Implement basic MSA functions

PATCH Applied David Hildenbrand 3 participants 8 messages Sep 20, 2017 – Sep 22, 2017

lore

Share & tools

lore

David Hildenbrand

author

2017-09-20 15:30 UTC #1 lore raw

PATCH v3 1/3 s390x/tcg: implement spm (SET PROGRAM MASK)

In reply to David Hildenbrand

Missing and is used inside Linux in the context of CPACF.

Reviewed-by Richard Henderson <[email protected]>

Signed-off-by David Hildenbrand <[email protected]>

---
 target/s390x/cpu.h         |  2 ++
 target/s390x/insn-data.def |  2 ++
 target/s390x/translate.c   | 11 +++++++++++
 3 files changed, 15 insertions(+)

target/s390x/cpu.h +2 -0

index 9b549dc491..4414485089 100644
@@ -292,6 +292,7 @@ extern const struct VMStateDescription vmstate_s390_cpu;
 #undef PSW_SHIFT_ASC
 #undef PSW_MASK_CC
 #undef PSW_MASK_PM
+#undef PSW_SHIFT_MASK_PM
 #undef PSW_MASK_64
 #undef PSW_MASK_32
 #undef PSW_MASK_ESA_ADDR

@@ -309,6 +310,7 @@ extern const struct VMStateDescription vmstate_s390_cpu;
 #define PSW_SHIFT_ASC           46
 #define PSW_MASK_CC             0x0000300000000000ULL
 #define PSW_MASK_PM             0x00000F0000000000ULL
+#define PSW_SHIFT_MASK_PM       40
 #define PSW_MASK_64             0x0000000100000000ULL
 #define PSW_MASK_32             0x0000000080000000ULL
 #define PSW_MASK_ESA_ADDR       0x000000007fffffffULL

target/s390x/insn-data.def +2 -0

index ad84c748e3..84233a456d 100644
@@ -755,6 +755,8 @@
     C(0xb2b8, SRNMB,   S,     FPE, 0, 0, 0, 0, srnm, 0)
 /* SET DFP ROUNDING MODE */
     C(0xb2b9, SRNMT,   S,     DFPR, 0, 0, 0, 0, srnm, 0)
+/* SET PROGRAM MASK */
+    C(0x0400, SPM,     RR_a,  Z,   r1, 0, 0, 0, spm, 0)
 
 /* SHIFT LEFT SINGLE */
     D(0x8b00, SLA,     RS_a,  Z,   r1, sh32, new, r1_32, sla, 0, 31)

target/s390x/translate.c +11 -0

index 5abd34fb34..59fde44d55 100644
@@ -3796,6 +3796,17 @@ static ExitStatus op_srnm(DisasContext *s, DisasOps *o)
     return NO_EXIT;
 }
 
+static ExitStatus op_spm(DisasContext *s, DisasOps *o)
+{
+    tcg_gen_extrl_i64_i32(cc_op, o->in1);
+    tcg_gen_extract_i32(cc_op, cc_op, 28, 2);
+    set_cc_static(s);
+
+    tcg_gen_shri_i64(o->in1, o->in1, 24);
+    tcg_gen_deposit_i64(psw_mask, psw_mask, o->in1, PSW_SHIFT_MASK_PM, 4);
+    return NO_EXIT;
+}
+
 #ifndef CONFIG_USER_ONLY
 static ExitStatus op_spka(DisasContext *s, DisasOps *o)
 {

Signature · 1 line

2.13.5

David Hildenbrand

author

2017-09-20 15:30 UTC #0 lore raw

v3 cover Implement basic MSA functions

Some leftover from "target/s390x: tcg improvments + MSA functions".

Implement all basic MSA (cpacf/crypto) instructions <= z13. Only provide the query subfunction (to query available subfunctions), no actual de/encryption yet. Good enough to unlock the STFL bits.

I have written kvm-unit-tests for MSA functions and for SPM/IPM. Will send them out soon. I use the following cpu model to test with an upstream kernel compiled for z10:

... -cpu qemu,mvcos=on,stfle=on,ldisp=on,ldisphp=on,\
              eimm=on,stckf=on,csst=on,csst2=on,ginste=on,\
              exrl=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on ...

