Skip to contentSkip to conversation

s390x: STFLE nested interpretation

PATCH Superseded Nina Schoetterl-Glausch 7 participants 41 messages Jun 20, 2024 – Oct 16, 2024

Subsystems (4)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover s390x: STFLE nested interpretation

v2 -> v3:
* pick up Ack (thanks Andrew)
* minor cosmetic change to rand generator
* add sie_is_pv function
* extend sie_is_diag_icpt to support pv, replace pv_icptdata_check_diag

Add a test case that tests the interpretation of STFLE performed by a nested guest using a snippet.
Also add some functionality to lib/, namely:
* pseudo random number generation (arch independent)
* exit (optionally with return code) from snippet (s390x)
* add function for checking diag intercepts, replacing

      pv_icptdata_check_diag (s390x)

v1 -> v2:
* implement SHA-256 based PRNG
* pick up R-b (thanks Claudio)
* change snippet exit API and implementation (thanks Claudio)
* add stfle-sie to unittests.cfg

Nina Schoetterl-Glausch (7):
lib: Add pseudo random functions
s390x: lib: Remove double include
s390x: Add sie_is_pv
s390x: Add function for checking diagnose intercepts
s390x: Add library functions for exiting from snippet
s390x: Use library functions for snippet exit
s390x: Add test for STFLE interpretive execution (format-0)

 Makefile                                |   1 +
 s390x/Makefile                          |   3 +
 lib/s390x/asm/arch_def.h                |  13 ++
 lib/s390x/asm/facility.h                |  10 +-
 lib/rand.h                              |  21 +++
 lib/s390x/pv_icptdata.h                 |  42 ------
 lib/s390x/sie.h                         |  18 +++
 lib/s390x/snippet-guest.h               |  26 ++++
 lib/s390x/{snippet.h => snippet-host.h} |  10 +-
 lib/rand.c                              | 177 ++++++++++++++++++++++++
 lib/s390x/sie.c                         |  58 +++++++-
 lib/s390x/snippet-host.c                |  42 ++++++
 lib/s390x/uv.c                          |   2 +-
 s390x/mvpg-sie.c                        |   2 +-
 s390x/pv-diags.c                        |  10 +-
 s390x/pv-icptcode.c                     |  13 +-
 s390x/pv-ipl.c                          |   9 +-
 s390x/sie-dat.c                         |  13 +-
 s390x/snippets/c/sie-dat.c              |  19 +--
 s390x/snippets/c/stfle.c                |  26 ++++
 s390x/spec_ex-sie.c                     |   2 +-
 s390x/stfle-sie.c                       | 134 ++++++++++++++++++
 s390x/uv-host.c                         |   2 +-
 s390x/unittests.cfg                     |   3 +
 24 files changed, 558 insertions(+), 98 deletions(-)
 create mode 100644 lib/rand.h
 delete mode 100644 lib/s390x/pv_icptdata.h
 create mode 100644 lib/s390x/snippet-guest.h
 rename lib/s390x/{snippet.h => snippet-host.h} (92%)
 create mode 100644 lib/rand.c
 create mode 100644 lib/s390x/snippet-host.c
 create mode 100644 s390x/snippets/c/stfle.c
 create mode 100644 s390x/stfle-sie.c

Range-diff against v2:
1: 6c869961 ! 1: baecabf2 lib: Add pseudo random functions

    @@ Commit message
                 for i in range(8):
                     yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")
     
    +    Acked-by: Andrew Jones <[email protected]>
         Signed-off-by: Nina Schoetterl-Glausch <[email protected]>
     
     
    @@ lib/rand.c (new)
     +	return rot(x, 17) ^ rot(x, 19) ^ (x >> 10);
     +}
     +
    -+enum alphabet { a, b, c, d, e, f, g, h, };
    ++enum alphabet { A, B, C, D, E, F, G, H, };
     +
     +static void sha256_chunk(const uint32_t (*chunk)[16], uint32_t (*hash)[8])
     +{
    @@ lib/rand.c (new)
     +	for (int i = 0; i < 64; i++) {
     +		uint32_t t1, t2;
     +
    -+		t1 = w_hash[h] +
    -+		     upper_sig1(w_hash[e]) +
    -+		     ch(w_hash[e], w_hash[f], w_hash[g]) +
    ++		t1 = w_hash[H] +
    ++		     upper_sig1(w_hash[E]) +
    ++		     ch(w_hash[E], w_hash[F], w_hash[G]) +
     +		     K[i] +
     +		     w[i];
     +
    -+		t2 = upper_sig0(w_hash[a]) + maj(w_hash[a], w_hash[b], w_hash[c]);
    ++		t2 = upper_sig0(w_hash[A]) + maj(w_hash[A], w_hash[B], w_hash[C]);
     +
    -+		w_hash[h] = w_hash[g];
    -+		w_hash[g] = w_hash[f];
    -+		w_hash[f] = w_hash[e];
    -+		w_hash[e] = w_hash[d] + t1;
    -+		w_hash[d] = w_hash[c];
    -+		w_hash[c] = w_hash[b];
    -+		w_hash[b] = w_hash[a];
    -+		w_hash[a] = t1 + t2;
    ++		w_hash[H] = w_hash[G];
    ++		w_hash[G] = w_hash[F];
    ++		w_hash[F] = w_hash[E];
    ++		w_hash[E] = w_hash[D] + t1;
    ++		w_hash[D] = w_hash[C];
    ++		w_hash[C] = w_hash[B];
    ++		w_hash[B] = w_hash[A];
    ++		w_hash[A] = t1 + t2;
     +	}
     +
     +	for (int i = 0; i < 8; i++)
2:  77319d3e = 2:  b30314eb s390x: lib: Remove double include
-:  -------- > 3:  f2af539b s390x: Add sie_is_pv
-:  -------- > 4:  c4331d19 s390x: Add function for checking diagnose intercepts
3:  e2c2cad8 ! 5:  a3f92777 s390x: Add library functions for exiting from snippet
    @@ lib/s390x/asm/arch_def.h: static inline uint32_t get_prefix(void)
     +
      #endif
     
    - ## lib/s390x/sie.h ##
    -@@ lib/s390x/sie.h: void sie_expect_validity(struct vm *vm);
    - uint16_t sie_get_validity(struct vm *vm);
    - void sie_check_validity(struct vm *vm, uint16_t vir_exp);
    - void sie_handle_validity(struct vm *vm);
    -+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag);
    - void sie_guest_sca_create(struct vm *vm);
    - void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
    - void sie_guest_destroy(struct vm *vm);
    -
      ## lib/s390x/snippet-guest.h (new) ##
     @@
     +/* SPDX-License-Identifier: GPL-2.0-only */
    @@ lib/s390x/snippet-guest.h (new)
     +	mb(); /* allow host to modify guest memory */
     +}
     +
    -+#endif
    ++#endif /* _S390X_SNIPPET_GUEST_H_ */
     
      ## lib/s390x/snippet.h => lib/s390x/snippet-host.h ##
     @@
    @@ lib/s390x/snippet-host.h: static inline void snippet_setup_guest(struct vm *vm,
     +void snippet_check_force_exit_value(struct vm *vm, uint64_t exit_exp);
      #endif
     
    - ## lib/s390x/sie.c ##
    -@@ lib/s390x/sie.c: void sie_check_validity(struct vm *vm, uint16_t vir_exp)
    - 	report(vir_exp == vir, "VALIDITY: %x", vir);
    - }
    - 
    -+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag)
    -+{
    -+	union {
    -+		struct {
    -+			uint64_t     : 16;
    -+			uint64_t ipa : 16;
    -+			uint64_t ipb : 32;
    -+		};
    -+		struct {
    -+			uint64_t          : 16;
    -+			uint64_t opcode   :  8;
    -+			uint64_t r_1      :  4;
    -+			uint64_t r_2      :  4;
    -+			uint64_t r_base   :  4;
    -+			uint64_t displace : 12;
    -+			uint64_t zero     : 16;
    -+		};
    -+	} instr = { .ipa = vm->sblk->ipa, .ipb = vm->sblk->ipb };
    -+	uint64_t code;
    -+
    -+	assert(diag == 0x44 || diag == 0x9c);
    -+
    -+	if (vm->sblk->icptcode != ICPT_INST)
    -+		return false;
    -+	if (instr.opcode != 0x83 || instr.zero)
    -+		return false;
    -+	code = instr.r_base ? vm->save_area.guest.grs[instr.r_base] : 0;
    -+	code = (code + instr.displace) & 0xffff;
    -+	return code == diag;
    -+}
    -+
    - void sie_handle_validity(struct vm *vm)
    - {
    - 	if (vm->sblk->icptcode != ICPT_VALIDITY)
    -
      ## lib/s390x/snippet-host.c (new) ##
     @@
     +/* SPDX-License-Identifier: GPL-2.0-only */
    @@ s390x/pv-diags.c
      #include <libcflat.h>
     -#include <snippet.h>
     +#include <snippet-host.h>
    - #include <pv_icptdata.h>
      #include <sie.h>
      #include <sclp.h>
    + #include <asm/facility.h>
     
      ## s390x/pv-icptcode.c ##
     @@
    @@ s390x/pv-icptcode.c
      #include <sclp.h>
     -#include <snippet.h>
     +#include <snippet-host.h>
    - #include <pv_icptdata.h>
      #include <asm/facility.h>
      #include <asm/barrier.h>
    + #include <asm/sigp.h>
     
      ## s390x/pv-ipl.c ##
     @@
    @@ s390x/pv-ipl.c
      #include <sclp.h>
     -#include <snippet.h>
     +#include <snippet-host.h>
    - #include <pv_icptdata.h>
      #include <asm/facility.h>
      #include <asm/uv.h>
    + 
     
      ## s390x/sie-dat.c ##
     @@
4:  67fbf0bb ! 6:  a1db588b s390x: Use library functions for snippet exit
    @@ s390x/sie-dat.c: static void test_sie_dat(void)
     
      ## s390x/snippets/c/sie-dat.c ##
     @@
    -  */
      #include <libcflat.h>
      #include <asm-generic/page.h>
    + #include <asm/mem.h>
     +#include <snippet-guest.h>
      #include "sie-dat.h"

5: 157079f2 ! 7: 0b89b3c6 s390x: Add test for STFLE interpretive execution (format-0)

    @@ s390x/stfle-sie.c (new)
     +}
     
      ## s390x/unittests.cfg ##
    -@@ s390x/unittests.cfg: extra_params = """-cpu max,ctop=on -smp cpus=1,drawers=2,books=2,sockets=2,cores
    +@@ s390x/unittests.cfg: file = sie-dat.elf
      
    - [sie-dat]
    - file = sie-dat.elf
    + [pv-attest]
    + file = pv-attest.elf
     +
     +[stfle-sie]
     +file = stfle-sie.elf

base-commit: 28ac3b10d6f982b1d9c2fe629f23d23ec5024b4f

Signature · 1 line
2.44.0
#1 lore raw

PATCH v3 1/7 lib: Add pseudo random functions

Add functions for generating pseudo random 32 and 64 bit values. The implementation uses SHA-256 and so the randomness should have good quality.
Implement the necessary subset of SHA-256.
The PRNG algorithm is equivalent to the following python snippet:

def prng32(seed):
from hashlib import sha256
state = seed.to_bytes(8, byteorder="big")
while True:
state = sha256(state).digest()
for i in range(8):
yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")

---

Notes:

    Since a PRNG with better quality was asked for I decided to use SHA-256
    because:
     * it is a standard, commonly used algorithm
     * high quality randomness is assured
     * the implementation can be checked against the spec
     * the implementation can be easily checked via comparison
    
    I tested the implementation in the following way:
    
    cat <<'EOF' > rand.py
    #!/usr/bin/python3
    
    def prng32(seed):
        from hashlib import sha256
        state = seed.to_bytes(8, byteorder="big")
        while True:
            state = sha256(state).digest()
            for i in range(8):
                yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")
    
    r = prng32(0)
    for i in range(100):
        print(f"{next(r):08x}")
    
    EOF
    
    cat <<'EOF' > rand.c
    #include <stdio.h>
    #include "rand.h"
    
    void main(void)
    {
    	prng_state state = prng_init(0);
    	for (int i = 0; i < 100; i++) {
    		printf("%08x\n", prng32(&state));
    	}
    }
    EOF
    cat <<'EOF' > libcflat.h
    #define ARRAY_SIZE(_a) (sizeof(_a)/sizeof((_a)[0]))
    EOF
    chmod +x rand.py
    ln -s lib/rand.c librand.c
    gcc -Ilib librand.c rand.c
    diff <(./a.out) <(./rand.py)
 Makefile   |   1 +
 lib/rand.h |  21 +++++++
 lib/rand.c | 177 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 199 insertions(+)
 create mode 100644 lib/rand.h
 create mode 100644 lib/rand.c
Makefile +1 -0
index 5b7998b7..3d51cb72 100644
@@ -28,6 +28,7 @@ cflatobjs := \
 	lib/printf.o \
 	lib/string.o \
 	lib/abort.o \
+	lib/rand.o \
 	lib/report.o \
 	lib/stack.o
 
lib/rand.h new file +21 -0
new file mode 100644
index 00000000..cdce8bd7
@@ -0,0 +1,21 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * (pseudo) random functions
+ *
+ * Copyright IBM Corp. 2024
+ */
+#ifndef _RAND_H_
+#define _RAND_H_
+
+#include <stdint.h>
+
+/* Non cryptographically secure PRNG */
+typedef struct {
+	uint32_t hash[8];
+	uint8_t next_word;
+} prng_state;
+prng_state prng_init(uint64_t seed);
+uint32_t prng32(prng_state *state);
+uint64_t prng64(prng_state *state);
+
+#endif /* _RAND_H_ */
lib/rand.c new file +177 -0

177 lines not on this page · raw message

Signature · 1 line
2.44.0
Acked-by on v3 1/7
#2 lore raw

Re: [kvm-unit-tests PATCH v3 1/7] lib: Add pseudo random functions

Quoted 6 lines · Nina Schoetterl-Glausch

Add functions for generating pseudo random 32 and 64 bit values. The implementation uses SHA-256 and so the randomness should have good quality.
Implement the necessary subset of SHA-256.
The PRNG algorithm is equivalent to the following python snippet:

Quoted 8 lines

def prng32(seed):
from hashlib import sha256
state = seed.to_bytes(8, byteorder="big")
while True:
state = sha256(state).digest()
for i in range(8):
yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")

Nice, could use this for powerpc radom SPR value tests (and probably other things too).

