_AA64MMFR0, 0x00101122);
     SET_IDREG(isar, ID_AA64MMFR1, 0);
     SET_IDREG(isar, CLIDR, 0x0a200023);
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
 
     /* From B2.4 AArch64 Virtual Memory control registers */
     cpu->reset_sctlr = 0x00c50838;
@@ -219,7 +219,7 @@ static void aarch64_a55_initfn(Object *obj)
     /* Ordered by B2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x84448004; /* L1Ip = VIPT */
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000010305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0000100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000000100001ull);
@@ -325,7 +325,7 @@ static void aarch64_a72_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 3, 64, 48 * KiB, 2);
     /* 1MB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 64, 1 * MiB, 7);
-    cpu->dcz_blocksize = 4; /* 64 bytes */
+    set_dcz_blocksize(cpu, 4); /* 64 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
@@ -352,7 +352,7 @@ static void aarch64_a76_initfn(Object *obj)
     /* Ordered by B2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x8444C004;
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000010305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0000100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000000100001ull);
@@ -424,7 +424,7 @@ static void aarch64_a78ae_initfn(Object *obj)
     /* Ordered by 3.2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x9444c004;
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000110305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0010100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000001200031ull);
@@ -517,7 +517,7 @@ static void aarch64_a64fx_initfn(Object *obj)
     cpu->ccsidr[1] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 4, 256, 64 * KiB, 2);
     /* 8MB L2 cache */
     cpu->ccsidr[2] = make_ccsidr(CCSIDR_FORMAT_LEGACY, 16, 256, 8 * MiB, 7);
-    cpu->dcz_blocksize = 6; /* 256 bytes */
+    set_dcz_blocksize(cpu, 6); /* 256 bytes */
     cpu->gic_num_lrs = 4;
     cpu->gic_vpribits = 5;
     cpu->gic_vprebits = 5;
@@ -673,7 +673,7 @@ static void aarch64_neoverse_n1_initfn(Object *obj)
     /* Ordered by B2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0x8444c004;
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64DFR0, 0x0000000110305408ull);
     SET_IDREG(isar, ID_AA64ISAR0, 0x0000100010211120ull);
     SET_IDREG(isar, ID_AA64ISAR1, 0x0000000000100001ull);
@@ -749,7 +749,7 @@ static void aarch64_neoverse_v1_initfn(Object *obj)
     /* Ordered by 3.2.4 AArch64 registers by functional group */
     SET_IDREG(isar, CLIDR, 0x82000023);
     cpu->ctr = 0xb444c004; /* With DIC and IDC set */
-    cpu->dcz_blocksize = 4;
+    set_dcz_blocksize(cpu, 4);
     SET_IDREG(isar, ID_AA64AFR0, 0x00000000);
     SET_IDREG(isar, ID_AA64AFR1, 0x00000000);
     SET_IDREG(isar, ID_AA64DFR0, 0x000001f210305519ull);
@@ -1011,7 +1011,7 @@ static void aarch64_a710_initfn(Object *obj)
     SET_IDREG(isar, CLIDR, 0x0000001482000023ull);
     cpu->gm_blocksize      = 4;
     cpu->ctr               = 0x000000049444c004ull;
-    cpu->dcz_blocksize     = 4;
+    set_dcz_blocksize(cpu, 4);
     /* TODO FEAT_MPAM: mpamidr_el1 = 0x0000_0001_0006_003f */
 
     /* Section B.5.2: PMCR_EL0 */
@@ -1113,7 +1113,7 @@ static void aarch64_neoverse_n2_initfn(Object *obj)
     SET_IDREG(isar, CLIDR, 0x0000001482000023ull);
     cpu->gm_blocksize      = 4;
     cpu->ctr               = 0x00000004b444c004ull;
-    cpu->dcz_blocksize     = 4;
+    set_dcz_blocksize(cpu, 4);
     /* TODO FEAT_MPAM: mpamidr_el1 = 0x0000_0001_001e_01ff */
 
     /* Section B.7.2: PMCR_EL0 */
@@ -1377,7 +1377,7 @@ void aarch64_max_tcg_initfn(Object *obj)
      * blocksize since we don't have to follow what the hardware does.
      */
     cpu->ctr = 0x80038003; /* 32 byte I and D cacheline size, VIPT icache */
-    cpu->dcz_blocksize = 7; /*  512 bytes */
+    set_dcz_blocksize(cpu, 7); /*  512 bytes */
 #endif
     cpu->gm_blocksize = 6;  /*  256 bytes */
 