Available on: https://github.com/davidhildenbrand/qemu.git s390x-msa

v2 -> v3:
- "s390x/tcg: add basic MSA features"
-- drop one check (fc >= 128) that can due to masking never happen

v1 -> v2:
- "s390x/tcg: implement spm (SET PROGRAM MASK)" -- use tcg_gen_extrl_i64_i32 + tcg_gen_extract_i32
- "s390x/tcg: move wrap_address() to internal.h" -- internal.h instead of cpu.h
- "s390x/tcg: add basic MSA features"
-- minor style fixes

David Hildenbrand (3):
s390x/tcg: implement spm (SET PROGRAM MASK)
s390x/tcg: move wrap_address() to internal.h
s390x/tcg: add basic MSA features

 target/s390x/Makefile.objs   |  2 +-
 target/s390x/cpu.h           |  2 ++
 target/s390x/cpu_models.c    |  4 +++
 target/s390x/crypto_helper.c | 65 ++++++++++++++++++++++++++++++++++++++++++
 target/s390x/helper.h        |  1 +
 target/s390x/insn-data.def   | 15 ++++++++++
 target/s390x/internal.h      | 14 +++++++++
 target/s390x/mem_helper.c    | 14 ---------
 target/s390x/translate.c     | 67 ++++++++++++++++++++++++++++++++++++++++++++
 9 files changed, 169 insertions(+), 15 deletions(-)
 create mode 100644 target/s390x/crypto_helper.c

Signature · 1 line

2.13.5

David Hildenbrand

author

2017-09-20 15:31 UTC #2 lore raw

PATCH v3 2/3 s390x/tcg: move wrap_address() to internal.h

We want to use it in another file.

Reviewed-by Philippe Mathieu-Daudé <[email protected]>

Reviewed-by Richard Henderson <[email protected]>

Signed-off-by David Hildenbrand <[email protected]>

---
 target/s390x/internal.h   | 14 ++++++++++++++
 target/s390x/mem_helper.c | 14 --------------
 2 files changed, 14 insertions(+), 14 deletions(-)

target/s390x/internal.h +14 -0

index bc8f83129a..70d2b87e55 100644
@@ -162,6 +162,20 @@ static inline uint8_t get_per_atmid(CPUS390XState *env)
            ((env->psw.mask & PSW_ASC_ACCREG) ?    (1 << 2) : 0);
 }
 
+static inline uint64_t wrap_address(CPUS390XState *env, uint64_t a)
+{
+    if (!(env->psw.mask & PSW_MASK_64)) {
+        if (!(env->psw.mask & PSW_MASK_32)) {
+            /* 24-Bit mode */
+            a &= 0x00ffffff;
+        } else {
+            /* 31-Bit mode */
+            a &= 0x7fffffff;
+        }
+    }
+    return a;
+}
+
 /* CC optimization */
 
 /* Instead of computing the condition codes after each x86 instruction,

target/s390x/mem_helper.c +0 ~~-14~~

index ec4760e390..a254613dc2 100644
@@ -122,20 +122,6 @@ static inline void cpu_stsize_data_ra(CPUS390XState *env, uint64_t addr,
     }
 }
 
-static inline uint64_t wrap_address(CPUS390XState *env, uint64_t a)
-{
-    if (!(env->psw.mask & PSW_MASK_64)) {
-        if (!(env->psw.mask & PSW_MASK_32)) {
-            /* 24-Bit mode */
-            a &= 0x00ffffff;
-        } else {
-            /* 31-Bit mode */
-            a &= 0x7fffffff;
-        }
-    }
-    return a;
-}
-
 static void fast_memset(CPUS390XState *env, uint64_t dest, uint8_t byte,
                         uint32_t l, uintptr_t ra)
 {

Signature · 1 line

2.13.5

David Hildenbrand

author

2017-09-20 15:31 UTC #3 lore raw

PATCH v3 3/3 s390x/tcg: add basic MSA features

In reply to David Hildenbrand

The STFLE bits for the MSA (extension) facilities simply indicate that the respective instructions can be executed. The QUERY subfunction can then be used to identify which features exactly are available.

Availability of subfunctions can also vary on real hardware. For now, we simply implement a CPU model without any available subfunctions except QUERY (which is always around).