Quoted 42 lines

---

Notes:

    Since a PRNG with better quality was asked for I decided to use SHA-256
    because:
     * it is a standard, commonly used algorithm
     * high quality randomness is assured
     * the implementation can be checked against the spec
     * the implementation can be easily checked via comparison
    
    I tested the implementation in the following way:
    
    cat <<'EOF' > rand.py
    #!/usr/bin/python3
    
    def prng32(seed):
        from hashlib import sha256
        state = seed.to_bytes(8, byteorder="big")
        while True:
            state = sha256(state).digest()
            for i in range(8):
                yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")
    
    r = prng32(0)
    for i in range(100):
        print(f"{next(r):08x}")
    
    EOF
    
    cat <<'EOF' > rand.c
    #include <stdio.h>
    #include "rand.h"
    
    void main(void)
    {
    	prng_state state = prng_init(0);
    	for (int i = 0; i < 100; i++) {
    		printf("%08x\n", prng32(&state));
    	}
    }
Quoted 8 lines
    EOF
    cat <<'EOF' > libcflat.h
    #define ARRAY_SIZE(_a) (sizeof(_a)/sizeof((_a)[0]))
    EOF
    chmod +x rand.py
    ln -s lib/rand.c librand.c
    gcc -Ilib librand.c rand.c
    diff <(./a.out) <(./rand.py)

Cool... you made a unit test for the unit tests. We could start a make check? :)

Thanks,
Nick

#3 lore raw

Re: [kvm-unit-tests PATCH v3 1/7] lib: Add pseudo random functions

Quoted 1 line · Nicholas Piggin

On Fri Jun 21, 2024 at 12:16 AM AEST, Nina Schoetterl-Glausch wrote:

[...]

Quoted 39 lines
Quoted 30 lines
    I tested the implementation in the following way:
    
    cat <<'EOF' > rand.py
    #!/usr/bin/python3
    
    def prng32(seed):
        from hashlib import sha256
        state = seed.to_bytes(8, byteorder="big")
        while True:
            state = sha256(state).digest()
            for i in range(8):
                yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")
    
    r = prng32(0)
    for i in range(100):
        print(f"{next(r):08x}")
    
    EOF
    
    cat <<'EOF' > rand.c
    #include <stdio.h>
    #include "rand.h"
    
    void main(void)
    {
    	prng_state state = prng_init(0);
    	for (int i = 0; i < 100; i++) {
    		printf("%08x\n", prng32(&state));
    	}
    }
Quoted 8 lines
    EOF
    cat <<'EOF' > libcflat.h
    #define ARRAY_SIZE(_a) (sizeof(_a)/sizeof((_a)[0]))
    EOF
    chmod +x rand.py
    ln -s lib/rand.c librand.c
    gcc -Ilib librand.c rand.c
    diff <(./a.out) <(./rand.py)
Quoted 2 lines

Cool... you made a unit test for the unit tests. We could start a make check? :)

I wouldn't complain about it, but my test is a bit hacky and I don't expect the code to get touched much.

Quoted 5 lines

Thanks,
Nick

#4 lore raw

Re: [kvm-unit-tests PATCH v3 1/7] lib: Add pseudo random functions

In reply to Nina Schoetterl-Glausch
Quoted 14 lines · Nina Schoetterl-Glausch

Add functions for generating pseudo random 32 and 64 bit values. The implementation uses SHA-256 and so the randomness should have good quality.
Implement the necessary subset of SHA-256.
The PRNG algorithm is equivalent to the following python snippet:

def prng32(seed):
from hashlib import sha256
state = seed.to_bytes(8, byteorder="big")
while True:
state = sha256(state).digest()
for i in range(8):
yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")

Quoted 4 lines

---

There was a discussion about potentially needing random numbers for RISC-V SBI tests, so I've queued this on riscv/sbi,
https://gitlab.com/jones-drew/kvm-unit-tests/-/commits/riscv%2Fsbi

Thanks,
drew

#5 lore raw

Re: [kvm-unit-tests PATCH v3 1/7] lib: Add pseudo random functions

In reply to Nina Schoetterl-Glausch

Quoting Nina Schoetterl-Glausch (2024-06-20 16:16:54)

Quoted 6 lines

Add functions for generating pseudo random 32 and 64 bit values. The implementation uses SHA-256 and so the randomness should have good quality.
Implement the necessary subset of SHA-256.
The PRNG algorithm is equivalent to the following python snippet:

Quoted 7 lines

def prng32(seed):
from hashlib import sha256
state = seed.to_bytes(8, byteorder="big")
while True:
state = sha256(state).digest()
for i in range(8):
yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")

Thomas, Andrew,
do you want to take this patch or is it OK it if comes with the whole series via s390x tree?

#6 lore raw

Re: [kvm-unit-tests PATCH v3 1/7] lib: Add pseudo random functions

Quoted 15 lines · Nico Boehr

Quoting Nina Schoetterl-Glausch (2024-06-20 16:16:54)

Quoted 6 lines

Add functions for generating pseudo random 32 and 64 bit values. The implementation uses SHA-256 and so the randomness should have good quality.
Implement the necessary subset of SHA-256.
The PRNG algorithm is equivalent to the following python snippet:

Quoted 7 lines

def prng32(seed):
from hashlib import sha256
state = seed.to_bytes(8, byteorder="big")
while True:
state = sha256(state).digest()
for i in range(8):
yield int.from_bytes(state[i*4:(i+1)*4], byteorder="big")

Quoted 3 lines

Thomas, Andrew,
do you want to take this patch or is it OK it if comes with the whole series via s390x tree?

It's already merged, commit e83373301c43 ("lib: Add pseudo random functions")

Thanks,
drew

#7 lore raw

PATCH v3 2/7 s390x: lib: Remove double include

In reply to Nina Schoetterl-Glausch

libcflat.h was included twice.

---
 lib/s390x/sie.c | 1 -
 1 file changed, 1 deletion(-)
lib/s390x/sie.c +0 -1
index 28fbf146..40936bd2 100644
@@ -14,7 +14,6 @@
 #include <sie.h>
 #include <asm/page.h>
 #include <asm/interrupt.h>
-#include <libcflat.h>
 #include <alloc_page.h>
 #include <vmalloc.h>
 #include <sclp.h>
Signature · 1 line
2.44.0
Reviewed-by on v3 2/7
#8 lore raw

Re: [kvm-unit-tests PATCH v3 2/7] s390x: lib: Remove double include

Quoted 4 lines · Nina Schoetterl-Glausch

libcflat.h was included twice.

Quoted 16 lines
---
  lib/s390x/sie.c | 1 -
  1 file changed, 1 deletion(-)
lib/s390x/sie.c +0 -1
index 28fbf146..40936bd2 100644
@@ -14,7 +14,6 @@
  #include <sie.h>
  #include <asm/page.h>
  #include <asm/interrupt.h>
-#include <libcflat.h>
  #include <alloc_page.h>
  #include <vmalloc.h>
  #include <sclp.h>
#9 lore raw

PATCH v3 3/7 s390x: Add sie_is_pv

In reply to Nina Schoetterl-Glausch

Add a function to check if a guest VM is currently running protected.

---
 lib/s390x/sie.h | 6 ++++++
 lib/s390x/sie.c | 4 ++--
 2 files changed, 8 insertions(+), 2 deletions(-)
lib/s390x/sie.h +6 -0
index c1724cf2..53cd767f 100644
@@ -281,6 +281,12 @@ void sie_expect_validity(struct vm *vm);
 uint16_t sie_get_validity(struct vm *vm);
 void sie_check_validity(struct vm *vm, uint16_t vir_exp);
 void sie_handle_validity(struct vm *vm);
+
+static inline bool sie_is_pv(struct vm *vm)
+{
+	return vm->sblk->sdf == 2;
+}
+
 void sie_guest_sca_create(struct vm *vm);
 void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
 void sie_guest_destroy(struct vm *vm);
lib/s390x/sie.c +2 -2
index 40936bd2..0fa915cf 100644
@@ -59,7 +59,7 @@ void sie(struct vm *vm)
 	/* When a pgm int code is set, we'll never enter SIE below. */
 	assert(!read_pgm_int_code());
 
-	if (vm->sblk->sdf == 2)
+	if (sie_is_pv(vm))
 		memcpy(vm->sblk->pv_grregs, vm->save_area.guest.grs,
 		       sizeof(vm->save_area.guest.grs));
 
@@ -98,7 +98,7 @@ void sie(struct vm *vm)
 	/* restore the old CR 13 */
 	lctlg(13, old_cr13);
 
-	if (vm->sblk->sdf == 2)
+	if (sie_is_pv(vm))
 		memcpy(vm->save_area.guest.grs, vm->sblk->pv_grregs,
 		       sizeof(vm->save_area.guest.grs));
 }
Signature · 1 line
2.44.0
Reviewed-by on v3 3/7
#10 lore raw

Re: [kvm-unit-tests PATCH v3 3/7] s390x: Add sie_is_pv

Quoted 3 lines · On Thu, 20 Jun 2024 16:16:56 +0200 Nina Schoetterl-Glausch <[email protected]>…

Add a function to check if a guest VM is currently running protected.

Quoted 44 lines
---
 lib/s390x/sie.h | 6 ++++++
 lib/s390x/sie.c | 4 ++--
 2 files changed, 8 insertions(+), 2 deletions(-)
lib/s390x/sie.h +6 -0
index c1724cf2..53cd767f 100644
@@ -281,6 +281,12 @@ void sie_expect_validity(struct vm *vm);
 uint16_t sie_get_validity(struct vm *vm);
 void sie_check_validity(struct vm *vm, uint16_t vir_exp);
 void sie_handle_validity(struct vm *vm);
+
+static inline bool sie_is_pv(struct vm *vm)
+{
+	return vm->sblk->sdf == 2;
+}
+
 void sie_guest_sca_create(struct vm *vm);
 void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
 void sie_guest_destroy(struct vm *vm);
lib/s390x/sie.c +2 -2
index 40936bd2..0fa915cf 100644
@@ -59,7 +59,7 @@ void sie(struct vm *vm)
 	/* When a pgm int code is set, we'll never enter SIE below. */
 	assert(!read_pgm_int_code());
 
-	if (vm->sblk->sdf == 2)
+	if (sie_is_pv(vm))
 		memcpy(vm->sblk->pv_grregs, vm->save_area.guest.grs,
 		       sizeof(vm->save_area.guest.grs));
 
@@ -98,7 +98,7 @@ void sie(struct vm *vm)
 	/* restore the old CR 13 */
 	lctlg(13, old_cr13);
 
-	if (vm->sblk->sdf == 2)
+	if (sie_is_pv(vm))
 		memcpy(vm->save_area.guest.grs, vm->sblk->pv_grregs,
 		       sizeof(vm->save_area.guest.grs));
 }
Reviewed-by on v3 3/7
#12 lore raw

Re: [kvm-unit-tests PATCH v3 3/7] s390x: Add sie_is_pv

In reply to Nina Schoetterl-Glausch
Quoted 3 lines · Nina Schoetterl-Glausch

Add a function to check if a guest VM is currently running protected.

Quoted 44 lines
---
 lib/s390x/sie.h | 6 ++++++
 lib/s390x/sie.c | 4 ++--
 2 files changed, 8 insertions(+), 2 deletions(-)
lib/s390x/sie.h +6 -0
index c1724cf2..53cd767f 100644
@@ -281,6 +281,12 @@ void sie_expect_validity(struct vm *vm);
 uint16_t sie_get_validity(struct vm *vm);
 void sie_check_validity(struct vm *vm, uint16_t vir_exp);
 void sie_handle_validity(struct vm *vm);
+
+static inline bool sie_is_pv(struct vm *vm)
+{
+	return vm->sblk->sdf == 2;
+}
+
 void sie_guest_sca_create(struct vm *vm);
 void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
 void sie_guest_destroy(struct vm *vm);
lib/s390x/sie.c +2 -2
index 40936bd2..0fa915cf 100644
@@ -59,7 +59,7 @@ void sie(struct vm *vm)
 	/* When a pgm int code is set, we'll never enter SIE below. */
 	assert(!read_pgm_int_code());
 
-	if (vm->sblk->sdf == 2)
+	if (sie_is_pv(vm))
 		memcpy(vm->sblk->pv_grregs, vm->save_area.guest.grs,
 		       sizeof(vm->save_area.guest.grs));
 
@@ -98,7 +98,7 @@ void sie(struct vm *vm)
 	/* restore the old CR 13 */
 	lctlg(13, old_cr13);
 
-	if (vm->sblk->sdf == 2)
+	if (sie_is_pv(vm))
 		memcpy(vm->save_area.guest.grs, vm->sblk->pv_grregs,
 		       sizeof(vm->save_area.guest.grs));
 }
#13 lore raw