target/arm/tcg/helper-a64.c +1 -1
index ba1d775d818e..045a00b43373 100644
@@ -792,7 +792,7 @@ void HELPER(dc_zva)(CPUARMState *env, uint64_t vaddr_in)
      * (which matches the usual QEMU behaviour of not implementing either
      * alignment faults or any memory attribute handling).
      */
-    int blocklen = 4 << env_archcpu(env)->dcz_blocksize;
+    int blocklen = 4 << get_dcz_blocksize(env_archcpu(env));
     uint64_t vaddr = vaddr_in & ~(blocklen - 1);
     int mmu_idx = arm_env_mmu_index(env);
     void *mem;
target/arm/tcg/mte_helper.c +2 -2
index bb48fe359b8c..20f8351d8676 100644
@@ -545,7 +545,7 @@ void HELPER(stzgm_tags)(CPUARMState *env, uint64_t ptr, uint64_t val)
      * i.e. 32 bytes, which is an unreasonably small dcz anyway,
      * to make sure that we can access one complete tag byte here.
      */
-    log2_dcz_bytes = env_archcpu(env)->dcz_blocksize + 2;
+    log2_dcz_bytes = get_dcz_blocksize(env_archcpu(env)) + 2;
     log2_tag_bytes = log2_dcz_bytes - (LOG2_TAG_GRANULE + 1);
     dcz_bytes = (intptr_t)1 << log2_dcz_bytes;
     tag_bytes = (intptr_t)1 << log2_tag_bytes;
@@ -945,7 +945,7 @@ uint64_t HELPER(mte_check_zva)(CPUARMState *env, uint32_t desc, uint64_t ptr)
      * i.e. 32 bytes, which is an unreasonably small dcz anyway, to make
      * sure that we can access one complete tag byte here.
      */
-    log2_dcz_bytes = env_archcpu(env)->dcz_blocksize + 2;
+    log2_dcz_bytes = get_dcz_blocksize(env_archcpu(env)) + 2;
     log2_tag_bytes = log2_dcz_bytes - (LOG2_TAG_GRANULE + 1);
     dcz_bytes = (intptr_t)1 << log2_dcz_bytes;
     tag_bytes = (intptr_t)1 << log2_tag_bytes;
target/arm/tcg/translate-a64.c +0 -0
index 08b21d7dbfa6..c72611e55e45 100644

target/arm/tcg/translate-a64.c static void aarch64_tr_init_disas_context(DisasContextBase *dcbase, · line 10718

Quoted 8 lines
target/arm/tcg/translate-a64.c +1 -1
@@ -10712,7 +10712,7 @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase,
     dc->vec_stride = 0;
     dc->cp_regs = arm_cpu->cp_regs;
     dc->features = env->features;
-    dc->dcz_blocksize = arm_cpu->dcz_blocksize;
+    dc->dcz_blocksize = get_dcz_blocksize(arm_cpu);
     dc->gm_blocksize = arm_cpu->gm_blocksize;
 
 #ifdef CONFIG_USER_ONLY

Thanks

Eric

#2 lore raw
Quoted 6 lines · Eric Auger

Hi Connie,

Quoted 2 lines · Cornelia Huck

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize

what do you mean by EL specific part?

DZP is depending on the current EL (see DC ZVA; it might trap for EL0/EL1.)

Quoted 5 lines

you may simply say that at the moment only bs field is handled as part of cpu.dcz_blocklen
so you replace all the users of that field by accessors to the isar.idreg[] storage

Hm, I'm not sure I understand this sentence -- what I wanted to say is that we care about DCZID_EL0.BS and moving DCZID_EL0 to idregs is made easier by a simple wrapper for that.

Quoted 6 lines
Quoted 6 lines

via a wrapper working on DCZID_EL.BS.

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

(...)

Quoted 7 lines

target/arm/cpu.h struct ARMCPUClass { · line 1179

Quoted 5 lines
target/arm/cpu.h +1 -0
@@ -1178,6 +1176,19 @@ struct ARMCPUClass {
     ResettablePhases parent_phases;
 };
 
+static inline uint8_t get_dcz_blocksize(ARMCPU *cpu)

While at it I would replace dcz_blocksize by dczid_el0_bs to be more explicit

Can do; I wanted to make the diff more straightforward to use.

Quoted 12 lines

target/arm/cpu.h struct ARMCPUClass { · line 1181

Quoted 2 lines
target/arm/cpu.h +2 -0
+{
+    return cpu->isar.idregs[DCZID_EL0_IDX] & 0xf;

extract64?

target/arm/cpu.h struct ARMCPUClass { · line 1189

Quoted 8 lines
target/arm/cpu.h +8 -0
+}
+
+static inline void set_dcz_blocksize(ARMCPU *cpu, uint8_t bs)
+{
+    uint64_t dczid = cpu->isar.idregs[DCZID_EL0_IDX];
+
+    /* keep dzp unchanged */
+    cpu->isar.idregs[DCZID_EL0_IDX] = (dczid & ~0xf) | bs;

deposit64?

Can change.

Quoted 5 lines
Quoted 5 lines
target/arm/cpu.h +2 -0
+}
+
 /* Callback functions for the generic timer's timers. */
 void arm_gt_ptimer_cb(void *opaque);
 void arm_gt_vtimer_cb(void *opaque);

(...)

Quoted 8 lines

target/arm/helper.c static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri)

Quoted 5 lines
target/arm/helper.c +0 -0
index 27ebc6f29b82..8dfeaff25350 100644
@@ -3324,7 +3324,8 @@ static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri)

the name of the function seems to indicate you read the whole DCZID but I see the DZP bit is tweaked below. Do you get why we can't return the raw id reg?