As all MSA functions behave quite similarly, we can use one translation handler for now. Prepare the code for implementation of actual subfunctions.

At least MSA is helpful for now, as older Linux kernels require this facility when compiled for a z9 model. Allow to enable the facilities for the qemu cpu model.

Signed-off-by David Hildenbrand <[email protected]>

---
 target/s390x/Makefile.objs   |  2 +-
 target/s390x/cpu_models.c    |  4 +++
 target/s390x/crypto_helper.c | 65 ++++++++++++++++++++++++++++++++++++++++++++
 target/s390x/helper.h        |  1 +
 target/s390x/insn-data.def   | 13 +++++++++
 target/s390x/translate.c     | 56 ++++++++++++++++++++++++++++++++++++++
 6 files changed, 140 insertions(+), 1 deletion(-)
 create mode 100644 target/s390x/crypto_helper.c

target/s390x/Makefile.objs +1 -1

index 9615256d81..c88ac81e84 100644
@@ -1,6 +1,6 @@
 obj-y += cpu.o cpu_models.o cpu_features.o gdbstub.o interrupt.o helper.o
 obj-$(CONFIG_TCG) += translate.o cc_helper.o excp_helper.o fpu_helper.o
-obj-$(CONFIG_TCG) += int_helper.o mem_helper.o misc_helper.o
+obj-$(CONFIG_TCG) += int_helper.o mem_helper.o misc_helper.o crypto_helper.o
 obj-$(CONFIG_SOFTMMU) += machine.o ioinst.o arch_dump.o mmu_helper.o diag.o
 obj-$(CONFIG_KVM) += kvm.o
 obj-$(call lnot,$(CONFIG_KVM)) += kvm-stub.o

target/s390x/cpu_models.c +4 -0

index 5169379db5..1051a16ece 100644
@@ -825,6 +825,7 @@ static void add_qemu_cpu_model_features(S390FeatBitmap fbm)
         S390_FEAT_STFLE,
         S390_FEAT_EXTENDED_IMMEDIATE,
         S390_FEAT_EXTENDED_TRANSLATION_2,
+        S390_FEAT_MSA,
         S390_FEAT_EXTENDED_TRANSLATION_3,
         S390_FEAT_LONG_DISPLACEMENT,
         S390_FEAT_LONG_DISPLACEMENT_FAST,

@@ -841,6 +842,9 @@ static void add_qemu_cpu_model_features(S390FeatBitmap fbm)
         S390_FEAT_STFLE_49,
         S390_FEAT_LOCAL_TLB_CLEARING,
         S390_FEAT_STFLE_53,
+        S390_FEAT_MSA_EXT_5,
+        S390_FEAT_MSA_EXT_3,
+        S390_FEAT_MSA_EXT_4,
     };
     int i;

target/s390x/crypto_helper.c new file +65 -0

65 lines not on this page · raw message

target/s390x/helper.h +1 -0

index 4b0290774e..75ba04fc15 100644
@@ -115,6 +115,7 @@ DEF_HELPER_4(cu21, i32, env, i32, i32, i32)
 DEF_HELPER_4(cu24, i32, env, i32, i32, i32)
 DEF_HELPER_4(cu41, i32, env, i32, i32, i32)
 DEF_HELPER_4(cu42, i32, env, i32, i32, i32)
+DEF_HELPER_5(msa, i32, env, i32, i32, i32, i32)
 
 #ifndef CONFIG_USER_ONLY
 DEF_HELPER_3(servc, i32, env, i64, i64)

target/s390x/insn-data.def +13 -0

index 84233a456d..d09f2ed538 100644
@@ -941,6 +941,19 @@
 /* UNPACK UNICODE */
     C(0xe200, UNPKU,   SS_a,  E2,  la1, a2, 0, 0, unpku, 0)
 