PATCH v3 4/7 s390x: Add function for checking diagnose intercepts

In reply to Nina Schoetterl-Glausch

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

---
 lib/s390x/pv_icptdata.h | 42 --------------------------------
 lib/s390x/sie.h         | 12 ++++++++++
 lib/s390x/sie.c         | 53 +++++++++++++++++++++++++++++++++++++++++
 s390x/pv-diags.c        |  8 +++----
 s390x/pv-icptcode.c     | 11 ++++-----
 s390x/pv-ipl.c          |  7 +++---
 6 files changed, 76 insertions(+), 57 deletions(-)
 delete mode 100644 lib/s390x/pv_icptdata.h
lib/s390x/pv_icptdata.h deleted +0 -42
deleted file mode 100644
index 4746117e..00000000
@@ -1,42 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/*
- * Commonly used checks for PV SIE intercept data
- *
- * Copyright IBM Corp. 2023
- * Author: Janosch Frank <[email protected]>
- */
-
-#ifndef _S390X_PV_ICPTDATA_H_
-#define _S390X_PV_ICPTDATA_H_
-
-#include <sie.h>
-
-/*
- * Checks the diagnose instruction intercept data for consistency with
- * the constants defined by the PV SIE architecture
- *
- * Supports: 0x44, 0x9c, 0x288, 0x308, 0x500
- */
-static bool pv_icptdata_check_diag(struct vm *vm, int diag)
-{
-	int icptcode;
-
-	switch (diag) {
-	case 0x44:
-	case 0x9c:
-	case 0x288:
-	case 0x308:
-		icptcode = ICPT_PV_NOTIFY;
-		break;
-	case 0x500:
-		icptcode = ICPT_PV_INSTR;
-		break;
-	default:
-		/* If a new diag is introduced add it to the cases above! */
-		assert(0);
-	}
-
-	return vm->sblk->icptcode == icptcode && vm->sblk->ipa == 0x8302 &&
-	       vm->sblk->ipb == 0x50000000 && vm->save_area.guest.grs[5] == diag;
-}
-#endif
lib/s390x/sie.h +12 -0
index 53cd767f..6d1a0d6e 100644
@@ -287,6 +287,18 @@ static inline bool sie_is_pv(struct vm *vm)
 	return vm->sblk->sdf == 2;
 }
 
+/**
+ * sie_is_diag_icpt() - Check if intercept is due to diagnose instruction
+ * @vm: the guest
+ * @diag: the expected diagnose code
+ *
+ * Check that the intercept is due to diagnose @diag and valid.
+ * For protected virtualisation, check that the intercept data meets additional
+ * constraints.
+ *
+ * Returns: true if intercept is due to a valid and has matching diagnose code
+ */
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag);
 void sie_guest_sca_create(struct vm *vm);
 void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
 void sie_guest_destroy(struct vm *vm);
lib/s390x/sie.c +53 -0

59 lines not on this page · raw message

s390x/pv-diags.c +3 -5
index 3193ad99..6ebe469a 100644
@@ -9,7 +9,6 @@
  */
 #include <libcflat.h>
 #include <snippet.h>
-#include <pv_icptdata.h>
 #include <sie.h>
 #include <sclp.h>
 #include <asm/facility.h>
@@ -32,8 +31,7 @@ static void test_diag_500(void)
 			size_gbin, size_hdr, SNIPPET_UNPACK_OFF);
 
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x500),
-	       "intercept values");
+	report(sie_is_diag_icpt(&vm, 0x500), "intercept values");
 	report(vm.save_area.guest.grs[1] == 1 &&
 	       vm.save_area.guest.grs[2] == 2 &&
 	       vm.save_area.guest.grs[3] == 3 &&
@@ -45,7 +43,7 @@ static void test_diag_500(void)
 	 */
 	vm.sblk->iictl = IICTL_CODE_OPERAND;
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == PGM_INT_CODE_OPERAND,
 	       "operand exception");
 
@@ -57,7 +55,7 @@ static void test_diag_500(void)
 	vm.sblk->iictl = IICTL_CODE_SPECIFICATION;
 	/* Inject PGM, next exit should be 9c */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == PGM_INT_CODE_SPECIFICATION,
 	       "specification exception");
 
s390x/pv-icptcode.c +5 -6
index d7c47d6f..bc90df1e 100644
@@ -13,7 +13,6 @@
 #include <smp.h>
 #include <sclp.h>
 #include <snippet.h>
-#include <pv_icptdata.h>
 #include <asm/facility.h>
 #include <asm/barrier.h>
 #include <asm/sigp.h>
@@ -47,7 +46,7 @@ static void test_validity_timing(void)
 			size_gbin, size_hdr, SNIPPET_UNPACK_OFF);
 
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x44), "spt done");
+	report(sie_is_diag_icpt(&vm, 0x44), "spt done");
 	stck(&time_exit);
 	tmp = vm.sblk->cputm;
 	mb();
@@ -258,7 +257,7 @@ static void test_validity_asce(void)
 
 	/* Try if we can still do an entry with the correct asce */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x44), "re-entry with valid CR1");
+	report(sie_is_diag_icpt(&vm, 0x44), "re-entry with valid CR1");
 	uv_destroy_guest(&vm);
 	free_pages(pgd_new);
 	report_prefix_pop();
@@ -294,7 +293,7 @@ static void run_icpt_122_tests_prefix(unsigned long prefix)
 
 	sie(&vm);
 	/* Guest indicates that it has been setup via the diag 0x44 */
-	assert(pv_icptdata_check_diag(&vm, 0x44));
+	assert(sie_is_diag_icpt(&vm, 0x44));
 	/* If the pages have not been shared these writes will cause exceptions */
 	ptr = (uint32_t *)prefix;
 	WRITE_ONCE(ptr, 0);
@@ -328,7 +327,7 @@ static void test_icpt_112(void)
 
 	/* Setup of the guest's state for 0x0 prefix */
 	sie(&vm);
-	assert(pv_icptdata_check_diag(&vm, 0x44));
+	assert(sie_is_diag_icpt(&vm, 0x44));
 
 	/* Test on standard 0x0 prefix */
 	run_icpt_122_tests_prefix(0);
@@ -348,7 +347,7 @@ static void test_icpt_112(void)
 
 	/* Try a re-entry after everything has been imported again */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == 42,
 	       "re-entry successful");
 	report_prefix_pop();
s390x/pv-ipl.c +3 -4
index cc46e7f7..cd49bd95 100644
@@ -11,7 +11,6 @@
 #include <sie.h>
 #include <sclp.h>
 #include <snippet.h>
-#include <pv_icptdata.h>
 #include <asm/facility.h>
 #include <asm/uv.h>
 
@@ -35,7 +34,7 @@ static void test_diag_308(int subcode)
 
 	/* First exit is a diag 0x500 */
 	sie(&vm);
-	assert(pv_icptdata_check_diag(&vm, 0x500));
+	assert(sie_is_diag_icpt(&vm, 0x500));
 
 	/*
 	 * The snippet asked us for the subcode and we answer by
@@ -46,7 +45,7 @@ static void test_diag_308(int subcode)
 
 	/* Continue after diag 0x500, next icpt should be the 0x308 */
 	sie(&vm);
-	assert(pv_icptdata_check_diag(&vm, 0x308));
+	assert(sie_is_diag_icpt(&vm, 0x308));
 	assert(vm.save_area.guest.grs[2] == subcode);
 
 	/*
@@ -118,7 +117,7 @@ static void test_diag_308(int subcode)
 	 * see a diagnose 0x9c PV instruction notification.
 	 */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == 42,
 	       "continue after load");
 
Signature · 1 line
2.44.0
#14 lore raw

Re: [kvm-unit-tests PATCH v3 4/7] s390x: Add function for checking diagnose intercepts

Quoted 5 lines · On Thu, 20 Jun 2024 16:16:57 +0200 Nina Schoetterl-Glausch <[email protected]>…

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

[...]

Quoted 32 lines
lib/s390x/sie.c +24 -0
index 0fa915cf..d4ba2a40 100644
@@ -42,6 +42,59 @@ void sie_check_validity(struct vm *vm, uint16_t vir_exp)
 	report(vir_exp == vir, "VALIDITY: %x", vir);
 }
 
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag)
+{
+	union {
+		struct {
+			uint64_t     : 16;
+			uint64_t ipa : 16;
+			uint64_t ipb : 32;
+		};
+		struct {
+			uint64_t          : 16;
+			uint64_t opcode   :  8;
+			uint64_t r_1      :  4;
+			uint64_t r_2      :  4;
+			uint64_t r_base   :  4;
+			uint64_t displace : 12;
+			uint64_t zero     : 16;
+		};
+	} instr = { .ipa = vm->sblk->ipa, .ipb = vm->sblk->ipb };
+	uint8_t icptcode;
+	uint64_t code;
+
+	switch (diag) {
+	case 0x44:
+	case 0x9c:

lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 78

Quoted 10 lines
lib/s390x/sie.c +10 -0
+	case 0x288:
+	case 0x308:
+		icptcode = ICPT_PV_NOTIFY;
+		break;
+	case 0x500:
+		icptcode = ICPT_PV_INSTR;
+		break;
+	default:
+		/* If a new diag is introduced add it to the cases above! */
+		assert_msg(false, "unknown diag");

just a nit, but would it be possible to also print the diag number that causes the error?

otherwise looks good

#15 lore raw

Re: [kvm-unit-tests PATCH v3 4/7] s390x: Add function for checking diagnose intercepts

Quoted 50 lines · Claudio Imbrenda
Quoted 5 lines · On Thu, 20 Jun 2024 16:16:57 +0200 Nina Schoetterl-Glausch <[email protected]>…

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

[...]

Quoted 37 lines
lib/s390x/sie.c +29 -0
index 0fa915cf..d4ba2a40 100644
@@ -42,6 +42,59 @@ void sie_check_validity(struct vm *vm, uint16_t vir_exp)
 	report(vir_exp == vir, "VALIDITY: %x", vir);
 }
 
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag)
+{
+	union {
+		struct {
+			uint64_t     : 16;
+			uint64_t ipa : 16;
+			uint64_t ipb : 32;
+		};
+		struct {
+			uint64_t          : 16;
+			uint64_t opcode   :  8;
+			uint64_t r_1      :  4;
+			uint64_t r_2      :  4;
+			uint64_t r_base   :  4;
+			uint64_t displace : 12;
+			uint64_t zero     : 16;
+		};
+	} instr = { .ipa = vm->sblk->ipa, .ipb = vm->sblk->ipb };
+	uint8_t icptcode;
+	uint64_t code;
+
+	switch (diag) {
+	case 0x44:
+	case 0x9c:
+	case 0x288:
+	case 0x308:
+		icptcode = ICPT_PV_NOTIFY;
+		break;
+	case 0x500:
Quoted 8 lines

lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 78

Quoted 5 lines
lib/s390x/sie.c +5 -0
+		icptcode = ICPT_PV_INSTR;
+		break;
+	default:
+		/* If a new diag is introduced add it to the cases above! */
+		assert_msg(false, "unknown diag");

just a nit, but would it be possible to also print the diag number that causes the error?

Yes and easy

		assert_msg(false, "unknown diag 0x%x", diag);
Quoted 5 lines

otherwise looks good

#16 lore raw

Re: [kvm-unit-tests PATCH v3 4/7] s390x: Add function for checking diagnose intercepts

In reply to Nina Schoetterl-Glausch

lib/s390x/pv_icptdata.h

Quoted 21 lines · Nina Schoetterl-Glausch

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

---
  lib/s390x/pv_icptdata.h | 42 --------------------------------
  lib/s390x/sie.h         | 12 ++++++++++
  lib/s390x/sie.c         | 53 +++++++++++++++++++++++++++++++++++++++++
  s390x/pv-diags.c        |  8 +++----
  s390x/pv-icptcode.c     | 11 ++++-----
  s390x/pv-ipl.c          |  7 +++---
  6 files changed, 76 insertions(+), 57 deletions(-)

delete mode 100644 lib/s390x/pv_icptdata.h

lib/s390x/pv_icptdata.h deleted +0 -0
deleted file mode 100644
index 4746117e..00000000
@@ -1,42 +0,0 @@

There's a reason why I didn't put this in sie.c and I'm still torn on whether this should be in the lib or in s390x. It's not related to actually running snippets and managing them.

Quoted 7 lines
lib/s390x/pv_icptdata.h deleted +0 -7
-/* SPDX-License-Identifier: GPL-2.0-only */
-/*
- * Commonly used checks for PV SIE intercept data
- *
- * Copyright IBM Corp. 2023
- * Author: Janosch Frank <[email protected]>
- */

[...]

Quoted 5 lines
lib/s390x/sie.h +0 -0
index 53cd767f..6d1a0d6e 100644

lib/s390x/sie.h static inline bool sie_is_pv(struct vm *vm) · line 296

Quoted 10 lines
lib/s390x/sie.h +7 -0
@@ -287,6 +287,18 @@ static inline bool sie_is_pv(struct vm *vm)
  	return vm->sblk->sdf == 2;
  }
 
+/**
+ * sie_is_diag_icpt() - Check if intercept is due to diagnose instruction
+ * @vm: the guest
+ * @diag: the expected diagnose code
+ *
+ * Check that the intercept is due to diagnose @diag and valid.
+ * For protected virtualisation, check that the intercept data meets additional

virtualization

Quoted 9 lines
lib/s390x/sie.h +5 -0
+ * constraints.
+ *
+ * Returns: true if intercept is due to a valid and has matching diagnose code
+ */
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag);
  void sie_guest_sca_create(struct vm *vm);
  void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
  void sie_guest_destroy(struct vm *vm);
lib/s390x/sie.c +0 -0
Reviewed-by on v3 4/7
#17 lore raw

Re: [kvm-unit-tests PATCH v3 4/7] s390x: Add function for checking diagnose intercepts