See above -- the value of DZP depends on the current EL (and some other bits, which don't change.)

Quoted 6 lines

target/arm/helper.c static uint64_t aa64_dczid_read(CPUARMState *env, const ARMCPRegInfo *ri) · line 3327

Quoted 5 lines
target/arm/helper.c +1 -1
     if (aa64_zva_access(env, NULL, false) == CP_ACCESS_OK) {
         dzp_bit = 0;
     }
-    return cpu->dcz_blocksize | dzp_bit;
+

nit spurious NL

ok

Quoted 12 lines
Quoted 12 lines
target/arm/helper.c +1 -0
+    return cpu->isar.idregs[DCZID_EL0_IDX] | dzp_bit;
 }
 
 static CPAccessResult sp_el0_access(CPUARMState *env, const ARMCPRegInfo *ri,
target/arm/kvm.c +0 -0
index 0d57081e69fb..5d65f64addc6 100644
@@ -2020,6 +2020,13 @@ int kvm_arch_init_vcpu(CPUState *cs)
     }
     cpu->mp_affinity = mpidr & ARM64_AFFINITY_MASK;
 
Quoted 7 lines

target/arm/kvm.c int kvm_arch_init_vcpu(CPUState *cs) · line 2028

Quoted 6 lines
target/arm/kvm.c +6 -0
+    /*
+     * We currently do not need this, except for tcg. Should KVM gain support
+     * for accessing DCZID_EL0 via ONE_REG, we'll overwrite this below. Just
+     * set a dummy value that corresponds to the minimum value for FEAT_MTE2.
+     */
+    set_dcz_blocksize(cpu, 2);

it is not clear to me why we need that?

We currently don't use it for kvm, but I wanted to put something non-bogus there. Alternatively, we could moan if we try to access the reg under kvm.

Quoted 4 lines
Quoted 4 lines
target/arm/kvm.c +1 -0
+
     return kvm_arm_init_cpreg_list(cpu);
 }
 
#3 lore raw
In reply to Cornelia Huck
Quoted 7 lines · Cornelia Huck

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize via a wrapper working on DCZID_EL.BS.

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

That seems like an odd (unintended?) omission -- is it worth adding? (We would need to handle older kernels that don't expose it anyway, of course.)

thanks
-- PMM

#4 lore raw
Quoted 6 lines · Peter Maydell
Quoted 5 lines · Cornelia Huck

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize via a wrapper working on DCZID_EL.BS.

Quoted 6 lines
Quoted 2 lines

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

That seems like an odd (unintended?) omission -- is it worth adding? (We would need to handle older kernels that don't expose it anyway, of course.)

I'm not sure whether there's actually a usecase for KVM exposing this to the VMM - AFAICS, KVM doesn't do anything special for DC ZVA and friends, and doesn't tweak HCR_EL2.TDZ which would change behaviour.

#5 lore raw
Quoted 12 lines · Cornelia Huck
Quoted 9 lines · Peter Maydell
Quoted 7 lines · Cornelia Huck

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize via a wrapper working on DCZID_EL.BS.

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

Quoted 7 lines
Quoted 3 lines

That seems like an odd (unintended?) omission -- is it worth adding? (We would need to handle older kernels that don't expose it anyway, of course.)

I'm not sure whether there's actually a usecase for KVM exposing this to the VMM - AFAICS, KVM doesn't do anything special for DC ZVA and friends, and doesn't tweak HCR_EL2.TDZ which would change behaviour.

I guess the only one I can think of is to correctly fail migration from a source CPU with a DCZID_EL0.BS that doesn't match the one on the destination CPU. (We can't lie to the guest about the blocksize as part of "tell the guest it has a different CPU type from the actual host" unless we want to trap and emulate all the DC ZVA etc insns...)

thanks
-- PMM

#6 lore raw
Quoted 21 lines · Peter Maydell
Quoted 12 lines · Cornelia Huck
Quoted 9 lines · Peter Maydell
Quoted 7 lines · Cornelia Huck

This requires a bit of care, since we still have to handle the EL specific part (DCZID_EL0.DZP). Callers can set/access dcz_blocksize via a wrapper working on DCZID_EL.BS.

KVM currently does not support DCZID_EL0 via ONE_REG, and actually does not need to work with it, so provide a dummy value for now.

Quoted 7 lines
Quoted 3 lines

That seems like an odd (unintended?) omission -- is it worth adding? (We would need to handle older kernels that don't expose it anyway, of course.)

I'm not sure whether there's actually a usecase for KVM exposing this to the VMM - AFAICS, KVM doesn't do anything special for DC ZVA and friends, and doesn't tweak HCR_EL2.TDZ which would change behaviour.

Quoted 6 lines

I guess the only one I can think of is to correctly fail migration from a source CPU with a DCZID_EL0.BS that doesn't match the one on the destination CPU. (We can't lie to the guest about the blocksize as part of "tell the guest it has a different CPU type from the actual host" unless we want to trap and emulate all the DC ZVA etc insns...)

Agreed.

Might actually be best to assert that kvm code is not doing anything with the reg, rather than providing a dummy value. I'll respin.