+/* MSA Instructions */
+    D(0xb91e, KMAC,    RRE,   MSA,  0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KMAC)
+    D(0xb928, PCKMO,   RRE,   MSA3, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_PCKMO)
+    D(0xb92a, KMF,     RRE,   MSA4, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KMF)
+    D(0xb92b, KMO,     RRE,   MSA4, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_KMO)
+    D(0xb92c, PCC,     RRE,   MSA4, 0, 0, 0, 0, msa, 0, S390_FEAT_TYPE_PCC)
+    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(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)
+
 #ifndef CONFIG_USER_ONLY
 /* COMPARE AND SWAP AND PURGE */
     D(0xb250, CSP,     RRE,   Z,   r1_32u, ra2, r1_P, 0, csp, 0, MO_TEUL)

target/s390x/translate.c +56 -0

68 lines not on this page · raw message

Signature · 1 line

2.13.5

Richard Henderson

Reviewed-by on v3 3/3

2017-09-20 16:32 UTC #4 lore raw

Re: [Qemu-devel] [PATCH v3 3/3] s390x/tcg: add basic MSA features

Quoted 16 lines · David Hildenbrand

The STFLE bits for the MSA (extension) facilities simply indicate that the respective instructions can be executed. The QUERY subfunction can then be used to identify which features exactly are available.
Availability of subfunctions can also vary on real hardware. For now, we simply implement a CPU model without any available subfunctions except QUERY (which is always around).
As all MSA functions behave quite similarly, we can use one translation handler for now. Prepare the code for implementation of actual subfunctions.
At least MSA is helpful for now, as older Linux kernels require this facility when compiled for a z9 model. Allow to enable the facilities for the qemu cpu model.
Signed-off-by David Hildenbrand <[email protected]>

Quoted 9 lines

---
 target/s390x/Makefile.objs   |  2 +-
 target/s390x/cpu_models.c    |  4 +++
 target/s390x/crypto_helper.c | 65 ++++++++++++++++++++++++++++++++++++++++++++
 target/s390x/helper.h        |  1 +
 target/s390x/insn-data.def   | 13 +++++++++
 target/s390x/translate.c     | 56 ++++++++++++++++++++++++++++++++++++++
 6 files changed, 140 insertions(+), 1 deletion(-)
 create mode 100644 target/s390x/crypto_helper.c

Reviewed-by Richard Henderson <[email protected]>

Cornelia Huck

2017-09-21 08:23 UTC #5 lore raw

In reply to David Hildenbrand

Quoted 10 lines · On Wed, 20 Sep 2017 17:30:13 +0200 David Hildenbrand <[email protected]> wrote:

Some leftover from "target/s390x: tcg improvments + MSA functions".
Implement all basic MSA (cpacf/crypto) instructions <= z13. Only provide the query subfunction (to query available subfunctions), no actual de/encryption yet. Good enough to unlock the STFL bits.
I have written kvm-unit-tests for MSA functions and for SPM/IPM. Will send them out soon. I use the following cpu model to test with an upstream kernel compiled for z10:

Quoted 5 lines

... -cpu qemu,mvcos=on,stfle=on,ldisp=on,ldisphp=on,\
              eimm=on,stckf=on,csst=on,csst2=on,ginste=on,\
              exrl=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on ...

Available on: https://github.com/davidhildenbrand/qemu.git s390x-msa

I used the following cpu model:

-cpu qemu,eimm=on,ldisp=on,stckf=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on

and got

qemu-system-s390x: warning: 'msa5-base' requires 'kimd-sha-512'.
qemu-system-s390x: warning: 'msa5-base' requires 'klmd-sha-512'.

(It works anyway.)

However,

-cpu qemu,eimm=on,ldisp=on,stckf=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on,kimd-sha-512=on

gives

qemu-system-s390x: can't apply global qemu-s390-cpu.kimd-sha-512=on: Feature 'kimd-sha-512' is not available for CPU model 'z900', it was introduced with later models.

which is a bit confusing.

So I'm wondering whether providing msa5-base without the providing the algorithms above is sane - or whether it is just enough to satisfy some kernels.