In reply to Nina Schoetterl-Glausch
Quoted 136 lines · Nina Schoetterl-Glausch

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

---
 lib/s390x/pv_icptdata.h | 42 --------------------------------
 lib/s390x/sie.h         | 12 ++++++++++
 lib/s390x/sie.c         | 53 +++++++++++++++++++++++++++++++++++++++++
 s390x/pv-diags.c        |  8 +++----
 s390x/pv-icptcode.c     | 11 ++++-----
 s390x/pv-ipl.c          |  7 +++---
 6 files changed, 76 insertions(+), 57 deletions(-)
 delete mode 100644 lib/s390x/pv_icptdata.h
lib/s390x/pv_icptdata.h deleted +0 -42
deleted file mode 100644
index 4746117e..00000000
@@ -1,42 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/*
- * Commonly used checks for PV SIE intercept data
- *
- * Copyright IBM Corp. 2023
- * Author: Janosch Frank <[email protected]>
- */
-
-#ifndef _S390X_PV_ICPTDATA_H_
-#define _S390X_PV_ICPTDATA_H_
-
-#include <sie.h>
-
-/*
- * Checks the diagnose instruction intercept data for consistency with
- * the constants defined by the PV SIE architecture
- *
- * Supports: 0x44, 0x9c, 0x288, 0x308, 0x500
- */
-static bool pv_icptdata_check_diag(struct vm *vm, int diag)
-{
-	int icptcode;
-
-	switch (diag) {
-	case 0x44:
-	case 0x9c:
-	case 0x288:
-	case 0x308:
-		icptcode = ICPT_PV_NOTIFY;
-		break;
-	case 0x500:
-		icptcode = ICPT_PV_INSTR;
-		break;
-	default:
-		/* If a new diag is introduced add it to the cases above! */
-		assert(0);
-	}
-
-	return vm->sblk->icptcode == icptcode && vm->sblk->ipa == 0x8302 &&
-	       vm->sblk->ipb == 0x50000000 && vm->save_area.guest.grs[5] == diag;
-}
-#endif
lib/s390x/sie.h +12 -0
index 53cd767f..6d1a0d6e 100644
@@ -287,6 +287,18 @@ static inline bool sie_is_pv(struct vm *vm)
 	return vm->sblk->sdf == 2;
 }
 
+/**
+ * sie_is_diag_icpt() - Check if intercept is due to diagnose instruction
+ * @vm: the guest
+ * @diag: the expected diagnose code
+ *
+ * Check that the intercept is due to diagnose @diag and valid.
+ * For protected virtualisation, check that the intercept data meets additional
+ * constraints.
+ *
+ * Returns: true if intercept is due to a valid and has matching diagnose code
+ */
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag);
 void sie_guest_sca_create(struct vm *vm);
 void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len);
 void sie_guest_destroy(struct vm *vm);
lib/s390x/sie.c +42 -0
index 0fa915cf..d4ba2a40 100644
@@ -42,6 +42,59 @@ void sie_check_validity(struct vm *vm, uint16_t vir_exp)
 	report(vir_exp == vir, "VALIDITY: %x", vir);
 }
 
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag)
+{
+	union {
+		struct {
+			uint64_t     : 16;
+			uint64_t ipa : 16;
+			uint64_t ipb : 32;
+		};
+		struct {
+			uint64_t          : 16;
+			uint64_t opcode   :  8;
+			uint64_t r_1      :  4;
+			uint64_t r_2      :  4;
+			uint64_t r_base   :  4;
+			uint64_t displace : 12;
+			uint64_t zero     : 16;
+		};
+	} instr = { .ipa = vm->sblk->ipa, .ipb = vm->sblk->ipb };
+	uint8_t icptcode;
+	uint64_t code;
+
+	switch (diag) {
+	case 0x44:
+	case 0x9c:
+	case 0x288:
+	case 0x308:
+		icptcode = ICPT_PV_NOTIFY;
+		break;
+	case 0x500:
+		icptcode = ICPT_PV_INSTR;
+		break;
+	default:
+		/* If a new diag is introduced add it to the cases above! */
+		assert_msg(false, "unknown diag");
+	}
+
+	if (sie_is_pv(vm)) {
+		if (instr.r_1 != 0 || instr.r_2 != 2 || instr.r_base != 5)
+			return false;
+		if (instr.displace)
+			return false;
+	} else {

lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 96

Quoted 10 lines
lib/s390x/sie.c +10 -0
+		icptcode = ICPT_INST;
+	}
+	if (vm->sblk->icptcode != icptcode)
+		return false;
+	if (instr.opcode != 0x83 || instr.zero)
+		return false;
+	code = instr.r_base ? vm->save_area.guest.grs[instr.r_base] : 0;
+	code = (code + instr.displace) & 0xffff;
+	return code == diag;
+}

It looks like this transformation is equivalent for the PV case. You could put the switch into the sie_is_pv() branch? Otherwise looks okay.

Quoted 140 lines
lib/s390x/sie.c +1 -0
+
 void sie_handle_validity(struct vm *vm)
 {
 	if (vm->sblk->icptcode != ICPT_VALIDITY)
s390x/pv-diags.c +3 -5
index 3193ad99..6ebe469a 100644
@@ -9,7 +9,6 @@
  */
 #include <libcflat.h>
 #include <snippet.h>
-#include <pv_icptdata.h>
 #include <sie.h>
 #include <sclp.h>
 #include <asm/facility.h>
@@ -32,8 +31,7 @@ static void test_diag_500(void)
 			size_gbin, size_hdr, SNIPPET_UNPACK_OFF);
 
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x500),
-	       "intercept values");
+	report(sie_is_diag_icpt(&vm, 0x500), "intercept values");
 	report(vm.save_area.guest.grs[1] == 1 &&
 	       vm.save_area.guest.grs[2] == 2 &&
 	       vm.save_area.guest.grs[3] == 3 &&
@@ -45,7 +43,7 @@ static void test_diag_500(void)
 	 */
 	vm.sblk->iictl = IICTL_CODE_OPERAND;
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == PGM_INT_CODE_OPERAND,
 	       "operand exception");
 
@@ -57,7 +55,7 @@ static void test_diag_500(void)
 	vm.sblk->iictl = IICTL_CODE_SPECIFICATION;
 	/* Inject PGM, next exit should be 9c */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == PGM_INT_CODE_SPECIFICATION,
 	       "specification exception");
 
s390x/pv-icptcode.c +5 -6
index d7c47d6f..bc90df1e 100644
@@ -13,7 +13,6 @@
 #include <smp.h>
 #include <sclp.h>
 #include <snippet.h>
-#include <pv_icptdata.h>
 #include <asm/facility.h>
 #include <asm/barrier.h>
 #include <asm/sigp.h>
@@ -47,7 +46,7 @@ static void test_validity_timing(void)
 			size_gbin, size_hdr, SNIPPET_UNPACK_OFF);
 
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x44), "spt done");
+	report(sie_is_diag_icpt(&vm, 0x44), "spt done");
 	stck(&time_exit);
 	tmp = vm.sblk->cputm;
 	mb();
@@ -258,7 +257,7 @@ static void test_validity_asce(void)
 
 	/* Try if we can still do an entry with the correct asce */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x44), "re-entry with valid CR1");
+	report(sie_is_diag_icpt(&vm, 0x44), "re-entry with valid CR1");
 	uv_destroy_guest(&vm);
 	free_pages(pgd_new);
 	report_prefix_pop();
@@ -294,7 +293,7 @@ static void run_icpt_122_tests_prefix(unsigned long prefix)
 
 	sie(&vm);
 	/* Guest indicates that it has been setup via the diag 0x44 */
-	assert(pv_icptdata_check_diag(&vm, 0x44));
+	assert(sie_is_diag_icpt(&vm, 0x44));
 	/* If the pages have not been shared these writes will cause exceptions */
 	ptr = (uint32_t *)prefix;
 	WRITE_ONCE(ptr, 0);
@@ -328,7 +327,7 @@ static void test_icpt_112(void)
 
 	/* Setup of the guest's state for 0x0 prefix */
 	sie(&vm);
-	assert(pv_icptdata_check_diag(&vm, 0x44));
+	assert(sie_is_diag_icpt(&vm, 0x44));
 
 	/* Test on standard 0x0 prefix */
 	run_icpt_122_tests_prefix(0);
@@ -348,7 +347,7 @@ static void test_icpt_112(void)
 
 	/* Try a re-entry after everything has been imported again */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == 42,
 	       "re-entry successful");
 	report_prefix_pop();
s390x/pv-ipl.c +3 -4
index cc46e7f7..cd49bd95 100644
@@ -11,7 +11,6 @@
 #include <sie.h>
 #include <sclp.h>
 #include <snippet.h>
-#include <pv_icptdata.h>
 #include <asm/facility.h>
 #include <asm/uv.h>
 
@@ -35,7 +34,7 @@ static void test_diag_308(int subcode)
 
 	/* First exit is a diag 0x500 */
 	sie(&vm);
-	assert(pv_icptdata_check_diag(&vm, 0x500));
+	assert(sie_is_diag_icpt(&vm, 0x500));
 
 	/*
 	 * The snippet asked us for the subcode and we answer by
@@ -46,7 +45,7 @@ static void test_diag_308(int subcode)
 
 	/* Continue after diag 0x500, next icpt should be the 0x308 */
 	sie(&vm);
-	assert(pv_icptdata_check_diag(&vm, 0x308));
+	assert(sie_is_diag_icpt(&vm, 0x308));
 	assert(vm.save_area.guest.grs[2] == subcode);
 
 	/*
@@ -118,7 +117,7 @@ static void test_diag_308(int subcode)
 	 * see a diagnose 0x9c PV instruction notification.
 	 */
 	sie(&vm);
-	report(pv_icptdata_check_diag(&vm, 0x9c) &&
+	report(sie_is_diag_icpt(&vm, 0x9c) &&
 	       vm.save_area.guest.grs[0] == 42,
 	       "continue after load");
 
#18 lore raw

Re: [kvm-unit-tests PATCH v3 4/7] s390x: Add function for checking diagnose intercepts

Quoted 15 lines · Nicholas Piggin
Quoted 14 lines · Nina Schoetterl-Glausch

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

---
 lib/s390x/pv_icptdata.h | 42 --------------------------------
 lib/s390x/sie.h         | 12 ++++++++++
 lib/s390x/sie.c         | 53 +++++++++++++++++++++++++++++++++++++++++
 s390x/pv-diags.c        |  8 +++----
 s390x/pv-icptcode.c     | 11 ++++-----
 s390x/pv-ipl.c          |  7 +++---
 6 files changed, 76 insertions(+), 57 deletions(-)
 delete mode 100644 lib/s390x/pv_icptdata.h

[...]

Quoted 46 lines
Quoted 46 lines
lib/s390x/sie.c +46 -0
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag)
+{
+	union {
+		struct {
+			uint64_t     : 16;
+			uint64_t ipa : 16;
+			uint64_t ipb : 32;
+		};
+		struct {
+			uint64_t          : 16;
+			uint64_t opcode   :  8;
+			uint64_t r_1      :  4;
+			uint64_t r_2      :  4;
+			uint64_t r_base   :  4;
+			uint64_t displace : 12;
+			uint64_t zero     : 16;
+		};
+	} instr = { .ipa = vm->sblk->ipa, .ipb = vm->sblk->ipb };
+	uint8_t icptcode;
+	uint64_t code;
+
+	switch (diag) {
+	case 0x44:
+	case 0x9c:
+	case 0x288:
+	case 0x308:
+		icptcode = ICPT_PV_NOTIFY;
+		break;
+	case 0x500:
+		icptcode = ICPT_PV_INSTR;
+		break;
+	default:
+		/* If a new diag is introduced add it to the cases above! */
+		assert_msg(false, "unknown diag");
+	}
+
+	if (sie_is_pv(vm)) {
+		if (instr.r_1 != 0 || instr.r_2 != 2 || instr.r_base != 5)
+			return false;
+		if (instr.displace)
+			return false;
+	} else {
+		icptcode = ICPT_INST;
+	}
+	if (vm->sblk->icptcode != icptcode)
+		return false;
Quoted 8 lines

lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 96

Quoted 6 lines
lib/s390x/sie.c +6 -0
+	if (instr.opcode != 0x83 || instr.zero)
+		return false;
+	code = instr.r_base ? vm->save_area.guest.grs[instr.r_base] : 0;
+	code = (code + instr.displace) & 0xffff;
+	return code == diag;
+}

It looks like this transformation is equivalent for the PV case.

Yes, the PV case just has hardcoded values that we want to check.

Quoted 2 lines

You
could put the switch into the sie_is_pv() branch? Otherwise looks okay.

I want to validate diag for both PV and non PV.

Quoted 2 lines

Thanks!

#19 lore raw

Re: [kvm-unit-tests PATCH v3 4/7] s390x: Add function for checking diagnose intercepts

Quoted 74 lines · Nina Schoetterl-Glausch
Quoted 15 lines · Nicholas Piggin
Quoted 14 lines · Nina Schoetterl-Glausch

sie_is_diag_icpt() checks if the intercept is due to an expected diagnose call and is valid.
It subsumes pv_icptdata_check_diag.

---
 lib/s390x/pv_icptdata.h | 42 --------------------------------
 lib/s390x/sie.h         | 12 ++++++++++
 lib/s390x/sie.c         | 53 +++++++++++++++++++++++++++++++++++++++++
 s390x/pv-diags.c        |  8 +++----
 s390x/pv-icptcode.c     | 11 ++++-----
 s390x/pv-ipl.c          |  7 +++---
 6 files changed, 76 insertions(+), 57 deletions(-)
 delete mode 100644 lib/s390x/pv_icptdata.h

[...]

Quoted 46 lines
Quoted 46 lines
lib/s390x/sie.c +46 -0
+bool sie_is_diag_icpt(struct vm *vm, unsigned int diag)
+{
+	union {
+		struct {
+			uint64_t     : 16;
+			uint64_t ipa : 16;
+			uint64_t ipb : 32;
+		};
+		struct {
+			uint64_t          : 16;
+			uint64_t opcode   :  8;
+			uint64_t r_1      :  4;
+			uint64_t r_2      :  4;
+			uint64_t r_base   :  4;
+			uint64_t displace : 12;
+			uint64_t zero     : 16;
+		};
+	} instr = { .ipa = vm->sblk->ipa, .ipb = vm->sblk->ipb };
+	uint8_t icptcode;
+	uint64_t code;
+
+	switch (diag) {
+	case 0x44:
+	case 0x9c:
+	case 0x288:
+	case 0x308:
+		icptcode = ICPT_PV_NOTIFY;
+		break;
+	case 0x500:
+		icptcode = ICPT_PV_INSTR;
+		break;
+	default:
+		/* If a new diag is introduced add it to the cases above! */
+		assert_msg(false, "unknown diag");
+	}
+
+	if (sie_is_pv(vm)) {
+		if (instr.r_1 != 0 || instr.r_2 != 2 || instr.r_base != 5)
+			return false;
+		if (instr.displace)
+			return false;
+	} else {
+		icptcode = ICPT_INST;
+	}
+	if (vm->sblk->icptcode != icptcode)
+		return false;
Quoted 8 lines

lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 96

Quoted 6 lines
lib/s390x/sie.c +6 -0
+	if (instr.opcode != 0x83 || instr.zero)
+		return false;
+	code = instr.r_base ? vm->save_area.guest.grs[instr.r_base] : 0;
+	code = (code + instr.displace) & 0xffff;
+	return code == diag;
+}

It looks like this transformation is equivalent for the PV case.

Quoted 6 lines

Yes, the PV case just has hardcoded values that we want to check.

Quoted 2 lines

You
could put the switch into the sie_is_pv() branch? Otherwise looks okay.

I want to validate diag for both PV and non PV.

To make sure it is one of listed cases, okay I missed that point. All those same diag numbers are valid for !PV? In that case it's fine.

Thanks,
Nick

#20 lore raw

PATCH v3 5/7 s390x: Add library functions for exiting from snippet

In reply to Nina Schoetterl-Glausch

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

---
 s390x/Makefile                          |  1 +
 lib/s390x/asm/arch_def.h                | 13 ++++++++
 lib/s390x/snippet-guest.h               | 26 +++++++++++++++
 lib/s390x/{snippet.h => snippet-host.h} | 10 ++++--
 lib/s390x/snippet-host.c                | 42 +++++++++++++++++++++++++
 lib/s390x/uv.c                          |  2 +-
 s390x/mvpg-sie.c                        |  2 +-
 s390x/pv-diags.c                        |  2 +-
 s390x/pv-icptcode.c                     |  2 +-
 s390x/pv-ipl.c                          |  2 +-
 s390x/sie-dat.c                         |  2 +-
 s390x/spec_ex-sie.c                     |  2 +-
 s390x/uv-host.c                         |  2 +-
 13 files changed, 97 insertions(+), 11 deletions(-)
 create mode 100644 lib/s390x/snippet-guest.h
 rename lib/s390x/{snippet.h => snippet-host.h} (92%)
 create mode 100644 lib/s390x/snippet-host.c
s390x/Makefile +1 -0
index 23342bd6..12445fb5 100644
@@ -111,6 +111,7 @@ cflatobjs += lib/s390x/css_lib.o
 cflatobjs += lib/s390x/malloc_io.o
 cflatobjs += lib/s390x/uv.o
 cflatobjs += lib/s390x/sie.o
+cflatobjs += lib/s390x/snippet-host.o
 cflatobjs += lib/s390x/fault.o
 
 OBJDIRS += lib/s390x
lib/s390x/asm/arch_def.h +13 -0
index 745a3387..db04deca 100644
@@ -504,4 +504,17 @@ static inline uint32_t get_prefix(void)
 	return current_prefix;
 }
 
+static inline void diag44(void)
+{
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}
+
 #endif
lib/s390x/snippet-guest.h new file +26 -0
new file mode 100644
index 00000000..3cc098e1
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Snippet functionality for the guest.
+ *
+ * Copyright IBM Corp. 2023
+ */
+
+#ifndef _S390X_SNIPPET_GUEST_H_
+#define _S390X_SNIPPET_GUEST_H_
+
+#include <asm/arch_def.h>
+#include <asm/barrier.h>
+
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}
+
+#endif /* _S390X_SNIPPET_GUEST_H_ */
lib/s390x/snippet-host.h renamed from lib/s390x/snippet.h +7 -3
similarity index 92%
rename from lib/s390x/snippet.h
rename to lib/s390x/snippet-host.h
index 910849aa..230b25b0 100644
@@ -1,13 +1,13 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 /*
- * Snippet definitions
+ * Snippet functionality for the host.
  *
  * Copyright IBM Corp. 2021
  * Author: Janosch Frank <[email protected]>
  */
 
-#ifndef _S390X_SNIPPET_H_
-#define _S390X_SNIPPET_H_
+#ifndef _S390X_SNIPPET_HOST_H_
+#define _S390X_SNIPPET_HOST_H_
 
 #include <sie.h>
 #include <uv.h>
@@ -144,4 +144,8 @@ static inline void snippet_setup_guest(struct vm *vm, bool is_pv)
 	}
 }
 
+bool snippet_is_force_exit(struct vm *vm);
+bool snippet_is_force_exit_value(struct vm *vm);
+uint64_t snippet_get_force_exit_value(struct vm *vm);
+void snippet_check_force_exit_value(struct vm *vm, uint64_t exit_exp);
 #endif
lib/s390x/snippet-host.c new file +42 -0

42 lines not on this page · raw message

lib/s390x/uv.c +1 -1
index 723bb4f2..c9fffbfc 100644
@@ -18,7 +18,7 @@
 #include <asm/uv.h>
 #include <uv.h>
 #include <sie.h>
-#include <snippet.h>
+#include <snippet-host.h>
 
 static struct uv_cb_qui uvcb_qui = {
 	.header.cmd = UVC_CMD_QUI,
s390x/mvpg-sie.c +1 -1
index d182b49a..3aec9d71 100644
@@ -19,7 +19,7 @@
 #include <alloc_page.h>
 #include <sclp.h>
 #include <sie.h>
-#include <snippet.h>
+#include <snippet-host.h>
 
 static struct vm vm;
 
s390x/pv-diags.c +1 -1
index 6ebe469a..a686c688 100644
@@ -8,7 +8,7 @@
  *  Janosch Frank <[email protected]>
  */
 #include <libcflat.h>
-#include <snippet.h>
+#include <snippet-host.h>
 #include <sie.h>
 #include <sclp.h>
 #include <asm/facility.h>
s390x/pv-icptcode.c +1 -1
index bc90df1e..82fd4246 100644
@@ -12,7 +12,7 @@
 #include <sie.h>
 #include <smp.h>
 #include <sclp.h>
-#include <snippet.h>
+#include <snippet-host.h>
 #include <asm/facility.h>
 #include <asm/barrier.h>
 #include <asm/sigp.h>
s390x/pv-ipl.c +1 -1
index cd49bd95..2acf7d0a 100644
@@ -10,7 +10,7 @@
 #include <libcflat.h>
 #include <sie.h>
 #include <sclp.h>
-#include <snippet.h>
+#include <snippet-host.h>
 #include <asm/facility.h>
 #include <asm/uv.h>
 
s390x/sie-dat.c +1 -1
index f0257770..9e60f26e 100644
@@ -16,7 +16,7 @@
 #include <alloc_page.h>
 #include <sclp.h>
 #include <sie.h>
-#include <snippet.h>
+#include <snippet-host.h>
 #include "snippets/c/sie-dat.h"
 
 static struct vm vm;
s390x/spec_ex-sie.c +1 -1
index fe2f23ee..0ad7ec08 100644
@@ -13,7 +13,7 @@
 #include <asm/arch_def.h>
 #include <alloc_page.h>
 #include <sie.h>
-#include <snippet.h>
+#include <snippet-host.h>
 #include <hardware.h>
 
 static struct vm vm;
s390x/uv-host.c +1 -1
index 55b46446..87d108b6 100644
@@ -15,7 +15,7 @@
 #include <sclp.h>
 #include <smp.h>
 #include <uv.h>
-#include <snippet.h>
+#include <snippet-host.h>
 #include <mmu.h>
 #include <asm/page.h>
 #include <asm/pgtable.h>
Signature · 1 line
2.44.0
#21 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

Quoted 9 lines · On Thu, 20 Jun 2024 16:16:58 +0200 Nina Schoetterl-Glausch <[email protected]>…

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

[...]

Quoted 14 lines
lib/s390x/asm/arch_def.h +13 -0
+static inline void diag44(void)
+{
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}
+
 #endif

[...]

lib/s390x/snippet-guest.h line 24

Quoted 11 lines
lib/s390x/snippet-guest.h new file +11 -0
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}

why not adding "memory" to the clobbers of the inline asm? (not a big deal, I'm just curious if there is a specific reason for an explicit mb())

[...]

#22 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

Quoted 34 lines · Claudio Imbrenda
Quoted 9 lines · On Thu, 20 Jun 2024 16:16:58 +0200 Nina Schoetterl-Glausch <[email protected]>…

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

[...]

Quoted 14 lines
lib/s390x/asm/arch_def.h +13 -0
+static inline void diag44(void)
+{
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}
+
 #endif

[...]

Quoted 15 lines

lib/s390x/snippet-guest.h line 24

Quoted 11 lines
lib/s390x/snippet-guest.h new file +11 -0
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}

why not adding "memory" to the clobbers of the inline asm? (not a big deal, I'm just curious if there is a specific reason for an explicit mb())

Mostly a matter of taste I guess.
The diag functions are just convenience wrappers, doing nothing but executing the diag.
force_exit is a protocol between the host and guest that uses the diags and adds additional semantics on top.
In theory you could have other use cases where the diags are just a timeslice yield.

Quoted 3 lines

[...]

Reviewed-by
#23 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

Quoted 51 lines · On Thu, 20 Jun 2024 19:16:05 +0200 Nina Schoetterl-Glausch <[email protected]>…
Quoted 34 lines · Claudio Imbrenda
Quoted 9 lines · On Thu, 20 Jun 2024 16:16:58 +0200 Nina Schoetterl-Glausch <[email protected]>…

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

[...]

Quoted 14 lines
lib/s390x/asm/arch_def.h +13 -0
+static inline void diag44(void)
+{
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}
+
 #endif  

[...]

Quoted 15 lines

lib/s390x/snippet-guest.h line 24

Quoted 11 lines
lib/s390x/snippet-guest.h new file +11 -0
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}  

why not adding "memory" to the clobbers of the inline asm? (not a big deal, I'm just curious if there is a specific reason for an explicit mb())

Quoted 6 lines

Mostly a matter of taste I guess.
The diag functions are just convenience wrappers, doing nothing but executing the diag.
force_exit is a protocol between the host and guest that uses the diags and adds additional semantics on top.
In theory you could have other use cases where the diags are just a timeslice yield.

fair enough

Quoted 4 lines
Quoted 3 lines

[...]

#24 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

Quoted 54 lines · Claudio Imbrenda
Quoted 51 lines · On Thu, 20 Jun 2024 19:16:05 +0200 Nina Schoetterl-Glausch <[email protected]>…
Quoted 49 lines · Claudio Imbrenda
Quoted 9 lines · On Thu, 20 Jun 2024 16:16:58 +0200 Nina Schoetterl-Glausch <[email protected]>…

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

[...]

Quoted 14 lines
lib/s390x/asm/arch_def.h +13 -0
+static inline void diag44(void)
+{
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}
+
 #endif  

[...]

lib/s390x/snippet-guest.h line 24

Quoted 11 lines
lib/s390x/snippet-guest.h new file +11 -0
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}  

why not adding "memory" to the clobbers of the inline asm? (not a big deal, I'm just curious if there is a specific reason for an explicit mb())

Quoted 8 lines
Quoted 6 lines

Mostly a matter of taste I guess.
The diag functions are just convenience wrappers, doing nothing but executing the diag.
force_exit is a protocol between the host and guest that uses the diags and adds additional semantics on top.
In theory you could have other use cases where the diags are just a timeslice yield.

fair enough

Sorry I missed these comments. I still had a question about the barriers (e.g., one needed before the diag()?)

Quoted 7 lines
Quoted 4 lines
Quoted 3 lines

[...]

#25 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

In reply to Nina Schoetterl-Glausch
Quoted 50 lines · Nina Schoetterl-Glausch

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

---
 s390x/Makefile                          |  1 +
 lib/s390x/asm/arch_def.h                | 13 ++++++++
 lib/s390x/snippet-guest.h               | 26 +++++++++++++++
 lib/s390x/{snippet.h => snippet-host.h} | 10 ++++--
 lib/s390x/snippet-host.c                | 42 +++++++++++++++++++++++++
 lib/s390x/uv.c                          |  2 +-
 s390x/mvpg-sie.c                        |  2 +-
 s390x/pv-diags.c                        |  2 +-
 s390x/pv-icptcode.c                     |  2 +-
 s390x/pv-ipl.c                          |  2 +-
 s390x/sie-dat.c                         |  2 +-
 s390x/spec_ex-sie.c                     |  2 +-
 s390x/uv-host.c                         |  2 +-
 13 files changed, 97 insertions(+), 11 deletions(-)
 create mode 100644 lib/s390x/snippet-guest.h
 rename lib/s390x/{snippet.h => snippet-host.h} (92%)
 create mode 100644 lib/s390x/snippet-host.c
s390x/Makefile +1 -0
index 23342bd6..12445fb5 100644
@@ -111,6 +111,7 @@ cflatobjs += lib/s390x/css_lib.o
 cflatobjs += lib/s390x/malloc_io.o
 cflatobjs += lib/s390x/uv.o
 cflatobjs += lib/s390x/sie.o
+cflatobjs += lib/s390x/snippet-host.o
 cflatobjs += lib/s390x/fault.o
 
 OBJDIRS += lib/s390x
lib/s390x/asm/arch_def.h +2 -0
index 745a3387..db04deca 100644
@@ -504,4 +504,17 @@ static inline uint32_t get_prefix(void)
 	return current_prefix;
 }
 
+static inline void diag44(void)
+{

lib/s390x/asm/arch_def.h static inline uint32_t get_prefix(void) · line 518

Quoted 10 lines
lib/s390x/asm/arch_def.h +10 -0
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}

Would you add a "memory" clobber to these maybe? In theory I think gcc can move even volatile asm around unless there are depdendencies. Maybe I am overly paranoid.

Quoted 22 lines
lib/s390x/asm/arch_def.h +1 -0
+
 #endif
lib/s390x/snippet-guest.h new file +14 -0
new file mode 100644
index 00000000..3cc098e1
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Snippet functionality for the guest.
+ *
+ * Copyright IBM Corp. 2023
+ */
+
+#ifndef _S390X_SNIPPET_GUEST_H_
+#define _S390X_SNIPPET_GUEST_H_
+
+#include <asm/arch_def.h>
+#include <asm/barrier.h>
+
+static inline void force_exit(void)

lib/s390x/snippet-guest.h line 24

Quoted 10 lines
lib/s390x/snippet-guest.h new file +10 -0
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}

You have barriers here, but couldn't the diag get moved before a prior store by the guest?