[Double-checking the various dependencies looks impossible without access to IBM doc, anyway - I just hope you got it right back then :)]

David Hildenbrand

author

2017-09-21 12:38 UTC #6 lore raw

Quoted 19 lines · Cornelia Huck

Quoted 10 lines · On Wed, 20 Sep 2017 17:30:13 +0200 David Hildenbrand <[email protected]> wrote:
Some leftover from "target/s390x: tcg improvments + MSA functions".
Implement all basic MSA (cpacf/crypto) instructions <= z13. Only provide the query subfunction (to query available subfunctions), no actual de/encryption yet. Good enough to unlock the STFL bits.
I have written kvm-unit-tests for MSA functions and for SPM/IPM. Will send them out soon. I use the following cpu model to test with an upstream kernel compiled for z10:
Quoted 5 lines
... -cpu qemu,mvcos=on,stfle=on,ldisp=on,ldisphp=on,\
              eimm=on,stckf=on,csst=on,csst2=on,ginste=on,\
              exrl=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on ...
Available on: https://github.com/davidhildenbrand/qemu.git s390x-msa

Quoted 9 lines

I used the following cpu model:
-cpu qemu,eimm=on,ldisp=on,stckf=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on
and got
qemu-system-s390x: warning: 'msa5-base' requires 'kimd-sha-512'.
qemu-system-s390x: warning: 'msa5-base' requires 'klmd-sha-512'.

Yes, I am aware of that and it is on my todo list. Shouldn't matter for now as it is completely optional and not enabled as default.

Quoted 11 lines

(It works anyway.)
However,
-cpu qemu,eimm=on,ldisp=on,stckf=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on,kimd-sha-512=on
gives
qemu-system-s390x: can't apply global qemu-s390-cpu.kimd-sha-512=on: Feature 'kimd-sha-512' is not available for CPU model 'z900', it was introduced with later models.
which is a bit confusing.

Due to the way we hack up the qemu cpu model, allowing to set features that are theoretically not possible (just to have some way of enabling them before we can change the qemu model to something never).

Quoted 8 lines

So I'm wondering whether providing msa5-base without the providing the algorithms above is sane - or whether it is just enough to satisfy some kernels.
[Double-checking the various dependencies looks impossible without access to IBM doc, anyway - I just hope you got it right back then :)]

We can take care of that later (if at all relevant).

Signature · 3 lines

Thanks,

David

Cornelia Huck

Applied

2017-09-22 11:34 UTC #7 lore raw

In reply to David Hildenbrand

Quoted 34 lines · On Wed, 20 Sep 2017 17:30:13 +0200 David Hildenbrand <[email protected]> wrote:

Some leftover from "target/s390x: tcg improvments + MSA functions".
Implement all basic MSA (cpacf/crypto) instructions <= z13. Only provide the query subfunction (to query available subfunctions), no actual de/encryption yet. Good enough to unlock the STFL bits.
I have written kvm-unit-tests for MSA functions and for SPM/IPM. Will send them out soon. I use the following cpu model to test with an upstream kernel compiled for z10:
... -cpu qemu,mvcos=on,stfle=on,ldisp=on,ldisphp=on,\
              eimm=on,stckf=on,csst=on,csst2=on,ginste=on,\
              exrl=on,msa-base=on,msa3-base=on,msa4-base=on,msa5-base=on ...
Available on: https://github.com/davidhildenbrand/qemu.git s390x-msa
v2 -> v3:
- "s390x/tcg: add basic MSA features"
-- drop one check (fc >= 128) that can due to masking never happen
v1 -> v2:
- "s390x/tcg: implement spm (SET PROGRAM MASK)" -- use tcg_gen_extrl_i64_i32 + tcg_gen_extract_i32
- "s390x/tcg: move wrap_address() to internal.h" -- internal.h instead of cpu.h
- "s390x/tcg: add basic MSA features"
-- minor style fixes
David Hildenbrand (3):
s390x/tcg: implement spm (SET PROGRAM MASK)
s390x/tcg: move wrap_address() to internal.h
s390x/tcg: add basic MSA features

Quoted 12 lines

 target/s390x/Makefile.objs   |  2 +-
 target/s390x/cpu.h           |  2 ++
 target/s390x/cpu_models.c    |  4 +++
 target/s390x/crypto_helper.c | 65 ++++++++++++++++++++++++++++++++++++++++++
 target/s390x/helper.h        |  1 +
 target/s390x/insn-data.def   | 15 ++++++++++
 target/s390x/internal.h      | 14 +++++++++
 target/s390x/mem_helper.c    | 14 ---------
 target/s390x/translate.c     | 67 ++++++++++++++++++++++++++++++++++++++++++++
 9 files changed, 169 insertions(+), 15 deletions(-)
 create mode 100644 target/s390x/crypto_helper.c

Thanks, applied.