Silly question since I don't understand the s390x arch or snippet design too well... the diag here causes a guest exit to the host. After the host handles that, it may resume guest at the next instruction? If that is correct, then the barrier here (I think) is for when the guest resumes it would not reorder subsequent loads from guest memory before the diag, because the host might have modified it.

Quoted 69 lines
lib/s390x/snippet-guest.h new file +2 -0
+
+#endif /* _S390X_SNIPPET_GUEST_H_ */
lib/s390x/snippet-host.h renamed from lib/s390x/snippet.h +7 -3
similarity index 92%
rename from lib/s390x/snippet.h
rename to lib/s390x/snippet-host.h
index 910849aa..230b25b0 100644
@@ -1,13 +1,13 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 /*
- * Snippet definitions
+ * Snippet functionality for the host.
  *
  * Copyright IBM Corp. 2021
  * Author: Janosch Frank <[email protected]>
  */
 
-#ifndef _S390X_SNIPPET_H_
-#define _S390X_SNIPPET_H_
+#ifndef _S390X_SNIPPET_HOST_H_
+#define _S390X_SNIPPET_HOST_H_
 
 #include <sie.h>
 #include <uv.h>
@@ -144,4 +144,8 @@ static inline void snippet_setup_guest(struct vm *vm, bool is_pv)
 	}
 }
 
+bool snippet_is_force_exit(struct vm *vm);
+bool snippet_is_force_exit_value(struct vm *vm);
+uint64_t snippet_get_force_exit_value(struct vm *vm);
+void snippet_check_force_exit_value(struct vm *vm, uint64_t exit_exp);
 #endif
lib/s390x/snippet-host.c new file +28 -0
new file mode 100644
index 00000000..44a60bb9
@@ -0,0 +1,42 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Snippet functionality for the host.
+ *
+ * Copyright IBM Corp. 2023
+ */
+
+#include <libcflat.h>
+#include <snippet-host.h>
+#include <sie.h>
+
+bool snippet_is_force_exit(struct vm *vm)
+{
+	return sie_is_diag_icpt(vm, 0x44);
+}
+
+bool snippet_is_force_exit_value(struct vm *vm)
+{
+	return sie_is_diag_icpt(vm, 0x9c);
+}
+
+uint64_t snippet_get_force_exit_value(struct vm *vm)
+{
+	struct kvm_s390_sie_block *sblk = vm->sblk;
+
+	assert(snippet_is_force_exit_value(vm));
+
+	return vm->save_area.guest.grs[(sblk->ipa & 0xf0) >> 4];

lib/s390x/snippet-host.c line 38

Quoted 10 lines
lib/s390x/snippet-host.c new file +10 -0
+}
+
+void snippet_check_force_exit_value(struct vm *vm, uint64_t value_exp)
+{
+	uint64_t value;
+
+	if (snippet_is_force_exit_value(vm)) {
+		value = snippet_get_force_exit_value(vm);
+		report(value == value_exp, "guest forced exit with value (0x%lx == 0x%lx)",
+		       value, value_exp);

This is like kvm selftests guest/host synch design, which is quite nice and useful.

lib/s390x/snippet-host.c line 41

Quoted 3 lines
lib/s390x/snippet-host.c new file +3 -0
+	} else {
+		report_fail("guest forced exit with value");
+	}

Guest forced exit without value? And do you also need to check for non-value force exit to distinguish from a normal snippet exit?

Thanks,
Nick

#26 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

Quoted 29 lines · Nicholas Piggin
Quoted 28 lines · Nina Schoetterl-Glausch

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

---
 s390x/Makefile                          |  1 +
 lib/s390x/asm/arch_def.h                | 13 ++++++++
 lib/s390x/snippet-guest.h               | 26 +++++++++++++++
 lib/s390x/{snippet.h => snippet-host.h} | 10 ++++--
 lib/s390x/snippet-host.c                | 42 +++++++++++++++++++++++++
 lib/s390x/uv.c                          |  2 +-
 s390x/mvpg-sie.c                        |  2 +-
 s390x/pv-diags.c                        |  2 +-
 s390x/pv-icptcode.c                     |  2 +-
 s390x/pv-ipl.c                          |  2 +-
 s390x/sie-dat.c                         |  2 +-
 s390x/spec_ex-sie.c                     |  2 +-
 s390x/uv-host.c                         |  2 +-
 13 files changed, 97 insertions(+), 11 deletions(-)
 create mode 100644 lib/s390x/snippet-guest.h
 rename lib/s390x/{snippet.h => snippet-host.h} (92%)
 create mode 100644 lib/s390x/snippet-host.c

[...]

Quoted 25 lines
Quoted 25 lines
lib/s390x/snippet-guest.h new file +19 -0
new file mode 100644
index 00000000..3cc098e1
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Snippet functionality for the guest.
+ *
+ * Copyright IBM Corp. 2023
+ */
+
+#ifndef _S390X_SNIPPET_GUEST_H_
+#define _S390X_SNIPPET_GUEST_H_
+
+#include <asm/arch_def.h>
+#include <asm/barrier.h>
+
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
Quoted 8 lines

lib/s390x/snippet-guest.h line 24

Quoted 5 lines
lib/s390x/snippet-guest.h new file +5 -0
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}

You have barriers here, but couldn't the diag get moved before a prior store by the guest?

Yeah, makes sense to add another before.

Quoted 7 lines

Silly question since I don't understand the s390x arch or snippet design too well... the diag here causes a guest exit to the host. After the host handles that, it may resume guest at the next instruction? If that is correct, then the barrier here (I think) is for when the guest resumes it would not reorder subsequent loads from guest memory before the diag, because the host might have modified it.

[...]

Quoted 5 lines
Quoted 5 lines
lib/s390x/snippet-host.c new file +0 -0
new file mode 100644
index 00000000..44a60bb9

[...]

Quoted 12 lines

lib/s390x/snippet-host.c line 38

Quoted 8 lines
lib/s390x/snippet-host.c new file +8 -0
+void snippet_check_force_exit_value(struct vm *vm, uint64_t value_exp)
+{
+	uint64_t value;
+
+	if (snippet_is_force_exit_value(vm)) {
+		value = snippet_get_force_exit_value(vm);
+		report(value == value_exp, "guest forced exit with value (0x%lx == 0x%lx)",
+		       value, value_exp);

This is like kvm selftests guest/host synch design, which is quite nice and useful.

Quoted 5 lines

lib/s390x/snippet-host.c line 41

Quoted 3 lines
lib/s390x/snippet-host.c new file +3 -0
+	} else {
+		report_fail("guest forced exit with value");
+	}

Guest forced exit without value?

It's this way round so the output reads:

FAIL: guest forced exit with value

What's after the colon is what failed and the message is the same for PASS/FAIL. Indeed a bit confusing.

Quoted 2 lines

And do you also need to check for non-value force
exit to distinguish from a normal snippet exit?

No, the function does just this check and if you need to handle more complicated situations you need to do that in the caller.

Quoted 3 lines

Thanks,
Nick

#27 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

In reply to Nina Schoetterl-Glausch
Quoted 151 lines · Nina Schoetterl-Glausch

It is useful to be able to force an exit to the host from the snippet, as well as do so while returning a value.
Add this functionality, also add helper functions for the host to check for an exit and get or check the value.
Use diag 0x44 and 0x9c for this.
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.

---
 s390x/Makefile                          |  1 +
 lib/s390x/asm/arch_def.h                | 13 ++++++++
 lib/s390x/snippet-guest.h               | 26 +++++++++++++++
 lib/s390x/{snippet.h => snippet-host.h} | 10 ++++--
 lib/s390x/snippet-host.c                | 42 +++++++++++++++++++++++++
 lib/s390x/uv.c                          |  2 +-
 s390x/mvpg-sie.c                        |  2 +-
 s390x/pv-diags.c                        |  2 +-
 s390x/pv-icptcode.c                     |  2 +-
 s390x/pv-ipl.c                          |  2 +-
 s390x/sie-dat.c                         |  2 +-
 s390x/spec_ex-sie.c                     |  2 +-
 s390x/uv-host.c                         |  2 +-
 13 files changed, 97 insertions(+), 11 deletions(-)
 create mode 100644 lib/s390x/snippet-guest.h
 rename lib/s390x/{snippet.h => snippet-host.h} (92%)
 create mode 100644 lib/s390x/snippet-host.c
s390x/Makefile +1 -0
index 23342bd6..12445fb5 100644
@@ -111,6 +111,7 @@ cflatobjs += lib/s390x/css_lib.o
 cflatobjs += lib/s390x/malloc_io.o
 cflatobjs += lib/s390x/uv.o
 cflatobjs += lib/s390x/sie.o
+cflatobjs += lib/s390x/snippet-host.o
 cflatobjs += lib/s390x/fault.o
 
 OBJDIRS += lib/s390x
lib/s390x/asm/arch_def.h +13 -0
index 745a3387..db04deca 100644
@@ -504,4 +504,17 @@ static inline uint32_t get_prefix(void)
 	return current_prefix;
 }
 
+static inline void diag44(void)
+{
+	asm volatile("diag	0,0,0x44\n");
+}
+
+static inline void diag9c(uint64_t val)
+{
+	asm volatile("diag	%[val],0,0x9c\n"
+		:
+		: [val] "d"(val)
+	);
+}
+
 #endif
lib/s390x/snippet-guest.h new file +26 -0
new file mode 100644
index 00000000..3cc098e1
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Snippet functionality for the guest.
+ *
+ * Copyright IBM Corp. 2023
+ */
+
+#ifndef _S390X_SNIPPET_GUEST_H_
+#define _S390X_SNIPPET_GUEST_H_
+
+#include <asm/arch_def.h>
+#include <asm/barrier.h>
+
+static inline void force_exit(void)
+{
+	diag44();
+	mb(); /* allow host to modify guest memory */
+}
+
+static inline void force_exit_value(uint64_t val)
+{
+	diag9c(val);
+	mb(); /* allow host to modify guest memory */
+}
+
+#endif /* _S390X_SNIPPET_GUEST_H_ */
lib/s390x/snippet-host.h renamed from lib/s390x/snippet.h +7 -3
similarity index 92%
rename from lib/s390x/snippet.h
rename to lib/s390x/snippet-host.h
index 910849aa..230b25b0 100644
@@ -1,13 +1,13 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 /*
- * Snippet definitions
+ * Snippet functionality for the host.
  *
  * Copyright IBM Corp. 2021
  * Author: Janosch Frank <[email protected]>
  */
 
-#ifndef _S390X_SNIPPET_H_
-#define _S390X_SNIPPET_H_
+#ifndef _S390X_SNIPPET_HOST_H_
+#define _S390X_SNIPPET_HOST_H_
 
 #include <sie.h>
 #include <uv.h>
@@ -144,4 +144,8 @@ static inline void snippet_setup_guest(struct vm *vm, bool is_pv)
 	}
 }
 
+bool snippet_is_force_exit(struct vm *vm);
+bool snippet_is_force_exit_value(struct vm *vm);
+uint64_t snippet_get_force_exit_value(struct vm *vm);
+void snippet_check_force_exit_value(struct vm *vm, uint64_t exit_exp);
 #endif
lib/s390x/snippet-host.c new file +18 -0
new file mode 100644
index 00000000..44a60bb9
@@ -0,0 +1,42 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Snippet functionality for the host.
+ *
+ * Copyright IBM Corp. 2023
+ */
+
+#include <libcflat.h>
+#include <snippet-host.h>
+#include <sie.h>
+
+bool snippet_is_force_exit(struct vm *vm)
+{
+	return sie_is_diag_icpt(vm, 0x44);
+}
+
+bool snippet_is_force_exit_value(struct vm *vm)
+{

lib/s390x/snippet-host.c line 28

Quoted 10 lines
lib/s390x/snippet-host.c new file +10 -0
+	return sie_is_diag_icpt(vm, 0x9c);
+}
+
+uint64_t snippet_get_force_exit_value(struct vm *vm)
+{
+	struct kvm_s390_sie_block *sblk = vm->sblk;
+
+	assert(snippet_is_force_exit_value(vm));
+
+	return vm->save_area.guest.grs[(sblk->ipa & 0xf0) >> 4];

Hmm, you have a nice instr struct that you made earlier and now you're back to mask and shift... What about exposing that struct and add a function to create it so you could do grs[sblk_to_instr(sblk).r1] here... Just a thought.

Thanks,
Nick

#28 lore raw

Re: [kvm-unit-tests PATCH v3 5/7] s390x: Add library functions for exiting from snippet

On Tue, 25 Jun 2024 12:57:11 +1000
"Nicholas Piggin" <[email protected]> wrote:

[...]

Quoted 2 lines

Hmm, you have a nice instr struct that you made earlier and now you're back to mask and shift... What about exposing that struct and add a

this is actually a nice idea, we could make a union of the structs for the instruction formats we need, and then we can do what you propose here

Quoted 1 line

function to create it so you could do grs[sblk_to_instr(sblk).r1]

probably something like _INSTR(sblk).${FORMAT}.r1

Quoted 1 line

here... Just a thought.

and maybe we could do this for the kernel as well

Quoted 3 lines

Thanks,
Nick

#29 lore raw

PATCH v3 6/7 s390x: Use library functions for snippet exit

In reply to Nina Schoetterl-Glausch

Replace the existing code for exiting from snippets with the newly introduced library functionality.

---
 s390x/sie-dat.c            | 11 +++--------
 s390x/snippets/c/sie-dat.c | 19 +------------------
 2 files changed, 4 insertions(+), 26 deletions(-)
s390x/sie-dat.c +3 -8
index 9e60f26e..c8f38220 100644
@@ -27,23 +27,18 @@ static void test_sie_dat(void)
 	uint64_t test_page_gpa, test_page_hpa;
 	uint8_t *test_page_hva, expected_val;
 	bool contents_match;
-	uint8_t r1;
 
 	/* guest will tell us the guest physical address of the test buffer */
 	sie(&vm);
-	assert(vm.sblk->icptcode == ICPT_INST &&
-	       (vm.sblk->ipa & 0xff00) == 0x8300 && vm.sblk->ipb == 0x9c0000);
-
-	r1 = (vm.sblk->ipa & 0xf0) >> 4;
-	test_page_gpa = vm.save_area.guest.grs[r1];
+	assert(snippet_is_force_exit_value(&vm));
+	test_page_gpa = snippet_get_force_exit_value(&vm);
 	test_page_hpa = virt_to_pte_phys(guest_root, (void*)test_page_gpa);
 	test_page_hva = __va(test_page_hpa);
 	report_info("test buffer gpa=0x%lx hva=%p", test_page_gpa, test_page_hva);
 
 	/* guest will now write to the test buffer and we verify the contents */
 	sie(&vm);
-	assert(vm.sblk->icptcode == ICPT_INST &&
-	       vm.sblk->ipa == 0x8300 && vm.sblk->ipb == 0x440000);
+	assert(snippet_is_force_exit(&vm));
 
 	contents_match = true;
 	for (unsigned int i = 0; i < GUEST_TEST_PAGE_COUNT; i++) {
s390x/snippets/c/sie-dat.c +1 -18
index 9d89801d..26f045b1 100644
@@ -10,28 +10,11 @@
 #include <libcflat.h>
 #include <asm-generic/page.h>
 #include <asm/mem.h>
+#include <snippet-guest.h>
 #include "sie-dat.h"
 
 static uint8_t test_pages[GUEST_TEST_PAGE_COUNT * PAGE_SIZE] __attribute__((__aligned__(PAGE_SIZE)));
 
-static inline void force_exit(void)
-{
-	asm volatile("diag	0,0,0x44\n"
-		     :
-		     :
-		     : "memory"
-	);
-}
-
-static inline void force_exit_value(uint64_t val)
-{
-	asm volatile("diag	%[val],0,0x9c\n"
-		     :
-		     : [val] "d"(val)
-		     : "memory"
-	);
-}
-
 int main(void)
 {
 	uint8_t *invalid_ptr;
Signature · 1 line
2.44.0
Reviewed-by on v3 6/7
#30 lore raw

Re: [kvm-unit-tests PATCH v3 6/7] s390x: Use library functions for snippet exit

Quoted 4 lines · On Thu, 20 Jun 2024 16:16:59 +0200 Nina Schoetterl-Glausch <[email protected]>…

Replace the existing code for exiting from snippets with the newly introduced library functionality.

Quoted 70 lines
---
 s390x/sie-dat.c            | 11 +++--------
 s390x/snippets/c/sie-dat.c | 19 +------------------
 2 files changed, 4 insertions(+), 26 deletions(-)
s390x/sie-dat.c +3 -8
index 9e60f26e..c8f38220 100644
@@ -27,23 +27,18 @@ static void test_sie_dat(void)
 	uint64_t test_page_gpa, test_page_hpa;
 	uint8_t *test_page_hva, expected_val;
 	bool contents_match;
-	uint8_t r1;
 
 	/* guest will tell us the guest physical address of the test buffer */
 	sie(&vm);
-	assert(vm.sblk->icptcode == ICPT_INST &&
-	       (vm.sblk->ipa & 0xff00) == 0x8300 && vm.sblk->ipb == 0x9c0000);
-
-	r1 = (vm.sblk->ipa & 0xf0) >> 4;
-	test_page_gpa = vm.save_area.guest.grs[r1];
+	assert(snippet_is_force_exit_value(&vm));
+	test_page_gpa = snippet_get_force_exit_value(&vm);
 	test_page_hpa = virt_to_pte_phys(guest_root, (void*)test_page_gpa);
 	test_page_hva = __va(test_page_hpa);
 	report_info("test buffer gpa=0x%lx hva=%p", test_page_gpa, test_page_hva);
 
 	/* guest will now write to the test buffer and we verify the contents */
 	sie(&vm);
-	assert(vm.sblk->icptcode == ICPT_INST &&
-	       vm.sblk->ipa == 0x8300 && vm.sblk->ipb == 0x440000);
+	assert(snippet_is_force_exit(&vm));
 
 	contents_match = true;
 	for (unsigned int i = 0; i < GUEST_TEST_PAGE_COUNT; i++) {
s390x/snippets/c/sie-dat.c +1 -18
index 9d89801d..26f045b1 100644
@@ -10,28 +10,11 @@
 #include <libcflat.h>
 #include <asm-generic/page.h>
 #include <asm/mem.h>
+#include <snippet-guest.h>
 #include "sie-dat.h"
 
 static uint8_t test_pages[GUEST_TEST_PAGE_COUNT * PAGE_SIZE] __attribute__((__aligned__(PAGE_SIZE)));
 
-static inline void force_exit(void)
-{
-	asm volatile("diag	0,0,0x44\n"
-		     :
-		     :
-		     : "memory"
-	);
-}
-
-static inline void force_exit_value(uint64_t val)
-{
-	asm volatile("diag	%[val],0,0x9c\n"
-		     :
-		     : [val] "d"(val)
-		     : "memory"
-	);
-}
-
 int main(void)
 {
 	uint8_t *invalid_ptr;
Reviewed-by on v3 6/7
#31 lore raw

Re: [kvm-unit-tests PATCH v3 6/7] s390x: Use library functions for snippet exit

In reply to Nina Schoetterl-Glausch
Quoted 3 lines · Nina Schoetterl-Glausch

Replace the existing code for exiting from snippets with the newly introduced library functionality.

Is like for like AFAIKS

Quoted 71 lines
---
 s390x/sie-dat.c            | 11 +++--------
 s390x/snippets/c/sie-dat.c | 19 +------------------
 2 files changed, 4 insertions(+), 26 deletions(-)
s390x/sie-dat.c +3 -8
index 9e60f26e..c8f38220 100644
@@ -27,23 +27,18 @@ static void test_sie_dat(void)
 	uint64_t test_page_gpa, test_page_hpa;
 	uint8_t *test_page_hva, expected_val;
 	bool contents_match;
-	uint8_t r1;
 
 	/* guest will tell us the guest physical address of the test buffer */
 	sie(&vm);
-	assert(vm.sblk->icptcode == ICPT_INST &&
-	       (vm.sblk->ipa & 0xff00) == 0x8300 && vm.sblk->ipb == 0x9c0000);
-
-	r1 = (vm.sblk->ipa & 0xf0) >> 4;
-	test_page_gpa = vm.save_area.guest.grs[r1];
+	assert(snippet_is_force_exit_value(&vm));
+	test_page_gpa = snippet_get_force_exit_value(&vm);
 	test_page_hpa = virt_to_pte_phys(guest_root, (void*)test_page_gpa);
 	test_page_hva = __va(test_page_hpa);
 	report_info("test buffer gpa=0x%lx hva=%p", test_page_gpa, test_page_hva);
 
 	/* guest will now write to the test buffer and we verify the contents */
 	sie(&vm);
-	assert(vm.sblk->icptcode == ICPT_INST &&
-	       vm.sblk->ipa == 0x8300 && vm.sblk->ipb == 0x440000);
+	assert(snippet_is_force_exit(&vm));
 
 	contents_match = true;
 	for (unsigned int i = 0; i < GUEST_TEST_PAGE_COUNT; i++) {
s390x/snippets/c/sie-dat.c +1 -18
index 9d89801d..26f045b1 100644
@@ -10,28 +10,11 @@
 #include <libcflat.h>
 #include <asm-generic/page.h>
 #include <asm/mem.h>
+#include <snippet-guest.h>
 #include "sie-dat.h"
 
 static uint8_t test_pages[GUEST_TEST_PAGE_COUNT * PAGE_SIZE] __attribute__((__aligned__(PAGE_SIZE)));
 
-static inline void force_exit(void)
-{
-	asm volatile("diag	0,0,0x44\n"
-		     :
-		     :
-		     : "memory"
-	);
-}
-
-static inline void force_exit_value(uint64_t val)
-{
-	asm volatile("diag	%[val],0,0x9c\n"
-		     :
-		     : [val] "d"(val)
-		     : "memory"
-	);
-}
-
 int main(void)
 {
 	uint8_t *invalid_ptr;
#32 lore raw

PATCH v3 7/7 s390x: Add test for STFLE interpretive execution (format-0)

In reply to Nina Schoetterl-Glausch

The STFLE instruction indicates installed facilities. SIE can interpretively execute STFLE.
Use a snippet guest executing STFLE to get the result of interpretive execution and check the result.

---
 s390x/Makefile           |   2 +
 lib/s390x/asm/facility.h |  10 ++-
 s390x/snippets/c/stfle.c |  26 ++++++++
 s390x/stfle-sie.c        | 134 +++++++++++++++++++++++++++++++++++++++
 s390x/unittests.cfg      |   3 +
 5 files changed, 174 insertions(+), 1 deletion(-)
 create mode 100644 s390x/snippets/c/stfle.c
 create mode 100644 s390x/stfle-sie.c
s390x/Makefile +2 -0
index 12445fb5..7c38d66a 100644
@@ -44,6 +44,7 @@ tests += $(TEST_DIR)/exittime.elf
 tests += $(TEST_DIR)/ex.elf
 tests += $(TEST_DIR)/topology.elf
 tests += $(TEST_DIR)/sie-dat.elf
+tests += $(TEST_DIR)/stfle-sie.elf
 
 pv-tests += $(TEST_DIR)/pv-diags.elf
 pv-tests += $(TEST_DIR)/pv-icptcode.elf
@@ -129,6 +130,7 @@ snippet_lib = $(snippet_asmlib) lib/auxinfo.o
 $(TEST_DIR)/mvpg-sie.elf: snippets = $(SNIPPET_DIR)/c/mvpg-snippet.gbin
 $(TEST_DIR)/sie-dat.elf: snippets = $(SNIPPET_DIR)/c/sie-dat.gbin
 $(TEST_DIR)/spec_ex-sie.elf: snippets = $(SNIPPET_DIR)/c/spec_ex.gbin
+$(TEST_DIR)/stfle-sie.elf: snippets = $(SNIPPET_DIR)/c/stfle.gbin
 
 $(TEST_DIR)/pv-diags.elf: pv-snippets += $(SNIPPET_DIR)/asm/pv-diag-yield.gbin
 $(TEST_DIR)/pv-diags.elf: pv-snippets += $(SNIPPET_DIR)/asm/pv-diag-288.gbin
lib/s390x/asm/facility.h +9 -1
index a66fe56a..2bad05c5 100644
@@ -27,12 +27,20 @@ static inline void stfl(void)
 	asm volatile("	stfl	0(0)\n" : : : "memory");
 }
 
-static inline void stfle(uint64_t *fac, unsigned int nb_doublewords)
+static inline unsigned int stfle(uint64_t *fac, unsigned int nb_doublewords)
 {
 	register unsigned long r0 asm("0") = nb_doublewords - 1;
 
 	asm volatile("	.insn	s,0xb2b00000,0(%1)\n"
 		     : "+d" (r0) : "a" (fac) : "memory", "cc");
+	return r0 + 1;
+}
+
+static inline unsigned long stfle_size(void)
+{
+	uint64_t dummy;
+
+	return stfle(&dummy, 1);
 }
 
 static inline void setup_facilities(void)
s390x/snippets/c/stfle.c new file +26 -0
new file mode 100644
index 00000000..eb024a6a
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright IBM Corp. 2023
+ *
+ * Snippet used by the STLFE interpretive execution facilities test.
+ */
+#include <libcflat.h>
+#include <snippet-guest.h>
+
+int main(void)
+{
+	const unsigned int max_fac_len = 8;
+	uint64_t res[max_fac_len + 1];
+
+	res[0] = max_fac_len - 1;
+	asm volatile ( "lg	0,%[len]\n"
+		"	stfle	%[fac]\n"
+		"	stg	0,%[len]\n"
+		: [fac] "=QS"(*(uint64_t(*)[max_fac_len])&res[1]),
+		  [len] "+RT"(res[0])
+		:
+		: "%r0", "cc"
+	);
+	force_exit_value((uint64_t)&res);
+	return 0;
+}
s390x/stfle-sie.c new file +134 -0

134 lines not on this page · raw message

s390x/unittests.cfg +3 -0
index 3a9decc9..f2203069 100644
@@ -392,3 +392,6 @@ file = sie-dat.elf
 
 [pv-attest]
 file = pv-attest.elf
+
+[stfle-sie]
+file = stfle-sie.elf
Signature · 1 line
2.44.0
#33 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

Quoted 6 lines · On Thu, 20 Jun 2024 16:17:00 +0200 Nina Schoetterl-Glausch <[email protected]>…

The STFLE instruction indicates installed facilities. SIE can interpretively execute STFLE.
Use a snippet guest executing STFLE to get the result of interpretive execution and check the result.

[...]

s390x/stfle-sie.c line 35

Quoted 3 lines
s390x/stfle-sie.c new file +3 -0
+struct guest_stfle_res {
+	uint16_t len;
+	uint64_t reg;

you don't really use reg, do you?

s390x/stfle-sie.c line 42

Quoted 7 lines
s390x/stfle-sie.c new file +7 -0
+	unsigned char *mem;
+};
+
+static struct guest_stfle_res run_guest(void)
+{
+	struct guest_stfle_res res;
+	uint64_t guest_stfle_addr;

uint64_t tmp;

s390x/stfle-sie.c line 48

Quoted 6 lines
s390x/stfle-sie.c new file +6 -0
+
+	sie(&vm);
+	assert(snippet_is_force_exit_value(&vm));
+	guest_stfle_addr = snippet_get_force_exit_value(&vm);
+	res.mem = &vm.guest_mem[guest_stfle_addr];
+	memcpy(&res.reg, res.mem, sizeof(res.reg));

memcpy(&tmp, res.mem, etc);

s390x/stfle-sie.c line 49

Quoted 1 line
s390x/stfle-sie.c new file +1 -0
+	res.len = (res.reg & 0xff) + 1;

(tmp & 0xff) + 1

etc

s390x/stfle-sie.c line 50

Quoted 1 line
s390x/stfle-sie.c new file +1 -0
+	res.mem += sizeof(res.reg);

(you could just do res.mem++ if you had declared it as uint64_t * instead of unsigned char *)

Quoted 3 lines
s390x/stfle-sie.c new file +3 -0
+	return res;
+}
+

[...]

Quoted 4 lines
s390x/stfle-sie.c new file +4 -0
+int main(int argc, char **argv)
+{
+	struct args args = parse_args(argc, argv);
+

s390x/stfle-sie.c line 131

Quoted 10 lines
s390x/stfle-sie.c new file +10 -0
+	if (!sclp_facilities.has_sief2) {
+		report_skip("SIEF2 facility unavailable");
+		goto out;
+	}
+
+	report_info("PRNG seed: 0x%lx", args.seed);
+	prng_s = prng_init(args.seed);
+	setup_guest();
+	if (test_facility(7))
+		test_stfle_format_0();

since STFLE is literally the feature you are testing, maybe you can just skip, like you did for SIEF2?

Quoted 14 lines
s390x/stfle-sie.c new file +3 -0
+out:
+	return report_summary();
+}
s390x/unittests.cfg +3 -0
index 3a9decc9..f2203069 100644
@@ -392,3 +392,6 @@ file = sie-dat.elf
 
 [pv-attest]
 file = pv-attest.elf
+
+[stfle-sie]
+file = stfle-sie.elf
#34 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

Quoted 11 lines · Claudio Imbrenda
Quoted 6 lines · On Thu, 20 Jun 2024 16:17:00 +0200 Nina Schoetterl-Glausch <[email protected]>…

The STFLE instruction indicates installed facilities. SIE can interpretively execute STFLE.
Use a snippet guest executing STFLE to get the result of interpretive execution and check the result.

Quoted 8 lines

[...]

s390x/stfle-sie.c line 35

Quoted 3 lines
s390x/stfle-sie.c new file +3 -0
+struct guest_stfle_res {
+	uint16_t len;
+	uint64_t reg;

you don't really use reg, do you?

No, and I don't think I will either, must be some vestige.

[...]

Quoted 10 lines
Quoted 10 lines
s390x/stfle-sie.c new file +9 -0
+int main(int argc, char **argv)
+{
+	struct args args = parse_args(argc, argv);
+
+	if (!sclp_facilities.has_sief2) {
+		report_skip("SIEF2 facility unavailable");
+		goto out;
+	}
+
Quoted 8 lines

s390x/stfle-sie.c line 131

Quoted 5 lines
s390x/stfle-sie.c new file +5 -0
+	report_info("PRNG seed: 0x%lx", args.seed);
+	prng_s = prng_init(args.seed);
+	setup_guest();
+	if (test_facility(7))
+		test_stfle_format_0();

since STFLE is literally the feature you are testing, maybe you can just skip, like you did for SIEF2?

Yeah, that's better.

Quoted 16 lines
Quoted 15 lines
s390x/stfle-sie.c new file +3 -0
+out:
+	return report_summary();
+}
s390x/unittests.cfg +3 -0
index 3a9decc9..f2203069 100644
@@ -392,3 +392,6 @@ file = sie-dat.elf
 
 [pv-attest]
 file = pv-attest.elf
+
+[stfle-sie]
+file = stfle-sie.elf
#35 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

In reply to Nina Schoetterl-Glausch
Quoted 6 lines · Nina Schoetterl-Glausch

The STFLE instruction indicates installed facilities. SIE can interpretively execute STFLE.
Use a snippet guest executing STFLE to get the result of interpretive execution and check the result.

This one is beyond me... one minor thing, you could move the prng patch to just before this one.

s390x/snippets/c/stfle.c line 3

Quoted 11 lines

+++ b/s390x/snippets/c/stfle.c

s390x/snippets/c/stfle.c new file +7 -0
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright IBM Corp. 2023
 
+++ b/s390x/stfle-sie.c
@@ -0,0 +1,134 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright IBM Corp. 2023

Time to flip your calendar page? :)

Thanks,
Nick

#36 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

In reply to Nina Schoetterl-Glausch

Quoting Nina Schoetterl-Glausch (2024-06-20 16:17:00) [...]

lib/s390x/asm/facility.h static inline void stfl(void) · line 30

Quoted 10 lines
lib/s390x/asm/facility.h +1 -1
index a66fe56a..2bad05c5 100644
@@ -27,12 +27,20 @@ static inline void stfl(void)
        asm volatile("  stfl    0(0)\n" : : : "memory");
 }
 
-static inline void stfle(uint64_t *fac, unsigned int nb_doublewords)
+static inline unsigned int stfle(uint64_t *fac, unsigned int nb_doublewords)

Why unsigned int?

[...]

Quoted 5 lines
s390x/snippets/c/stfle.c new file +0 -0
new file mode 100644
index 00000000..eb024a6a

[...]

s390x/snippets/c/stfle.c line 19

Quoted 10 lines
s390x/snippets/c/stfle.c new file +10 -0
+int main(void)
+{
+       const unsigned int max_fac_len = 8;
+       uint64_t res[max_fac_len + 1];
+
+       res[0] = max_fac_len - 1;
+       asm volatile ( "lg      0,%[len]\n"
+               "       stfle   %[fac]\n"
+               "       stg     0,%[len]\n"
+               : [fac] "=QS"(*(uint64_t(*)[max_fac_len])&res[1]),

Out of curiosity:

Q = Memory reference without index register and with short displacement S = Memory reference without index register but with long displacement

Which one is it?

And: is long displacement even appropriate here?

The cast also is hard to understand. Since this is not super high performance code, do we just want to clobber memory so this gets a bit easier to understand?

Quoted 1 line

+ [len] "+RT"(res[0])

Same question about RT as above.

[...]

Quoted 5 lines
s390x/stfle-sie.c new file +0 -0
new file mode 100644
index 00000000..a3e7f1c9

[...]

s390x/stfle-sie.c line 49

Quoted 11 lines
s390x/stfle-sie.c new file +11 -0
+static struct guest_stfle_res run_guest(void)
+{
+       struct guest_stfle_res res;
+       uint64_t guest_stfle_addr;
+
+       sie(&vm);
+       assert(snippet_is_force_exit_value(&vm));
+       guest_stfle_addr = snippet_get_force_exit_value(&vm);
+       res.mem = &vm.guest_mem[guest_stfle_addr];
+       memcpy(&res.reg, res.mem, sizeof(res.reg));
+       res.len = (res.reg & 0xff) + 1;

If I'm not mistaken, you subtracted 1 in the guest. Here you add it again. Is there a particular reason why?

#37 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

Quoted 10 lines · Nico Boehr

Quoting Nina Schoetterl-Glausch (2024-06-20 16:17:00) [...]

Quoted 8 lines
lib/s390x/asm/facility.h +0 -0
index a66fe56a..2bad05c5 100644
@@ -27,12 +27,20 @@ static inline void stfl(void)
        asm volatile("  stfl    0(0)\n" : : : "memory");
 }
 
Quoted 4 lines

lib/s390x/asm/facility.h static inline void stfl(void) · line 30

Quoted 2 lines
lib/s390x/asm/facility.h +1 -1
-static inline void stfle(uint64_t *fac, unsigned int nb_doublewords)
+static inline unsigned int stfle(uint64_t *fac, unsigned int nb_doublewords)

Why unsigned int?

The return value is 1-256, the size of the type is a bit arbitrary I suppose.

Quoted 21 lines

[...]

Quoted 5 lines
s390x/snippets/c/stfle.c new file +0 -0
new file mode 100644
index 00000000..eb024a6a

[...]

s390x/snippets/c/stfle.c line 19

Quoted 10 lines
s390x/snippets/c/stfle.c new file +10 -0
+int main(void)
+{
+       const unsigned int max_fac_len = 8;
+       uint64_t res[max_fac_len + 1];
+
+       res[0] = max_fac_len - 1;
+       asm volatile ( "lg      0,%[len]\n"
+               "       stfle   %[fac]\n"
+               "       stg     0,%[len]\n"
+               : [fac] "=QS"(*(uint64_t(*)[max_fac_len])&res[1]),

Out of curiosity:

Quoted 4 lines

Q = Memory reference without index register and with short displacement S = Memory reference without index register but with long displacement

Which one is it?

Ups, just short displacement actually.

Quoted 10 lines

And: is long displacement even appropriate here?

The cast also is hard to understand. Since this is not super high performance code, do we just want to clobber memory so this gets a bit easier to understand?

Quoted 1 line

+ [len] "+RT"(res[0])

Same question about RT as above.

Long, but providing a short displacement should be fine too. Not sure if there is any benefit to letting the compiler choose.

Quoted 8 lines

[...]

Quoted 5 lines
s390x/stfle-sie.c new file +0 -0
new file mode 100644
index 00000000..a3e7f1c9

[...]

Quoted 14 lines

s390x/stfle-sie.c line 49

Quoted 11 lines
s390x/stfle-sie.c new file +11 -0
+static struct guest_stfle_res run_guest(void)
+{
+       struct guest_stfle_res res;
+       uint64_t guest_stfle_addr;
+
+       sie(&vm);
+       assert(snippet_is_force_exit_value(&vm));
+       guest_stfle_addr = snippet_get_force_exit_value(&vm);
+       res.mem = &vm.guest_mem[guest_stfle_addr];
+       memcpy(&res.reg, res.mem, sizeof(res.reg));
+       res.len = (res.reg & 0xff) + 1;

If I'm not mistaken, you subtracted 1 in the guest. Here you add it again. Is there a particular reason why?

No, it's the direct result of STFLE on register 0.

maintainer
#38 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

Quoted 9 lines · Nina Schoetterl-Glausch
Quoted 8 lines · Nico Boehr

Quoting Nina Schoetterl-Glausch (2024-06-20 16:17:00)

And: is long displacement even appropriate here?

The cast also is hard to understand. Since this is not super high performance code, do we just want to clobber memory so this gets a bit easier to understand?

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

+ [len] "+RT"(res[0])

Same question about RT as above.

Long, but providing a short displacement should be fine too. Not sure if there is any benefit to letting the compiler choose.

There are some older gcc compilers around which fail to compile if you specify T for long displacement, but the compiler sees that a short displacement would work. So please specify RT to avoid such compile bugs.

See https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=3e4be43f69da

#39 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

In reply to Nina Schoetterl-Glausch

Quoting Nina Schoetterl-Glausch (2024-09-02 16:24:53)

Quoted 26 lines · Nico Boehr
Quoted 10 lines

Quoting Nina Schoetterl-Glausch (2024-06-20 16:17:00) [...]

Quoted 8 lines
lib/s390x/asm/facility.h +0 -0
index a66fe56a..2bad05c5 100644
@@ -27,12 +27,20 @@ static inline void stfl(void)
        asm volatile("  stfl    0(0)\n" : : : "memory");
 }
 
Quoted 4 lines

lib/s390x/asm/facility.h static inline void stfl(void) · line 30

Quoted 2 lines
lib/s390x/asm/facility.h +1 -1
-static inline void stfle(uint64_t *fac, unsigned int nb_doublewords)
+static inline unsigned int stfle(uint64_t *fac, unsigned int nb_doublewords)

Why unsigned int?

The return value is 1-256, the size of the type is a bit arbitrary I suppose.

Quoted 8 lines

[...]

Quoted 5 lines
s390x/snippets/c/stfle.c new file +0 -0
new file mode 100644
index 00000000..eb024a6a

[...]

Quoted 10 lines
Quoted 10 lines

s390x/snippets/c/stfle.c line 19

Quoted 10 lines
s390x/snippets/c/stfle.c new file +10 -0
+int main(void)
+{
+       const unsigned int max_fac_len = 8;
+       uint64_t res[max_fac_len + 1];
+
+       res[0] = max_fac_len - 1;
+       asm volatile ( "lg      0,%[len]\n"
+               "       stfle   %[fac]\n"
+               "       stg     0,%[len]\n"
+               : [fac] "=QS"(*(uint64_t(*)[max_fac_len])&res[1]),

Nina, do you mind sending a new version where we have the constraints simplified, e.g. with just a memory clobber?

#40 lore raw

Re: [kvm-unit-tests PATCH v3 7/7] s390x: Add test for STFLE interpretive execution (format-0)

Quoted 27 lines · Nico Boehr

Quoting Nina Schoetterl-Glausch (2024-09-02 16:24:53)

Quoted 26 lines · Nico Boehr
Quoted 10 lines

Quoting Nina Schoetterl-Glausch (2024-06-20 16:17:00) [...]

Quoted 8 lines
lib/s390x/asm/facility.h +0 -0
index a66fe56a..2bad05c5 100644
@@ -27,12 +27,20 @@ static inline void stfl(void)
        asm volatile("  stfl    0(0)\n" : : : "memory");
 }
 
Quoted 4 lines

lib/s390x/asm/facility.h static inline void stfl(void) · line 30

Quoted 2 lines
lib/s390x/asm/facility.h +1 -1
-static inline void stfle(uint64_t *fac, unsigned int nb_doublewords)
+static inline unsigned int stfle(uint64_t *fac, unsigned int nb_doublewords)

Why unsigned int?

The return value is 1-256, the size of the type is a bit arbitrary I suppose.

Quoted 8 lines

[...]

Quoted 5 lines
s390x/snippets/c/stfle.c new file +0 -0
new file mode 100644
index 00000000..eb024a6a

[...]

Quoted 13 lines
Quoted 10 lines
Quoted 10 lines

s390x/snippets/c/stfle.c line 19

Quoted 10 lines
s390x/snippets/c/stfle.c new file +10 -0
+int main(void)
+{
+       const unsigned int max_fac_len = 8;
+       uint64_t res[max_fac_len + 1];
+
+       res[0] = max_fac_len - 1;
+       asm volatile ( "lg      0,%[len]\n"
+               "       stfle   %[fac]\n"
+               "       stg     0,%[len]\n"
+               : [fac] "=QS"(*(uint64_t(*)[max_fac_len])&res[1]),

Nina, do you mind sending a new version where we have the constraints simplified, e.g. with just a memory clobber?

Will do